Tense and Aspect in the Languages of Europe (Empirical Approaches to Language Typology : Eurotyp 20-6)

Osten Dahl (Editor)

Tense and Aspect in the Languages of Europe

Empirical Approaches to Language Typology EUROTYP

Editors Georg Bossong Bernard Comrie

Mouton de Gruyter Berlin • New York

20-6

Tense and Aspect in the Languages of Europe

edited by Osten Dahl

Mouton de Gruyter Berlin . New York

2000

№

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin.

® Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress

Cataloging-in-Publication-Data

Tense and aspect in the languages of Europe / edited by Osten

Dahl. p. cm. — (Empirical approaches to language typology ; 20-6) One of nine vols, published as part of the Typology of Languages in Europe (Project^ Includes bibliographical references and index. ISBN 311015752 7 (alk paper) 1. Europe — Languages — Tense. 2. Europe — Languages — Aspect. I. Dahl, Osten. II. Typology of Languages in Europe (Project) III. Series. P380.T46 2000 415-dc21 00-029180

Die Deutsche Bibliothek —

Cataloging-in-Publication-Data

Tense and aspect in the languages ol Europe / ed. by Osten Dahl. - Berlin ; New York : Mouton de Gruyter, 2000 (Emp rical approaches to language typology 20 : EUROTYP 6) ISBN 3-11-015752-7

© Copyright 2000 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting: Christoph Eyrich, Berlin. Printing: WB-Druck, Rieden/Allgäu. Binding: Lüderitz ÔC Bauer, Berlin. Printed in Germany.

ЮЙМС ; ' ; . ; ï i * ;

ч

General preface

The present volume is one of a series of nine volumes in which the results of the European research project "Typology of Languages in Europe" (EURO­ TYP) are published. The initiative for a European project on language typology came from a proposal jointly submitted to the European Science Foundation (ESF) by Johannes Bechert (University of Bremen), Claude Buridant (University of Strasbourg), Martin Harris (University of Salford, now University of Man­ chester) and Paolo Ramat (University of Pavia). On the basis of this proposal and following consultations with six experts the Standing Committee for the Humanities of the ESF decided to organize a workshop (Rome, January 1988), in which this idea was further explored and developed. The results of this workshop (published by Mouton, 1990) were sufficiently encouraging for the Standing Committee to appoint a preparatory committee and entrust it with the tasks of drawing up a preliminary proposal, of securing interest and participation from a sufficiently large number of schol­ ars and of finding a suitable programme director. The project proposal formu­ lated and sent out by Simon Dik (University of Amsterdam) as chair of this committee met with very supportive and enthusiastic reactions, so that the Standing Committee for the Humanities recommended the funding of a plan­ ning stage and the General Assembly of the ESF approved a year zero (1989) for an ESF Programme in Language Typology. During this planning phase all major decisions concerning the management structure and the organisation of the work were taken, i. е., the selection of a programme director, the selection of nine focal areas around which the research was to be organized, the selection of a theme coordinator for each theme and the selection of the advisory committee. The first task of the programme director was to draw up a definitive project proposal, which was supplemented with individual proposals for each theme formulated by the theme coordinators, and this new proposal became the basis of a decision by the ESF to fund the Programme for a period of five years (1990-1994). Language typology is the study of regularities, patterns and limits in crosslinguistic variation. The major goal of EUROTYP was to study the patterns and limits of variation in nine focal areas: pragmatic organization of discourse, constituent order, subordination and complementation, adverbial construc­ tions, tense and aspect, noun phrase structure, clitics and word prosodie sys­ tems in the languages of Europe. The decision to restrict the investigation to

vi

General preface

the languages of Europe was imposed for purely practical and pragmatic reasons. In the course of the project an attempt was made, however, to make as much sense of this restriction as possible, by characterizing the specific features of European languages against the background of non-European languages and by identifying areal phenomena {Sprachbünde) within Europe. More specifically, the goals of the EUROTYP project included the following: — to contribute to the analysis of the nine domains singled out as focal areas, to assess patterns and limits of cross-linguistic variation and to offer explanations of the patterns observed. — to bring linguists from various European countries and from different schools or traditions of linguistics together within a major international project on language typology and in doing so create a new basis for future cooperative ventures within the field of linguistics. More than 100 linguists from more than 20 European countries and the United States participated in the project. — to promote the field of language typology inside and outside of Europe. More specifically, an attempt was made to subject to typological analysis a large number of new aspects and domains of language which were uncharted territory before. — to provide new insights into the specific properties of European languages and thus contribute to the characterization of Europe as a linguistic area {Sprachbund). — to make a contribution to the methodology and the theoretical foundations of typology by developing new forms of cooperation and by assessing the role of inductive generalization and the role of theory construction in language typology. We had a further, more ambitious goal, namely to make a contribution to lingustic theory by uncovering major patterns of variation across an important subset of languages, by providing a large testing ground for theoretical controversies and by further developing certain theories in connection with a variety of languages. The results of our work are documented in the nine final volumes: Pragmatic Organization of Discourse in the Languages of Europe (edited by G. Bernini) Constituent Order in the Languages of Europe (edited by A. Siewierska) Subordination and Complementation in the Languages of Europe (edited by N . Vincent) Actance et Valence dans les langues d l'Europe (edited by J. Feuillet) Adverbial Constructions in the Languages of Europe (edited by J. van der Auwera) Tense and Aspect in the Languages of Europe (edited by 0 . Dahl)

General preface

VU

Noun Phrase Structure in the Languages of Europe (edited by F. Plank) Clitics in the Languages of Europe (edited by H. van Riemsdijk) Word Prosodie Systems in the Languages of Europe (edited by H. van der Hülst) In addition, the EUROTYP Project led to a large number of related activities and publications, too numerous to be listed here. At the end of this preface, I would like to express my profound appreciation to all organizations and individuals who made this project possible. First and foremost, I must mention the European Science Foundation, who funded and supported the Programme. More specifically, I would like to express my appreciation to Christoph Mühlberg, Max Sparreboom and Geneviève Schauinger for their constant and efficient support, without which we would not have been able to concentrate on our work. I would, furthermore, like to thank my colleague and assistant, Martin Haspelmath, and indeed all the participants in the Programme for their dedication and hard work. I finally acknowledge with gratitude the crucial role played by Johannes Bechert and Simon Dik in getting this project off the ground. Their illness and untimely deaths deprived us all of two of the project's major instigators. Berlin, September 1995

Ekkehard

König, Programme

Director

preface

J

This volume contains about twenty papers which represent the work of the EUROTYP Theme Group on Tense and Aspect. (The final versions were submitted in 1997, and no substantial updates have been undertaken since then.) I want to thank here first and foremost the authors of the papers, not only for their work but also for the patience they have shown during the long and complex editing process. In addition to the authors, several other people participated in our group meetings and contributed greatly to the discussions: Joan Bybee, Bernard Comrie, Martin Haspelmath, Ekkehard König, Barbara Moltzer, Vladimir Nedjalkov, Nina Niissalo, Svenka Savic, Suzanne Schlyter. Obviously, the number of people who have helped us with information on various languages, in particular by filling out our rather extensive questionnaires, is much larger, making it impossible to enumerate them here. A general thanks is extended to them on behalf of all the members of our group. Also, we thank all people who were helpful in the organization of the meetings, in particular the ESF staff in Strassbourg, who were always been willing to assist us when needed. For contributing to the major task of converting our manuscripts into a neat printed volume, we thank Susan Long, who corrected our English, Bernard Comrie and Georg Bossong, the series editors, and the staff of Mouton de Gruyter. Finally, two persons deserve special mention here: the late Simon Dik, without whom there surely would have been no EUROTYP program, and Larissa Bister, our goddaughter, whose birth coincided with the final group meeting in Helsinki. To them we dedicate this volume. Stockholm, March 2000

Osten Dahl

Contents

J

General preface

v

Preface

i*

General Papers Osten Dahl The tense-aspect systems of European languages in a typological perspective Lars Johanson Viewpoint operators in European languages

3

27

Pier Marco Bertinetto and Denis Delfitto Aspect vs. Actionality: Why they should be kept apart

189

Eva Hedin The type-referring function of the Imperfective

227

RolfThieroff On the areal distribution of tense-aspect categories in Europe

265

Future Time Reference Osten Dahl The grammar of future time reference in European languages

309

Eva Hedin Future marking in conditional and temporal clauses in Greek

329

Osten Dahl Verbs of becoming as future copulas

351

The Perfect Jouko Lindstedt The perfect - aspectual, temporal and evidential

365

xii

Contents

Osten Dahl and Eva Hedin Current relevance and event reference

385

Mario Squartini and Pier Marco Bertinetto The Simple and Compound Past in Romance languages

403

Hannu Tommola On the perfect in North Slavic

441

Nina Graves Macedonian - a language with three perfects?

479

Marja Leinonen and Maria Vilkuna Past tenses in Permic languages

495

The Progressive Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot The progressive in Europe

517

Pier Marco Bertinetto The progressive in Romance, as compared with English

559

Karen H. Ebert Progressive markers in Germanic languages

605

Hannu Tommola Progressive aspect in Baltic Finnic

655

Casper de Groot The absentive

693

Case Studies Eva Agnes Csato Some typological features of the viewpoint and tense system in spoken Northwestern Karaim

723

Karen H. Ebert Aspect in Maltese

753

4

Contents

xiii

i

Appendices 1 The Future Time Reference Questionnaire

789

2 The Perfect Questionnaire

800

3

810

Questionnaire on the Progressive Aspect

4 List of abbreviations used in interlinear glosses

819

5

822

List of working papers

Indices Subject index

827

Language index

834

Author index

841

r

General Papers

г

Osten Dahl

The tense-aspect systems of European languages in a typological perspective

1. General According to the original EUROTYP program proposal, Theme Group 6, with the grammatical categories of tense and aspect as its domain, would focus "on the study of grammaticalization processes as manifested in European languages, and on the identification, description and explanation of tendencies peculiar to the tense-aspect systems of European languages". The following issues were singled out for special attention in the work of the Theme Group: (i) diachronic paths of development; (ii) identification and explanation of areal phenomena; (iii) in-depth studies of individual languages; (iv) language acquisition. Except for the last item, which was planned more as a possible point of contact with other research projects, these issues have all in fact been in focus in the work of our group, something that is hopefully reflected in this volume. The work of the group relied both on the general tradition of tense-aspect research and on earlier contributions of the group members themselves. I shall comment on these two in turn. It is natural that the linguistic phenomena traditionally subsumed under the labels "tense" and "aspect" should have attracted the attention of scholars early on, given their salience in the grammars of most natural languages and their intimate relationship with central cognitive categories. Anyone who undertakes a study in this area has to cope with the burden of an enormous tradition. Paradoxically, however, our EUROTYP Theme Group had relatively little previous work to build on, compared to some other groups in the program. The reason is that, whereas there is an abundant literature on the tense-aspect systems of individual languages as well as works of a general theoretical character, more directly typologically oriented research on tense and aspect is relatively scarce, in spite of notable exceptions such as Friedrich (1974), Comrie (1976,1985), Chung & Timberlake (1985), and Ultan (1978). There are a number of obvious factors behind this scarcity: the predominantly semantic nature of the problems and the difficulty in finding a suitable framework in which different systems can be compared. Given that several members of the group had considerable research experience within the field, it was hardly to be expected (nor desired) that their theoretical ori-

4

Osten Dahl

entation would be totally homogeneous. Seniority being the most objective criterion of order, precedence goes no doubt to Lars Johanson's approach, presented already in his thesis on Turkish aspect (Johanson 1971) and further developed in his contribution to this volume. Since I was given the responsibility to organize the work within the group, it will not be too difficult to discern a bias in the original proposal and the ensuing publications towards the claims made in my earlier work as presented in Dahl (1985), my joint paper with Joan Bybee (Bybee & Dahl 1989), and the methodology used in the investigations behind those publications. Less visible in this volume, although also influential, has also been the Functional Grammar of Simon Dik, the late EUROTYP Chairman. Naturally, the work within the group also builds on the earlier investigations of tense-aspect systems of individual languages or language families by group members such as Pier Marco Bertinetto (Romance, particularly Italian), Karen Ebert (Germanic, particularly Frisian), Casper de Groot (Hungarian), Eva Hedin (Greek), Jouko Lindstedt (Slavic, particularly Bulgarian), Rolf Thieroff (German), and Hannu Tommola (contrastive studies Finno-Ugrian.Slavic). We defined three "focal areas" for the work within the group: (I) Future Time Reference; (II) The Perfect; (III) The Progressive. The last two focal areas thus had a major "gram type" as defined below as their object of study. The first focal area, on the other hand, looked at grammatical marking in a semantically defined domain. Still, of course, there was a salient "gram type" also in Focal Area I, namely the future. Methodologically, the three areas were organized in similar ways, the central empirical tool being a questionnaire. Within Focal Area I, a relatively large number of descriptive sketches of individual languages were produced.1 The rest of this introductory chapter will present, as a general background, an outline of the theoretical assumptions behind my own approach to the typology of tense-aspect systems and some of the typological and areal generalizations that can be made about those systems. In addition, the contents of the volume will be summarized.

2. Notes on the methodology of typological investigations In language typology, methodological issues have been a somewhat neglected area, although lately, questions about language sampling and the use of different kinds of data have become more topical. In large-scale typological research, the following main types of data are available: - primary data elicited by questionnaires and similar methods - primary data from corpora of different kinds - secondary data in the form of previously existing descriptions of the languages in question

The tense-aspect systems of European languages in a typological perspective

5

All three methods have their advantages and drawbacks. One extremely important consideration in typological research is cost in the broad sense of the word - cost in terms of money, other material and immaterial resources and, above all, time. Crudely expressed, in order to get anywhere at all, you have to have "quick and dirty" methods of data collection. Both the questionnaire method and the method of collecting data via grammars should be evaluated in this light. There is really no conflict between the large-scale approach and the in-depth study of individual languages. A large-scale typological investigation necessarily has to be superficial but gives a perspective on the phenomena found in individual languages that you cannot get by looking at them just one by one. Even if most linguists probably agree about this, large-scale typological investigations are still sometimes met with a certain skepticism, especially with regard to the possibility of bias in the heuristics. It is of course true that in some sense you always have to have some idea what you are looking for already in the beginning of a search, and that your expectations will necessarily bias how you interpret data. (Cf. the famous example of the ozone hole at the South Pole, which was initially neglected because the data were filtered away as being too extreme.) A translation questionnaire, that is, a questionnaire in which native speakers of a language L are asked to translate expressions from some other language into L, samples a grammatical domain in a way that has to be guided by the investigator's initial knowledge of the domain. However, the validity of this sampling is not untestable: if the questionnaire, when applied to a language, fails to elicit examples of forms listed in descriptions ofthat language, it is clear that the questionnaire has to be somehow modified. Conversely, the adequacy of a grammatical description is tested when a questionnaire is applied to the language: if forms turn up that are not listed in the grammar, we know that the grammar is not adequate. The translation questionnaire method for investigating the use of grammatical and lexical items in languages has one great advantage, which is perhaps not always appreciated, and that, in my opinion, makes up for a number of its drawbacks. It relies on the notion of translational equivalence, which has the nice property that it is operationally definable and thus independent of any linguistic theory, preconceived or otherwise. An utterance in a language can be said to be translationally equivalent to an utterance in another languages if the two utterances are both given as responses to the task of translating one and the same utterance in a third language. The assumption is that translational equivalence in a large set of contexts will be a strong indicator of synonymy, but synonymy is a notion which can only be dealt with within a given semantic theory, and there is also no guarantee that there are no disturbing factors in the translation process. Translational equivalence thus means that two expressions are actually translated the same way by informants, not that there ls necessarily any deeper relationship between them. The relationship between the notions of translational equivalence and synonymy would be rather similar to that be-

/ 6

Osten Dahl

• г 4; ••• ,_.uv>M'-i M Ï ;

tween acceptability and grammaticality, as these notions are commonly understood in linguistic theory. Let us look at a concrete (albeit constructed) example. Suppose that we compare a Swedish and a German translation of Shakespeare's works and that we find that the English word mean corresponds to either mena or betyda in Swedish and to either meinen or bedeuten in German. Suppose further that we find that whenever the Swedish translator uses mena, the German uses meinen, and whenever the Swedish translator uses betyda, the German uses bedeuten. We are then entitled to say that with respect to these two corpora, the words mena and meinen (or betyda and bedeuten) are translational equivalents. The point here is that we can make this statement without making any claims about the meanings of the words involved. This does not mean, of course, that the fact that the words are translationally equivalent is irrelevant to a description of their meaning; on the contrary, it forms a good point of departure for a further study of them. In a similar way, finding that two grammatical forms are translational equivalents with respect to certain questionnaire material may be highly relevant to the understanding of these forms, but it does not presuppose that we have characterized the meaning of the forms in question in advance. It is thus possible to speak for example of perfects or progressives in different languages on the basis of translational equivalence data without having a theory of the semantics of the perfect or the progressive. The typologist's dream is to have large tagged multilingual corpora of translated texts in which the distribution of various items could be compared systematically. A corpus-based investigation would have the advantage of eliminating the risk of bias in the material due to theoretical preconceptions. On the other hand, it is of course much more costly. For most languages of the world, the question is not so much how we could create such corpora but rather whether they will be there for us to investigate in a generation or two. Realistically, then, the translation questionnaire method will be with us for some time.

3.

"The Bybee & Dahl approach"

In the end of the 70's, I initiated a data-oriented investigation of tense-aspect systems in a large number of languages. Using a translation questionnaire of about 160 sentences, we gathered data about 64 languages in what could probably most aptly be called a "convenience sample". At the same time, Joan Bybee (together with Revere Perkins and William Pagliuca) conducted an investigation of verbal morphological categories in a controlled sample of 50 languages, using extant grammatical descriptions as the main source of information. The results from these projects were published simultaneously, in Dahl (1985) and Bybee (1985). In spite of the differences in methodology, the results obtained were strikingly similar. In 1989, Joan Bybee

The tense-aspect systems of European languages in a typological perspective

f

and I published a joint paper in which we tried to integrate the approaches. Recently, Bybee, Perkins, and Pagliuca have presented their joint work in a monograph (Bybee et al. 1994), which is at present the most complete and up-to-date treatment of tense, aspect and modality in a grammaticalization perspective. Although there are differences in details, and sometimes in emphasis, between the individual works listed here, and between the views of the individual researchers, there is a sufficiently well-developed common core for it to make sense to talk about a "Bybee & Dahl approach". The interpretation I give in this chapter is my own, however. (Cf. also Chapter 1, "Theoretical Background" in Bybee et al. 1994.) The B&D approach differs from most other treatments of tense and aspect in that the basic units of description are not "the category of tense" and "the category of aspect" but rather what we call grams 2 , i.e., things like Progressive in English, the Passesimple in French etc. Notions like tense, aspect, and mood are seen as ways of characterizing the semantic content of grams, or domains from which their meanings are chosen, but do not, in the typical case, represent structurally significant entities in grammatical systems. Many, if not most, grams combine elements from several domains in their semantics, and it is the rule rather than the exception that grams that would traditionally be treated as belonging to the same category behave very differently with respect to how they are expressed in a language. The term "gram" is intended to be used on a language-specific level, that is, a gram belongs to the grammar of an individual language rather than to the general theory of human languages. In this volume, we follow the practice introduced in Comrie (1976) and write names of language-specific grammatical items with initial capitals, and this practice applies also to grams. We thus speak, for example, of the English gram Progressive. An important tenet of the B&D approach, however, is that tense-aspect grams can crosslinguistically be classified into a relatively small set of types. In a universal theory of grammar, then, the relevant unit is the crosslinguistic gram type, the manifestations of which at the language-specific level is the individual gram. Such gram types should not be thought of as absolute entities - characters chosen from a universal "gram alphabet" - but rather as the statistically most probable clusterings of properties in "grammatical space", or alternatively, as relatively stable points along the paths of development that grams take in the course of grammaticalization processes, to be further discussed below. Nor should gram types be identified with 'notional" or "semantic categories". It is true that what keeps the grams of a certain gram type together is primarily their semantics, but it is essential that the gram type is not equal to a notion or concept but is a type of grammatical element, which can also be characterized as to its expressional properties: each gram type has a typical mode of expression, directly related to its position in grammaticalization processes. Later on in this paper, I will discuss a further notion, that of a gram family, which finds its application primarily in areal linguistics. A gram family is basically a set of

8

Osten Dahl

language-specific grams that can be hypothesized to have arisen through one and the same historical process - either by being inherited from a common parent language or as a result of language contact. Gram families, then, differ from gram types in having a location in time and space rather than being universally available, as the latter are.

4.

Grammaticalization

In the B&D approach, we see the study of the grammaticalization (or grammaticization) processes which give rise to tense-aspect grams as an integral part of the general study of those systems, closely intertwined with and often inseparable from the synchronic description. This is not the place to review the rapidly growing literature on grammaticalization: some recent general works that should be mentioned are Lehmann (1982, 1985), Hopper & Traugott (1993), and Heine, Claudi & Hiinnemeyer (1991). I shall instead briefly summarize some of the important properties of grammaticalization processes, as they have been identified in the literature, but from a perspective coloured by my own research experience. The "classical" definition of grammaticalization is the one given already by Antoine Meillet in 1912, viz. that "grammaticalization" denotes those diachronic processes by which lexical items develop into grammatical items. Such an understanding of the notion may seem too narrow, however. The emergence of fixed word order, for instance, would only be subsumable under grammaticalization when the position of morphemes which are on their way to becoming grammaticalized is concerned. Yet, we would want to see such processes as a unitary phenomenon. A more generous definition of grammaticalization would generalize it to all processes by which grammatical phenomena develop." With respect to the processes that interest us here, namely those that feed tense-aspect systems, the classical understanding of grammaticalization is adequate for the majority of all cases. I shall therefore concentrate my discussion on those. When a lexical item grammaticalizes, changes affect both its content and its form. There is no unanimity in the literature concerning the nature of the semantic changes that are involved in grammaticalization. According to one popular view, grammaticalization essentially means semantic bleaching, that is, the semantic content of the item is partly or wholly lost. Another view emphasizes the role of semantic processes such as metaphor in grammaticalization. A possible synthesis of these might differentiate between the early stages of a grammaticalization process (e.g., the development of full verbs into auxiliaries), which are in many respects rather like lexical semantic change in general, and where metaphor, metonymy and similar processes may play essential roles, and the later stages (e.g., the development of past tenses from perfects), for which terms like semantic bleaching may be more appropriate (Hopper & Traugott 1993, Chapter 4).

The tense-aspect systems of European languages in a typological perspective

9

Semantic bleaching in general increases the domain of applicability of an item, and thus may lead to an increase in frequency. The same effect may also be the result of another process, namely that of obligatorization. The property of being obligatory in certain semantically or syntactically defined contexts is often mentioned as characteristic of grammatical elements. In the area of tense and aspect, we may for instance contrast two kinds of progressives: those with obligatory use, as the English Progressive construction, and those which are normally only optional, as the progressive constructions found in most other Germanic languages (see Ebert's paper "Progressive markers in Germanic languages" in this volume). In many cases, the crucial property is not so much absolute obligatoriness as something that could perhaps be labeled independence of relevance considerations, that is, that the use of a certain item is governed by factors other than carrying new and relevant information in the utterance context or not. For instance, past tense morphemes in most European languages may not always be obligatory - there are various uses of the present for past time reference - but they are by and large used irrespective of whether the temporal information they contain is necessary for the intended message or not. In particular, grammaticalized morphemes tend to be used even if they duplicate the information carried by some other element in the sentence (in the case of a tense morpheme, temporal adverbials and the like). In fact, it may sometimes be more difficult to omit a past tense marking when it is in principle redundant than when it is not: the combination of a present tense and a deictic adverbial like yesterday is felt as a deviation from the norm. Both semantic bleaching and the diminished reliance on relevance considerations lead to a general decrease in communicative motivation of an item. Thus, when an item is grammaticalized, its content becomes less significant to the communication. As a concrete example of this we may take the development of perfects into pasts. As noted in Dahl (1985: 138), one may postulate a hierarchy of definiteness in temporal reference, such that, ascending it, the probability of using a perfect diminishes. We may distinguish three or perhaps four steps in this hierarchy, with respect to the point in time at which a situation is located: (i) an indefinite time-point in the past, (ii) a time-point located by an overt time adverbial, (iii) a definite time-point presupposed in the context, (iv) a definite time point defined by a narrative context. The communicative motivation for a tense morpheme marking past time reference is arguably less the more easily the time reference is derivable from the context. The hierarchy thus represents a scale of diminishing communicative motivation. The development °f a perfect into a past, as it can be witnessed for instance in present-day German, proceeds along this hierarchy, allowing perfect marking for more and more definite time reference. The connection between grammaticalization and communicative motivation is something that has not always been appreciated fully. One reason may be that there !s at least seemingly a conflict between the decrease in communicative motivation

10

Osten Dahl

and the functionalist idea that much of language change is motivated by factors having to do with the communicative function of the items involved. In my view, it is obvious that if we are adequately to describe the "functions" of grammatical items, we need a much wider interpretation of "function". I shall return to this question shortly. The diachronic perspective makes possible a re-evaluation of the role of prototypes in the semantics of grammatical items. In Dahl (1985), gram types were said to be definable in terms of their semantics, and the different manifestations of a gram type were supposed to share the same prototype. Diachronically, the prototypical uses ought to be the oldest ones, from which the others have developed. To the extent that grammaticalization involves shifts in meaning rather than just 'semantic bleaching', that does not exclude the prototype of a gram changing, however. It is reasonable to assume that English will no longer shares its prototype with its origin, the Germanic verb willan 'to want'. It is also possible that the processes referred to as 'semantic bleaching' also tend to make the prototype less salient. One interesting phenomenon that is best understood in the diachronic perspective is that of what can be called "doughnut grams" , namely grams whose domain has no focus, that is, no prototypical uses. Doughnut grams are in fact quite frequent, and arise naturally in grammaticalization processes whenever two or more grams are generated out of the same source, at differing times. The older gram then has its centre invaded by the younger one, but keeps the periphery for the time being. Typical cases are progressives that develop into imperfectives and then have to yield their old territory to a new progressive formation, resulting in the seemingly paradoxical situation of a progressive having no progressive uses. Doughnut grams are in fact a special case of the more general phenomenon of residual grams, that is, grams whose domain has been reduced by the invasion of another gram. Such developments, in which grams lose rather than gain territory, may look like counterexamples to the generalizations of grammaticalization theory, but are only apparently so, in that such losses are the secondary result of some other well-behaved grammaticalization process. However, it is not excluded that such secondary processes may result in shifts of meaning, that is, that what was in the earlier situation a secondary use becomes a primary one. This would appear to be the case for example with subjunctives, which are often residual categories with original indicative meaning. Probably as a consequence of the semantic or functional changes, grammaticalized items undergo reduction processes of various sorts. Equally important, however, is the loss of autonomy, which, with a maximally brief characterization, means that what is originally an independent word turns into a modification of another word. 'Modification' should be understood here in a very general sense: it may be both linear, i.e., expressed by affixation, or nonlinear, i.e., by various other processes, such as stem alternations (e.g., ablaut and umlaut) or change in prosodie patterns. In fact, decrease in linearity of expression can be seen as one important factor in

The tense-aspect système of European languages in a typological perspective

11

Figure 1. A donut gram

grammaticalization - nonlinear modifications are characteristic of advanced stages of grammaticalization. An idea that has played an important role in the discussion of grammaticalization is that of its unidirectionality (Hopper & Traugott 1993, Chapter 5). It is not exactly clear, however, what the claim that grammaticalization is unidirectional implies, and what would constitute counterevidence to it. It is often taken to mean that there are no processes by which grammatical morphemes develop into lexical ones. This version of the claim is probably untrue but also fairly uninteresting. A more relevant question is whether the individual processes that instantiate grammaticalization are reversible or not. For instance, consider the following statement: "a category can shift from PROG to IPFV or vice versa".5 What the unidirectionality hypothesis tells us is that there should be no "vice versa": imperfectives should not turn into progressives. But this statement again needs qualification. It should be noted that on the whole, it is very hard to exclude in a principled way a certain historical development, that is, to establish conclusively that a synchronic state A can never be followed by a synchronic state B. In particular, if we are talking about the possibility of reversing a certain process, it is hard to exclude that some combination of processes might lead to what looks like such a reversal. We might imagine, for instance, that an imperfective might end up as a progressive as a result of there being a number of new grams which happen to take over exactly the nonprogressive uses of the imperfective. But this still does not mean that we have to Postulate an inverse of the progressive —y imperfective grammaticalization process in the theory. A somewhat more troublesome situation is perhaps the following. Suppose there is a language community in which some kind of grammaticalization takes place, e.g., a Progressive develops into an imperfective, but in a geographically restricted fashion, resulting in a dialect split, where the gram in question becomes an imperfective in dialect A but remains a progressive in dialect B. Suppose further that due to factors having to do with prestige and other extralinguistic factors, the speakers of dialect A give up most of their dialectal features and adopt what is essentially dialect B. It

12

Osten Dahl

seems that in such a situation, the grammaticalization process may be reversed in the sense that the speakers of dialect В stop using the gram in question as an imperfective and revert to the less grammaticalized stage where it is only a progressive. What this constructed example shows is the borderline between language change and language shift is extremely problematic. On one hand, we could argue that it is not dialect A that is changing, rather, its speakers are shifting to dialect B. On the other hand, it is clear that changes that we would like our theory to account for clear cases of grammaticalization - also often spread in a rather similar fashion, by speakers adopting forms from neighbouring dialects with high prestige. There are in fact attested examples of historical developments in which there seems to be a reversal of a grammaticalization process, and which might be ac­ counted for by an explanation of the "sociolinguistic" kind just sketched. In older stages of High German, the perfect auxiliary could be omitted, especially in subor­ dinate clauses, as in the following example: (1)

German (W. Goethe, Faust I, Vorspiel auf dem Theater) Ihr beiden, die ihr mir so oft, In Not und Trübsal, beigestanden,... 'Ye two that have so often stood by me In time of need and tribulation ... ' (G. M. Priest's translation)

In other languages, e.g., Slavic, auxiliary drop (or perhaps rather: copula drop) shows up as one part of the process by which perfects develop into pasts. In Modern German, however, the process has been reversed in the sense that it is in general no longer possible to omit the perfect auxiliary. (Ironically, the principle that the perfect auxiliary may be omitted in subordinate clauses was borrowed in written Swedish, where it has survived and is still operative.) What we have to conclude, I think, in order to maintain the unidirectionality thesis, is that it has to be seen as operating on a fairly high level of abstraction. We cannot exclude that courses of events that look exactly like the reversal of some grammaticalization process sometimes take place. However, we should still be able to do without such reversed processes as independent constructs in our theory. Given the prominent place of various kinds of reduction - semantic and phonological - in grammaticalization processes, it is somewhat tempting to view grammaticalization in general as "linguistic attrition". I think it is important also to emphasize the positive aspects of grammaticalization: that the object of study is the build-up of grammar, with the focus on systems of inflection. Such systems are a widespread although not universal - feature of human languages, and it is reasonable to assume that we cannot explain their existence only in terms of the wearing-down of lexical material. Rather, we have to assume that inflection serves a function of its own in language. Exactly what that function is remains to be elucidated, like the question of whether we are somehow genetically predisposed to learning inflections.

The tense-aspect systems of European languages in a typological perspective

13

One issue that has not yet been taken seriously in the study of grammaticalization is whether the items that are subject to grammaticalization processes are really animals of the same kind, in particular, whether a concept like "morpheme" is suitable to cover them all. The concept of a morpheme, a child of structuralism, more or less tacitly presupposes an "item and arrangement" (IA) model of language. Applying the IA model to inflectional morphology is in a way forcing it to look the way it would look if it were just like syntax. In many ways, the two other models discussed in Hockett's classic paper (1958), "item and process" (IP) and "word and paradigm" (WP) seem more suited for more complex inflectional systems. If this is the case, however, it is a serious challenge for grammaticalization theory to describe how morphemes are integrated into systems which do not really consist of morphemes. I shall return to this somewhat abstract statement in a while and give a more concrete illustration of what I mean. Bybee et al. make the following statement in the introductory chapter to their book: "We do not take the structuralist position that each language represents a tidy system in which units are defined by the oppositions they enter into and the object of study is the internal system the units are supposed to create. Rather, we consider it more profitable to view languages as composed of substance - both semantic substance and phonetic substance." (1994: 1). It is of course a little risky to associate very specific positions with a large and heterogeneous movement like that of linguistic structuralism, and it may be debated whether you have to believe in "tidy systems" to be a structuralist. The following points may therefore be seen either as marking categorical differences or just shifts of emphasis between grammaticalization-based theories and traditional structuralist approaches: - A grammaticalization-based approach will tend to emphasize precisely the substantive similarities - both in expression and in content - between grammatical items in different languages. The system-internal relations that characterize an item will be seen as resulting from its substantive properties rather than the other way round. When subscribing to the Saussurean slogan that there is nothing in language but differences, linguists have had such phenomena in mind as the fact that the meaning of an unmarked member of an opposition depends on the meaning of the marked member. More precisely, the nonuse of a certain obligatory marking signals that the conditions for that marking are not fulfilled. However, it is important to see that in order to determine this effect, we have to formulate those conditions in the first place. In other words, the paradigmatic relations do in fact presuppose the substantive properties of the items that enter into them. - Grammaticalization processes tend to give rise to situations that do not easily lend themselves to a description in terms of binary oppositions. Thus, grams tend to expand from a point of origin in a wave-like fashion, (metaphorically speaking) chasing each other along a path of development. If one can talk of an "opposition" between an older and a younger gram on the same path, it is rather a secondary

14

Osten Dan!

effect of the relative positions of the grams. Indeed, due to the multidimensionality of the grammaticalization process, it may not be possible to establish a systematic semantic difference between two such grams. - Like other work inspired by prototype theory, this approach rejects the idea of "invariant meanings" and does not postulate a sharp borderline between conventional and contextual interpretations.

5. Grammaticalization clines Grams show gradualness both synchronically and diachronically. Synchronically, the use of a gram tends to be obligatory in the central (prototypical, focal) uses and optional in the peripheral ones, with sinking propensity of use as we go outwards. We can then talk about grammaticalization clines, that is, ordered sets of contexts along which the frequency of grams decreases monotonically. Good examples of such clines are the Romance de-andative future constructions, discussed in the introduction to Part II of this volume. Such clines may of course involve several dimensions, and most probably do in the majority of cases. Diachronically, the propensity to use a gram in a given context also rises gradually. But we also have to include the geographical point of view here. Since linguistic changes of the kind exemplified by grammaticalization tend to spread outwards from a centre of innovation, the propensity to use a gram in a certain context will decrease as we move away from that point. Reducing grammatical space and real space to one dimension each, we may display a theoretical model of a grammaticalization cline as in Figure 2. In real life, the slopes will probably be less smooth. Still, we could take the graph to be an idealized model of, for instance, the use of the Passato Prossimo in Italian, as described by Squartini & Bertinetto in their paper in Part III of this volume.

6. Gram types in tense-aspect systems Figure 3 shows the major gram types that tend to show up in tense-aspect systems and the most common grammaticalization paths that connect them. I have divided the gram types into core gram types and peripheral gram types, depending on their typical degree of grammaticalization. The core gram types are those that as a rule have morphological (mainly inflectional) modes of expression, and which are also in general characterized by being more or less obligatory in their central uses. The peripheral ones are predominantly expressed periphrastically. The most common inflectional tense-aspect gram types in the world's languages are imperfective, perfective, past and future. Indeed, it is rather hard to find an inflectional tense-aspect system that lacks all four of them. Of these, the first three

The tense-aspect systems of European language* ia a typological perspective

IS

propensity of gram use 100 80 60 40 20

s \ • \ 7 \

Distance from " , 4 centre of , innovation

Distance from prototype

Figure 2. A grammaticalization cline

ADDITIONAL TYPES: narrative, experiential, remoteness markers, evidentiality markers LOCATIONAL CONSTRUCTIONS

iterativ

-v-

Core gram-types

progressive

habitual

(mainly inflectional) resultative

»

/A'ALREADr perfect^T >*» -FINISH'

perhlcäve *•*!*•

\

past

4Ï

futuroids lULUrUIUä

> Peripheral

INTENTION

MOVEMENT CHANGE

gram-types (mainly periphrastic) 4.

—'

Figure 3. Major tense-aspect gram types

definitely have predominant inflectional marking; the future is a bit more questionable in this regard. The imperfective and perfective are problematic in other respects, which we now turn to. In most tense-aspect gram types, the marking relations are fairly clear. We have auxiliaries, particles, and affixes marking for example progressive constructions or forms, but there are no morphemes marking nonprogressivity. When it comes to

16

Osten Dahl

perfectivity and imperfectivity, on the other hand, we find both perfective and imperfective markers. In structuralist terms, we cannot identify one of the members of the opposition as the unmarked one. Moreover, perfective and imperfective verb forms tend to be distinguished from each other by rather more complex devices than many other items in morphology. In languages from all over the world, we find that morphological processes such as ablaut, consonant gradation, reduplication, infixes etc. are used to create stem alternations between perfective and imperfective forms to an extent not found anywhere else in tense-aspect systems. Also, there is often a considerable amount of lexical idiosyncrasy: you cannot predict from one verb to another how the opposition is going to be realized. In view of all this, I shall introduce a new term for the grammatical entity represented by the distinction between perfective and imperfective: I shall label it a hypergram type, more specifically the perfectivity hypergram type, since it appears to be one level higher than the gram types we are talking of in other places in this book, and may in specific languages be realized as grams of different types. We might of course keep the structuralist term "opposition", but this might give the wrong associations. The interaction between aspectual and temporal elements in the semantics of the core gram types has far-reaching consequences for tense-aspect systems in general. More specifically, there is a coupling between notional perfectivity and past time reference, and notional imperfectivity and present time reference, in the following sense. States and on-going processes are most naturally thought of as holding at or going on at a specific point in time, at which they can be observed. This point in time will, in the default case, be the time of speaking. Completed events, on the other hand, are typically referred to after being completed. States and on-going processes, then, are connected with present time reference, while completed events are connected with past time reference. This connection shows up in tense-aspect systems in several ways: - Many systems (most of them outside Europe) treat different types of verbs in opposite ways: a zero-marked verb form is interpreted as having present time reference if it is stative and as having past time reference if it is dynamic. - In languages with a distinction between perfective and imperfective verb forms regardless of the marking relations between them - the perfective forms are in the majority of all cases restricted to past time reference, at least when appearing in asserted main clauses. - In those languages that in addition to the perfectivity hypergram also have a past, it is often (probably in the majority of all such languages) restricted to the imperfective, that is, it is what was called a PASTi in Dahl (1985) and a Past Imperfective in Bybee et al. (1994). In such a case, we get what is called a tripartite system in Bybee & Dahl (1989), which is found in a relatively large number of European languages.

The tense-aspect systems of European languages in a typological perspective TypeO (common)

No core categories

Type 1 (common)

IMPERFECTIVE

Type 2 (common)

NON-PAST

ТуреЗ (common)

I

1

PERFECTIVE

I

PAST

I

IMPERFECTIVE

NON-PAST

IMPERFECTIVE

1

NON-PAST

PERFECTIVE

PAST

Г Type 4 (less common)

17

I

PAST

1

PERFECTIVE

1

NON-PAST

PAST

Figure 4. Combinations of core gram types

Figure 4 shows what combinations of the core gram types, perfective, imperfective, and past, are generally found. A number of languages spoken in the eastern Part of Europe have tense-aspect systems that seem to fit Type IV systems, said in Figure 4 to be less common. The nature of "Slavic-style aspect"6 has been dis­ cussed in Dahl (1985) and Bybee & Dahl (1989). To sum up the essential points, the system found, e.g., in Russian differs from typologically more common mani­ festations of perfectivity (i) by being less narrowly tied up with time reference, (ii) °У displaying a somewhat different semantics, which seems more closely related to Aktionsart or actionality distinctions, traditionally connected with the inherent se­ mantics of the verb as a lexical item. These properties seem to be connected with the

18

Osten DaM

historical origin of the perfectivity distinctions in question, in that derivational processes and in particular the morphemes called "bounders" in Bybee & Dahl (1989), that is, telicity-inducing verb particles like up and out in English or prefixes like voz- and vy- in Russian. It seems that there is reason to argue that the "Slavicstyle" systems undergo a grammaticalization process the result of which is that they come closer to the more common perfectivity systems: there is thus a clear difference between Russian and some of the West and South Slavic languages in this regard and an even clearer difference relative to the non-Slavic European languages in which bounders are used as perfectivity markers (Latvian, Lithuanian, Hungarian). The last gram type treated as belonging to the core, the future, is discussed in detail elsewhere in this volume (Part 2), since it was in the centre of interest of one of the Focal Areas of the group. Actually, it is only the more advanced futures that deserve being lumped together with the core gram types; "younger" futures, that is, less grammaticalized grams that mark future time reference, for which we might coin the label futuroids, are better seen as belonging to the periphery of tense-aspect systems. On the periphery, we also find important gram types such as the progressive, the perfect and the habitual. The first two of these made up Focal Areas II and III of our group and accordingly, Parts 3 and 4 of this volume treat them from different perspectives. Here, it should be noted that the progressive and the perfect feed the core gram types, each from its direction: the progressive is a main source for marked imperfectives, while the perfect gives rise to perfectives and pasts, and in addition, to various other gram types, such as indirectives and hodiernal pasts.

7. The areal study of tense-aspect systems We shall here discuss areal phenomena on two levels, which we shall call the microlevel and the macrolevel respectively. The microlevel is the one that has been paid most attention in traditional areal linguistics, which centered on the notion of Sprachbund - a set of languages, geographically close but not necessarily genetically related, in which similar grammatical developments can be found. As will be argued in the introduction to Part II of this volume, Sprachbund phenomena are the rule rather than the exception in grammaticalization processes, in that most of these processes tend to spread over several geographically contiguous languages, giving rise to gram families, as defined above. The genetic distance between the members of a pair of languages involved in such a process may vary from one extreme to another - from closely related dialects to totally unrelated languages. It is plausible that the ease with which a process spreads is inversely correlated to this distance, but there is no reason to see influences that

The tense-aspect systems of European languages in a typological perspective

19

jump the borders between language families as being different in kind rather than just in degree from intrafamilial influences. Europe is an ideal domain for studying micro-areal phenomena, being of the right size to contain a number of "micro-areas" and having a well-documented linguistic history, making it possible to trace synchronic phenomena backwards. Such studies are right now becoming much easier with the advent of computerized corpora both for older and modern texts, although we have not been able to take advantage of these developments to any greater extent within the EUROTYP program. While the microlevel of areal typology can be defined as a level where the relations and mutual influences between individual languages are still discernible, the macrolevel concerns tendencies within larger groups of languages which may be up to the size of a continent. Areal phenomena of this kind have become the object of systematic study only recently, one reason for the newly awakened interest being the insight that many of the phenomena studied in typology have an uneven geographical distribution, which may distort the typologist's results if neglected. Tense-aspect turns out not to be an exception in this regard. Using my own database and that created by Joan Bybee and her collaborators in the GRAMCATS project, I looked at the global distribution of the major gram types. (For a fuller account of the investigation, see Dahl 1995). In Figure 5, the distribution of pasts and perfectives/imperfectives in the GRAMCATS sample is plotted. If we lump together pasts, past imperfectives and remoteness markings, we can see a clear concentration of those gram types in a few fairly well-delimited areas, the largest one covering the bulk of the western part of the Old World, excluding in particular West Africa. There are also clear concentrations in Australia, New Guinea and some other parts of Oceania, and more scattered occurrences in the Americas. Perfectivity distinctions are more evenly distributed. A few remarks on the relation between typological samples and areal phenomena are in order here. It can be said that, due to the way it is constructed, a sample like the one presented in Bybee et al. (1994) or the similar but larger sample of Nichols (1992) (which comprises 175 languages) is in fact rather unlikely to do justice to areal phenomena. The choice method makes it improbable that two geographically contiguous languages make it into the same sample. Also, the percentage of languages chosen is very low - the GRAMCATS sample comprises roughly one per cent of the world's languages, which means that each language in the sample is proxy for about one hundred languages. Any grouping that you can discern and that is large enough not to be due to chance will thus comprise several hundred languages. The conclusion is that no areal phenomenon that covers a smaller number of languages can be detected in this way, which, among other things, means that Europe (with its 175 languages) is not really a possible candidate for an area here. On the other hand, !t is under these circumstances all the more remarkable that the areal patterns seen in !gure 5 are so clear. A further conclusion to be drawn, then, is that areal influence

20

Osten Dahl

,;••>*< ••{уГ [—t] and the adverbial refers to each subevent. Then, +PAST (+INTRA) is also possible, e.g., French toujours réparait la voiture en trois heures 'always repaired the car in three hours'. 'For X time' expressions, which measure the temporal extension and identify the actional phrase as [-t], prefer +PAST (-INTRA), e.g., played chess for one hour, repaired the car for three hours. Exact indications of the outer measures of the event are mostly infertile with a +INTRA view, e.g., *When I entered, they were playing chess for an hour. Compare the natural use of +PAST (-AD) to present an event as extending over a period of time. This choice shows an essential difference between IPFV items of the +INTRA and -AD types. It is thus wrong to claim that 'for X time' necessitates a PFV view. 'For X time' may cooccur with +INTRA items if the actional content has a [+sen reading and the adverbial can be taken to refer to each subevent, e.g., French jouaient toujours aux échecs pendant une heure "They always played chess for а» hour'. Low-focals are generally used in such cases. Some high-focals can only b used for uni-occasional events, for example, the Italian Copula + Gerund periphra"

Viewpoint operate« in European languages

85

which is thus incompatible with durative adverbials. Languages with a focality nosition may use a higher item to envisage each subevent as ongoing, e.g., They re playing chess every evening for three years. If 'for X time' measures the total tension of a pluri-occasional global event, -INTRA is again natural, e.g., Turkish Senelerce, her hafta bana mektup yazdi 'For years and years he wrote to me every week'. 'Until t ' expressions are temporally delimiting in a similar way and choose -INTRA, e.g., played chess until dinner. +INTRA is possible if the actional content gets a [+ser] interpretation. The adverbial then determines the duration of each occurrence, e.g., Turkish Her aksam, saat ona kadar satranç oynuyorlardi 'Each evening they played / were playing chess until ten o'clock'. Here, +INTRA applies to a pluri-occasional global event composed of temporally delimited but aspectually not characterized subevents. Again, languages with a focality opposition may use a higher item to envisage each subevent as ongoing, e.g., Each evening they were playing chess until ten o'clock. The type 'from tx to t y ' behaves similarly, if it measures the event itself, e.g., played chess from 9 p.m. to midnight. If it indicates O, the intraterminal viewpoint, it is of course compatible with +INTRA, e.g., From 9 to 10 p.m. (= as long as observed), they were playing chess. In many languages, 'since t x ' expressions refer to an event that is still going on and viewed intraterminally at O, 't x ' determining its initium, e.g., Italian ballava da mezzogiorno, Spanish bailaba desde el mediodia 'had been dancing since noon'. The abtemporal expression 'since X time', which measures the time between the initium and O, may be used similarly (Bertinetto, this volume). Compare the corresponding uses of -AD.

7.3. Degrees of focality Intraterminals display higher and lower focality degrees. Focality concerns the concentration (focus) of the psychological interest on the situation obtaining at O, the core of "nunc". All intraterminals refer to events that are relevant within an interval uding 0> but they may differ with respect to the relative narrowness of the range of vision determined by "nunc" (Johanson 1971: 130-134). The 'presentness' is re or less focal: from the narrowest idea of an interval confined to the immediProximity of О to broader ideas of an expanded and even infinitely expandable P en odofti m e. ind ° C a l i t y ' S 3 s c a l a r n o t i o n ' a n d focality values are relative. The focality values of ^. l ual items are also subject to constant diachronic change. If focality degrees tem ora the Г P l validity of events, it is only in a relative sense. For example, m r in t, P° ariness conveyed by high-focals is not restricted to events going on at О m st ° concrete and unbroken sense. / am writing a book can be felicitously

86

Lars Johanson

uttered at a moment when no concrete writing or preparatory work is going on. Are you still playing the guitar? may be asked at a moment of total musical inactivity and only imply a relative temporariness, delimiting an expanded 'presentness' from a still broader one. A higher focality degree may also be chosen to express a higher internal dynamicity, intensity or actional density of what is going on at O. Such ef­ fects depend on the conceptualization of the actional content. For example, actional density is less likely to be produced with [—dyn] actional phrases. There have been many attempts to classify items of different focality degrees in European languages and at defining them in a general way by means of ontological and situational characteristics. The results are unsatisfactory, partly due to the lack of clear criteria for the contexts in which the types are supposed to occur. It has not been possible to set up situational types in terms of the focality degrees used to de­ scribe them, not even to establish absolute functional stations such as "progressive" in a straightforward way. Focality oppositions do not distinguish 'processes' (nontransitional events involving internal gradual change) from 'states' (non-transitional events not involving any internal gradual change). The ontological and situational characterization basically depends on actionality and is ultimately a matter of in­ teraction of aspect - of higher and lower focality degrees - with different types of actional content. As a rough classification of 'presentness' ascribable to events going on at O, it might be assumed that the range of vision can be (i) narrow, (ii) expanded, or (iii) open. It may thus concern (i) uni-occasional events, basically confined to the im­ mediate proximity of О and actually performed there; (ii) uni- or pluri-occasional events, not confined to the immediate proximity of O, but actually performed there; or (iii) uni- or pluri-occasional events, in principle being the case at O, but not actu­ ally performed there. The tentative focality scale adopted here has three positions or cardinal degrees: relatively high focality ( HF ), relatively low focality (LF) and nonfocality (N1P). The classification is based on notions developed in Johanson (1971), where different de­ grees of focality ("Prägnanz") were discerned, and mirrors empirically well-known variants of intraterminals. It is not incompatible with the one found in Comrie (1976: 25), where an "imperfective" aspect is subdivided into "habitual" and "continuous", the latter category being further subdivided into "nonprogressive" and "progressive". High-focal roughly corresponds to "progressive", low- focal to "nonprogressive continuous", and nonfocal to "habitual". However, my classification only applies to IPFV items of the intraterminal kind. European nonadterminals do not distinguish linguistically relevant focality degrees. Terms such as "progressivity" and "habituality" will be avoided here, since they are often used in aspectological literature in a substantial sense (e.g., for short-term and long-term states) that has little in common with focality degrees. The meaning of "habituality" in everyday language might also give rise to confusion, suggesting

Viewpoint operators in European languages

8f

that 'habit' and 'repetition' are necessary features. NonfocaHty does not only cover +INTRA operating on [+ser] actional phrases. On the other hand, as noted above, habits do not necessarily require +INTRA presentation as obtaining at an O. The focality degrees will not be referred to as different "aspects". They are neither special viewpoint operators nor actional content categories. Since the same is obviously true of Comrie's subtypes, there is no justification for Bache's assessment of "progressivity" and "habituality" as inherent meanings in the sense of "Aktionsarten", or for his remark that Comrie's subcategories are "all definable in terms of inherent meanings rather than as subdivisions of imperfectivity" (1982: 60). 7.3.1.

Focality types

Intraterminals of high focality, +INTRAHF items, focus on the core of "nunc" and events going on there, suggesting a 'narrow presentness'. They tend to stand for uni-occasional events, i.e., one single occasion of performance, and are particularly suited to convey the impression of internal dynamics, gradual movement towards the finis. These properties are typical of PROG ("progressive"), often considered to be a subtype of IPFV, e.g., was writing. According to Dik, the English Progressive expresses one facet of what may be covered by the "imperfective" of other languages, though the semantic content gets a more specific interpretation (1989: 188-189). Since clear criteria for "progressive situations" are not available, we shall not try to establish a high-focal category delimited from other intraterminals on the basis of situations. The question 'What is X doing right now?' may be interpreted in terms of different degrees of narrowness and answered with more than one item. It is not uncommon that two items of different focality degrees both qualify as PROG according to a given situational criterion or that an item qualifying for PROG in one test fails in the next one. Intraterminals of low focality, +INTRALF items, do not focus strongly on the core of "nunc" but denote, in the sense of an 'expanded presentness', single or repeated, uni-or pluri-occasional events going on there. What is often referred to as a "general present" is characterized by low-focal intraterminality and does not constitute an independent aspect. As noted above, nonfocal -PAST items are non-oppositive +INTRA° items representing the intraterminal notion in a rather feeble way. This subtype of intraterminality is also represented by Imperfects such as French écrivais °r Turkish yaziyordu 'was writing, wrote'. It is apt to present an event as a 'continuous unfolding process' and thus corresponds to Comrie's negatively defined "continuousness", namely an "imperfectivity" that is not "habituality" (1976: 26). However, * do not consider it possible to establish an absolute low-focal category on the basis °f unequivocal situational criteria. Nonfocal intraterminals, expressing +INTRANF, do not focus on the core of nunc" but denote, in the sense of an 'open presentness', single or repeated, uni-

88

Lars Johanson

or pluri-occasional events obtaining there. The event is conceived of as being valid in a regular or characteristic way, without being actually performed at the very mo­ ment of introspection. It may thus also be an extratemporal event, an event-type or a general property. Nonfocals roughly correspond to Comrie's "habituais" which "describe a situation which is characteristic of an extended period of time" (1976: 27-28). Note, however, that it is not the objective extension in time that decides whether the event is 'characteristic' or not, as Bache's interpretation of Comrie's classification seems to suggest (1982: 61). Nonfocal non-pasts, -PAST (+INTRAoNF) items, thus exhibit uses that clearly burst the narrow introspective view. These so-called "general" or "unmarked" Presents can be used for uni-occasional events in temporally limited progress at 0 s , pluri-occasional events, temporally unlimited events, events referred to as a type, universally valid facts (gnomic use), past events (historical use), fictitious events, foreseen, scheduled, planned, intended events (futurate use), potential events, ability to perform a given action, etc. In all this, they are rather similar to —PAST (—AD) items. 7.3.2.

Coverage of situations

Focality degrees are distributed across language-specific items in rather different ways. As for non-past items, there are, for example, "concrete" Presents represent­ ing high and low focality, "usual" or "general" Presents representing low focality or nonfocality, combined "Present-Futures", general items covering the whole range of current, habitual and timeless events, and "potential" Presents representing nonfo­ cality with modal values (e.g., disposition). For example, both the Georgian general Present and the Turkish potential Present may express timeless events: Georgian Dedamica mzis garsemo brunavs, Turkish Diinya, giines etrafinda doner 'The earth turns round the sun'. A lower item on the focality scale is semantically more general than a higher one, indifferent towards the notion of a higher focality. It may thus cover referential areas of a higher item, i.e. be used in cases that allow characterization in terms of higher focality or require a higher item in some other language. If a language has two items of different focality, the lower one may roughly cover ongoing events that are more precisely expressed by the higher one, e.g. English wrote Э was writing. A relatively low +INTRA item in language A may cover situations that require a higher item in language B, e.g., French écrivait Э English was writing. This is also true of items that are indifferent towards intraterminality, e.g., German schrieb D English was writing. Nonfocals may roughly cover situations more accurately expressed by lowand high-focals, e.g., French fume, Modern Greek kapnizi 'smokes'. Many nonfocals with partly modal functions may also cover higher areas. All these items must, however, be distinguished from special markers of pluri-occasionality, disposition,

/ Viewpoint operators in European languages

89

etc. (Section 4.2). The latter cannot cover any higher referential areas, as a praesens I imperfectum usuale, generale or potentiate can. Though intraterminals generally operate on global events, high-focals may operate on the subevent level, i.e. apply to subevents of a pluri-occasional global event, e.g., Each time we flew home, he was reading a book, She always calls me when I am working, Portuguese О Joào esta sempre afumar 'John is always smoking' (Oliveira & Lopes 1995: 109, footnote 16). If two items of different focality degrees are available, the lower one may apply to the global event and the higher one to subevents (see "Prägnanzüberlagerung", 'focality superimposition', Johanson 1971: 267-268). Operating on the subevent level is typical of preaspectuals and early high-focals, e.g., Finnish Han soittaa aina, кип olen siivoamassa '(S)he always calls when I am cleaning'. This behaviour is rather natural given the fact that the items develop from modes of action, which modify single basic actions.

7.4. Focality oppositions Numerous European languages exhibit thoroughly grammaticalized focality opposi­ tions. The area of oppositions in this rigorous sense includes most Romance, Turkic, Iranic, and Caucasian languages, English, Icelandic, Irish, Basque, Maltese, Kalmyk, etc. The absolute degrees vary: some oppositions concern higher, others lower focal­ ity. On the other hand, many European languages such as Slavic, Finno-Ugrian and most of the Germanic ones make little use of highly grammaticalized devices for dis­ tinguishing focality degrees. However, many of them possess preaspectual devices representing lower degrees of grammaticalization and not integrated into the core of the aspectotemporal systems. It may sometimes seem difficult to estimate the degree of grammaticalization and thus to delimit the areas of focality oppositions. The items signalling higher focality show varying degrees of generalized use ("obligatoriness") and combinability with tenses. Many are used rather restrictively. If all such devices are considered, the areas of focality oppositions will practically cover the whole of Europe. Some of them will be commented upon under 8.8-9. In the non-past stratum, English progressives express high-focal intraterminality within the opposition -PAST (+INTRA0 (+FOQ) is writing vs. -PAST (+INTRA0 (—FOC)) writes. In the simple past stratum, however, the high-focal is the only +INTRA item of the opposition: +PAST (+INTRA) was writing vs. +PAST (-INTRA) wrote. The opposing simple Present and Past tenses are not "perfectives", just items characterized by a lower degree or absence of focality. The highly devel­ oped English Progressive system is unique within the Germanic group, but Irish has a strikingly similar basic system. In the non-past stratum, there is an oppo­ sition of a high-focal progressive Present, e.g., -PAST (+INTRA0 (+FOC)) ta ag scriobh 'is writing', and a so-called "usual" or "habitual" Present tense, e.g., -PAST

90

Lars Johanson

(+INTRA° (—FOC)) scriobhann 'writes'. In the simple past stratum, however, the high-focal is, as in English, the only +INTRA item: +PAST (+INTRA) bhi ag 61 'was drinking' vs. 4-PAST (—INTRA) d'ol 'drank'. Similarly, in Basque, analytical verbs form one single +INTRA item in the simple past stratum, +PAST (+INTRA), e.g.Joaiten nintzen 'I went, was going' vs. +PAST (—INTRA^'oa« nintzen 'I went'. In the non-past stratum, the few synthetical verbs left from an older system may oppose a -PAST (+INTRA0 (+FOC)) item to a -PAST (+INTRA0 (-FOC)) one, e.g., egoiten naiz 'I am staying' vs. banago 'I stay' (Haase 1994: 285). The situation is similar with some preaspectual or early +INTRA distinctions, e.g., Estonian -PAST (+INTRA0 (+FOC)) on tegemas 'is doing' vs. -PAST (+INTRA° (-FOC)) teeb 'does', but +PAST (+INTRA) oli tegemas 'was doing' vs. +PAST (-INTRA) tegi 'did'. Kurylowicz assumes a corresponding Hittite opposition appiskizzi 'is taking' vs. epzi 'takes' (1956: 26). The types just mentioned are, however, rather marginal in the overall European picture. Focality oppositions more frequently involve two +INTRA items in both strata, a phenomenon sometimes addressed as grammaticalization of IPFV and PROG in one and the same language. For example, Romance languages except French and Romanian employ, apart from their +INTRA0 Presents and +INTRA Imperfects, thoroughly grammaticalized devices signalling higher focality, e.g., items of the type Auxiliary + Gerund / Infinitive. There are thus -PAST (+INTRA° (+FOC)) vs. -PAST (+INTRA0 (-FOC)) oppositions such as Italian sta cantando 'is singing' vs. canta 'sings, is singing', Spanish esta hablando 'is speaking' vs. habla 'speaks, is speaking', and Portuguese esta afalar or est'afalando 'is speaking' vs. fala 'speaks, is speaking'. There are corresponding +PAST (+INTRA (+FOC)) vs. +PAST (+INTRA (—FOC)) oppositions such as stava cantando 'was singing' vs. cantava 'sang, used to sing, was singing', estaba hablando 'was speaking' vs. hablaba 'spoke, used to speak, was speaking'. The +PAST high-focals are used rather extensively in the languages where they occur. The situation is somewhat different in Turkic and Iranian languages. For example, Turkish has oppositions between higher and lower focal items, e.g., diismekte[dir] 'is falling' vs. düsüyor 'is falling, falls', düsmekteydi 'was falling' vs. düsüyordu 'was falling'; compare the corresponding Kirmanji oppositions dikevîye 'is falling' vs. dikeve 'is falling, falls', diketîye 'was falling' vs. diket 'was falling'. The lower items are very frequently used. All Turkic and most of the Iranian languages also display -PAST (+INTRA° (+FOC)) vs. -PAST (+INTRA° (-FOC)) oppositions including still lower focals, e.g., Azerbaijani yaz'ir 'is writing' vs. yazar 'writes, will write', Tati mibaftœn 'is weaving' vs. mibafu 'weaves, will weave'. Corresponding +PAST (+INTRA (+FOC)) vs. +PAST (+INTRA (-FOC)) oppositions are Azerbaijani yazïrdï 'was writing' vs. yazardï 'wrote, used to write, would write', Tati mibaftœn bu 'was weaving' vs. mibaft 'wove, would weave', Talysh händedä be 'was reading' vs. ahändi 'read, used to read, would read'.

Viewpoint operators in European languages

91

Focality oppositions are found in many other languages. Examples of —PAST (+INTRA0 (±FOC)) oppositions are Sami lœ bâradœmen 'is eating' vs. bârat 'eats, is eating', Kalmyk umsjana 'is reading' vs. umsna 'is reading, reads'. Northeast Caucasian (Nakh-Dagestanian) examples are Chechen (Nakh) molusvu 'is drinking' vs. molu 'is drinking, drinks', Khvarshi kulse goli 'is throwing' vs. kulse 'is throwing, throws', Akhvakh qwarere godi 'is writing' vs. qwariri 'writes' and Archi zari siarsi vi q'onq' T am writing a book (just now)' vs. zari siar q'onq' T write a book (in general)'. Certain Caucasian languages use different copulas - an "esse concretum" У and an "esse abstractum" - for different focality degrees, e.g., Tabasaran +PAST (+INTRA (+FOC)) licurayi < licuri ayi 'was going' vs. +PAST (+INTRA (-FOC)) licuji < licuri vuyi 'went (generally)'. Note that, in spite of their material similarity to English Progressives, the Lithuanian so-called Continuatives (periphrases 'be' + Present participle) do not regularly convey high-focal intraterminal meaning, but rather tend to express imminence, 'be about to'. Thus, buvo berasqs is normally interpreted as 'was about to write', and not as 'was (already) writing'. Renewal of focality is often observed earlier in —PAST than in +PAST items. Thus, Gagauz has one —PAST item of higher and one of lower focality, verer 'gives, is giving' vs. verir 'gives, will give', whereas there is only one established +PAST item, veri(r)di 'gave, was giving'. In some languages, +PLUR (pluri-occasionality) and +DISP (disposition) mark­ ers may combine with members of focality oppositions. The marked member of the English opposition +PAST (+PLUR (+INTRA (+FOC))) used to be writing vs. +PAST (+PLUR (+INTRA (-FOC))) used to write signals that the first actant referent was, on each occasion, found in the middle of the given event. Sim­ ilarly, Irish pluri-occasionality constructions can distinguish "between a progres­ sive event which is taking place at the moment of speaking and a similar event which is performed regularly over a longer period of time" (Ö Baoill 1994: 202), e.g., -PAST (+PLUR (+INTRA°(+FOC))) bionn ag scriobh 'is usually writing' and +PAST (+PLUR (+INTRA (+FOC))) bhiodh ag scriobh 'used to be writing'. In Lithuanian, +PLUR may also combine with the so-called Continuative, e.g., budavo berasqs 'used to be about to write', budavo beparasqs 'used to be about to write to completion'. The Maltese +DISP (disposition) marker ihm signals, with the Progressive +INTRA° (+FOC), disposition to high-focal intraterminality, e.g., -PAST (+DISP (+INTRA° (+FOC))) ikun qed jikteb 'will be writing', and, with the Imperfect +INTRA° (-FOC), it signals disposition to low-focal intraterminality, e.g., -PAST (+DISP (+INTRA0 (-FOC))) ikun jikteb 'he will write'.

92

Lars Johanson

7.5. Oppositions of higher and lower focality The degree of focality signalled in the oppositions may be higher or lower. While no absolute graduation is possible, a rough distinction between relatively high focality (HF) and relatively low focality (LF) may be useful. In some oppositions, the higher item basically stands for a narrow presentness and is not readily used for a expanded or open presentness, e.g., Icelandic —PAST (+INTRA° (+FOCHF)) er ад lésa bôkina 'is reading the book', er sofandi 'is asleep' vs. -PAST (+INTRA0 (-FOC)) léser bôkina 'is reading, reads the book'. Some Romance focals are more restricted than others. For example, Italian copula + Gerund items seem to represent a higher focality than motion verb + Gerund or copula + locative Infinitive items. Turkic languages of Europe display oppositions of a relatively low focality, e.g., Turkish -PAST (+INTRA0 (+FOCLF)) yaziyor 'writes, is writing' vs. -PAST (+INTRA0 (-FOC)) yazar 'writes, will write'; +PAST (+INTRA (+FOCLF)) yaziyordu 'was writing, wrote' vs. +PAST (+INTRA (—FOC)) yazardi 'wrote, would write' (Johanson 1971: chapter 5, 1994: 261-262). A similar picture is found in some non-Turkic languages standing under strong Turkic influence, e.g., Lezgian: -PAST (+INTRA0 (+FOCLF)) fizwa 'goes, is going' vs. -PAST (+D4TRA0 (-FOC)) fida 'goes, will go'; +PAST (4-INTRA (+FOCLF)) fizwaj 'was going, went' vs. +PAST (+INTRA (~FOC))fidaj 'went, would go'. The unmarked -FOC items represent a highly general, 'open presentness', which includes pluri-occasionality and also type-reference, genericity, extratemporality, potentiality, disposition, inclination, tendency, intention, etc. The nonfocal value is thus partly modal. While being neither Habituais nor Futures, the items in question are, by virtue of their value, easily interpreted as having habitual or future time reference. This is a type found in several European languages. A relatively low-focal Maltese intraterminal, signalled by the Imperfect 4- qed, is opposed to a nonfocal intraterminal, expressed by the bare Imperfect, e.g., -PAST (+INTRA0 (+FOC)) qedjikteb 'he is writing' vs. -PAST (+INTRA° (-FOC)) jikteb 'he writes, will write'; +PAST (+INTRA (+FOC)) kien qedjikteb 'he was writing' vs. +PAST (+INTRA (-FOC)) kien jikteb 'he wrote, would write'. The nonfocal is often taken for a habitual, though it also gets modal and future time readings. It is a defocalized item similar to the Turkic type mentioned. In some cases, the qed-marked item is used in a still less focal sense, i.e. for pluri-occasional subevents, albeit with a meaning of temporariness, e.g., qedjorqod 'is (usually) sleeping (now as opposed to earlier)'. This occurs with initiotransformatives forming high-focal +POST items of the type rieqed 'has fallen asleep' = 'is asleep, is sleeping' (see 10.3.1.2). Northeast Caucasian (Nakh-Dagestanian) languages such as Andi, Avar, Archi, and Lezgian possess similar nonfocals of the type -PAST (+INTRA° (-FOC)). The Lezgian -da form has general and partly modal functions interprétable as 'habitual',

Viewpoint operators in European languages

93

'future', etc. It has therefore been characterized as a "Future" with a "future / habitual polysemy" (Haspelmath 1994: 276). The above-mentioned Khvarshi item kulse 'throws, will throw' displays analogous uses. In several Iranian languages such as Talysh and Tati, the +PAST (+INTRA (-FOC)) items get similar habitual and modal (potential, counterfactual) readings, e.g., Talysh ahândi 'read, used to read, would read'. The further the focal item develops along the defocalization path (7.8), the stronger the modal meaning of the unmarked member will be. Certain languages distinguish more than two degrees of focality. In particular, high focality can be renewed, when the lowest item has developed so far along the defocalization path that it rather belongs altogether to the modal area. Some Turkic languages of the area have renewed high focality by means of periphrases with auxiliary verbs such as yati'r 'is lying' and turur 'is standing'. Besides the Nogai opposition -PAST (+INTRA° (+FOC)) yazayatïr 'is writing' vs. -PAST (+INTRA° (-FOC)) yazadï 'is writing, writes', there is a third lower item that corresponds formally to Turkish yazar but surpasses it with respect to modal content, —PAST (+MOD) yazar 'will, may, tends to, is likely to write'. Compare Karachai ala turadï 'is taking', aladi 'is taking, takes', alïr 'will [etc.] take'. There are corresponding +PAST items, e.g., Karachai ala tura edi 'was taking', ala edi 'was taking, took', alïr edi 'would [etc.] take'. Turkish also has, besides yaziyor and yazar, a third item which, at least in certain discourse types, represents a higher degree of focality, yazmakta(dir) 'is writing', yazmaktaydi ' was writing'. Similar trichotomies may be said to exist in Daghestanian languages, if their so-called continuatives, signalling 'is, was still V-ing' are taken to represent a high degree of focality. Albanian may express different degrees of focality by (i) copula + intraterminal gerund, (ii) the combination po + Present, and (iii) the plain Present. The narrowest "nunc" is expressed by the high-focal (i), e.g., është duke punuar 'is just working'. Degree (ii) is defined as nongeneral ("merkmalhaft nicht-generell", Buchholz & Fiedler 1987: 168) and is functionally rather close to the Turkish Present tense in -iyor. There are also +PAST items corresponding to both (i) and (ii).

7.6. Sources and shapes of intraterminals Intraterminals, like postterminals, usually emerge as high-focals by way of grammaticalizing statal expressions. The creation seems to be based on relatively few models, largely identical to the construction types of statal, progressive, or cursus-specifying modes of action mentioned under 4.2. In general, the verb takes a non-finite - participial, gerundial or infinitival - form, mostly provided with a copula. Typically, the point of departure is a nominal form °r the verb, e.g., writing, a preaspectual item outside the verb conjugation and orig»lally usable as an adjective. The aspectualization turns it into a genuinely verbal

94

Lars Johanson

form, e.g., is writing, directly related to write. Compare the preaspecrual development to postterminals (8.9). Some examples are the Late Latin esse periphrases, the English Progressive, e.g., Old English wœs writende 'was writing', Icelandic er sofandi 'is asleep', as well as the types represented by Khvarshi kulse goli 'is throwing' and Chechen molus vu 'is drinking' (intraterminal converb + 'is'). The diachronic development often seems to start with meanings of permanence and to pass through iterative stages. The relation between the viewpoint and the cursus is often expressed by locative metaphors of inessive or adessive nature ('be in', 'be at'), e.g., the older English construction is at V-ing, Finnish on lukemassa 'is reading', Danish er vedat arbejde 'is working', Dutch is aan het lezen 'is reading', Icelandic er ад lésa 'is reading', Basque egoiten da 'is staying', Portuguese esta a ir or esta indo 'is going', Turkish çahsmakta[dir] 'is working', Armenian gnum ë 'is going', Talysh kàrda (< kàrdedä) 'is doing', Lezgian fizwa 'is going', Bagvalal ig'iyax ira 'is doing'. Inessive constructions meaning 'in[side]' obviously provide the most natural expression of intraterminality. In Basque, the copula is 'be' with intransitives and 'have' with transitives (verbs requiring the ergative). Typical statives such as the postural verbs 'stand', 'sit' and 'lie' are often used. Many 'be' copulas actually originate in such verbs; see Italian sta cantando 'is singing' and the adessive Portuguese type esta a escrever 'is writing'. Some Turkic intraterminals go back to actional periphrases based on body position metaphors and consisting of converb in -a + turur 'is standing, is situated', e.g., Chaghatay qïladur (< *qïla turur) 'is doing', Chuvash yulat' (< *qala turur) 'is staying'. Kalmyk uses M 'stand, be' combined with the converb in -d. As we have noted, some modern Turkic languages of Europe have renewed high focality by means of yat'ir 'is lying'. The Maltese focalizing element qed goes back to qieghed, a participle of qaghad 'stay, sit'. Compare similar actional constructions in Germanic, e.g., Danish sidder og skriver 'sits / is writing', Swedish star och tittar 'stands / is looking'. Metaphors of movement are also well suited to express a 'dwelling' in the cursus. Romance languages such as Italian, Spanish, Catalan, and Portuguese use desemanticized locomotive items of the type 'go', 'come' -f- Gerund. Turkish ahyor 'is taking' goes back to an actional periphrasis converb + yorir ('moves'). Azerbaijani alir and Gagauz alär are of similar origin. Older intraterminals are usually known to us as synthetical formations and often exhibit more reduced, less transparent shapes with less clearly identifiable sources. Synchronically, some Present items even appear to have zero-markers, e.g., Adyghe sä-txi 'I write'. The etymologies of all these older items will not be dealt with here. Suffice it to say that some of them began their career as iteratives, duratives, intensives, etc., and then became statals and — T-markers. Indo-European seems to have renewed its present formations successively by means of statal items. Classical Greek so-called Present stems, e.g., épheuge 'escaped, was escaping', are opposed

! Viewpoint operators in European languages

95

to Aorist stems, e.g., éphyge 'escaped'. On this basis, Modern Greek has developed a full-fledged intraterminality opposition, enlarging the old stem formation to the effect that most verbs have a +INTRA stem opposed to a -INTRA stem, e.g., ghrafvs. ghraps- 'write'. There are also suppletive pairs such as trogh- vs.fagh- 'eat'. The Latin Imperfect legêbat 'was reading' has its regular reflexes in all Romance languages, e.g., Portuguese lia. The Slavic imperfective type in -ajç, which came to oppose the old Present, was of iterative origin. Iranian languages form intraterminality with elements such as mi-, di- prefixed to the stem, e.g., Kirmanji +PAST (+INTRA) dixebiti 'was working' vs. +PAST (-INTRA) xebitî 'worked'. Turkic languages formed older intraterminals in -(V)r, e.g., Turkish ahr 'takes' (later on generally renewed by means of periphrases).

7.7. The way to intraterminality Many European languages display items that have progressive-like meanings without being full-fledged aspectual +INTRAHF items. The development of high-focal intraterminals often starts with preaspectuals that preserve older iterative, durative or continuative meanings. Numerous progressive-like constructions in Germanic, Finno-Ugrian, Turkic, etc. exhibit uses atypical of grammaticalized +INTRAHF items. Some periphrases express 'be in the course of, e.g., French être en train de, Romanian afi in curs de, Danish vœre i gang med. Others use verbs meaning 'hold', e.g., Norwegian bokmâl jeg holder pa med à skrive, Yiddish ikh halt in shraybn 'I am writing'. Hungarian displays a progressive-like device based on inversion, e.g., ment ki 'was going out' vs. kiment 'went out'. It is created by placing the +T-marker of a transformative in a postverbal position and adding a specific accent pattern ('ment 'ki), e.g., éppen irta aid a levelet 'was just signing the letter' vs. éppen alâirta a levelet 'just signed the letter'. The Ossetic —T-marker -cœi also has similar progressivelike uses, e.g., racœicïdi 'was going out' vs. racïdi 'went out'. When a preaspectual progressive-like item develops into an early aspectual -t-INTRA™7 item, there are stages where it exhibits both actional and aspectual features. For cases of apparent ambiguity in Turkic, see Johanson (1995). The process in question must also be distinguished from subsequent defocalization processes (7.8), which, in some cases, might even seem to take the reverse direction. Examples of these bidirectional tendencies will be given below. Some languages possess special high-focal continuative items signalling that an intraterminal state has begun and is continued, 'is, was still V-ing', e.g., Catalan segueix treballant 'is still working' (seguir 'follow, continue' + Gerund). Daghestanian items of this kind carry markers such as Andi -guza, Archi -mat, Lezgian -z-ma, e.g., Lezgian ksuzma 'is still falling asleep' as against the non-continuative high-focal ksuzwa 'is falling asleep' (Haspelmath 1994: 269). Kalmyk uses non-

96

Lars Johanson

past and past continuatives meaning 'keeps, kept V-ing', 'is, was still V-ing', e.g., köalä bilä 'was still working'. The Lithuanian so-called Continuatives have different functions, but high-focal continuativity in the non-past stratum may be expressed by combining be- with the particle te-, e.g., tebèra (be)rasas 'is still writing' (Mathiassen 1996). Preaspectuals are less grammaticalized in the sense of having a less generalized use and being easily replaced by unmarked items to refer to ongoing events, even within the so-called incidential schema. This is the case with most Germanic progressive periphrases. Not even items such as war am Essen 'was eating' in spoken German varieties are generalized enough to be included in the core systems. Their optional use for ongoing events indicates that they are less grammaticalized (Johanson 1975: 150). Among similar items are Baltic Finnic inessive constructions, e.g., Estonian oli tegemas 'was doing', and the Hungarian progressive device based on inversion, e.g., ment ki 'was going out'. Several Romance items, e.g., the Italian stava per periphrasis, are often easily replaced by simple Imperfects of lower focality. Some other typical differences may be observed. Preaspectuals often combine with 'in X time' adverbials, whereas +INTRAHF items do not. Many preaspectuals such as Latin esse periphrases, Germanic postural statals, Swedish hâlla pâ att I och constructions, Turkic postverb constructions with tur- 'stand', may express durativity in addition to their statal meaning, whereas -l-INTRA™3 items do not. Preaspectuals may also apply to [—dyn] actional contents, whereas +INTRAHF items require dynamicity. Combination with [—dyn] may yield derogative connotations of impatience with the intensity or perseverance of the event. Such meanings result from durativity, i.e. an actional notion. Preaspectual postural verb constructions may preserve some of the original meaning of the auxiliary and thus be restricted to actions performed in a certain body position, e.g., Dutch zit te kijken 'is sitting and watching', Finnish istuu työskentelemässä 'is sitting and working'. In +INTRAHF items, the auxiliaries are desemanticized and thus may be interchangeable. For Turkic postverbial constructions, see Johanson 1992: 30, 1995. Some preaspectuals tend to be closely connected with agentivity, intention, and purpose. -(-INTRA1115 items, as genuine viewpoint operators, do not require the event to be purposeful and thus easily combine with non-intentional actional phrases. There are locative-purposive items implying 'dwelling somewhere in order to V , in which an intentional notion ('with the purpose of V-ing') is added to the meaning of ongoing event ('be V-ing'), e.g., the Finnish construction Copula + -ma-Infinitive + Inessive. For so-called absentive constructions ('be away V-ing'), see Ebert (this volume), De Groot (this volume), and Tommola (this volume a). Locative-purposive constructions are by definition infertile with non-agentive, non-intentional actional phrases. Imminential constructions such as German war im Begriff zu schreiben or Lithuanian buvo berasqs 'was about to write' often seem to reside in the immediate neighbourhood of high-focal intraterminals or to move in their direction.

Viewpoint operators in European languages

97

While preaspectual items may occur in habitual contexts, i.e. be compatible with pluri- occasionality, +INTRAHF items tend towards uni-occasionality. The copula 4- gerund constructions of the Ibero-Romance languages are not restricted to unioccasionality and thus preserve the old stage found in Late Latin esse periphrases (Dietrich 1973). The same seems to be true of the Finnish locative-purposive construction. However, a subsequent defocalization will increase the tendency towards pluri-occasionality, which is an example of the bidirectional tendencies mentioned above. 7.7.1.

Peripheral periphrases

Whereas the distribution of +INTRA items is restricted, their components may exhibit a relatively high combinability and occur in various peripheral items of a lower degree of grammaticalization. For example, the elements of Late Latin esse periphrases have higher combinability than the high-focal items of modern Romance languages (Dietrich 1973). As mentioned above, some languages use intraterminal elements referring to events that are foreseen to be the case at a posterior O 2 , i.e., -PAST (+PRO) items may combine with items of intraterminal nature, e.g., English will be writing, Modern Greek tha ghrâfi 'will be writing'. Similar combinations are possible with certain Romance progressives, e.g., Italian stare + Gerund, the Lithuanian Continuative, the Turkish -mekte olacak periphrasis, etc. Peripheral items of this kind cannot be dealt with at length in the present overview. Note, however, that auxiliaries involved in complex constructions may themselves take part in +INTRA and +POST oppositions. The latter will be dealt with in Section 8.9. As is well known, intraterminal participles also take part in lexicalization processes of adjective formation, e.g., English charming. The whole combinatorics of such elements must be omitted here. One important point should, however, be mentioned. In the +INTRAHF items dealt with so far, the copula is a non-dynamic auxiliary ('is', 'was'). They imply non-dynamic, inessive readings in the sense of 'dwelling in the cursus'. Some languages exhibit apparently similar constructions with dynamic auxiliaries ('becomes', 'became'), in the illative sense of 'entering into the cursus', e.g., English get going, Turkish gelir ol- 'set out' (literally 'get coming'). The dynamic statals resulting from these combinations have a rather modest distribution in modern European languages, ihe Turkish type does not, in the contemporary language, express an entering into a uni-occasional event and is thus not high-focal (Johanson 1971: 190-193). Similar differences are also found in certain Romance constructions. In the Poruguese periphrasis estava a escrever 'was writing', which expresses a past dwelling ln an intraterminal state, the auxiliary is a non-dynamic +PAST (+INTRA) item. In esteve a escrever 'wrote', however, the auxiliary is a dynamic +PAST (-INTRA) 1e m. Constructions of the latter kind are indeed not intraterminals, since the event is

~э

98

Lars Johanson

not envisaged in progress at an O, but nonintraterminals, rather functioning as delimitatives in the sense of 'come to spend a certain time V-ing'. They can consequently be used with mensural 'until t x ' and 'for X time' adverbials, e.g., Spanish estuvo escribiendo todo el dia 'wrote the whole day'. Dynamic postterminal constructions will be dealt with in Section 8.9. 7.7.2. The Greek development The Greek development from a preaspectual actionality duality into a +INTRA op­ position is particularly interesting. The Classical Greek Present vs. Aorist distinction was, as we have noted, originally an actional one, but it developed into an aspectual one, the nontransformatives turning into intraterminals and the transformatives into nonintraterminals. The Classical Greek duality is not yet an IPFV opposition in the sense of a viewpoint distinction. The Aoristic [+t] items imply a crucial limit. The [+tf] items imply a final one and may thus produce a so-called "effective mean­ ing", e.g., drew 'gained, acquired', épeise 'persuaded'. The [+ti] items imply an initial limit and may thus yield an "ingressive meaning", e.g., edâkryse 'burst out crying', ekhârë 'got joyful', ebasileuse 'became regent'. On the other hand, the [—t] items do not imply any limit and thus have a more general meaning of mere occupation with the action, e.g., edâkrye 'cried, was crying', ékhaire 'rejoiced', ebasileue 'reigned'. Classical Greek [—t] items are therefore more easily used for single accomplished events than Modern Greek +INTRA items. Even [—t] items such as épeithe 'persuaded / has persuaded' only denote engagement in the action and do not signal either conation ('tried / has tried to persuade') or full accomplishment ('convinced/ has convinced', 'succeeded / has succeeded in persuading'). The interacting Aorists have a marked [+t] meaning and consequently a more limited frequency. Thus, [—t] epeskeûaze rather means 'was / has been busy with the production', whereas [+t] epeskeûase means 'produced / has produced'. These past items are the only eventoriented ones, since the Perfect was still a high-focal postterminal. In early narratives, we often find sets of [+t] items denoting discrete events, e.g., éteke 'gave birth to' or élipen 'handed over', and a concluding [—t] item, e.g. tikte 'gave birth to' or lelpe 'handed over'. The [—t] items are not inceptives (Koller 1951), since they do not indicate the beginning of the actions expressed by tiktein 'give birth to' and leipein 'leave, let'. They are more likely to highlight the consequences of the event at O s (Ruipérez 1954: 86) or, more precisely, to suggest the opening of a consequence ("une ouverture possible sur un développement ultérieur"; Seiler 1993: 28). The use is based on the simple fact that the Imperfect has the default value [—t], whereas the Aorist is transformative, i.e. [+tf] or [+ti]. Since transformatives by definition imply a transition, the Aorist may suggest a situational change in the propulsive sense: 'did (and then)', which is not possible with the

Viewpoint operators in European languages

99

unmarked, nontransformative Imperfect. The unmarked [—t] item is also used for constative functions, since Classical Greek lacks a low-focal +POST item for such purposes. The development in Homeric, Classical, and Hellenistic Greek cannot be dealt with here, but it should be noted that Modern Greek exhibits a clearly different opposition with a wider applicability and essentially changed values. The actionality of the verb pairs has developed into corresponding viewpoint values, though the orientation of the system is inverted. The Imperfect has developed into a +INTRA item with a narrower use, the Aorist into a -INTRA item with a broader use. This is a more definite change to a genuine viewpoint opposition limited to "la vue d'un procès" (Seiler 1993: 27).

7.8. Defocalization of intraterminals High-focal intraterminals tend to develop into less focal items, acquiring more general and finally modal (prospective, subjunctive, etc.) functions along the line -blNTRA™ > +INTRALF > +INTRANF > +MOD. Cases of semantic fusion in the sense of a general +INTRAF item and other combined items are frequent. This development is a defocalization drift and not a process of perfectivization. It is often accompanied by morphological reduction. As we have noted, newly created marked items are more likely to be expressed periphrastically than older and unmarked ones, which often have reduced shapes and less clearly identifiable sources. The diachronic development of high-focals expanding their range of uses to assume lower functions is rather well-known. When items of higher degrees are degraded, the higher focality is renewed by new items that leave the old items with lower functions. Innovations of higher focality encroaching on the area of —PAST (+INTRAoLF) items constantly lead to changes of whole systems. Old "worn out" intraterminals may also finally disappear. Compare the similar encroachment upon -PAST (+POSTLF) items by high-focal postterminals (8.10). The development from high-focals into more general Presents and Imperfects is known from many languages. Maltese jikteb 'writes', together with kiteb 'has written, wrote' reflecting the West Semitic ±INTRA opposition (cf. Standard Arabic yaktubu vs. kataba), goes back to a +INTRAHF item, a participial stative of nommai origin, which later on penetrated the low-focal area of a former Imperfect, thus assuming the whole +INTRA range. It is now a +INTRALF item, the high-focality having been renewed by qed jikteb 'is writing'. The Slavic pripëkajç 'burn', originally an iterative and then a renewer of higher focality, drove the old Present pripekç mto the modal domain of prospectivity (Kurylowicz 1956: 28). In Irish, Scots Gaelic and Welsh, periphrases have pushed the old simple Presents into the areas of lowand nonfocality. In Turkic, former high-focals have become low-focals, pushing the

100

Lars Johanson

former low-focals, the so-called Aorists, into the nonfocal and modal domains (Johanson 1971: 131-139). In Armenian, the originally high-focal hum ë 'is hearing' has been degraded to a low-focal and relegated the old Present he 'hears' into nonfocal and modal functions. Persian mikunad 'is doing, does' has similarly pushed the former low-focal kunad into the modal area. The Lezgian intraterminal in -z(a)wa, c.g.,fizwa 'is going', which once renewed high focality, is now a low-focal, while the item in -da, e.g.,fida 'will go, goes', has assumed nonfocal ('habitual') and even modal ('future') functions, possibly under Azerbaijani influence (Haspelmath 1994: 276). Old intraterminals have vanished in several languages. In Basque, the older Present item is used with only about a dozen of verbs (Haase 1994: 279, 284.). The old Kartvelian so-called Permansive, belonging to the Aorist system, once had lowand nonfocal functions (habitual, general, timeless events), but is absent in modern literary Georgian, where the lower functions are now fullfilled by the Imperfect. Defocalization often seems to take place earlier with —PAST than with +PAST items. Turkic -f PAST items of lower focality have generally preserved more indicativity than the corresponding —PAST items. High-focal intraterminals are mostly confined to certain types of internal phase structure. While preaspectual progressives of an actional nature may also apply to [ dyn] actional contents, genuine -HNTRA™1 items start with [+dyn] actional phrases and are generally infertile with [—dyn]. However, when an -HNTRA1^ item begins to step down the focality scale, it tends to tolerate weaker dynamicity again. When it has expanded to cover all actionality classes including [+tf, +mom], it has eo ipso ceased to be a high-focal. Similarly, while preaspectual progressives are often also used with non-agentive actional contents, -t-INTRA™1 items seem to start in agentive contexts. The compatibility with non-agentive actional phrases increases further down the focality scale. Finally, while preaspectuals may be compatible with pluri-occasionality, -(-INTRA™5 items tend towards uni-occasionality. In the course of a subsequent defocalization, a tendency towards pluri-occasionality is observed. These cases are examples of the bidirectionality mentioned, involving one development from actionality into -l-INTRAHF and a reverse development within the defocalization process. It has already been noted that operating on the subevent level is typical of preaspectuals developed from modes of action, which modify single basic actions, and that this option may also be observed with + INTRA™3 items. Lower intraterminals operate on global events. When a former low-focal yields its central function to an expanding high-focal, it may be left with a residue that is nonfocal and even more or less modal (Johanson 1971: 138-139). It implies disposition (potentiality, inclination, attitude, etc.) and is thus easily interprétable as having habitual or future time reference, though it 1s neither a Habitual nor a Future. As has already been noted, Turkic languages of Europe possess dispositive items such as Turkish yazar 'will write, tends to write .

Viewpoint operators in European languages

101

+PAST markers turn them into dispositive pasts and counterfactuals, e.g., yazardi 'would write'. Items situated at the end of the defocalization path display strongly modal meanings, e.g., Tatar bafir 'will go' (vs. bara 'is going, goes'), Chuvash yule 'will remain' (vs. yulat' 'is remaining, remains'). As we have also noted, the Lezgian so-called Future in -da (Haspelmath 1994: 271, 276) is claimed to exhibit "future / habitual polysemy". This is.not surprising, since it is a nonfocal item with modal functions interprétable as future time reference. It would be more astonishing if it were a Future that expressed habituality. As Haspelmath remarks, the corresponding +PAST item "cannot, of course, express past future situations", and thus only has "the past habitual reading" (1994: 271). This is, however, a restriction we would not expect from a Future. Genuine -PAST (+PRO) items usually have +PAST (+PRO) counterparts, prospectives-in-past such as Turkish yazacakti 'would write, was going to write', Bulgarian stese da cete 'would read, was going to read'. Tendencies towards defocalization are observed in different parts of the linguistic map of Europe. Many of today's low-focals are results of the development -PAST (+INTRAHF) > -PAST (+INTRALF). Turkish, whose -iyor items have undergone this development, is far from unique in this respect, but rather typical of an eastern area in which defocalization and renewal of focal intraterminality have taken place in particularly systematic and observable ways.

7.9. Loss of ±INTRA categories ±INTRA distinctions emerge, undergo various changes and may finally vanish. When studying such developments, it is important to distinguish changes of formal markers from cases where functional oppositions as such cease to exist. A phenomenon of the former kind, observed in Romance and other languages, is that a former —PAST (+POST) item, e.g., French passé composé, takes over a nonintraterminal function from an older +PAST (-INTRA) item, e.g., French passé simple (Section 8.10.2). A phenomenon of the latter kind is the loss of the +PAST (±INTRA) distinction, signalled by the Imperfect and the Aorist, in most Slavic languages. This destruction started with the loss of the Imperfect, whereas the Aorist has usually been preserved lo nger as a formal item, before being replaced by a former -PAST (+POST) item. The + PAST (±INTRA) distinction is only maintained in South Slavic languages, ulgarian, Macedonian and conservative Serbo-Croatian varieties, e.g., Bulgarian S edase 'was looking' vs. gleda 'looked', Serbo-Croatianpisase 'was writing, wrote' s - Pisa 'wrote'. Its use in written Croatian is stylistically marked as archaic. Though °me Serbo-Croatian dialects have preserved the Aorist, the Imperfect has usually ^appeared and the opposition therefore ceased to function. Thus, the spoken stanar d Serbian "Aorist" is just a +PAST item. There are even signs of a merger between

102

Lars Johanson

Aorist and Present tense forms. Interestingly enough, Italo-Croatian has preserved the Imperfect but lost the Aorist. Here, the old opposition is carried on by means of an item developed from a former -PAST (+POST) item and taking over the +PAST (—INTRA) part. This situation is obviously due to contact with Italian, which has a similar ±INTRA distinction. The +PAST (±INTRA) distinction also seems to be endangered in spoken Romanian, where both the Aorist {perfectul simplu) and the Imperfect (imperfectul) tend to be replaced by the former -PAST (+POST) item (per­ fectul compus). The Sorbian +PAST (±INTRA) distinction remained relatively long, but is now lost. The early reduction of verbal categories in East and West Slavic languages and the loss of the +PAST (±INTRA) distinction laid the ground for the development of the +AD distinctions found there. The intraterminal and nonintraterminal items of the past strata vanished. The inherently intraterminal non-past items left underwent a defocalization process down to the nonfocal stage. The postterminal categories to be dealt with in Section 8.10 underwent a corresponding defocalization process down to the nonfocal stage. On this basis, the new idiosyncratic viewpoint distinction of adterminality was established (9.4).

8. Postterminality 8.1. Definition

У

The next viewpoint opposition to deal with is that of postterminality vs. nonpostterminality. Postterminality, +POST, envisages the event after the transgression of its relevant limit, post terminum. Transgressing the limit means going beyond it and not only reaching it in the adterminal sense. Nonpostterminality, —POST, disregards this view. Postterminals are used to describe events as observed from different orienta­ tion points situated after the initium or the finis, depending on the phase structure type. The initium is the relevant limit of initiotransformatives and nontransformatives, and the finis is the relevant limit of finitransformatives. +POST items may thus present an event from an orientation point outside its cursus. Postterminality is, like intraterminality, a basic mode of presentation emerging from the narrow perspective of the primary "nunc", and as such a natural way of conceptualizing the flow of events. In its focal forms, the postterminal perspective is tied to an О at which no event is grasped in its totality. No past phase of the event appears in the range of vision so that it might be regarded directly. This indirect way of presentation implies that the event is, entirely or partly, already out of sight. As with all aspects, however, the phases not envisaged are only "concealed", i.e. latent. Postterminals do not negate the realization of what they exclude from their more or less narrow view.

Viewpoint operators in European languages

103

+POST focuses the attention on a situation obtaining beyond the relevant limit, where the event, whether totally or partially past, is still relevant in one way or an­ other, i.e. extends right up to O, has effects relevant to O, or allows a conclusive judgment at O. The relevance of the event to О does not mean that it terminates at O. Nor does the general ^fPOST definition include notions such as 'change of sit­ uation', 'state', 'result', or 'recentness'. If +POST items suggest a change leading to a new situation, this is an effect of transformativity and dynamicity. Whether or not the postterminal situation is conceptualized as a state depends on how strongly it is focused, i.e. on the focality degree. If the action itself is telic and resultative, the postterminal situation can be said to result from it. Even if the event has al­ ready disappeared from the range of vision, it may have left traces, vestiges, effects observable there. Thus, the well-known fact that children usually acquire —PAST (+POST) items (e.g., British English and Swedish Perfects) earlier than other past tenses is explained by the naturalness of the postterminal perspective on past events with perceptible effects. However, this kind of resultativity is not part of the general definition. Nor is it decisive whether such possible effects of the event have already been overruled. Finally, it is also unessential whether the event is temporally located immediately prior to О or in a more remote past. +POST items of some focality are compatible with and often accompanied by relational adverbs indicating the diagnostic dimension and meaning 'already' = 'not later than O', e.g., has already left. In languages lacking focal postterminals, such lexical markers give a diagnostic interpretation to other past items.

8.2. Postterminals as anteriors While +POST can of course be interpreted in temporal terms, it cannot simply be defined as anteriority relative to O, in the sense of "the position of E relative to R" (Reichenbach 1947: 297). It does not just form 'relative tenses' denoting events that are completed before O. Whether the whole event or just part of it may be interpreted as having occurred prior to О depends on the internal phase structure. The feature +POST is an instruction to situate the relevant limit of the event posterior to an O, the point of relevance. While +POST has a natural affinity with the representation of past events, it is not equal to the feature +PAST. A -PAST (+POST) item can, by virtue of its viewpoint value, refer to an event partly or wholly prior to O s . It consequently involves a component of anteriority as opposed to -PAST (+INTRA) items, which allows it to compete with +PAST (-POST) items in the simple past stratum. On the other hand, H does not signal remoteness in the sense of +PAST items. Focal +POST items typically remain in the realm of the given deictic centre and do not establish own deictic centres, as +PAST (-INTRA) and +PAST (-POST) items do.

104

LarsJohanson

Comrie's description of the "perfect" (PF) as a relation between a state and a previous situation (1976: 52, 56, 62) may seem to concern an event-external time relation and thus to be a tense definition according to his own criteria. It has been criticized as contradictory that an aspect is taken to describe something as being prior to some point on the time-axis. The contradiction disappears if it is not the event itself that is situated anterior to O s , but if the postterminal perspective on it is situated at O s . The definition of PF as a +POST item does not vitiate the distinction between aspect and tense. Anterior items can be interpreted more or less diagnostically or historically, i.e. as focusing more on the О (orientation point) or on the L (localization point) of the temporal relation. Postterminals tend to be diagnostic and are thus mostly used as non-narrative devices. Focal —PAST (+POST) items are rather typical of the syn­ chronic report, and their occurrence in narrative discourse is relatively low. Focal -fPOST items are, due to their indirect way of viewing events, non-propulsive, illsuited to advance the plot in the way historical items do. When used in narratives, they typically represent events which, at '4opic time" (O2), are accessible to retro­ spection only and thus do not belong to the narrative event chain at a given discourse level. This of course does not mean that the events they refer to are unessential for the narrative in question. With decreasing focality, +POST items become increas­ ingly historical, without focusing explicitly on the localization point. They often refer to events outside strict temporal and sequential settings and are less inclined to combine with expressions that specify the occasion on which the event has taken place.

8.3. Postterminality oppositions Postterminals are extremely widespread. Even all known créole systems exhibit +POST and —POST items, though the latter are less temporalized. In the temporalized European systems, items expressing natural viewpoints such as intra- and postterminality get contrastive values only if they have competing categories to interact with in the same temporal stratum. This is, for example, the case when a language has both a primarily oriented postterminal-in-present and a corresponding nonpostterminal past item. Such oppositions are found in many European languages: English, North Germanic, Modern Greek, Albanian, Bulgarian, Macedonian, Iranian, Armenian, Kartvelian, Finno-Ugrian, Turkic, etc. A competing nonpostterminal past item opposed to a -PAST (+POST) item will be written as +PAST (-POST). Examples of the oppositions of the type -PAST (+POST) vs. +PAST (-POST) are Icelandic hefur talad 'has spoken' vs. taladi 'spoke', Faroese er komin 'has come' vs. кот 'came', Norwegian har skrevet 'has written' vs. skrev 'wrote', Swedish har rest bort 'has gone on a journey ' vs. reste bort 'went on a journey', Sami lœ bâradan

Viewpoint operators in European languages

105

'has eaten' vs. bâradii 'ate, has eaten', Estonian on teinud 'has done' vs. tegi 'did', Vepsian tonu 'has brought' vs. toi 'brought', Low (Meadow) Mari ludin 'has read' vs. ludo 'read', vozen 'has written' vs. voz'is 'wrote', High (Hill) Mari siren 'has written' vs. sirïs 'wrote', Udmurt m'inäm 'has gone' vs. mi'niiz 'went', Komi-Zyryan munorna 'has gone' vs.^munis 'went', Latvian ir rakstïjis 'has written' vs. rakstiïja 'wrote', Standard Italian ha cantato 'has sung' vs. canto 'sang', Spanish ha hablado 'has spoken' vs. hablö 'spoke', Catalan ha vist 'has seen' vs. va veure 'saw', Basque mintzatu dut I have spoken' vs. mintzatu nuen 'I spoke', Bulgarianxodil e 'has gone' vs. xodi 'went', Modern Greek éxipési 'has fallen', épese 'fell, has fallen', Albanian ka vrarë 'has killed' vs. vrau 'killed', Armeniangrac ë 'has written' vs. grec 'wrote', Sorani nûsîwa 'has written' vs. nûsî 'wrote', Kalmyk urns} 'has read' vs. umsv 'read', Karachai alyandi 'has taken' vs. aldï 'took', Lezgian fenwa 'has gone' vs. fena 'went', Chechen mella 'has drunk' vs. melira 'drank', Georgian damiceria 'I have written it' vs. davçere T wrote it'. —PAST (+POST) items may differ a great deal from each other with respect to how systematically they are used. Thus, the Lithuanian item yra rasçs 'has written', is relatively rare in everyday usage, the Simple Past rase 'wrote', being the dominant item in the simple past stratum. This is an effect of a relatively high degree of focality of the -PAST (+POST) item (Section 8.5). Similarly, since the Irish -PAST (+POST) is a high-focal item, ta éis ...a scriobh 'has just written', the +PAST (—POST) item is generally also used in cases where English uses the Perfect or the Pluperfect, e.g., scriobh 'has written / had written / wrote'. The artificial language Esperanto has corresponding devices, e.g., —PAST (+POST) estas skribinta 'has written' vs. +PAST ( -POST) skribis 'wrote', though it would be difficult to determine the range of their use. Postterminals-in-present, often referred to as Present Perfect, Anterior Present, etc., invite the addressee to situate the relevant limit of the event before a non-past O. Unless there are contextual indications to the contrary, О will by default be identified as Os, i.e. as having primary orientation. Postterminals used with this orientation convey events whose relevant limit has been transgressed at Os (E* before Os), e.g., Modern Greek éxifiji 'has left'. Thus, by virtue of their viewpoint value, they may refer to an event prior or partly prior to O s , and, as such, compete with +PAST (—POST) items in the simple past stratum. +PAST (-POST) items are characterized by the absence of postterminality, which makes them more event-oriented and suitable as narrative pasts. Since their perspective is not tied to an О that narrows the range of vision, they freely survey the flow °f events and may refer directly, in a historical way, to an event at the very interval °f its realization, independently of its relevance to any simultaneous or following O. и a +PAST (-POST) item is not intraterminal, it may present the historical event in­ cluding its limits in an integral way, suggesting, unless the contrary is signalled, the °ccurrence of the event as an unanalyzed totality. Note that this is possible without

106

Lars Johanson

the marked totality explicitly signalled by +AD items, which highlight the crucial limit. A historical +PAST (—POST) item emancipates the event from the deictic centre of O s , creating a deictic centre of its own, an O 2 somewhere in the text world. +PAST (—POST) items are thus typically used for situating events and often refer to events conceived of as occurring on specific past occasions. Consequently, they are less compatible with markers of indefinite, unrestricted or generalized time ('ever', 'always', 'never', etc.) or with O-implying adverbs meaning 'already', 'not yet', 'just', 'now', etc., e.g., Norwegian allerede, French déjà, Italian già, Spanish ya, Finnish jo, Lithuanian jau 'already'. However, they readily combine with temporal expressions specifying the interval of realization, e.g., Norwegian d0de i gar 'died yesterday'. +PAST (—POST) items do not break the linear successivity, since they are propulsive, capable of narrating chains of events and advancing the plot in narrative discourse. They may, of course, also refer to isolated events outside such chains. The decisive point is that they present an event in a more direct way than focal —PAST (+POST) items do. +PAST (-POST) items are thus not "simple Pasts", usable for all past events. The tasks just sketched clearly result from the value —POST, i.e. independence of an О situated after the relevant limit. They differ considerably from items that may present past events both in a historical and a diagnostic way, e.g., Hungarian lopott, Maltese seraq, Turkish çaldi 'stole, has stolen'. Two types of +PAST (—POST) items will be distinguished below, one plain type and one +PAST (-POST (-INTRA)) type (11.6.2-3). Some languages exhibit similar oppositions in the pre-past stratum, distinctions between a non-dynamic-diagnostic item ("plusquamperfectum status") and a dynamic-historical one ("plusquamperfectum actionis"). One Romance example is the obsolete French opposition Pluperfect vs. Past anterior, e.g., avait fait vs. eut fait 'had done'. As the loss of the Simple past also led to the loss of the Past anterior, the Pluperfect was the only item to cover the whole pre-past stratum (diagnostic and historical functions). In many spoken varieties, however, the Past anterior has been replaced by a "passé surcomposé" a eu fait (literally 'has had done'), which, according to Harris, may be used as an optional "perfectivity" marker, e.g., quand il a eu payé 'as soon as he had finished paying' (1988: 229). Albanian has an analogous distinction between a Pluperfect and an "Aorist II", e.g., kishte hapur vs. pat hapur 'had opened' (Buchholz & Fiedler 1987: 130-131). Similar oppositions are Udmurt rriinäm val vs. mïniiz val 'had gone', veram val vs. veraz val 'had said', Chuvash kilпёссё vs. kilseccë 'had come', Turkish gitmi§ti 'had gone' vs. gittiydi 'went (once, long ago, etc.)', ~Lezgim fenwaj 'had gone' vs.fenaj 'went earlier' (cf. fenwa 'has gone' vs.fena 'went'; Haspelmath 1994: 267).

Viewpoint operators in European languages 8.4.

107

4-POST0 items

The О of the postterminal perspective may also, like the О of the intraterminal per­ spective, be separated from the time of encoding and situated at any other point on the time-axis. For example, it may be determined relative to another event (taxis). If the vantage point is a foreseen O 2 , the perspective is projected into the future, e.g., French Ce soir, j'ai terminé la lettre 'This evening, I will have finished the letter', Dutch WanneerX komt, is Yal vertrokken 'When X arrives, Y will have left'. In such cases, the postterminal-in-present is a -PAST (+POST0) item, as it lacks a competing historical item with a contrastive value. Postterminals-in-past (referred to as Pluperfects, Past Perfects, Anterior Pasts, Past Anteriors, Preterit Anteriors, etc.), convey events whose relevant limit is transgressed at a past O 2 , e.g., Classical Greek egegraphei 'had written', Modern Greek ixe ghrâpsi 'had written', Icelandic hafdi talad 'had spoken', Sami lœi bâradan 'had eaten', Estonian oli teinud 'had done', Vepsian ol'i tonu 'had brought', Low (Meadow) Mari vozen île 'had written', ludin île 'had read', Udmurt veram val 'had said', Komi-Zyryan munöma völi 'had gone', Latvian bija rakstljis 'had written', Latin cantäverat 'had sung', Catalan havia vist 'had seen', Italian aveva cantato 'had sung', Romanian mersese 'had gone', Portuguese partira or tinhapartido 'had left', Bulgarian bese daval 'had given', Tatar baryan idë 'had gone', Nogai barip edi 'had gone', Karachai alyan edi 'had taken', Chuvash kalanaccë 'had said', Turkish yemisti 'had eaten', Albanian kishte vdekur 'had died', Serbian bio je isao 'had gone', Kirmanji hatibû 'had come', Sorani nûsïbu 'had written', Chechen mellëra 'had drunk'. In general, so-called Pluperfects do not have the same semantic structure as socalled Perfects. They differ from them by signalling +PAST, but they are not just "Perfects related to the past", temporally transposed -PAST (+POST) items, the only difference residing in the orientation points (O2 vs. Os). Most European Pluperfects are +PAST (+POST0) items since they cover the whole pre-past stratum and lack a competing historical item with a contrastive value. They thus do not share the special properties of the corresponding -PAST (+POST) items, which have a more restricted use due to the competition with +PAST (-POST) items. For example, the Portuguese Pluperfect is not subject to the restrictions typical of the Perfect (8.10.2, 10.3.2). While a -PAST (+POST) item competes with a +PAST (-POST) item, there are seldom oppositions of the type Perfect-in-past vs. Preterite-in past (for exceptions, see °-3). There may also be more substantial differences between the non-past and past 'terns in question. Thus, Turkish has a Pluperfect in -misti, e.g., gitmisti 'had gone', whereas the corresponding simple finite form in -mis is a predominantly indirective lte m, e.g., gitmis 'has [apparently] gone'. e

A +PAST (+POST0) item invites the addressee to situate the relevant limit of the vent before a past O2 (E* before O 2 ). In X had left when Y came, the transgression

108

Lars Johanson

has taken place at the O2 marked by the adverbial when he came, whereas in X left when Y came the relevant limit, due to the +PAST (—POST) value of left, coincides with the O 2 . A time expression combined with a +PAST (+POST0) item may refer diagnostically to the O 2 , e.g., At two o'clock, X had [already] written the letter, or historically to the event time, e.g., X had written the letter at two o'clock. +PAST (+POST0) items are often used in narratives, when the chain of events is broken by an explanation or a regress that conveys a past or partly past event. +PAST (+POST0) items may also be employed without a secondary orientation to an O 2 . This allegedly "expressive" use "instead of" a simple Past indicates non-relevance or even nonvalidity at O s (8.5.2). Some reduced systems use one single general +POST0 item to cover both the simple past and the pre-past stratum, e.g., Hungarian irt 'has written, wrote, had written'. The lack of a +PAST marker may be made up for by lexical disambiguaters meaning 'earlier', 'just (before)', 'already', etc. Pluperfects are lacking in many Slavic languages, in particular in West, East, and Western South Slavic. Some of them may still use +PAST markers in certain registers, e.g., Ukrainian buv xodiv 'had gone', Belarusan pryjsou byu 'had come', Polish kupil byl 'had bought' (almost extinct). Numerous languages that do possess a Pluperfect often substitute simple postterminals for it in past narratives (for Slavic, see Maslov 1980: 54, 58-59). A +PAST marker of a Pluperfect may itself be a +PAST (+INTRA) item, e.g., French avait 'had', était 'was', a +PAST (—POST) item, e.g., Danish havde 'had', var 'was', or a +PAST (-POST (-INTRA)) item, e.g., English had, was. It may also be a -PAST (+POST0) item, in which case the main verb is a -PAST (+POST0) item as well, e.g., Maltese kien kiel 'had eaten', Yiddish iz gehat avekgeforn 'had gone away', hot gehat geshribn 'had written', Swiss German (Züritüütsch) isch ggange gsii 'had gone', Serbian bio je isao 'had gone'. Languages in which certain varieties have generalized Past tenses of the —PAST (+POST0) type, while others have not (8.10.2), may exhibit two corresponding Pluperfects, e.g., German hatte geschrieben and hat geschrieben gehabt 'had written'.

8.5. Degrees of focality Anterior items can be interpreted more or less diagnostically or historically. A diagnostic reading is O-oriented, focusing the attention on О and stressing the relevance of the event at this point, often supported by adverbials of the type 'already'. A historical reading is event-oriented, focusing on L and stressing the relevance of the event at the time of its realization, answering questions such as "When did it occur?' • However, the two dimensions are seldom neatly distinguished by morphosyntactic means. Most anterior items represent both sides, the reported event and its validity atO.

Viewpoint operators in European languages

109

Postterminals tend towards diagnostic readings, referring to events that are rel­ evant to an interval including O, e.g., has broken the window vs. the more eventoriented broke the window. However, they can also be more O-oriented themselves, focusing less rigidly on the postterminal situation. They thus display higher and lower degrees of focality,.concentration (focus) of the psychological interest on the situation obtaining at O, the core of "nunc". The range of vision may extend from the narrowest idea of an interval confined to the immediate proximity of О to much broader views. It has already been emphasized that focality is a scalar notion, that focality values are relative, and that the choice also depends on the conceptualization of the actional content. Still, it might be roughly assumed that events whose relevant limits precede О may be presented by +POST items in a (i) narrow, (ii) expanded or (iii) open way. The range may be (i) confined to O, the event being conceived of as relevant at О only; (ii) not confined to O, while the event is conceived of as still relevant at O; or (iii) not confined to O, while the event is not necessarily conceived of as relevant at O. Since the scalar focality of +POST items depends on the relative narrowness of the range of vision, it is reminiscent of the focality of +INTRA items. For example, socalled "resultatives" and "progressives" might be claimed to represent high focality in the postterminal and intraterminal dimensions respectively. In both dimensions, the basic question is to what degree the event is conceived of as part of the "O world". Since focality values are relative, we do not assume absolute functional stations such as "resultative", "perfect", etc. Nevertheless, a tentative focality scale with three cardinal degrees will be adopted here: relatively high focality (HF), relatively low focality (LF), and nonfocality (NF). Note again that the nonfocals are non-oppositive +POST0 items that represent the postterminal notion in a weak way. The interpretation of a +POST item in terms of diagnostic and historical orienta­ tion depends on its focality degree. By and large, high-focals might be said to cor­ respond to "statives" and "resultatives", low-focals to "perfects" and "constatives", and nonfocals to more general pasts. However, we assume gliding transitions and areas of overlap between the postulated types. The focality degrees are distributed to items in different ways, and individual items may also represent more than one degree. All language-specific +POST items are subject to constant functional shifts along the focality scale. Labels such as "perfect" and "resultativity" will be avoided, since they are often used in a more substantial sense. It seems impossible to establish absolute categories on the basis of unequivocal situational criteria. Our scheme does not reflect any attempt at classifying postterminals by means of fixed ontological and situational characteristics. Such interpretational types are a matter of interaction of as pect (of higher and lower focality) with different types of actional content (Section One word on focality degrees and coverage of situations must be added. Since a OST item of lower focality is more general than a higher one, it may be used in

p

110

Lars Johanson

cases where some other language permits or requires a higher item. It may cover ref­ erential areas of a higher item, i.e. be used instead of it to refer to a given situation, without expressing a higher degree of focality. In languages that do not distinguish focality degrees, items indifferent towards such distinctions may cover situations that would require higher focality in another language, e.g., German hat geschrieben Э English has written. A nonfocal item in language A may roughly cover situations that would require a low-focal item in language B, e.g., French a écrit D English has written. Lower focality does not exclude higher interpretations, e.g., stative readings of sentences such as has arrived = 'is now here' or Spanish ha comprado una casa 'has bought a house' = 'owns a house'. To render the static situation described in is already asleep I awake, individual languages may use high-focals, low-focals, nonfocals, or items indifferent to postterminality. A general Past covers the whole range of anteriority. The South German Past tense may cover situations that require focality in other languages. Thus, ist verschwunden 'has disappeared / disappeared' may correspond to three Swedish items representing the simple past stratum, —PAST (-l-POST^) ärförsvunnen 'is gone', -PAST (+POSTLF) har försvunnit 'has disappeared', and +PAST ( — POST) försvann 'disappeared'. Higher items may also cover lower levels in the sense of involving them logically. Thus, Russian vypito 'is drunk up' implies that the liquid in question has been drunk or was drunk on some occasion^). 8.5.1. High-focal postterminality -(-POST107 items are strongly O-oriented, putting high focus on the postterminal state obtaining at О after the transgression of the relevant limit of the event. Their range of vision is narrow, restricted to what is still relevant of the event in the "O world". High-focals are basically restricted to transformatives, i.e. [+tf] and [+ti] actional phrases. The high focality of the Indo-European Perfect is attested in Classical Greek, e.g., téthnëke 'is dead', gégraphe 'has written'. Other examples of —PAST (-(-POST105) items are Hittite appan harzi 'has taken', Lithuanianyra raSes 'has written', yra atèjes 'has come [and is here]', Romany hi mulo 'is dead', Turkish ölmiis bulunuyor 'has died, is dead', Karachai ketibdi 'is gone', East Armenian gnacac ë 'is gone', Kalmyk untsn 'is asleep', Akhvakh cankohe gudi 'has fallen ill', Agul aq'unaa 'is / has done', Basque joana da 'is gone'. Corresponding +PAST (+POST0 ) items are, for example, Hittite appan harta 'had taken', Lithuanian buvo atvykes 'had arrived [and was there]', Romany his mulo 'had died, was dead', Turkish ölmüs bulunuyordu 'had died, was [already] dead', Karachai ketib edi 'was gone', Kalmyk suusn bilä 'was sitting', Eastern Armenian gnacac er 'was gone'. What is often referred to in terms such as "l'état résultant d'une action accomplie (Kurylowicz 1956: 26) is thus not taken here to be an independent aspect but just to represent a high degree of focal postterminality. High-focals denote a dwelling in

Viewpoint operators in European languages

111

a postterminal state created by transgressing the relevant limit of the event and still prevailing at O. They often imply that something - the referent of the subject or of an object - has such properties at О that it can be concluded that the relevant limit, leading to this state, has already been transgressed at that point. However, readings with respect to 'result' vary according to lexical contents. For example, as Ruipérez shows, the pertinent value of the Classical Greek perfect cannot be "la consideration del estado résultante" (1954: 60). Since high-focals signal property-like states, the number of verbs occurring with them is usually limited to a rather restricted set. Among verbs of a suitable actional content are many denoting spatial and mental transformations, e.g., 'come', 'go out, 'hang up', 'lean', 'open', 'pass by', 'put', 'sit down', 'call to mind', 'convince', 'decide', 'discover', 'fall asleep', 'get accustomed', 'get tired', 'persuade', 'understand', 'wake up'. The posttransformational state may be reversible or not, e.g., German ist geöffnet 'is opened', ist zerrissen 'is torn'. The postterminal property-like states tend to be perceptible. For example, the event may be manifested by nothing more than traces left at O. This is the source of the well-known indirective meanings connected with many postterminals (8.7). High-focals may also be used for states not preceded by any corresponding real telic event ("quasi-resultative use"), e.g., Russian Dom okruzen lesom 'The house is surrounded by forest'. 8.5.1.1.

-PAST (-l-POST)™3 items

-PAST (+POST)HF items indicate that the posttransformational state obtains at Os and may thus get present-like readings, in particular with initiotransformatives, e.g., Classical Greek Perfect dédie '[has got frightened and] is frightened', kéktëtai '[has acquired and] owns', Archi ustullit q'owdili wi 'is sitting on the chair' (Kibrik 1977: 195). As is well known from school grammars, the Classical Greek Perfect corresponds to a Latin "perfectum praesens" and not to a "perfectum historicum". The Irish high-focal easily gets non-past interpretations, e.g., ta leabhar scriofa aige 'has written a book' or 'has a book written' (Ö Baoill 1994: 208). The high-focal Old Georgian Perfect is often considered a Present tense. Svan -PAST (+POST)HF items behave syntactically like Presents (Deeters 1930: 181-182). High-focal postterminals of Baltic languages are usually called "Compound Presents". Analogously, +PAST (-bPOST)101 items indicating that the state obtains at a past O2 may get imperfective-like readings. The syntactic behaviour of high-focals thus tends to be similar to that of intraterminals. However, -t-POST™3 items differ from INTRA™3 items by their reluctance to be used for future time reference. The highest focal uses imply a dwelling in the postterminal state without regard to the preceding telic event, however recent. Less high uses allow reference to the ev ent itself. While paying due attention to such differences, we shall not try to set U P absolute categories such as "statives" vs. "resultatives" (Nedjalkov 1983) on the

"1

112

LarsJohanson

[—t]. Constative uses are also typical of many varieties of American Spanish. The Turkish finite item in -mis is not a "resultative" Perfect but has, besides its inferential uses, also low-focal constative functions (Johanson 1971: chapter 8). The corresponding Azerbaijani -mis, however, has more "perfect-like" uses than the Turkish item, probably due to influence from the Persian -PAST (+POST)LF item, e.g., karda ast 'has done' = yapmis (Johanson 1988: 249). The same is true of some Kipchak-Turkic -gän items. Certain languages possess special negated low-focal constatives with the meaning of a perfectum nondum facti 'has not yet done'. 8.5.3. Nonfocal postterminality Nonfocal postterminals, -bPOST1415 items, do not focus on O, but denote a wideopen anteriority, not tying the event to the "O world" and not implying any particular O-relevance. -PAST (+POSTNF) items are -PAST (+POST0) items, since they lack competing nonpostterminals, general anteriors covering the whole diagnostichistorical range, usable as general retrospective items with primary as well as secondary orientation. Differing from focals by being suited for direct event-orientation, they are used as propulsive items in narratives and combine freely with specific time expressions. They denote single or repeated, uni- or pluri-occasional past events. With respect to actionality, the same tendencies are found as with focals: [+t] actional phrases are more suited to highlight the effect of the event than [—t] actional phrases. -PAST (-t-POST1^) items require special lexical markers to distinguish the diagnostic dimension from the historical one. The use of specific time expressions suggests historical readings, e.g., German hat um sechs Uhr gefrühstückt 'had breakfast at six o'clock', whereas devices of the 'already' type exclude them, e.g., hat schon gefrühstückt 'has (already) had breakfast', Czech uz napsal ten dopis 'has (already) written the letter'. Constative readings may be suggested by expressions of the types once', 'ever', 'already' (e.g., German schon, schon einmal, jemals, Hungarian mar, valaha), e.g., HungarianLâttad mâr a kutyâmat? 'Have you ever seen my dog?'. Nonfocals are found in most Slavic languages, in South German, Yiddish, Hungarian, several Romance varieties such as colloquial French, northern Italian dialects, colloquial Romanian, Romansh, etc. Some examples from verbs meaning 'write' are Serbo-Croatian ye pisao, Italian ha scritto, German hat geschrieben, French a écrit, Maltese kiteb. Some are general +POST° items that also cover the pre-past stratum, e -g-, Polish przeczytal 'has read, read, had read', Hungarian irt 'has written, wrote, had written'. Very many Pluperfects are nonfocals, i.e. +PAST (+POSToNF) items,

120

Lars Johanson

e.g., Dutch had gewerkt 'had worked', Estonian oli kirjutanud 'had written', Talysh händäS be 'had read', Kirmanji kiribû 'had done'. Nonfocals emerge from focal postterminals due to loss of nonpostterminals (8.10.2). They function as pasts capable of primary orientation and also maintain their original postterminal capability of secondary orientation. While sometimes called "temporal Perfects", they are not low-focal postterminals any more. Even Slavic past items taking part in ±AD oppositions, e.g., Russian (na)pisal 'has written, wrote, had written', are former focal postterminals that have lost their focality and become -PAST (+POST0) = +PAST items. These +PAST items might thus also be represented as -PAST (+POST0 (+AD)) and -PAST (+POST0 (-AD)) items, respectively.

8.6. Oppositions of higher and lower focality Postterminals may form language-specific oppositions with respect to the degree of focality: -PAST (+POST (+FOC)) vs. -PAST (+POST (-FOC)). The degrees signalled may be higher or lower. Examples of ±FOC oppositions in which the higher item is a -PAST (-fPOST107) one ('the posttransformative state is still prevailing at O s ') and the lower item a -PAST (+POSTLF) are English is gone vs. has gone, Irish ta leabhar scriofa aige 'has a book written' vs. ta tar eis leabhar a scriobh 'has [just] written a book', Icelandic er dottid 'has fallen (and is now lying)' vs. hefur dottid 'has fallen', French est maigri 'is emaciated' vs. a maigri 'has grown thin', Finnish ovat väsyneitä 'are tired' vs. ovat väsyneet 'have got tired', Swedish är bortrest 'is gone on a journey' vs. har rest bort 'has gone on a journey', North Macedonian e dojden 'has come (and is here)' vs. ima dojdeno 'has come', Albanian ështe i vdekur 'is dead' vs. ka vdekur 'has died', Karachai ketibdi 'has left (and is now gone)' vs. ketgendi 'has gone', Armenian grac ë 'has written' vs. grel ë 'has written', Lezgian fenma 'has, is still gone' vs. fenwa 'has gone', Chechen vaxana bu 'is gone' vs. vaxana 'has gone', Basque joana da 'is gone' vs.joan da 'is, has gone'. There are also ±FOC oppositions with passive items such as North German ist gestohlen 'is stolen (and now missing)' vs. ist gestohlen worden 'has been stolen'. Passive constructions of higher focality often contrast with active ones of lower focality, e.g., English is convinced vs. has convinced, Georgian icereba igi 'this is written' vs. dauçeria 'has written it', Finnish [hänet] on kammattu '[(s)he] is / has been combed' vs. on kammanut 'has combed'. Items of different voice do not, however, form proper oppositions on a common basis of comparison. It was stated above that most European Pluperfects are general +PAST (+POST0) items, since competing nonpostterminals are lacking. They are thus both diagnostic and historical, exhibiting static, experiential and other uses. Nevertheless, they may be more or less focal. Corresponding focality oppositions of +PAST (+POST°) items

Viewpoint operators in European languages

121

are, for example, English +PAST (+POST0 (+FOC)) was gone vs. +PAST (+POST0 (—FOC)) had gone, French était maigri 'was emaciated' vs. avait maigri 'had grown thin', Karachai ketib edi 'had left (and was still gone)' vs. ketgen edi 'had gone', Armenian gnacac ër 'was gone' vs. gnacel ër 'had gone', Lezgian fenmaj 'had, was still gone' vs. fenwaj 'had gone'. In Kalmyk, past forms of higher focals -sn bilä contrast with forms of the postterminal in -/, which, however, mostly lack the +PAST marker bilä 'was'. While items such as the Latin Perfect cantavit have lost their postterminality and turned into nonfocals, focality might seem to be more resistent in Pluperfects such as cantaverat 'had sung', but the latter is a general +PAST (+POST°) item, lacking contrasting items and thus covering all pre-past situations. Though the corresponding former —PAST (+POST) items have developed into nonfocals, Pluperfects of this kind may be largely restricted to diagnostic uses, highlighting the posttransformational state, rather than being used to refer in a historical way to pre-past events. Thus, Western South Slavic Pluperfects are nowadays restricted to [-R] actional phrases and diagnostic uses (cf. Gvozdanovic 1995).

8.7. Indirective postterminals The problem of postterminal and ex-postterminal items conveying indirectivity must be mentioned here, though it cannot be dealt with at length. As is well known, certain languages possess evidential categories signalling that a given utterance is based on indirect evidence. Indirectives, often referred to as "reportive", "narrative", or "relative" items, are used to express different kinds of dissociation from the event, cognitive or emotional distance to it, non-testimonial reference, assumption, doubt, irony, surprise, etc. While they play central parts in Altaic, Uralic, and several other languages, they are marginal in Indo-European, though represented in certain regions such as the Baltic, Pontic, Caucasian, and Central Asian areas. Certain types of indirectives are closely connected with postterminality. Focal postterminals often get more or less clear readings of indirective evidentiality. —PAST (+POST) items convey an indirect perspective, expressing the postterminal state emerging from a past or partially past event, and may thus also be used indirectively for past events whose occurrence is only inferred from information available at Os. For example, -PAST (+POST) items of the eastern Finno-Ugrian languages are characterized as "perfecta praesentis" etc. (Serebrennikov 1960: 59, 165), but have additional indirective meanings, e.g., Komi-Zyryan munoma 'has [apparently] gone'. The connection between the postterminal state and the event is then established by way of inference: the situation at "nunc" is such that it may be concluded that the relevant limit of the event has already passed. The indirect view creates an element of distance and uncertainty concerning the actual realization of the event.

122

Lars Johanson

The source of information, the basis on which the conclusions regarding the past or partly past event are drawn, may vary. It may be inference from perception or present knowledge, a conclusion based on given facts, effects, vestiges, or hearsay, reports of other persons. This semantic link between indicative and modal uses is a general typological one. Many focal postterminals activate this latent property and exhibit both aspectual and more or less vague inferential uses. Focal postterminals tend towards indirective readings in various Indo-European languages, from Sanskrit down to modern languages such as Armenian and Scandinavian. For inferentiality suggested by PF items in different languages, see Comrie (1976: 108-110). A recent example is the new Macedonian possessive Perfect consisting of 'have' + participle, which may also get modal readings, e.g., ima napraveno 'has [presumably] done' (cf. Graves, this volume). In some Italo-Albanian varieties, the construction corresponding to the literary Albanian Perfect has modal meanings expressing possibility or probability, e.g., ka vdekur 'may have died', 'has presumably died' (Breu 1991: 55). In Balkan varieties of Romany, the postterminal active participle may be used as an inferential past item. Similar phenomena may be found in the use of the Perfect of certain Kurdish varieties, obviously under Turkish influence. But the indirective shades of meaning occurring with focal postterminals are generally vague and unstable. The indirective capacities are often overstated in the literature, for example, in claims regarding presumptive or inferential uses of Scandinavian Perfects "about a probable event in the past" (Haugen 1976: 80). The modal meaning is often produced by other factors. Tendencies towards indirectivity are often supported by inferential or dubitative expressions, e.g., particles such as Swedish nog 'probably, presumably' or Armenian erevi 'probably' used with the low-focal postterminals in question. However, certain languages possess comparatively clear-cut and stable indirectives developed from postterminals. Such items are often found in Turkic, Mongolian, Finno-Ugrian, Kartvelian, etc. They have left the focal stage of postterminality and developed into nonfocal indirectives that may also be used historically and serve, in narratives accounting for unwitnessed past events, as propulsive units of the discourse basis. Unlike focals, they thus also combine freely with specific time expressions. Since references to the very circumstances of the event, notably the temporal localization, are not favoured by a more or less strongly O-dependent focality, they tend to block indicative readings and to urge indirective interpretations. Thus, the indirect postterminal perspective is reinterpreted as indirective meaningThis development may already start with inferential uses of low-focals, but in the nonfocal indirective types, the ability to represent the historical dimension is fully developed. This development is typical of Turkic finite items in -gän, -mis, -iptir, etc., which have indirective and certain constative uses. The situation is partly similar in Mongo-

Viewpoint operators in European languages

123

lian. The Kalmyk low-focal in -/may refer to past events in an inferential or reportive way and also express the typical additional meanings of unexpectedness, etc. Former Perfects of the Bulgarian and southwestern Macedonian area show similar functions and are thus obviously not PF items any more. The Albanian so-called admirative goes back to a focal postterminal, differing from it by the reverse order of the 'have' auxiliary and the participle. It displays the expected nuances of emotional distance, disbelief, irony, astonishment, and is also used narratively. The presence of an indirective marker of non-attested evidence means that the opposing items get more categorical readings. Moreover, if a postterminal has turned into a nonfocal indirective, it lacks a nonpostterminal item to compete with in the non-modal past stratum. Thus, Turkic +PAST (-INTRA) items in -di often lack pure postterminals to compete with and thus have a rather wide range of use. Since Turkish gitmi§ mainly displays indirective uses, e.g., 'went / has gone' + 'apparently', the item gitti 'went / has gone' is employed for almost all past events that do not call for indirective and / or intraterminal marking. The opposition with indirective items does of course not imply that -di only expresses events witnessed by the speaker. In the literary language, the addition of -dir ('is' < turur 'stands) to -mis eliminates the indirectivity and creates an indicative item in -mistir. Since -dir corresponds to the Bulgarian copula e, the formal and functional situation seems to be rather similar in the third person: xodil = gitmis, xodil e = gitmistir (Johanson 1971: 305). In spoken varieties, however, -mistir is a presumptive item, which allows for an opposition within the evidential field: gitmi§tir 'went, has gone' + 'presumably' vs. gitmis 'went, has gone' + 'obviously, reportedly'. A similar phenomenon in Latvian is that past indirectives used in dubitative functions carry the auxiliary esot, but may omit it in narrative functions. Many languages, Turkic, Mongolian, Finno-Ugrian, Caucasian, Nuristan languages, Tibetan, etc., possess indirective particles (reportive particles, distance particles) that may produce indirectives corresponding to different aspectotemporal and modal items. Turkic languages possess generalized, temporally indifferent indirectivity markers of the types imis < ärmis and ekän < ärkän, which should not be confused with the postterminal types. Bulgarian bil may function as a generalized marker of indirectivity in a rather similar way. The overall structure of the Bulgarian indirective system indicates considerable Turkic influence, non-Oghuzic patterns being the most probable primary models (Johanson 1998). The Baltic Finnic and Baltic indirective markers are somewhat different, though also of a participial nature. They produce special -PAST and +PAST items, e.g., °i the Estonian indirective mood ("kaudne köneviis") söitvat 'reportedly goes', olev at soitnud 'has reportedly gone'. The Latvian "relative mood" is based on the nitely used intraterminal participle in -ot « * -ant), e.g., -PAST rakstot 'reportedly rites', +PAST esot rakstïjis 'has reportedly written'. Lithuanian has corresponding «ems, -PAST rasqs, +PAST esq rasçs.

124

Lars Johanson

8.8. Sources and shapes of postterminals Postterminals have various though largely similar sources. Many can be traced back to periphrases, which tend to turn into more reduced and synthetic shapes as a concomitant of the semantic development. The old Indo-European Perfect, a high-focal postterminal, was already synthetic in Old Indie and Classical Greek. In some Turkic languages, the analytic item -ib turur (postterminal converb 4- turur 'stands') has boiled down to -b. Analytic and synthetic items may coexist in one system. Thus, Arumanian has a synthetic -PAST (+POST) item, but an analytic +PAST (+POST0) item. Most European postterminals can be traced back to periphrastic essive or possessive constructions, based on the metaphorical models 'dwelling in a postterminal phase' (essive) and 'having something in a postterminal phase' (possessive). There is no evidence that they go back to complétives or to constructions with auxiliaries derived from directional and movement verbs. There are also dynamic versions, 'entering into a postterminal phase' (essive) and 'getting something into a postterminal phase' (possessive). The postterminal phase is signalled by a non-finite verbal form, e.g. a participle or a converb. Typically, the point of departure of the development is a nominal form of the verb, a preaspectual item usable as an adjective, an adverb or a noun, and not part of the verb conjugation, e.g., Latin scriptum, Hungarian irva 'written'. The aspectualization turns it into a genuinely verbal form, e.g., scriptum est 'is written', directly related to scribere 'write'. A further example of an item going back to a déverbal nominal form outside the conjugation is the Armenian item in -ac (Kozinceva 1983: 215). Compare the preaspectual development into intraterminals (7.7). Certain languages use locative metaphors such as 'being (j ust ) after m e event', e.g., Irish ta éis a scriobh 'has just written' (= 'is after writing'). An older item of possessive origin is the Akkadian Stative (Rundgren 1961: 59). The auxiliary of possessive items is often a conjugated transitive verb of the 'have' type, whereas essive items are mostly provided with a copula that is a conjugated form of an auxiliary 'be', 'be situated', 'stand', e.g., Hittite es-, Latin esse, Old English beo-, Old Slavic byti, Armenian em, Turkic tur-, Kalmyk bä-, Lezgian awa, Sami icet. Essive items with different "nomina perfecti", postterminal participles, converbs, gerunds, are used in practically all languages, e.g., Indo-European, Finno-Ugrian, Turkic, Mongolian, Caucasian, Basque. A typical model is the Archi combination postterminal suffix -w- + converb + copula, i.e. -PAST (+POST) -li i, +PAST (+POST°) -li edi. Some examples of active readings are Dutch is I was vertrokken 'has / had left, Finnish on I oli tullut 'has /had come', Latvian ir I bija rakstïjis 'has / had written', Sorbianje / be pil 'has / had drunk', Livonian urn I vol' lu'ggën 'has / had read', Abaza dcaxjatl 'has gone', dcaxjan 'had gone', Kirmanji ketiye 'has fallen', ketibû 'had fallen', Talysh avïm gätä T have taken it' vs. avi'm gätä be 'I had

Viewpoint operators in European languages

125

taken it' (ergative), Lezgian ksanwa 'has fallen asleep', ksanwaj 'had fallen asleep'. Slavic has in general used the old participle in -I, but some modern varieties employ other active postterminal participles, e.g., in the North Russian focal polucivsi 'has received' (Trubinskij 1984: 216). Some items of the essive type are diathetically ambiguous (8.5.1.2). Some get passive interpretations, e.g., Polish jest napisany 'is / has been written', Czech je napsân 'is / has been written', German ist gestohlen 'is / has been stolen', Faroese er gj0rt 'is done', Irish ta leite 'is read'. In Georgian, some postterminals are periphrases with a passive participle + copula 'be', whereas others have developed from stative passives, e.g., miçeria 'is written for me'. Many modern languages, Baltic, FinnoUgrian, Turkic and others, consistently distinguish active and passive postterminal participles, Latvian ir I bija uzrakstïjis 'has / had written down', ir I bija uzrakstïts 'has been / had been written down'; cf. active vs. passive distinctions in participles such as Estonian -nud vs. -tud, Tatar -gän vs. -elgän, Esperanto -ita vs. -inta. IndoEuropean postterminal participles may not only be diathetically ambiguous, but also have present or past interpretations according to their internal phase structure and thus seem temporally ambiguous (10.3.2). Auxiliaries of essive items often occur in reduced shapes, mirroring their level of grammaticalization. They may also be lacking altogether. The copula of the old Perfect is lost in most Slavic languages, e.g., Polish pisal (Old Polish pisal jest), preserved in Bulgarian, Macedonian, and Sorbian, and can be omitted in certain cases in some languages, e.g., Serbo-Croatian. The copula may often be absent in postterminals-in-present but present, at least optionally, in postterminals-in-past, e.g., Ukrainianxodiv 'has gone' vs. buvxodiv 'had gone', Vepsian tonu 'has brought' vs. ol'i tonu 'had brought', Low (Meadow) Mari vozen 'has written' vs. vozen île 'had written', Maltese kiteb 'has written' vs. kien kiteb 'had written', Kalmyk irsn 'has come' vs. irsn bilä 'had come'. Finally, it may be lacking in the 3rd person of postterminals-in-present but be present elsewhere. This is usually the case with Turkic non-verbal copulas developed from personal pronouns, e.g., in -gän, -mis, -p items. With postterminals-in-present, the absence and presence of the 3rd person copula may also distinguish indirective meanings from non-indirective ones. Compare the similar situation in Bulgarian and the copulaless form of the Latvian "relative mood" as a reportive Past, e.g., rakstïjis 'is reported to have written'. The possessive type presents the referent of the first actant as the possessor of the Posttransformational state. A déverbal nominal denoting a posttransformational state !s reinterpreted as a verbal structure and incorporated into an inflectional paradigm. Une widespread subtype makes use of a possessive verb meaning 'have'. The type "as broken a window is thus the result of a reanalysis of has a broken window. ( he subject = possessor remains in the nominative case. Representatives of this have' auxiliary are, e.g., Latin habere, Old Icelandic hafa, Old English habb-, Italian av ere, Portuguese ter, Sardinian âere, Romanian avea, Macedonian ima, Modern

126

Lars Johanson

Greek éxo, Albanian ka. The auxiliary may be contracted, e.g., Romanian a < are, and even omitted. The latter is possible in subordinated clauses in earlier German and still in modern Swedish, e.g., Swedish brevet jag skrivit 'the letter I have / had written'. The possessive type is common to many European languages, e.g., spoken Latin habet scriptum 'has written', French a donné 'has given', Italian ha cantato 'has sung', Dutch heeft gelezen 'has read', Modern Greek éxi ghrâpsi 'has written', Albanian ka hapur 'has opened', Macedonian ima dojdeno 'has come'. Some Slavic languages, notably colloquial varieties, exhibit more recent items of this type. (See Maslov 1988: 80-85; for Polish, Czech, Slovak, Serbo-Croatian, see Galton 1976; for northern Russian dialects, see Panzer 1984: 115-127). One of the Georgian Perfect types is a combination of a participle with 'have', e.g., mokluli mqavs 'I have killed him'. In this paper, we shall not discuss contact-induced developments of the possessive type in European languages, but just note that it is even possible that periphrases such as the Old Icelandic and Basque ones are structural copies of the Latin habet factum type. The Basque low-focal analytical construction (present auxiliary + participle) may be an innovation originating from spoken Latin. A second possessive subtype, presenting the possessor as the first actant, is found in ergative languages lacking a possessive verb meaning 'have'. Like the first one, this subtype also goes back to a déverbal nominal denoting a posttransformational state, reanalyzed as a verbal structure and incorporated into an inflectional paradigm. The possessor of the nominal construction appears as the first actant marked by an oblique case, e.g., Kirmanji min tu dîtî T saw you'. Older Indo-European languages display a rather clear-cut distribution of essive and possessive items. At their high-focal stages, the items in question are restricted to certain actional phrases compatible with the original meanings of the paraphrases. The essive type is primarily used with intransitives, and the possessive type with transitives, e.g., Hittite pan eszi 'is gone', pan esta 'was gone', hatran harzi 'has written', hatran harta 'had written'. The auxiliaries tend to agree with the first actant. Examples of this are the Latin high-focal with habere, the Old English one with habb- 'have', the Old Icelandic ones with hafa and the corresponding esse, beoand vera items. At later stages of development, generalization phenomena can be observed, and the distribution of essive and possessive items varies considerably across languages. Transitives get passive meanings with the essive type, e.g., is written. Transformative intransitives often choose the essive type, e.g., Dutch is gearriveerd 'has arrived'. Nontransformative intransitives may choose the possessive type, e.g., Dutch heeft geslapen 'has slept', or the essive type, e.g., Yiddish iz geslofn 'has slept'. (Yiddish, however, does not use the essive type for all nontransformative intransitives.) The possessive type has generalized in English and Swedish low-focals, e.g., Swedish har forsvunnit 'has disappeared', har skrivit artikeln 'has written the article'. If essive constructions with intransitives are available, they are high-focal, e.g., English is

Viewpoint operators in European languages

127

gone, Swedish ärförsvunnen 'has disappeared, is lost'. Compare high-focal possessive constractions with transitives such as Swedish har artikeln skriven 'has finished the article, has the article in a written state'.

8.9. Peripheral postterminal items Many components found in postterminals show a high combinability and also occur in constructions that indicate a 'dwelling in' or 'entering into' a posttransformational phase without functioning as viewpoint operators of the core systems. These less generalized peripheral items exhibit uses atypical of fully grammaticalized postterminal items and must be assigned +POST0 values, as they do not compete with nonpostterminals in their respective temporal strata. Some are only actional, specifying postterminal states related to hypothetical actions rather than expressing events. Certain preaspectual items may be difficult to distinguish from viewpoint operators. For some Turkic cases of apparent ambiguity, see Johanson (1995). Baltic Finnic has preaspectual high-focals consisting of a non-finite item (past participle or infinitive + case) and an auxiliary going back verbs meaning 'become', 'remain', etc. As for prospective postterminals, -PAST (+PRO (+POST0)) items ("Future perfect", "Futurum exactum", "Future anterior", etc.), instruct the addressee to situate the relevant limit of the event prior to an O2 foreseen at O s , e.g., Latin scripserit 'will have written', Icelandic тип hafa talad 'will have spoken', French aura vu 'will have seen', Modern Greek tha éxi ghrâpsi, Bulgarian ste e dal 'will have given', Italian avrà cantato 'will have sung', Romanian va fi venit 'will have come', Portuguese terâ ido 'will have gone', Maltese ikun kiel 'will have eaten', Armenian gnacac ë linelu 'will be gone', Turkish ölmüs olacak 'will have died'. Correspondingly, +PAST (+PRO (+POST0)) items instruct the addressee to situate the relevant limit of the event before an О foreseen at a past O 2 , e.g., Modern Greek tha ixe ghrâpsi 'would have written', Bulgarian stese da e eel 'would have read', Armenian gnacac ër linelu 'would be gone', Turkish gitmis olacakti 'would have gone'. Since no competing nonpostterminals are available, these combinations are not simply —PAST (+POST) items transposed into different time strata. They are absent in many languages, notably in those possessing defocalized -PAST (+POST0) items. As was noted above, auxiliaries involved in complex constructions may themselves take part in +INTRA and +POST oppositions. Examples of the latter are we English distinctions has been vs. was and has become vs. became. The —PAST (+POST) items may, as expected, be used to avoid the implication of a specific occasion and do not readily combine with specific time expressions, whereas the competln g +PAST (-POST) items are historically oriented. Combined with items signalling a +INTRA or +POST perspective, the -PAST (+POST) auxiliary expresses that the Perspective on the event has been valid. The +PAST ( -POST) auxiliary suggests

128

Lars Johanson

that the perspective was valid on some past occasion. For example, it is possible to distinguish between 'has dwelled in an intra- or postterminal state' and 'dwelled in an intra- or postterminal state', e.g., Estonian on olnud tegemas 'has been doing' vs. oli tegemas 'was doing', Swedish har varitgjord vs. var gjord. Correspondingly, dynamic auxiliaries taking part in ±POST oppositions allow the speaker to distinguish between 'has entered' and 'entered' an intra- or postterminal state, i.e. to express that the perspective on the event has become valid or became valid on some past occasion. Certain peripheral constructions involving +POST0 elements have developed into more central items. Most of them are inessive and illative constructions based on the non-dynamic and dynamic copula types 'be' and 'become' and implying 'dwelling in' and 'entering into' a postterminal state. Dynamic intraterminal constructions were mentioned in Section 7.7.1. Only a few possible postterminal counterparts can be hinted at here. Illative-dynamic poststatals have a modest distribution in European languages, often restricted to lexicalizations, e.g., become interested, get married. There are also genuinely paradigmatic items. Combinations implying 'entered a postterminal state' ('became having V-ed') are East Armenian gnacac eyav 'went', Turkish açmi§ oldu 'opened'. The dynamic auxiliaries are +PAST (—POST) items of verbs meaning 'be', 'become', or 'have'. The combinations in questions are not dissimilar to older past anteriors such as French eut fait 'had done' or Italian fu uscito 'had gone out', whose auxiliaries were +PAST ( —POST (—INTRA)) items. Passive combinations are, e.g., Estonian sai tehtud, Swedish blev gjord, Turkish yapilmis oldu 'got done'. Dynamic items of this kind often have the meaning of 'eo ipso', e.g., Turkish gitmis oldu 'thus / therewith went' (Johanson 1971: 312). Unlike the corresponding non-dynamic items, they are incompatible with adverbials meaning 'already', which refer to a state obtaining at O. This is also true of the Romance anterior past types, e.g., French fut parti 'had left'. Combinations provided with postterminal auxiliaries and implying 'has entered a postterminal state' are, for example, East Armenian gnacac ë eye/ 'has gone', Estonian on saanud tehtud 'has become done'. Among combinations implying 'had entered a postterminal state' are for example, literary Italian ebbe cantato 'had sung', Turkish gitmis olmustu 'had gone', East Armenian gnacac ër eye/ 'had gone'. As was noted above, +PLUR and +DISP markers (signalling pluri-occasionality, habituality, disposition) may combine with ±POST in the sense of 'has usually done' and 'tends to have done', e.g., Lithuanian budavo rasçs 'used to have written', Armenian gnacac ë linum 'is usually gone', Maltese ikun kiteb, Turkish yazmis olur 'will / tends to have written'.

Viewpoint operators in European languages

129

8,10. Defocalization of postterminals Postterminals are subject to successive functional shifts with gradually diminishing degrees of focality, increasing historical orientation, weaker O-relevance. Items characterized by a more diagnostic orientation get more event-oriented functions. This defocalization drift is taken here to include the diachronic steps (1) high focality -> (2) low focality —> (3) nonfocality. Semantic fusion in combined items is often observed. High-focals tend to develop into less focal items, acquiring increasingly general (historical) and finally modal (irreal, conditional) functions or disappearing. In particular, innovations of higher focality encroaching on the area LF 0 f -PAST (+POST ) items lead to constant system changes; compare the similar encroachment upon -PAST (+INTRAoLF) items by high-focal intraterminals (7.8). Well-known examples of the development are found in Classical Greek, where the old high-focal Perfect developed into a diagnostic + historical item, or in Persian, where a high-focal [mana] krtam 'I have done' developed via a low-focal [man] kart into a nonfocal kardam (Kurylowicz 1956: 29-30). The evolution of Romance Perfects includes corresponding diachronic stages; see, for example, the scheme laid down in Harris (1982). Turkic defocalization developments have been dealt with in Johanson (1971, 1993, etc.). The origin of Maltese kiteb 'has written, wrote', opposed to jikteb 'writes' (cf. Standard Arabic - INTRA kataba and +INTRA yaktubu) is an old -bPOST1111 item (corresponding to the Akkadian so-called Permansive), which later on defocalized to a -PAST (+POSTNF) item and pushed the old Imperfect (yaktub) into modal functions. The high focality has been renewed by participial -f-POST101 items of the type rieqed 'is asleep' (with a limited set of verbs). Standard Arabic creates +POST (—INTRA) items by means of the particle qad, e.g., qad kataba 'has written'. Graves (this volume) presents three Macedonian items of a postterminal nature or origin, occurring side by side and opposed to each other in various ways in the individual dialects: (1) a recent essive item formed with a passive participle, (2) a possessive construction, and (3) an old essive item formed with an active participle. These constructions do not constitute three PF items in a synchronic sense, but hold different positions along the focality scale. In the North, Hem 1 is high-focal, e.g., umren e 'is dead', 2 is limited to possessive cases, and 3 is low-focal. In the South-West, item 1 is losing its focality, 2 is low-focal, and 3 is a n onfocal indirective. Item 2 is generally gaining ground from 3 and has replaced it "i some dialects. Defocalization is, as we have noted, due to semantic generalization, leading from narrower to increasingly broader uses. Items developing along such lines become capable of covering more and more situations, taking over functions of older, lower nerns, gaining ground from them and eventually superseding them. Even if higher 0c als thus may seem to "take the place" of lower ones, they do not replace them

130

Lars Johanson

in the sense of assuming the same values. A higher item with relatively specialized functions typically extends its use to subsume both its own old functions and certain functions of a lower item. Narrow uses are typical of young items created to renew the expression of a ceein function. High-focals represent earlier stages of semantic development than lower items, which have a wider functional range and the ability to cover functions typical of higher items. It would be misleading to refer to defocalization as perfectivization. The development of PF items into "perfective pasts" in languages with a PFV vs. IPFV opposition (Bybee & Dahl 1989:74) is not a result of defocalization. First, the development of a -PAST (+POST)tem into a PFV of the nonintraterminal kind depends on the presence of an opposing +INTRA item and does not belong to the defocalization process. Secondly, defocalization never leads to the emergence of PFV items of the adterminal kind. This seems natural, since a device denoting the transgression of a limit is not likely to become a device denoting the mere attainment of the same limit. There is no defocalizjlion path leading from a PF item to a perfective of a Slavic type. Certain Slavic +T-marked postterminals have developed into +AD pasts, but their adterminality derives from their transformativity, not from the former postterminal ("Perfect") valut. A further reason to distinguish PFV items of the adterminal and the nonintraterminal kind is thus that their paths of development are quite different. 8.10.1. From high-focal to low-focal Some of the numerous shifts from high-focal to low-focal postterminality have already been mentioned. For example, the Old Georgian Perfect was high-focal, whereas the modern Georgian one is more event-oriented and vacillates between diagnostic and historical uses. Armenian exhibits two items representing the two stages of development (cf. Maslov 1988). It was also noted that, when higher focals are defocalized, high-focal postterminality is often renewed by new periphrastic items; compare the analogous renewal of high-focal intraterminality. Thus, the former high value of Persian krtam 'done' was renewed by kartak am T have done', which developed intokarda am (Kurytowicz 1956: 29-30). The latter item does not express 'Tactionperfective pure et simple", but is simply a low-focal —PAST (+POST) item. The Turkic items it -gän and -mis seem to have been renewers of focality, possibly filling the placeof an older finite item in -ip (Kormusin 1984: 44). As they turned into low-focals, high focality was renewed by periphrases such as -ip turur, -mis turur and -gän mr, which later on developed into constative low-focals themselves (Johanson 1993, 1995). In most Turkic languages, the originally high-focal postterminal in -mist, not used any more. Interestingly enough, however, several modern Iranian languages, when accommodating copies of Turkic verbs to native

Viewpoint operators in European languages

131

morphosyntactic frames, use do and be constructions plus a -miS form of the copied (typically transformative) verb, e.g., Zaza kandirmiS kärd 'persuaded' (cf. Doerfer 1993). The Mongolian items in -jugu, -gsan and -luga also seem to have been high-focals. In Literary Mongolian, the item in -luga expressed "a fact of completed action, and the situation thus established" (Gr0nbech-Krueger 1955: 36). The corresponding Kalmyk item in -la exhibits constative, complexive functions. Whereas Turkic, Mongolian, Iranian, and some Caucasian languages display rich renewing developments of focal postterminals, Germanic, Romance, Slavic and Finno-Ugrian languages are rather unproductive in this respect. For example, since the postterminal periphrasis esmï + participle in -/ was defocalized and finally replaced the Aorist, no North Slavic language has developed any reasonably grammaticalized new focal category. It has already been stated that, when high-focals become low-focals, they are not restricted to [+t] actional phrases any more, but rather operate freely on all actional types. An interesting fact is the weak defocalization of the Calabrian and Sicilian Perfects, which have retained a higher degree of focality and still only operate on transformatives. The ambivalent diathetic orientation has been given up in some languages. Compare, for example, the Old Turkic type yïrtïq 'torn' (preserved in Turkish yirtik) with modern Turkic passive forms such as Turkish yirtdmis 'torn'. In several Indo-European languages, possessive 'have' items have expanded to operate on more actional types than they did as high-focals, gradually encroaching on the territory of essive items, e.g., English is gone > has gone. There are several examples of this development in Romance languages, e.g., Spanish somos idos > hemos ido 'we have gone' (cf. also Catalan, Portuguese and Romanian).The defocalization of the Indo-European possessive type was also accompanied by other syntactic changes. The participle lost its adjectival character and its agreement with the object, became part of the verb phrase and mostly changed its place in the sentence, e.g., Old Icelandic hefir bôkina lesna 'has the book in a read-through state' > hefir lesit bôkina 'has read the book'. This development is also assumed for Late Latin: habet litteras scriptas 'has the letter in a written-down state' > habet scriptum litteras 'has written the letter'. Constative readings are already found in the highest parts of the low-focal sector. Graves (this volume) reports that even the new Macedonian high-focal may, though not characteristically, be used with experiential meaning. Thus, if -PAST (+POST) items are observed to have "non-resultative" uses with [-t] actional phrases, we cannot exclude the possibility that they have retained a higher focality degree; cf. Tommola's discussion (this volume b) of past tenses in Old Russian.

132

8.11.

Lars Johanson

From low-focal to nonfocal

Focal postterminals are, as we have noted, found in large parts of Europe. They are by no means limited to some maritime areas in the western parts of the continent, but are also present in Finno-Ugrian, Turkic, Caucasian, and other languages. On the other hand, the drift to nonfocality is also a widespread phenomenon and not an areal feature limited to a coherent inner part of the western half of the continent. This ultimate defocalization coupled with the disappearance of a former +PAST (—POST) item ("preterite loss") has been described in various ways. It is traditionally often said that an original "perfect" comes to be used as a "perfectum historicum" and finally turns into a "historical tense" or a "past tense". If focal postterminality is thought to be reference to two temporal strata, transition from low-focality to nonfocality is taken to be a loss of this double reference. The development is sometimes also conceived of as a retrospective expansion of the present space of time. The change from focal postterminals to pasts is a universal tendency also reflected in ontogenetic development, namely in child language acquisition (see Antinucci & Miller 1976, cf. Givön 1982: 151). The generalization of a low-focal and the disappearance of the corresponding —POST item means the loss of the ±POST opposition. Postterminals involved in this defocalization process may be more or less progressive. An item X is more progressive than an item Y, if it takes on Os-independent functions in more contexts than Y does. There are generally transitional stages with promiscuous uses of the former +POST vs. -POST items and with residual higher functions observable in the +POST items. The tendency towards total defocalization is observed in many older languages including Old Indie, Latin, etc. As for the Classical Greek Perfect, the tendency led, from the Alexandrian period on, to promiscuous use and transition to a nonfocal item. The Perfect took a step down the focality scale, assumed event-oriented functions, fused with the Aorist and was superseded by it (Browning 1983: 30). The result of similar developments is found in several modern European languages. As for Slavic, it is present in Russian, Belarusan, Ukrainian, Czech, Polish, Rusyn, and also largely characteristic of Slovene, Croatian as well as most Serbian varieties. The development has been delayed in Sorbian, possibly under German influence. However, in Lower Sorbian, the former +PAST (-POST) item has practically disappeared. Upper Sorbian and literary Lower Sorbian exhibit promiscuous use, the socalled Perfect being a -PAST (-t-POST1^) item that can always be substituted for the former +PAST (-POST) item (Faßke 1981: 262-263). Among Germanic languages, German and Yiddish are most strongly affected. The Dutch Perfect is found in the lowest part of the low-focal sector, being more progressive than the English or Scandinavian counterparts but more conservative than the Southern German one, since it has not yet ousted the +PAST (-POST) item. The Afrikaans counterpart is clearly

Viewpoint operators in European languages

133

more progressive. Romance postterminals exhibit various degrees of defocalization. Their progressivity seems to increase along a line stretching from Galician and Portuguese to Castilian Spanish, Catalan, and Occitan and to Italian, French, Romanian, and Romansh. Thus, the Spanish type ha hablado 'has spoken', in the middle of this continuum, is observed to occupy more and more of the functional territory of the type hablô 'spoke'. On Sardinian, see Bossong (1993). Nonfocals are also found in Hungarian, Maltese, Romany and varieties of Albanian. In several other languages, postterminals seem to be encroaching on the territory of nonpostterminals, e.g., the Estonian Perfect, possibly due to Russian influence. The functional distribution often exhibits a good deal of regional variation. In several languages, the loss of ±POST oppositions predominantly affects certain regional varieties, e.g., South German, North Italian, Daco- and Istro-Romanian, or Gheg dialects of Albanian. In dialects tending towards defocalization, the ±POST opposition is mostly rare in colloquial language (Italian, Serbo-Croatian, Romanian) but preserved in more formal registers and stylistically marked contexts. The old Catalan +PAST (-POST) item (escrivi 'wrote') is also mostly limited to written and literary registers. However, the latter reduction is not due to defocalization of the -PAST (+POST) item (ha escrit 'has written'), but to the introduction of the -f PAST (-POST) periphrasis va + infinitive (va escrime 'wrote'). Moreover, if a ±POST opposition is given up, it is not always the former -PAST (4-POST) item that generalizes. It is also possible that the former +PAST (-POST) item develops into a more general item referring to past events both historically and diagnostically (Section H.6.1). With the change to nonfocality, the meaning of a postterminal generalizes to cover the widest range of diagnostic and historical interpretations. The loss of the specific meaning component of Os relevance makes the items compatible with more contexts. However, the generalization does not lead to the precise expression of more functions. If a relatively specialized item X extends its use to subsume both its old functions and the functions of an item Y, this generalization leaves both the old X and Y functions without adequate expressions. Unless focality is renewed by some other item, the nonfocal may also continue to cover cases of Os relevance. Thus, even if the Russian Past might be claimed to cover meanings typical of a "Perfect" (Maslov 1980: 51, 53), these meanings are not explicitly expressed. It can only be concluded that this generalized Past is also used in cases where some other language Would use a focal postterminal. A nonfocal can never replace a +PAST (-POST) ttem in the sense of taking over its value. While it may refer to the event historically, rt is no genuine 'temps historique'. Only if focal postterminality is renewed by other ^eans may it evolve into a nonpostterminal past. The statement that a low-focal postterminal, e.g., in the sense of a PF item, devel°Ps into a nonfocal is problematic if the latter is taken to be a PFV item. Thus, in the Functional Grammar as initiated by Dik (e.g., 1989), this would mean that a higher

134

Lars Johanson

operator ("perfect") develops into a lower one ("perfective"), which is at variance with other known diachronic tendencies. But this is only a pseudo-problem. First, nonfocal postterminals are, per se, not PFV items in an adterminal or a nonintraterminal sense. Secondly, even if they develop into PFV items in new oppositions, the problem of a higher operator developing into a lower one does not arise. In a Func­ tional Grammar model of this kind, PFV should, as I have argued, be taken to be a higher operator on a par with PF (Johanson 1996). When low-focals lose their focality, the development seems to start with [-И] actional phrases, whereas residual focal uses may be observed with [—t] actional phrases. For the development of the Old French Perfect, see Schwenter (1994). The uses with [—t] actional phrases have constative readings, which may mean that constativity is the last domain given up by focal postterminals. Note that the American English Perfect, while obviously losing some of its focal uses, is still stable in con­ stative functions. The relatively weakly defocalized Portuguese Perfect, which was formerly also used with [+t] actional phrases, is now essentially restricted to [—t]. The situations are partially similar in Galician and varieties of American Spanish. Constative readings may thus be transitory to the nonfocality stage. Constative inter­ pretations are possible with all low-focal items and with both [+t] and [—t] actional phrases. On the other hand, constative interpretations without "resultative" shades of meaning are typical of +POSTLF x [—t]. It is obvious that they are only a result of interaction with actional values and do not represent any independent cardinal stage in the defocalization drift. Sensitivity to temporal distance may be decisive for the first step leading from +POSTLF to +POSTNF. This step is often taken in contexts referring to events tem­ porally remote from 0 s . Comrie supposes a gradual "relaxation of the degree of recentness required for the use of the Perfect" to have been a key part of the Ro­ mance development (1976: 61). Postterminals of the lowest focality degree are often used in ways that seem to motivate definitions as "remote Past", "Tempus der Fer­ ndistanz", or "mythic Past", e.g., Turkic -mis and -gärt items, constatives in -ipdir etc., Kalmyk constatives in -la. There have been similar losses of postterminal oppositions of the pre-past stratum. Varieties in which a former —PAST (+POST) is defocalized are also likely to exhibit corresponding general pluperfects. However, the loss of the opposition may be accompanied by compensatory developments. As we noted, the loss of the French Simple past led to the loss of the Past anterior. In certain spoken varieties, however, the latter has been replaced by the "passé surcomposé" a eu lu. Similar products are the South German Pluperfect hat geschrieben gehabt 'had written', the Serbian type bio je isao 'had gone', and the Albanian so-called Perfect II, typical of Gheg varieties whose former Perfect has become a nonfocal, e.g., ka pasë qenë 'had been' (Buchholz & Fiedler 1987: 133). In varieties of this kind, other types of supercompound items for the pre-past and even prepre-past stratum are also found (literally

Viewpoint operators in European languages

135

'had had V-ed'), e.g., French avait eu vu 'had seen', German hatte gelesen gehabt 'had read', Albanian Gheg dialects kishte pasë ardhë 'had come'.

9. Adterminality 9.1. Definition The last viewpoint oppositions to be dealt with in more detail are those of adterminality vs. nonadterminality, present in North, East and Western South Slavic languages. They are of an aspectual nature, though intricately interrelated with actionality. ± AD distinctions represent rather atypical varieties of the general PFV vs. IPFV distinction assumed by many linguists. Adterminality (+AD) is signalled by perfectives. However, the presence of a viewpoint value +AD in a given verb presupposes a partner verb with the same lexical meaning, e.g., Russian + ADpostroit' vs. —AD stroit' 'build'. Thus, not all morphological perfectives signal +AD. Hermann even avoided the term "Aspekt" because of its use for Slavic perfectives and imperfectives, which are not always, as he noted, "subjektiv geschieden" (1927: 228). As already stated, many perfective verbs represent modes of action and are thus not ±AD partners. On the other hand, many verb pairs are genuine ± AD partners although they may be translated into other languages by different lexemes, for example, Russian dobivat'sja by English strive after and dobit'sja by attain. The opposition is absent in verbs that do not form aspectual pairs. Such verbs are referred to as imperfectiva tantum and perfectiva tantum (e.g., Belarusan zaminac' 'disturb' and abnarodvac' 'publish', respectively) or are even claimed to represent both aspects (e.g., arandavac' iease'). Adterminality, +AD, envisages the event ad terminum, in the attainment of the relevant limit of its actional content. Since it operates on transformatives, this limit is a crucial one, generally the terminus finalis. By contrast, nonadterminality, —AD, disregards the attainment of a relevant limit. +AD denotes that the transformation is brought about, whereas —AD does not deny or exclude it. Note that, as always with viewpoint notions, the phases that are not highlighted are only latent, not necessarily inexistent. Adterminality is not identical to transformativity, which only implies a limit to attain but not the very attainment of this limit. Whereas [+t] implies the presence of the crucial limit in the actional content, +AD signals its realization in an event. The difference is thus by no means a gradual one. Much confusion has been caused by the compatibility of [+t] and +AD as well as their special interconnections in Slavic systems. As noted above, it is necessary to distinguish the viewpoint operator (aspect) from the operandum (actional content) but possible to assume that [+t] and +AD may merge into portmanteau markers. Thus, Russian-type perfectives

136

Lars Johanson

are combined +AD- and +T-markers, expressing +AD x [+t]. The corresponding imperfectives signal neither +AD nor [+t], thus standing for a nonadterminal perspective on events. By definition, systems of this kind lack the combination *+AD x t-t]. How does our definition relate to other assumptions regarding aspect oppositions of the Russian type, as discussed from Jakobson (1932) on? The [-K] actional content of perfectives in the sense of a possible transformation rather seems to correspond to what is sometimes called the 'event-unit' (Barentsen 1985: 59-60: "handelingseenheid"). +AD has a certain affinity with the feature 'totality', which presents the 'event-unit' as one total whole. But the ±AD opposition does not concern 'completion' to the effect that —AD characterizes an event as unfinished. —AD neither affirms nor negates completion. The aspectological literature offers numerous examples of —AD verbs in sentences implying that the event has really been brought to an end, e.g., Polish czytal tç ksiçzkç 'has read this book'. Nor is a well-defined 'result' of the event a pertinent feature of + AD. Both 4- AD and —AD may get a more or less resultative reading according to the context. The idea of asymmetry in the markedness structure - + AD being the semantically marked and -AD the semantically unmarked member of the opposition - is of basic importance and conforms to a long tradition going back to A. X. Vostokov (Jakobson 1932). For example, it excludes the possibility that morphologically unmarked imperfectives such as Russian pisat ' 'write' are taken to possess zero markers signalling a positive aspectual content. —AD represents negation of the + AD value and irrelevance towards it, often simply implying occupation with the event. The —AD past is often said to have a 'simple denotative' or 'general factual' function, the event being referred to in a generalized manner. Comrie attests that Russian -AD items can be used when "the speaker is simply interested in expressing the bare fact that such and such event did take place, without any further implications, and in particular without any implication of progressive or habitual meaning" (1976: 113). This statement does not mean that imperfectives signal 'totality' and is thus not incompatible with Comrie's totality view of perfectivity. According to Dahl (1985: 76), Comrie's claim that IPFV pays essential attention to the internal structure clashes with the idea that it sometimes expresses "the bare fact that such and such an event did take place". —AD is widely used for events dissociated from a sequential setting, e.g., Czech Tuhle knihu cetla ' She has read this book'. It is only natural that the direct adterminal view of the limit is more fertile in sequential settings, but there is no reason to claim that the opposition is neutralized outside them. Since the attained transformation signalled by + AD suggests a transition to a new situation beyond the transformation, +AD is typically used for temporally situated events conceived of as linked to a preceding and / or following event, as leading from one situation to another. On the other hand, —AD is rather indifferent to situational change. The relevance of

Viewpoint operators in European languages

13?

sequential connection as a feature of discourse organization will be further discussed below. It is, however, important to emphasize that ± AD distinctions are relevant with non-set events as well, the decisive point being whether the event is envisaged in its very attainment of the crucial limit or not. It is often claimed that +AD typically denotes single events and that negation is more characteristic of —AD than of + AD. Such features concern the actional content, the object of aspectual characterization. As we have seen, a serial reading as well as a negation may lead to recategorization [+t] > [—t]. Since [-tt] is not typical of actional phrases expressing single actions, + AD does not typically refer to global events with subevents. However, perfectives of certain Slavic languages, e.g., Czech, behave differently with respect to the representation of repeated events and may also be used for [+ser] readings.

9.2.

Temporalization

A +PAST (+AD) item normally situates the +AD perspective at an L that is anterior to Os and coincides with the attainment of the crucial limit, e.g., Russian napisal 'wrote, has written', Belarusan/?rocytau, Polishprzeczytal 'read, has read, had read'. +PAST (—AD) items, which are not concerned with any crucial limit, only signal anteriority, e.g., Russian pisal 'wrote, has written, was writing, has been writing, had written, had been writing', Belarusan cytau, Polish czytal 'read, has read, was reading, has been reading, had read, had been reading'. Most Slavic languages only possess one single past tense, which thus has several English translation equivalents. The non-past stratum is more complicated. One problem concerns presentness related to Os. —PAST situates the aspectual perspective at a point that is not anterior to 0 s . With —PAST (—AD) items, this point can coincide with O s , the natural vantage point from which an ongoing event is observed, e.g., Russian piset pis'mo, Polish pisze list 'is writing, writes a letter'. The final limit is not included in the view offered here, no matter how extended the event is. —PAST (—AD) items such as SerboCroatian pise 'writes, is writing' are characterized by the natural intraterminality of Os-presentness and might thus even be represented as a -PAST (+INTRA0 (-AD)) items. Like other nonfocal -PAST (+INTRA0) items, they display usages that burst the narrow "nunc" perspective, being used for uni-occasional events in progress at Os, pluri-occasional events, temporally unlimited events, events referred to as a type, universally valid facts, past events, ficticious, scheduled, intended, potential events, etc. -PAST (+AD) cannot be applied to events in progress at Os. The meaning of pectotemporal items of this structure is incompatible with the description of something going on at encoding time. +AD cannot be applied to events current at "nunc", as it is incapable of envisaging a present cursus. It might be thought that the presenas

138

LarsJohanson

tation of the event as a totality necessarily implies Os -posteriority of the crucial limit to attain. More adequately expressed, however, adterminality envisages the very attainment of a future crucial limit and thus projects that limit directly into the future. It is a well-known fact that perfective presents of Russian, Polish, Czech, etc. do not express current events any more, but are regularly used with future - not necessarily 'immediate' future - time reference, which has become their main function. The socalled incapability of presentness of —PAST (+AD) items ("Gegenwartsunfähigkeit des perfektiven Präsens") has been taken to be a central criterion of aspect. +AD is totally incapable of primary deictic presentness, cannot be anchored in the present time sphere. Some kinds of presentness which will not be discussed here, namely those represented by so-called historical and performative presents, are less dependent of the Os perspective and may allow a -PAST (+AD) presentation envisaging the crucial limit. Thus, —PAST (+AD) items actually serve as a special kind of prospectives. If they interact, as Russian napiset 'will write', with periphrastic imperfective futures such as budetpisat' 'will write, will be writing', they may be said to function as —PAST (+PRO (+AD)) items, opposed to -PAST (+PRO (-AD)) ones. Compare the Modern Greek Aoristic Future, e.g., tha ghrâpsi 'will write', which is a -PAST (+PRO (-INTRA)) item opposed to a -PAST (+PRO (+INTRA)) item, e.g., tha ghrqfi 'will write, will be writing'. It differs considerably from the Slavic —PAST (+AD) type of future time reference in signalling prospectivity by means of a special marker tha and not directly envisaging the future attainment of a crucial limit.

9.3. Combinability The application of ±AD oppositions is impossible in certain contexts. Thus, the use of +AD is blocked with phasal verbs such as 'begin', e.g., Russian *nacinaet napisat'pis'mo 'starts writing a letter'. It is possible to begin an action that contains a crucial limit, but not to begin the attainment of this limit. This is clear evidence of the aspectual nature of +AD. The combinability of ±AD with time expressions has already been briefly commented on in Section 7.2.4. —AD items exhibit certain restrictions with punctuality but may combine with punctual expressions that do not refer to the crucial limit, e.g., Russian v dva casa rabotai 'was working at two o'clock'. However, they do not combine with expressions meaning 'suddenly', e.g., *vnezapno stojal tam 'stood there suddenly'. This is one of several differences between IPFV items of the +INTRA and -AD types. —AD items may, like intraterminals, cooccur with adverbials expressing that the initium is not later than О ('already') or that the finis is not earlier than О ('still')> Russian uzepisal 'was already writing', vse esce pisal 'was still writing'. However,

Viewpoint operators in European languages

139

.4-AD in uze napisal would suggest pre-past: 'had already written'. 'In X time' expressions (Russian za X vremja), which measure accomplishment time and identify the actional phrase as [+t], require +PAST (+AD) items, which subsume the value [+t], e.g., Russian poëinil masinu za tri casa 'repaired the car in three hours'. Compare the similar use of +PAST (-INTRA). Abtemporal 'since X time' expressions (Russian uze X vremja), which measure the time between the initium and O, require -AD, e.g., Russian uze dva goda tarn zu 'had been living there for two years'. Compare the corresponding uses of +INTRA items. 'X times' expressions may be limiting, and +PAST (+AD) can thus be used to represent the whole global event with its final limit, e.g., Russian napisal mne tri raza 'wrote to me three times'. Compare the similar use of +PAST (-INTRA). However, if the final limit is unimportant to highlight, +PAST (—AD) may also be used, e.g., pisal mne tri raza, which would of course be impossible with a +PAST (—INTRA) item. 'For X time' expressions, which measure the temporal extension and identify the actional phrase as [—t], require, in languages such as Russian and Polish, —AD items, which subsume the value [—t]. To present a past event as extending over a period of time, Russian uses +PAST (-AD), e.g., dolgopisal 'wrote for a long time'. As we have seen, however, this is a case where +PAST (-INTRA) items are used, since a -f INTRA view is infertile with exact indications of the outer measures of the event. Bulgarian, as expected, uses the imperfective +T-unmarked Aorist, e.g., stoja dälgo na prozoreca 'stood for a long time at the window'. It was noted above that these different choices show an essential incongruity between IPFV items of the +INTRA and —AD types and that 'for X time' expressions by no means necessitate a "perfective" view. On the other hand, a durative event may be presented as a totality by means of delimitative or perdurative modes of action: 'spent [a certain time period] V-ing'.

9.4.

ThewaytoiAD

It has already been noted that the sources of ±AD systems are transformatives and nontransformatives, represented morphologically by +T-marked perfective and —Tunmarked imperfective verb stems respectively, and grouped together into pairs. +AD items are thus formed by means of prefixes or suffixes from stems that eo ipso become —AD items. There are also cases of suppletivism, e.g., Polish +AD wziac vs. —AD brae 'take'. Secondary imperfectives (see below) are formed with suffixes from perfective stems or with vowel changes in the primary stem, e.g., Polish zarobic -»• zarabiac 'gain'. Many aspectologists assume a stable basic distinction of perfectivity vs. imperfectivity, present in all older and modern Slavic languages. With respect to grammatical values, this alleged stability or unity is highly questionable. There is no

140

Lars Johanson

unitary "Slavic style aspect", and general definitions of a Slavic perfectivity such as "l'action dans sa totalité, comme un point, en dehors de tout devenir" (Saussure 1916: 161-162) conceal substantial differences. Modem Slavic languages use formal perfectives for both +T-marking and adterminality. While the dichotomy serves ±AD distinctions in languages such as Russian, it is at least known to be "less aspectual" in certain West and South Slavic languages. It will be argued in the present contribution that the type found in Bulgarian belongs to the domain of actionality. The two types represent different semantic values and cannot be said to be essentially equivalent. There is thus also little justification for assuming a "Sprachbund", an areal group of Slavic, Baltic, German, and Hungarian, which would have grammaticalized "perfectivity" in a uniform way. The Slavic languages do not represent a unified type, and German does not even employ +T-marking in the systematic way as Hungarian, Bulgarian, Lithuanian, Georgian or Ossetic does. The systems mentioned certainly display basic morphological analogies. For example, the Slavic formation types are partly paralleled in Baltic languages. Of course, a "Sprachbund" does not need to be homogeneous, and may display different degrees of development in its parts. However, the different interpretations of the morphological data at the semantic level leave us with a poor basis for establishing linguistically interesting common features. +T-marking, even if it is systematically applied, remains within the domain of actional content. On the other hand, +T-markers may develop diachronically into viewpoint operators. The exclusively +T-marking perfectives represent an earlier stage of semantic development than the +AD-signalling ones found in certain Slavic languages. The Russian-type development led from (i) lexeme derivation to (ii) -+-Tmarking to (iii) aspect formation. A mode of action without aspectotemporally determining force developed into a viewpoint operator; an actional distinction turned into a ±AD opposition. The perfectives extended their function from marking actional phrases for [+t] to determining them aspectually as +AD. The +T-markers came to mark -I-AD x [-И], a development into what Mourek (1895) called "true perfectivity". The marked category not only implied a crucial limit to attain but also envisaged the attainment of this limit. The viewpoint operator imposed an additional semantic constraint on the meaning it operated on. A portmanteau marker combined transformativity, which just aims for totality, with adterminality, which actually in­ volves it. The unmarked items became nonadterminals, implying no crucial limit to attain and consequently no attainment. This decisive step is often ignored in the discussion. It has not been taken ш Hungarian, Latvian, Lithuanian, Kartvelian, Ossetic, etc. Though the +T-marking systems of these languages are often referred to as perfective vs. imperfective dis­ tinctions, they have not developed into ±AD oppositions. The Bulgarian distinction is still an actional one, the perfective functioning as a +T-marker (marker of "pi"e'

Viewpoint operators in European languages

14J

del'nost' "; Tommola 1984). The presence of an imperfective Aorist and a perfective Imperfect in Bulgarian has sometimes been used as an argument against the aspectual nature of the Imperfect vs. Aorist opposition. On the contrary, it should rightly be considered evidence against treating the perfective vs. imperfective dichotomy as a viewpoint distinction. Similarly, the Hungarian +T-marking preverbs are not aspectual markers or "perfectivizers" in an adterminal sense. The distinction between megir 'writes (to finish writing)' and ir 'writes' is one of actional content. It has been suggested that a Russian-type aspect system is in the process of development in Hungarian. Comrie, who rightly considers Hungarian less developed than Russian with respect to a perfective vs. imperfective opposition system, supposes that the Hungarian preverbs are developing into markers of perfectivity (1976: 9 3 94). However, Csato (1994: 232-237) shows that +T-marking is decisive in assigning correct interpretations to the Hungarian aspectotemporal forms and "provides a better characterization of the function of Hungarian verbal prefixes (which derive transformatives) than does their characterization as perfectivizing" (Comrie 1994: 299). Though the difference between actional and aspectual categories is by no means gradual, there may of course be gradual diachronic developments from pure +Tmarking into +AD. The starting point of the Slavic development is a Proto-Slavic system comprising the aspectotemporal categories Present, Perfect, Imperfect, Aorist, and, in addition, a rather unrestricted [±t] duality (perfective vs. imperfective). This system was restructured in most Slavic languages with the loss of the Imperfect, the Aorist, and the -PAST (±INTRA) opposition holding between these two items. Moreover, the Perfect mostly lost its -PAST (+POST) function and developed into a +PAST item. The [±t] distinction, semantically very different from the ±INTRA and ±POST distinctions, could not compensate for their loss or, as is sometimes claimed, take over their roles. In languages such as Russian, the [±t] distinction developed into a ±AD distinction, in the past strata in the following way: (i) pisal -PAST (+POST) x [-t] > +PAST (-AD); (ii) napisal -PAST (+POST) x [+t] > +PAST (+AD). These changes seem rather natural in the light of the affinities between certain actional and aspectotemporal categories. Thus, +INTRA items such as Imperfects have affinities with [-t], and -INTRA items such as Aorists with [+t]. As already stated, [+t] was constitutive for the Classical Greek Aorist. +T-marking may be more important with some aspectotemporal categories than with others. It is hardly astonishing if the affinities also lead to mergers. A system of free combinability of viewpoint operators and actional contents may change to the effect that [+t] is largely restricted to -INTRA, and [-t] to +INTRA. Thus, the Slavic Aorist is, in languages that have preserved it, predominantly perfective. In Sorbian, the Aorist is used with Perfectives and the Imperfect with imperfectives (Sewc 1968: 171-172). These past e nses hold an intermediary position between an older Slavic type, represented by

142

LarsJohanson

Bulgarian, and a new type, represented by Russian. Their development is a parallel to the merger of transformativity with adterminality in languages of the Russian type.

9.5. Differences between +AD- and +T-markers The two types of perfective vs. imperfective distinctions, as represented, for example, by Russian and Bulgarian, differ clearly in their functions. ±AD signals aspect based on actionality, since +AD presupposes and subsumes the actional value [+t]. This does not imply any dominance of actional values; the aspectual values always have scope over the actional ones. The Bulgarian-type use of perfectives and imperfectives is purely actional and concerns the [±t] parameter. Here, too, as we shall see, aspectoactional parameters interact to the effect that the viewpoint values (±INTRA, ±POST) have scope over the actional ones. A few comments should be made here on the use of morphological perfectives and imperfectives as ±AD and ± T and the differences between them. First of all, adterminal items display a much more general and systematic use than +T-markers do. It has already been noted that one indication of a completed grammaticalization process is a higher degree of generality of use. There are thus never one-to-one correspondences between adterminals and +Tmarkers. For example, though Ossetic +T-marked items are often claimed to be equivalent to Russian perfectives, fcek-kodta 'did' corresponds to sdelal, whereas the simplex kodta corresponds to both délai and sdelal (Miller 1962: 139). Another difference mentioned above is the blocking of +AD with phasal verbs such as 'begin', as it is not possible to begin the very attainment of a crucial limit. A +T-marker easily combines with such verbs, since it is quite natural to start an action directed towards a crucial limit. Thus, Polish verbs expressing the beginning or end of an activity or process combine with imperfectives. The Lithuanian perfective exhibits far fewer such constraints, and Hungarian + T-marked actional phrases also rather freely occur with phasal verbs, e.g., Lithuanian èmé atprasti 'began to give up (a habit)', Hungarian kezdett megbolondulni 'began to go crazy'. Many speakers also accept Hungarian kezdett megirni 'started to write (to completion)'. All this shows us, again, that the Lithuanian and Hungarian items in question are not +AD items. 9.5.1.

Capability of future time reference and incapability of presentness

It was also stated above that -PAST (+AD) cannot apply to an event in progress and gets future time reference if related to Os, e.g. Russian napisetpis'mo 'will write a / the letter', perejdet most 'will cross the bridge', Czech napfse 'will write'. It should, however, be noted that all finitransformatives easily suggest future time reference with —PAST items. At O s , a telic event can only be viewed in its course or during

Viewpoint operator! in European languages

143

its ongoing preliminaries, and this view implies that a possible accomplishment can only occur in the future. In other words, a goal-oriented event is bound to the Os perspective with the natural implication of posteriority. This kind of future reference is, for example, a natural reading of Classical Greek Present forms of [+t] verbs, e.g., néomai 'go, come'. This is universally valid for the non-past of finitransformatives, in particular for [+tf, +mom] actional phrases. The possible preliminaries are observed in the deictic present, but the crucial limit must lie, if anywhere, in the future. The question "What is going on right now?" is answered more satisfactorily by nontransformative items. Thus, to express 'is writing a letter', Lithuanian uses raso laiska, i.e., with the +T-unmarked ("imperfective") item raso rather than the +T-marked ("perfective")paraso. Finnish uses the —T-marked lukee kirjaa 'is reading / reads the book' (with partitive object), whereas the -T-unmarked lukee kirjan 'reads and finishes reading the book' (with the object in the total case) has future time reference. However, the decisive criterion of +AD is not the capability of future time reference but the total incapability of presentness. The notion of adterminality in the primary deictic "nunc" perspective means a more direct future projection. The object of observation is, again, not an event moving towards a crucial limit, but it is the very attainment of that limit. Note that Lithuanian perfectives such as parasyti 'write (to completion)' do not share the total "incapability of presentness" typical of Russian or Polish perfectives such as napisat', napisac. Their main non-past function is not to express prospectivity. The Latvian so-called perfective Present behaves similarly. In the same way, the Present of a -I-T-marked Hungarian actional phrase just suggests that the possible completion can only take place in the future, e.g., megirja a levelet 'writes / is writing (to complete) the letter'. Similarly, the Bulgarian perfective Present does not display the future time reference typical of -PAST (+AD) items. To express prospectivity, languages such as Lithuanian, Hungarian and Bulgarian use special items. Hungarian uses the fog Future with both transformatives and nontransformatives, e.g., —PAST (+PRO) x [+t] megfogja irni a levelet 'will write (to complete) the letter'. Bulgarian possesses a special Future, consisting of the particle ste + Present and predicting events of both transformative and nontransformative actional contents, e.g., -PAST (+PRO) x [+t] ste doide 'will come'. Georgian —PAST items determined by +T-markers get readings in the sense of does (to completion)' and tend, as expected, towards future time reference, e.g., dacer 'tu l'écriras', in which case they may be translated by Russian perfectives, e.g., napises' 'you will write' (Vogt 1971: 183-184). Hungarian non-past items such as atmegy a hidon 'crosses the bridge' can, like German überschreitet die Brücke, exPress an ongoing telic event, projecting its conclusion into the future. If the Russian Perfectives were +T-markers, we would expect them to behave similarly. However, Perejdet most 'will cross the bridge' adds the +AD meaning, directing the attention directly to the event in its actual conclusion. Note also that the Hungarian and Ger-

144

Lars Johanson

man items, unlike the Russian ones, combine with explicit +PRO markers, e.g., ât fog menni a hidon and wird die Brücke überschreiten 'will cross the bridge'. 9.5.2.

+ AD-marking of global events and +T-marking of subevents

A further important criterion is how perfectives relate to global events and possible subevents. Adterminality envisages the whole event as such, whether it consists of one single event or subevents. +T-marking, however, may also be used for the actional characterization of the subevents. As we have seen, a transformative actional phrase may be quantitatively reinterpreted as [+ser], which removes the crucial limit and makes it more capable of presentness. If it still contains a +T-marker, it is in order to assign transformativity to the subevents. Thus, in their primary reading as denoting Os-oriented single events, Finnish and Estonian transformative actional phrases with total objects are interpreted as having future time reference, e.g., Estonian küpsetab koogi 'will bake a cake'. When a [+ser] reading removes the limitation and makes the whole actional phrase more capable of presentness, the +T-marking total object specifies the actional properties of the subevents. If Russian perfectives were +T-markers, they would be expected to function similarly. However, they do not signal [+t] of subevents, but operate, as -t-AD markers, on the actionality of the global event. It is the attainment of the crucial limit of the global event - not of the subevents - that is envisaged by a + AD operator. Os-oriented +T-marked non-past items typically refer to an ongoing global event that is not completed itself but consists of telic subevents. Thus, the Bulgarian combination -PAST (+INTRA°) x [+t] mainly expresses repeated events. Similarly, Lithuanian, Latvian, and Hungarian perfectives can refer to ongoing global events, the +T-marking only signalling the transformativity of the subevents, e.g., Latvian katru dienu uzraksta pa vëstulei, Hungarian minden nap megir egy levelet 'writes (and finishes writing) a letter every day'. Here, the function of the "perfectives" is not to denote habituality, but to characterize the subevents actionally. The pluri-occasional reading arises from the [+ser] reinterpretation. The pluri-occasional global event, seen in progress at O s , is a set of events, each of which is actionally characterized as [+t]. The Bulgarian, Lithuanian, Latvian, and Hungarian perfectives are thus not +AD categories. Only +T-marking perfectives may express the telic character of subevents. For the sentence just cited, the imperfective Present would be the only choice in Russian: kazdyjden'piset odnopis'mo. Here, the —AD aspect operates on a [+ser] actional content used to denote a pluri-occasional global event. There is no device to mark the subevents for [+t] actionality. Adverbials expressing unlimited repetition such as 'every day' (Lithuanian kasdien, Russian kazdyj den', etc.) combine with +T-marked actional phrases but not with -PAST (+AD) items. If Russian perfectives were strictly +T-marking, they would also be used to assign transformativity to subevents. If the Bulgarian, Lithuanian, Latvian,

1 Viewpoint .operators in European languages

145

and Hungarian "perfectives" were adterminals, they would only be used to envisage the attainment of the endpoint of the global event, which is incompatible with the present perspective.

9.5.3.

Imperfectivization

Slavic secondary imperfectives derived from prefixed stems may, as we have seen, function as — T-markers. Obligatory secondary imperfectivization is sometimes considered the major criterion for Slavic aspect grammaticalization, and it indeed seems to provide arguments for the presence of ±AD oppositions. In languages such as Russian, —T-marked items of this kind have developed into aspectual imperfectives. From a simple imperfective pisat' 'write', a perfective perepisat' 'rewrite, copy' is derived by means of a transformativizing preverb that also modifies the lexical meaning. The non-past item perepisët 'will rewrite' exhibits the incapability of presentness to be expected from a -PAST (+AD) item. As should also be expected, a secondary imperfective item pere-pis-yv-at', derived from pere-pisat', does not share this incapability, since it is a —AD item: perepisyvaet 'rewrites'. The Lithuanian +T-marked per-rasyti 'rewrite' corresponds to both +AD perepisat' and —AD pere-pisyvat' and is thus obviously not a +AD item. The item per-ras-inèti 'rewrite, copy (repeatedly)' is still an iterative here. Georgian uses for 'rewrite, copy' a +T-marked verb, which, like all finitransformatives, naturally suggests future reference in non-past items, e.g., gadacers 'will rewrite'. The Russian secondary imperfective Present, however, is a —AD item with full capability of presentness, e.g., pere-pisyvaet 'rewrites, is rewriting'. The Georgian lexeme is just a transformative, and it even lacks a lexically adequate nontransformative equivalent, the only [—t] counterpart available being the simple cers 'writes, is writing'.

10. Interaction of viewpoint operators with phase structure types 10.1. Typical realizations The interdependence of viewpoint operators and the actional contents they apply to has been emphasized above. The choice of aspectoactional combinations produces different types of interactional readings. Inherent phase structure properties in terms °f [±t], [±dyn], [±mom] and other distinctions determine the realizations. The typical realizations of post-, intra- and adterminality with finitransformative, initiotransformative and nontransformative actional phrases can be roughly illustrated as follows: \

146

Lars Johanson

[+tf]

Postterminality | g>

H-ti]

®

1

[-t]

x

1

Intraterminality | g>

Adterminality I ®

x

The relevant limit of nontransformatives is marked with x, and the crucial transformational limit of transformatives with ®. The symbol " stands for the viewpoint, and dots (...) mark the variable space within which it may be situated. Grammatical viewpoint categories always interfere with actional semantics, and it is important to describe the specific connections. Different aspects operating on one and the same actional content produce different actional readings. As we have noted, some of the interactions in narrative discourse allow textual interpretations of 'entry', 'dwelling' and 'exit'. Such readings are, as has been stressed, not identical to the functions of phasal modes of action, e.g., ingressivity. The latter signal properties of a given action as such, whereas aspects signal the conceptualization of an event as an occurrence of the same action, as action taking place. Viewpoint operators do not select certain phases, as phasal devices do. They do not pick out certain portions of the actional content, but may just give prominence to phases typical of the respective actional classes. Thus, actional distinctions cannot compensate for viewpoint distinctions, as their functions are different. For example, a system without a ±INTRA or ± AD opposition does not express any PFV vs. IPFV duality in a systematic way. Certain complementary distributions of viewpoint operators over actional classes do not prove any functional identity. Even elaborate systems of actional devices cannot make up for the lack of viewpoint operators. They may have roughly similar effects with respect to the representation of non-transitional and transitional events in discourse, but cannot express viewpoint meanings. Viewpoint semantics does not only serve to describe events in terms of 'situation' vs. 'situation change'.

10.2.

Realizations of ±INTRA

The ±INTRA opposition imposes viewpoint values on actional phrases of different phase structures. It has been stressed above that [±t] is not equal to ±INTRA and that, for example, the Bulgarian imperfective vs. perfective and Imperfect vs. Aorist distinctions are of different kinds. If perfective and Aorist are both defined as expressing "boundedness", whereas imperfective and Imperfect are both taken to express "non-boundedness", the combinations "imperfective Aorist" and "perfec"

n

Viewpoint Operators in European languages

141

tive Imperfect" get contradictory meanings that are impossible to account for. The combinations of ±INTRA and [±t], the latter sometimes signalled by presence vs. absence of +T-marking, do not represent complex aspectual meanings, but different predictable aspectoactional realizations of viewpoint values operating on different actional contents. The interactional meanings become unintelligible unless the basic categories are defined independent of each other. 10.2.1.

+INTRA

-f INTRA highlights the cursus of the actional phrase meaning in a much more explicit way than the —AD kind of IPFV. The view 'from within' yields textual readings of 'dwelling in the event'. +PAST (+INTRA) items are thus not propulsive, but have a ruptive effect in narratives, denoting events that form a temporal or explicatory background for the main events. Their textual behaviour is often similar to that of +PAST (-AD) items. Interaction of this cursus- oriented aspect with inherent phase structures may yield different processual, progressive, stative, preliminary and other readings. 10.2.1.1.

+INTRA x [-t]

Intraterminality is most naturally applied to nontransformatives. The values +INTRA and [—t] are highly compatible, though by no means identical. +INTRA needs a cursus of some saliency to operate on, which is the case with [—t] items. Events expressed with [—t] items may of course be conceived of as having limits, though none is crucial. +INTRA x [—t] views the event within these limits, without any emphasized beginning or end. It is not enough to define the result of this interaction as a mere 'state', since introspection is an essential element. Still, +PAST (+INTRA) x [-t] may exhibit a textual behaviour roughly similar to that of +PAST (-AD) x [—t]: cf. Modern Greek éghrafe ghrâmata 'wrote / was writing letters' and Russian pisalpis'ma. +INTRA signals that something is going on at an O. The combination +PAST (+INTRA) x [—t] tells us that a portion of the action is already achieved at Os. Dik's observation on English progressives holds for +INTRA x [—t] in general: if it is truly said that, at some point of time, an atelic state of affairs obtains, for example, he is painting, it may be concluded at some later point that this state of affairs "has obtained": he has painted (1989: 94). Whether a past event expressed by +PAST (±INTRA) can afterwards be interpreted as having occurred or not is a matter of actional content, the decisive factor being compatibility with the notion of partial realization. +INTRA combines with both [-t, -l-dyn] and [-t, -dyn] actional phrases. With he latter, it denotes a static dwelling in the event, without internal progression, e.g., rench savait, Turkish biliyordu 'knew'. With [-t, +dyn] actional phrases, it may

148

Lars Johanson

yield readings of a dynamic dwelling in the event, with possible internal progression, e.g., French travaillait, Modern Greek dhuleve, Turkish çahsiyordu 'was working'. Note that these readings are not necessarily progressive in the sense of a high-focal intraterminality. +INTRA can operate on an actional content that is homogenized in the sense of [-t-tf] + [+hom] > [—t], e.g., Finnish on korjaamassa kelloa 'is repairing the clock', where the actional phrase korjata kelloa is homogenized by the partitive, and the +INTRAHF operator marked by -massa. It can also operate on an actional content that is homogenized in the sense of [-btf] 4- f+ser] > [~fj> in which case the intraterminally viewed global event consists of repeated and possibly pluri-occasional subevents. Pluri-occasionality yields processual readings of so-called habitual iteration, to which we shall return. As we have noted, +INTRA is not meaningful with specifications of the entire temporal extension, e.g., French *écrivait deux heures. (An adverbial such as pendant deux heures would rather measure the period of intraterminal observation.) The Bulgarian imperfective Imperfect, which signals +PAST (+INTRA) x [—t], is not used in sentences denoting uni-occasional events with delimitative 'for X time' adverbials. This restriction is only intelligible if the Imperfect is defined as +INTRA and not simply as IPFV, since —AD items freely combine with 'for X time' adverbials. With pluri-occasional [+ser] readings, imperfective Imperfects are also compatible with 'for X time' expressions, if the specification of the temporal extension refers to each subevent, e.g.,peese pesenta tri minuti 'used to sing the song for three minutes [on each occasion]' (Lindstedt 1985: 205). +INTRA x [—t] does not combine with 'in X time' adverbials when the reference is uni-occasional. Under [+ser] readings, however, the combination is possible: in this case, the accomplishment time of each telic subevent is measured, e.g., Bulgarian +PAST (+INTRA) x [+t] + [+ser] > [—t] peese pesenta za tri minuti 'used to sing the song in three minutes [on each occasion]'. 10.2.1.2.

+INTRA x [+ti]

Since [+ti] actional phrases denote both a transformation and a following posttransformational phase, the combination +INTRA x [+ti] signals an introspective view of an event excluding its crucial starting point, the very transformation. +PAST (+INTRA) x [+ti] may thus highlight the cursus phase of the lexical meaning, the posttransformational statal phase of the actional content, e.g., Modern Greek katalâvene 'understood (= was aware of)', krivôtan 'was hiding, hid (= kept himself / herself out of sight)', Turkish duruyordu 'was standing, stood'. These readings contrast with -INTRA readings suggesting initial attraction (10.2.2.1.1). Note that the non-dynamic posttransformational phase highlighted by +INTRA x [+ti] combinations may be objectively identical to the one highlighted by corre-

Viewpoint operators in European languages

149

sponding +POST x [+ti] combinations, e.g., is hiding = has hidden, understands = has understood, Classical Greek rhigéô T shudder [with fear] = érriga 'I am terrified', Turkish oturuyor 'is sitting' = oturmustur 'has sat down' (cf. 10.3.1.2). The two options may differ with respect to focality. As was noted bove, Maltese qedmarked intraterminals are less focal than high-focal +POST items of the same initiotransformative verb, e.g., qedjorqod 'is (usually) sleeping' vs. rieqed 'has fallen asleep, is asleep'. Note also that the distinction possible between (i) hid and (ii) was hiding becomes impossible if the +INTRA item is high-focal and thus does not apply to all actional phrases. The English past item understood is ambiguous between (i) and (ii), since *was understanding is blocked for (ii). In languages lacking ±INTRA distinctions, [+ti] actional phrases are systematically ambiguous, e.g., German erkannte (i) 'became aware of, (ii) 'was aware of. 10.2.1.3.

+INTRA x [+tf]

The combination +INTRA x [+tf] signals an introspective view of an event excluding its crucial endpoint, the very transformation. It is found in all ±INTRA languages. Thus, the Georgian "determined Imperfect" denotes an event "qui se déroulait dans le passé vers un but, un terme" (Vogt 1971: 189). Among the Kartvelian languages, Svan readily applies +T-marking to intraterminal transformatives. +INTRA x [+tf] is also expressed by the +T-marked Bulgarian perfective Imperfect. However, +Tmarking is relatively dispensable in Present tenses. Thus, Lithuanian may use the simple Present, e.g., mirsta 'dies', [eina 'enters'. The Kartvelian use of +T-marking preverbs in the Present emerged rather late. The combination does not tell us whether the crucial limit is actually reached after the viewpoint or not. This fact has often been misunderstood. It has been called an "imperfective paradox" that an utterance containing an "imperfective telic" verb can be true even if it is known that the endpoint was never reached (Dowty 1977). It would be more correct to state that if a telic event expressed by a [+tf] item is said to be in progress at some point, it cannot later be concluded with certainty that » has been achieved (cf. Dik 1989: 94). The combination +INTRA x [+tf] may thus produce preliminary readings of partial realization, preparation for completion, 'mminence, inclination, propinquitivity, conativity, etc. (Johanson 1971: 202-206). he readings vary according to the semantics of the actional phrase and the context. It may seem that intraterminality is incompatible with momentaneous finitransform atives and that the latter are thus monoaspectual, combinable with -INTRA only. ^ere appears to be little need for +INTRA x [+tf, +mom], an inside view of an ct ional content consisting of a limit alone, without a salient cursus to look into. The c °mbination +INTRA x [+tf, +mom] is nevertheless used, referring to the time Mediately before the crucial limit of which the action consists. It may thus have

150

Lars Johanson

,? [-t]

Intraterminality may, as we have noted, operate on [+ser] actional phrases denoting pluri-occasionality. The reading [+ser] may be due to a quantitative reinterpretation without any overt signal. The objective property of repetition is not a sufficien

Viewpoint operators in European languages

151

reason to use a +INTRA operator, but the decisive criterion is the inside view, which can be applied to both uni- and pluri-occasional global events. The global event is presented by +INTRA regardless of the actional properties of the subevents. If the latter are telic and basically represented by [+t] actional phrases, the combination +INTRA ? [+t] + [+ser] > [-t] is used. Examples of +PAST (+INTRA) ? [+tf] + [+ser] x [-t] are Turkish her gün saat altida kalhyordu, Bulgarian stavase vseki den v sest casa 'got up every day at six o'clock'. The basic phase structure may be [+tf, +mom], e.g., Romanian exploda о bomba 'a bomb used to explode', Modern Greek to évriske 'used to find it'. The basic structure may also be [+ti], e.g., Turkish her gün orada oturuyordu 'sat / was sitting there every day'. +T-marking languages may use the presence and absence of marking to characterize the subevents. Absence basically stands for global events whose subevents are atelic. Thus, Bulgarian +T-unmarked +PAST (+INTRA) items, imperfective Imperfects, are regularly used for intraterminally viewed global events consisting of atelic subevents. An intraterminal perspective on a global event consisting of telic subevents is usually expressed by +T-marked +INTRA items. The distribution of functions in this aspectoactional interaction is clear-cut: -flNTRA marks the global event aspectually, and +T marks the actionality of each subevent for [+t]. Thus, a Bulgarian or Macedonian +T-marked +PAST (+INTRA) item, a perfective Imperfect, may apply an intraterminal perspective to a whole set of unrestrictedly repeated telic subevents, e.g., in Bulgarian subordinate clauses such as veski pat stom ja poglednese 'each time he glanced at her'. Bulgarian -PAST (+INTRA0) x [+t], the perfective Present, typically expresses pluri-occasional telic events. +T-marked Imperfects are typically used in the same way in Kartvelian, e.g., Georgian and Svan. Italo-Croatian uses +T-marked +PAST (+INTRA) items to express an intraterminal view on a sequence of telic events, e.g., Sa vrzasajasuc, rivasa nonda, sa skinasa a njimi ponesasa mblika gor 'He used to ride away, arrive, get off and bring milk up tothem'(Breu 1992: 110). A similar view on global events with repeated telic subevents is achieved with +T-marked Present items, e.g., Bulgarian sutrin izleze, vecer se värne 'leaves in the morning and returns in the evening' (Stankov 1976: 31); compare the Lithuanian example already quoted: kasdien parasè po vienq laiskq 'writes (and finishes writing) a letter every day'. The Lithuanian or Bulgarian kind of "perfectivity" is thus by n o means incompatible with present time reference. There is nothing contradictory about a sequence of telic events viewed in the natural intraterminality of a Present le nse. On the other hand, it is clear that ±AD languages cannot use their quite different kind of "perfectivity" in such cases, but only -PAST (-AD) items. However, the type of aspectoactional interaction discussed here is relatively lim•ted. Bulgarian +T-marked habitual Imperfects are rather rare and can be replaced by u nmarked ones (Maslov 1959: 272). Thus, the latter may also be used for intratermi-

152

Lars Johanson

in

trnt-,", i" •

nally viewed global events consisting of telic subevents. Italo-Croatian +T-marked Imperfects are replaced by unmarked ones in the speech of the younger generation (Breu 1992: 115-116). Such simplifying steps towards the expression of pure aspect at the expense of actional characterization of subevents bring the systems closer to the Romance and Turkic ones. This simplification is formally opposite to the one that has taken place in ±AD languages, where the +T-marking perfectives are formally maintained but have come to function as +AD operators. On the other hand, the developments are similar inasmuch as + AD also expresses aspect without regard to actional properties of possible subevents. 10.2.1.5. Combinations with -f-INTRA10" Intraterminals of different focality degrees interact differently with certain actional contents. High-focals are especially sensitive to the phase structure properties [±ser], [±dyn] and [±mom]. +INTRA items naturally start as high-focals with [—t, -t-dyn] and [+ti] actional phrases. If they spread to combine with the remaining phase structure classes, [4-tf, —mom], [—t, —dyn] and [+tf, +mom], they are eo ipso lowfoe als. +INTRAHF items are usually more restricted with respect to [+ser] than low- and nonfocals. Preaspectuals often reject pluri-occasional readings, e.g., German war am Trinken 'was drinking (on one occasion)'. The English progressive was drinking tends more to uni-occasionality than Spanish bebia or Turkish içiyordu 'was drinking, drank'. If used pluri-occasionally, it suggests a narrower presentness in the sense of temporariness, e.g., was drinking coffee every day. +INTRAHF prefers to operate on [+dyn] actional phrases. High-focal intraterminality is particularly fertile with dynamic contents implying internal evolution. H-INTRA^ items are often strongly preferred to represent ongoing dynamic - transitional or dynamic non- transitional, i.e. processual - events implying gradual change, particularly progress observable in gradually produced effects, e.g., is improving. Such combinations are often accompanied by expressions of graduality and speed, meaning 'little by little', 'slightly', 'rapidly', etc. To represent ongoing events expressed by [—dyn] actional phrases, items of lower focality are sufficient, e.g., the English Simple Present. A simple present is often sufficient to imply a narrower "nunc" if the event is of a mental nature. Irish so-called stative verbs such as silim 'I think, I am thinking' may have PROG meaning without being used in the special -(-INTRA™1 form. Combinations such as English *was knowing, Maltese *qed jaf '*is knowing' are usually blocked. Most preaspectual items show the same constraint, e.g., German *war am Wissen 'was knowing'. Even in cases of rather narrow temporariness, many Romance and Germanic -f INTRA1111 items refuse to operate on [—dyn] verbs of spatial location such as 'hang', 'sit', 'stand'. The same holds for postural verbs in some Turkic languages (Johanson 1995). Preaspectual progressives

Viewpoint operators in European languages

153

used with [—dyn] actional contents often have derogative connotations. Some highfocals avoid passives, since non-agentive roles are often linked with [—dyn] actional contents. When the expression of high-focal intraterminality is renewed in a language, it does not affect [—dyn] actional phrases. On the other hand, when older intraterminals lose focality, becoming more and more restricted to nonfocal or modal uses, +INTRA x [-dyn] combinations tend to retain low-focal functions longer than oth­ ers. The former East Armenian Present tense is modal today, except with a few verbs, e.g., gitem T know'. The present-day use of Turkish so-called Aorist, formerly a cat­ egory of higher focality, is often less modal with [-dyn] verbs than with others, e.g., ister 'wants', bilir 'knows'. Similarly, the Lezgian -da form preserves its older meaning with verbs such as k'anda 'wants', cida 'knows' (Haspelmath 1994: 276). Such residual meanings obviously result from aspectoactional interaction. It is, however, necessary to stress the relative nature of the ability of high-focals to operate on more or less dynamic actional contents. Some relatively high focals more readily accept less dynamic actional contents than others. The Icelandic vera 'be' + Present Participle periphrasis is less restrictive than the vera ad + Infinitive construction, e.g., er sofandi 'is asleep'. Portuguese high-focals are relatively open, e.g., esta a gostar 'is liking' (Oliveira & Lopes 1995: 107). Maltese high-focals combine with most [—dyn] actional phrases to express temporariness. The same holds for certain English high-focals, e.g., is feeling young. There is no absolute 'unmarked present time context' in which high-focals are obligatory. If high-focals of two languages differ in their use in certain cases of low actional dynamicity, this means that the defocalization drift has advanced further in one of the items, not necessarily that the events denoted are conceptualized differently in terms of 'state', 'process', etc. High-focal intraterminality is fertile with non- momentaneous transformatives and less fertile with momentaneous ones, though it may yield imminential or propinquitive readings (see below). The combination -l-INTRA101 x [+tf, —mom] occurs in many languages with preliminary readings: a preliminary phase preparing for the attainment of the crucial limit is viewed as going on at an O, e.g., Icelandic er (alveg) ad sofna 'is (just) falling asleep'. But -I-INTRA1117 x [+t] may also refer to events likely to occur in the immediate future without any meaning of ongoing preliminar­ ies, the intraterminal viewpoint being situated immediately prior to the crucial limit ('is / was close to V-ing', 'is / was about to V ) . In some languages, -(-INTRA1113 items are the only device for expressing imminence with transformatives. Imminen­ tial readings often result from -blNTRA™3 operating on [+mom] actional contents, e g-, Icelandic er ад fara 'is leaving / about to leave', Danish er ved at кфге 'is deParting / about to depart', Portuguese esta a ganhar 'is winning / about to win'; cf. Cornrie's 'prospective' aspect (1976: 64-65). Transformative phasal verbs such as °egin and finish either do not combine with high-focals at all or get imminential read-

154

Lars Johanson

ings, e.g., Icelandic vera ад + byrja 'begin'. Some high-focals are ambiguous with respect to imminence and ongoing process, e.g., is leaving, whereas others are un­ equivocally imminential, e.g., German ist im Begriff abzufahren 'is [on the point of] leaving'. Certain preaspectuals prefer the feature [-(-controlled] to express an ongo­ ing process, whereas non-agentive and non-purpose actional phrases tend to express imminence, e.g., Swedish Tâget hâller pâ att gâ 'The train is about to leave' (not *'already leaving'). High-focal intraterminals may thus yield prospective meaning and eventually develop into prospectives. Preaspectual high-focals often suggest propinquitive readings ('be close to V-ing, but not V ) , usually with [+t] actional phrases, e.g., Danish var ved at d0, men ... 'was on the verge of dying, but ...'. The Lithuanian construction 'was' + beprefixed Present Participle typically expresses a thwarted event (Mathiassen 1996). For Yiddish ikh halt baym shraybn T am about to write', see Ebert, this volume.

10.2.2.

-INTRA

Nonintraterminals, which combine with all kinds of actional phrases, are unmarked for intraterminality and thus have negatively defined values. They present an integral view of the event, suggesting, unless the contrary is stated otherwise, that the event is carried out and comes to an end. I have referred to them as "pseudoperfectives", since they differ essentially from adterminals, "perfectives proper", but exhibit a partly similar textual behaviour (Johanson 1996). +PAST (-INTRA) items are propulsive and thus used in narrative texts to indicate the succession of events. Interacting with +INTRA and +POST items, they may express situation changes occurring on the basis of ongoing situations. On the other hand, they are not explicitly limit-oriented as -f-AD items are. They do not give any special prominence to limits in the meaning of the actional phrase and may indeed suggest rather cursus-oriented readings. This fact is often ignored in attempts at reducing +AD and -INTRA to a common PFV denominator under the assumption that IPFV items are cursus-oriented and PFV items limit-oriented. 10.2.2.1.

-INTRA x [+t]

The values -INTRA and [+t] are, though not identical, highly compatible with each other. As -INTRA does not deny the inclusion of the crucial limit into the view, the combination +PAST (-INTRA) x [+t] thus typically refers to accomplished past telic events. Transformative -INTRA items are partly similar to adterminals in their textual behaviour. Nonintraterminality is sufficient to produce this similarity. In narrative discourse, —INTRA x [+t] may suggest readings of 'entry' into the event and 'exit' from it ("terminal attraction"; Johanson 1971: 211), but there may also be more cursus-oriented readings.

Viewpoint operators in European languages 10.2.2.1.1.

155

-INTRA x [+TI]

With initiotransformatives, —INTRA items may suggest, among other interpretations, an initial attraction, putting an accent on the initial crucial point and highlighting the transformation. +PAST (-INTRA) x [+ti] combinations with initium-oriented 'entry' readings are, for example, Turkish yam 'lay down', bildi '(suddenly) knew', Modern Greek katâlave 'understood (= became aware of)', kriftike 'hid (= put himself / herself out of sight)', East Armenian atec '(suddenly) hated, began to hate'. These readings contrast with the corresponding +INTRA readings dealt with above (10.2.1.2). Some Italo-Croatian verbs, e.g., mat 'know' and imat 'have', have copied the [+ti] phase structure of corresponding Italian verbs, e.g., sapere, avere, and thus get analogous initium-oriented readings with +PAST (-INTRA) items ('learned', 'got'), where other Slavic languages such as Russian or Serbian use inceptive modes of action. Verbs such as cut 'hear' und vit 'see' are also used as [+ti] verbs, e.g., sa cuja 'heard', sa vidija 'caught sight of (Breu 1992: 115). As was stressed above, however, initial attraction should not be confused with explicitly inceptive or ingressive modes of action. It is an important fact that +PAST (-INTRA) x [+ti] items also have more cursus-oriented readings without special attention given to the initium, e.g., Turkish oturdu 'sat (for a period of time)'. It would thus be wrong to claim that it is the function of —INTRA items to select the initium with verbs of this type, as is often assumed for PFV items. The idea that the ±INTRA opposition produces pairs of stative vs. ingressive meaning appears to be an unjustified objectivization of the aspect distinction in question. 10.2.2.1.2. -INTRA x [+TF] With finitransformatives, —INTRA items may suggest, among other interpretations, a final attraction, putting a certain accent on the final crucial point and thus highlighting the transformation. +PAST (-INTRA) x [+tf] combinations with finis-oriented 'exit' readings are, for example, Modern Greekpéthane 'died', éxtise éna spi'ti 'built a house', Italo-Croatianponija 'brought', Maltese hareg 'went out', Turkish kitabi okudu 'read the book', Portuguese escreveu a carta 'wrote the letter'. These are all examples of -INTRA x [+tf, -mom]. As for the combination -INTRA x [+tf, +mom], it can, since there is no salient cursus, only refer to the crucial limit of which the event consists, e.g., Portuguese ganhou 'won', Modern Greek vrike 'found'. But it is not monoaspectual, as the combination +INTRA x [+tf, +mom] is also possible. -INTRA x [+tf] combinations do not produce PFV items of the +AD type, explicitly highlighting the finis. They just report that the event has obtained, not excluding the crucial limit from the view. The narrative behaviour of +PAST (-INTRA) x [+tf] is partly similar to that of +PAST (+AD) items. Both types are propulsive and

156

Lars Johanson

• «i. >>лv.'o'V"•:

may express a transition to a new situation. This is considered a main PFV property. But -INTRA is not a PFV of the +AD type, highlighting the crucial limit. +PAST (—INTRA) x [+tf] does not select the final phase, as PFV pasts are supposed to do. The accent on the finis should not be confused with implicitly signalled egressivity. The possible interpretation of the event as an accomplished whole results from the inherent phase structure. The Maltese Simple Past expresses completion with transformatives only. Modern Greek éghrapse éna ghrâma 'wrote a letter' corresponds to Russian +PAST (+AD) napisal pis 'mo only in the case of undivided object reference. Hungarian Pasts operating on [+tf] actional phrases differ clearly from +PAST (+AD) realizations and are interpreted by De Groot as "complete / perfective" situations (1984). A past tense of afinitransformative requires no qualified terminality to suggest readings of 'completion', 'accomplishment', 'achievement'. Readings of 'result', 'consumption', 'exhaustiveness', etc. are also effects of contextually determined transformati vi ty (Johanson 1971: 208-213). The question whether [+tf] is supported by explicit +T-marking or not is relevant for languages such as Lithuanian, Bulgarian, Macedonian, Italo-Croatian, and Georgian. Thus, the Bulgarian perfective Aorist contains two features that can be assigned distinct functions: the viewpoint operator -INTRA gives a nonintraterminal view of an event whose actionality is specified by +T-marking. The Georgian Aorist, a +PAST (-INTRA) item opposed to the +PAST (+INTRA) Imperfect, exhibits two variants, of which only one is a +T-marked transformative, e.g., daçera 'he wrote (and finished writing)', and an unmarked form cera (of very limited use today). The item is thus part of two oppositions, one aspectual and one actional: +INTRA x [—t] cerda 'was writing' -INTRA x [+t] daçera 'wrote' vs. -INTRA x [-t] çera 'wrote' The opposition between +T-marked and -t-T-unmarked -INTRA items is typical of Georgian and Laz (Zan), but not of Svan, which is consistent in +T-marking transformatives. +T- marking of +PAST (-INTRA) x [+tf] combinations is sometimes not needed, since the actional phrase is obviously [+tf] anyhow, e.g., Lithuanian mire 'died', pradêjo 'began'. However, +T-marking is often required with actional phrases denoting actions that may be either telic or atelic, e.g., Lithuanian parasé 'wrote' (and finished writing)' as opposed to a less goaldirected rase 'wrote'. As was noted above, Italian influence has led to decreasing +T-marking of some verb categories in Italo-Croatian. Note again that a +T-marker, unlike the viewpoint operator -f AD, operates on basic events. In cases of -INTRA x [+t] + [+ser] > [-t], i.e., when nonintraterrninality is applied to the expression of a global event consisting of repeated occurences of telic subevents, +T-marking naturally signals the actionality of each subevent, e.g.. Bulgarian napravi tova njakolkopäti 'did this several times'.

Viewpoint operators in European languages

157

As we have noted, 'in X time' adverbials, which measure accomplishment time, naturally combine with transformatives but do not cooccur with [—t] actional phrases under uni-occasional readings. When +PAST (-INTRA) is applied to an actional phrase determined by an 'in X time' adverbial, the actionality is thus usually [+tf], e.g., did it in ten minutes. Delimitative and perdurative modes of action require a specification of the time interval filled with the event and thus produce transformative actional phrases meaning 'spend [a certain time period] V-ing', the basic action eing atelic. The time-specifying expression is no mensural adverbial of the 'for X time' type. For example, the +T-marker of the Bulgarian perfective Aorist signals the [+tf] actionality of a temporally delimited atelic event, e.g., +PAST (-INTRA) x [+tf] pospa dva casa 'slept for two hours'. Since the actional phrase is [+tf], coresponding Russian delimitatives and perduratives only choose the +AD aspect, e.g., PAST (+AD) tamprozil tri goda 'spent three years there'. Compare +PAST (-AD) terns based on nontransformatives and provided with 'for X time' adverbials, e.g., zil tri goda tam 'lived there for three years'. 10.2.2.1.3.

-INTRA x [-T]

Nonintraterminals may also operate on nontransformatives, which is not possible with the adterminal type of PFV items. -INTRA x [-t] suggests integral readings of realization of a portion of the actional content without special attention to limits. Even an event lacking salient limits has its initium, cursus and finis; none is excluded from the view. As a result of the weak terminality value of —INTRA and the vague phase structure of [—t], the combination allows various initium-, cursus- and finisoriented readings, e.g., '(and then) read', 'read (for some time)', 'read (and then)'. It thus has a propulsive function in narrative discourse. In its general textual behaviour, however, it is more similar to the nonadterminal type of IPFV than to the adterminal type of PFV. Even —INTRA x [—t, —dyn] is indeed a possible combination, e.g., French contint 'contained', Turkish bekledi 'waited', Modern Greek élipse 'lacked'. Static verbs are often claimed to be "monoaspectual", i.e., not to combine with PFV items. The combinability may be reduced, e.g., in Modern Greek, where a few [—t, —dyn] verbs such as kséro 'know', periéxo 'contain', periméno 'wait' lack Aorist forms (Sasse 1991b: 15), but, on the whole, the constraints concern the +AD type of PFV rather than the -INTRA type. Since —INTRA items lack the marked holistic nature supposed to be typical of PFV, they cannot stress totality as the adterminal counterparts do. There is thus no basis for a clear-cut 'progressivity' vs. 'totality' duality with [-t, +dyn] actional Phrases. -INTRA x [-t] expresses a portion of nontransformative actional content without any marked phase meaning. Due to the negative viewpoint value, it may get slightly limit-orientated readings, a weak initial or final attraction. Thus, items such

158

Lars Johanson

as Turkish agladi and Albanian qau (from agla- and qan 'weep') may put a certain accent on the initium, '(and then) cried, burst into tears', and thus be translated by inceptives, e.g., Russian zaplakal 'began to cry'. The combination often expresses the mere occurrence of a past event of some duration ('for a certain time'), e.g., Modern Greek dhûlepse 'worked', Turkish okudu 'read', Italian scrissi 'wrote', Maltese kiteb 'wrote'. Italo-Croatian +T-unmarked +PAST (-INTRA) items may get 'for a while' readings, e.g., dop ja sa rabija (Jena mal) 'then I worked (a little)' (Breu 1992: 115). This is a kind of cursus-orientation that does not present the cursus from a point of view established in the middle of it. —INTRA x [—t] does not specify whether the entity denoted by a possible direct object is totally affected (finished, consumed, etc.) or not. Undivided reference to the entity is neither signalled nor excluded. Without a limiting object, the actional phrase remains nontransformative and cannot be +T-marked. In the Bulgarian sentence Cetoxme knigi ot dva do cetiri casa 'We read books from two to four o'clock' (Lindstedt 1985: 149), the [—t] actional content of ceta remains -f-T-unmarked, since it is not limited to [+t] either by the object or by the adverbial specifying the temporal extension of the event. The event is characterized as having occurred in an integral and not goaldirected way, which excludes +T-marking. There is no reason for "perfective" interpretations of such cases. The —INTRA type of PFV, e.g., Modern Greek +PAST (-INTRA) éghrapse ghrdmata 'wrote letters', is thus rather similar to the -AD type of IPFV, e.g., Russian +PAST (-AD) pisal pis'ma. This is quite natural, since both represent an unmarked way of presenting a [—t] event, which is not the case with the + AD type of PFV. +PAST (—INTRA) x [—t] is naturally compatible with 'for X time' expressions of temporal extension, specifying the length of the portion of action, e.g., Maltese raqad il-gurnata koll-ha 'slept the whole day', Kalmyk xoyr casas avn tavn cas kürtl gazet umsv 'read newspapers from two to five o'clock'. Note that the actional content is not limited by such expressions but remains nontransformative. A wellknown example is Classical Greek ebasileuse triâkonta été 'reigned for thirty years', French régna trente ans, Bulgarian caruva trijset godini, Turkish otuz yd hüküm sürdü, etc. With durational 'for X time' adverbials, +T-marking languages such as Bulgarian and Georgian typically dispense with +T-marking and use unmarked, unqualified +PAST (-INTRA) x [-t] items. +PAST (—INTRA) x [—t] may represent a nonintraterminally viewed atelic global event consisting of an unlimited or limited series of atelic or telic subevents. Readings with telic subevents presuppose recategorization [+t] + [+ser] > [-t], e.g., Bulgarian cesto se napiva 'often got drunk' (Ivancev 1971: 133-134), na dva päti minava 'passed twice' (Maslov 1981: 250). The reason for choosing +PAST (—INTRA) is certainly not that the series was discontinued afterwards, which is equally possible with +PAST (+INTRA), but rather that the global event is not viewed from inside. As was stressed above, not even habitually repeated events need to be

Viewpoint operators in European languages

159

presented from an internal point of view but can also be envisaged in an integral way. -INTRA x [-t] naturally excludes +T-marking. As for actions that may be conceived of in different ways, +T-unmarked verbs are used to distinguish atelic conceptualizations from goaldirected ones, e.g., Lithuanian uses rasé 'wrote', Bulgarian jado 'ate'. Lack of +T-marking only concerns the actional content. It may also mean that transformativity is not an essential or fertile notion with a given actional content and context. Bulgarian plain Aorists such as spa 'slept' are simply nonintraterminals suggesting, as opposed to Imperfects, an integral view of the event. Verbs that refuse +T-markers have contents that are less fertile with transformativity. An item such as objadva 'had dinner' simply lacks a +T-marker signalling transformativity, while its aspectual meaning arises from the —INTRA value. Telicity of the event thus does not necessarily require +T-marking of the actional phrase that denotes it. +T-unmarked +PAST (-INTRA) items are ideal for expressing general-factual meanings without stressing goal-directedness and accomplishment, for signalling mere occurrence or occupation with the action ('has been / was engaged in V-ing'). +T-marking may be dispensed with in constative, experiential, characterizing, type-referring utterances, in contexts where notions of completion and situation change are not essential ('has [some time in the past] V-ed'). +PAST (-INTRA) without +T-marking in a case such as Bulgarian Dnes uciteljat me nakazva 'Today the teacher punished me' (Stankov 1980: 97) implies that the event is referred to in an integral, generalized way and that no crucial limit needs to be pointed out in the given context. +T-unmarked +PAST (-INTRA) items may also stand for telic events with subsequently cancelled results, e.g., Koj e otvarjal cantata mi? 'Who opened my bag [which is no longer open]?' (Stankov 1976: 14-15). There is no need to stress the attainment of a transformation the result of which has been reversed. Note that +PAST (-AD) items, which do not signal any crucial limit either, may be used to refer to telic events in similar ways. The situation may be similar in languages lacking systematic ±INTRA and ±AD distinctions. In Hungarian, an accomplished telic event may well be referred to without +T-marking, e.g., tegnap irta a levelet 'wrote the letter yesterday'. The +T-marked version megirta a levelet 'wrote (and finished) the letter' adds the goal-directedness of the accomplished action. +Tunmarked items are allowed, if the notion of goal-directedness is unessential.

10.3. Realizations of

±POST

Next, the interactions between +POST items and inherent phase structure types will be commented on. Their main discourse-pragmatic functions in narratives are ruptoon and retrospection for purposes of temporal or explicatory backgrounding. It

160

Lars Johanson

should be kept in mind that viewpoint operators are not defined as devices selecting phases of the lexical meaning of actional phrases. With finitransformatives, the phase highlighted by +POST is never part of the lexical meaning. Thus, 'have died' is not part of die. With initiotransformatives, however, the posttransformative viewpoint may well be situated in the statal phase of the lexical meaning. Thus, 'having sat down' is part of the content of Turkish otur- 'sit down + sit'. Most of the cases discussed below are realizations of high-focal postterminality with certain phase stracture types. Some of them will be compared to corresponding combinations with high-focal intraterminality. The different values and their realizations are interrelated and understandable within the respective systems, but they cannot easily be explained by means of fixed functional stations based on situation types. If such analyses lead to the result that closely related realizations turn up at different stations such as 'resultative' and 'progressive', their interconnections become invisible. —POST items will be largely disregarded, as they combine freely with different actional phrases and the resulting combinations do not exhibit any striking peculiarities. Note, however, that +PAST (-POST) items {French passé simple, Italian passato remoto, etc.), which are often taken to be PFV items, readily combine with [—t] actional phrases, whereas this is excluded with PFV items of the adterminal kind.

10.3.1.

+POST x [+t]

Postterminality is particularly suited for transformatives. Thus, Armenian, -bPOST^ x [+t] items typically combine with [+t] verbs such as halvel 'thaw', barkanal 'get angry', as well as with expanded [+t] actional phrases containing basic [—t] verbs as anel 'do', grel 'write'. -fT-marking languages may use -fT-marked items. However, as we have noted, +T-marking may be more or less important with different aspectotemporal items. Marking is often less relevant for highly focal postterminals that pay no or little attention to the action behind the postterminal state. The Classical Greek Perfect could easily dispense with +T-markers void of lexical meaning. Whereas the Aorist apéthane 'died (off)' from (apo)thnêisk- 'die (off)' was +T-marked, the Perfect might lack -t-T-marking: téthnëke 'has died, is dead'. Under the influence of other forms, however, +T-markers were often attached to the Perfect forms as well. Similarly, the Georgian passive high-focal postterminal çeria 'it is written remains unmarked, whereas the more event-oriented Perfect almost always carries a +T-marking preverb. As we have noted above, when low-focals lose their focality, the development typically begins with [+t] actional phrases.

Viewpoint operators in European languages 10.3.1.1.

161

+POST x [+tf]

+POST x [+tf] produces a perspective post terminum finalem. Since +POST items can denote an event whose relevant limit is situated earlier than that of the preceding item, the combination with [-t-tf] may reverse the linear order of events. If the action is conceptualized as having an inner goal, completive readings are produced. The inner goal may be signalled by +T-marking, e.g., Bulgarian perfective Perfect —PAST (+POST) x [+tf] napisal e 'has written'. +PAST (+POST) x [+t) suggests that the event has effects relevant to a past O 2 . High-focals refer to the posttransformative state, e.g., Classical Greek kéklëtai '[has been named and] is called', gégraptai 'is [already] written', egégrapto 'was [already] written', kéktëtai 'has acquired' > 'possesses', Maltese jismu 'is named', fieq 'is healed', East Armenian lavacac ë 'is healed', Karachai ketibdi 'has left (and is still gone)', Romany his anphandlo 'was tied'. Low-focals may, even when they denote telic events, lack +T-marking in contexts where the constative, summarizing, type-referring statement of the occurrence of the event itself is more essential than the notion of completion, e.g., Bulgarian vizdal e 'has (at least once) seen it'. This is typical of postterminally characterized events not conceived of as occurring in particular settings. Compare the analogous use of +PAST (-AD) items instead of +PAST (+AD) ones and of +T-unmarked +PAST (-INTRA) items instead of +T-marked ones.

10.3.1.2.

+POST x [+ti]

Combinations of postterminality with initiotransformatives produce a perspective post terminum initialem. A [+fi] actional phrase denotes both a transformation and a following posttransformational phase. The +POST x [+ti] combination is often misunderstood as an interaction of PFV with "stative" verbs. However, [+ti] verbs are neither stative nor inchoative. The combination is only intelligible as the effect of a postterminal view of an actional content that combines a dynamic transformative phase with a non-dynamic posttransformational one. The latter may thus be objectively identical to the phase highlighted by a corresponding intraterminal. The relation is particularly clear with high-focals. -f-POST™3 x [+ti] expresses a state that has come about through an initial transformation. This state may correspond to the cursus of the same actional phrase, the state highlighted °У a +INTRA x [+ti] item. A statement meaning 'is in the state of having sat down, !s seated' may refer to the same objective situation as a statement meaning 'is in the state of sitting'. Similar equations are 'has learnt' = 'knows', 'has got fond of = hkes', 'has stood up' = 'is standing', 'had lain down' = 'was lying'. Some Males e initiotransformatives have special high-focal participial items that focus on the Posttransformational state obtaining at O, thus covering situations that might alter­ natively be expressed by high-focal +INTRA items, e.g., rieqed 'has fatten asleep'

162

Lars Johanson

о; t < :щи

= 'is asleep'. English examples are has hidden = is hiding, has leaned = is leaning, has understood = understands. Verbs occuring in +POST x [+ti] combinations often express both the acquisition and the possession of positions and properties, e.g., 'get, have', 'become similar, resemble', 'call to mind, remember', 'catch sight, look', 'come to believe, believe', 'get accustomed, be accustomed', 'fall asleep, be asleep', 'get filled, be full', 'grasp, hold', 'learn, know', 'get hope, hope'. As we have noted, [4-ti] verbs are frequent in some European languages and rare in others. Classical Greek Perfects representing a posttransformational state are, e.g., dédorke 'looks, is looking', gégëthe 'is glad', éoike 'resembles'. They often denote the same phase as the corresponding Present items, e.g., dérketai 'looks', gëtheï 'is glad', or lack a corresponding Present (éoike). Similar examples are Armenian nstac ë 'has sat down' = 'is sitting' from nstel 'sit down, sit', barkacac ë 'has got angry' = 'is angry' from barkanal 'get, be angry', kangnac ë 'has stood up' = 'is standing' from kangnel 'stand up, stand', Archi osdili i 'is standing' from ocis 'stand still' and axuli i 'is lying' from axas 'lie down' (Kibrik 1983: 113), Albanian ka qëndruar 'has stopped' = 'is standing' (= qëndron), Talysh h'itä be 'had fallen asleep = was sleeping' (Miller 1953: 173) from hue 'fall asleep, sleep', riista be 'has sat down = was sitting' from hüte 'sit down, sit'. A Turkish example is oturmustur 'has sat down' - oturuyor 'is sitting' from otur'sit down, sit'. Among other Turkish verbs are dur- 'stop, stand', inan- 'come to believe, believe', yat- 'lie down, lie', uyu- 'fall asleep, sleep'. Corresponding Kalmyk verbs are kevt- 'lie down, lie', med- 'get to know, know', su- 'sit down, sit', unt- 'fall asleep, sleep', zogs- 'stop, stand', e.g., unt-sn 'is asleep', untsn bilä 'was asleep', suusn 'is sitting', suusn bilä 'was sitting'. Compare Maltese so-called Imperfects with certain [+ti] verbs, e.g.Jaf 'knows', jixbah 'resembles' (Borg 1981: 157) and Romany items such as hi besdo 'is sitting', hi tardo 'is standing', hi tsido 'is lying'. High-focal "present state" uses of +POST x [+ti] may, as is well known, develop into Present items. Germanic PreteritePresents, formed from originally strong verbs, have Present forms going back to + P 0 S T H F x |-_|_ t j^ e g ^ Gothic wait 'knows', Old Icelandic veit 'knows', a 'possesses', тип 'remembers', skal 'shall', English can, Norwegian vet (cf. Classical Greek oïde 'knows'). The effect of high-focal postterminality on [+ti] verbs as well as the close connection with corresponding intraterminals remains unintelligible as long as the analysis is based on translations into languages lacking [+ti] verbs. -PAST (+POST) and +PAST (+POST) items combined with [+ti] are often taken to be non-pasts and simple pasts respectively, whereas the corresponding combinations with [+tij are regarded as past and past-before-past items respectively. The Kalmyk postterminal item in -/ is sometimes claimed to express both past and present events ("Geschehenes", "Geschehendes") without clear rules. This variation is, however,

Viewpoint operators in European languages

163

regular and due to the inherent phase structure. Finitransformatives such as omsj or irj are interpreted as 'has / had dressed' = 'is / was dressed' (compare Turkish giyinmistir I giyinmisti) or 'has / had come' = 'is / was (t)here' (compare Turkish gelmistir I gelmisti). On the other hand, initiotransformatives such as suj are inter­ preted as 'has / had sat down' = 'is / was sitting' (compare Turkish oturmugtur I oturmustu). In the same way, the Kalmyk item in -ad band is sometimes thought to have both PF and progressive meaning ('has done', 'is doing'), depending on the verb type, but it always expresses a dwelling in the postterminal state, the only variable being the phase structure. Similarly, the Maltese postterminal active participle expresses a posttransformational state with certain verbs. With [+ti] verbs such as raqad 'fall asleep, sleep' and rikeb 'mount, ride', we get -PAST (+POSTHF) x [+ti] realiza­ tions, e.g. rieqed 'is sleeping' < 'has fallen asleep', riekeb 'is on horseback' < 'has mounted'. The same situation applies to motion verbs meaning 'get moving, move on' such as mexa 'set out, walk'. Thus, miexi 'is walking' represents a +POSTHF aspect of a [+ti] actional content and refers to a posttransformational phase of this content, obtaining as a state at 0 s . It may be tempting to claim that the participle encodes 'progressivity' in such cases. However, although progressive interpretations are suggested by the English translations ('dressed' > 'wearing', etc.), this is not a progressive item in the sense of a high-focal intraterminal. -I-POST*111 items differ from -(-INTRA101 items by lacking internal dynamicity and not being capable of fu­ ture time reference. +POST x [+ti] combinations competing with +INTRA x [+ti] ones lack the graduality necessary for implying internal dynamicity. For example, in Armenian, only the latter type combines with expressions such as aveli и aveli 'more and more', kamac-kamaç 'little by little'. 10.3.2. +POST x [-t] When a +POST operator applies to a [—t] actional content, the initial limit is the relevant one. With the combination -PAST (+POST) x [-t], the event has at least begun prior to Os and may also overlap it. +PAST (+POST) x [-t] suggests a past О . If an actional phrase such as read a book is interpreted as [—t], lacking an inner goal of the action, its combination with -PAST (+POST) implies has read (some Pages of) a book. +T-marking languages dispense with +T-marking in such cases. Bulgarian thus uses its imperfective Perfect for -PAST (+POST) x [-t]. But there are also certain constraints on the combination +POST x [—t]. The ban °n high- focal postterminals with nontransformatives has been mentioned already. There is little need for representing a state post terminum if a crucial limit is lacking. In a few languages, however, items known as high-focal postterminals can also be u sed with [—t] verbs, e.g., Kalmyk kelsn bilä 'was speaking'. The relevant limit is, as e *pected, the initium. The nontransformatives are thus treated very much like initio-

164

Lars JohansoB

•»' я > ; • ;;t; fc

transformatives. Similar situations are found in several Eurasian languages outside the boundaries of Europe. Certain traces in Indo-European languages are also reminiscent of this phenom­ enon. Here, some postterminal participles may get different temporal readings ac­ cording to their internal phase structure. Passive postterminal participles may get nonanterior readings with [—t] verbs, e.g., German ist geliebt 'is loved', but anterior readings with [+t] verbs, e.g., ist gefanden 'has been found'. If we use Esperanto as a metalanguage, this gives the following German-Esperanto equivalences: ist ge­ fanden = estas trovita 'has been found', war geschrieben - estis trovita 'had been found', ist geliebt = estas amata 'is loved', war geliebt - estis amata 'was loved'. The participles of [—t] verbs are thus clearly postterminal in the sense of the initium being the relevant limit. (Note, however, that languages such as English and French have developed passives with which [+t] verbs also get nonanterior readings, e.g., is found, est trouvé.) The applicability of +POST to [—t] increases with lower focality degrees, but there are often restrictions on combinations with [—t, —dyn] actional phrases. This is, for example, true of the Modern Greek possessive Perfect. +POST items are also excluded with Archi verbs such as hubus 'blow', qebus 'dance', and arhas 'think' (Kibrik 1983: 113). As was noted above, when low-focals lose their focality, [—t] actional phrases are usually the last to be affected. Thus, the Portuguese and Galician Perfects are largely restricted to [—t], with constative uses summarizing the event at O s . These are clear examples of an interaction with an actional value. As we have seen, [—t] combinations often produce constative ("experiential") readings of lowfocals. Finally, defocalized -PAST (+POST° ( -INTRA)) items such as the Maltese Perfect are, of course, fully combinable with [—t] actional phrases. They are often classified as PFV items, though their properties are very different from those of PAST (+AD) items.

10.4. Realizations of ±AD There is a good deal of interdependence between ±AD and [±t] actionality. As we will see, there is no freedom of choice if the global event is referred to with a [+tf, + m o m ] or a [—t] actional phrase. The former case requires +AD, the latter case —AD. Adverbials expressing temporal punctuality such as Russian vdrug 'suddenly naturally require + AD. Incompatibility with such adverbials is, however, only typical of the - A D variety of IPFV, not of the +INTRA variety (Section 7.2.4). 10.4.1.

+AD x [+t]

Adterminality suggests strong limit-orientation, envisaging the event in its attainment of the crucial limit. It thus presupposes [+t].

Viewpoint operators in European languages

165

+AD is particularly suited to operate on finitransformatives: +AD x [+tf, —mom] gives strong prominence to the finis and may also yield readings that imply an exit from the event. The actional phrase denotes an action heading for completion, and the viewpoint operator signals the attainment of it. The combination does not favour [+ser] readings. An actional content referring to the lowest quantity of a telic event requires -I-AD, e.g., Polish umarl 'died'. Russian perfectives such asposlat' 'send', razbudit' 'waken' or vstaf 'get up' express a total event and its conclusion. The +AD meaning requires that the actional content of napisal pis 'ma be interpreted as [+tf] including a limiting object, a set of letters referred to as a whole: 'wrote (and finished) (the) letters' (undivided reference to the set). As we have noted, [+tf] does not combine with 'for X time' adverbials. Thus, +AD x [+tf] napisalpis'mo 'wrote the (whole) letter' cannot be modified by adverbials such as dolgo 'for a long time' or dva casa 'for two hours'. Since [+tf, +mom] actional contents lack a salient cursus, +AD x [+t, +mom] envisages nothing more than the crucial limit itself in the moment of transformation. There is no —AD option. +AD x [4-ti] gives strong prominence to the initium, signalling the attainment of the initial crucial limit. It may thus suggest the entry into the event, e.g., Russian opersja na stol 'leant over the table'. For the realizations mentioned, terms such as "resultative", "egressive", "punctual", "ingressive", and "inceptive" should be avoided, since they easily lead to confusion with modes of action (4.2). 10.4.2. *+AD x [-t] Adterminality does not operate on nontransformatives: *+AD x [—t]. This is an intelligible constraint on the freedom of aspect choice, since there is no use for items signalling the attainment of a limit not present in the actional content. Thus, [—t, —dyn] actional phrases only combine with —AD. Russian [—t, —dyn] verbs, whose lexical meanings do not imply clearly delimited extensions, are imperfectiva tantum, for example, prinadlezat' 'belong', sostojat' 'consist', mat' 'know'. However, this is no reason to take [—t, —dyn] actional phrases to be "monoaspectual" in principle, in the sense of IPFV being the only option. It has already been noted that -INTRA x [-t, -dyn] is a possible combination. The situation may seem different with [—t, +dyn] actional phrases. Russian perfectives can express past events such as prozil tri goda v Moskve 'spent three years m Moscow'. The possibility of choosing prozil here instead of zil is often regarded as evidence for mutual independence of aspect and actionality. However, this option requires a delimitative or perdurative item that provides the actional content with a crucial limit indicating the minimal- maximal extension. Perduratives such as prozil combine with mensural expressions specifying the length of the whole period of tome. Such modes of action constitute [+tf] actional phrases, which are naturally compatible with +AD. The resulting combinations sum up the event as a totality

166

Lars Johanson

reaching the final crucial limit. Adterminality is thus by no means incompatible with expressions of duration as such. However, there is no option *+AD x [—t, +dyn]. This fact sharply distinguishes +AD from PFV items of the —INTRA type, which suggest integral readings with [—t, +dyn] actional phrases. 10.4.3.

-AD x [-t]

Nonadterminality envisages the event without highlighting a crucial limit. —AD items exclude limit-orientation and suggest a cursus-orientation which, however, is much weaker than with +INTRA items. In its textual behaviour, a +PAST (-AD) item such as Russian pisal 'wrote' thus corresponds both to the intraterminals was writing, écrivait and to the nonintraterminals wrote, écrivit. Nonadterminality has a natural affinity with nontransformativity. If the global event referred to is expressed with a [—t] actional phrase, —AD is normally required. Applied to [—t, —dyn] phrases, —AD presents the event as a static stretch of action without any salient natural limits, e.g., Russian visel 'hung'. The combination is typically expressed by imperfectiva tantum. This does not mean that static nontransformatives are "aspectually indifferent" in general, as claimed by Bache (1982: 69). The claim that they are only referred to by IPFV items is not valid for ±INTRA languages, which use them freely with —INTRA items. Combinations of —AD with [—t, +dyn] actional phrases present the event as a stretch of action, without suggesting totality or attainment of any limit. On the other hand, they do not yield the 'progressive' readings possible with IPFV items of the +INTRA type. In narrative discourse, -AD x [-t, -fdyn] does not suggest any 'entry' or 'exit', but merely 'occurrence (for a period of time)'. It often simply denotes the lexical content, notably a content with which + AD is rather infertile. Since +AD implies the involvement of a crucial limit, it is natural to interpret —AD items as implying no such limit. The actional content of Russian pisal ('wrote') pis'mo 'letter' is thus prima facie interpreted as [—t], without a limiting object referred to as an undivided entity. But —AD is also used where the mere occurrence of the event, not its full performance or completion, is important. The event is referred to in a generalized manner without highlighting any phase. Such readings often occur in contexts where the event is conceived of as taking place outside a particular setting. Such an unqualified, "characterizing" manner of presentation is what might be expected from the unmarked member of a ±AD opposition. In similar cases, ±INTRA languages use unqualified -INTRA pasts, and languages with +T-marking dispense with this device. Compare, for example, Russian +PAST (-AD) x [-t, +dyn] Ту pisal ej? with Turkish (-INTRA) x [-t, +dyn] Ona yazdin mi? 'Have you written to her?'As t+t] is incompatible with 'for X time' adverbials, any actional phrase is iden­ tified as [—t] by occurring with them. Russian —AD x [—t] zil tri goda v Moskve 'lived for three years in Moscow' has a 'for X time' adverbial but no limiting object.

Viewpoint operators in European languages

lof

The event is characterized as simply having occurred for a certain period of time. In citai knigi ot dvux do cetyrex casov 'read books from two to four o'clock' (cf. Thelin 1978: 33), the action denoted by the [-t] verb citat' is not limited to [+tf] by the object. The interactional meaning +PAST (—AD) x [—t, +dyn] is 'a stretch of action took place' without an idea of a crucial limit implied. The temporal extension is specified by the 'for X time' adverbial ot dvux do cetyrex casov 'from two to four o'clock'. With [—t, +dyn] verbs, both +AD and —AD are possible and their viewpoint values realized in a particularly clear way. If both +AD x [-t, +dyn] and -AD x [—t, +dyn] can be used to refer to a given situation, e.g., skazal and govoril 'said', the difference only lies in the view presented. It would be absurd to claim that the opposition is "neutralized" here or that —AD is used "instead of" +AD. Since -AD does not highlight any limit, a - AD x [-t, +dyn] case such as Russian pisal pis'mo may certainly be used when the speaker is concerned with the internal structure of the event. It may textually even correspond to English 'was writing a / the letter' or German 'schrieb an einem / an dem Brief. But it does not present the event as a continuous, unfolding process of gradual achievement of the result (Maslov 1948: 313). —AD items do not, as +INTRA items do, describe the event with specific attention to its internal structure. Their way of presenting events is not particularly "unfolding". On the contrary, it is essential for their use that they do not suggest any intraterminal meaning. Because the viewpoint is not explicitly intra termines, —AD readily combines with 'for X time' statements, e.g., pisal dva casa 'wrote for two hours', carstvoval tridcat' let 'ruled for thirty years'. This is unusual with +INTRA items, since the idea of introspection is incompatible with the indication of the total temporal extension. When —AD operates on [+ser] actional phrases denoting pluri-occasionality, the reading [+ser] may be overtly marked or not. The global event is presented by -AD regardless of the actional properties of the subevents. If the latter are telic and basically represented by [+t] actional phrases, we get the combination —AD x [+t] + [+ser] > [—t], e.g., Russian citaipo knige v den' 'read a book a day', procityval knigu 'repeatedly read through a / the book'. The subevents may also be atelic, e.g., Russian +PAST (-AD) x [-t] + [+ser] > [-t] citai ètu knigu kazdyj den' 'read in this book every day' orpocityval knigu 'repeatedly read a little in a / the book'. 10.4.4.

-AD x t+t]

While naturally operating on nontransformatives, -AD can also apply to transfurinatives. The reason is that it expresses a perspective that neither considers nor negates the limits of the event. It does not signal non-completion, but may suggest this as a negative value. Even so, the combination -AD x [+fj is perfectly logical, since heading for a crucial limit is not equal to reaching it.

168

Lars Johanson

Whereas the Russian-type perfectives are portmanteau -I-AD- and +T-markers, the corresponding imperfectives are neither +AD- nor +T-markers and thus less distinct. Besides standing for —AD x [—t], they may also express a nonadterminal view of a non-momentaneous telic event. Our model includes the combination —AD x [-И], where the actional phrase is left +T-unmarked, though it denotes a telic event. This means that the combination +PAST (-AD) x [+t] may also be used for fully ac­ complished events. It neither highlights the transformation itself, nor excludes the unabridged achievement of the event. For example, Russian pisal pis'mo 'write a / the letter' may not only be interpreted as + PAST (-AD) x [—t], a process excluding the attainment of an inner goal, but also as +PAST (—AD) x [+t], an event including the total involvement of a limiting object. In ±INTRA languages, the latter reading is typical of +PAST (-INTRA) items. Though +AD is often claimed to be obligatory for telic events, the combination —AD x [+t, —mom] is perfectly possible. With heterogeneous actional contents that imply both a statal and a dynamic phase, the Russian-type imperfectives get statal interpretations. With initiotransformatives, perfectives highlight the initium, e.g., opersja 'leant', whereas imperfectives may give a similar prominence to the resulting state, e.g., opiralsja 'leant, was leaning'. These cases represent a contrast +AD x [+ti] vs. -AD x [-Hi]. As for finitransformatives, +AD highlights the finis, whereas —AD x [+tf] may, unless the event is conceived of as momentaneous, give a certain prominence to the preceding state. -AD does not combine with [+tf, +mom] actional contents, as they only consist of the crucial limit that is to be attained. If —AD is applied to verbs that are basi­ cally [+tf, +mom], the nontransformativizing feature [+ser] is added. The typical interactional meaning 'repetition' is an effect of this recategorization [+tf] + [+ser] = [ - t ] . Russian imperfectives such as naxodit' 'find' and vkljucat' 'switch on' im­ ply repetition if the basic event is conceived of as momentaneous. Thus, the aspect distinction has important actional implications with [+tf, +mom] actional phrases. Several Russian [+tf, +mom] verbs are unpaired perfectives and offer no —AD op­ tion, e.g., zaplakat' 'start to cry'. Russian +PAST (—AD) items can be used to refer to telic events the result of which has subsequently been cancelled. Thus, otkryval dver' '(had) opened the door' may be interpreted as a so-called 'twoway action' with the implication that 'the door was closed again'. Such readings are also possible with momentaneous telic events, which demonstrates a certain affinity with [+ser] readings, but the common feature is simply that no crucial limit is highlighted. The claim that such imperfectives are in reality perfectives (Ferrand 1982) seems to be based on the equation of 'perfective with 'telic'. Some possible cases of -AD x [+t] are problematic. It seems that -AD x [+tf> —mom] may get processual and preliminary readings suggesting a process that pre­ cedes the transformation. These readings are typical of Russian imperfectives such as brat' 'take', davat' 'give', padat' 'fall', posylat' 'send'. Such verbs might be

Viewpoint operators in European languages

169

claimed to differ from their perfective counterparts, vzjat' 'take' etc., by being nontransformative. However, they express a nonadterminal view of telic events, and their actional content might be taken to be transformative as well, the only difference from the perfectives residing in the viewpoint. Forms such as bral 'took' would thus not only represent —AD x [—t], but also the case —AD x [+t], though without +T-marking. By contrast with their perfective counterparts, vzjal etc., forms such as bral 'took' get preliminary readings that do not include the attainment of the point of transformation of 'taking', 'giving', 'falling', 'sending', etc., but only a stretch of action heading for the transformation. They may even express conation, the attempt to perform the telic action and to reach the crucial limit, e.g., ubezdal 'tried to convince'; cf. Polish doganial go, ale nie dogonil 'tried to pass him, but did not succeed' (Koschmieder 1934). These imperfecfives do not express "nonevents", but just events excluding the crucial limit from the range of view. However, they are unusual both in meaning and shape. Their behaviour is similar to +INTRA items, and they differ from their perfective counterparts by stem-internal markers rather than by the absence of preverbs. It is possible that they should rightly be considered marked members of limited ±INTRA oppositions within an otherwise ±AD-dominated system. In that case, umiral would be a +PAST (+INTRA) x [+tf] item meaning 'was in the process of dying' as opposed to +PAST (—INTRA) x [+tf] item umer 'died (ceased to live)'.

11. Main aspectotemporal types 11.1. Types discussed Our discussions have so far focused on items marked for -HNTRA, +POST and +AD and their respective opposition partners. If we disregard differences with respect to high and low focality, the above analyses yield a rather limited number of aspectotemporal types, which are listed here together with more traditional - and partly misleading - designations: -PAST (+INTRA°) Present

-PAST (+POST) Perfect

-PAST (-AD) Imperfective Present

+PAST (+INTRA) Imperfect

+PAST (-POST) Preterite

-PAST (+AD) Perfective Present

+PAST (-INTRA) Aorist

+PAST (+POST°) Pluperfect

+PAST (-AD) Imperfective Past +PAST (+AD) Perfective Past

170

Lars Johanson

«»к

This list does not exhaust the number of main aspectotemporal types to be dis­ cerned in European languages. Some more complex ones, combining positive and negative viewpoint values, have already been touched upon above in a general way. The major additional types will be listed below.

11.2.

+PAST (-POST (+INTRA))

Besides the intraterminal type +PAST (+INTRA), there is a -f-PAST (-POST (+INTRA)) type that also takes part in a postterminal opposition, e.g., English was writing, opposed to +PAST (-POST (-INTRA)) wrote and -PAST (+POST (-INTRA)) has written. +PAST (+INTRA) items are semantically more general, since they are also used for past events that a ±POST language would express postterminally.

11.3.

-PAST (+POST (-INTRA))

Besides the postterminal type -PAST (+POST), there is a -PAST (+POST (-INTRA)) type that also takes part in an intraterminal opposition, e.g., English has written, Armenian grel ё, High (Hill) Man siren. -PAST (+POST) items are se­ mantically more general, being used for past events that a ±INTRA language would express intraterminally. Some -PAST (+POST (-INTRA)) items are semantically more general than others, for example, Estonian on teinud 'has done', since the competing +PAST (-POST (+INTRA)) item is less strongly grammaticalized in the sense of having a rather limited use, e.g., Estonian oli tegemas 'was doing'. Some -PAST (+POST (-INTRA)) items have a relatively restricted use themselves, e.g., Irish ta tar éis leabhar a scriobh 'has (just) written a book' as opposed to the +PAST (—POST (—INTRA)) item scriobh se leabhar 'wrote / has written a book'.

11.4.

-PAST (+POST (+INTRA))

In addition to +PAST (+INTRA) and +PAST (-POST (+INTRA)), there is a third intraterminal type, -PAST (+POST (+INTRA)), taking part in both a postterminal and an intraterminal opposition. Thus, the English so-called Continuative Perfect has been writing is opposed to -PAST (+POST (-INTRA)) has written and +PAST (-POST (+INTRA)) was writing. The complex construction serves to express a particular kind of interaction in which a +POST notion operates upon a +INTRA notion to express the Os relevance of an intraterminally viewed event. The English type combines a progressive periphrasis with postterminal morphology, using a ±POS distinction of the 'be' auxiliary: -PAST (+POST) has been vs. -t-PAST (-POST) was.

Viewpoint operators in European languages

171

Note that this type, although taken to be a "y" item by Kurylowicz (1956: 26), is by no means a high-focal in the sense of —PAST (+POSTHF). Thus, has been dying is not equal to Classical Greek téthnëke 'is dead'. -PAST (+POST (+INTRA)) is a special kind of postterminal, signalling that the point transgressed at Os is the O2 of the intraterminal perspective. The event is viewed diagnostically and as having continuing relevance to the primary deictic centre Os. The intraterminal view highlights an internal portion of the event at an O2 consisting of one or more intervals of unspecified localization. O 2 , the vantage interval of +INTRA, is situated within the limits of the event, and O s , the vantage point of +POST, after the beginning of this interval. Since the localization is unspecified, this item, like other focal postterminals, does not readily combine with specific time expressions. The interpretation of the postterminal element as PFV and of the intraterminal element as IPFV can only lead to the conclusion that -PAST (+POST (+INTRA)) items are aspectually self-contradictory or at least ambivalent. The kind of retrospective intraterminality conveyed here suggests that the event, which has begun at some time in the past, continues right up to Os and that the event itself, its effects, or its concomitants overlap this point. A sentence such as has been examining the case easily suggests, since it only refers to the involvement, that, despite all efforts in the past, the final limit of the examination has not yet been reached at Os. But has examined the case may also imply Os-inclusion, since a +POST operator only requires the initial limit to be transgressed unless the actional phrase is unequivocally finitransformative. However, -PAST (+POST (+INTRA)) items tend more strongly towards incomplete readings. The choice of an item that only highlights internal parts of a given event may, though the main function is mere introspection, get the pragmatic implicature that only part of the event has taken place, so that it is still unfinished at Os. Similar contrasts are found in Irish. A -PAST (+POST (-INTRA)) construction such as in ta tar éis leabhar a scriobh 'has (just) written a book' presents the whole event without isolating or highighting any part of it. The choice of -PAST (+POST (+INTRA)) ta tar éis a bheith ag scriobh leabhair 'has been writing a book' adds an extra dimension and focuses on the intervening periods of writing (Ö Baoill 1994: 209). While -PAST (+POST (+INTRA)) items operate on all kinds of actional phrases, nontransformatives and initiotransformatives are preferred. Restriction to the latter Would mean consistent Os-overlapping and the possiblity of describing Os as the Mentation point for both the +POST and the +INTRA view.

172

11.5.

Lars Johanson

.,v>.-,•,, ,f,;

Pluperfects

Besides simple +PAST (+POST0) items, e.g., German hatte geschrieben 'had written', postterminal items involved in an intraterminal opposition may occur in the pre-past stratum, e.g., English +PAST (+POST0 (+INTRA)) had been writing and +PAST (+POST° (-INTRA)) had written. Distinctions between non-dynamic and dynamic items in the pre-past stratum have been mentioned above (8.3).

11.6.

Historical pasts

European languages exhibit different types of historical pasts, event-oriented, propulsive preterites that may refer to events conceived of as occurring on specific past occasions. Some of them are commonly considered PFV items, whereas others are called "simple Pasts". However, they are of rather different kinds. Most items are complex ones, deriving their values from several oppositions. Only one type is a general past in the sense of being applicable to any past event. -(-AD, —INTRA, and —POST are obviously appropriate values for constituting historical pasts. +AD makes an item propulsive by highlighting the crucial limit and thus changing the situation. —POST makes an item particularly event-oriented and suited to refer to events conceived of as occurring on specific past occasions, in particular settings, pinpointed at specific intervals. —INTRA makes it particularly suited to present the event in an integral way. An item interacting with both a +PAST (+INTRA) and a -PAST (+POST) item is thus ideally suited to envisage an event directly and in its totality. Qualified historical items of this type often cooccur with definite past time adverbials. They are typically incompatible with markers of indefinite time and of unrestricted or very long periods, as well as with adverbials of the types 'already' ('the relevant limit is not later than O') and 'not yet' ('the relevant limit is later than O'). However, clear-cut +AD, -INTRA or -POST values are not required for the historical functions. Many items are suited as narrative pasts, for narrating sequences of discrete past events under the general conditions of linear successivity, without being "perfectives" in any qualified sense. The mere absence of intraterminality allows presentation of the event in an integral way. Several events presented one after another in a nonintraterminal past tense are most naturally interpreted as a sequence in the sense of 'did x, and [then] did y'. Nonfocal postterminals and other generalized items are potentially event-oriented and applicable to specific past occasionsNote that a past that does not compete with a postterminal can be both diagnostic and historical, covering all the focality degrees of a postterminal. On the other hand, a past that does not compete with an intraterminal can cover the focality degrees of an intraterminal. Thus, +PAST (-POST) Danish skrev, +PAST (-AD) Russian

Viewpoint operators in European languages

173

pisal, and Hungarian +PAST irt may all stand for a limited or extended "nunc" ('was writing, used to write, wrote'). The sources of nonpostterminals, nonintraterminals, and more general items (without any ±INTRA or ±POST commitment) will not be discussed here in detail. Many items have long developments behind them, being expressed by bound elements with reduced shapes typical of old forms representing late stages of grammaticalization. In some cases, the sources are unidentifiable. Indo-European past items are marked by an augment, originally an independent word (< adverb), proclitically added to the verbal form. Turkic and Mongolian historical items in -di and in -ba[i\ are old forms with highly eroded material shapes that are difficult to etymologize. If the item in -di goes back to a verbal noun provided with a possessive suffix, it might ultimately be traced back to a possessive high-focal postterminal. Similar possessive pasts are found in Finno-Ugrian and Tungus languages (Benzing 1988: 48-49). But defocalization as such does not naturally lead to the creation of -POST items. There is no evidence that European +PAST (-POST) items that do not interact with +INTRA pasts have developed from postterminals (cf. the discussion in Bybee et al. 1994: 81-85). Nonfocal postterminals can only develop into -POST items, if new focal postterminals are created to interact with them. Nor is there evidence that European nonpostterminals go back to the same sources as PFV items of the adterminal kind, i.e. to complétives or constructions with auxiliaries derived from directional and movement verbs. 11.6.1.

+PAST (-INTRA)

Aorists of the +PAST (-INTRA) type, e.g., colloquial French a écrit 'wrote, has written', only interact with a +INTRA category and are indifferent to ±AD and ±POST. They may present past events both in a historical and a diagnostic way, covering the domains of English -PAST (+POST) and +POST (-PAST) items, e.g., Turkish yazdi or Maltese kiteb 'has written, wrote'. Some Pasts of this type are, however, opposed to a Remote Past, e.g., the Kabardian item in -a (vs. the Remote Past in -a-ya-). In a system involving a ±INTRA distinction, a -PAST (+POST) item may lose its focality and generalize into a +PAST (+POST0 (-INTRA)) item, taking over a function of the old superseded +PAST (-INTRA) item. The ±INTRA opposition is thus continued after a formal substitution of the unmarked member. For example, we Latin Perfect cantavit, serving to express both 'has sung' and 'sang', developed into a nonfocal postterminal and a nonintraterminal, covering the whole range of past u ses except for cases that motivate an intraterminal view (cantabat). Similarly, most Romance reflexes of an originally high-focal postterminal type habet cantatum, e.g., ^orth Italian ha cantato, have generalized to nonfocals, taking over the -INTRA v alues from old +PAST (-POST (-INTRA)) items, e.g., canto and other reflexes °r cantavit. Certain varieties, however, have preserved the passé simple, passato

,

174

Lars Johanson

remoto, perfectul simplu, etc., as the —INTRA member of the opposition. In many non-Romance languages, the ±INTRA opposition similarly consists of an Imperfect and a +PAST (+POST0 (-INTRA)) item, e.g., in Italo-Croatian, where the old Aorist has disappeared and the former Perfect has lost its focality. It has already been noted that, if a ±POST opposition is given up, it is not always the former -PAST (+POST) item that generalizes. The former +PAST (-POST) item may also develop into a more general item referring to past events both historically and diagnostically. In some South Italian dialects, the use of the former +PAST (—POST) item (Simple Past) has generalized at the expense of the former —PAST (+POST) item (Perfect), e.g., Calabrian/и 'became, has become' (also instead of è stato 'has become'; Bosco 1924: 13). While the European Spanish type -PAST (+POST) ha hablado is encroaching on the functional territory of +PAST (-POST) hablö, the opposite tendency is observed in varieties of South American Spanish, e.g., in Chilean Spanish, where -PAST (+POST) (preterite perfecto compuesto) is being replaced by +PAST (-POST) (pretérito indefinido). Similar tendencies are also met with in American English, where the Simple Past seems to be gaining ground from the Perfect. However, the Perfect is still present in these varieties. In Turkish, on the other hand, the item in -di, e.g., yazdi 'wrote', has almost generalized as a nonintraterminal past, whereas the finite item in -mis mainly fulfills indirective functions, e.g., yazmis 'has [apparently] written' and the item in -mistir is, at least in the spoken language, a presumptive, e.g., yazmistir 'has [presumably] written' (Section 8.7). A 4-PAST (-INTRA) item that does not compete with a postterminal can be both historical and diagnostic, covering all the focality degrees of a postterminal without being marked for them. Turkish items such as bitti or kalmadi, operating on transformatives, do not only mean 'ended, has ended' and 'did not remain, has not remained', respectively, but may also refer to the posttransformational state without regard to the event: bitti 'is finished, over', kalmadi 'is all gone, used up'. Similar examples are Latin +PAST (—INTRA) items ("Perfects") such as novit and cônsuëvit, which may mean 'knows' and 'is accustomed to', respectively. 11.6.2.

+PAST(-POST)

+PAST (-POST) items interact with a +POST category but are indifferent to ±AD and ±INTRA. This plain type is present in a few Germanic and Finno-Ugrian languages, e.g., Norwegian skrev 'wrote', Finnish kirjoitti 'wrote', Latvian rakstïja 'wrote', Dutch werd geschreven 'was written', Swedish skrevs 'was written'. Similar items are found in Slavic varieties that have preserved the old Aorist but lost the old Imperfect, e.g., Serbian pisa 'wrote'. One of the two Basque Past tenses, formed from synthetical verbs only, is indifferent to ±INTRA and thus belongs to this category; e.g., ba-nengoen 'I stayed'. It can always be used when the analytic3

Viewpoint operators in European languages

175

constructions would necessitate a choice between +PAST (-POST (+INTRA)) and +PAST (-POST (-INTRA)) items (Haase 1994: 284). 11.6.3.

+PAST (-POST (-INTRA))

Nonpostterminals frequently compete with a +INTRA category as well, which produces a +PAST (-POST (-INTRA)) type of nonpostterminal and nonintraterminal items marked for anteriority. Thus, English wrote is opposed to the items +PAST (-POST (+INTRA)) was writing and -PAST (+POST (-INTRA)) has written. Other examples of +PAST (-POST (-INTRA)) items are Modern Greek éghrapse 'wrote, has written', French écrivit, Portuguese morreu 'died, has died', Spanish entra 'entered' , Catalan perde 'lost', Romanian cantä 'sang', Albanian vrau 'killed', Bulgarian vleze 'entered', Sami bâradii ' ate, has eaten', High (Hill) Mari sins 'wrote', Tatar hardi 'went', Karachai aldi 'took', Kalmyk kelv 'spoke', Tati baft 'wove', Kirmanji hat 'came', Armenian grec, Lezgian kxena 'wrote'. In several languages, e.g., French, Italian, Romanian, Serbian, Albanian, the type is limited to certain varieties (8.10.2). The Basque item ("Perfective Past") is restricted to analytical verbs (Haase 1994: 282). The languages in question thus possess three items covering the simple past stratum: a +PAST (+INTRA) Imperfect, a -PAST (+POST) Perfect, and a +PAST (-POST (-INTRA)) Aorist or Simple Past. This is also a frequent pattern in IndoEuropean languages bordering on the European area, e.g., Persian +PAST (+INTRA) mïkard 'was doing, did', -PAST (+POST) karda ast 'has done' and +PAST (-POST (—INTRA)) kard 'did'. Creole systems, e.g., Romanian ones, possess non-temporalized "neutral" -POST (-INTRA) items implying anteriority with transformatives and nonanteriority with nontransformatives. 11.6.4.

+PAST

A plain, generalized type with the simple value +PAST, not interacting with any other viewpoint category and signalling anteriority only, is found in some languages such as Hungarian, South German, Yiddish, and Ossetic, e.g., South German hat geschrieben 'has written'. The Hungarian and Ossetic items are even general pasts, e g-, irt 'has written, wrote, had written', Ossetic dzirdta 'spoke, has spoken, had spoken'. These +PAST items void of ±INTRA, ±POST and ±AD values cover all kinds of Past events. They combine with transformatives to the effect that undivided reference implies completion. If the referent of a direct object is totally involved in the actional content, it is limiting, i.e. understood to be totally affected with the realnation of the event, for example, Hungarian kiolvasta a konyvet 'read the (whole) °ook' (De Groot 1984). In the diagnostic dimension, they cover different focality

176

Lars Johanson

degrees of postterminals. Items operating on transformatives may often be used in adjective-like ways, denoting the posttransformational state, e.g., Hungarian fâradt +?ASTfâradt 'has endeavoured' —» 'is tired'. Hungarian szokott 'is accustomed' is a lexicalized former high-focal postterminal, now functioning as a special expression of habituality. Considered diachronically, these pasts are +PAST (-POST0) items. It was noted above that —PAST (+POST) items may lose their focality, generalize and oust former +PAST (-POST) items. Several unqualified +PAST items have once been -PAST (-POST (-INTRA)) items, before the ±POST and the ±INTRA oppositions in question were lost. The defocalized Romanian Perfect seems to be developing this way, taking over functions of the vanishing Imperfect (Haase 1995: 142). On the other hand, plain pasts may of course develop into +PAST (-INTRA), +PAST (-POST), or +PAST (-POST (-INTRA)) items, if new ±INTRA or ±POST distinctions are created. 11.6.5.

+PAST(-AD)

Most Slavic languages have lost their ±INTRA oppositions, and their former Perfects have turned into nonfocals without competing ±INTRA or ±POST items. Some of these Perfects have served as the basis for the creation of adterminality oppositions. Russian pisal 'wrote, was writing' and Czech spal 'slept, was sleeping' are basically postterminals that have lost their focality, generalized -PAST (+POST0) = +PAST items, which have acquired the structure +PAST (—AD). They do not take part in any other viewpoint opposition. If ±AD items are the only past items in a system, they are both diagnostic and historical. Though European nonadterminals lack linguistically relevant focality degrees, they do cover the various degrees specified by postterminals. Thus, +PAST (-AD) items are often used for what is expressed by -PAST (+POSTLF) in ±POST languages. This is natural, since in both cases the very attainment of the relevant limit is unessential: the postterminal conveys the view after the limit has been reached, and the nonadterminal denotes a past event without reference to a limit. A past that does not compete with a +INTRA item can also cover all the focality degrees of an intraterminal. Thus, +PAST (—AD) Russian pisal may stand for a more or less limited or extended presentness in the past: 'was writing', 'wrote', 'used to write'. 11.6.6.

+PAST(+AD)

+PAST (+AD) items only signal adterminality and do not interact with other viewpoint categories. An item such as Polish napisal or Czech napsal 'wrote' is basically a postterminal that has lost its focality but is marked for +AD. It is both historica and diagnostic, covering all the degrees of focal postterminals. Thus, Russian + PA^

Viewpoint operators in European languages

177

(+AD) ustal does not only mean '(has / had) got tired' but may also denote the posttransformational state without referring to the event: 'is tired'.

11.7. Relations to PF, IPFV, PFV Let us finally add a few summarizing words on how the main aspectotemporal types relate to crosslinguistic types such as PF, IPFV and PFV. Even if PF may be subsumed under the +POST domain, it only represents one sector of it. Many postterminals do not fit into the PF type. On the other hand, many items that appear in typical PF uses, e.g., +PAST (-INTRA) or +PAST items, are not postterminals. Postterminality has a broader crosslinguistic occurrence and higher degrees of independence, stability and combinability. PF items tend to occur in limited context types. On the other hand, they are constantly threatened by defocalization. It is not easy to establish PF as a universal type, particularly since 'current relevance' does not seem to be a sufficient criterion for all language-specific categories claimed to belong to it. Given its vagueness, it might not have acquired the status of a cardinal crosslinguistic type, had it not been for the English Perfect, which shows a remarkable correspondence to it. PF is sometimes taken to constitute a special case of PFV to the effect that PFV signals completion of the event at Os, whereas PF adds the notion of current relevance. However, not all PFV items satiate the criterion 'completion at O s ', and the only common temporal property of PFV and PF seems to be: 'at least E* is reached at 0 s ' . At any rate, it should be clear that postterminality cannot be considered a subspecies of PFV. The relationship between the main aspectotemporal types and the IPFV vs. PFV dualism can be summarized as follows. IPFV roughly corresponds to two kinds of items with the values +INTRA and -AD, respectively. Their common property, manifested in similar textual behaviour, is a view of the event without regard to boundaries and with more attention to the internal structure. However, the two IPFV types differ from each other in essential respects. Most IPFV representatives are marked +INTRA categories; very few are unmarked —AD categories. Whereas +INTRA items view the event within the lim4s, —AD items attract attention to the internal structure by negating or ignoring the notion of an attained limit. The inclusive semantic relationship of IPFV and PROG 0l % applies to the +INTRA type. The -AD type does not include any high-focal or Progressive subcategory. PFV roughly corresponds to two kinds of items with the values +AD and -INTRA, respectively. The unifying property, manifested in similar textual behaviour, is the yiew of the event as a totality including its beginning and its end, without attention u s internal structure. Both types are holistic, capable of presenting an integral

178

Lars Johanson

view of the event, and thus propulsive. Despite the analogies, however, the two types differ from each other in many respects. +AD pasts function as PFV items by virtue of a positive value implying a transformation, —INTRA pasts by negating the introspective meaning. Despite all similarities, the general distribution of the two types of IPFV vs. PFV oppositions tends to be different (Johanson 1971: 88-100). As we have noted, -AD items have much in common with —INTRA items, both being suited for the mere statement of the occurrence of the event in a 'general factual' or 'simple denotative' sense. The conflation of several "perfective" types has a long history, beginning with Curtius' confusing equation of the Classical Greek Aorist with Slavic perfectives (1846, cf. Ruipérez 1954: 34). However, the semantic differences between the two opposition types are clearly motivated by their different values and cannot be explained as marginal Slavic idiosyncrasies (Maslov 1973: 82; Bondarko 1983: 117). They are certainly not satisfactorily accounted for by reference to the different morphological (inflectional vs. derivational) make-up. Nor is it sufficient as a synchronic definition to refer to "source determination", pointing out that the specific grammatical meaning of + AD perfectives "follows directly from their richer lexical meanings" (Bybee et alii 1994: 89). What we observe here are substantial differences, based on different linguistic values, rather than minor details of realization. Moreover, as we have seen, several other Past types in European languages are capable of suggesting a direct and integral view of the event and are thus used for propulsive narration of sequences of past events. Such similar roles should not lead us to equate the linguistic values. Only +PAST (+AD) is marked, the other categories ("unqualified terminal pasts") taking up the space left over by different competing marked categories in their systems. Some are —POST and —INTRA items opposed to +POST and +INTRA items. Others have no +POST or +INTRA items to contrast with, and are thus indifferent to post- and intraterminality respectively. If "perfectivity" is defined in this unqualified sense, the + AD type cannot be claimed to be typical of it. The different types discerned above may be summarized as follows, listed in descending order according to their degree of explicitness with respect to direct and integral presentation: +PAST (+AD), +PAST (-POST (-INTRA)), +PAST (-POST), +PAST (-INTRA), +PAST.

11.8. System types, areal distribution The viewpoint operator systems dealt with above are of varying complexity. Very often, older stages of development are relatively rich, whereas a reduction is observed in modern ones. Thus, the Proto-Slavic aspectotemporal system was compta" hensive, including ±INTRA, ±POST, and ±PAST oppositions that yielded a -PAST (+INTRA°) Present, a +PAST (+INTRA) Imperfect, a +PAST (-POST (-INTRA))

Viewpoint operators in European languages

179

Aorist, a -PAST (+POST) Perfect, and a +PAST (+POST0) Pluperfect. There was also an actional +T-marking perfective vs. imperfective opposition, rather freely combinable with the members of the aspectotemporal oppositions. While most Slavic languages have lost this richness, Bulgarian and Macedonian have preserved and developed it. Italo-Croatian has a ±INTRA opposition Imperfect vs. Aorist (< Perfect) and a +T-marking system, whose values combine to form perfective Imperfects +INTRA x [+t], imperfective Imperfects +INTRA x [—t], perfective Aorists -INTRA x [+t], and imperfective Aorists -INTRA x [—t]. As for the Romance languages, Latin displays a richer system than its modern daughters. The rich Classical Greek system has been transformed into a not less rich Modern Greek system. Germanic systems are usually restricted to ±POST oppositions and lack systematic -|-T-marking. Western Finno-Ugrian systems are rather similar to them, but have also developed different additional devices. As we have seen, this is even true of the considerably reduced and "apparently extremely simple Hungarian system" (Comrie 1994: 299), which has, for example, developed a systematic +T-marking. Numerous modern languages of the Caucasian region have elaborate systems of distinctions in the ±POST and ±INTRA dimensions. On the other hand, peculiarly scanty systems are found in some languages, e.g., within the Daghestanian Andi group (Godoberi, Karata, etc.). Among the Iranian languages, Ossetic has a strikingly simple system, comprising one —PAST (+INTRA0) and one +PAST item. Turkic systems are, right from the Old Turkic period, highly well-equipped, including ±PAST, ±INTRA, and ±POST oppositions. Many more recent Turkic languages have also developed a rather systematic actional +T- and —T-marking, which also affects members of aspectotemporal oppositions. Questions of areal distribution, though occasionally touched upon above, will not be dealt with systematically in the present contribution. A comparison of the individual areas of reasonably grammaticalized ±INTRA, ±POST, and ±AD categories yields isoglosses that do not constitute coherent geographic areas. Both ±INTRA and ±POST are predominantly found in languages of peripheral areas, Icelandic, English, Irish, Ibero-Romanian, Bulgarian, Macedonian, Modern Greek, eastern FinnoUgrian, Turkic, Caucasian, Mongolian, etc. Languages void of well grammaticalized ±INTRA, ±POST, or ±AD categories (German, Hungarian, etc.) occupy a small, rather central area. Systems essentially restricted to ±POST (Scandinavian, Faroese, Frisian, Dutch, Baltic Finnic, etc.) are found farther north than systems essentially limited to ±INTRA (French, Italian, Romanian, Maltese, Turkish, etc.). The Slavic languages void of other categories than ± AD (mainly East and West Slavic) are distributed over a huge area in the eastern and central parts of Europe. The distribution is not very strongly correlated to genetic groups. Though there are highly stable systems, e.g., within the Turkic group, contact-induced changes have played an important part. The languages of Western Europe exhibit certain structural similarities that can only go back to long contact. Some are obviously due

180

Lars Johanson

to influence of dominant colonial languages, in particular Latin. There are several typical spread zones, e.g., north, south, and west of the Black Sea as well as in vast parts of Eastern Europe, areas of long-standing contact between Indo-European, Turkic, Finno-Ugrian, and other elements. Some contact-induced changes imply copying (some kind of "borrowing") of new items, whose effects depend on what items are already present in the recipient language. The Mari system is clearly influenced by the Turkic languages of the Volga region, and there is a considerable Azeri impact on Talysh and on the Daghestanian Lezgian group. The Upper Sorbian system is rather similar to the German one. Irish has a ±INTRA, ±POST system that is strongly reminiscent of the English system. However, no evidence can be found for the alleged Romance influence on the Maltese system, which is very close to those of other Arabic dialects (Ebert, this volume). Many changes start with frequential copying, i.e. contact-induced decrease or increase in the use of a given item (Johanson 1992: 182-183). For example, the increase in the use of Latvian -(-PAST (—POST) items instead of compound —PAST (+POST) items may be due to influence from Russian, where +PAST (±AD) items cover all past events (Mathiassen 1996: 15). A system may also change so strongly that it comes very close to the model system. The Basque system has changed in a way that mirrors the oppositions - and in part the formal structure - of the systems of neighbouring Romance languages (Comrie 1994: 299). There are also long chains of contact-induced changes, e.g., the widespread loss of —PAST (+POST) and —PAST (+POST (-INTRA)) items and their replacement by +PAST, +PAST (-INTRA) and +PAST (±AD) items in Eastern and Central Europe. Despite all differences in details, the overall similarities of the systems of viewpoint operators in European languages are rather striking. Occasionally, the available data of certain non-Indo-European languages, in particular Basque and some Caucasian languages spoken in residual zones, give an impression of underlying old deviant structures. However, it is by no means the task of the present contribution to try to trace documented structures back to prehistorical ones.

References Aerts, Willem J. 1965 Periphrastica. An investigation into the use oféînai and ékhein as auxiliaries or pseudoauxiliaries in Greek from Homer up to the present day. Amsterdam: Hakkert. Anderson, Lloyd B. 1982 "The 'perfect' as a universal and as a language-specific category", in: P. J. Hopper (ed•), Tense - aspect: between semantics and pragmatics. (Typological Studies in Languag 1.) Amsterdam-Philadelphia: John Benjamins, 227-264. Antinucci, Francesco & Ruth Miller 1976) "How children talk about what happened", in: Journal of Child Language 3: 167-189-

Viewpoint operators in European languages

181

N Aronson, Howard I. 1967 'The grammatical categories of the indicative in the contemporary Bulgarian literary language" in: To honor Roman Jakobson, I. (Janua Linguaram, Series Maior 31.) The Hague-Paris: Mouton, 82-98. Bache, Carl 1982 "Aspect and Aktionsart: Towards a semantic distinction", Journal of Linguistics 18: 57-72. Barentsen, Adriaan A. 1985 'Tijd', 'aspect' en de conjunctie рока. Over betekenis en gebruik van enkele vormen in het moderne Russisch. Amsterdam. [No indication of publisher]. Benzing, Johannes 1985 Kalmückische Grammatik zum Nachschlagen. (Turcologica 1.) Wiesbaden: Harrassowitz. 1988 Kritische Beiträge zur Altaistik und Turkologie. (Turcologica 3.) Wiesbaden: Harrassowitz. Bertinetto, Pier Marco 1987 "Structure and origin of the 'narrative' imperfect", in: A. Giacalone Ramat, O. Carruba & G. Bernini (eds.), Papers from the 7th International Conference on Historical Linguistics. Amsterdam, Philadelphia: Benjamins, 71-85. this volume "The progressive in Romance, as compared with English". Bertinetto, Pier Marco & Denis Delfitto this volume "Aspect vs. actionality. Some reasons for keeping them apart". Bertinetto, Pier Marco, Karen Ebert, Casper de Groot ' this volume "The progressive in Europe". Bickerton, Derek > 1975 Dynamics of a creole system. New York: Cambridge University Press. Bondarko, Aleksandr V. r 1983) Principy funkcional'noj grammatiki ivoprosy aspektologii. Leningrad: Nauka. Borg, Albert J. 1981 A study of aspect in Maltese. (Linguistica Extranea 15.) Ann Arbor: Karoma. Bosco, Umberto 1924 Esercizi di traduzione dai dialetti delta Calabria. (Dal dialetto alia lingua.) Torino. Bossong, Georg 1993 "Innovative Tendenzen im sardischen Vokalsystem", in: Johannes Kramer & Guntram A. Plangg (eds.), Verbum Romankum. Festschrift für Maria Iliesen. Hamburg: Helmut Buske, 337-352. Breu, Walter 1985 "Handlungsgrenzen als Grundlage der Verbklassifikation", in: Werner Lehfeldt (ed.), Slavistische Linguistik 1984. München: Otto Sagner, 9-34. 1991 "Das italoalbanische 'Perfekt' in sprachvergleichender Sicht", in: Francesco Altimari et alii (eds.), Atti del Congresso Internazionale di Studi sulla Lingua, la Storia e la Cultura degli Albanesi d'Italia. Celuc-Rende, 51-66. 1992 "Das italokroatische Verbsystem zwichen slawischem Erbe und kontaktbedingter Entwicklung", in: Tilmann Reuther (ed.), Slavistische Linguistik 1991. Referate des XVII. Konstanzer Slavistischen Arbeitstreffens Klagenfurt-St. Georg/Längsee 10.-14.9.1991. (Slavistische Beiträge 292.) München: Otto Sagner. 93-122. Browning, Robert '983 Medieval and Modern Greek. (2nd edition.) Cambridge: Cambridge University Press. Br ugmann, Karl 1900 Griechische Grammatik. Lautlehre, Stammbildungs- und Flexionslehre und Syntax. (3rd edition.) München: Beck.

182 1904

Lars Johanson

wtHuwi mi
196

Pier Marco Bertinetto and Denis Delfitto

incompatible with "until tx" due to their basically non-durative nature; however, to make things simpler, in this paper we shall only consider accomplishment verbs as illustrations of [+telic] predicates. As to the aspectual point of view, the adverbial expression examined here allows only for terminativity. To prove this, consider the following examples (cf. below for further qualifications concerning sentences (b) and (d).

(1) a.

/= atelic, terminative/ Mary danced until midnight, b. /= atelic, nonterminative/ (??)Mary was dancing until midnight, с /= detelicized, terminative/ Mary painted the wall until midnight, d. I- detelicized, nonterminative/ (??)Mary was painting the wall until midnight.

»

«

The question marks accompanying sentences (lb) and (Id) are within parentheses, for the English progressive allows for a "prospective" or "future time reference" reading (to be interpreted here as future-in-the-past), in which case these sentences would be grammatical. However, and not surprisingly given the inclinations of the adverbials employed, this would turn the aspectual value into terminative. Apart from this idiosyncratic detail, connected with the specific properties of the English progressive,10 the same actional and aspectual observations made above hold for Italian (and Romance languages in general), with the additional qualification that the Imperfect might be admitted in these contexts with a habitual meaning. This might seem to contradict the basically terminative nature of this adverbial. In fact, it does not do so, because one typical feature of the habitual aspect is that it admits terminative adverbials as long as they simply modify each single occurrence of the event, rather than the whole event. Indeed, the core of the habitual aspect lies in the indeterminacy of the total number of occurrences, rather than in the indeterminate duration of each occurrence. This is enough to guarantee its basic interminativity. Thus, one can find Italian utterances such as:

(2) a.

Italian /= atelic, habitual/ Ogni sabato sera, Maria ballava fino a mezzanotte. every Saturday evening Mary dance.IMPF until midnight 'Every Saturday evening, Mary used to dance until midnight.'

•:«»>1сг->< f b.

Aspect vs. Actionality

197

/= detelicized, habitual/ Ogni sabato sera, Maria puliva il giardino fino a every Saturday evening, Mary clean:IMPF the garden until mezzanotte. midnight 'Every Saturday evening, Mary used to clean the garden until midnight.'

where it is clear that the number of occurrences constituting the whole event is unde­ termined, for we do not know how many Saturdays are involved. Consequently, the whole event cannot be viewed terminatively, despite the terminative characterization of the single occurrences. We can make sense of this apparent paradox by saying that a habitual situation consists of a nonterminative macroevent composed of a series of terminative microevents. The adverbial "until tx" in (2) refers of course to the single occurrences, which are thus perfectly compatible with its aspectual character. These observations extend to other adverbials, as we shall see. The basically terminative nature of this adverbial is proved by the fact that in Italian the progressive periphrasis is totally excluded in these contexts. Spanish, however, is an interesting case in this respect, for in this language we may find the Simple Past with the progressive periphrasis, as in: (3) a.

b.

Spanish I- atelic, terminative/ Maria estuvo bailando hasta la media Mary was:SP dancing until the middle 'Mary kept dancing until midnight.' /= detelicized, terminative/ Maria estuvo pintando la pared hasta Mary was:SP painting the wall until 'Mary kept painting the wall until midnight.'

noche. night

la media noche. the middle night

This would not be possible with the Imperfect (except, as already noted for Italian, where it has habitual meaning). This is an important feature of Spanish (and IberoRomance languages in general), to which we shall return in Section 3.2. 2.1.2. From tx to ty Next, consider the adverbial type "from tx to ty" (It. "da tx a ty" Sp. "desde tx hasta l y )• Here, Italian and Spanish essentially behave as before, while English is more flexible, and tolerates to some extent nonterminative sentences, which sound accept­ able to many speakers (as suggested by the diacritic): (4)

%Mary was dancing from 10 p.m. to midnight.

198

Pier Marco Bertinetto and Denis Delfitto

The insertion of further temporal specifications enhances the acceptability of these sentences: (5)

Yesterday, during the maths class, i.e. from 9 to 10, Mary and John were playing cards; in fact, they went on playing even afterwards.

Note here that the relationship between adverbial and verb is only indirect, owing to the intermediation of a temporal clause, which attenuates the impact of the competing features. Thus, it is no wonder that in such contexts the Romance Imperfect, and even to some extent the progressive periphrasis, is not excluded: (6)

Italian Ieri, dalle 2 alle 3, quando tu credevi che stesse yesterday from 2 to 3 when you thought that was:SUBJ studiando, in realtà Maria giocava I stava giocando a studying in fact Mary play:IMPF / was:IMPF playing at tennis. tennis 'Yesterday, from 2 to 3 o'clock, when you thought that she was studying, Mary was in fact playing tennis.'

We may thus state that this adverbial, besides being atelic from the actional point of view, is tendentially terminative in Italian and Spanish, although not to the utmost degree. This is true even more in English, perhaps because of the ambiguous aspectual nature of the Simple Past (see Section 3.2 for further comments on this point). Unsurprisingly, Spanish admits, as in the previous case, the Simple Past in the progressive periphrasis. 2.1.3. Since tx Let us now consider the adverbial type ''since t x " (It. "da t x " Sp. "desde t^'). hi the Romance languages, this adverbial combines with both telic and atelic verbs, provided they are used nonterminatively, as we gather from: (7) a.

/= atelic, terminative/ *Maria hallo da mezzogiorno. Mary dance:SP since noon b. I- atelic, nonterminative/ Maria ballava da mezzogiorno (quando...) Mary dance.IMPF since noon (when) 'Mary was dancing since noon (when . . . ) '

п.г'!д]

ij.iüpectvs. Actionality

199

ч

c.

/= telic, terminative/ *Maria dipinse la parete da mezzogiomo. Mary paint: SP the wall since noon d. /= detelicized, nonterminative/ Maria dipingeva la parete da mezzogiomo (quando...) Mary paint:IMPF the wall since noon (when) 'Mary was painting the wall since noon (when ...)'

English is slightly different because in (b) and (d) one prefers the Past Perfect Pro­ gressive (cf. Mary had been dancing/painting the wall since noon). However, doing so amounts to using an essentially nonterminative device, for the conclusion of the event is not necessarily envisaged. Thus, the ultimate result does not change. Span­ ish behaves exactly like Italian (cf. Maria bailabal*bailôlpintaba la pared/*pintô la pared desde el mediodià). However, it must be remarked that in this language the progressive periphrasis with the Simple Past may be allowed here, as in the preceding cases (cf. Maria estuvo bailando desde el mediodia), provided there is a contextual presupposition which fixes a temporal limit, such as hasta las ocho "until 8 o'clock". But in the latter case we would in practice obtain the adverbial type "from tx to ty" which we examined in Section 2.1.2. This shows that "desde tx" has indeed a strong nonterminative value.

2.2. Type II adverbials 2.2.1.

/nXTime

The second series of temporal adverbials that we will examine is characterized by the presence of quantified (conventional) units of time. Let us begin with "in X Time"; e.g., in one hour, in two weeks, in six months. This demands telic verbs, and is associated with the terminative aspect only, as can be seen in:

(8) a.

/= atelic, terminative/ *Mary danced in two hours, b. I- atelic, nonterminative/ *Mary was dancing in two hours, с I- telic, terminative/ Mary painted the wall in two hours, d. /= telic, nonterminative/ *Mary was painting the wall in two hours.

200

Pier Marco Bertinetto and Denis Delfitto

2.2.2. For X Time Next, consider the adverbial type "for X Time" (It. "per X Tempo", Sp. "durante X Tiempo"): (9) a.

/= atelic, terminative/ Mary danced for two hours. b. /= atelic, nohterminative/ % Mary was dancing for two hours. c. /= detelicized, terminative/ Mary painted the wall for two hours. d. /= detelicized, nonterminative/ %Mary was painting the wall for two hours.

From the actional point of view, this is clearly a detelicizing adverbial. Indeed, (9c) does not mean that the event is carried through to the end. However, the situation regarding aspectual values is slightly complicated in English (as to Romance languages, see below). Although this adverbial type clearly prefers terminative contexts, sentences (9b) and (9d) receive contrasting judgements by the speakers, as shown by the diacritic used.1 They appear rather clumsy, at least in isolation, to some scholars (e.g., Mittwoch 1988, Hatav 1989), but acceptable to others (e.g., Vlach 1981). It is useful to compare these sentences with a truly nonterminative situation, such as: (10)

When I came in, Mary was dancing/painting the wall.

The difference in this sentence compared to (9a) and (9c) is as follows. Sentence (10), where the temporal clause provides the evaluation time, suggests an indeterminate situation, as far as the continuation of the event beyond the relevant moment is concerned. Indeed, for all we know, the event may have gone on beyond that moment, or it may have stopped right then. Sentences (9a) and (9c), on the other hand, suggest that the (relevant portion of the) event is entirely confined within the interval considered; it may of course be resumed later on, but that should explicitly be stated in a further sentence. Thus, sentences (9a) and (9c) combine an atelic and (respectively) detelicized situation with a terminative view. As already noted with (5-6), added temporal specifications seem to improve the situation when the nonterminative aspect is employed (cf. 11-12), although this is not always the case (cf. 13): (11)

% While I was sleeping, John was running for an hour.

(12)

% Yesterday I was told that, while we were in class, John was running for an hour.

(13)

*During the class, Mary was playing cards with John for 10 minutes.

Aspect vs. Actionality

201

Sentence (13) is ungrammatical, due to the fact that the "for X Time" adverbial indicates a short interval included into another relatively short one, so that we are in­ escapably bound to view the situation terminatively (i.e., by considering the terminal point of the event), and this is clearly incompatible with the Past Progressive. Sen­ tences (11-12), acceptable for some speakers, might be taken as counterexamples to this claim. Note however that when we add a punctual temporal clause, as in (14b), the incompatibility with the adverbial becomes quite clear, although the same sen­ tence without this adverbial type sounds perfectly natural (cf. 14a). Clearly, in these cases the punctual temporal clause helps us to focus on a particular moment of the event, which is thus viewed nonterminatively. Note further that the ungrammaticality of ( 14b) is not simply due to a possible conflict between the duration indicated by the "for X Time" expression and the punctual temporal clause, since (14c) sounds nat­ ural. However, and not surprisingly, due to the preceding punctual temporal clause, the Simple Past of the last sentence takes an ingressive meaning (a subspecies of the terminative aspect), which is alien to the Past Progressive:

(14) a. When I came in, John was running. b. *When I came in, John was running for an hour. с When I came in, John ran for an hour.

Thus, (14c) enhances the conclusion formulated above, according to which "for X Time" adverbials refer to a basically terminative situation. The acceptability, at least for some speakers, of (9b) and (9d), as well as (11-12), is probably due to the am­ biguous aspectual nature of the English Simple Past (cf. Section 3.2 for further com­ ments). Indeed, when the context forces a nonterminative interpretation, as in (14ab), the presence of this adverbial is unanimously rejected. Now consider Romance languages. In Italian the Imperfect of the progressive pe­ riphrasis is quite inappropriate with the adverbial "per X Tempo". The same applies to the bare Imperfect, unless it is understood in the habitual meaning. Spanish be­ haves like Italian as far as the Imperfect is concerned, but unlike Italian it may exploit the Simple Past in the progressive periphrasis, which sounds perfectly appropriate, due to its terminative character; cf. Maria estuvo bailando durante dos horns 'Mary w as dancing for two hours' (lit.: 'Mary was:SP dancing.'). This is the same sort of situation we noted above for "until tx" and "from tx to ty"; and indeed this does not c ome as a surprise, because the aspectual and actional properties of all these types °f adverbial are exactly the same. To sum up, the actional and aspectual inclinations of "for X Time" adverbials are as follows: atelic, terminative.

202

Pier Marco Bertinetto and Denis Delfitto

2.2.3. It. "da X Tempo", Sp. "desde X Tiempo", Fr. "depuis X temps" Let us now analyse the adverbial It. "da X Tempo" (Sp. "desde X Tiempo", F r "depuis X Temps"). This type is quite common in Romance languages, much less so in English. The only possible equivalent is "for X Time" apparently identical to the adverbial examined in Section 2.2.2. But, as we shall see, there are differences. The adverbial "da X Tempo" is compatible with both telic and atelic verbs, and with both terminative and nonterminative tenses, although with different meanings, and (saliently) with the exclusion of one combination of these features, namely *[—telic, +terminative]. It should be noted that this adverbial is not found in Ro­ mance with the Simple Past, whereas it may easily be with the Compound Past (morphologically equivalent to the English Present Perfect, but not strictly identical from the semantic point of view). Indeed, Bertinetto (1986) regards this adverbial as a powerful diagnostic tool for discriminating, within the terminative aspect, between perfectal tenses and purely aoristic ones. Consider the following Italian examples, keeping in mind that in these contexts the Compound Past takes a strictly perfectal value: (15)

Italian a. /= atelic, terminative/ *Maria ha ballato da due ore. Mary has.PRS danced since two hours b. /= atelic, nonterminative/ Maria ballaval stava ballando da Mary dance:IMPF was:IMPF dancing since 'M. had been dancing for two hours.' с I- telic, terminative/ Maria ha dipinto la parete da due Mary has:PRS painted the wall since two 'M. finished painting the wall two hours ago.' d. /= detelicized, nonterminative/ Maria dipingeval stava dipingendo la Mary paint:IMPF/ was.IMPF painting the ore. hours 'Mary had been painting the wall for two hours.'

due ore. two hours

ore. hours

parete da due wall since two

The actional value in (15d) may properly be defined as "aspectually detelicize because the adverbial is in itself compatible with telic verbs, as shown by С Indeed, the detelicization is entirely due here to the nonterminative aspect (rerne1" the so-called 'imperfective paradox'), rather than to the adverbial. Compare, on

Aspect vs. Actionality

203

ther hand ^ c ) ' w n e r e the verb is detelicized independently of the aspectual value "cf.alsothecommentinfn-9)The diffefent meaning of the adverbial, depending on the aspectual value, is quite oarent. V^ith terminative tenses, it indicates the interval elapsed between the end f the event and the evaluation time. With nonterminative tenses (specifically, with the progresS*ve aspect), it measures the time elapsed between the beginning of the event and th e evaluation time (nothing is implied, of course, as to the conclusion of the event). This explains the exclusion of (15a). This sentence looks weak because atelic predicates are not suitable for indicating that the event referred to is directed towards a s p e c m c goal; thus, it is not easy to determine from where one should start measuring the elapsed time (unless the final moment, which obviously exists, is explicitly given, as with the periphrasis ha finite di ballare 's/he has finished danc­ ing', which i s based on a telic verb). Sentence (15b), on the other hand, is perfectly natural, because durative events obviously have an initial, psychologically salient, moment. As noted above, English makes much more limited use of this type of adverbial in contexts corresponding to those exemplified in (15). It is only allowed with stative verbs and Perfect tenses. However, if these appear in the progressive form, then the actional restriction is relaxed, i.e., non-stative verbs may appear: (16) a.

I- -bstative/ Mary has/had known John for two months, b. /= + stative/ (*)Mary knew John for two months, с /= —stative; detelicized/ ??Mary has/had painted the wall for two hours, d. /= —stative; detelicized/ Mary has/had been painting the wall for two hours.

2.3. Type щ adverbials 2-3.1. Already similar ca se j s offered by the adverb already (It. già, Sp. yd), which belongs to Pe of temporal adverbials. This combines with all possible actional and ^Pectual tyjj eS; b u t a g a i n w i t h different meanings:12 e n e x t tv

' a'

'^ atelic, terminative/ Mary already danced the polka. U-e., some other time in the past] '^ atelic, nonterminative/ Mary was already dancing the polka, when I came.

204

Pier Marco Bertinetto and Denis Delfitto с

d.

/= telic, terminative/ Mary already painted this wall. [i.e., some other time in the past] /= detelicized, nonterminative/ Mary was already painting this wall, when I came.

As noted regarding (15d), the detelicization of the predicate occurring in (17d) does not depend on the direct contribution of the adverb, but is a mere product of the progressive aspect (remember again the 'imperfective paradox'), because the adverb is not in itself incompatible with telic verbs, witness (17c). Note, furthermore, that although already may be used in conjunction with both aspectual values, the meaning it acquires in terminative sentences (as indicated in the comments attached to (17a) and (17c)) must be regarded as a derived one, available only in particular contexts. This is shown by sentences like the following, where the terminative tense sounds inappropriate: (18)

??During the maths class, Mary already played cards with John.

2.3.2. Still The use of the adverb still (It. ancora, Sp. todavia) is more limited. It does not combine with terminative tenses, although it is indifferent to the actional value con­ sidered here:13 (19) a.

/= atelic, terminative/ *Mary still danced the polka, before leaving, [i.e. again] b. /= atelic, nonterminative/ Mary was still dancing the polka, when I came, с /= telic, terminative/ *Mary still painted this wall, before leaving, [i.e. again] d. /= detelicized, nonterminative/ Mary was still painting this wall when I came.

Actually, Italian and Spanish are more liberal in this respect, because they allow for a resemanticization of the adverbs ancora and todavia, which in conjunction with terminative tenses mean "again, another time", as shown by the following examples (to be compared with 19a): (20) a.

Italian Maria ballo ancora la polka, prima di andarsene. Mary dance.SP still the polka before of go.INF

Aspect vs. Actionality b.

ЭДЗ

Spanish Maria bailo todavia la polka, antes de irse. Mary dance:SP still the polka before of go:INF 'Mary danced the polka again, before leaving.'

2.4. Type IV adverbials The next type of adverbials, which may be called adverbials of 'graduality' (gradu­ ally, little by little and the like), is also relevant, as these treat actional and aspectual properties as two independent parameters. Namely, they admit both major aspectual values, but are restricted to telic predicates.14 Thus, they are unaffected by aspectual values, just as still and already are unaffected by the feature [± telic]. Consider:

(21) a.

/= atelic, terminative/ *Mary danced gradually, b. /= atelic, nonterminative/ *Mary was dancing gradually, с /= telic, terminative/ Mary painted the wall gradually, d. /= telic, nonterminative/ Mary was painting the wall gradually.

Besides presenting adverbials of graduality corresponding to the English ones (gradualmente, a poco a poco), Italian exhibits a peculiar adverbial {man mano) with more constrained properties. This is found only with a combination of telic verbs and non­ terminative (specifically, progressive) aspect:

(22) a.

/= telic, terminative/ ??Maria dipinse man mano la parete. Mary painf.SP little by little the wall 'Mary painted the wall little by little.' b. /= telic, nonterminative/ Mentre Ugo puliva il giardino, Maria dipingeva while Ugo clean:IMPF the garden Mary paint:IMPF man mano la parete. little by little the wall 'While Ugo was cleaning the garden, Mary was little by little painting the wall.'

206

Pier Marco Bertinetto and Denis Delntto

3. Aspectual and actional properties as independent entities 3.1. Synopsis It may be useful, at this point, to recapitulate what we have observed in Section 2. For simplicity's sake, we shall tabulate only data referring to English, with the single exception of the adverbial type It. "da X Tempo" (which also stands for similar constructions in other languages, such as Sp. "desde X Tiempo" or Fr. "depuis X Temps"). Note that the diacritic ' ± ' has two different meanings: either that the given adverbial type is truly indifferent to the specific value of the diacritic, or that the adverbial is compatible with both values, but with separate readings. In the latter case, we insert the diacritic within parentheses for clarity. Furthermore, we make use of the exclamation mark to suggest that the given adverbial type shows the tendency indicated to a high degree: Table 1. Actional and aspectual properties of selected English (and Romance) temporal ad­ verbiale. Actionality

Aspect

- telic! - telic! - telic!

+ terminative! + terminative — terminative!

+ telic! - telic! ± telic

+ terminative! + terminative (±) terminative

± telic ± telic

(±) terminative — terminative

+ telic!

± terminative

Type I "until t x " "from tx to t„" "since t x " rpe II "in X Time" "for X Time" It. "da X Tempo" фе III already still ^pelV adverbials of graduality

It should be noted that with It. "da X Tempo", the combination *[-telic, +terminative] is excluded.15 As to the detailed differences between English on the one side, and Italian and Spanish on the other, the reader is directed to the preceding discus­ sion. To sum up briefly: in Italian and Spanish the adverbs corresponding to still are [(±)terminative]; in addition, all adverbials demanding the features configuration [—telic, +terminative] accept in both languages the Imperfect with habitual mean­ ing, while in Spanish they also tolerate the progressive periphrasis with the Simple Past.

i««

Aspect vs. Actionality

20f

Table 1 clearly shows that aspectual and actional values are independent enti­ ties, as claimed at the outset, and contra the opinion defended in for instance Sasse (1991). Indeed, the various adverbials behave quite differently with respect to these two semantic categories. Consider for instance still. This is neutral to the [± telic] distinction, but is clearly selective with regard to aspectual values. Conversely, adverbials of graduality are very tolerant with aspectual values, but quite selec­ tive with actional ones. These are of course the most relevant cases, because they diverge in the sharpest way; but the varying behaviour of the remaining adver­ bials is quite revealing. Thus, "until t x ", "from t x to t," and "for X Time", which are [-telic!, +terminative(!)], contrast neatly with "in X Time", which is [+telic!, -bterminative!], and so forth. We shall return to this in Section 4.2.

3.2. Progressive periphrasis and terminative aspect One topic needing clarification is a peculiarity of Spanish alluded to above. Like English, this language has preserved the possibility (once also exhibited by Italian) of combining the progressive periphrasis with both terminative and nonterminative tenses (see Section 3.3). Unlike English, however (but like any Romance language), Spanish is morphologically endowed with a typically nonterminative tense, called "Imperfect". This provides the Spanish progressive periphrasis with additional ex­ pressive power, in contradistinction to both Italian (which accepts only nontermina­ tive tenses with this periphrasis) and English (which lacks a markedly nonterminative tense, like the Romance Imperfect). As observed above, this has specific consequences for all adverbials of the type [—telic, -bterminative], such as "until t x ", "from t x to t," and "for X Time": (23) a. b. с d. e.

English Italian Italian Spanish Spanish

/= /= /= /=

%Mary was dancing for two hours, Imperfect/ *Maria stava ballando per due ore. Simple Past/ *Maria stette ballando per due ore. Imperfect/ *Maria estaba bailando durante dos horas. Simple Past/ Maria estuvo bailando durante dos horas.

Here, Italian yields ungrammatical sentences (cf. 23b-c), while English behaves rather ambiguously (cf. the discussion relating to examples 9-14 above). As shown by the diacritic in (23a), the progressive does not sound entirely appropriate to a number of speakers, owing to the purely terminative character of the adverbial. Span­ ish, on the other hand, makes a very neat distinction in these contexts (cf. 23d-e). The progressive periphrasis with the Imperfect is unacceptable (just as in Italian), whereas with the Simple Past it sounds absolutely appropriate. 16 Clearly, Spanish !s able to counteract the nonterminative orientation of the progressive periphrasis by means of an explicit morphological tool. This effect, on the other hand, does not

,

208

Pier Marco Bertinetto and Denis Delfitto

extend to the adverbial type "in X Time" (*Mar(a estuvo pintando la pared en dos horas; cf. 8d for translation). However, as we saw above, the latter adverbial demands a strictly telic situation; and this is not ensured by the progressive periphrasis, which inevitably determines a detelicization of the predicate. The reason why English encounters some difficulties in (23a) is presumably a direct consequence of the ambiguous aspectual characterization of the Simple Past. It should indeed be noted that this tense is regularly used in contexts where Romance languages would normally employ the Imperfect: (24) a. b. (25)

Italian I- Imperfect/ Maria era bruna e aveva gli occhi azzurri. /= Simple Past/ Mary was dark-haired and had blue eyes. Italian /= Imperfect/ II cervo corse, corse, raggiunse la tribu dei cervi che vedendolo con un uomo sulle corna un po' lo sfuggivano, un po' gli s'avvicinavano curiosi. (I. Calvino) 'The deer ran, ran, reached the deer troup which, seeing it with a man on its antlers, escaped (*were escaping) or approached (*were approaching) it in turn with eagerness.'

Obviously, in (24) the progressive could not be employed because of the stative character of the verb; however, it is worth noting that it is precisely in contexts such as those that Romance languages tend to make almost exclusively use of the Imperfect, reserving the Simple Past for highly marked stylistic registers.17 As to (25), the English progressive would sound inappropriate even though the verb is not stative. Thus, the English Simple Past turns out to be appropriate both for truly terminative contexts, where Romance languages would also normally employ the Simple Past, and for truly nonterminative contexts, where Romance languages employ the Imperfect (as in 24-25). This suggests an explanation for the divergent reactions of English speakers, some of whom tend to reject sentences such as (23a), presumably because they regard the periphrastic construction "was + VERB-ing" as genuinely nonterminative. Those who accept such sentences, on the other hand, seem to consider this construction as a sort of syncretic tool, combining the meanings of both Sp. "estaba + Gerund" and "estuvo + Gerund". We believe that this phenomenon has still further consequences: indeed, attempts to provide a semantic treatment for the progressive have so far seriously underestimated this fact (Bertinetto and Delfitto, 1996). One point on which Spanish and English converge is the possibility of combining the progressive periphrasis with Perfect morphology (in our view, a subspecification

>-'rt»t| th::.r- ! ^yawi^>щ-»;> *еГ

-

(78)

Xtizi / éxtize éna spfü se dhfo build:(IPFV):PRS / build:IPFV:PST INDEF house in two mères. days 'He can/could build a house in two days.'

(79)

Plénete / plenötan wash:(IPFV):PRS:PASS / wash:IPFV:PST:PASS '(This item) should be hand-washed.'

sto xéri. in_DEF hand

The borderlines between habitual, generic, potential and other uses of the Imperfective are not very clear. What they have in common, however, is the qualifying character. No instantiation of a situation is considered. Rather, some entity is characterized in one way or other by being connected to some situation, which is thus naturally viewed non-temporally as a situation type. In Russian, there are some similar contexts where the Perfective is used. This has been referred to as the graphic-exemplary use of the Russian Perfective. (80)

Russian (Rassudova 1984: 114)

,bs

Obratites' к moemu bratu, on vam vsegda turn to my:DAT brother:DAT he you:DAT always pomozet. help:PFV:PRS 'Turn to my brother; he'll always help you.' It should be noted, however, that this use of the Perfective is limited to contexts with present-time reference. Since the Russian Perfective Present is used as a Future, (80) could be interpreted as corresponding to the use of the Future in, for example, En­ glish (cf. the translation). The overlapping of the Present and the Future in contexts of this kind is natural, since the situation referred to may be seen in a more or less general perspective, as exemplified in (81) and (82). (81)

If you ask him, he'll help you.

(82)

He (is such a person who) helps people.

The Perfective would thus correspond to the first perspective, the Imperfective to the second.

The type-referring function of the Imperfective

255

5.6. Schedule and time point expressions-type and token time The situation referred to in schedule expressions like (83) may also be described as a potential situation. (83)

To tréno févji / éfevje se pende DEF train leave:(IPFV):PRS / leave:IPFV:PST in five leptâ. minutes 'The train leaves/was leaving in five minutes.'

It is not the actual leaving (the situation token) taking place five minutes later (if it ever takes place) that is predicted. What is communicated is that five minutes later is/was the time for departing (the situation type). The scheduling use sometimes comes close to the habitual use. This is the case in (84), which, in fact, is ambiguous. (84)

To kalokéri to proino tréno tha févji pjo noris DEF summer DEF morning- train FUT leave: PFV more early 'In the summer the morning train will leave/be leaving earlier.'

The train will be leaving earlier either according to some summer schedule, which will be valid, or just because the lazy conductor will manage to get up in time during the summer. As was noted in Section 2.4, a specific time point may qualify a situation although the situation itself is not referred to as a unique instantiation. This is exemplified in the schedule cases. In five minutes in (83) denotes a specific time point, but it does not denote the time point for some specific instantiation of leaving (by some specific train at the time point t), only the time point when - according to some schedule - 'there should be departing'. When the temporal qualifier has the first function it could be said to denote "token time" (time for a situation token) and when it has the second function "type time" (time for a situation type - represented by a situation token or not). This differentiation is important in explaining examples like the following, where the situation is not a potential one as in the schedule cases but actually did occur at some time (cf. (25) repeated as (85) below):16 (85)

To 1945 pandrev6tan. DEF marry:IPFV:PST 'In 1945 he married.'

256

Eva Hedin

(86)

Sta ikosi mu xrönja spudhaza vévea .0 in_DEF twenty me:GEN years study:IPFV:PST naturally акбта, sta ikosiéna éperna to ptixîo still in_DEF twenty-one take:IPFV:PST DEF diploma i mu, ke sta ikosidhio éfevgha ja ti me.GEN and in_DEF twenty-two leave:IPFV:PST for DEF Mési Anatoli. Middle East 'In my twenties I was still studying of course, when I was twenty-one I graduated and when I was twenty-two I left for the Middle East.'

(87)

Sarandapénde xronja prin, séna nosokomio, ksepsixuse forty-five years ago inJNDEF hospital die:IPFV:PST énas Iskios - ârostos ke perifovos ke monos. INDEF Shadow sick and frightened and alone 'In a hospital, forty-five years ago, a Shadow passed away - sick and frightened and alone.'

Although the temporal modifiers refer to some specific time, the situations denoted by the imperfective verb forms are not considered as the specific instantiations occurring at that time. It is not so much a question of what actually concretely took place at this specific time in the past (the token, the instantiation as such) as of pointing out the time for something (somebody's marriage, graduation, departure or death). The situation referred to is thus viewed in a more abstract way as a whole (as 'marrying', 'graduating', 'departing' or 'dying'; cf. the discussion in 2.5), as a non-temporal situation type. (Cf. also section 3 above on focusing on the circumstances instead of the verb.)17

5.7. Imperfective with verbs of communication In the literature on Russian and Classical Greek it has been pointed out that verbs of saying are often used with the Imperfective in contexts where one might expect the Perfective. This is to some extent also true of Modern Greek, where verbs denoting communication in some contexts occur in the Imperfective:

The type-referring function of the Imperfective

257

(88)

["Makâri na katâpina ti ghlösa mu ! ' ] Aftö éleje xthes stin Peanfa о X. pu that say:IPFV:PST yesterday in_DEF Peanfa DEF COMP ipostirize tin proighumeni Paraskevi ôti claim:IPFV:PST DEF last Friday COMP ['If only I had swallowed my tongue!'] 'X said this yesterday in Peania, he who last Friday claimed that... '

(89)

["Néi ânthropi, nées idhées ke thavmastâ apotelésmata",] sxöljaze xthes athlitikos parâghondas, commenf.IPFV'.PST yesterday athletic representative [anaferömenos sto proxthesinö dérbi ton eonion andipalon, OlimpiakuPanathinaiku]. '["New people, new ideas and wonderful results",] an athletic representative commented yesterday, [referring to the derby the other day between the eternal antagonists Olympiakou-Panathinaikou.]'

Again, what is central is not that which actually concretely took place at the time of the utterance in question - the instantiation, somebody said something or commented on something at some point in time-but what (non-temporal situation type - saying, commenting, uttering, claiming, etc.) was instantiated (without focusing on the instantiation as such). Rather than 'yesterday X said such and such', what is communicated is something like 'yesterday there was saying such and such by X' or 'X was the agent of making such and such an utterance'. Compare also the following sentences. In (90) the speaker wonders about the source of information (without considering any instantiation of saying), whereas (91) could be uttered by a school teacher who wants to know who in the class (actually, a few seconds ago) uttered the nasty comment he just heard when writing on the blackboard. (90)

Pjos (to) éleje? who (it:ACC) say:IPFV:PST 'Who said (that)?'

(91)

Pjos to ipe aftö? who it:ACC say:PFV:PST that:ACC 'Who said that?'

The fact that verbs of saying often occur in the Imperfective may partly be due to its Possibility to denote two sides of the utterance, namely the concrete act of speaking ; °r writing on one hand, and its "performative" function on the other. One example \ « the difference in interpretation of the verb say, which apart from its more concrete meaning of uttering, may have the meaning 'tell', 'inform', 'express an opinion',

258

Eva Hedin

'maintain', 'claim' etc. (by speaking, signing or writing). In some contexts, where the content of the utterance and its source is what is important, not the particular instantiation of uttering, type-focusing instead of token-focusing of the situation may be a way to underline this more "abstract" meaning by shifting focus away from the "concrete" token.

5.8. Scripts and scenarios In Slavic aspectology it has been pointed out (Vinogradov 1947: 558, cited in Leinonen 1982: 195) that there is a connection between normality or usualness of a situation and the Imperfective in verb phrases like 'to have breakfast', 'to have dinner', 'to have tea'. (92)

Russian My zavtrakali v vosem' casov. we have_breakfast:IPFV:PST:PL at eight hour:PL:GEN 'We had breakfast at eight.'

Leinonen (1982: 195) refers to this kind of expressions as representing a script, i.e., "a predetermined, stereotyped sequence of actions that define a well known situation" (Schank & Abelson 1977, cited by Leinonen 1982: 196). She considers the script to be a variant of the scenario structure of the kind discussed in Newton (1979), where it is argued that "the basic schematic form of one common class of multiple-event expressions" may be paraphrased 'Every time p, then q'. He uses the term "scenario" to refer to such "statements of constant conjunctions of occasions" (1979: 139). A scenario structure may be overt as in (93) or covert as in (94). (93)

Every time he sees me, he insults me.

(94)

John always drinks coffee.

In an overt scenario expression the antecedent is specified whereas in a covert one it is only implied. The antecedent for (94) would be something like 'whenever he drinks anything (non-alcoholic)' (Newton 1979: 140). The link between scripts and scenarios would according to Leinonen be repetition, which is a condition for stereotyping. However, the scenario structures describe "one class of multiple-event expressions", namely the habitual use of the Imperfective, and I find it hard to see how a script expression like had breakfast in (92) could fit into this class. That a situation is frequently repeated may be a condition on the world for the situation to be comprehended as a stereotype, but repetition is not what is expressed by the stereotyped phrase.

ч The type-referring function of the Imperfective

259

The connection between script expressions of the kind referred to above and the Imperfective could be explained by the type-referring function of the Imperfective. When referring, for instance, to somebody's eating some food in the morning we may say either He greedily ate three portions of bacon and eggs or He had breakfast and somebody's drinking tea in the afternoon we may refer to as He carefully sipped a hot cup of tea or He had tea speaking about the same situation. In the first case it would typically be the actual situation (the instantiation in time of eating some food or drinking some tea) that is referred to. In the second the situation is viewed as type and the breakfast and the tea is referred to as non-specific, only defining the type of situation (as 'breakfasting' or 'tea-drinking'). In these examples the Russian imperfective verb sometimes corresponds to the English lexical construction with have instead of eat or drink etc. Consider the following example from Mehlig (1979: 156, cited by Leinonen 1982: 165): (95)

Russian Segodnja v sem' casov Ivan prinimal lekarstvo. today at seven hour:PL:GEN Ivan take:IPFV:PST medicine 'Today at seven Ivan took his medicine.'

According to Mehlig, the adverbial does not have the function of localizing the sit­ uation, rather it is part of the proposition. He means that it is predicated about Ivan that he normally takes his medicine at seven o'clock. Again, I find it hard to see that 'Ivan's medication at seven' is referred to as a habit. However, by referring to the situation as a stereotype, using the Imperfective, one may infer that it is part of a habit, something that is normally done (in this case by Ivan). (92) and (95) could also be given as examples of time point Imperfective discussed above, the time points giving type-time rather than token-time. (The normal) breakfasting/medicating was executed at eight/seven. Thus, the reason why the Imperfective is used for stereotype situations would be that these are referred to as situation types. In fact, it could be argued that, in some sense, stereotype situations are exactly what the Imperfective always refers to. To refer to a situation as a static, non-temporal type means to consider it as a non-temporal, invariant whole, that is, as a stereotype.

6.

Summary

In this paper it has been proposed that the functional difference between the Perfective and the Imperfective could be described in terms of type and token reference. With the Imperfective, situations are considered in a static, non-temporal perspective a s types (neglecting any instantiation), as opposed to the Perfective, which is used

260

Eva Hedin

for reference to situations as tokens or instantiations occurring in time (in a temporal, "left-to-right" perspective). Some typical uses of the Imperfective have been described and reinterpreted against the background of its type-referring function. Not only in non-specific con­ texts as in the generic, habitual or potential uses (where no specific situation is fo­ cused on) may the Imperfective refer to the situation as a type, but also in contexts where it is possible to focus referentially on some situation and this situation conse­ quently could be referred to in both ways. Thus, according to the description proposed here, in a sentence like О Jânis xtîzi/éxtize éna spiti 'John is/was building a house' (that is, an example of the frequent continuative use of the Imperfective), the situation is referred to as a nontemporal situation type, as 'house-building', not as an instantiation of it in time, as some house actually coming into existence at the reference time or later. Presupposition seems to be important for the choice of the Imperfective in Modern and Classical Greek as well as in Russian for instance. When the situation is known or given by the context, what is given is exactly the type of situation, the denotative content of the verb phrase. If this situation is referred to as the presupposed situation, it is naturally the situation as type that is considered. The connection between negation and the Imperfective which has been observed may be a secondary one depending on this link between presupposition and the Imperfective. The ingressive and the signal uses of the Imperfective may be described as referring to a situation as an execution of some situation type typically given by the context (for instance, as 'Execute (the) Verb-ing!' rather than 'Verb!'). This may naturally give rise to an ingressive interpretation. The neutral function of type reference of neglecting the instantiation of the situation (for instance neither stating nor denying its actual accomplishment) explains why the conative interpretation may occur with many transitional verb phrases in the Imperfective. That a conative interpretation is not the only possible reading of these imperfective verb forms, however, is shown by the fact that the same forms in other contexts may have the opposite interpretation, viz. that of completion. The schedule and time point Imperfectives are used for reference to time points as the time for some abstract situation type (referred to as type time as opposed to token time), not to the actual concrete instantiation at this time point (in case there is one). Verbs of communication sometimes occur in the Imperfective "pro Perfective". This is connected with the more abstract function that some verbs of communication may have apart from referring to some concrete act of uttering. This abstract function would favour type-focusing reference to the act of communication, where a concrete token-focusing interpretation is avoided by using the Imperfective. Finally, the use of the Imperfective with phrases of the type 'have tea', 'have breakfast', well-known from Slavic aspectology, is discussed. It is connected to the

The type-referring function of the Imperfective

261

function of the Imperfective as type-referring and thus in some sense always referring to situations as stereotypes.

Acknowledgment Financial support for the work presented has been received from the Swedish Council for Research in the Humanities and Social Sciences (HSFR).

Notes 1. For a definition of the term "bounded", see 2.2. 2. For clarity, the term "Modern Greek" is used throughout this paper instead of just "Greek" as in the rest of the volume. 3. Forsyth (1970: 82) claims that the simple denotative function use reflects "the most basic and general function of the imperfective verb: to identify the type of action, naming it without reference to the question of its 'perfectivity' or otherwise". Although the first part of this description looks very much like the one proposed in this paper, the second part makes it differ crucially from it. Thus, according to Forsyth, when the Imperfective is used with the simple denotative function, it has a function which, although basic, is distinct from its function in other contexts (where it is non-perfective). Although he considers the function of the Imperfective to "identify the type of action" as basic, he does not consider it as the function of the Imperfective as opposed to the Perfective. Perfectivity and non-perfectivity are treated like another kind of opposition, which is neutralized in the simple denotative function of the Imperfective. 4. Examples are from Modern Greek if there is no indication of language. 5. In some other (extralinguistic) context (TMAQ: 14) could get the Imperfective in the answer provided the sense of the question is not 'What happened after dinner?' but 'Which activity did he choose to occupy himself with after dinner' having e.g. earlier discussed different possible alternatives (such as wash the dishes, make a phone call or write a letter). 6. In the Greek correspondence to (20) the perfective would be used: Pjos to stolise; Who it decorate:PFV:PST It should be noted, however, that the Greek example, like the English translation of (20), contains an explicit object. This syntactic structure, which is the only natural one in Greek, would, according to my informants, allow both aspects in the Russian example. 7. According to Bakker the situation referred to also has to start immediately. According to his description the givenness of the type of action is thus one of the factors determining the aspect choice in this kind of context, immediacy another (cf. 6.3). 8. Translation by C F . Smith (Loeb Classical Library no. 169 [1976]. W. Heinemann Ltd: London & Harvard University Press: Cambridge, Massachusetts.).

262

Eva Hedin

9. The example is from Lindstedt (1985). Rassudova notes that the context with an adverb meaning 'once' is a very favourable context for the "general-factual" meaning (Rassu­ dova 1984: 71 ) of the Imperfective in Russian. It is also a typical context for the Modern Greek Pluperfect. 10. A typical way of referring to situations as types is to state their existence within a tempo­ ral domain. Experiential interpretations, for example, tend to occur with temporal frame expressions (cf. Dahl & Hedin, this volume). 11. Translation by Walter Miller (Loeb Classical Library no. 52 [1968]. W. Heinemann Ltd: London & Harvard University Press: Cambridge, Massachusetts). This is also an ex­ ample of the use of the Imperfective to refer to a given situation, since the verb is used earlier in the context. 12. Translations of Xenophon quotations are by Carleton L. Brownson (Loeb Classical Li­ brary no. 89 [1968]. W. Heinemann Ltd: London & Harvard University Press: Cam­ bridge, Massachusetts). 13. I find the explanation of Rijksbaron (1984: 18) a bit strained, according to which, in examples of this kind, "the value [non-completed] of the imperfect serves to direct the attention to the consequences of the completion of the action". It is hard to see why non-completion should lead to the interpretation of result. 14. Translation by С F. Smith (Loeb Classical Library 169 [1976]. W. Heinemann Ltd: Lon­ don & Harvard University Press: Cambridge, Massachusetts). 15. It should be noted that in the English translation by J.H. Vince (Loeb Classical Library no. 299 [1964]. W. Heinemann Ltd: London & Harvard University Press: Cambridge, Massachusetts.) the second (Imperfective) Imperative Anagîgnôske is left out altogether. This way of rendering the imperfective verb form may perhaps be interpreted as illustrating the signal function of the Imperfective referred to above (cf. 27). The reader may imagine some extralinguistic behaviour - a gesture for instance - taking on the signalling function. 16. Given that the Imperfective always denotes situation types, this has to be the function of all time point expressions combined with it. The interpretation will differ. If in 1945 in (85) is changed to at that very moment the sentence would more naturally have the interpretation 'At that very moment he was being married' (if the context tells us that at that time he and his bride were standing in front of the priest). Also in this case, however, the time point expression would denote the type time and not refer to the time for a situation token (cf. 2.3). 17. Cf. the use of the Imperfective in similar constructions in French (cited by Pollak I960: 145-151): Il fut nommé ministre, et deux ans après il mourait 'He was appointed a minister, and two years later he died' (p. 150). Et la semaine suivante, il s'embarquait à Brest 'And the following week, he embarked at Brest' (p. 150). Reparti de Las Palmas le ..., В. arrivait à ... le 23 décembre 'Having returned from Las Palmas on ..., В. arrived in ... on the 23rd of December' (p. 151). Pollak (in a footnote on p. 147) refers to Gamillscheg (1957: 407) as the only one of

The type-referring function of the Imperfective

263

the cited authors who does not consider this use of the Imperfective as "pittoresque" or expressing a "lebhafte Vorstellung". According to Gamillscheg the Imperfective is used for the following reasons: "In den Sätzen, die die Zeitbestimmung enthalten, ist diese vordringliche Mitteilung. Dadurch bekommt das im Imperfekt ausgedrückte Geschehen den Charakter eines Nebenumstandes, wenn auch logisch betrachtet eine unabhängige neue Mitteilung vorliegt (emphasis added)." A direct correspondence to the Modern Greek example (87) is the introduction to a chapter on Chateaubriand in a history of literature (cited by Pollak 1960: 148): Le 4 septembre 1768, naissait à Saint-Malo, dans la sombre rue des Juifs, le chevalier François de Chateaubriand. 'On September 4, 1768, was born at Saint-Malo, in the dark street of the Jews, the chevalier François de Chateaubriand.' Also in Italian the Imperfective is used in a similar way: In quello stesso anno nasceva a Firenze Dante Alighieri. 'In the same year Dante Alighieri was born in Florence.'

References Bakker, Willem Frederik 1965 "The aspect of the imperative in modern Greek", Neophilologus 49: 89-103. 1966 The Greek imperative. Amsterdam: Hakkert. Comrie, Bernard 1976 Aspect. An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge University Press. Dahl, Osten 1981 "On the definition of the telic-atelic (bounded-nonbounded) distinction", in: P. Tedeschi & A. Zaenen (eds.), Syntax and semantics 14: Tense and aspect. New York: Academic Press, 79-90. 1985 Tense and aspect systems. Oxford and New York: Blackwell. Dahl, Osten & Fred Karlsson 1976 "Verbal aspects and object marking: A comparison between Finnish and Russian", International Review of Slavic Linguistics 1: 1-30. Dahl, Osten & Eva Hedin this volume "Current relevance and event reference". Dowty, David R. 1979 Word meaning and Montague grammar. Dordrecht: D. Reidel. Forsyth, James 1970 A grammar of aspect: Usage and meaning in the Russian verb. Cambridge: Cambridge University Press. Gamillscheg, Ernst 1957 Historische Französische Syntax. Tübingen: Max Niemeyer Givön, Talmy 1978 "Negation in language: Pragmatics, function, ontology", in: P. Tedeschi & A. Zaenen (eds.), Syntax and semantics 9: Pragmatics. New York: Academic Press, 69-112.

264

Eva Hedin

Hedin, Eva 1987

On the use of the perfect and the pluperfect in modern Greek. Acta Universitatis Stockholmiensis: Studia Graeca Stockholmiensia VI. Stockholm: Almqvist & Wiksell.

Leech, Geoffrey 1974 Semantics. Harmondsworth: Penguin (5th edition, 1978). Leinonen, Marja 1982 Russian aspect, "Temporal'naja lokalizacija" and definiteness/indefiniteness. Helsinki. Lindstedt, Jouko 1985 On the semantics offense and aspect in Bulgarian. Slavica Helsingiensia 4. University of Helsinki, Helsinki. Mackridge, Peter 1985 The modern Greek language. Oxford: Clarendon (2nd paper back edition, 1989). Mehlig, Hans Robert 1979 "Überlegungen zur Funktion und Determinierung der Aspekte im Russischen", in: Jochen Raecke & Christian Sappok (eds.), Referate des VI. Konstanzer Slavistischen Arbeitstreffens, Tübingen 26.-29. Sept. 1978. (Slavistische Beiträge, Band 133, Slavistische Linguistik 1978). München: Otto Sagner, 151-169. Newton, Brian 1979 "Scenarios, modality, and verbal aspect in Modern Greek", Language 5: 139-167. Pollak, Wolfgang 1960 Studien zum 'Verbalaspekt' im Französischen. (Österreichische Akademie der Wissenschaften, Philosophisch-historische Klasse, Sitzungsberichte 233, 5.) Wien: Rudolf M. Rohrer. Rassudova, Olga Petrovna 1984 Aspectual usage in modern Russian. Moscow: Russky Yazyk. Rijksbaron, Albert 1984 The syntax and semantics of the verb in classical Greek. An introduction. Amsterdam: J. С Gieben. Ruijgh, С J. 1985 "L'emploi 'inceptif ' du thème du présent du verbe grec", Mnemosyne XXXVIH, Fasc. 1-2: 1-61. 1991 "Les valeurs temporelles des formes verbales en grec ancien", in: Jadranka Gvozdanovic & Theo A. J. M. Janssen (eds.), The function of tense in texts. Amsterdam: North-Holland, 197-217. Schank, Roger С & Robert P. Abelson 1977 "Scripts, plans and knowledge", in: P. N. Johnson-Laird & P. С Wason (eds.), Thinking. Readings in Cognitive Science. Cambridge: Cambridge University Press, 421^132. Sicking, С M. J. 1991 "The distribution of aorist and present tense stem forms in Greek, especially in the imperative". Glotta 69: 154-170. Smyth, Herbert Weir 1920 Greek grammar. Cambridge, Massachusetts: Harvard University Press (8th edition, 1973). Stunovâ, Anna 1993 A contrastive study of Russian and Czech aspect: Invariance vs. discourse. Ph.D. dissertation, University of Amsterdam, Amsterdam. Schwyzer, Eduard 1975 Griechische Grammatik. Zweiter Band. Syntax und syntaktische Stilistik. (4th edition.) München: С. Н. Beck.

.С:ь - - , , , ч

h-mb,,.,:-r •»в

RUB

K m

>Fri LSrb

Dut storm

HSrb

flOin

Cl

toi

Blr Tid Dkr

Ir Hng Baq

Nit

slvi SCr

Rum

»rt Spa

Ctl

Stlt Big Alb

I

ata

P 1. The languages of the sample

Irk

tea

268

RolfThieroff

*0

2. Describing tense and aspect systems 2.1. Morphosyntactic categories and categorizations Taking English as an example, the following forms and constructions are traditionally viewed as belonging to the paradigm of the verb: sing, sang, has sung, had sung, will sing, would sing, will have sung, would have sung, is singing, was singing, has been singing, had been singing, will be singing, would be singing, will have been singing, would have been singing. If we look at these forms, statements like the following impose themselves: all forms are finite, as opposed to further verbal forms like (to) sing, sung, singing. Two forms consist of the main verb only, one which we could call unmarked (sing), one which we might provisionally call "marked" (in this case by ablaut: sang). Four forms contain the elements have + past participle of the main verb; the element have appears either in a finite form - again, either unmarked (have) or marked (in this case by the allomorph -d, functionally equivalent to the ablaut: had) - or in the infinitive (in will have sung and would have sung). Finally, four forms contain the elements will + infinitive; will is always finite, twice unmarked and twice marked (ablaut + -d). The infinitive is either the infinitive of the main verb (in will/would sing) or the infinitive of have + the past participle of the main verb (in will/would have sung).3 It is obvious that these eight forms or constructions are combinations of (the presence or absence of) the three elements [marked], [have + past participle], and [will + infinitive]. To get a picture of the entire paradigm, one more element has to be added, namely be + -ing. This element can be combined with each of the eight forms, thus yielding a paradigm of a total of 16 forms. This can be depicted as in Table 2.

Table 2. Paradigm offiniteverb forms in English unmarked -[will + inf.] +[will + inf.] non-progressive —[have + PP] sings +[have + PP] hassung

"marked" -[will + inf.] + [will + inf.]

will sing sang will have sung had sung

would sing would have sung

progressive — [have + PP] is singing will be singing was singing would be singing +[have + PP] has been singing will have been had been singing would have been singing singing

On the areal distribution of teoee-eepect categories in Europe

markedness

/\ +

-

have + will

will + inf.

/\

/\

+ -

+

be + -ing

/4 -

+

-

Figure 1. Categories and categorizations offiniteverb forms in English

aspect p r e s e n t / / / /I future perfect past/ / /past perfect future/ perfect

progressive non-progressive

indicative conditional

Figure 2. Traditional classification of thefiniteverb forms in English

What we have now is a classification of the 16 verb forms of English, assigning each form to four different formal categories. Each of these categories is a member of a class with two members. We could also say that each form has been classified or categorized with respect to four classes or categorizations. This is shown in Figure 1. Figure 1 represents the morphosyntactic categories and categorizations of the verb in English. Morphosyntactic categories thus are the result of a classification of a paradigm according to formal properties. Each verb form is classified or categorized with respect to four categorizations: markedness of the finite part, presence or absence of have + past participle, presence or absence of will + infinitive, presence or absence of be + -ing. Classifying or categorizing a form with respect to four categorizations implies that the form has four categories. For example, sing belongs to the categories —[marked], —[have + past participle], —[will +infinitive], —[be + -ing]. Would sing has the categories + [marked], —[have +past participle], +[will -binfinitive], -[be +-ing] etc. If we now try to replace the category names in Table 2 by the traditional notions, we run into problems. At first sight we could replace "unmarked" by Present, "marked" by Past, have + past participle by Perfect, will + infinitive by Future and be + -ing by Progressive. However, "marked" will + infinitive is not called Past Future (analogous to Past Perfect) in grammars of English, but "Conditional". This means that, obviously, not each form with the feature "marked" is regarded as Past, nor is each form with the feature will + infinitive regarded as Future. Furthermore, not every categorization in Figure 1 is regarded as a categorization HI traditional grammars. Instead, both Past and Future and often also Perfect are subsumed under a categorization "tense" as opposed to the categorization "aspect" with the two categories progressive and non-progressive. Finally, would + infinitive is re-

270

RolfThieroff

garded as belonging to an extra categorization, "mood". A traditional classification of the forms thus would have to be reconstructed as in Figure 2. The reason for the rather different analyses in Figure 1 and Figure 2 is that in one analysis only the forms are considered, whereas in the other the meanings of the forms are taken into consideration as well. The crucial question now is how the analysis of the forms can be brought together with the meanings associated with these forms.

2.2. Compositionality

•

Whereas it is obvious that the morphological make-up of the forms in Tab!* 2 is strictly compositional, this is less clear for their semantics. In fact, some scholars have proposed systems in which the complex, analytic tense forms of English (and other languages) are treated as indivisible units for the purposes of semantic interpretation. In such systems the present perfect, the past present [sic], the present perfect progressive, and so on, are each given a special treatment. Yet other scholars have argued that the progressive, the perfect, and each of the tenses require separate treatments, so that the analysis of the present perfect progressive, for example, follows from, or consists in, that of the present plus that of the perfect plus that of the progressive. This requires that the treatment of such complex forms be at least weakly compositional, in the sense that semantic interpretation of the expression as a whole consists in semantic interpretation of its constituent parts. (Binnick 1991: 257f.) Binnick continues by saying that "intuitively a compositional approach makes sense", and this is the approach taken in the present chapter. This means that forms which are morphologically compositional are also regarded as semantically compositional as long as no obvious differences in the basic meaning speak against a compositional analysis. With regard to the basic meaning, following Kortmann, the following assumptions are made: [...] we need to make a strict separation between semantics and pragmatics. A compositional account can only be concerned with what is asserted and not with what is inferred. It can only be concerned with intension, not with extension [...], that is with meaning invariants out of context and not with different uses to which a grammatical category is put. Nevertheless [...] the basic meanings which a compositional analysis postulates for individual categories must underlie, or at least be compatible with, the various uses to which the relevant categories can be put. Thus whatever interpretation we may want to give a present perfect or past perfect in a given sentence, the invariant meaning we postulate for the category perfect as such must be compatible with this interpretation

On the areal distribution of leme-aspect categories in Europe

271

[...]. However, in view of the many idiomatic uses that often exist for individual cate­ gories there is widespread agreement that it suffices if this requirement is met only by the basic, non-idiomatic uses [...]. (Kortmann 1995: 184f.) A compositional analysis of complex forms thus does not exclude certain specific restrictions on the use of the elements of these forms in certain combinations. For example, the semantic description of the category anterior ("perfect") as E < R (point of event before point of reference) holds for the Present Perfect in English, although, in contrast to Past Perfect and Future Perfect, the Present Perfect cannot be collocated with adverbials denoting a specific time in the past (see 3.1, 3.2).4 Neither is the compositional analysis excluded by the fact that certain complex forms may have additional uses not entirely predictable from the meanings of their elements. Thus a compositional analysis has to account for certain idiosyncrasies occurring when certain categories are combined in one form. The problem of compositional vs. non-compositional analyses is also raised in Dahl (1985: 67). Dahl discusses arguments both for and against a compositional analysis. As for pluperfects, the author comes up with the solution to count them "as instances of both the cross-linguistic category PERFECT and PLUPERFECT". We have now reached a point where we can modify the scheme given in Figure 1. Instead of referring to the categories by writing down the morphological elements by which they are constituted (Table 2), we can give names to the categories which also give an indication of their semantic content (Figure 3). The names of categorizations are simply derived from the names of the categories. Henceforth, names of marked categories will be abbreviated, with capital letters. In figures, categorizations are printed in boldface. For a justification of the terms preterite (PRET) and anterior (ANT instead of perfect) see 3.1. The term "past" is used as a cover term for the categories PRET and IMPF. As has been mentioned, in English each verb form belongs to four different cate­ gories. However, instead of referring to the form sang as "preterite, non-anterior (i.e., unmarked for anteriority), non-future (i.e., unmarked for futurity), non-progressive (i.e., unmarked for progressivity)", it will be sufficient to refer to this form as the preterite of sing, i.e., by mentioning the marked category, implying that sang is not marked with regard to anteriority, futurity, or progressivity. In this sense, the term 'preterite' is an abbreviation of [+preterite, —anterior, —future, —progressive]. Ac!

!

I pastness

1 anteriority

1 futurity

1 progressivity

RET unm

ANT unm

FUT unm

PROG unm

p

Figure 3. Morphosyntactic categories of thefiniteverb in English

к

272

RolfThieroff

/

cordingly, with 'preterite progressive' we refer to the form categorized as [+preterite, —anterior, —future, +progressive], with 'preterite future' to the form categorized as [+preterite, -anterior, -(-future, -progressive] etc. In contrast, verb forms with the category PRET are all verb forms with the preterite marker (i.e., sang, was singing, had sung, would sing etc.). Note that in traditional tense names there are some incongruences. The form with the categories [—preterite, +anterior, —future, —progressive] is traditionally called the 'Present Perfect', whereas the form with the categories [—preterite —anterior, -(-future, —progressive] is simply called 'Future', instead of 'Present Future', although this Future is opposed to the preterite future in the same way as is the present anterior to the preterite anterior. It should have become clear by now how the term 'present' is to be understood. With 'present' we refer to the form which is categorized [—preterite, —anterior, —future, —progressive], i.e., 'present' is used as a term for the maximally unmarked form. In combinations such as 'present anterior', 'present progressive', 'present' has the meaning 'non-preterite'. However, in order to avoid misunderstandings, I shall refer to the entirety of the forms with the category [—preterite] as the non-preterite forms (instead of present forms). Similarly, the nonprogressive forms are all forms belonging to the category [—progressive] etc. For the term 'pluperfect', see 3.3.

2.3. Non-compositional analyses Despite what has been said in the previous section, a non-compositional semantic analysis of morphological composite forms is not generally excluded. A case in point is Modern Basque, whose verbal paradigm consists of analytical verb constructions only. The paradigm of the forms is represented in Figure 4, rendering the forms listed in Table 3. However, the meanings of these forms are far less compositional than their morphology is. Haase indicates the tense-aspect oppositions in Table 4.

auxiliary unmarked

marked

participle unm

inessive verbal noun

+ -a(k)

+ -ko Figure 4. Morphosyntactic categories of thefiniteverb in Basque I

On the area! distribution past development, stages 1-3 present anterior > past development, stage 3 GPST (stage 4)

Map 3. Development of present anteriore

Finally, in stage 4, represented by the north Slavic languages except Upper Sorbian and by Slovene, neither a combination of the former ANT with PRET/IMPF nor with FUT is possible any longer. The new category, which covers the four semantic fields displayed in Figure 6, is called general past (GPST). Whereas Map 3 shows the coherent area where present anteriors are currently developing into more general pasts, an area in the centre of the area under investi­ gation, Map 4 shows in a sense the reverse picture: here those languages are high­ lighted which do have a present anterior which is exclusively or predominantly used ш current relevance contexts. In other words, these are the languages possessing the perfect gram in.the sense of Bybee & Dahl. As Lindstedt (this volume) points out, 'the perfect has become a peculiar 'maritime category' in Modern Europe - most °f the languages and dialects with a stable perfect are situated on the fringe of the continent".13 This is confirmed by Map 4.

286

RolfThieroff

кг» L3rb

Storm USrb

a arm

j>ol

B l r

Yid

ci

i

B*q

Spn

stable present anterior ("maritime perfect") Map 4. Stable present an tenors

3.3.

Pluperfects

In all languages of the sample having either the categories ANT and PRET or ANT and IMPF (which by definition implies the existence of AOR), these categories can be combined, yielding a pluperfect, i.e. a form whose meaning can roughly be de­ scribed as referring to an event which took place prior to a point of reference which in turn is situated prior to the time ofspeech (E < R & R < S). As has been mentioned before, in languages with the IMPF vs. AOR distinction, it is always the IMPF that is used to construct the pluperfect. In addition, in some Romance languages, there is also a marginal AOR + ANT pluperfect (cf. Figure 8 and Table 7). In all languages having a pluperfect, the core meaning is composed of the semantics of its categories, i.e., the core meaning is anteriority (due to ANT) to a point of reference in the past (due to PRET or IMPF). In addition, some special uses of pluperfects may occur, as described in Dahl (1985: 144ff.). In addition to the PRETflMPFf ANT pluperfects a second way of constructing pluperfects occurs in the languages of the sample, viz. the so-called supercompound forms, i.e., forms resulting from applying the perfect operator twice to the same verb. Supercompound forms are only attested in languages where the present anterior is

On the areal distribution of tense-aspect categories in Europe

287

undergoing or has undergone the development toward a more general past category. While this is not astonishing, given that in these languages the PRETYIMPF tends to be abolished, it is not possible to predict at which stage of the present anterior development supercompound forms occur, nor whether they occur at all. In the languages of the sample, supercompound forms exist in French, in some Northern Italian dialects and in some varieties of Romanian (Squartini & Bertinetto, this volume), in Standard German, Southern German, Yiddish, Albanian and in the three Slavic languages Serbo-Croatian, Slovene, and Polish. In French, Standard German, and Albanian, the supercompound forms coexist with the (older) ANT+PRETVIMPF pluperfects; in French and German these forms are rare and are not accepted by normative grammars; in Albanian, the ANT+ANT pluperfect is again restricted to the north-east Gheg dialects (Buchholz & Fiedler 1987: 133). In Slovene and Polish14, the supercompound pluperfect is the only pluperfect (and to my knowledge normally not called supercompound). In Serbo-Croatian the supercompound form seems to be the only existing pluperfect today (Rehder 1991: 52; Gvozdanovic 1995).15 Since in these languages AOR and IMPF have been lost (Polish and Slovene) or almost lost (Serbo-Croatian), their pluperfects have to be described as GPST+GPST. In the remaining languages undergoing (Upper Sorbian) or having undergone (Northern Slavic, Hungarian) the present anterior > past development, supercompound forms are not attested. Finally, it should not go unnoted that the Romanian pluperfect is not compositional morphologically, but rather a continuation of the Latin inflectional pluperfect. Still, semantically, the Romanian pluperfect does not differ from the pluperfects of the other languages described, i.e. it can be analyzed as being composed semantically of the features "past" and "anterior", but not morphologically of the categories IMPF and ANT. This also applies to the synthetic pluperfect {cantara 'he had sung') in Portuguese, which is "grammatically equivalent" with the analytic pluperfect (tinha cantado 'he had sung') and "is effectively limited to written forms of European Portuguese" (Parkinson 1988: 150f; cf. also Hundertmark-Santos Martins 1982: 189). Note as well that the existence of such synthetic pluperfect forms is one of the reasons for Dahl establishing the cross-linguistic category of PLUPERFECT (cf. 2.3). On Map 5 the languages with a supercompound pluperfect are shown along with the present anterior > past languages and the few languages lacking a pluperfect altogether.

4.

Futures

Under the heading "European future gram families" the morphology of the future ! categories in the European languages is described in Dahl (this volume b). It bej comes clear from Dahl's description that with the possible exception of what Dahl

288

RolfThieroff

Big

j3 past development, stages 1-3 pluperfect: supercompound forms no pluperfect

Map 5. Supercompound pluperfects

calls the "futureless area", there are future markers in all European languages considered by Dahl, to which we may add Maltese, Turkish, and Armenian, which also possess future markers.1 As for the "futureless area", Dahl takes "the obligatory use in (main clause) prediction-based contexts as a main criterion for the grammaticalization" (Dahl, this volume b), and given that this does not hold for the Finno-Ugrian and Germanic languages except English, he claims that these languages constitute a futureless area. Indeed in these languages in general, future time reference can be referred to with the unmarked form (the present), in other words, in general the future is not obligatory in sentences with future time reference. However, in all languages there are special devices to refer to future time reference which can (optionally) be used in future contexts.17 In traditional descriptions of Finnish and Estonian, a future category is not mentioned (for Finnish see Hakulinen 1957, Olli 1958, Fromm 1982, Karlsson 1983). However, in both languages there are devices too which are (optionally) used to refer to future events (see Tommola 1992a, 1992b, 1994; Metslang & Tommola 1995). In any case there can be little doubt that in both Finnish and Estonian the

On the area! distribution of tense-aspect categories in Europe

Spn

О О

Ctl

Big

TrJt

289

past devel­ opment (see 3.2, Table 8). In addition, due to the lack of PRET, there is no future-

к

290

RolfThieroff

in-the-past in Southern German, Yiddish, and Hungarian (stage 3), nor in Romanian and Serbo-Croatian (stage 2), nor in Upper Sorbian (stage 1). Finally, the combination of FUT and PRET is lacking in Lithuanian and in the language with the least grammaticalized FUT, Estonian. Whereas the languages lacking a future-in-the-past belong to one coherent area, languages without ANT+FUT are isolated in Europe. With the exception of Sorbian, all are situated at the fringes of the area investigated. Again, for Finnish and Estonian the lack of ANT+FUT can be explained by the weakly grammaticalized FUT in these languages. In Irish it is the anterior which is only weakly grammaticalized, which may explain the lack of the combination there. In Basque and Armenian, however, both ANT and FUT are well-established categories, and a reason for the impossibility of combining both in one verbal form cannot be given.

5.

Aspects

5.1.

PFV:IPFV

One of the major issues of the literature on aspect in the last few decades is the question whether the AOR:IMPF opposition discussed in 3.1 and the aspectual opposition found in, e.g., the Slavic languages are instances of the same cross-linguistic opposition perfective:imperfective, and if so, how differences in the use of the respective forms can be accounted for. Whereas Slavicists often claim that only the Slavic opposition deserves the name of aspect and only the members of this opposition should be called "perfective" and "imperfective", recent typological research has come to the insight that it is rather the Slavic opposition which is atypical cross-linguistically. So for Dahl [t]he question that arises is whether the Perfectivity/Imperfectivity opposition in Russian, Polish, Czech and Bulgarian should be subsumed under PFV:IPFV at all. In particular for Bulgarian, where Perfective/Imperfective exists alongside the Aorist/NonAorist opposition, this would appear a natural conclusion. (Dahl 1985: 85) Indeed, since, "although there is also a rather high correlation between the two systems in Bulgarian, one can find a sufficient number of disharmonie choices, sucn as Imperfective Aorists" (Bybee & Dahl 1989: 87), it seems to make little sense to claim that both oppositions belong to the same cross-linguistic categories. Similarly» Bertinetto & Delfitto argue that "if boundedness [i.e., the Slavic "aspect"; R.T.] w e r a purely aspectual phenomenon, it would be hard to understand how it could eve develop in Bulgarian, which has independent (and abundant) aspectual devices at i disposal" (Bertinetto & Delfitto, this volume).18 Thus, the existence of imperfecta

}

On the areal distribution of teese-aspect categories in Europe

291

Aorists and perfective Imperfects in Bulgarian is considered a first, crucial argument for the assumption that AOR:IMPF are distinct from the Slavic style opposition, which I shall henceforth call the PFV:IPFV opposition. Some other arguments are listed in Bybee & Dahl 1989. The authors argue that the Slavic aspectual systems differ from the tripartite systems in their origins, their semantics, their means of expression and their relation to other parts of the system of verbal grams such as tense. (Bybee & Dahl 1989: 85) As for the origins, Bybee & Dahl remark that in Slavic languages "one can see a tendency for bounders [such as up in English eat up; R.T.] to become grammaticized as aspectual markers" (p. 86), which "makes the Slavic aspectual systems much more derivational in their character than the tripartite systems which are typically inflectional" (p. 87). Another important feature of the Slavic style opposition is "that morphologically the opposition between perfective and imperfective aspect is almost wholly independent of the category of tense in Russian" (p. 87). In fact, the categories AOR and IMPF on the one hand and IPFV and PFV on the other hand are radically different also with regard to their compositionality (see 3.1).19 Finally, the authors point out that in (1)

[What did your brother do after dinner yesterday?] He wrote letters

cited from Dahl (1985), all the Slavic languages use the imperfective aspect, whereas normally a perfective form is used, which "strongly suggests that there is a systematic semantic difference between the two kinds of perfectivity distinctions" (Bybee & Dahl 1989: 88). However, somewhat surprisingly, the authors finally come to the conclusion that "the similarity between the perfective meaning evolved historically from bounders and that which evolves from periphrastic constructions (i.e., perfects) is strong evidence for the validity of universal gram-types for perfective aspect" (Bybee & Dahl 1989: 89). Another scholar discussing the similarities and differences between the two oppositions is Lindstedt, who, with regard to (2), uttered in the same context as (1), observes: "After all, Slavonic, Romance and many other languages agree in using w hat is called the perfective in [(2)], and sentences of this kind form a clear crosstinguistic focus of the postulated category" (Lindstedt 1995). ")

Russian On napisal pis'mo. he write:PFV:SG letter.ACC 'He wrote a/the letter'

292

RolfThieroff

However, Lindstedt then proposes to make a distinction between a "material bound", which is crucial for the Slavic perfective, and a temporal bound, crucial for Romance aspect. This distinction also underlies the analysis in Bertinetto & Delfitto (this volume) and at least the notion of material bound can be connected with Johanson's 'adterminality', "envisaging the event in the attainment of its relevant limit: ad terminum" (Johanson, this volume). Thus, following Johanson, Bertinetto & Delfitto, Feuillet (1983), and re-interpreting the facts presented by Dahl, Bybee & Dahl and Lindstedt, I conclude that the Slavic style aspects PFV and IPFV are categories different from the categories AOR and IMPF of the Southern European languages. Whereas much has been written about the categories PFV and IPFV in the Slavic languages, it is less clear in which other languages these categories are found. From Europe, Bybee & Dahl cite Latvian and Lithuanian, Hungarian, and Georgian as languages with "a tendency for bounders to become grammaticized as aspectual markers" (Bybee & Dahl 1989: 86). As for Lithuanian, the existence of the PFV.IPFV opposition seems to be commonly accepted (cf. Sliziené 1995). With regard to Hungarian, rather different views can be found in the literature. In Bânhidi, Jokay & Szabö (1975) no aspects or aspect markers are mentioned at all. Abondolo (1987: 587) only remarks that "like preverbs in Slavonic languages, Hungarian coverbs are connected with aspect". For Dahl, things are less evident ("it appears that we are not dealing with a systematic manifestation of PFV"; "many verbs do not seem to have any aspectually marked counterparts"; Dahl 1985: 86) and the Hungarian forms get the label "PFVd?" (Dahl 1985: 71). De Groot claims that Hungarian does have the PFV:IPFV opposition, the prefix meg- being a perfectivizer (De Groot 1989: 7). Csato claims that the Hungarian meg-forms represent an aktionsart distinction rather than perfective aspect. For a detailed discussion see Csato (1994: 232-237). Still a tendency for bounders to become grammaticalized as aspect markers cannot be denied for Hungarian. Therefore, on Map 7 (see 5.3) Hungarian is shown as belonging to the PFV:IPFV area. Whereas the difference between the AOR:IMPF opposition in the Romance languages, Bulgarian and Armenian on the one hand and the PFV:IPFV opposition in the Slavic languages on the other hand is quite clearcut, there remains one problematic case: In Modern Greek, the position labelled AOR:IMPF clearly has features of both the Romance and the Slavic-style oppositions. Whereas the perfective and imperfective past are used to a large extent like aorist and imperfect in the Romance languages, the Greek perfective can be combined with the future, but not with the anterior, which is the opposite of the situation in the Romance languages, Bulgarian and Armenian. At the same time, the Greek opposition is also rather different from the Slavic one: Whereas the Slavic perfectives are built derivationally, the Greek perfective is constructed inflectionally, and with regard to the semantics Hedin mentions that the boundedness distinction, which plays "an important role for instance

On the areal distribution of tense-aspect categories in Europe

293

in the Russian aspect system", "cannot [...] be crucial for aspect choice in Modern Greek" (Hedin 1995: 236). On the other hand, Mackridge notes that there is "an important group of verbs which exist only in the imperfective" (Mackridge 1985: 103), a feature which is reminiscent of Russian imperfective verbs without perfective partners, whereas I am not aware of verbs which cannot have both aorist and imperfect. Note finally that the Greek perfective past also occurs in current relevance contexts (see Dahl & Hedin, this volume), a usage which is not possible with the prototypical aorists of the other languages mentioned. For further discussion see Thieroff (1995: 19-20).

5.2. Progressives The categories progressive (PROG) and habitual (HAB) are commonly regarded as special cases of imperfectivity. Thus, for example, Comrie classifies progressive and habitual as subdivisions of the imperfective aspect (Comrie 1976: 24-40). In Johanson's terminology, IPFV, PROG, and HAB are regarded as having in common the same "viewpoint aspect", namely intraterminality. The difference between the three categories is described as a difference of degrees of focality: Johanson's nonfocality corresponds roughly to our IMPF, his high focality to our PROG, and his low focality to our HAB (see Johanson, this volume). Similarly, Bertinetto & Delfitto (this volume) say that the progressive and the habitual "belong" to the imperfective aspects. As to how PROG is to be distinguished from the imperfective aspect, Dahl (1985: 92f.) mentions the following features: (i) In contradistinction to PFV:IPFV [i.e., AOR:IMPF; R.T.], which is strongly correlated with the distinction between past and non-past time reference, PROG is usually independent or almost independent of time reference - in other words, it is used both of the present, the past and the future [...] (ii) PROG is quite infrequently extended to habitual meaning [...] (iii) [...] PROG is normally used only of dynamic - that is, nonstative - situations [...] Dahl also notes that "PROG and PFV:IPFV are very different in the ways in which they are marked - PROG [...] is rather consistently marked periphrastically, whereas [• -.] the marking of PFVTPFV is usually much less straightforward" (p. 93).21 The progressive is one of the focal areas of the theme group on tense and aspect. For detailed information about the progressive in the Germanic languages, the reader is referred to Ebert (this volume a), about the evolution in the Romance languages to Bertinetto (this volume) and about the progressive constructions in Finnish and Estonian to Tommola (this volume).

294

RolfThieroff

The progressive area is the largest aspectual area, covering the western half of Europe but reaching to the east as fans Estonian and (with reservations) Lithuanian. In a large area in the south and southwest, the AOR:IMPF languages and the PROG languages overlap: with the exception of Romanian, all Romance languages have both the AOR-.IMPF opposition and PROG. Within the languages of the progressive area, at least two subgroups have to be distinguished, as is also observed by Comrie: In some languages, the distinction between progressive and nonprogressive meaning by means of progressive and nonprogressive forms is obligatory, whereas in others the use of the specifically progressive forms is optional, i.e., the nonprogressive form does not exclude progressive meaning. (Comrie 1976: 33). English, Irish, West Frisian, Icelandic, and Maltese belong to the first (relatively small) group of languages, where the progressive is grammaticalized in a way that it has to be used obligatorily in the appropriate contexts, at least in some tenses. In English, the grammaticalization of the progressive has progressed the farthest, which can be seen from the fact that the English progressive can be combined with all tenses, including the perfect, and with the habitual periphrasis (as in John used to be writing poems; cf. Comrie 1976: 33). Next to the English progressive is the Icelandic progressive. According to Ebert (1996), "the Icelandic progressive [...] is fully grammaticalized and underlies none of the restrictions formulated for the other Germanic languages. Similarly to the English progressive, Icelandic vera ad V has developed various special functions" (cf. also Kress 1982: 159-162). In Maltese, the progressive cannot be combined with the anterior. For Irish, see 5.3. Similarly to perfects, which universally tend to become pasts (cf. 3.2), progressives, at least unless they are combined with marked tenses, tend to develop into aspectually unmarked forms, in other words, present progressives often become simple presents (or at least imperfective forms; cf. Bybee & Dahl 1989: 82f.). Such a development has occurred in English and Maltese, where the morphologically simple present has with most verbs (exceptions are static verbs) a habitual meaning and in unmarked present time contexts the progressive is obligatory for dynamic situations (for Maltese, cf. Ebert this volume b). A similar development can be observed for telic verbs in Icelandic, which have future time reference with the unmarked finite form (the Present) and present time reference with the progressive (cf. Bonner 1995). The remaining Germanic languages, the Romance languages (with the exception of Romanian), Finnish and Estonian belong to the second group. In these languages, there exist (sometimes several different) progressive periphrases, which are however optional. The situation can be compared to the futures in German, Dutch, Frisian, Icelandic, Finnish and Estonian. In general, the progressives in these languages can be combined with all tense categories (as in the languages of the first group), although there are a few restrictions, especially with ANT and with AOR, as to

On the areal distribution of tense-aspect categories in Europe

295

example in Italian, where the progressive periphrasis cannot be combined with the Passato prossimo (present anterior), nor with the Passato remoto (AOR). In Lithuanian there are forms with progressive meaning as well, but here PROG is restricted to 1) the past and 2) the non-anterior forms. Thus the situation in Lithua­ nian differs considerably from both the languages of the first and of the second group. Note that Lithuanian is the only language with both the PFV:IPFV opposition and PROG. Finally, there are two languages whose former progressives can no longer be regarded as instances of PROG. The development from a progressive into the unT marked category has almost come to an end in Basque, where the former progressive is the unmarked present in Modern Basque, with the exception of about sixteen verbs which still can be used with the older present form, which no longer exists for the other verbs (Haase 1994: 283ff.). This is the reason why Basque is outside the pro­ gressive area on Map 7. A similar case can be observed in Turkish, where in the present a former PROG morpheme (yor) constitutes the unmarked present today and has become a marker of IMPF in combination with past and future (see Johanson 1994: 261f.).

5.3. Habituais The last tense-aspect category to be treated in the present chapter is the category habitual (HAB), the only category not belonging to the "major gram-types" in Bybee & Dahl (1989).22 Indeed, of the languages of the sample, there are only a few languages with a commonly accepted habitual category. According to Comrie (1976: 27f.), "[t]he feature that is common to all habituais [...] is that they describe a situation which is characteristic of an extended period of time, so extended in fact that the situation referred to is viewed not as an incidental property of the moment but, precisely, as a characteristic feature of a whole period." Dahl (1985: 97) illustrates the properties of HAB as follows: It appears that the cases where HAB is typically used are those in which the adverb usually is possible in English [...] These involve quantification over a set of occasions which is given explicitly or by context. For instance, in (Q.20) ", the set of occasions consists of the time intervals immediately following upon the breakfasts the speaker's brother had last summer. The use of HAB indicates that what is expressed in the sentence took place in the majority of those occasions. Such sentences differ from generic ones by their lack of lawlikeness. Of course, as for notional perfectivity, imperfectivity, and progressivity, it is pos­ sible to express the meaning of habituality in all languages. However, with regard to the expression of habituality, it seems to be more difficult to decide 1) whether a given expression has to be regarded as a fully grammaticalized morphosyntactic i. к

296

RolfThieroff

•#

^

category obligatory in the appropriate contexts, and hence as a form belonging to the verbal paradigm; 2) whether it is a weakly grammaticalized category, optional in the appropriate contexts but consisting of a construction whose meaning is not predictable from its elements (like the progressives in most Germanic and Romance languages); or 3) whether we simply are dealing with a lexical expression. For exam­ ple, in Dahl (1985: 96) German pflegen + infinitive and Swedish bruka + infinitive are mentioned as instances of HAB, but at least for the German construction the traditional view that pflegen + infinitive is a lexical expression seems to be more appropriate. Bertinetto (1996) mentions four different habitual periphrases in Ital­ ian (solere, essere solito, esser uso, aver I'abitudine di + infinitive), but the mere existence of so many expressions is an indication that there is no grammaticalized HAB in Italian. So, in this section, though it is not excluded that there are (weakly grammatical­ ized) habitual constructions in other languages of the sample, fully grammaticalized habitual categories are only accounted for in six languages: Czech, Irish, Lithuanian, English,25 Yiddish, and Upper Sorbian. In the first three, HAB is expressed morpho­ logically, i.e., by a bound marker - Czech -va- (Kucera 1981: 178); Irish -dhl-nn (Ö Siadhail 1989: 178); and Lithuanian -da- (Slizienè 1995). In the latter, HAB is expressed periphrastically - in English by used to + infinitive; in Yiddish byfleg + infinitive (Aronson 1985: 177, Kiefer 1994); and in Upper Sorbian by a form of by'be' + /-participle, formally identical with the conditional (Faßke 1981: 253, 266; Lötzsch 1995). Whereas the Czech infix -va- can be used with past and non-past forms, Lithuanian -da- occurs only in combination with PRET, and the English and Yiddish constructions have past time reference too.26 For Irish, the situation is not quite clear. Ö Siadhail (1989: 177) states that "habituality is only apparent when it is combined with a past tense", but this is only true for the finite main verb. The verb Ы 'be' has a habitual form (bim 'I am usually') distinct from the non-habitual (faim 'I am') also in the present, and since the progressive is constructed periphrastically with a finite form of bi, there is also a habitual and a non-habitual present progressive in Irish. And finally, the so-called impersonal form has a habitual form too, marked by the infix -f-.27 On Map 7, the three aspectual distinctions discussed in this section are represented together with the AOR-.IMPF opposition discussed in З.1. 28 Note that, while the phenomena discussed so far are with very few exceptions found in coherent areas - there is a PRET, an AOR:IMPF and a GPST area, a PFV: IPFV and a PROG area, an area of the present anterior > past development and a "stable present anterior" area - the (fully grammaticalized) habitual is the only category which is not found in one area, but seems to be randomly distributed ove Europe, though there are more habituais in the east than elsewhere. It is also wort mentioning that the habitual is not correlated with genetic language groups. Or u1

On the «real distrifeHtian of lease-aspectcategories in Europe

Q^> @ a

S *-

AOR:IMPF PFV:IPFV HAB PROG PROG highly grammaticalized

J95f

>' !

i!

Map 7. Aspectual areas six languages with a fully grammaticalized habitual, one is Celtic, two are Germanic, two are Slavic, and one is Baltic.

6.

Conclusion

If we take the maps presented in the previous sections together, several areas of similar tense-aspect systems can be identified. Perhaps the most striking correlation is depicted on Map 8, which shows that the overwhelming majority of the languages with a stable perfect also have PROG, and «at the overwhelming majority of languages with PROG also have a stable perfect. Exceptions are French (though only the spoken varieties) and varieties of German a nd Italian, which have PROG but belong to the area of the present anterior > past development, and Basque, where the former PROG has developed into imperfective

298

£3

RolfThieroff

>o

stable present anterior ("maritime perfect") PROG

Map 8. Correlation of PROG and perfect

forms. Exceptions with a stable perfect, but lacking PROG, are the four languages in the very south-east of the area investigated, namely Greek, Bulgarian, Turkish, and Armenian. On Map 9, this latter group appears again (area 6), being the only group of languages with the opposition AOR:IMPF, a stable perfect, and lack of PROG. Three more areas with the feature of having a stable perfect can be defined. In the south-west, an area with three languages has the AOR:IMPF opposition and PROG, in the north an area with eight languages has the categories PRET and PROG and a weakly grammaticalized FUT, and in the north-west an area with four languages has PRET and a highly grammaticalized PROG. In the north-east, an area with the category GPST, the PFV:IPFV opposition and absence of PROG can be identified. The last area shown on Map 9, covering the centre of the area investigated, has a status different from the aforementioned five areas. It is the area already identified ш section 3.2 (see Map 3), i.e., the area with the present anterior > past development, should be noted that this development is the only feature common to all languages о this area, which means that there are more differences between these languages than

On the areal distribution oftome-aspectcategories in Europe

299

1 ANT:PRET, Pefect, highly grammaticalized PROG 2 ANT:PRET, Perfect, PROG, weakly grammaticalized FUT 3 GPST, PFV:IPFV, no PROG 4 present anterior > past development in different stages 5 AOR:EvIPF, Perfect, PROG 6 AORMPF, Perfect, no PROG Map 9. Major areas of tense-aspect systems

the given picture suggests. Thus, French, Italian, Serbo-Croatian, Romanian, and Albanian belong to the languages with an AOR:IMPF opposition, German and Upper Sorbian have PRET, and Hungarian has neither. Upper Sorbian and Serbo-Croatian have the PFV:IPFV opposition, the others don't, and German, French, Italian, and Yiddish have PROG, whereas the other languages lack this category. With regard to the criteria given on Map 9, only four languages cannot be assigned to one area. Basque differs from Area 5 on Map 9 in that it has no PROG and Maltese ш that is has not the PFV:IPVF opposition. Lithuanian differs from Area 2 in that it «as a fully grammaticalized FUT and the PFV.IPFV opposition, and Karaim29 differs fom Area 3 in that it has neither the PFV:IPFV opposition, nor GPST. Note finally that, of course, the languages of each area still differ in various re­ a c t s . Only with regard to the features indicated do they belong to the same area. closer investigation of the tense-aspect categories and an inclusion of the mood

300

RolfThieroff

categories would possibly lead to other, probably more numerous and smaller, areas, but this is a task which has to be left to the post-EUROTYP era.

Notes 1. In Southern Germany as well as in Northern Italy a distinction has to be made between 1) the dialects (called "vernaculars" elsewhere in this volume) spoken in these regions and 2) the regional varieties of the standard language. Note that "Southern German" (SGrm) and "Northern Italian" (Nit) stand for the dialects of these regions, not for the local varieties of the standard language. 2. Cf. also Dahl (this volume a). For descriptions of languages not included in this chapter see Johanson (this volume). 3. Cf. also Dahl (1985: 67): "[...] the English sentence He would have been swimming contains at least the following TMA markers: the auxiliary will, the suffix -d, the auxiliary have (in combination with a past participle) and the copula be (in combination with an -ing form)." 4. Similarly, Squartini & Bertinetto (this volume) describe idiosyncrasies of the present anterior in Portuguese, which do not hold for the category anterior when combined with other tense aspect categories. 5. There is also a small class of about 16 verbs with synthetic forms (Haase 1994: 283-287). These verbs are neglected in the following discussion. 6. For the inclusion of non-finite verb forms, see Thieroff (1994b: 129-131). 7. For the (non-)obligatoriness of FUT in the Germanic languages, see 4 and Dahl (this volume b). For other future constructions than the ones displayed in Table 6, see Dahl (this volume b). 8. For Bulgarian and Serbo-Croatian, a second categorization of perfectivity has to be added (see 5.1) 9. For a discussion whether the aorist with non-past time reference is to be regarded as a subjunctive, see Mackridge (1985: 102ff.). 10. Note that the northern Italian dialects differ from the remaining languages of the area by having neither the ANT:PRET nor the ANT:AOR:IMPF system; see below. 11. Since there is no preterite in Southern German, the same form probably would have to be categorized as one modal category, e.g., "conditional". 12. Note the difference between the stages of development of present anteriors shown m Table 8 and the four stages in the Romance languages proposed by Harris (1982), referred to by Squartini & Bertinetto (this volume). Harris' stage I lies chronologically before stage 0 in Table 8: in this stage, havelbe + past participle is still a resultative and not yet a present anterior. Harris' stage II represents a special development of this form (in Europe restricted to Portuguese), which is not part of the "aonstic drift", as is argued by Squartini & Bertinetto. Thus it is Harris' stage III which corresponds to stage 0 in Table 8, whereas stages 1 and 2 in Table 8 correspond to Harns stage IV. Stage 3 in Table 8 is not covered by Harris, since in this stage the aorist ha

On the areal distribution of tense-aspect categories in Europe

13. 14. 15. 16.

17. 18. 19.

20. 21.

22. 23. 24.

25. 26.

30J

completely vanished. Finally, stage 4 in Table 8 is not reached by any Romance language. In addition to the languages displayed on Map 4, Lindstedt mentions North German and South Italian dialects; these are not taken into consideration in the present chapter. In Polish, the pluperfect "is extremely rare and is found only in a formal literary style" (Stone 1987: 363). In Kotyczka (1987) the pluperfect is not even mentioned. However in Corbett (1987: 404) an anterior + past pluperfect is still mentioned. In Maltese there are two devices for the expression of future time reference, the auxiliary ikun and the particle sa, which function differently with different verb classes (cf. Ebert this volume b). Both are regarded here as markers of FUT, though ikun also has other functions. For details see Ebert, this volume b, Fabri (1995) and Thieroff (1995a: 3 1 33). For a discussion of cases with an obligatory FUT in German, see Thieroff (1992: 125— 128). Cf. also Feuillet (1983: 74). This is not entirely clear in Thieroff (1994a). There, AOR and IMPF were treated as if they were composed of two categories, remoteness + perfective and remoteness + imperfective respectively. However, this is neither in conformity with the semantics of AOR and IMPF, nor with their morphology. Note that this criterion also holds for the difference between the AOR:IMPF and the PFV:IPFV oppositions. See above. Distinguishing the categories AOR and IMPF on the one hand from the categories PFV and IPFV on the other hand, we can say more about the marking of these categories: AOR and IMPF are usually marked morphologically, PFV and IPFV derivationally, by what Bybee & Dahl call "bounders" (see above). In Johanson (this volume), habitual constructions are not even regarded as belonging to "viewpoint aspects". Question 20 of Dahl's questionnaire: "What your brother usually DO after breakfast last summer?" The verb in capitals stands for the uninflected lexeme. Pflegen, together with drohen 'threaten' (Die Wand drohte einzustürzen 'the wall threatened to collapse), scheinen 'seem' {Karl scheint zu schlafen 'Karls seems to be asleep') and versprechen 'promise' (Der Sommer verspricht schön zu werden 'The summer promises to be beautiful') belongs to a small class of verbs which syntactically behave like modal auxiliaries (dürfen 'may', können 'can', müssen 'must' etc.), with the only difference that they take the infinitive with the particle zu 'to' (Erpflegt zu kommen 'He usually comes' vs. Er darf kommen 'He may come'). For an exhaustive discussion of the syntax of pflegen, drohen, scheinen, versprechen (called "Halbmodalverben", 'semi-modal auxiliaries' by the author), see Eisenberg (1994: 382-385). Bertinetto (1996) comes to the conclusion that the English used to construction is not really a habitual, but his arguments cannot be discussed here. For this reason, Dahl labels the English used to construction as HABPAST, which "is used as a label for categories which are mainly used for habitual sentences with past time reference and are not analysable as consisting of HAB ... combined with a regular past tense" (Dahl 1985: 100), whereas the Czech iterative verb is labelled "HAB".

302

Rolf Thieroff

27. For further details, see Ö Siadhail (1989: 177f), 6 Baoill (1994), Thieroff (1994a: 1720). 28. Note that the Northern Italian dialects differ from the remaining languages of the AOR: IMPF-area in that they do not have the category ANT. 29. For a full description of the tense-aspect system of Karaim, see Csato (this volume).

References Abondolo, Daniel 1987 "Hungarian", in: Bernard Comrie (ed.), 577-592. Aronson, Howard I. 1985 "Oiaspect in Yiddish", General Linguistics 25: 171-188. Bânhidi, Zoltân, Zoltân Jôkay & Dénes Szabô 1975 Lehrbuch der ungarischen Sprache. Munich: Hueber. Bertinetto, Pier Marco 1992 "Lestratture tempo-aspettuali dell'italano e dell'inglese a confronto", in: L'Europa linpistica: contatti, contrasti, affinità di lingue. Società di linguistica itaüana. SLI30. Rone: Bulzoni. 1996 "Ltperifrasi abituali in italiano ed in inglese", Studi Orientali e Linguistici 6, 1995/96: 117-133. Ulis volume "The progressive in Romance, as compared with English". Bertinetto, Pier Marco & Denis Delfitto this volume "Aspect vs. Actionality. Why they should be kept apart". Bertinetto, Pier Marco, Valentina Bianchi, Osten Dahl & Mario Squartini (eds.) 1995 Temporal Reference, Aspect and Actionality. Vol. 2: Typological Perspectives. Torino: Rosenberg & Sellier. Beyrer, Arthur, KlatijBochmann & Siegfried Bronsert 1987 Grammatik der rumänischen Sprache der Gegenwart. Leipzig: VEB Verlag Enzyklopädie. Binnick, Robert I. 1991 71« and the Verb. A Guide to Tense and Aspect. New York & Oxford: Oxford University Press. Bonner, Maria 1995 "Zum Tempussystem des Isländischen", in: Rolf Thieroff (ed.), 37-58. Buchholz, Oda & Wilfried Fiedler 1987 Ahnische Grammatik. Leipzig: Verlag Enzyklopädie. Bybee, Joan L. 1985 Morphology. A study of the relation between meaning and form. Amsterdam & Philadelphia.' Benjamins. Bybee, Joan L. & Osten Dahl 1989 "The Creation of Tense and Aspect Systems in the Languages of the World", Studies m Language 13: 51-103. Comrie, Bernard 1976 Aspect. An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Comrie, Bernard (ed.) 1987 Tk World's Major Languages. London & Sydney: Croom Helm. Corbett, Greville 1987 "Serbo-Croat", in: Bernard Comrie (ed.), 391^409.

/ On the aieal distritaüo* ofteese-aspectcategories in Europe

303

Csatö, Éva Agnes 1994 'Tense and Actionality in Hungarian", in: Rolf Thieroff & Joachim Ballweg (eds.), 231-246. this volume "Some typological features of the viewpoint aspect and tense system in spoken Northwestern Karaim". Dahl, Osten 1985 Tense and Aspect Systems. Oxford & New York: Basil Blackwell. ', this volume a 'The tense-aspect systems of European languages in a typological perspective". this volume b'The grammar of future time reference in European languages". I Dahl, Osten & Eva Hedin this volume "Current relevance and event reference". Dardano, Maurizio & Pietro Trifone 1985 La lingua italiana. Morfologia. Sintassi. Fonologia. Formazione delle parole. Lessico. Nozioni di linguistica e sociolinguistica. Bologna: Zanichelli. Ebert, Karen H. 1996 "Progressive aspect in German and Dutch", Interdisciplinary Journalfor Germanic Linguistics and Semiotic Analysis 1: 43-62. this volume a "Progressives in Germanic languages", this volume b "Aspect in Maltese". Eisenberg, Peter 1994 Grundriß der deutschen Grammatik. 3rd edition, Stuttgart: Metzler. Fabri, Ray 1995 "The tense and aspect system of Maltese", in: Rolf Thieroff (ed.), 327-343. Faßke, Helmut 1981 Grammatik der obersorbischen Schriftsprache der Gegenwart. Morphologie. Bautzen: Domowina. Fenn, Peter 1987 A semantic and pragmatic examination of the English perfect. Tübingen: Narr. Feuillet, Jack 1983 "Methodologische Probleme des Aspekts", in: Norbert Reiter (ed.). Ziele und Wege der Balkanlinguistik. Wiesbaden: Harrassowitz. 66-80. Fromm, Hans 1982 Finnische Grammatik. Heidelberg: Winter. Groot, Casper de 1989 Predicate structure in a Functional Grammar of Hungarian. (Functional Grammar Series 11.) Dordrecht: Foris. Gvozdanovic, Jadranka 1995 "Western South Slavic tenses in a typological perspective", in: Rolf Thieroff (ed.), 181— 193. Haase, Martin 1994 "Tense and aspect in Basque", in: Rolf Thieroff & Joachim Ballweg (eds.), 279-292. 1995 'Tense, aspect and mood in Romanian", in: Rolf Thieroff (ed.), 135-152. Hakulinen, Lauri 1957 Handbuch der finnischen Sprache. ]. Band. Wiesbaden: Harrassowitz. Harris, Martin B. "82 "The 'past simple' and 'present perfect' in Romance", in: Martin B. Harris & Nigel Vincent, Studies in the Romance verb. London: Croom Helm. Harris, Martin & Nigel Vincent (eds.) "88 The Romance Languages. London: Routledge. Hedin, Eva 1995 'The tense aspect system of Modern Greek", in: Rolf Thieroff (ed.), 233-252.

i

304

RolfThieroff

=*««;••,?{;;.'

im^M'i>G _

Hundertmark-Santos Martins, Maria Teresa 1982 Portugiesische Grammatik. Tübingen: Niemeyer. Johanson, Lars 1994 "Türkeitürkische Aspektotempora", in: Rolf Thieroff & Joachim Ballweg (eds.), 247266. this volume "Viewpoint operators in European languages". Karlsson, Fred 1983 Finnische Grammatik. Hamburg: Buske. Kiefer, Ulrike 1994 "Die Tempusformen im Jiddischen", in: RolfThieroff & Joachim Ballweg (eds.), 135148. Klein, Wolfgang 1992 "The present perfect puzzle", Language 68: 525-552. 1994 Time in language. London & New York: Routledge. Kortmann, Bernd 1995 "Compositionality and the perfect", in: Wolfgang Riehle & Hugo Keiper (eds.). Anglistentag 1994 Graz. Proceedings. Volume XVI. Tübingen: Niemeyer. 183-199. Kotyczka, Josef 1987 Kurze polnische Sprachlehre. Berlin: Volk und Wissen. Kress, Bruno 1982 Isländische Grammatik. München: Hueber. Kucera, Henry 1981 "Aspect, markedness, and trj", in: Philip Tedeschi & Annie Zaenen (eds.). Syntax and Semantics 14: Tense and Aspect. New York: Academic Press, 177-189. Lindstedt, Jouko 1995 "Understanding perfectivity - understanding bounds", in: Bertinetto et al. (eds.), 95103. this volume "The perfect - aspectual, temporal and evidential". Lötzsch, Ronald 1995 "Das sorbische Tempussystem", in: RolfThieroff (ed.), 167-179. Mackridge, Peter 1985 The Modern Greek Language. A Descriptive Analysis of Standard Modern Greek. Oxford: Clarendon. Mallinson, Graham 1988 "Rumanian", in: Martin Harris & Nigel Vincent, (eds.), 391-419. Metslang, Helle 1994 Temporal relations in the predicate and the grammatical system of Estonian and Finnish. (Oulun yliopiston Suomen ja Saamen kielen laitoksen tutkimusraportteja 39). Oulu: Oulun Yliopisto. Metslang, Helle & Hannu Tommola 1995 "Zum Tempussystem des Estnischen", in: Rolf Thieroff (ed.), 299-326. Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago: The University of Chicago Press. Ö Baoill, Donall P. 1994 'Tense and aspect in Modern Irish", in: RolfThieroff & Joachim Ballweg (eds.), 201216. Ö Siadhail, Mfcheâl 1989 Modern Irish. Grammatical Structure and Dialectal Variation. Cambridge etc.: Cambridge University Press. Olli, John B. 1958 Fundamentals of Finnish Grammar. New York.

On the areal distribution of (вше-aspect categories in Europe

305

Parkinson, Stephen 1988 "Portuguese", in: Martin Harris & Nigel Vincent (eds.), 131-169. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik 1985 A comprehensive grammar of the English language. London & New York: Longman. Render, Peter 1991 "Das Serbokroatische", in: Peter Rehder (ed.), Einführung in die slavischen Sprachen. Darmstadt: Wissenschaftliche Buchgesellschaft, 46-60. Reichenbach, Hans 1947 Elements of symbolic logic. New York: The Free Press & London: Collier-Macmillan. Slizienè, Nijolé 1995 'The Tense System in Lithuanian", in: Rolf Thieroff (ed.), 215-232. Squartini, Mario & Pier Marco Bertinetto this volume "The Simple and Compound Past in Romance languages". Stone, Gerald 1987 "Polish", in: Bernard Comrie (ed.), 348-366. Stump, Gregory 1985 The semantic variability of absolute constructions. Dordrecht: Reidel. Thieroff, Rolf 1992 Das finite Verb im Deutschen. Tempus-Modus-Distanz. (Studien zur deutschen Grammatik 40) Tübingen: Narr. 1994a "Inherent Verb Categories and Categorizations in European Languages", in: Rolf Thieroff & Joachim Ballweg (eds.), 3^15. 1994b "Das Tempussystem des Deutschen", in: Rolf Thieroff & Joachim Ballweg (eds.), 119134. 1995a "More on Inherent Verb Categories in European Languages", in: Rolf Thieroff (ed.), 1-36. Thieroff, Rolf & Monika Budde 1995 "Are Tense and Aspect Categories?", in: Bertinetto et al. (eds.), 47-62. Thieroff, Rolf (ed.) 1995b Tense Systems in European Languages П. (Linguistische Arbeiten 338) Tübingen: Niemeyer. Thieroff, Rolf & Joachim Ballweg (eds.) 1994 Tense Systems in European Languages. (Linguistische Arbeiten 308) Tübingen: Niemeyer. Tommola, Hannu 1992a "The Marking of Future Time Reference in Estonian", in: Osten Dahl, Casper de Groot, & Hannu Tommola (eds.), Future Time Reference in European Languages I. Bulgarian, Estonian, German, Hungarian, Continental Scandinavian, ZüritüUtsch. (EUROTYP Working Papers VI:2.) Stockholm. 12-28. 1992b "The Marking of Future Time Reference in Finnish", in: Osten Dahl, Casper de Groot, & Hannu Tommola (eds.), Future Time Reference in European Languages II. Dutch, Finnish, Modern Greek, Italian, Lezgian, East Slavic, Turkish. (EUROTYP Working Papers VI:3.) Stockholm. 12-28. 1994 "Zum Tempus im Finnischen", in: Rolf Thieroff & Joachim Ballweg (eds.), 219-229. this volume "Progressive aspect in Baltic-Finnic". Wheeler, Max W. ^88 "Catalan", in: Martin Harris & Nigel Vincent (eds.), 170-208.

Future Time Reference

V

•-

,tl*
gonna). In Dahl (1985), the constructions mentioned here were tentatively subsumed under a cross-linguistic gram-type 'prospective' (cf. also Comrie 1976 for a similar treatment). The evidence for the existence of a such a gram-type as distinct from early futures in general is somewhat shaky, however. The de-andative constructions seems to have developed relatively late. In English, 11 spread in the 17th century. S 5

' -

Germanic de-obligative construction (SHALL)

°gnates of the English auxiliary shall (with the original meaning 'to owe', henceh SHALL) are found in most Germanic languages and are or have been used

0rt

320

Osten Dahl

as future-marking devices over a rather extensive part of the Germanic territory, including earlier stages of High German, where sollen was used more frequently than it is today. According to OED, sceal was used already in Old English to express prophesies and the like. There are clear differences in the domain of use between the Germanic languages, however: it is only in English and Dutch (including Afrikaans) that SHALL is used for prediction-based FTR, although it should be added that in English, this use is rather restricted since shall is mainly used in the 1st person. In the Scandinavian languages, SHALL is restricted to intention-based FTR and obligational meanings closer to its original sense.

5.6. De-venitive constructions Under these headings, we treat constructions involving verbs with the meaning 'come'. These include two gram families - one in Scandinavia and one mainly comprising a number of Romansh dialects - which exhibit striking similarities and seem to have arisen roughly at the same time but which still cannot be assumed to be related in view of the geographical distance between them. The evidence suggests that these constructions may be the result of a path of development that has not been properly described in the literature on grammaticalization. This motivates discussing them in somewhat more detail. 5.6.1.

Scandinavian

This gram family comprises Continental Scandinavian - Danish, Norwegian, and Swedish - and (somewhat marginally) Finnish. Its rise and spread appears to be relatively recent - the first attested examples in Swedish go back to the 17th century. The original form of the construction in Scandinavian (preserved in Norwegian and Danish and attested in older Swedish) is

(14)

kommer til(l) at(t) come:INF to INFM

In Swedish, the preposition till has been dropped, and there is a tendency in th< present-day language to also drop the infinitive marker. In Finnish, the verb tulla is combined with the Illative case of the so-called 3rd Infinitive (suffix -maari). Tn construction has been looked upon with some suspicion as being a Swedish calque-

The grammar of future time reference in European language! 5.6.2.

321

Swiss

In the Romansh dialects spoken in Switzerland, there exists a construction which is formally and perhaps also semantically very similar to the Scandinavian de-venitive (Ebneter 1973). The earliest attested examples are from the 16th century. The construction now seems receding but is still preserved in a number of dialects. An example from Surmiran (surmeirisch): (15)

Romansh/Surmiran Té vinst arjk a vçkr s i va la razurj. you come:PRS:2SG yet to see:INF that I have DEF right 'You will yet see that I am right'

In some dialects, this construction appears mainly in examples of the type 'It's going to rain', e.g., Oberengadin: (16)

Romansh/Oberengadin a veil a plçver. it come:PRS to rain:INF 'There is rain on the way / it will rain.

Analogues to this are also found in Swiss German (Ebneter 1973: 242, Bickel 1992): (17)

Züritüütsch Es chunt cho rägnen. it come SP rain 'There is rain on the way / it will rain.'

These de-venitive constructions are somewhat peculiar among future-referring grams in that they are primarily used for prediction-based rather than intention-based FTR. It is therefore of some interest to consider how they have come about, and we shall now look at the source that has been proposed in the literature (see, for example, Ebneter 1973). In many Germanic languages, there is a construction which is formally quite similar but whose semantics is distinct, as shown by the following English example: (18)

I came to hate him.

(18) could perhaps be best paraphrased as 'By various causes I was led to hate him'. What is notable is thus that it expresses something that it is not under the control 01 the subject, in other words, a non-volitional process. This suggests that the construction might relatively easily be extended, when used with future time reference, 0 express predictions in general. Cf. (19) as an example that could serve as an ln termediate step towards such an extension.

322 ( 19)

Osten Dahl

ж«

кЯ

You will come to hate him.

If this account is correct, it creates a problem for the claim made by Bybee, Pagliuca â Perkins (1994: 270) that "all modal and movement future sources begin with human agents and move from the expression of the intentions of that agent to the expression of prediction".4 At any rate, there is no evidence to suggest that the Germanic de-venitives ever expressed intention. (In that case, we would have to assume that it has now lost this use, which would also create difficulties for the theory.) Rather, we have to conclude that we are dealing with a separate grammaticalization path, which in the attested cases ends in a gram with a focus in prediction-based FTR. Whether further developments from this point are possible is something that only coming research can decide. nw .

5.7. North European de-volitive construction (WILL) Descendants of the Proto-Germanic willan 'want', henceforth WILL, are used as FTR markers in a restricted number of Germanic languages, mainly in the North Sea area: English {will), Danish and Norwegian Bokmâl (vil), Faroese (vil), Frisian (wal) and Yiddish (vet). The area may earlier have also included High German. In English, the use of WILL for both intention-based and prediction-based FTR was well established already in the Old English (Anglo-Saxon) period, and has (in combination with SHALL) become what is probably the most grammaticalized future marker in the Germanic languages. In Scandinavian, WILL is much more marginal - it seems to have spread later and never reached Swedish. Yiddish is a special case in that WILL seems to have fused with werden (first person forms such as ixvel T will' are supposedly derived from willan while other forms such as er vet 'hewill' seem to come from werden, although a general analogy to the regular verb paradigm is not excluded).

5.8, Circum-Baltic 'become' In Modern High German, the most common FTR device is the construction werden + Infinitive. Werden is identical to the verb for 'become'. The details of the origin of this construction are somewhat controversial. In dialects and older stages of the standard language, there is an alternative construction, where the main verb has the form of a Present Participle, and according to a widespread theory, this is the original variety. This hypothesis is rejected in the detailed study by Saltveit (1962), where it is found that both constructions already existed in Old High German, although the one with the Infinitive did not become frequent until the end of the Middle Ages. According to Saltveit, there was a semantic difference between the two constructions,

The grammar of future time reference in European languages

323

in that the participial one had an inchoative interpretation and the infinitival one a modal interpretation. The Infinitive construction has not become rooted in all dialects in the High German area - for instance, it is still felt to be alien in Swiss German (Bickel 1992). In Yiddish, on the other hand, it is one of the sources for the fused future paradigm. The Estonian saama construction, which is a slightly marginal FTR device, is regarded as a caique on the German werden future.

5.9. Slavic perfective present In the West and East Slavic languages, the non-past forms (Present tense) of perfective verbs are primarily used with present time reference. In general, they cannot be used to refer to events that take place at the moment of speech. For a discussion of the origin of this state-of-affairs, see below.

5.10. Balkan de-volitive construction Most languages in the Balkan area have a construction derived from a verb meaning 'want'. There are two different types, however: one in which the marker is an uninflected particle, and one where it is an inflected auxiliary. The first type appears in Modern Greek (tha + subjunctive), Bulgarian (ste), Macedonian (ke) and Albanian (do + subjunctive). The second is found in Romanian (voi) and Serbian/Croatian (eu, ces, ée). This gram family is one of the classical examples of a Sprachbund phenomenon, covering four different branches of Indo-European.

5.11. Balkan 'have' future In some Balkan languages - Bulgarian (ima da), Gheg Albanian (кат те + infinitive) and Romanian (am sä) - there is an alternative future construction based on the verb 'have', thus most probably an original obligational construction. Its range of uses in the different languages is not quite clear from the sources. The Bulgarian construction is different from the others syntactically in that ima does not agree with the subject. The Ukrainian inflectional imperfective future may also belong to this gram fam%, although it has gone further in grammaticalization and also differs aspectually from the others.

324

Osten Dahl

иаТ

v

5.12. Slavic copular constructions These are combinations of a copula and a participle or the like used as FTR devices. The clearest cases are in Slavic: (i) the imperfective futures formed with the stem bçd-lbud- and an /-participle or infinitive in North (West and East) Slavic, and (ii) the Slovenian future construction bo- + /-participle. The North Slavic constructions are relatively late in origin. The infinitive construction seems to have spread from the Czech area at the end of the 13th century, shows up in Polish, Belarusan, and Ukrainian at the end of the 14th century and in Russian in the 15th. Although this rather nice chronology strongly suggests an areal development, Knzkovâ (1960), from whose monograph these dates are taken, seems rather skeptical about all hypotheses implying areal influences with regard to the rise of the North Slavic imperfective futures. She also takes a skeptical stance against linking this construction with the German werden future.

5.13.

'Take, seize'

There are at least two cases of FTR markers which are derived from verbs with meanings like 'seize' and 'take' viz. Hungarian/og ( fut t
sitB

''

The relation between the two situations in the first case is a temporal one, orienting them to the same future time, whereas the relation represented in the second figure is aspectual, expressing order (cf. section 2 above).

4. Future-marking and imperfectivity We have seen above that imperfective verb phrases are typically (or even obligato­ rily) future-marked in temporal FTR-clauses. This should probably be seen against the background of the function of the imperfective aspect. When an imperfective verb phrase is used to refer to a future situation, it normally needs some reference time, more or less explicit, a time point or a time frame to which it may be related. This could be why the Imperfective almost automatically gets the future-marking in the subordinate. On the other hand, it also seems to be the case that the imperfec­ tive (including FUT PF) is comparatively frequent in the main clause governing a future-marked temporal clause. See, e.g., FTRQ: 17 and 26:

344

Eva Hedin

(FTRQ: 17) Ötan tha jirisis tha éxo ghrâpsi. when FUT return:PFV:2SG FUT have:IPFV:lSG writerPFF afto to ghrâma this DEF letter 'When you return, I will have written this letter.' (FTRQ: 26) Ötan tha ftâsi to fajitô tha me krio. when FUT arrive-.PFV DEF food FUT be:IPFV cold 'When he arrives, the food will be cold.' In the material only 6 of 39 subordinates with a PFV SUBJ were linked to a main clause with a verb phrase in the imperfective aspect. Of 10 PFV FUT clauses 4 were subordinated to a main clause with an imperfective verb phrase. For the IPFV FUT / FUT PF clauses the figure was 7 of 10 and for PRS/PF 5 of 10. Even if the material is too limited to allow any conclusions on the basis of these figures, at least it seems that imperfective aspect in the main clause is less frequent in combination with the PFV SUBJ than with the FUT in the subordinate. Reference to a future state (in a wide sense) in the main clause thus seems to favour the future-marking of the subordinate. This, too, can be interpreted as support for the hypothesis that future-marking in the subordinate has the function of establishing some future reference time. One reason for future-marking the temporal clause when referring to a state in the main clause could be namely the need of a reference time for the state of affairs described. Consider the following example: (30)

When I come home my mother will be happy.

The most natural interpretation is that the mother will become happy as a result of the speaker's coming home. Her happiness is directly related to the arrival of her child. Another possible interpretation, however, is that it so happens that the mother is (already) happy (for some other reason) at the time of the speaker's arrival. In the second case, the subordinate gets the function of reference time for the state described in the main clause. Compare FTRQ: 17 and 18 where the Greek informant chose the PFV FUT in the first but not in the second. In both cases the main clause has a Future Perfect. (FTRQ: 17) [Talking to someone who is leaving in a while:] When you RETURN, I WRITE this letter(= I FINISH it already at that time). (FTRQ: 18) [Said as an order by a teacher leaving the classroom:] When I RETURN, you WRITE this assignment (=You FINISH it by then).

Future marking in conditional and temporal clauses in Greek

345

A possible reason for this choice could be that in the first sentence the reference time is more obvious than in the second. (Note the instruction = I FINISH it already at that time.) In the second the teacher tells the class to finish (have finished) the work before he comes back, but does not as obviously refer to the future state of the work already being done. It is interesting to note that a similar differentiation was made by one of the French informants who chose the Future in FTRQ: 17 (with the Future Perfect in the main clause), but the Present in FTRQ: 18: (FTRQ: 17) Quand tu reviendras y'aurai fini d'écrire cette lettre. (FTRQ: 18) Quand je reviens, vous devez avoir fini ce travail.

5. Causal/conditional interpretations of 'when'-clauses The semantics of the temporal conjunction meaning 'when' in many languages seems to allow for a causal interpretation in some contexts, particularly with perfective verbs. The degree to which one is inclined to give the temporal clause a causal meaning may well be due to different pragmatic factors. For a sentence like (31) it is not easy to find a direct causal interpretation whereas (32) is equally hard to give a purely temporal interpretation. (31)

When the sun sets we will stop working.

(32)

When the sun sets it'll get dark.

As has already been mentioned, in some contexts as in (30) both interpretations are possible. Greek, however, seems to have the possibility of excluding a causal interpretation when the interpretation is not obvious from contextual and pragmatic factors, by future-marking the temporal clause. Consider the following example: (33)

Ötan érthi о Jânis tha îmaste dhéka when come:PFV:SUBJ DEF Jannis FUT be:IPFV:lPL ten edho mésa. here inside 'When Jannis comes we will be ten in here.'

One interpretation of (33) is thus: 'When Jannis comes, we will be (become) ten persons in the room (as a consequence of his arrival)'. However, another possible interpretation is the following: 'At Jannis' arrival (at the time point when he arrives) we will (already) be ten persons in the room', 'At his arrival Jannis will find ten persons in the room'. If the 'when'-clause is future-marked, however, it seems that the sentence more easily gets the second interpretation.

1

346 (34)

Eva Hedin Ötan tha érthi o Jânis tha imaste dhéka when FUT come:PFV DEF Jannis FUT be:IPFV:lPL ten edhö mésa. here inside 'When (by the time) Jannis comes we will be ten in here.'

Let us consider the two "types of simultaneity" characteristic of the two kinds of 'when'-clauses discussed above. The future-marked 'when'-clauses were said to refer to some future time functioning as reference time (A and В at t) whereas the nonmarked clauses express order ( В at A). While the first (temporal) relation could be called contingent the second (aspectual) shows some kind of necessity or implica­ tion; В is temporally determined by A itself: When A, then В (whenever that is). The interpretation of this "temporal implication" (when-then) as a "logical implication" (if-then) is natural (e.g. 35), but not necessary (e.g. 36): (35)

Otan tin akusis tha jelâsis. when it:ACC hear:PFV:SUBJ:2SG FUT laugh:PFV:2SG 'When you hear it you'll laugh.'

(36)

Otan siköso to xeri thâ arxisis, étsi? when lift:PFV:SUBJ:lSG DEF arm FUT start:PFV:2SG so 'When I lift my arm you'll start, okay?'

6. Modal interpretations Setatos has pointed out (1985: 180-181) that for Greek, there seems to be a modal difference between the future-marked conditional and temporal clauses and those that are not marked for future. The future-marked clauses seem to convey a feeling of stronger certainty: He describes both the temporal and the conditional clauses with the subjunctive as expressing something "possible, if he ever comes", and the correspondent clauses with the Future as something "natural, as we expect". For the conditional clauses the 'if it is the case' element of the future-marked clauses referred to above (in Section 2) may perhaps contribute to such an interpretation, i.e., the implicit reference to the truth of some prediction would make the conditional clause more assertive than it normally is. As for the temporal clauses, the difference in modal interpretation could perhaps be conveyed by the differences in temporal reference in the two cases. Reference to a future situation as occurring at some specific time ('at time t when') may convey a feeling of stronger certainty as for its occurring, i.e., the temporal definiteness may be understood as a modal definiteness. Consider the following two examples:

Future marking in conditional and temporal clauses in Greek (37) a.

b.

Ôtan érthi when come:PFV:SUBJ po. say:PFV:lSG Otan tha érthi when FUT come:PFV po. say:PFV:lSG 'When Daddy comes I'll

347

о patéras tha tu to DEF father FUT hinr.GEN if.ACC

о patéras tha tu to DEF father FUT him:GEN it.ACC

tell him.'

In both cases we do in fact - by using the conjunction ôtan 'when' in this context (instead of 'if e.g.) - implicitly assert that we expect (presuppose) that Father will come at some more or less definite occasion. Whether we use the Perfective Subjunctive or the Perfective Future, we express some expectation that Father will come. However, only when we use the Future do we anticipate a specific future time for the arrival. With the Subjunctive the arrival is not defined temporally. But the indefiniteness is actually concerned only with the temporal reference. In conditional as well as in temporal clauses, the Future as indicative and temporal naturally gives a different modal status to the utterance compared to the nontemporal Subjunctive.

7. Conclusion As was noted in the introduction, future-marking in conditional clauses and in temporal clauses in Greek cannot be described in the same way. Temporal clauses seem to be more easily marked for future than conditional clauses. Specific temporal reference in the temporal subordinate seems to favour future-marking. In both cases, however, when the subordinate clause is marked for future there seems to be some weaker dependency between this clause and the main clause. This is most typically observed in the obligatorily future-marked cases where the 'when'-clause is a nonrestrictive relative clause. Such cases, where the subordinate clause is less dependent and has a more assertive function, could be a possible starting point for a future gram to enter this kind of context from which it may expand further in a grammaticalization process to other temporal and conditional clauses.

Acknowledgment Financial support for the work presented has been received from the Swedish Council tor Research in the Humanities and Social Sciences (HSFR).

348

Eva Hedin

ь.й

Notes 1. The Modern Greek Subjunctive is a category that has raised some controversy. Morpho­ logically, it exists only in the perfective aspect (for instance, tha ghrâpso), whereas the Imperfective Subjunctive has to be syntactically defined. Compare thélo na ghrâpso 'I want to write (PFV)' and thélo na ghrâfo 'I want to write (IPFV)', where the form ghrâfo is morphologically identical with the Present Indicative ghrâfo 'I write/am writing' and its function as a Subjunctive is marked only by the presence of the particle na (in the glosses: SUB J). For further discussion see Veloudis, I & Philippaki-Warburton 1983. 2. This example was pointed out to me by loannis Veloudis, Thessaloniki, as a counterexample to my somewhat simplified view at the time of the temporal order relation discussed here as one of sequence (which is typically but not necessarily true). 3. PFF = Aparemfato or Perfect formant, which combines with the verb éxo to form the Greek Perfect periphrasis. 4. Note also that the main clause has imperfective aspect. Cf. 4 below. 5. Not all subordinates were followed by a main clause and some verbs have the same form in the perfective and the imperfective. This is why the figures do not correspond to those in Table 2 above.

inn

References

Blass, Friedrich & Albert Debranner 1970 Grammatik des neutestamentlichen Griechisch. (13th edition.) Göttingen: Vandenhoeck & Ruprecht. Bybee, Joan L. & Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51-103. Comrie, Bernard 1982 "Future time reference in the conditional protasis", Australian Journal of Linguistics 2: 143-152. Dahl, Osten 1985 Tense and aspect systems. Oxford and New York: Blackwell. Hedin, Eva 1992 "Present with future time reference in Modern Greek", in: EUROTYP Working Papers, Series VI, no. 1. August 1992. Magnien, Victor 1912 Le futur grec. Paris: Champion. Papazafefri, Ioanna 1992 Lâthi sti xrtsi lis ghlâsas mas [Mistakes in the use of our language]. Athina: Smili. Setatos, Michail 1985 "Tropikötitestorimatosstin kinî neoelinikf ' [Modalities of the verb in Modern Greek], in: Studies in Greek Linguistics, Proceedings of the 6th annual meeting of the Department of Linguistics, Faculty of Philosophy, Aristotelian University of ThessalonikiThessaloniki: Kiriakidis, 175-182.

Future marking in сожШкхт! and temporal clauses in Greek

349

Veloudis, I. & Philippaki-Warburton, I. 1983 "I ipotaktiki sta néa elinikâ" [The Subjunctive Mood in Modern Greek], in: Studies in Greek Linguistics. Proceedings of the 4th Annual Meeting of the Department of Linguistics, Faculty of Philosophy, Aristotelian University of Thessaloniki. Thessaloniki: Kiriakidis, 151-169.

••:>

SM.

Osten Dahl

Verbs of becoming as future copulas

In many languages of Northern Europe, there is a strong tendency to use the present tense of verbs meaning 'become' when talking of states in the future. This issue has been somewhat neglected in the literature, although scattered cursory statements are found (cf., e.g., Saltveit 1962; cf. also the brief discussion in Bybee, Perkins & Pagliuca 1994: 261). In this paper, I shall look at the extent of this phenomenon and try to elucidate its nature. Of course, it seems quite natural that the distinction between being and becoming should blur with respect to the future, since the prototypical situation involves both the state itself and the event that marks its beginning. Thus, the sentences (la) and (lb) are equivalent in most contexts. (1) a. b.

You'll soon be a big boy. You'll soon become a big boy.

In Swedish, the most natural way of rendering the same content would be by simply using the present tense of bli 'become' : (2)

Swedish Du blir snart en stor pojke. you:SG become:PRS soon a big boy 'You will soon be a big boy.'

The use of blir in (2) might be explained by appealing to the general tendency to use the present tense of telic verbs for future time reference in Swedish (as in most other Germanic languages). Other cases are not equally easily dealt with. Consider the following Swedish example, which should be thought of as occurring in the context of a weather forecast: (3)

Swedish Det blir kallt imorgon. it become:PRS cold:NT tomorrow 'It will be cold tomorrow.'

352

Osten Dahl

What is notable in this example is that there is no implication of change. (3) can be used even if it is cold when it is uttered (it sounds even better if ocksa 'too' is added). Notice that in English, It will get/become cold tomorrow is not natural in such a situation, and It will get cold tomorrow too hardly makes any sense. Likewise, in Russian, (4) would unequivocally imply that the weather will change: (4)

Russian Zavtra stanet xolodno. tomorrow become:PF:PRS cold 'It will become cold tomorrow.'

On the other hand, it is not always possible to translate will be as blir. The following is an example: (5)

Swedish Barnen kommer att vara mycket sömniga när pappa i, child:PL come INFM be:INF very sleepy when father •.»-, kommer. come:PRS 'The children will be very sleepy when father arrives.'

In (5), substituting blir for kommer att vara 'will be' changes the interpretation: it would mean that Father's arrival makes the children sleepy, or at least that they become sleepy when he arrives, whereas (5), as it stands, is naturally interpreted as implying that the children are already sleepy when he arrives. Analogously, (6) suggests a causal relationship between our arrival in Gothenburg and the weather: (6)

Swedish Vädret blir fint när vi kommer fram till weathenDEF become:PRS fine when we come PRT to Göteborg. Gothenburg 'The weather will be fine when we arrive in Gothenburg.'

In these examples, then, the use of blir does seem to imply that there is an event which changes a state of affairs. We must therefore look more closely at the examples which do not have this implication, to see what the conditions are. Consider the following Swedish example: (7)

Swedish Den här festen blir nog trevlig. this here party become:PRS surely pleasant 'This party will surely be pleasant.'

1

Verbf of becoming as future copulas

353

This clearly does not mean that the party first is unpleasant and then becomes pleasant. Rather, what it means is that the party will display the quality of being pleasant when it takes place. A possible translation into English would be: (8)

The party will turn out pleasant.

(7) represents what could be seen as a prototypical example of this extended use of bli. An essential feature seems to be that the property concerned is one that the entity in question acquires when it comes into existence or develops. This makes it natural for the subject of such sentences to denote events and similar types of entities. The use of 'become' found in (7) is not restricted to future time reference. Cf. (9), where we are talking of a party that has already taken place. (9)

Swedish Festen blev trevlig. party:DEF become:PST pleasant 'The party became pleasant.'

In other words, what we are here dealing with is, in principle, a special use of verbs of becoming rather than a future tense marker. But there is still a clear link to future time reference. The use of a verb of becoming in an example like (9) relates the state of affairs in question to some observer's previous expectations.1 When talking of the future, the speaker and/or the listener fill the role of this observer. This makes the use particularly natural with future time reference. It may be mentioned here that in Finnish, a formal distinction between the core uses of becoming and the 'turn out' use, in that the predicate of the 'turn out' construction takes the nominative or partitive case rather than the translative. Examples like the following are commonly cited (Setälä 1973: 30, Almqvist 1990)2 (10)

Finnish Puuro(sta) tuli mustaa. porridge(:ELAT) become:PST black:PRTV 'The porridge turned out black, i.e., when it was ready.'

(11)

Finnish Puuro tuli mustaksi. porridge:NOM become:PST black:TRNSL 'The porridge became black, i.e., it changed from some other colour.'

354

Osten Dahl

,«vr'

What about the weather examples, with which we started the discussion? Most of these examples are constructed with a dummy subject, and it seems that the question of coming into existence does not make a lot of sense here. It might be argued that in saying that it will be cold tomorrow, we are really thinking of 'the weather tomorrow' as something that does not yet exist. Sentences expressing location present special problems. In Swedish, one cannot in general use bli in sentences like the following: (12)

Swedish *Blir du pâ Institutionen imorgon? become:PRS you:SG at department:DEF tomorrow 'Will you be at the department tomorrow?.'

Rather, one would use är 'am, are, is' or kommer att vara 'comes to be' : (13)

Swedish Är du / be:PRS you.SG kommer du att vara pâ institutionen imorgon? come:PRS you:SG INFM be:INF on department:DEF tomorrow 'Are you/Will you be at the department tomorrow?'

Likewise, in German, one could not say (14) but would have to add sein, as in (15): (14)

(15)

German Wirst du morgen im Institut? FUTAUX:2SG you tomorrow in_DEF department German Wirst du morgen im Institut sein? FUTAUX:2SG you tomorrow in_DEF department be:INF

This is fairly natural, since Swedish bli and German werden do not in general function as inchoatives in locative constructions: a movement verb like Swedish komma 'come' is used instead. However, there are at least two classes of counterexamples to the claim that bli is excluded from locative constructions in Swedish. One might in fact say: (16)

Swedish Jag blir hemma imorgon. I become :PRS at home tomorrow 'I will be at home tomorrow.'

Veil» of becoming as future copulas

355

A closer look reveals that bli here has its original meaning, 'remain'. In German, where the verb for 'become', werden, does not have this ambiguity, it cannot be used in translating (16). Rather, bleiben 'remain' is used: (17)

German Ich bleibe morgen zu Hause. I become :PRS tomorrow at home 'I will stay at home tomorrow.'

Another type of counterexample came to my attention when I saw the following headline in a German newspaper (about three years before the meeting actually took place, so it was clearly future time reference at the time): (18)

German Die Katholiken-Tagung 1994 ist in Dresden, the Catholic meeting 1994 be:PRS in Dresden 'The 1994 Catholic meeting will be in Dresden.'

This is not possible in Swedish, given the context. A Swedish newspaper would say [what would be ungrammatical in German (Thieroff 1992)]: (19)

Swedish Katolikmötet 1994 blir i Dresden. Catholic meeting:DEF 1994 become:PRS in Dresden 'The 1994 Catholic meeting will be in Dresden.'

We may note that the subject in ( 19) is an event and thus of a type that would typically be expected to occur with the extended use of 'become' verbs already discussed. But there also seem to be pragmatic principles at work. Using är 'is' instead of blir 'becomes' in such a sentence would be much more natural in some other contexts. The crucial feature of (19) seems to be that it reports a decision, as a piece of 'hot news'. If we are talking about something that has been scheduled for a long time, är 'is' is possible: (20)

Swedish Nästa sammanträde är pâ torsdag. next meeting be.PRS on Thursday 'The next meeting will be on Thursday.'

The 'hot news' reading is also possible in a future-in-the-past context, e.g.,

356

(21)

Osten Dahl

Ь < hs"..'

Swedish Mötet blev den 24 maj. meeting:DEF become:PST DEF 24 May '(It was decided) that the meeting will/would be on May 24.'

Notice that (21) can be felicitously uttered both before and after May 24. Further­ more, it is actually ambiguous between the future-in-the past reading and the inter­ pretation 'The meeting took place, as it turned out, on May 24' (cf. above) - another illustration of the complexities of the future-related uses of 'become' verbs. In addition to verbs meaning 'become', similar extended uses may also be found with verbs meaning 'get, receive'. The following is an example from Swedish, which should be interpreted as saying that my children will have fair hair when they are born (or soon after): (22)

Swedish Mina barn kommer nog att fâ / far nog my child:PL come:PRS probably to get 1 get:PRS probably ljust hâr. fair hair 'My children will probably have fair hair.'

Among the Germanic languages, Swiss German should be mentioned as a special case where there are several verbs of change corresponding to become in English, all of which may be used to express future time reference in the ways discussed here. Compare the following examples from Züritüütsch (Bickel 1992: 78): (23)

Züritüütsch Er isch Leerer. he PRS teacher 'He is a teacher.'

(24)

Züritüütsch Er wîîrt Leerer, he FUT teacher 'He will be a teacher.'

(25)

Züritüütsch Es isch ehalt, it be.PRS cold 'It is cold.'

(nominal predicates)

(impersonal adjectival stative predicates)

Vint» of becoming as future copulii

Ц7

(26)

Züritüütsch .

Admittedly, this may be a fairly easily cancelled implicature (cf. 36). It may also be noted that according to many speakers of Hungarian, the literal translation of (5) into that language gives the same impression as the Swedish original, namely that the children become sleepy only after Father has arrived: (38)

Hungarian A gyerekek âlmosak lesznek, amikor az apjuk the child:PL sleepy LESZ:3PL when the father:POSS.3PL hazajön. come_home:PRS:3SG 'The children become/will be sleepy when father arrives.'

Thus, even in Hungarian, the treatment of 'becomes' as a future copula is somewhat problematic. Looking at the geographical distribution of the extended 'become' phenomenon, we find that it coincides largely with what is called the North European "futureless" area in Dahl (this volume). What then about the third language family involved in this area, viz. Slavic? In most Slavic languages, there is a paradigm formed from the stem bqd-, which functions as a future copula. It is used as an auxiliary in the periphrastic imperfective in West and East Slavic, but it is both significantly older and has a wider distribution than this construction. It was current already in Old Church Slavonic and there is some evidence that it had an inchoative interpretation there. In Havrânek (1980: 114) the following example is given: (39)

Old Church Slavonic (Mark 4: 32) Vuzdrastefï i bpdetu bole vïsëx grow:PRS and become:PRS great:COMP all:GEN:PL zelii. herb:GEN:PL 'It groweth up, and becometh greater than all herbs.'

;}

bqdetu here translates Greek ginetai 'becomes'. (Admittedly, 'will be' would not be impossible in this context, so the evidence for bqdetu being a regular inchoative verb is perhaps not totally clear.). The etymology given for example in Havrânek (1980: 114) for forms like bqdetu is that they are 'a typical ingressive present with an n infix' from the Indo-European root s/bhif. If this is correct, it would show that

360

Osten Dahl

the path from 'become' to 'will be' is a possible one, and that it has been manifested within the North European "futureless" area. It is thus possible that even if the Germanic and Baltic-Finnic verbs of becoming cannot be regarded as future copulas, the extensions of their use that we can observe represent the first step in such a grammaticalization path. One thing that remains to be elucidated is the precise relation between the extended 'become' verbs and the rise of periphrastic future constructions involving such a verb as an auxiliary, like the German werden construction (Saltveit 1962) or the North Slavic periphrastic imperfective future.

Notes 1. It may be noted that Romance languages often use the (perfective) Simple Past of the verb 'to be' in similar contexts. Consider the following French example from a corpus of newspaper texts: Les résultats de cette enquête furent surprenants 'The results of this survey were surprising' (Wiberg 1995: 195). 2. According to Almqvist, the elative case (puurosta) rather than the nominative (puuro) is normal in modern written Finnish. The generalization of the elative subject construction in Finnish is in itself an interesting grammaticalization process which is, however, outside the scope of this paper. For a discussion, see Almqvist (1990). 3. In the present, a zero copula is the normal choice in constructions; however, forms of vol- are used, e.g., in locative constructions. 4. The root л/Ыгй- in itself is often said to have 'become' as one interpretation, but this does not appear directly relevant to the issue here.

References

Almqvist, Ingrid 1990 "Varför det är lättare att bli femtio an att bli gammal pâfinska",in: Ingrid Almqvist, Per-Erik Cederholm, Jarmo Lainio (eds.), Fràn Pohjolas pörten till kognitiv kontakt. Vänskrifl till Erling Wände den 9 maj 1990. Stockholm: Department of Finnish, Stockholm University, 15-28. Bickel, Balthasar 1992 "The Marking of Future Time Reference in Züritüütsch". Future Time Reference in European Languages I. EUROTYP Working Papers VI: 2. Bybee, Joan, Revere Perkins & William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. Csatö Johanson, Eva 1992 "The Marking of Future Time Reference in Hungarian", in: Future Time Reference in European Languages I. EUROTYP Working Papers VI: 2. Dahl, Osten this volume "The grammar of future time reference in European languages".

Visit» of becoming as future copules

361

x Havrânek, Bohuslav (ed.) 1980 Etymologicky slovnuc slovanskych jazyku: slova gramatickâ a zajména. Sv. 2. Praha: Academia. Karoly, Sândor 1972 "The grammatical system of Hungarian", in: Lorând Benko & Imre Samu (eds.), The Hungarian Language. The Hague: Mouton. Saltveit, Laurits 1962 Studien zum deutschen Futur. Die Fügung 'werden mit dem Partizip des Präsens' und 'werden mit dem Infinitiv ' in ihrer heutigen Funktion und in ihrer geschichtlichen Entwicklung. Bergen: Norwegian Universities Press. Setälä, Eemil N. 1973 Suomen kielen lauseoppi. Helsinki: Otava. Thieroff, Rolf 1992 "The Marking of Future Time Reference in German". Future Time Reference in European Languages. EUROTYP Working Papers Ш: 2. Tompa, Josef 1968 Ungarische Grammatik. The Hague: Mouton. Wiberg, Lars-Erik 1995 Le passé simple, son emploi dans le discours journalistique. Stockholm: AlmqVMt A Wiksell.

The Perfect

.'.Ox1 fs::;;;,'>i:. ;..-,,-

K.....

.••.>">.'*&•

'-.st

öd?

Jouko Lindstedt

The perfect - aspectual, temporal and evidential 1.

Introduction

The perfect has found its way from grammars of Classical Greek and Latin to those of modern European languages - as a term. But it is usually described as part of language-specific tense and aspect systems; there have not been many attempts to explore its nature as a cross-linguistic category, and it is often not even asked whether the "Perfects" of languages A and В are really manifestations of the same typological feature at all, or only happen to share the same name for obscure historical reasons. The perfect and its development in various European languages was one of the focal research areas of the EUROTYP Tense and Aspect Theme Group. The point of departure was Dahl's (1985: 129-153) important result that a cross-linguistic cate­ gory of perfect can be identified empirically, without a preconceived definition of its semantics. The perfects of various languages centre on certain prototypical examples like the following (Dahl 1985: 131); the uninfected verb should be replaced with a properly inflected verb form in each language under investigation: (1)

[A says: I want to give your brother a book to read, but I don't know which. Is there any of these books that he READ already? - В answers:] (Yes,) he READ this book.

Obviously, the English Perfect (as in He has read this book, or He's read this one) would be a good candidate for an instance of this cross-linguistic gram type (for a definition of the term, see Dahl, this volume). Material about the perfect and related categories was collected by means of a ty­ pological questionnaire, referred to as the Perfect Questionnaire - PFQ for short in the articles of this section (see Appendix 2). The first part contains 88 items, each consisting of one or more sentences to be translated with the help of contextual information (for translation questionnaires, see Dahl 1985, and Dahl, this volume); Part II, to be answered by a linguist, contains analytical questions. So far the ques­ tionnaire has been completed in over thirty languages,1 and for some of these, such a s Dutch, Italian and Macedonian, as well as the Serbo-Croatian area,2 dialectal or individual variation has been taken into account to some extent. Maslov (1990) gives the following notional definition of the perfect: "an aspectotemporal form of the verb, expressing a present state as a result of a preceding action 0r change, and/or expressing a past action, event or state that is somehow important

Ж

366

Jouko Lindstedt

to the present and is considered from the present point of view, detached from other past facts" (my translation). Actually the first part of the definition rather applies to a gram type called resultative (see below); important elements in the definition of the perfect proper are (1) the relevance of a past situation from the present point of view and (2) detachment from other past facts, i.e., non-narrativity. According to Comrie (1976: 52), "the perfect indicates the continuing present relevance of a past situation". The notion of current relevance (CR) is further explicated by Dahl and Hedin (this volume). In their recent major work on the grammaticalization of tense, aspect and modality, Bybee, Perkins & Pagliuca (1994) define the gram type anterior, which is essentially the same as the perfect discussed here. Their definition is, however, broader in that it also provides for past anteriors and future anteriors: "an anterior signals that the situation occurs prior to reference time and is relevant to the situation at reference time [...] Anteriors may occur with past or future tense marking" (p. 54). Past perfects (or pluperfects) and future perfects (futura exacta) share several properties with present perfects, but they also have some special uses and characteristics of their own, and they will not be discussed in this chapter, although PFQ does contain several items pertaining to them. In PFQ, definitions have been operationalized: a language possesses a perfect if it has a gram, associated with the verb, that is used in most of the first seven examples-which illustrate different kinds of CR of past situations-but is not used in the following four examples, consisting of short narratives.3 This means that there is a negative criterion as well: the perfect is not a narrative tense, and therefore the Latin and Serbo-Croatian Perfects, for instance, do not belong to the cross-linguistic type "perfect" at all-the Latin Perfect is a perfective past tense, and the Serbo-Croatian Perfect is a general past tense, or preterite. In terms of Greenberg's (1978: 75-76) distinction between the stability and frequency of typological features, the perfect is a gram type that is frequent, that is to say, likely to appear in different languages, but unstable, as it often tends to be lost. More often than not, it does not disappear as a form but becomes something else - a general past tense, for instance.

2. From resultative to current relevance Bybee and Dahl (1989: 67-68) list four typical diachronic sources of the perfect in the languages of the world: (i) copula + past participle of the main verb; (ii) possessive constructions involving a past participle of the main verb; (iii) main verb + particle meaning 'already'; (iv) constructions involving verbs like 'finish' or 'cast aside • The two latter sources are by their semantics completive constructions, whereas the two first ones, common in European languages, are resultatives (Bybee, Perkins *

The perfect - aspectual, temporal and evidential

367

Pagliuca 1994: 53-74). A European perfect deriving from a possessive construction often involves a transitive verb meaning 'to have'; if this is the case, it can be called a 'have' perfect. A copula-based perfect is a 'be' perfect. The distinction between resultatives and perfects has been established in linguistics only recently, largely owing to the important collective work edited by Nedjalkov (1988/1983). Resultatives "signal that a state exists as a result of a past action" (Bybee, Perkins & Pagliuca 1994: 54). Traditionally the resultative is often subsumed under the category of perfect as a special "statal" variant (as in Maslov's definition quoted above), or it is simply called a "stative". For the criteria of distinguishing resultatives from perfects, the reader is referred to Nedjalkov & Jaxontov (1988), Dahl (1985: 133-135), Bybee & Dahl (1989: 68), and Bybee, Perkins & Pagliuca (1994: 63-687), but the most important single difference should be mentioned here: only resultatives combine with adverbs of unlimited duration, such as 'still' or 'as before'. In English, it is not possible to say *She has still gone (if still is used in its temporal meaning)-compare with the resultative construction She is still gone. In Finnish, the perfect and its resultative source coexist, but they are formally differentiated in the plural: (2)

Finnish He o-vat väsy-nee-t. they be-PRS:3PL become:tired-PP-PL:NOM 'They have become tired.'

(3)

Finnish He o-vat väsy-ne-itä. they be-PRS:3PL become:tired-PP-PL:PRTV 'They are tired.'

(2')

Finnish *He o-vat yhä väsy-nee-t. they be-PRS:3PL still become:tired-PP-PL:NOM

(3 )

Finnish He o-vat yhä väsy-ne-itä. they be-PRS:3PL still become:tired-PP-PL:PRTV 'They are still tired.'

*ne Finnish resultative construction, besides accepting an adverb meaning 'still', follows the syntactic rule of ascriptive6 sentences whereby the predicate adjective is usually in the partitive case when the subject is plural but does not refer to a pair of "ings; the perfect conserves a petrified older syntactic rule and puts the participle mto the nominative.

368

Jouko Lindstedt

The perfect is typically a periphrastic gram, being formally close to its resultative (or completive) source. An important exception seems to be the old Indo-European Perfect, as attested in Classical Greek and Old Indie. A newer inflectional perfect in statu nascendi is the active resultative construction in North Russian dialects (Trubinskij 1988: 394; Tommola, this volume): (4)

Russian (dialectal) On den'gi poluci-vsi he money receive-PST:GER 'He has received the money.'

This resultative gram - which has already assumed some features of a real perfect, cf. Nedjalkov & Jaxontov (1988: 42-43) - is not periphrastic for the simple reason that there is no present tense copula in Russian; in the past tense, the copula is regularly expressed in the resultative, too. It is possible that what look like inflectional perfects usually come into existence as copula-less ascriptive structures. Semantically, the change from resultative into perfect means the generalization of meaning from "current result" to "current relevance". Lexically this is reflected in the spread of the gram from telic to atelic verbs (cf. Bybee, Perkins & Pagliuca 1994: 68-69; Dahl & Hedin, this volume). As the semantic connection between resultativity and CR is easy to grasp, the central use of the perfect is often called the "resultative perfect". But since the "still test" and similar criteria are operational in showing whether the transition from resultative proper to perfect has taken place, I shall speak rather of the "CR perfect" instead. A CR perfect is a perfect in its most central, prototypical meaning. Is it a tense or an aspect category? In a gram-based approach this is not an essential question. Classifying the perfect among aspects or among tenses would not significantly add to our understanding of it. But if tenses express the temporal location of situations, and aspects their temporal shape, it can at least be said that the rise of a CR perfect is associated with the loosening of the aspectual properties of its resultative (or completive) source. As noted by Bybee (1985: 160), "it seems to resemble a tense more than an aspect, since it does not affect the internal temporal contours of the situation".7 The central aspectual parameter of boundedness comes in two varieties: when telic situations reach their natural end-points, we can speak of material bounds (as in Susan built a beautiful house); when a situation, not necessarily telic, is limited temporally, a temporal bound is reached (as in Susan slept for nine hours; see Lindstedt 1995; cf. also Bertinetto & Delfitto, this volume). A material bound entails a temporal bound, but not vice versa. Resultativity presupposes a material bound, whereas CR only presupposes a temporal bound: a sentence like Someone has been here implicates that the presence of that someone has ceased, but it does not make "being here" a telic situation.

The perfect - aspectual, temporal and evidential

969

3. From current relevance to indefinite past The CR perfect shades into what is usually called the experiential (or existential) perfect. In English, these two types are formally differentiated only in rare cases like the following (cf. Comrie 1976: 58-59): (5)

Mary has gone to Paris.

(6)

Mary has been to Paris.

In (5), the fact of Mary's having gone to Paris may be relevant to the present state of affairs in various ways, but typically the sentence implicates at least that she is not present. The experiential perfect of (6) only expresses that the past situation in which Mary went to Paris is more indirectly part of the present state of affairs, most notably through Mary, who perhaps now knows what Paris is like. In its narrower definition, an experiential perfect presupposes an animate agent, since it expresses that "certain qualities or knowledge are attributable to the agent due to past experiences" (Bybee, Perkins & Pagliuca 1994: 62). In a broader definition, it only means that "a given situation has held at least once during some time in the past leading up to the present" (Comrie 1976: 58). This notion is further explicated by Dahl & Hedin (this volume), who call it "type-focusing event reference" (as opposed to "token-focusing event reference", as in Mary went to Paris). The experiential perfect is a way of referring to a past situation without referring to a particular occasion, that is to say, it is characterized by non-specific past time reference. From this point of view, it is more tense-like than the CR perfect, being an indefinite past tense which typically occurs in questions and negated assertions with 'ever'-type adverbials. However, CR and experientiality do not exclude each other; in examples like (1), elements of both can be discerned. This gradual transition of the perfect from the aspectual to the temporal domain means that it comes to be linked to an integrative way of viewing the past (Lindstedt 1983). The narrative tenses - the tenses of Weinrich's (1963) erzählte Welt - primarily refer to specific occasions, to the past moments defined by the internal structure of each narrative. If all possible past occasions are spoken about, the dynamic worlds of narratives must be replaced by a single static world, Weinrich's besprochene Welt, comprising the past, present and future. In the perfect's world, all past situations are still present - first through their results, then because they themselves become parts of this "extended now" (cf. McCoard 1978). However, experientiality as such is neither a necessary nor a sufficient condition tor a gram to be called a perfect. Experientials proper are a distinct gram-type, described by Dahl (1985: 139-144); the Japanese -ta koto ga am is a well-known example (see also Dahl & Hedin, this volume).

370

Jouko Lindstedt

In Slavic languages without a distinct perfect gram, the CR perfects of other languages are usually rendered with Perfective Past forms, whereas experiential perfects can often be translated with Imperfective Pasts. In Bulgarian, which does possess a perfect, a Perfect of a Perfective verb is interpreted as a CR perfect, whereas experiential Perfects are usually Imperfective (see Lindstedt 1985: 223-229; Lindstedt 1995: 99-100). Compare: (7)

Bulgarian A vie käde ste cu-l-i tova? but you:PL where be:2PL hear:PFV-PP-PL that:NT 'Where did you [polite] hear that?' (CR perfect)

^*

An evasive answer would be: (8)

Bulgarian Cu-va-1-a säm. hear-IPFV-PP-F be:lSG 'I've just heard.' (experiential perfect)

Although the experiential meaning may become dominant in the perfect, historically it is usually secondary and derives from the CR meaning. The perfect of a particular language may well be compatible with specific past time adverbials. Sentence (9) is ungrammatical in English, to be sure: (9)

*I have woken up at 4 o'clock this morning.

However, a perfect would be possible - though not the only alternative - in Finnish and Bulgarian, for instance. This is because there is an obvious CR reading - 1 woke up so early that I am now tired.8 According to Dahl (1985: 137-138), Swedish seems to occupy an intermediary position: a specific time adverbial can combine with the perfect if it is in the information focus. I assume that the degree of incompatibility of specific time adverbials with the perfect in a particular language shows to what extent it has become a dominantly experiential form and, therefore, a kind of tense. A possible next stage in this development would be the total loss of the CR reading, and this is what may be taking place in American English. If a child asks his or her parent, "Can I go now?', the counter-question (10a) strikes one as belonging to an American parent, whereas (10b) is the preferred alternative in British English (PFQ: 5): (10) a. b.

Did you do your homework? Have you done your homework?

The perfect - aspectual, temporal and evidential

371

1 am not aware of any extensive studies of this difference, but its existence, as a tendency rather than a rule, is generally taken for granted.9 I assume that the experiential function has been preserved much better in American English than the CR function has; thus, both British and American speakers would normally ask, for instance, Have you (ever) met my sister?; the sentence Did you meet my sister? would refer to a specific occasion or interval. The American perfect can be considered to be more tense-like than its British counterpart. To return to Bulgarian, at closer examination the situation in this language is similar to the American English situation in that CR does not require a Perfect form to be used at all; as a matter of fact, the Aorist could well be used in translating (10): (11)

Bulgarian (PFQ: 5) Na-pisa li si domasno-to? PFV-write:AOR:2SG Q RFL:DAT homework-DEF.NT 'Have you done your homework?'

Although the Perfect would also be possible here, it is actually difficult to show that the Modern Bulgarian Perfect has a CR meaning because, as was mentioned, what look like CR perfects are always instances of the Perfective Aspect. As the CR meaning can be expressed with a Perfective Aorist, but not with an Imperfective Perfect, it must be the Perfective, not the Perfect, that is the primary carrier of the CR meaning. On the other hand, the Perfect is obligatory in experiential contexts, a fact which is also reflected in native grammarians' traditional name for this gram, minalo neopredeleno vreme 'past indefinite tense'. Hannu Tommola and Nina Niissalo (p.c.) point out that the same situation is observed in Slavic as early as in Old East Slavic ("Old Russian") texts: experientiality requires the Perfect, but in the CR meaning, both Perfective Perfects and Perfective Aorists are possible. It is thus possible that the CR meaning of the Slavic Perfect grew weaker early in the Proto-Slavic period when verbal aspect came into being.

4. The perfect becomes a narrative tense When a perfect can be used as a narrative tense (with the possible exception of evidential contexts, see below), it has ceased to be a perfect. This is what has happened m the majority of Slavic languages, as well as in South German and North Italian dialects, in Sardinian (Georg Bossong, p.c.) and spoken French. As a matter of fact, the perfect has become a peculiar "maritime category" in Modern Europe - most of the languages and dialects with a stable perfect are situated on the fringe of the continent: the Baltic Finnic languages, Scandinavian languages, North German dialects, English, Portuguese, Spanish, South Italian dialects, Greek, Albanian, Macedonian a nd Bulgarian. (Cf. Thieroff, this volume, Map 4.)

I

372

Jouko Lindstedt

The macroareal picture is completed by the observation that at the corners of this maritime fringe, Portuguese and Modern Greek exhibit what look like "young perfects", with relatively strong restrictions on what counts as CR. As can be seen in the article of Squartini & Bertinetto (this volume), the Portuguese Perfect {perfeito, also known as perfeito composto) is mainly used to express situations or series of repeated situations that span from the past up to the present. As a matter of fact, PFQ material obtained from one Portuguese informant suggests that it is rather the adverb jà 'already, now' that may be grammaticalizing as the real perfect marker in this language.10 As for the Modern Greek Perfect, restrictions on its use are less drastic, but Dahl & Hedin (this volume; cf. also Hedin 1987: 80-86) point out that in CR contexts, the Aorist can be used -just as in Bulgarian, cf. (10) above. In those contexts the Perfect focuses more on the present state of affairs and less on the past event itself than the Aorist does. Thus, the Greek Perfect is still relatively close to its resultative source - this can also be seen in the fact that it can only be formed from the Perfective ("Aorist") stem of the verb. The gradual development of the German Perfect from a perfect into a general past tense seems to fit what Dahl (this volume) describes as a two-dimensional grammaticalization cline: both the geographical distance from the southern centre of innovation and the functional distance from the old prototypical uses of the perfect (such as CR) bear on the propensity for Perfect use (see Latzel 1974 and Thieroff, this volume on the use of the German Perfect in different contexts). In those Romance dialects in which the perfect has become a narrative tense, it is now a perfective past, opposed to the imperfective past (Squartini & Bertinetto, this volume). This is an instance of semantic generalization, for the CR meaning presupposes boundedness. Semantic bleaching of this kind is typical of later stages of a grammaticalization process (Hopper & Traugott 1993: 87-93; the notion of bleaching was introduced by v.d. Gabelentz 1891: 241-242). The change whereby the perfect of most Slavic languages has become a general past with no aspectual value of its own must at least be partly due to the fact that aspect came to be expressed by other means in Slavic. The perfect did not become a perfective past, since perfectivity vs. imperfectivity could be expressed with verbal affixes, independently of tense marking. A still stronger assumption would be that the loss of the perfect vs. narrative past distinction itself was caused by the rise of this aspect opposition, as the new Perfective, with a completive source, was able to assume resultative and CR functions. As suggested by Dahl & Hedin (this volume), the development from a perfect into a more general past tense "may at least partly be interpreted in terms of gradual relaxation of the requirements of current relevance". It is more difficult to say whether there is also a path from experientiality (the indefinite past function) to a general past. Such a process cannot be observed in American English, but histori-

The perfect - aspectual, temporal and evidential

373

cal data from Slavic suggest that it is at least a typological possibility if the Slavic Perfect had already lost most of its CR functions before the other past tenses disappeared. At any rate, it is an interesting question why a CR form tends to displace a nonCR form even when the CR distinction is lost. It should be noted that this is not what happened to the old Indo-European inflectional resultative / perfect in Greek, Latin or Germanic; in those languages, the Perfective Past or Simple Past got the upper hand, though formally incorporating parts of the old perfect morphology (as in the reduplicated Perfective Pasts of Latin, or in the -k- Aorists of Modern Greek; for Germanic, see Bybee, Perkins & Pagliuca 1994: 77-78). Thus, the morphological type of the perfect - periphrastic vs. inflectional - may have some bearing on its propensity to displace a simple past tense; the mechanism seems to be the same whereby case forms are often replaced by adpositional phrases in various languages. In terms of grammaticalization processes, the perfect is typically a younger gram than the other (narrative) past tenses-less synthetic in its expression, more marked as to its semantic content. Its functional expansion and the concomitant semantic bleaching open a space for a new perfect to appear-an example of a grammaticalization cycle known as layering (Hopper & Traugott 1993: 124-129). Thus, in some Slavic languages where the old perfect has become a general past tense, a new perfect is coming into existence. Such a development is observed in Czech, where the verb mit 'to have' enters with the past passive participle into constructions like Mas vycistëné zuby? 'Have you brushed your teeth?' (Bybee & Dahl 1989: 70), Mâm ûlohu napsanou 'I have completed my task' (Maslov 1988: 80), mit knihu rozectenou 'to have a book half-read' (Short 1993: 487; see also Tommola, this volume, section 7). Another possible path of evolution mentioned in connection with the perfect becoming a narrative tense is the path leading through the so-called perfect of recent past. As pointed out by Dahl (1985: 136), the development from the CR meaning into the recent-past meaning is natural because "a recent event is more likely to have a persistent result than a distant one". Sometimes a perfect acquires a hodiernal interpretation: events of the same day can be referred to by it, even in connected narrative. This is illustrated in the PFQ material from a speaker of Alicante Spanish:

(12) a.

Alicante Spanish (PFQ: 8) [Do you know what happened to me just an hour ago?] Yo estaba andando en el bosque. De pronto, hepisado una culebra. Me ha mordido en la pierna. He cogido una piedra y se la he tirado a la culebra. Se ha muerto.

374

Jouko Lindstedt b.

(PFQ: 9) [Do you know what happened to me yesterday?] Yo estaba andando en el bosque. De pronto, pisé una culebra. Me mordiô en la pierna. Cogi una piedra y se la tiré a la culebra. Se murio. 'I was walking in the forest. Suddenly I stepped on a snake. It bit me in the leg. I picked up a stone and threw it at the snake. It died.'

The italicized forms in (12a) are Perfects; in (12b), Perfective Pasts are used instead. (The questionnaire was administered by Scott Schwenter; see Schwenter 1994.) An analogous distinction between PFQ: 8 and PFQ: 9 is made in Catalan, which is in areal contact with Alicante Spanish. The recent-past function of the perfect in Spanish, Catalan, Occitan and pre-Modern French is discussed by Comrie (1976: 60-61), Dahl (1985: 125) and Bybee, Perkins & Pagliuca (1994: 101-102); Comrie (1976: 61) suggests that "[g]radual relaxation of the degree of recentness required for the use of the Perfect seems to have been a key part of the development of the Perfect in many Romance languages to oust the Simple Past completely". However, the study of Squartini & Bertinetto (in this volume) on the Romance perfect does not suggest that recency has had such a central role in this development, except for a limited area in Southwest Europe, with Catalan as its centre. A more general Central and East European trend, based on areal contacts and the typological expansiveness of formally and semantically marked grams, may account for the Romance, Slavic and German development alike.

5. Evidentiality and South Slavic perfects The South Slavic linguistic area consists of a series of closely connected dialects, grouped nowadays into five or six standard languages." The Proto-Slavic Perfect, a periphrastic formation of the 'be' type, has been preserved as a form in all of the area except for some South Macedonian dialects, but almost everywhere it has been subject to grammatical changes. However, there are two epicentres of these developments, and the respective outcomes are far from identical (Lindstedt 1993, 1994). Starting from the Slovene territory in the northwest, the Perfect simply became a general past tense, ousting first the Imperfect, then the Aorist. These latter grams have completely disappeared in Slovene (excepting some dialects spoken mainly in Italy); in Croatian and Bosnian standard languages they have only survived as literary archaisms, apart from some expressive functions of the Aorist that have nothing to do with narration. In Serbian, the Aorist and Imperfect are more frequent, but the Perfect is already the prevalent narrative tense here, too (Savic 1991). The old system in which the Perfect is only a non-narrative tense is preserved in some Mon-

The perfect - aspectual, temporal and evidential

375

tenegrin dialects, whose system is sometimes reflected even in modern fiction written by authors from that area (Lindstedt 1994: 39). From the other end of the South Slavic area, a grammatical change has spread that has made the Perfect an evidential form, or rather built an evidential system out of the Perfect. Evidential, to be distinguished from mood, is, according to Trask (1993 s.v.), "a grammatical category occurring in some languages by which all statements (and sometimes other sentence types) are overtly and obligatorily marked to indicate the source of the speaker's evidence for her/his utterance" (see also Chafe & Nichols 1986, and Willett 1988). There are no well-established terms for different types of evidentials.12 I propose the term indirective for the most widespread evidential gram type, expressing that the speaker has not witnessed the situation he or she is speaking about, but knows of it from hearsay or other kinds of indirect evidence. Other grammatical and semantic terms needed are reportative, which should be confined to the hearsay case alone,13 and inferential, referring to statements made on the basis of inference, not hearsay. There are two major areas in Europe where grammaticalized evidentiality distinctions are common. One is the Baltic region, comprising the Baltic languages Lithuanian and Latvian and the Finnic languages Estonian and Livonian. The other area can be called the Black Sea area, as it consists of languages around this sea, though it stretches farther to Central Asia (Haarmann 1970; Dahl 1985: 149-153; Friedman 1986; Johanson 1992: 244-246). The indirectives of this area are generally regarded as having arisen due to Turkish influence during the Ottoman reign. However, the tendency to develop indirectives from various sources is typical not only of Turkish, but of the whole Turkic stock, and the area also extends from the Black Sea to regions where such Uralic languages as Komi and Udmurt, farther north, are spoken (Leinonen & Vilkuna, this volume). It has even been called into question whether Turkish was really the primary Turkic source of the indirectives now found in the Balkans (Johanson ms.). An interesting question is whether the two indirective areas really form a single "Eurasian isogloss" as suggested by Haarmann (1970). But as evidentiality has not been grammaticalized in East Slavic (see Tommola, this volume, section 4.6, for PFQ material), there are no data connecting the Baltic region with the east and southeast. Nothing is known about the grammatical structures of those extinct Baltic and Uralic languages of Central Russia that might have formed a bridge between the two areas. In several languages of the Black Sea area, a resultative or a perfect has been the main diachronic source of the indirective. We can assume that the inferential meaning forms the semantic link here. Drawing inferences from the visible results °f a non-witnessed event is a natural extension of resultativity or CR; it is resultativity the other way round, as it were (Comrie 1976: 108-110; Bybee, Perkins & Pagliuca 1994: 95-96). The inference in sentence (13), for instance, can be expressed by a

376

Jouko Lindstedt

perfect even in some European languages outside the Black Sea area, though not in English: (13)

(PFQ: 69) [Investigating a burglary, seeing footprints beneath a window:] The thief entered (must have entered) the house by this window.

Swedish is among the languages that can use the perfect: (14)

Swedish (PFQ: 69) Tjuv-en ha-r komm-it in genom det här thief-DEF have-PRS come-SUP in through DEF.NT here fönstr-et. window-DEF:NT

The Scandinavian inferential perfect is discussed by Haugen (1972) and, with abundant examples from Swedish, by Kinnander (1973). The use also extends to Finnish. As for the German Perfect in the same sentence {Durch dieses Fenster ist der Dieb [in das Haus] eingedrungen), its inferential value is more difficult to assess, for, as noted above, this tense has extended its use towards the preterite; but Weinrich (1964: 84-86) observes that the German Perfect is typically used in historical prose when past facts are commented on and explained, not only related in a sequence, and this can be seen as a natural extension of the inferential use. As a further step in grammaticalization, inferentiality can be extended to the reportative meaning and other types of indirect evidence, so that a real indirective comes into existence (Bybee, Perkins & Pagliuca 1994: 96-97, 105; cf. also Dahl 1985: 152-153, and Nedjalkov's "typological appendix" in Plungian 1988). Another possible connection between the perfect and indirective is the fact that participles often enter into subordinate constructions expressing reported speech. Participles used as predicates in subordinate sentences seem to be the main source of indirectives in the Baltic area (Ikola 1953: 62-63; Schmalstieg 1988: 117-121), and some partially similar developments are perhaps not excluded in the Black Sea area, either. On this grammaticalization path it is the reportative meaning that forms the basis of an indirective category. Let us now have a closer look at the South Slavic indirectives. In Bulgarian, the Perfect has given rise to an evidential form that can be characterized as an Indirective Aorist. Later, other indirective tenses have developed according to the structural model of the Indirective Aorist, such as the Indirective Imperfect (also used as the Indirective Present tense) and the Indirective Future. The distinction between the Perfect and the Indirective has been one of the most difficult questions for Bulgarian grammarians since the past century, and no definitive solution has been reached yet(The description I gave in Lindstedt 1985 is not satisfactory, either.)

The perfect - aspectual, temporal and evidential

377

The main descriptive problem is that the Perfect and the Indirective Aorist are not formally differentiated. The feature often mentioned in normative grammars, viz. the omission of the auxiliary in the third person of the Indirective, is not an unambiguous marker: the auxiliary is usually omitted when Indirective tenses are used in connected narrative, but this is not an absolute rule; and there are other uses of the Indirective forms that do not formally differ from the Perfect at all (Lindstedt 1993: 4 3 ^ 8 ; Lindstedt 1994: \A-\1). Auxiliary drop in certain discourse types is also observed in the neighbouring Serbo-Croatian area, but it is not connected with evidentiality (Savic, this volume). Despite the appearance of the Indirective forms, the Bulgarian Perfect has also retained its old meanings, notably the experiential (Lindstedt 1993: 49-50; 1994: 49_50). The sentence Mary has been to Paris would be translated into Bulgarian with the Perfect, and so would the sentence / have been to Paris, where the meaning is far from indirective - the speaker has certainly witnessed his or her own visits to Paris. The developmental path of the Bulgarian Perfect has clearly bifurcated: one of its two main functions is now experiential and non-narrative, the other is indirective and narrative. The linking CR function has grown weak, as noted in section 2 above; what we have here, then, is a kind of "doughnut gram" as denned by Dahl (this volume). In Macedonian, the old 'be' perfect has acquired an indirective meaning, as in Bulgarian, but it is not so clearly polysemous owing to the appearance of a new 'have' perfect modelled on Greek and Arumanian. There is also a third, mainly resultative construction, so that 'I have come' can be expressed in three ways (the 1SG forms with a masculine subject are given):

sum dosol imam dojdeno sum dojden

'I am' + past participle active, M.SG. 'I have' + past participle passive, NT.SG 'I am' + past participle passive, M.SG.

Friedman (1976; 1977: 52-99) and Graves (this volume) have investigated the areal and functional distribution of these grams. In Northern Macedonia the sum dosol type dominates in both the perfect and indirective meaning; this is closest to the Bulgarian situation. In the southernmost part of the language area, only imam dojdeno is used. In between, in Ohrid dialects, the old perfect is only used as an indirective, and the new type imam dojdeno serves as an ordinary perfect. As for the sum dojden type, it is still mainly a resultative proper, used in sentences like Umren e 'he is dead' (lit. 'died is', PFQ: 3). However, in southwest dialects it is °n its way to becoming a perfect, which is shown by the fact that even with certain transitive verbs the subject may be the agent of the action:14

378 (15)

Jouko Lindstedt Macedonian Jas sum jad-en-a. I be:lSG eat-PPP-F 'I've been eaten'; SW Macedonian: 'I've eaten.'

i *

The development of the three Macedonian grams can be schematically presented as follows, depicting the synchronic situation as an on-going process, with different grams covering different stretches of the grammaticalization path: Resultative

Perfect

Indirective

Loss

sum dojden imam dojdeno

'

sum dosol However, this diagram should not be interpreted as a prediction that the newer types of perfect are also bound to become indirectives in due course. Although this is of course possible, it is not the most probable development typologie ally; and areally it is not clear whether the Balkan area still favours the appearance of new indirectives, for the Turkic influence has almost vanished.

6. The path of the perfect Although expressing the current relevance of a past situation is the central and prototypical meaning of the perfect, I know of no perfects that only have this function. I propose the following tentative universal: If a gram has the CR meaning, it also has at least one of the following meanings: resultative; experiential (indefinite past); inferential; reportative. If the central or sole meaning of a gram is resultative or indirective (inferential and/or reportative), it is not yet a perfect, or no longer a perfect. If the meaning of a gram has strongly shifted towards indefinite past, as in the American English Perfect or in the non-evidential branch of the Bulgarian Perfect, it is perhaps a matter of definition whether the term "perfect" should still be used. Given that CR and experientiality do not logically exclude each other, and often one and the same utterance contains elements of both, I am inclined to retain the name "perfect" for these cases. It should, however, be pointed out that the two American English speakers who completed PFQ used the Perfect in only about half of the defining items.1 Bybee, Perkins & Pagliuca (1994: 79) list 30 "old anteriors" from meir crosslinguistic sample that have developed various further uses in addition to CR. They

The perfect — aspectual, temporal and evidential

379

also present a table of 57 "grams with anterior as their only use (young anteriors)" (1994: 64-65), but the caption is misleading: grams with other uses besides CR have been admitted to the table, if those uses (such as completive) were considered "indicative of early rather than later stages of development" (1994: 63). Moreover, experientiality or the function of anterior continuing (I've been waiting for him for an hour) was not frequently mentioned at all in the reference material used (1994: 62), so one cannot conclude that they are really absent from the grams of the table. Still another table (1994: 66) presents several grams that combine the resultative and "anterior" (CR) meanings. One can reach the conclusion that current relevance is not only a function, but also a junction: grammaticalization paths from different sources, such as resultative and completive, converge here; and further paths lead from here to experiential (indefinite past), perfective past, general past, and also to indirective (cf. Bybee, Perkins & Pagliuca 1994: 105, and also Anderson 1982: 240). Every concrete gram is only a passer-by at this junction, still having its tail in the resultative, or pushing its head towards new aspectual, temporal or evidential meanings.

Notes 1. More precisely, Part I has been completed in over thirty languages by one or more informants. The descriptive second part does not exist for all of these languages. The material will be placed on an Internet file server, possibly in the public domain if resources allow. 2. The Serbo-Croat(ian) area is the Central South Slavic dialect area between Slovene, on the one hand, and Bulgarian and Macedonian, on the other. The standard languages of this area are now called Croatian, Bosnian (see Halilovic 1991) and Serbian; they are all based on a dialect group called Neo-Stokavian. 3. It is important to comprehend the nature of this method. The crucial items in PFQ do not represent notional categories chosen a priori, because they draw upon Dahl's (1985) empirical results describing the actual clustering of grammatical categories across languages. 4. This fact need not prevent these tenses from being called Perfects in the grammars of the respective languages by force of tradition. This applies especially for languages such as Dutch and Croatian/Bosnian/Serbian, in which the old Perfect has become a narrative tense and, therefore, has ceased to be a perfect proper, yet the old narrative tense or tenses still survive as (stylistically marked) alternatives. 5. Thus, Romance perfects have mainly developed from Latin constructions involving habeo 'have' or teneo 'hold' > 'have', e.g., Navem paratam habeo 'I have a ship prepared' (Squartini & Bertinetto in this volume; Maslov 1988: 73; cf. Bybee & Dahl 1989: 72). Similar constructions occur in other language groups but, on the whole, 'have' perfects are a distinctively European phenomenon because "expressing 'having' by a transitive verb appears to be an areal trait of Western and Southern Europe" (Bybee & Dahl 1989: 98, fn. 7).

380

Jouko Lindstedt

6. For the term "ascriptive", see Lyons (1977: 437-438). 7. But she also points out that the perfect often has distinct forms for the present and past tense (as in English). 8. Notice that specific time adverbials can sometimes combine with resultatives in various languages, e.g., Swedish Boken är skriven 1950 'the book was written in 1950' (lit. "is written"). Nedjalkov & Jaxontov (1988: 54) present the German resultative Diese Stadt ist im Jahre 1794 gegründet 'this town was founded in 1794' (lit. "is founded"), but not all native speakers are happy with this example. Temporal adverbials used with resultatives typically specify the time of the resulting state, not that of the past event. 9. Unfortunately studies on "the English Perfect" do not always specify what variety of English is investigated. 10. The informant inserted jâ into all of the diagnostic sentences from PFQ: 1 to PFQ:7, and into several other typical perfect contexts as well, but did not use the perfeito. 11. See note 2. 12. A source of confusion is that the term "evidential" is often used as a name of an evidential gram ("the evidential in language X"), just as if the imperative, for instance, were called "the mood". 13. Dahl's (1985: 149) "quotative", though used in Amerindian studies and now adopted by Trask, is not felicitous, because to quote means 'to repeat the exact words someone has said or written' ; in reportatives, only the content is repeated. Dahl now prefers the term "reportative", too (p.c.). 14. For this criterion, see Nedjalkov & Jaxontov (1988: 23). 15. Both American English informants used the Perfect in PFQ: 1, PFQ:4 and PFQ:7, one of them also in PFQ:6 (for the items, see the Appendix). An informant of British English used the Perfect in all defining items from PFQ:1 to PFQ:7 (in PFQ:3 only as a second alternative besides the Stative / resultative He is dead).

References

Anderson, Lloyd B. 1982 "The 'Perfect' as a universal and as a language-particular category", in Paul J. Hopper (ed.), 227-264. Bertinetto, Pier Marco & Denis Delfitto this volume "Aspect vs. actionality. Why they should be kept apart". Bybee, Joan L. 1985 Morphology: A study of the relation between meaning and form. Amsterdam & Philadelphia: Benjamins. Bybee, Joan L. & Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51-103. Bybee, Joan & Revere Perkins & William Pagliuca 1994 The evolution of grammar: Tense, aspect, and modality in the languages of the wor Chicago and London: University of Chicago Press. Chafe, Wallace & Johanna Nichols (eds.) 1986 Evidentiality: The linguistic coding of epistemology. (Advances in Discourse Processes 20.) Norwood, New Jersey: Ablex.

the perfect - aspectual, temp«ili ttt|;cvidenüal

3*1

^omrie, Bernard 1976 Aspect. (Cambridge Textbooks in Linguistics.) Cambridge: Cambridge University Press, tomrie, Bernard & Greville G. Corbett (eds.) 1993 The Slavonic Languages. London & New York: Routledge. Jahl, Osten 1985 Tense and aspect systems. Oxford & New York: Basil Blackwell. this volume "The tense-aspect systems of European languages in a typological perspective". )ahl, Osten & Eva Hedin this volume "Current relevance and event reference". J 'riedman, Victor A. 1976 "Dialectal synchrony and diachronic syntax: The Macedonian perfect", in Sanford B. Steever & al. (eds.), Papers from the parasession on diachronic syntax, Chicago Linguistic Society, 96-104. 1977 The grammatical categories of the Macedonian indicative. Columbus, Ohio: Slavica. 1986 "Evidentiality in the Balkans: Bulgarian, Macedonian, and Albanian", in Chafe & Nichols (eds.), 168-187. abelentz, Georg von der 1891 Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. [Linguistics: Its tasks, methods and present findings.] Leipzig: Weigel. jraves, Nina this volume "Macedonian - a Language with Three Perfects?" ireenberg, Joseph H. 1978 "Diachrony, synchrony, and language universals", in Greenberg (ed.), 61-91. ireenberg, Joseph H. (ed.) 1978 Universals of human language, I: Method & theory. Stanford, California: Stanford University Press, laarmann, Harald 1970 Die indirekte Erlebnisform als grammatische Kategorie. Eine eurasische Isoglosse. [The indirect evidential form as a grammatical category: An Eurasian isogloss.] (Veröffentlichungen der Societas Uralo-Altaica 2.) Wiesbaden: Otto Harrassowitz. ïalilovic, Senahid 1991 Bosanski jezik. [The Bosnian language.] (Biblioteka Bosanski krug.) Sarajevo: Biblioteka Kljucanin. laugen, Einar 1972 'The inferential perfect in Scandinavian: a problem for contrastive linguistics", The Canadian Journal of Linguistics 17: 132-139. ledin, Eva 1987 On the use of the perfect and the pluperfect in Modern Greek. (Acta Universitatis Stockholmiensis, Studia Graeca Stockholmiensia 6.) Stockholm: Almqvist & Wiksell. topper, Paul J. (ed.) 1982 Tense-aspect: Between semantics and pragmatics. Amsterdam: Benjamins, topper, Paul J. & Elizabeth Closs Traugott 1993 Grammaticalization. (Cambridge Textbooks in Linguistics.) Cambridge: Cambridge University Press, kola, Osmo 1953 Vironja liivin modus obliquuksen historiaa. [On the history of the modus obliquus in Estonian and Livonian.] Helsinki: Suomalaisen Kirjallisuuden Seura. (Suomi 106: 4; with a German summary: "Zur Geschichte des estnischen und livischen Modus obliquus".)

382

Jouko Lindstedt

Johanson, Lars 1992

manuscript

-

r.m

Strukturelle Faktoren in türkischen Sprachkontakten. [Structural factors in contacts with Turkic languages.] (Sitzungsbenchte der Wissenschaftlichen Gesellschaft an der Johann Wolfgang Goethe-Universität. Frankfurt am Main, XXIX, 5.) Stuttgart: Franz Steiner. "Zum Kontakteinfluß türkischer Indirektive". [On the contact influence of the Turkic indirective.]

Kinnander, Bengt 1973 "Perfektum i 'sekundär' användning" [On the perfect in a 'secondary' use.], Nysvenska studier 53: 127-172. Latzel, Sigbert 1974

"Zum Gebrauch der deutschen Vergangenheitstempora. Zwei Studien" [On the use of the German past tenses. Two studies.], in Hermann Gelhaus & Sigbert Latzel, Studien zum Tempusgebrauch in Deutschen. (Institut für deutsche Sprache, Forschungsberichte, 15.) Mannheim, 169-348. Leinonen, Мала & Maria Vilkuna this volume "Past tenses in Permian languages". LES = Lingvisticeskij ènciklopediceskij slovar' [Encyclopedic dictionary of linguistics] 1990 Moskva: Sovetskaja ènciklopedija. Lindstedt, Jouko 1983 "The past is present: notes on the perfect tense", in Fred Karlsson (ed.), Papers from the Seventh Scandinavian Conference of Linguistics, I. (University of Helsinki, Department of General Linguistics, Publications 9.) Helsinki. 246-259. 1985 On the semantics of tense and aspect in Bulgarian. (Slavica Helsingiensia 4.) Helsinki. 1993 "Za razvitieto na juzboslavjanskija perfekt" [On the development of the South Slavic perfect], Studia Slavica Finlandensia 10: 31-54. 1994 "On the development of the South Slavonic Perfect", in Three papers on the perfect, 32-53. (EUROTYP Working Papers VI: 5). 1995 "Understanding perfectivity - understanding bounds", in Pier Marco Bertinetto, Valentina Bianchi, Osten Dahl & Mario Squartini (eds.), Temporal reference, aspect and actionality, vol. 2: Typological perspectives. Torino: Rosenber & Sellier. 95-103. Lyons, John 1977 Semantics 1-2. Cambridge: Cambridge University Press. McCoard, Robert W. 1978 The English perfect: Tense-choice and pragmatic inferences. (North-Holland Linguistic Series 38.) Amsterdam: North Holland. Maslov, Jurij S. 1988 "Resultative, perfect, and aspect", in Nedjalkov (ed.), 63-85. Maslov, Ju. S. 1990 "Perfekt" [Perfect], in LÈS, p. 372. Nedjalkov, Vladimir P. (ed.) 1988 Typology of resultative constructions. (Typological Studies in Language 12.) Amsterdam & Philadelphia: John Benjamins. [An enlarged translation of Tipologija retul'tativnyx konstrukcij (rezul'tativ, Stativ, passiv, perfekt). Leningrad 1983: Nauka.] Nedjalkov, Vladimir P. & Sergej Je. Jaxontov 1988 "The typology of resultative constructions", in Nedjalkov (ed.), 3-62. Plungian, Vladimir A. 1988 "Resultative and apparent evidential in Dogon", in Nedjalkov (ed.), 481^193. (With a "Typological appendix" by V. P. Nedjalkov.)

The perfect - aspectual,temporaland evidential Savic, Svenka 1991

383

"Pragmaticni aspekti vremena u naraciji u srpskohrvatskom standardnom jeziku" [Prag­ matic aspects of tenses in naratives in the Serbo-Croatian standard language], GodiSnjak Filozofskog fakulteta и Novom Sadu 20: 149-155. this volume "Discourse features of the truncated perfect in spoken Serbo-Croatian". Schmalstieg, William R. 1988 A Lithuanian historical syntax. Columbus, Ohio: Slavica. Schwenter, Scott A. 1994 "The grammaticalization of an anterior in progress: Evidence from a peninsular Spanish dialect", Studies in Language 18: 71-111. Short, David 1993 "Czech", in Comrie & Corbett (eds.), 455-532. Squartini, Mario & Pier Marco Bertinetto this volume 'The simple and compound past in Romance languages". Thieroff, Rolf this volume "On the areal distribution of tense-aspect categories in Europe". Tommola, Hannu this volume "On the perfect in North Slavic". Trask, R. L. 1993 A dictionary of grammatical terms in linguistics. London & New York: Routledge. Trubinskij, Valentin I. 1988 "Resultative, passive, and perfect in Russian dialects", in Nedjalkov (ed.), 389-409. Weinrich, Harald 1964 Tempus: Besprochene und erzählte Welt. Stuttgart: W. Kohlhammer. Willen, Thomas 1988 "A cross-linguistic survey of the grammaticization of evidentiality", Studies in Language 12: 51-97.

>«,АИ

Russian U zen-y stol'ko u2e pro vas PREP wife-GEN so_much already about PRON.2PL:ACC na-slys-an-o. PFV-hear-PPP-NT 'My wife has heard so much of you already.' Russian (non-standard) U syn-a institut za-konc-enl-o. PREP son-GEN institute PFV-finish-PPP:M/-NT 'My son has higher education [has finished the institute].'

Possessive and other resultatives tend to be emphatic, something that Bybee et al. (1994: 54) specifically attach to "complétives". Therefore it is interesting that verbs with a completive meaning are frequent in resultative constructions. 27 In the standard North Slavic languages possessive constructions employ either a verb for 'having', or (in Russian) a copula construction with a preposition phrase indicating the "possessor". 28 The occurrences found in the PFQ responses can be divided into two, or even three, categories: a) those using adjective complements (62), and b) those having a verbal (participle) form as the complement (63). The adjective constructions remain statives and can hardly grammaticalize. The participle constructions can show agreement of the participle with the object in number, case and gender (with numerals, the participle form may be governed by it; see 13), or the form of the participle is not inflected. In the latter case the development towards a perfect gram has become possible. Possessive constructions were found in the Czech (13,23, and 62a), Russian (63) and Upper Sorbian (62b) PFQ responses: 29

(62) a.

b.

(63)

Czech (PFQ: 05) Ma-s hotovy ukol? have-2SG ready assignment 'Have you done your homework?' Upper Sorbian (PFQ: 70) Ne, won drje z-më-je je hizo hotow-e. no he sure PFV-have-3SG PRON.ACC.PL already ready-PL 'No, he'll have finished it [them] already.'

Russian (PFQ: 44) U menja sobra-n-o nie. 200 kukol. PREP I:GEN gafher:PFV-PPP-NT already 200 doll:GEN.PL 'I have collected some two hundred dolls so far.'

0 * A * p * c t in North Slavii

4éS

7.4. Gerund construction In Northern dialects of Russian the copula + Present Gerund construction has been regarded as a potential new perfect.30 In the standard (spoken) language it appears, however, only with some typical verbs with a resultative state meaning, and these expressions have remained stylistically marked. To some extent this construction is a systemic feature of the tense-aspect grammar of the speakers of certain dialects. In the language of dialect speakers Imperfective verbs also occur (64b), while Perfectives are found in the standard language, often with just a few typical verbs, like the one in (64a). (64) a.

Russian On by-l vy-pi-vsi. he COP-PST out-drink-GER.PST 'He was drank' Russian (non-standard) My davno ne spa-msi, ne e-msi. we long NEG sleep-GER.PST NEG eat-GER.PST 'We have neither slept nor eaten for a long time.'

In some Russian dialects the gerund construction is used as an object resultative, in spite of the actual active semantics of the gerund itself (see Trubinskij 1984: 156— 186): (65)

Russian (non-standard) Pol po-my-vsi. floor PFV-wash-GER 'The floor is / has been washed.'

7.5. Other strategies31 In a context where the object is topicalized and the verb is focused, the experiential perfect reading seems to be pragmatically preferred to the narrative reading. (66)

Russian (PFQ: 02) Net, èt-u knig-u ona pro-ëita-1-a. no this-ACC.F book-ACC she PFV-read-PST-F 'She has read this one (all the way through).'

466

Hannu Tommola

Leinonen (1994: 143) reiterates Crystal's (1966: 8, fn. 2) observation about the role of intonation in the interpretation of the English Perfect in the sentence I've been to the Old Vic!. The reading is experiential, if the verb form (been) is stressed, but the stress on the Old Vic refers to a recent situation. It is well known that the Russian Imperfective Past has a so-called obscefakticeskoe znacenie, general factual meaning, or a simple denotative use that occurs if just the existence of a past action is stated. Then it is the verb that is stressed. The theme-rheme structure cannot in written text be rendered by intonation, but the same function is partly taken over by the word order. The experiential meaning does not belong to the imperfective aspect domain alone (see above, 46-47). In questions as in (67a) (experiential) and (67b) (resultative), the verb has to be stressed; otherwise they may not have the perfect meaning: (67) a.

b.

Russian (PFQ: 04) Vy kogda-nibud' vstreca-l-i mo-ju sestr-u? PRON.2PL ever meet-PST-PL my-ACC.F sister-AGC 'Have you met my sister?' Russian (PFQ: 05) A ty s-dela-l urok-i? and/but PRON.2SG PFV-do-PST lesson-PL 'Have you done your homework?'

Some informants tended to specify the context, and with this pragmatic addition, example (68), for instance, is an unambiguous (nonnarrative) context in Russian. (68) a.

b.

7.5.1.

Finnish (PFQ: 16) Ole-n nous-sut / Nous-i-n tänään neljä-ltä. COP-1SG stand_up-PP / -PST-1SG today four-ABL Russian (PFQ: 16) Delo v torn, cto ja prosnulsja v 4 matter PREP that:LOC that I wake_up:PST PREP 4 cas-a. hour-GEN '[It's because] I woke up at 4 o'clock today.' A Russian experiential

In Tommola (1986: 47, 48, 58) the function of the Russian modal auxiliary prixodilos' - Past of the Imperfective prixodit'sja 'have to' - was observed in uses that do not imply any modal element, but rather an experiential meaning. If in (69a) some modal nuance might be included, in (69b) prixodilos' is absolutely a pure experiential auxiliary:

OB Ae perfect in North Slavic (69) a.

b.

467

Russian (Tommola 1986: 48) Tebe prixodilos' celovek-a PRON.2SG:DAT happen:IPFV:PST:NT (hu)man-ACC ubi-va-t"! kill-IPFV-INF 'Have you ever killed a man (a human being)?' Ob et-om nam uze prixodilos' PREP it-LOC we:DAT already happen:IPFV:PST:NT pisa-t': write-INF... 'I have already written about this [as follows]: ... '

Some prerequisites for grammaticalization are there - the lexical meaning of the auxiliary has totally disappeared - but some are lacking; the construction is op­ tional. The same content can be communicated by using the simple Imperfective Past (Ту ubival kogda-nibud' celoveka, and Ob ètom my uze pisali). It is noteworthy, however, that the experiential meaning tends to require adding the adverbial kogdanibud' 'ever' in (69a), if the auxiliary is deleted. In (69b) 'already' can be taken as a sufficient experiential marker. 7.5.2.

'Already'

In the sample of Dahl (1985: 129-130), there is at least one language (Yoruba) that has grammaticalized a particle (ti) having the basic meaning 'already' as a perfect marker. The data recorded by Bybee et al. (1994: 64) add to these languages Inuit and Buli, in which "young anteriors" (= perfects) are of this same origin. Relational adverbs meaning 'already' and 'just' are typically found both in progressive and current state contexts, when "all is ready", i.e., an activity 'is already going on', or the (beginning of a new) state 'is (has been) already achieved'. If relative tense forms (perfect, pluperfect, future perfect) are not available, in temporal sentences 'already' is often needed in the main clause to indicate the order of the actions, since temporal connectors like 'when' tend to be neutral as to the mutual anteriority, simultaneity or posteriority of the actions denoted by the finite verbs (see above, 22-23, 30-31). As was mentioned above, in some languages 'already' also occurs frequently in experiential perfect contexts. It may be a later development, and the obligatoriness °f this use remains to be explored. But at any rate 'already' is a potential perfect barker, rather than a resultative marker, for it is attested to in all the different perfect Us es. Nevertheless, it cannot, for example, in Russian be regarded as compulsory, ev en though it is logically necessary to disambiguate temporal sentences.35 On the other hand, in Upper Sorbian the results of PFQ show that hizo 'already' is extensively used manifestly in experiential meaning. The use of 'already' could here be

468

Hannu Tommola

regarded as being due to German influence, for the German schon is typically used in experiential meaning (also, if not primarily, with the simple Past form). It may also appear paradoxical that of all the North Slavic languages, it is precisely Upper Sorbian, which has retained the perfect/simple past formal opposition in the tense system, that makes the largest use of 'already' as a perfect marker. As a matter of fact, neither the Sorbian nor the German Perfect are typical perfect grams, and the Upper Sorbian simple past tense forms are more or less literary narrative devices, whose main function is stylistic. In the sentences to be translated in PFQ, 'already' figured - in a kind of English metalanguage - four times (in the context descriptions additionally two times). In the English responses it was used four to five times, in the Scandinavian languages only three times, in Finnish five times, and so on. Therefore the considerable overuse of 'already' in some languages that do not have a tense qualifying as a perfect gram is important. Most striking is the amount of items with 'already' in Upper Sorbian (the word for 'already' used 24 times), but it is remarkable also in Hungarian (18), German (14), and Czech (14). Ukrainian (with 9 to 14 occurrences), Russian (9 to 12), Lower Sorbian (7 to 11 ) and Fering ( 10) are less obvious cases, but the difference in the basic PFQ sentences (and in the frequencies in "perfect languages") is still considerable.36

8. Conclusions The Slavic periphrastic perfect has in East Slavic become a synthetic (bound) form, while in West Slavic it has with some variation retained the auxiliary as an alternative means to indicate the subject in the nonthird person forms. The different stages that the individual languages have undergone in the development of the formal expression have, however, very little importance as far as the meaning of these forms is concerned. In all North Slavic languages the old perfect has ceased to be a perfect and become a universal past (as a matter of fact, also in Sorbian, even though simple past tenses exist). In the light of the cross-linguistic data about grammatical TA categories (Bybee et al. 1994: 105) this evolution is not unexpected: resultative -» anterior (= perfect) -> past. The only problem is the view maintained indirectly by some scholars that, for instance, in OES the perfect was only resultative, thus suggesting that the development proceeded directly from resultative to past. In section 6, an attempt was made to reject this claim - a claim which was, in fact, never formulated in modern terms oi cross-linguistic research. Because of another change in the TA system that presumably took place simultaneously with the loss of the perfect/past distinction, there are two pasts, a Perfective and an Imperfective. The very complicated process that still awaits explanation cannot be discussed here. It is possible that the perfective grams

On the perfect in North Slavic

469

N

are a further development of what Bybee et al. (1994: 54) call complétives. The imperfectivization and the rise of the binary aspect system is the most mysterious question in the development in Slavic. Iteratives are possible foregoers of the Imperfective verbs. Attempts to discover signs of new developments in the material of the modern North Slavic languages have not been very fruitful. However, imperfective pasts seem to expand into the domain of the experiential (perfect). Some signs of lexical rendering of this meaning were also found. Resultative constructions are not rare in North Slavic languages. They typically coincide with passive expression. Other than passive resultatives were found in the Russian, Czech and Upper Sorbian PFQ data. Trubinskij (1983,1984; see also Kuz'mina 1971) has shown that there are different tense systems in Northern Russian dialects. First, there are dialects, where a Gerund perfect is used (polpomyvsi 'the floor is / has been washed'). This is not a "general" perfect (anterior), but rather a resultative. Second, there are dialects that use the Passive Past Participle to denote the same type of resultative states. Third, a most interesting group of dialects shows a possessive construction with the Passive Past Participle that occurs in clear opposition to the "aoristic" use of the Perfective Past; i.e., there is a tense system as presented in (70): » (70) a. b. с d.

Perfective Past Perfective Perfect Imperfective Past Imperfective Perfect

on uexal и nego uexano onexal и nego exano

'he left' 'he has left' 'he travelled' 'he has travelled'

In those dialects where this kind of system prevails, the old perfect (the Past forma of the standard language) has really become a "doughnut" category. It no longer expresses the perfect meaning that was once its original and only meaning. One observation that may turn out to be significant is that several languages use the relational temporal adverb 'already' strikingly often. The extreme case is Upper Sorbian, and from the languages of the sample, the second in frequency of use of already' is a non-Slavic language, Hungarian, which also lacks a perfect tense or aspect. There is also a very high frequency in German and Czech, and, indeed, it is tempting to assume here an areal phenomenon. 9

« Addendum: The case of Kashubian (and Slovincian)

After I finished my work on this article Osten Dahl made me aware of the descripJ ons of past tense in Kashubian, one of the languages (spoken by ca. 150.000 per­ sons) forming the Lechitic group of West Slavic (the others being Polish, and the e xtinct Polabian). Kashubian is sometimes considered a dialect of Polish. Reasons

470

Hannu Tommola

exist, however, to assume a separate linguistic entity, which is supported by the interest some linguists, notably Friedrich Lorentz, have shown to this language from the 19th century onwards. Indeed, the linguistic identity of Slovincian dialects, separate from Kashubian, has been discussed (Lorentz 1903, Mikkola 1897). A chapter on Kashubian (Cassubian) is included in Comrie's and Corbett's volume on Slavic languages, while, for example, they choose to discuss Rusyn (Ruthenian) only together with Ukrainian (Comrie & Corbett (eds.) 1993: 996). There are three past tense forms in Kashubian: one synthetic, and two periphrastic constructions. In the absence of questionnaire data on Kashubian the actual usage of these forms remains somewhat unclear. Anyhow, two of the forms are regular exponents of the ProtoSlavic Perfect, one with and one without the auxiliary 'be' (cf. Sorbian and Russian, respectively, in section 3 above). They are both general past tense forms, the one with the auxiliary mostly used by older people. The third one is a periphrastic construction that involves the verb miec 'to have' and the passive past participle. Interestingly, neither Stone (1993: 777), nor Breza & Treder (1981: 133) discuss their possible specific semantic function. In fact, all the examples provided attest a perfect meaning. What is not commented on in these sources, either, although mentioned and evident from the examples given, is that, unlike all the other Slavic languages, this auxiliary combines also with the active past /-participle. Somewhat unexpectedly, Breza & Treder ( 1981: 133) remark about the Russian-type past (without the auxiliary) that it does not need to be a borrowing from German, whereas they do not say that the 'have'-construction apparently is a German caique. This is, instead, explicitly noticed for Slovincian by Lorentz (1903: 11), while also he ignores the idiosyncrasy of the construction combining 'have' and the /-participle. Lorentz speaks explicitly of the Perfect and Pluperfect considering these constructions both in Kashubian and in Slovincian (Lorentz 1925: 192; 1903: 302-303). But, without discussing the functional relationship between the different tenses, he also points out that the (standard North Slavic type) past tense forms of Kashubian are used with the perfect meaning as one of their functions ("fungiert als Perf. Präsens"; Lorentz 1925: 192). Thus, it seems to be the case that there are specific perfect forms in Kashubian, although they are not totally grammaticalized. Another type of Perfect, which is also a transparent German caique, is mentioned by Stone (1993: 777): In the case of intransitive verbs of motion, this tense is formed with the auxiliary bee 'to be' (instead of miec 'to have'). The participle then agrees in gender and number with the subject: Ta bialka je precz jidzonô. 'The (or that) woman has gone away.'

On the perfect in North Slav*

f7l

Acknowledgements I am indebted to Alexandr and Ljudmyla Dulicenko, Natalija Kozinceva, Jakub Lapatka, Helena Leheckovä, Helle Metslang, Fejsa Mihajlo, Madlena Norberg, Erlingur Sigurösson, Sonja Wölke, and Joanna Zach-Blonska for their kind help with their respective native languages. I also express my gratitude to Osten Dahl and Jouko Lindstedt for many useful remarks on earlier versions of the paper.

Notes 1. To some extent, the aspect distinction between the simple past forms is replaced by a new aspect opposition that is not restricted to past tenses. The whole TA-system has changed radically, which has had considerable influence, for example, on the expression of future time reference. In the new verb system the perfect/nonperfect distinction lacks systematic marking, and the original Imperfect/Aorist opposition is not quite rendered by the new Imperfective Past and Perfective Past opposition either. 2. Problems have arisen as researchers - especially of Old Russian - have used a set of semantic features to describe "the Perfect proper" that do not, as a matter of fact, comprise anything other than the resultative perfect (see section 6). 3. The interpretation suggested here is that the perfectivizing effect of the prefix (pro'through' = PFV) is cancelled by the involvement of a secondary imperfective suffix (-yra- = IPFV), while the lexical meaning of the prefix is retained. 4. A verb form is glossed Perfective in three different ways: (i) the prefix is marked as Perfective (as in s-dela-t' PFV-do-INF); (ii) the (semelfactive) suffix is marked as Perfective (either genuine momentary verbs like.pryg-nu-t' jump-PFV-INF, or Perfectives like otdox-nu-t' rest-PFV-INF); (iii) the lexeme as a whole is marked Perfective (e.g., prefixed verbs as vy-jt-i outgo:PFV-INF, or simplexes like skaza-t' say:PFV-INF). Imperfective simplexes are not explicitly glossed as such, whereas secondary Imperfectives are always marked out (as the suffix in vy-bras-yva-f out-throw-IPFV-INF, or the stem in vy-xodi-t' out-go:IPFV-INF). Otherwise lacking Perfective gloss is sufficient indication of imperfectivity. 5. The 1st and 2nd person clitics, though, are not necessarily attached to the main verb. Thus, at least mysmy/wyscie robili 'we/you did/have done' are possible instead of the more frequent robilismy/robiliscie, whereas analoguous forms in the 2nd person singular (tys robit) are rare, and the 1st person singular {jam robil) is obsolete (Joanna ZachBlonska, p.c.). 6. Whether the variant of Rusyn (also called Ruthenian) that our PFQ material reflects has to be regarded as a West Slavic (because of similarities, for example, with Slovak), or an East Slavic language (Ukrainian), or perhaps something else, is an interesting question of its own that will not be addressed here. Obviously, it is suitable to treat it as a North Slavic language. The material for PFQ was supplied by three speakers of the dialect

/; 472

Hannu Tommola

spoken in Vojvodina, Yugoslavia, and made available by Svenka Savic. There are more, at least twenty or thirty thousand, Rusyn speakers in Ukraine. 7. This pattern occurs, for example, in Armenian, Fering, German, Italian, Norwegian, Swedish, and possibly in Man (Cheremis). It is possible in English and Finnish, too. A notable exception is Modern Greek, where the Perfect is also used in PFQ:28. There is no variation in Lower Sorbian, either, and three Macedonian informants also use only the Perfect. In Icelandic a resultative construction is used consistently, not only in PFQ:2729, but also in PFQ:30-31. PFQ:29 was rendered in Icelandic:

Nei, nun er ekki ennbâ kom-in. no she COP NEG yet come-PP.F 'No, she hasn't / didn't [isn't] come back yet.'

•.«}
£)

"Specification and English tenses", Journal of Linguistics 2: 1-34. / Tense and Aspect Systems. Oxford: Basil Blackwell.

Grammatik der obersorbischen Schriftsprache der Gegenwart. Morphologie. Verfaßt von Helmut Faßke unter Mitarbeit von Siegfried Michalk. Budyäin: Domowina. Gorskova, Kflavdija] V[asil'evna] & G[eorgij] A[leksandrovic] Xaburgaev 1981 Istoriceskaja grammatika russkogo jazyka. Moskva: Vyäsaja äkola. Graves, Nina ., this volume "Macedonian - a language with three perfects". Havrânek, Bohuslav & Alois Jedlicka 1981 Ceskä mluvnice. Praha: Statnfpedagogické nakladatelstvf. 4. vydânf, prepracované (1st ed. 1959). Issatschenko [Isacenko], Alexander 1980-83 Geschichte der russischen Sprache, 1-2. Heidelberg: Winter. Istrina, Evgenija Samsonova 1919-21 "Sintaksiceskie javlenija Sinodal'nogo spiska Novgorodskoj letopisi", Izvestija Otdelenija Russkogo Jazyka i Slovesnosti 24: 2 [1919/1923], 1-172; 26 [1921/1923]: 207239. Ivanov, Valerij Vasil'evic 1982 "Istorija vremennyx form glagola", in: Ruben Ivanovic Avanesov & Valerij Vasil'evic Ivanov (eds.), Istoriceskaja grammatika russkogo jazyka. Morfologija. Glagol. Moskva: Nauka, 25-131. Jakobson, Roman 1948 "Quelques remarques sur l'édition critique du Slovo, sur sa traduction en langues modernes et sur la reconstruction du texte primitif", in: Henri Grégoire & Roman Jakobson & Marc Szeftel (eds.), La Geste du Prince Igor'. Épopée russe du douzième siècle [= Annuaire de l'Institut de Philologie et d'Histoire Orientales et Slaves, 8 (1945-1947)], 5-37. Janas, Pëtr 1976 Niedersorbische Grammatik (ßr den Gebrauch der Sorbischen Erweiterten Oberschule). Budysin: Domowina. Jespersen, Otto 1924 The philosophy of grammar. New York: Allen & Unwin. Johanson, Lars this volume "Viewpoint operators in European languages". Kallen, Jeffrey L. 1991 "Sociolinguistic variation and methodology: after as a Dublin variable", in: Cheshire, Jenny (ed.), English around the world. Sociolinguistic perspectives. Cambridge: Cambridge University Press. 51-74. Klenin, Emily 1993 "The Perfect Tense in the Laurentian Manuscript of 1377", in: Robert A. Maguire & Alan Timberlake (eds.), American Contributions to the Eleventh International Congress of Slavists. Bratislava, August-September 1993. Literature. Linguistics. Poetics. Columbus, Ohio: Slavica Publishers, 330-343. Knjazev, J[urij] P[avlovic] 1983 "Rezul'tativ, passiv i perfekt v russkom jazyke", in: V. P. Nedjalkov (ed.), 149-160. Kurylowicz, Jerzy 1965 Indoeuropejskie "perfectum" w slowianskim. Studia z filologii polskiej i stowianskiej, 5. Warszawa: PAN.

On the perfect in North Slavic

477

Kuz'mina, I[rina] B[orisovna 1971 "Predikativnoe upotreblenie pricastnyx form", in: I[rina] Bforisovna Kuz'mina & Elena Vasil'evna Nemcenko (eds.), Sintaksis pricastnyx form v russkix govorax. Moskva, 16223. Leinonen, Marja 1994 "Interpreting the perfect: the past as explanation", in: Susanna Shore & Maria Vilkuna (eds.), SKY 1994 [= 1994 Yearbook of the Finnish Linguistic Society]. Helsinki: Suomen kielitieteellinen yhdistys, 135-156. Lindstedt, Jouko 1994 "On the Development of the South Slavonic Perfect", in: Three Papers on the Perfect, 32-53 (EUROTYP Working Papers VI: 5). this volume "The perfect - aspectual, temporal and evidential". Lorentz, Friedrich 1903 Slovinzische Grammatik. St. Petersburg: Buchdruckerei der Kaiserlichen Akademie der Wissenschaften. 1925 Geschichte der pomoranischen (kaschubischen) Sprache. Grundriß der slavischen Philologie und Kulturgeschichte [1]. Berlin & Leipzig: Walter de Gruyter & Co. Maslov, Jurij Sergeevic 1949 "K voprosu о proisxozdenii posessivnogo perfekta", in: Ucénye zapiski LGU, No 97, serija filologiceskix nauk, vyp. 14: 76-104 (also in Maslov 1984: 224-248). 1983 "Rezul'tativ, perfekt i glagol'nyj vid", in: V.P. Nedjalkov (ed.), 41-54. 1984 Ocerki po aspektologii. Leningrad: Izdatel'stvo Leningradskogo universiteta. 1987 "Perfektnosf ", in: A.V. Bondarko (ed.), 195-209. 1990 "Perfekt", in: Victorija Nikolaevna Jarceva (ed.), Lingvisticeskij ènciklopedileskij slovar'. Moskva: Sovetskaja ènciklopedija, 372. Mikkola, J[ooseppi] Jfulius] 1897 "K izuceniju kasubskix govorov. I. Neäkol'ko zamëtok po kaäubskim govoram v sëverovostocnoj Pomeranii". Izvêstija Otdëlenija russkago jazyka i slovesnosti Imp. Akad. Nauk, II, kn. 2. Sankt-Peterburg. 400-428. Nedjalkov, Vfladimir] P[etrovic] (ed.) 1983 Tipologija rezul'tativnyx konstrukcij (rezul'tativ, stativ, passiv, perfekt). Leningrad: Nauka. Nedjalkov, Vladimir] P[etrovic] & Sergej E. Jaxontov 1983 "Tipologija rezul'tativnyx konstrukcij", in: V.P. Nedjalkov (ed.), 5—41. Niissalo, Nina 1994 Upotreblenie proäedsix vremennyx form v drevnerusskom jazyke Kievskogo perioda. Unpublished Master's thesis. University of Tampere. Slavonic Philology. Remneva, Marina Leont'evna 1988 Literaturnyj jazyk Drevnej Rusi. Moskva: Izdatel'stvo Moskovskogo Universiteta. Safranov, S. 1866 Russkij sintaksis. Riga: Izdanie knigoprodavca Kimmelja. Saxmatov, Aleksej Aleksandrovic 1941 Sintaksis russkogo jazyka. Leningrad: Uêpedgiz. Schooneveld, Cornelis H. van 1959 A semantic analysis of the Old Russian finite preterite system. The Hague: Mouton & Co. Sewc, Hinc * 96 8 Gramatika hornjoserbskeje rëce. 1. zwjazk. Fonematika i morfologija. Budyäin: Domowina.

478

Hannu Tommola

Sobolevskij, Aleksandr Ivanovië 1962 Lekcii po istorii russkogo jazyka. Izd. 4e, repr. S'-Gravenhage. [1st pub., Moskva 1907]. Smotrickij, Meletij 1619/1979 Grammatiki slavenskija praviinoe sintagma. Faksim. ed. by V. V. Nimêuk & V.M. Rusanovskij. Kiev: Nauka dumka. Stone, Gerald 1993 "Cassubian", in Bernard Comrie & Greviile G. Corbett (eds.), The Slavonic languages. London and New York: Routledge, 759-794. Thieroff, Rolf 1994 "Passives, Perfects, Resultatives, and Statives", in: Three Papers on the Perfect, 1-20 (EUROTYP Working Papers VI: 5). Tommola, Hannu 1981 "On the semantics of 'situations' and 'events'", in: Terminologie et traduction. Tome В [= Vaasan korkeakoulun julkaisuja. Tutkimuksia 80, Philologie 7]. Vaasa: University of Vaasa, 80-119. 1986 Aspektual'nost' vfinskom i russkom jazykax [= Neuvostoliittoinstituutin vuosikirja 28]. Helsinki: Neuvostoliittoinstituutti. 1993 '"Perfektnoe znacenie' v russkom jazyke", in: Jouko Lindstedt & Pekka Pesonen (eds.), Studio Slavica Finlandensia in Congressu XI Slavistarum internationali Bratislavae anno МСМХСШ oblata [= Studia Slavica Finlandensia, Tomus X]. Helsinki: Venäjän ja Itä-Euroopan instituutti, 134—141. Trubinskij, Valentin Ivanovic 1979 "O russkom razgovomom posessivnom perfekte", Severnorusskie govory, vyp. 3. Leningrad, 154-173. 1983 "Rezul'tativ, passiv i perfekt v nekotoryx russkix govorax", in: V.P. Nedjalkov (ed.), 216-226. 1984 Ocerki russkogo dialektnogo sintaksisa. Leningrad: Izdatel'stvo Leningradskogo universiteta. LTspenskij, Boris Andreevic 1987 Istorija russkogo literatumogo jazyka (XI-XVII vv.). Sagners Slavistische Sammlung 12. München: Verlag Otto Sagner. Vinogradov, Viktor Vladimirovic 1986 Russkij jazyk (Grammaticeskoe ucenie о slove). Moskva: Vyssaja skola. 3rd ed. (1st ed. 1947, Moskva: Ucpedgiz). Zitomirskij, K. G. 1915 Molox XX veka. [Pravopisanie]. Moskva: Trad.

Nina Graves

Macedonian - a language with three perfects?

1.

Introduction

Macedonian is a relatively small South Slavic, Balkan language which became a literary language only after the Second World War. For a long time it was considered to be a dialect or western variant of Bulgarian. The Macedonian literary language is based on the central dialects (Bitola - Veles - Prilep), but it has also liberally adopted forms from other dialects. It seems that the southwestern dialects are those most influenced by other Balkan languages, and thus are the gateway for accepting new forms. From these dialects forms spread to other dialects. The focus of this paper is on the perfect tenses of Macedonian and the fact that Macedonian grammars describe three different ways to form a perfect: 1. 2. 3.

The 'be' Perfect, formed with the auxiliary 'to be' 4- the Past Active Participle. The 'have' Perfect, formed with the auxiliary 'to have' + the Past Passive Participle. The Third Perfect, formed with the auxiliary 'to be' + the Past Passive Participle.

The aim of this article is to describe the characteristic use of each form in various contexts. The origin of these perfects will be briefly explained. Functional similarities and differences will be pointed out. The descriptions draw from both normative grammar and dialectal variation. Special attention is given to the southwestern dialect of Ohrid, since in this dialect the use of these three perfects is the most well-defined, with each one of them having its own characteristic semantic value. For this article I had seven informants, three of whom were assistants at the Department of Macedonian Language at the University of Kiril and Metodi in Skopje. All three are from Skopje and speak the a northern central dialect, referred to hereafter as a northern dialect. On the map they are indicated by the numbers 1, 2 and 3 - The other four informants are students of English language at the same university. They come from various dialectal areas: Kavadarci (number 4 on the map), Sveti Nikole (5), Prilep (6) and Ohrid (7).

480

Nina Graves

y—»

.6

7

_

fj

Map 1. Geographical distribution of the informants

2. The origin of different perfects in Macedonian 2.1. The 'be' Perfect The 'be' Perfect is formed with the auxiliary sum 'to be' + the Past Active Participle. This form is descended from Proto-Slavic, where the perfect was formed by the present of the auxiliary 'to be' (byti) + the past resultative participle, whose formation is possible only from the Aorist stem. There are two main formal differences between the perfect in Proto-Slavic (as evidenced in Old Church Slavonic) and present-day Macedonian. The first is that the past participle can nowadays be formed not only from the Aorist stem but also from the Imperfect stem. The second difference is that in current usage the auxiliary is lost in 3rd person singular and plural. There are three clear stages in the process of auxiliary loss: 1. 2. 3.

The auxiliary occurs in all persons in both singular and plural. Some kind of confusion arises in the use of the auxiliary in the 3rd person. The auxiliary disappears completely from the 3rd person forms.

The loss of the auxiliary first occurred in western dialects of Macedonian. From there it spread to all the other dialects until it recently became grammaticalized. Even nowadays, however, there are some dialects (e.g., Tikves) that are still in the second stage, and one can hear confusion in the use of the auxiliary:

Macedonian - »«language with three perfects ? (1) a.

b.

481

on e vikal he be:3SG shout:PPA:M 'He has shouted.' (Koneski 1986: 197-198) oni e vikale they be:3SG shout:PPA:PL 'They have shouted.'

The standard language uses no auxiliary in such 3rd person constructions; if it did, however, a plural auxiliary would be expected: oni se vikale. The 'be' Perfect is most widely used in northern and eastern dialects, which are most influenced by the neighboring language Serbian, where the 'be' Perfect is the only type of perfect. The 'be' Perfect occurs over most of the Macedonian linguistic territory, though it has entirely lost its use as a perfect at least in the dialect of Bobosica and some dialects of Kostur. There seems to be a similar tendency in such southwestern dialects as Ohrid. In these dialects the 'be' Perfect has been replaced by the 'have' Perfect (Koneski 1986: 197). This will be discussed below. The use of the 'be' Perfect in literature, especially in poetry, is quite limited. In prose, though, it is the most often used of the three perfect forms occurring in Macedonian. In a study of the various possibilities of translating the English Present Perfect into Macedonian (Arsova-Nikolic 1973: 157), it was shown that, in the majority of cases, the Present Perfect in English is translated into Macedonian by sum 'to be' + /-participle. The author considers this tendency to be due to the semantic and formal similarity of these two perfects, even though the English Present Perfect is in the form of a 'have' Perfect.

2.2. The 'have' Perfect The 'have' Perfect is formed with the auxiliary ima 'to have' + the Past Passive Participle, which is always neuter. It can be formed from both Imperfective and Perfective verbs. It is difficult to determine when this form of the perfect first appeared, but one of the earliest texts in which a form similar to the 'have' Perfect exists is a manuscript from the monastery of Krnino (now in the Kicevo region), written in 1706: ,:

>•'

Thus, the actual reliability of the information is not the issue, but rather whether the speaker presents the information as his own "responsibility". Note that variations in word order do not affect the interpretations. The following context illustrates possible grounds for the selection between PSTl and PST2 and the role of inference. The question presents a supposition based on the present state, and PST2 is normally used, as in (3a). PSTl can be used in the question if the action is presupposed and the question focuses on the actor, as in (3b). The answer is, however, in PSTl, as the speaker normally has direct evidence of his own actions. (3) a.

b.

c.

Komi (PFQ: 37) [It is cold in the room. The window is closed.] Te vos'tlömyd ösin'sö? you open:PST2 window:ACC 'Did you open the window [and close it again]?' ('perhaps, in my absence') Te vos'tlin ösin'sö? you open.PSTl window:ACC 'Did you open the window [and close it again]?' ('Was it you who ...?') Da, me vos'tli sijös. yes I openrPSTl it:ACC 'Yes, I opened it.' ('I did and was conscious of doing it.')

In the following example, the narrator, watching a man sawing wood, draws a conclusion about his past actions. Without the frequentative suffix, PST2 would simply mean that the action seems to be or must be completed: the frequentative affix invokes, as a pragmatic implication, an "experiential" meaning of the past tense. (4)

Komi (V. Beznosikov/Vojvyv kodzuv 5/1963) Tödts'ö: mortys pil'itts'yvlöma n'in, i, tydalö, n'e is_felt man:3SG saw:FREQ:PST2 already and it_seems not ötts'ydys'. once 'It is obvious that the man has sawed before, and more than once.'

Past tenses in Permic languages

499

If the present state shows a result that is the only evidence of a past event, PST2 is used even with temporal adverbials: (5)

Komi (PFQ: 14) [It is morning. A wakes up, looks out of the window and sees that the courtyard (or the street) is wet.] Vojnas zeröma. night:INESS.3SG rain:PST2 'It has rained at night.'

An equally unambiguous PST2 context is PFQ: 69, where one encounters a broken window and says, 'The thief has entered by this window'. Perhaps a typical example of how the past tenses work in Udmurt texts can be seen in the following sample from a story about a schoolboy called Mat'i. The main storyline, here represented by (6a), (d) and (g), is rendered with PST1, although "historical" present is common as well, as in (6c). We will later turn to the resultative form in (6b). Of immediate interest at this point are (6e) and (f), where the author has Mat'i infer a past event, the falling of snow, from its results, and use? PST2. The example also shows how PST2-type time reference can relate to a previous event as well as to a present one. No pluperfect is used here, as it would be in Finnish. (6) a. b. с d. e. f. g.

к

Udmurt (corpus) Tsukaz'ejaz vaz' ik sajkaz. next_morning early PRT wake_up:PSTl:3SG Anajez koskemyn in'i. mother:3SG leave:PART:INESS already Suzerez iz'e na. sister:3SG sleep.PRS still Mat'i ukno dory myniz: kytse kuaz'? M window to go:PSTl:3SG what_like weather Ujin lymy us'em, at_night snow fall:PST2 kud-og intyosy l'ukjos puks'ilTam. some places:ILL snowdrift:PL sit:PST2.PL Dis'as'kysa pijas kuasen urame potiz. dress:GER boy skr.INST streef.ILL go_out:PSTl:3SG '[Mat'i] woke up early the next morning. His mother is already gone. His sister is still sleeping. Mat'i went to the window: what was the weather like? It had snowed during the night, snowdrifts had formed in some places. The boy dressed and went out to the street on his skis.'

500

Marja Leinonen and Maria Vilkuna

The same is true for Komi: PST2 is inserted into the past event series just as easily as into the historical present:

(7)

Komi (Ëilina & Sorvaceva 1971: 53) Ses'a ji olöm mys't'i bara loktis joma-baba. Sie then PRT life after again come:PSTl witch she vetlöma kuz'n'its'aö. go:PST2 smithy:ILL 'After some time, the witch came again, (it appears) she had been to the smithy.'

Serebrennikov (1960) calls this use the 'impressive perfect', and notes that in some other languages it corresponds to the pluperfect. That is, the PST2 event precedes the past event currently in focus, leaving a resulting state that is in force at the stage that the speaker is referring to. In a narration cast in PST1, then, PST2 typically implies a "break" in the sequence of events. Things are different when PST2 is used in narration. This does not happen much in modern fiction, but it was the norm in the 1912 Udmurt New Testament (Svjatoe evangelie)3 and occurs quite often in Udmurt dialect samples, where elderly speakers are interviewed by students, and the main interests are tradition, past events and tales the interviewee has heard people tell during his life. (This is why Udmurt speakers sometimes characterize PST2 as pertaining to "old" events, even though the tense is not a remote past.) Personal experiences are rendered with PST1 (with frequent present tense and compound tenses). An example:

(8) a. b. с d. e.

Udmurt (Kel'makov & Saarinen 1994: 238)4 odik kysnomurtlen kartez vojne byrem. one woman:GEN husband:3SG war.TLL die:PST2.3SG solen bydes jurt nylpiosys kyl'ilT'am. s/he:GEN whole house child:PL:3SG remain:PST2.3PL okpol dzytadz'e t'at'azy bertem. once evening father:3PL return:PST2.3SG Van'zy s'is'kyny puks'ilTam. all:3PL eat:INF sit:PST2.3PL ogezlen vilkez us'em. one:3SG:GEN fork:3SG fall.PST2.3PL

Past tenses in Permic languages f.

50t

dzök ule mykyrkkem no, otyn, pe, iskal pydjos. table under bend_down:PST2 and there PRT cow foot:PL 'One woman's husband died in the war. She was left with a house full of children. One evening their father returned. They all sat down to eat. One of them dropped his/her fork. S/he bent down under the table and there, they say, were cow's hooves.'

The Komi PST2 is less common in narration, although fairytales and other past sequenced events can be presented as hearsay. In actual practice, PST1, historical present or future tenses are used, and PST2 indicates a break in the main line of narration. The Permic PST2 also covers such extensions of the notion of indirectivity as sudden revelations or reinterpretations of one's experience or action: something, surprisingly, turns out to be something else. Having supposed something to have been the case, one is distanced from one's conviction by direct perception or other people's opinions. The inferential of surprise is easiest to separate from resultative and non-witnessed meanings in atelic verbs in 1st and 2nd persons. An example:

(9)

Komi (Juskov 1970: 97) Mis'a, gaskö, te munin n'in. A tani na I_think perhaps you go:PSTl:2SG already but here still völömyd. be:PST2:2SG 'I thought you had already gone, but here you still are.'

Analogous Udmurt examples can be found in tales where the characters' honesty is tested (e.g., Kel'makov & Saarinen 1994: 276). In the following example, the main character of a folk tale, a rabbit, realizes his methods of fighting against his enemies have been misguided and says:

(10)

Udmurt (corpus) Eh, mon suz'i vylis'kem. oh I fool COP:PST2.1SG 'Oh, I have been foolish.'

Sudden revelation need not be evaluative; in the following, its object is the speaker's °wn unconscious action:

502 (11)

Maija Leinonen and Maria Vilkuna

"

9

Udmurt (Kel'makov & Saarinen 1994: 194) pöjasal ke, pöjasal, dyr, no, pöjalTas'kon trick:COND if trick:COND sure but trick_making pujyme gurte kel'tis'kem bag:ACC.lSG village:ILL leave:PST2.1SG 'I would indeed trick you, but I left my trick bag at home. [Said by trickster to a passer-by who has asked to see one of his tricks.]'

Recent research on evidentiality provides us with similar conceptual categories based on concrete linguistic forms. Thus, connected with indirect evidence is also the idea of the 'unprepared mind' (surprise, doubt) presented for evidentials in Turkish (Aksu-Koç & Slobin 1986: 1673) with an apt characterization: Normal experience is characterized by premonitory consciousness of the contents of coming moments [...] When a mind is unprepared, however, events cannot be assimilated at once. The speaker stands back, saying, in effect, by use of the [...] form: 'It seems that I am experiencing such-and-such' or 'It seems that such-and-such must have taken place.' [...] The event has become apparent through its consequences, or through verbal report; or the experienced event is radically different from the consciousness that preceded the experience. (Aksu-Koç & Slobin 1986: 164) 'Surprise' does not equal unexpectedness per se, but rather the 'unprepared mind', the speaker stating a fact that goes against possible preconceptions of an opposite state of affairs. Having noted that the 'sudden revelation' reading of PST2 is often most explicit in sentences with one of the speech act participants as subject, we should now take a closer look at 1st person PST2 in general, something that is perhaps too easily ruled out if one only considers the most concrete variants of indirectivity. Evidentiality in its "witnessed" sense is obviously not frequently associated with 1st person, as the speaker can usually be assumed to be conscious of his or her actions. There is indeed some controversy concerning the acceptability of PST2 1 st person forms in Komi; they do exist, but only in dialects (Cypanov 1992). In form, they are identical with the 3rd person verb forms s'etöma, s'etömaös', while Udmurt has developed a specific morphology in 1st persons. According to Batalova (1982: 141 ), the form is not found in Permyak. In Udmurt, the existence of the 1st person forms of PST2 is not in doubt, although they are infrequent. Examples were previously met in connection with "sudden revelation", but a speaker can also use 1st singular PST2 when reporting incidences during less than full consciousness, such as (12):

" Past tenses in Pennic languages (12)

503

Udmurt Mon ujbyrtis'kem ~ ujbyrtysa vetlis'kem. I sleepwalk:PST2.1SG sleepwalk.GER go:PST2.1SG 'I have walked in my sleep.'

A rather revealing narrative with "non-witnessed" PST2 can be found in Kel'makov and Saarinen (1994: 212). The speaker and his friend Igor had been teasing a ram. The story goes on: "The ram ran (PSTl) around and probably got tired (PSTl). Our ram was (PST2) wiser than us6: it ran (PSTl) towards me and hit (PSTl) its horns against my forehead." From this point on, the speaker goes on using PST2 for events during which he was unconscious: "I fell down immediately, but Igor ran away, frightened. I lay there a long time. My mother came from work and saw: I'm lying (PRS) in front of the shed. She quickly splashed me with water and took me into the house. But I still have (PRS) a mark on my forehead." Similar cases occur in Komi dialects. Baker (1983) points out that PST2 occurs if the speaker is somehow "distanced" from the event-does not remember it, perhaps did not take it into account. The examples are analogous to the Udmurt (10)—(12). The usual Russian translation resorts to the predicate okazyvaetsja 'it turns out'. The narrator's non-involvement is emphasized by Baker in the dialectal example Me vojnas ts'ets'yllama da pats'lö lontyllöma 'I must have got up in the night and fired the stove' (Baker 1983: 73-80). Lack of control of action on one hand, the state meaning on the other, in the 1st person forms in the dialects are pointed out in Bartens (1982). These nuances of meaning can clearly be accommodated under the heading of indirectivity. However, the Komi first person PST2 apparently need not be indirective, at least not in the spoken varieties. In the following section, we turn to the other facet of PST2, its resultativity, which distinguishes Komi from Udmurt.

3.

Resultativity

Cypanov (1992) reports further examples of the 1st person form from prose writers whose language reflects their dialectal background. The speaker merely reports on his present state, the result of a past action which he certainly must have been conscious of: (13)

Komi (Cypanov 1992: 26/B, Sahov, Vojvyv kodzuv 1990/8: 33) Tol'a, te talun stolövöjad ötnad vetly... Me Tol'a you today dining-room:ILL:2SG alone go.IMP I s'ojöma-juöma. eat:PST2.1 SG-drink:PST2.1SG 'Tolya, go to the diningroom alone today. I have eaten (and drunk).'

504

Marja Leinonen and Maria Vilkuna

In fact, the use of 1st person PST2 forms of certain verbs in certain contexts seems to be fairly normal in conversation. The speaker offers explanations, conclusions, motivations, as in the following example elicited from an informant: (14)

Komi Me uz'öma ~ uz'i n'in, tyrmas. I sleep:PST2 ~ sleep.PSTl already be_enough:FUT.3SG 'I have already slept enough.'

The range of Komi verbs that can be used with the 1st person forms without indirective meaning seems to be restricted. According to our non-conclusive elicitations, they signify either everyday activities-in addition to the above examples reflexives me myss 'öma, pyvs 'öma, pas 'tas 'öma T have washed myself, been to the sauna, got dressed', in the 'already, enough' context-or a few other resultative states, traditionally seen as participles, e.g., me ranitts'öma 'I am wounded'. Thus, it appears that the Komi PST2 has a use that should be characterized as Resultative in the sense of Nedjalkov & Jaxontov (1988): it signifies a state resulting from a past action, not past reference in a strict meaning. In principle, the construction is strictly intransitive, its subject representing the subject of an intransitive verb or the object of a transitive verb. In the latter case, the meaning is passive. Recall that the Permic past participle is neutral as to active or passive orientation (for this term, see Haspelmath 1994). We may compare the occasional 1st person "non-indirective" usage with Russian, where dialectal gerundial forms of generally perfective non-transitive verbs with the suffix -si are used as predicates in situations comparable to examples (13-14). Although the basic dialectal area where the forms are extensively used is rather far from Komi in the West, certain verb forms are widespread everywhere in Russia and are used in substandard urban speech as well. The most frequent verb forms, e.g., vypivsi 'drunk', odevsi 'clad', (ne) evsi '(not) eaten', (ne) pivsi '(not) drunk', (ne) spavsi '(not) slept' (Kuz'mina 1993: 142-146), are nearly lexicalized to signify a state - in the dialects where the form is used widely it has a clearly resultative meaning (cf. Trubinskij 1988). It would seem that in both languages, the exceptional verbs are either reflexive or transitive "active-oriented", i.e., the agent is saliently affected by the action (Haspelmath 1994: 161). Although the resultative meaning is based on the past participle form in both Komi and Udmurt, it is a point of difference between the languages. Udmurt does not resort to PST2 in cases like (13-14) but uses a specific resultative construction, whose predicate consists of the inessive form of the past participle. This form in itself denotes a present state and can be turned into a past by using one of the copular auxiliaries.

Past tenses in Permic languages

505

The two types of the resultative are illustrated below. In (15), it is formed from a transitive verb and takes on a prototypical passive meaning, as it is the patient argument whose state is predicated upon and therefore appears as the subject. In (16), however, the verb is lexically intransitive, and the only argument is the subject of the construction. The examples in (16) are translations of those Komi cases where the resultative meaning is at its clearest. (15) a.

b.

(16) a.

b.

с

r i *'

d.

Udmurt (corpus) Ukno us'temyn. window open:PART.INESS 'The window is open.' Ved' ta as'me udmurt kylyn goitemyn?! PRT this owmlPL Udmurt Ianguage:INST write:PART:INESS 'This is written in our own Udmurt language.' Udmurt Mon s'iis'kemyn. Köt tyremyn. I eaf.PART.INESS stomach fill:PART.INESS 'I have eaten. My stomach is full.' Mon dysetskemyn ~ iz'emyn n'i. I study:PART:INESS sleep .PART: INES S already 'I have already studied ~ slept (enough).' Mon mis'tis'kemyn, munts'oje vetlemyn. I wash(INTR):PART:INESS sauna.ILL go.PART.INESS 'I have washed, been to the sauna.' Mon ug kynmis'ky, mon sunyt dis'as'kemyn. I NEG:PRS catch_cold I warm dress(INTR):PART:INESS 'I won't catch a cold, I'm warmly dressed.'

The essential intransitivity of the Udmurt resultative construction is displayed by the fact that the corresponding form for Komi (13) cannot be formed from the usual transitive verb s 'iyny but from its intransitivized counterpart s 'is 'kyny 'to eat: to have a meal', as in (16a). As is to be expected, the Udmurt resultative requires that a change of state or affectedness can somehow be attributed to the subject of the construction. Thus, no examples of the form with the verb 'to sneeze' could be constructed. On the other hand, the affectedness interpretation is not always immediately obvious. Cases in Point are valamyn 'understood', todemyn 'known', or jaratemyn 'loved': that thing has been correctly understood, the thing is commonly known, or the person is much loved. By way of summary, it is instructive to consider the three following ways of saying °ne has caught a cold in Udmurt, using the verb kynmyny 'to get cold':

t

506

Marja Leinonen and Maria Vilkuna

(17) a. b. с

Udmurt PSTl: Mon kynmi. PST2: Mon kynmis'kem. Resultative: Mon kynmemyn.

,;

The PSTl variant places the event in a sequence experienced by the speaker: she is aware of the occasion when she caught the cold: T caught a cold'. The PST2 in (17b), with its 1st person subject, could mean 'I have caught a cold' with the implication T don't know when or where it happened'. Finally, (17c) is the normal way of reporting one's physical state: T have a cold'. At this point, it should be added that even Komi has a particular construction for the expression of resultativity. This is the impersonal form in (18) with the subject as a Possessor (Genitive case, sometimes also called Adessive); it appears to be more frequent than the simple PST2 construction in (14). For indirectivity or nonintentionality in the 1st person, the reflexive impersonal construction illustrated in (19) is used: (18)

Komi Menam uz'Öma n'in. I.GEN sleep:PST2 already T have already slept [no need for more].'

(19)

Komi (Ju§kov 1970: 175) Menam pöz'avs'öma ze. I:GEN sweat:RFL:PST2_3SG PRT T have become sweaty.'

Unlike the Udmurt resultative, however, the possessive construction also accepts transitive verbs. The patient may be marked as the surface subject as in (20b), or as the object, using the Accusative case as in (20c): (20) a. b. с

Komi (PFQ: 02) Sijö lydd'is n'in vot tajö knigasö. She read:PSTl already PRT that book:ACC So tajö knigays sylön lydd'öma n'in. PRT that book:3SG she.GEN read:PST2 already So tajö knigasö sylön lydd'öma n'in. PRT that book:ACC she:GEN read:PST2 already '[It seems that your sister never finishes books. - That is not quite true. ] She has read through this book (already).'

Past tenses in Permic languages

507

In Udmurt, this context systematically elicits PST1. Note that the Komi impersonal possessive construction is conspicuously similar to the North Russian dialectal construction with Accusative or Nominative objects, unvarying non-congruent past participle and Possessor (cf. Bartens 1982, Trubinskij 1988). Whether this similarity is a result of language contact remains unclear. In summary, the Komi PST2 names the result of a past event and, like the perfect of most Indo-European languages, refers predominantly to the moment of speech. It is resultative and/or indirective in specific contexts, though the lexical content of the verb predisposes the interpretation: telic verbs (2nd or 3rd person) produce resulting states, and one-argument situations with salient states do not leave much need for suppositions:

(21)

Komi (Leonik Palkin/Vojvyv kodzuv 5/1963: 36) Ton'a akan' kod': s'ödov jurs'isö mits'aa synalöma, Tonya doll like dark hainACC prettily comb:PST2.3SG kösa kyöma. plait plait:PST2.3SG 'Tonya is like a doll; her hair is prettily combed and plaited.'

Atelic verbs (notably 'to be'), on the other hand, produce no stable subsequent state, and thus bring forth the indirective interpretation, as in the following example (a). The same applies to telic verbs with agents which do not "carry" the subsequent state (b):

(22) a.

b.

Komi (Juskov 1970: 192) Germanly osjys'ömyd n'in. German:DAT boast:PST2:2SG already 'You have already been boasting to German (it turns out, as I hear etc.).' (JuSkov 1970: 40) Vera vajöma. Talun asyvnas mamys völi Vera bring:PST2:3SG. Today morning:INESS mother:3SG AUX1 pözalö, pyrali na ordö da. bake:PRS.3SG drop_in:PSTl.lSG they to when '[The heroine has brought the hero a pie to eat, he refuses because he already has some pies] Vera has brought (them). This morning her mother was baking, when I dropped in to visit them.'

к 508

Marja Leinonen and Maria Vilkuna

4. What about Perfect? Of the past tense forms in the Permic languages, it is obviously PST2 that is a can­ didate for the status of perfect. Being based on a past participle, PST2 contains some of the ingredients of the perfect, as it includes an anterior event resulting in a subsequent state (resultative) or a state of affairs inferred (indirective). The first seven questions in PFQ have been used as an operational test for the prototypical perfect (Lindstedt, this volume; see also Dahl 1985). Some confusion arises from the fact that so many of these are formulated in the first person, which rales out PST2 in practice. PST2 is always possible in these contexts, but at the cost of an indirective interpretation, which is not natural without further motivation. The first answer to the questions was always PST1, which supports the idea of PST1 as a default past tense. In Komi and Udmurt, the dimension of evidentiality thus clearly overrides the "present relevance" meaning characteristic of the prototypical perfect. In Udmurt, there is a possibility of solving the ambiguity presented by the Komi PST2, that is, indirective vs. resultative meaning, by choosing either PST2 or the resultative construction. Consider once more example (6). In (6b), the absence of the mother is expressed with the resultative, but in (6e, f), the appearance of snow is rendered with PST2. The difference between the two in this context is very small, but the resultative, which is nominal in form, provides a time-stable situation, while PST2 still refers to an event, a verbal concept. This is why speakers tend to experi­ ence the resultative variant as more remote or irreversible than the other one. One of the ingredients of perfect not treated so far is the so-called experiential meaning. It is of some interest that Udmurt has a special construction for this pur­ pose, which takes on some of the load of the prototypical perfect. This construction is not obligatory, though; questionnaire items like the following elicited PST1 an­ swers from informants, but when the Experiential construction was offered as an alternative, it was accepted. (23)

Udmurt (PFQ: 04) a.

b.

Ti mynes'tym apajme kukeno you(PL) I:ABL older_sister:ACC.lSG sometime pumitalT'ady-a? meet:PSTl-Q PumitalTamdy van'-a mynes'tym apajme? meet:PART:2PL COP.PRS-Q I:ABL older_sister.ACC.lSG 'Have you met my sister [at any time in your life up to now]?'

The Udmurt Experiential Perfect consists of the present or past tense copula and the past participle form of the main verb in a possessive construction; there is a

Past tenses in Permic languages

509

possessive suffix on the participle, and an optional genitive possessor-subject. Below, (24a) is an instance of the Experiential construction and (24b), of the simple 'have' construction. This is one of the many cases where some type of perfectivity is a grammaticalized 'have' construction. Metaphorically, at least, the connection is clear: 'You have a horse', and 'You have in your experience a past event of hearing about the revolutionaries.' (24) a.

b.

Udmurt (corpus) Tynad kylemed van'-a revol'utsion'erjos s'arys'? you:GEN hear:PART:2SG COP.PRS-Q revolutionary:PL about 'Have you heard about the revolutionaries?' Tynad valed van'-a? you:GEN horse:2SG COP.PRS-Q 'Do you have a horse?'

It should be made clear that, although this construction is quite similar to the Komi impersonal resultative, there is a difference that points to a different origin: in Komi, the existential copula is not used in Komi cases like (18). Moreover, Udmurt has other constructions consisting of the genitive, the past participle, and a "light" verb. For example, substituting the future copula luyny for the copula in (24a) yields the modal meaning 'can, is possible'. As the reader may have noted, the particle we have translated as 'already' (Komi n 'in, Udmurt (i)n 7) is fairly frequent in our examples. As 'already' is known to be a common source of the perfect (Bybee and Dahl 1989: 67-68), the question should be raised whether a similar grammaticalization process is going on or is to be expected in the Permic languages or whether the extensive use of the particle is due to Russian influence in a more superficial sense. However, its use is so wide and variable both in Komi and Udmurt that we must refrain from further discussion on this occasion.

5. Compound tenses Recall the two copular auxiliaries, völilval and völömlvylem. In Udmurt, and to a lesser degree in Komi, essentially all the tense and mood forms combine with the two copular auxiliaries, forming compound tenses with various interpretations, which are for the most part outside the scope of the present paper.7 However, we would like to conclude with a short survey of the remaining ways of referring to past events. A word on the syntactic structure of the compound tenses is in order here. An important feature of the Permic copulas is that they do not inflect for person or number like normal verbs, with the potential exception of AUX2 in simple predicative clauses ~ that is, outside the compound-tense system. An example of the copula in the 1st

510

Marja Leinonen and Maria Vilkuna

singular PST2 form was seen in (10) above. The Permic compound tenses are therefore structurally rather different from the Finnish and Germanic constructions, as can be seen by looking at (25a). It is the main verb that inflects - kalgis'ko is marked as both present and 1PL. The auxiliary remains the same when person/number changes. The following examples illustrate the difference between PST1 and the combination of present (PRS) and AUX in Udmurt. PST1 gives descriptions of situations and shows the events in sequence, and, as we saw, this can also be done with PST2. This contrasts with PRS+AUX, which here indicates a state functioning as the background of the subsequent action. (25) a.

b.

Udmurt (PFQ: 08) [Do you know what happened to me just an hour ago?] Mon n'uleskyn kalgis'ko val. I forest:INESS walk:PRS.lSG AUX1 T was walking in the forest.' Södtek sorys' kyj vyle l'ogis'kyny sed'ti. unexpectedly snake upon step:INF happen:PST1.1SG 'Suddenly I stepped on a snake.'

,

Varying the auxiliary adds the evidential dimension. As an extension of background states, PRS/FUT+AUX has a habitual use. This is common in the dialect texts when the interviewees describe the habits of earlier times. (Another, less frequent compound tense used for generic past is future with AUX, which is at home also in habitual uses.) PRS+AUX is also used for interrupted action: 'to be about to do something'. The habitual and "interrupted" uses are disambiguated by the context, as in the following: (26) a.

b.

Udmurt Jegit dyrjam, avtobus dugde val korkamy young time:ILL.lSG bus stop:PRS.3SG AUX1 house:lPL dory. at 'When I was young, the bus used to stop at our house.' Avtobus dugde val korkamy dory no, bus stop:PRS.3SG AUX1 house: 1PL at and koskiz. leave:PSTl:3SG 'The bus was about to stop at our house, but it went on.'

All these uses of PRS+AUX are found in Komi (KRS 1961, see völi; Serebrennikov 1963: 270-273), but actual occurrences are apparently less frequent. Serebrennikov

Past tenses in Permic languages

511

(1960) associates the Komi and Udmurt PRS+AUX pattern with the English past continuous, calling it 'durative, descriptive past'. Furthermore, Permic languages have a compound tense called "pluperfect" by Western scholars. It consists of PST2 and an auxiliary indicating the evidentiality dimension. The following Komi example shows that even the 1st person is natural here. The interpretation is purely resultative, but embedded in the past. (27)

Komi Kor sijö loktis, seki me völi sad'möma. when s/he come:PSTl then I be:PSTl wake_up:PST2 'When she came, I was/had woken up.'

In Udmurt, there are two choices: either PST2 (without apparent indirective meaning) or the resultative form may be used with the auxiliary. (28)

a. b.

Udmurt (PFQ: 76) [A's sister was not at home when A arrived. Question: Did you find your sister at home? A answers: No, I did not (find her).] So koskem val ~ vylem in'i. s/he leave:PST2.3SG AUX1 AUX2 already So koskemyn val ~ vylem in'i. s/he leave:PART:INESS AUX1 AUX2 already 'She had left (was gone).'

As a further complication, there is the combination of PST1 and AUX, which comes close to PST2+AUX in meaning, but is not identical to it. We will not discuss the meanings of the compound tenses any further, but one thing is clear: they are not narrative. The presence of the auxiliary always seems to signal a break in the narrative sequence, but how exactly this happens is a matter for future research.

6.

Conclusions

The central topic of this paper is the distinction between the two basic past tenses in the Permic languages. We have shown how two central semantic components of the perfect gram, viz. indirectivity and resultativity, can combine without giving rise to a perfect proper. This happens in Komi, whereas Udmurt makes a grammatical distinction between the two components. The clearest situation for the distinction is, naturally, when the speaker is the subject - the speaker knows best about his experience. It is for this reason that the 1 st person of PST2 has been considered to be non-existent in Komi. However, as the Udmurt examples amply prove, indirective meaning is far from excluded from the 1 st person, and the resultative interpretation of

512

Marja Leinonen and Maria Vilkuna

PST2 makes 1st person subjects even more natural. An essential difference between Komi and Udmurt is that the indirective undertones of PST2 cannot be escaped in the latter. Both languages have developed specific morphosyntactic means for the expression of resultativity. Beside the very productive resultative form, Udmurt also has an experiential construction and appears to resort to compound tenses more often than Komi.

Acknowledgements We are indebted to Jevgeni Cypanov, Valentina Ludykova and Sergei Gabov (Komi), Tatjana Krasnova, Aleksandr Skl'ajev and especially Bibinur Zaguljajeva (Udmurt) for sharing their native knowledge with us. The interpretations of the data remain our responsibility.

Notes 1. The Udmurt corpus was compiled by Pirkko Suihkonen and Bibinur Zaguljajeva. Approximately 9,000 orthographic sentences of prose (20th century fiction) are included, situated in the corpus server of the Department of General Linguistics, University of Helsinki. Examples from this corpus are marked "corpus". 2. Examples with no indication of source come from speaker interviews. 3. Predictably, the choice of past tense is a source of some debate in the ongoing Bible translation work (Marja Kartano, p.c.). 4. Some minor adjustments of dialect texts to the current transliteration system of the standard language have been made. 5. Note the particle pe indicating quotation or hearsay. This word and its Komi counterpart po combine with all tenses. 6. Note the "sudden revelation" use of PST2 here. 7. For example, the combination of imperative and AUX1 may be a polite request or a counterfactual advice of the type 'why didn't you . . . ? ' .

References

Aksu-Koç, Ayhan A. & Dan I. Slobin 1986 "A psychological account of the development and use of evidentials in Turkish", in: Wallace Chafe & Johanna Nichols (eds.) Evidentiality: The linguistic coding of epis temology. (Advances in Discourse Processes Vol. XX.) Norwood: Ablex Publishing Corporation, 159-167.

Baker, Robin W. 1983 "Komi Zyryan's second past tense", Finnisch- Ugrische Forschungen Band XLV, Hen 1-3 : 69-81.

in Permic languages Bartens, Raija 1982

513

"Die Dialektmonographien für das Komi-Syrjänische", Finnisch-ugrische Forschungen 44: 150-164.

Batalova, R. M. 1982 Areal'nye issledovanija po vostocnymfinno-ugorskimjazykam. [Areal investigations in Eastern Finno-Ugrian languages.] Moskva: Nauka. Bybee, Joan L. & Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51-103. Cypanov, Jevgenij 1992 "Dopolnenija к paradigme II prosedäego vremeni v komi jazyke" [Additions to the paradigm of the second past tense in Komi], Linguistica Uralica XXVIILl: 24-31. Dahl, Osten 1985

Tense and aspect systems. Oxford: Basil Blackwell.

Haarmann, Harald 1970 Die indirekte Erlebnisform als grammatische Kategorie. Eine eurasische Isoglosse. Veröffentlichungen der Societas Uralo-Altaica, Band 2. Wiesbaden: Otto Harrassowitz. Haspelmath, Martin 1994 "Passive participles across languages", in: Barbara Fox & Paul J. Hopper (eds.) Voice: Form and function. (Typological Studies in Language 27) Amsterdam/Philadelphia: John Benjamins Publishing Company, 151-178. KRS 1961

= Komi-russkij slovar'. [Komi-Russian dictionary.] Ed. V.l. Lytkin. Moskva: Gosudarstvennoe izdatel'stvo inostrannyx i nacional'nyx slovarej.

Kuz'mina, LB. 1993 Sintaksis russkix govorov v lingvogeograficeskom aspekte. [The syntax of Russian di» alects from the point of view of linguistic geography.] Moskva: Nauka. Lindstedt, Jouko this volume "The Perfect - aspectual and temporal and evidential." Nedjalkov, Vladimir P. & Sergej Je. Jaxontov 1988 "The typology of resultative constructions", in: V.P. Nedjalkov (ed.), 3-62. Nedjalkov, Vladimir P. (ed.) 1988 Typology of resultative constructions. (Typological studies in language 12). Amsterdam/Philadelphia: John Benjamins Serebrennikov B.A. I960 Kategorii vremeni i vida v finno-ugorskix jazykax permskoj i volzskoj grupp. [The categories of tense and aspect in the Permic and Volgaic groups of the Finno-Ugrian languages.] Moskva: Izdatel'stvo Akademii Nauk SSR. 1963

Istoriceskaja morfologija permskix jazykov. [Historical morphology of the Permic languages.] Moskva: Izdatel'stvo Akademii Nauk SSR.

àvjatoe evangelie ot Matfeja, Marka, Luki i Ioanna na udmurtskom jazyke 1912 [The Holy Gospel, according to Matthew, Mark, Luke and John in Udmurt.] Facsimile edition. Kazan': Central'naja tipografija. Trubinskij V.l. 1988

I

"Resultative, passive, and perfect in Russian dialects", in: V. P. Nedjalkov (ed.), 389410.

514

Marja Leinonen and Maria Vilkuna

Sources of examples

"

y

Juäkov, Gennadij 1970 Kuim tes. [Three comedies.] Syktyvkar: Komi kniznoe izdatel'stvo. Kel'makov, Valentin and Sirkka Saarinen 1994 Udmurtin murteet [Udmurt dialects]. Publications of the Department of Finnish and General Linguistics of the University of Turku No. 47 and Udmurt State University, Department of General and Finno-Ugrian Linguistics No. 2. Turku-Kevsk. Vojvyv kodzuv [Northern Star.] No. 5, 1963 A literary journal. Syktyvkar. Zilina, T. I. & V. A. Sorvaceva 1971 Obrazcy komi-zyrjanskoj reii. [Samples of Komi-Zyrian speech.] Syktyvkar: Komi kniznoe izdatel'stvo.

The Progressive

г

IS«,' Ц

I

-••'•••'

••'••••:

•;.?..;

в»^М:1ф-

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Kit-

w)

The progressive in Europe • 4:

1. Progressive: questionnaire and sample1 л 1.1.

Method

The progressive aspect is often identified with the imperfective aspect, rather than being treated as an aspect in its own right. When looking at the progressive aspect in the languages of Europe, one is inclined to take English as a starting point because English seems to be the preeminent language illustrating a fully grammaticalized progressive. On the basis of the use of the progressive in English, one could infer rules and relevant properties for describing "the" progressive in other languages. We decided to do otherwise. On the basis of a typological questionnaire (PROGQ), we established whether a language has a special form (which we call PROG) and whether such a form may be used in a number of specific cases. The informants for the different languages were asked to give the translation of the sentences after considering the contexts and extra information specified between slashes and square brackets. The first sentences were aimed at establishing whether a language has a dedicated PROG form or not. Consider the following sentences: (1)

(PROGQ: 1) [Somebody on the phone wants to know about Ann; the answer is: - Ann is next to me] She WORK [right now].

(2)

(PROGQ: 2) [A: What does Ann do every Saturday morning?] B: She CLEAN THE HOUSE/READ

The obvious difference between (1) and (2) is that in PROGQ: 1 the event is viewed as still in progress at the relevant moment (in this case, the speech moment), whereas in PROGQ: 2 no single moment is focussed on (the sentence indicates a habitual situation). In order to say that in a language there is a specific PROG form, it should be possible: (i) either to use a different form in these two sentences, or (ii) at least to have available in the first sentence an alternative form that is not avail­ able in the second.

518

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

A straightforward translation of sentences (1-2) into Dutch gives the following picture: (3) a.

Dutch (PROGQ: 1) Ze zit te werken. she sits to work 'She is working.' (PROGQ: 2) Dan maakt ze het huis schoon. then makes she the house clean 'She is cleaning the house then.'

:r

I

i 1•

i;

A construction as in (3b) can also be used to answer the question implicitly present in (1), whereas a construction as in (3a), based on a postural verb, would not be appropriate as an answer to the question that provides the context in (2). Consider: (4)

Dutch a. (PROGQ: 1) Ze maakt het huis schoon. she makes the house clean 'She is cleaning the house.' b. (PROGQ: 2) *Dan loopt ze het huis schoon te maken. then walks she the house clean to make 'She is then cleaning the house.'

On the basis of these observations one could consider Dutch to have a specific PROG form. Languages may also present special constructions, named "absentive" (ABSV) in De Groot (this volume), which are used to suggest that a given agent is remote from the deictic centre, performing a certain action. There are subtle analogies, as well as differences, between PROG and ABSV (cf. Section 3 below). In order to avoid the use of an ABSV instead of a PROG in the translation of sentences like PROGQ: 1, the context was specified with the locational phrase next to me. This type of locational phrase allows PROG but is incompatible with ABSV. As is well-known, languages may have a progressive/non-progressive opposition in other tenses than the Present, sometimes to the exclusion of the latter (cf. Section 2.8 below). For that reason, sentences similar to (1-2) were also provided with past time reference:

The progressive in Europe

019

(5)

(PROGQ: 3) [Last night at 8 o' clock,] when John came, Ann still WORK.

( 6)

(PROGQ: 4) Last year we [usually] CLEAN THE HOUSE on Saturdays [now we do it on Thursdays].

(7)

(PROGQ: 5) Last summer, John VISIT us three times.

Sentence PROGQ: 3 could contain PROG, whereas the other two sentences should not. PROGQ: 4 is a habitual sentence, while PROGQ: 5, by virtue of the numerical specification, is only compatible with a perfective viewpoint. As to PROGQ: 3, the following two assumptions were made: (i) If a language uses a PROG form in the present tense, it also allows the use of this j form in the past tense. I (ii) If a language does not use a PROG form in the present, the language may use it ', in the past tense. Ï In order to test the usage of the forms elicited, in fact to obtain some indication as I to their degree of grammaticalization, a number of sentences were concerned with !, the possible combination of particular grammatical categories, such as tense, mood, \ passive, causative etc. Section 2 briefly presents some of these topics. For a de| tailed report, cf. the chapters by Bertinetto, Ebert and Tommola in Part IV of this | volume.

1.2.

The sample

The potential sample of languages was the list of 150 languages of Europe, as established in the EUROTYP Guidelines. Since the investigation of the progressive started rather late in our project, it was not possible to get information on as many languages as we hoped. Questionnaires were only returned for the following languages: Albanian, Basque, Catalan, Danish, Dutch, English, Estonian, Fering, Finnish, Frash, Frysk, German, Greek, Hungarian, Icelandic, Irish, Italian, Kalmyk, Karachai, Maltese, Portuguese, Romanian, Russian, Rusyn, Serbo-Croatian, Spanish, Turkish, Swedish, and Ziiritüütsch.

13.

The data

For some languages, there was more than one informant. The answers of the different informants for one language sometimes showed considerable variation. This is due to the fact that most of the languages do not have a grammaticalized PROG. The

i

520

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

variation consists in the use of the Simple (imperfective) tenses in contrast to specific PROG forms on the one hand, and in the use of different PROG candidates on the other. Obviously, it would always be preferable to have a rather large number of informants for each language. The PROGQ consists of two parts (cf. Appendix). Part I contains 83 sentences to be translated. Part II, which is meant for linguists with a good knowledge of the language investigated, contains several questions related to grammar.

2. The morphosyntactic expression of PROG 2.1. Introduction Progressivity may be specified in several ways. See table 1 for a synoptic view of the morphosyntactic devices used in four major language families of Europe. Note that in this chapter we shall only consider morphosyntactic tools that are specialized for progressivity, or that have it as their main reading. We shall therefore leave out the progressive uses of simple tenses, although these are in some cases (e.g., in the Baltic Finnic domain) the devices most frequently employed to convey the idea of progressivity. One further possibility could consist in a specification outside the verb by means of an adverbial (such as at this moment). However, we did not take this kind of device into consideration, for it is quite unspecific. We only focussed on expressions related to the verb or the verb phrase. Thus, whenever we use the abbreviation PROG, this must be intended as referring to a specialized morphosyntactic device, rather than to the broad semantic notion of progressivity. In his typological study of the progressive, Blansitt (1975) proposed the following morphosyntactic classification: a. Affixal progressive markers b. Complex verb phrases as progressive signals i. verb phrases with a copula as auxiliary ii. verb phrases with a motion or postural verb as auxiliary iii. verb phrases with a pro-predicate (do-type) as auxiliary iv. verb phrases with a special progressive auxiliary verb If we apply this classification to our material, the languages of Europe mainly exhibit examples of complex verb phrases, the types (i), (ii) and (iv). In this section we will shortly discuss the morphosyntactic situation. A more detailed discussion of the data is given in the following chapters, dealing with subsets of the languages of Europe (Germanic, Romance, Baltic Finnic), as well as in the chapter on Maltese by Ebert.

The progressive in Europe

521

2.2. Affixal progressive Turkish is the only language of Europe which expresses PROG by means of an affix, the suffix -yor, as in: (8)

Turkish Cahsi-yor-du work-YOR.PST 'He was working.'

Some varieties of Arabic could be considered to have a préfixai progressive marker. Maltese, the only variety which falls inside the domain of the languages of Europe, uses an analytical form (cf. Ebert, this volume b).

' 2.3. Verb phrases with a copula as auxiliary One type of expression using a copula is the one where a copula combines with a Gerund. Examples of this type are Catalan, English, Italian, Brazilian Portuguese, Spanish, Kalmyk and Karachai. (9) a.

English Peter is writing a letter, b. Spanish ' Estaba hablando con una chica. wasilSG talk:GER with a girl T was talking with a girl.' с Italian Lei sta lavorando. she is work.GER 'She is working.'

We consider It. stare in expressions such as (9c) to be a copula, although this verb can be used as an independent lexical verb with the meaning of 'stand, be (in a certain location)'. In fact, in examples such as (9c) stare seems to have lost its lexical meaning. Nevertheless, it is fair to say that, except for some Central and Southern varieties of spoken Italian, this verb has not reached the full copula function typical °f Spanish estar, which can also be used with adjectival predicates. A second type combines a copula with an Infinitive or a related form, differently denominated in the various grammatical traditions (such as Supine or the like). Lan­ guages using this type of expression are Estonian, Finnish, Karelian, Sami, Livonian, and Vepsian. As can be seen from the list of languages, this type is limited to Finnic

i

522

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

languages. Consider for instance Finnish, which uses the so-called "third Infinitive" form with the inessive case ending: (10)

Finnish Minä ölen myy-mä-ssä lippuja. I am sell-3INF-INESS tickets.PRTV 'I am selling tickets.'

It is interesting to note that constructions with a copula followed by the Infinitive in Dutch, German, and Hungarian denote ABSV rather than PROG (cf. De Groot, this volume). Finnish (but not the other Finnic languages) is a special case, for the infinitival construction shown in (10) serves both purposes (cf. Sections 2.6 and 3^1 below; for Faroese, cf. Ebert, this volume a, ex. 43). A third type combines a copula with a prepositional phrase containing a nonfinite form of the verb. Examples of this are found in Breton (Hewitt 1985/86), Danish, Dutch, all varieties of Frisian, German, Icelandic, Italian, Portuguese, and Züritüütsch. Most of the prepositional phrases contain an Infinitive. They come in different forms (cf. table 1 for some details): (i) preposition + Infinitive (e.g. Breton, Icelandic, Italian) (ii) preposition + article + Infinitive (e.g. Dutch). (11) a.

b.

с

Icelandic Hun er aö vinna. she is at/to work:INF 'She is working.' Italian Lei sta a lavorare. she is at work:INF 'She is working.' Dutch Peter is aan het zwemmen. Peter is at the swimrINF 'Peter is swimming.'

Finally, the copula may also combine with lexical expressions of the type busy, be at work, be after, be under way etc. Examples are found, e.g., in Basque, Danish, Dutch, French, all varieties of Frisian, German, Swedish, and Züritüütsch, but they are sometimes reported as marginal types also in other languages (cf. table 1): (12)

Dutch Wim was bezig de stofzuiger te maken. Bill was busy the vacuum cleaner to repair 'Bill was busy repairing the vacuum cleaner.'

The progressive in Europe

523

2.4. Verb phrases with a motion or postural verb as auxiliary Expressions with a motion verb ('come' and 'go') as an auxiliary may sometimes be associated with progressive meaning. Note, however, that in a number of cases this type of construction indicates other meanings, such as ingressivity or near future (e.g. Dutch), future (French, Portuguese, Spanish), or past (Catalan). And even when it constitutes a true PROG device, it is often not interchangeable in the same contexts with the copular constructions. As a matter of fact, these constructions take on the "durative" rather than the "focalized" meaning of the progressive (cf. Section 3 for further qualifications). Examples of this are to be found in the Romance languages: Catalan, Italian, Portuguese and Spanish (as well as French, where it is, however, quite rare). (13) a.

b.

Spanish El calor venia durando demasiado the heat came last:GER too much 'The heat had been lasting too long.' Italian (Sardinian variety) (PROGQ: 52) Va dimenticando. goes forget:GER 'He is forgetting (names).'

Another type of motion verb ('walk', 'go around') is used in some Germanic languages (Danish, Dutch, and all varieties of Frisian). Consider: (14)

Dutch Marie loopt bloemen uit te venten Mary walks flowers out to hawk:INF 'Mary is hawking flowers.'

This type of expression is sometimes indistinguishable from the postural verb construction (cf. section 5 and Ebert, this volume a, section 2.5). The majority of the Germanic languages allow postural verbs as an auxiliary. Danish, Icelandic, Norwegian, and Swedish coordinate the postural verb with the finite form of the main verb. Consider: (15)

Swedish Han sitter och läser tidningen. he sits and reads newspaper-the 'He is reading a newspaper.'

°utch and all varieties of Frisian use a postural verb + infinitival marker + Infinitive:

524

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot ^

(16)

Frysk Ну sit in krante te lêzen. he sits in newspaper to read:INF 'He is reading the newspaper.'

2.5. Verb phrases with a special progressive auxiliary verb There are very few languages which use a dedicated verb as an auxiliary. Swedish and Yiddish are two examples. They express PROG with the help of a verb meaning 'hold': (17)

Swedish En ny kyrka hâller pâ att byggas. a new church keeps on to build'.PASS 'A new church is being built.'

2.6. Other types

rt

We have found other types of expressions which can be associated with progressive meaning. The first type is based on the use of a particle. A clear example of a language using a particle is Albanian. One way of expressing PROG in Albanian consists of the use of the emphatic particle po.4 Compare: (18) a.

b.

Albanian Laj. wash:lSG 'I wash.' Po laj. PTCL wash:lSG 'I am washing.'

We also consider Basque to use a particle, of the form ari. This particle, however, is originally the Infinitive form of a verb meaning 'keep doing something'. The use of this form requires the auxiliary izan which agrees with the subject, as well as the Imperfective Participle. For that reason, the auxiliary together with the form an could also be considered a compound verb (cf. King 1994: 383). (19)

Basque Zer idazten ari zara? what write:IPFV PTCL be:2SG 'What are you writing?'

The progressive in Europe

525

The other device we found is a syntactic device. In Finnish and Hungarian word order plays a role in the interpretation of PROG. In Finnish the use of the copula together with the "third Infinitive" with the inessive case ending usually expresses PROG. The same constraction may also express the ABSV. When an object is present, the two readings can be disambiguated by means of word order. Compare: (20) a.

b.

Finnish Minä ölen myy-mä-ssä lippuja. I am sell-3INF-INESS tickets:PRTV 'I am selling tickets.' (progressive) Minä ölen lippuja myy-mä-ssä. I am tickets.PRTV sell:3iNF:INESS 'I am off selling tickets.' (absentive)

Hungarian lacks any morphological marking for progressive. However, word order together with a specific intonation contour of the clause allows for a progressive interpretation (cf. Hetzron 1982, Kiefer 1994). Compare the following two examples, where < " > indicates focal stress and < > neutral stress: (21)

Hungarian a. Mari "le-vitte a bort, amikor csengetnek. Mary down-carried the wine when ring:3SG 'Mary carried down the wine when the doorbell rang.' b. ' Mari vitte le a bort, amikor csengetnek. Mary carried down the wine when ring:3SG 'Mary was carrying down the wine when the doorbell rang.'

2.7. Progressive and other "aspectual" distinctions PROG does not seem to be compatible with the so-called Perfective aspect, as present in the Slavic languages or Hungarian.6 In the sentences of the PROGQ where some form of PROG could be used, the Slavic languages tend to use the Imperfective (cf. fn. 36 in Bertinetto & Delfitto, this volume, for some detail). Hungarian might be considered an exception. In fact, the construction illustrated in (21b), allowing for the progressive interpretation, is based on the Perfective verb form, but the sentence does not express perfective aspect. The point is that this example lacks an element with focal stress to the left of the verb, which is a requirement for the perfective aspect to apply in Hungarian (the word order in (21a), on the other hand, necessar% involves a focalized element). Note, however, that Hungarian could hardly be considered to be a language with a genuine PROG construction.

526

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

The case of the Ibero-Romance languages is clearly different. The auxiliary estar has an imperfective and a perfective conjugation: both the imperfective and the perfective Past can be used with PROG (cf. Bertinetto, this volume; Olbertz 1998). Consider for instance: (22)

Spanish Estuve deambulando por las calles. was:SP:lSG stroll.GER about the streets 'I was strolling about the streets (i.e., for some time).'

However, these are aspectual distinctions of a different sort, which have little to do with the distinctions to be observed in Hungarian or in Slavic languages (cf. again fn. 5). Viewed from this perspective, example (22) does not represent an exception to the generalization above. Note further that Perfect tenses (generally expressed in European languages by the compound tenses), which constitute a particular subdomain of the perfective domain, may be associated with PROG devices in quite a number of languages (cf. below Sections 3 and 4, and Bertinetto, this volume, Ebert, this volume a, Tommola, this volume). Summing up, the progressive aspect is in principle compatible with both perfective and imperfective tenses, although it occurs most often with the imperfective ones. As to the languages presenting the distinction Perfective/Imperfective, PROG clearly favours the Imperfective predicates.

2.8. Progressive and tense distinctions Our material supports the assumption made by Blansitt (1975: 30) that there are never more tense distinctions in progressive than in non-progressive. On a broad typological scale, several languages allow a progressive/non-progressive opposition in the past tenses, but not in the present. For instance, the Hungarian construction presented in (21b) can only be found in the past. Another example comes from Lithuanian. While the Simple Present is ambiguously progressive/non-progressive, among the past and future tenses the simple forms of the copula buti combines with the Present Active Participle to form PROG (cf. Blansitt 1975: 20).

3. 3.1.

Typological observations Individuating the main types

The research conducted within EUROTYP made it possible to delineate, with a good amount of detail, a comprehensive picture of the typology of progressive constructions in European languages. However, since most of our data concern Western

The progressive in Europe

527

Europe, in this section we are going to restrict our observations to the Baltic Finnic, Germanic and Romance languages, where (as shown in Thieroff, this volume) the presence of specialized PROG devices is particularly well attested (but cf. Section 5 for a glance at Eastern Europe). Recall that, as stated in Section 2.1 above, the abbreviation PROG must be intended as referring to specialized morphosyntactic devices, rather than to the semantic notion of progressivity, for which the single languages may have at their disposal other grammatical tools. The different status of PROG in the various languages has obvious typological consequences. In languages like English, Irish, Icelandic or Maltese, it is quite likely that the respective PROG devices have reached a status of complete grammaticalization, considering that these are the only tools available to express the notion of progressivity. In other languages, however, the situation looks more complex. On the one hand, PROG constructions are not without competitors in that specific function, and not only in relation to simple imperfective tenses, but also with respect to the existence of more than one PROG device in one and the same language. On the other hand, these constructions may appear altogether to have a fairly low frequency in spontaneous linguistic usage. And here again the situation may vary. In Baltic Finnic, PROG seems to be used more in formal than in informal styles, whereas in the Romance and most of the Germanic area the reverse is true (at least as far as the most standard devices are considered). Table 1 is a synoptic presentation of the main morphological types to be found in the three groups considered here (cf. Bertinetto, this volume, Ebert, this volume a, and Tommola, this volume, for further qualifications concerning these data). Despite the multiplicity of meanings conveyed by these constructions in the various languages, three main types (or functions) may be isolated: (i.) "Focalized" progressive constructions (henceforth Foc-PROG), i.e. those expressing the notion of an event viewed as going on at a single point in time, here called "focalization point". The focalization point may be overtly expressed in the sentence, or else it may be recovered through the context, being the object of a presupposition. Needless to say, the focalization point does not exhaustively localize the event; it simply indicates a point in time overlapping the progressive event, while the actual duration of the latter remains indeterminate. (ü) "Durative" progressive constructions (henceforth Dur-PROG), i.e. those that are evaluated relative to a larger interval of time. Here again, however, the actual duration of the event remains indeterminate. Even when a durative temporal adverbial is present, this does not delimit the event but merely yields a vantage point from which the situation is observed. (üi) "Absentive" constructions (ABS V), i.e., those conveying the meaning of an event occurring in a place (characteristically reserved for a given purpose) displaced from the deictic centre.

528

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Table J. Morphosyntactic tools for expressing progressivity in four major language families of Europe A. GERMANIC languages, with the exception of English (from Ebert, this volume, with modifications)

Icelandic: Swedish Danish: Norwegian: Frasch (Wiidingh.): FeringOömrang: Frysk: Dutch: German: Züritüütsch:

Prepositional type

Postural type

Marginal types

er ad + INF hâlla pâ att + INF er ved at + INF

sit og + V sitter och + V sidder og + V

?

äs oont + INF

star oglligg(er)og + V sät toi sät an + INF

er i gang/ifœrd med at + INF driv(er) og + V äs bai tolän + INF

as uun 't + INF

sat tu + INF

as bi tu + INF

is oan V + INF is aan het + INF ist aml(beim) + INF isch am + INF

siet te + INF lit te + INF

is oan 'e gong mei + INF is bezig te + INF ist dabei zu + INF isch draa ze + INF

-

?

В. ROMANCE languages, plus English (from Bertinetto, this volume, with modifications)

Catalan: French: Italian:

copular AUX type

motion AUX type

Marginal types

ester + GER

anar + GER aller + GER andarehenire + GER

? être en train de + INF essere dietro a + INF

vir + GER

7

-

afi in curs de + INF

-

stare + GER stare a + INF Portuguese: estar a + INF estar + GER Romanian: Spanish: estar + GER English: be + V-ing

ir/andar/venir + GER?

-

-

С BALTIC FINNIC languages Estonian Finnish

olema + "V-ma +s be + SUPINE + INESS olla + V-mA +ssA be + 3rd-INF + INESS

NB: The same construction also exists in Karelian, Livonian, Vepsian (Eastern Baltic Finnic languages) and in Sami.

The progressive in Europe

im

529

D. ALTAIC languages Kalmyk Karachay Tatar

sim. converb type

participial type

-] bää-lyov-; CONV-SIM + 'be'/'move' -a tur-a torCONV-SIM + 'be, stand'

-Ja- -a (+ bää-) PCPL:IPFV (+ 'be')

-

NB: The same constructions as in Karachay and Tatar exist in most other Turkic languages, though functions vary somewhat from language to language.

The distinction between Foc-PROG and Dur-PROG is illustrated in particular in Bertinetto, this volume, with reference to Romance languages (cf. also Johanson, this volume, for the notion of focality), but it was already implicitly hinted at by Blansitt (1975).8 As to the notion ABSV, cf. De Groot, this volume. Suffice it to say that ABSV constructions exist in no fewer than eight European languages. Actually, in most cases ABSV devices differ morphologically from specifically PROG ones. However, in at least two languages (Finnish and Faroese) the two constructions coincide. It is therefore legitimate to examine this type in conjunction with PROG. In the final part of Section 4 we shall tackle the question of the proper relation between PROG and ABSV. To illustrate the issue, consider the following Finnish sentences, exemplifying the three main types, in the order given above (for Faroese, cf. Ebert, this volume a, ex. 43): (23)

Finnish a.

b.

Foc-PROG Kun lamppu putosi pöydälle, Hannu oli syömässä when lamp fell table:ALL Hannu was eat:INESS puuroa-nsa. porridge:PRTV-POSS:3SG 'When the lamp fell on the table, Hannu was eating his porridge.' Dur-PROG Viime tiistaina Nina oli pitämässä esitelmän Helsingissä. last tuesday Nina was keep:INESS talk Helsinki:INESS 'Last Tuesday, Nina gave a talk in Helsinki.'

530

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot С

ABSV Jouko oli lippuja myymässä. Jouko was tickets sell:INESS 'Jouko was off, selling tickets.'

Finnish is a somewhat exceptional language, for (as noted above) the same construction can be used in all three meanings. It is interesting, however, that this case exists because, as we shall see, there are a number of analogies between ABSV and PROG. Although we would not wish to claim that they are, in a broad perspective, typologically related, it is probably justified to put them side by side at least for comparative purposes. Indeed, it is often the case that both PROG and ABSV may occur in the same context, as in the following examples: (24) a.

b.

Italian PROG Aldo non è qui. Si sta allenahdo. Aldo not is here. RFL is training 'Aldo is not here. He is training.' ABSV Aldo non è qui. E' ad allenarsi. Aldo not is here, is at training:RFL 'Aldo is not here. He is off training.'

although, strictly speaking, they ultimately answer different questions ("What is X doing?" vs. "Where is X?"). By contrast, the opposition Foc-PROG vs. Dur-PROG does not always correspond to a morphological difference. The same constructions may often express both meanings. However, this is not the case with all PROG devices. For instance, the standard PROG constructions to be found in Standard Italian and French seem by and large restricted to the focalized interpretation. Conversely, some of the constructions to be found in certain Germanic languages (namely, those based on postural verbs) are most naturally interpreted in the durative meaning; and this is also true, in fact, of the Italian PROG variant built with the Infinitive instead of the Gerund (cf. Table 1)There is some evidence suggesting that the first examples of PROG constructions, in languages for which we have old testimonies, could be used in contexts which exclude a purely focalized reading. Consider the following texts, quoted from Dietrich (1973) and Bybee et al. (1994) respectively: (25) a.

Latin (Vulgata, John 10.40) [...] ubi erat Johannis prius baptizans [...] where was John before baptizing '[...] where John had previously been baptizing people [...]'

I

The progressive in Europe

531

Old English (iElfric, Lives, 1,11.52-55) Sume syndan creopende on eoröa mid eallum lichoman, swa some are creeping on earth with whole body, as swa wurmas doö. Sume gaö on twam fotum, sume on some on as worms do. some go on two feet, feower fotum, sume fleoö mid fyöerum. four feet, some fly with feathers 'Some (animals) creep on the^ earth with their body, just as worms do. Some walk on two feet, some on four, some fly with their wings.'

Of special interest is example (25b), for the PROG construction used there makes up (in Carlsson's 1978 terms) an "individual-level" predicate, i.e., a predicate exhibiting a "permanent" stative meaning, namely: 'some animals have the permanent property of creeping on the ground'. This suggests that even though these old constructions could appear in contexts, such as (26), which are compatible with a focalized reading, their overall semantic interpretation must have been different from the one that is available to their modern equivalents (note, however, that the focalized interpretation of (26), although very likely, is not necessarily the meaning intended by the writer):

(26)

Latin ( Vetus Latina, Coloss., 1.6) ... et est fructificans et crescens sicut in vobis ex qua and is fructifying and growing as in you since that die audistis et cognivistis gratiam dei in day hear:2PL:PST and learn:2PL:PST grace god:GEN in veritate." truth '... and it is giving fruit and growing in you since the day when you heard it and became acquainted with God's grace.'

Nevertheless, the existence of potentially focalized contexts indicates an easy line of development, which was taken by virtually all PROG devices, sometimes up to the extreme consequence of specializing as a purely focalized periphrasis, as in the case of the Italian Gerund PROG. By this, we do not want to suggest that the evolution undergone by the latter device shows the ultimate stage to be reached by these constructions. As is well known, there may be further stages (possibly reached at the end of an alternative developmental path), like the stage consisting of the complete ^interpretation of PROG as a general-purpose imperfective tool. See the following section for further comments on this.

532

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

3.2. Semantic and morphosyntactic classification Let us now have a close look at the data, with the aim of finding the analogies and the differences that concern the three main types listed above as (i-iii). Table 2 is an attempt at listing the behaviour of these three types with respect to a number of relevant features. It should be borne in mind that the labels "Foc-PROG" and "Dur-PROG" stand for broad semantic categories, rather than for independent mor­ phosyntactic constructions. Indeed, as we are going to show in Section 4, one and the same PROG construction may often correspond to more than one semantic cat­ egory. As to the correspondence between these broad semantic categories and the concrete manifestations of PROG, further qualifications will be provided in Section 4. The rest of this section will be a commentary on Table 2. The first cell of Table 2 relates to the locative content of the three types considered. As is well known, PROG constructions include, in one way or another, a locative morpheme. This may consist for instance of an auxiliary verb indicating existence or position (as in virtually every European PROG device), of an explicit marker of locativity (like the inessive case in Finnish PROG), or of a combination of more than one such morphemes (as again in Finnish PROG, which combines both of the above features). However, although the morphological structure of these constructions is based on a locative morpheme of some kind, the degree to which this meaning com­ ponent persists in each construction varies from case to case. It is obviously promi­ nent in ABSV devices, with the addition of an implication of remoteness/invisibility of the agent. This is indeed the distinctive feature of this construction. As to FocPROG, it should be observed that no residue of the original locative meaning persists in this type, as is proven by the possibility of using motion verbs, like in: Fred is go­ ing home. Note that motion verbs are admitted with the ABSV only when they indicate the specific activity that the agent is performing in the particular location in which s/he finds her/himself. For instance, German Hans ist laufen 'Hans is (away and is) engaged in the activity of running' could be said of somebody who regularly visits the running track and is currently away for that purpose. By contrast, verbs of directed motion, designating the locomotion towards the remote site, cannot appear with the ABSV. This restriction is absent with Foc-PROG, and this distinguishes it not only from the ABSV but, above all, from Dur-PROG, which is hardly compat­ ible with all sorts of motion verbs. This is notably the case with Germanic PROG constructions based on postural auxiliaries, which preserve to some extent their et­ ymological meaning (Ebert this volume a, Section З.1.). But even with Germanic constructions of the prepositional type (cf. Table 1), the availability of motion verbs is, generally speaking, quite weak (Ebert this volume a, Section З.1.). As to Finnish PROG, which has only recently been grammaticalized in the focalized meaning, it rejects directed movement verbs altogether, and employs instead nominal construc­ tions (Tommola, this volume).

'"\üh

:«J'Ï H «: The progressive in Europe

533

Table 2. Semantic properties (cells A-F) and morphological compatibilities (cell G) of the three types of periphrases considered in section 3: focalized progressive, durative progressive, absentive. The parentheses indicate that the given diacritic suggests a mere tendency, rather that a strong characterization. A. persistence of locative meaning compatibility with directed motion verbs remoteness / invisibility B. focalization point reading compatibility with non— durative verbs С. temporariness availability of determinate duration adverbials pragmatically defined duration D. compatibility with perfective tenses possible inclusive value of 'perfect' tenses E. compatibility with the habitual reading availability of interpretative meaning igentivity requirement compatibility with negation compatibility with stative verbs

FOC-PROG

DUR-FROG

ABSENTIVE

— +

(+) (—)

+ —

—

—

+

+

+

— —

+

0 —

+ +

(-)

—

—

..+

— —

+ +

+ —

(-)

+

+

(-)

+

—

+

-

(-)

+

(+)

—

(-)

(-) (-) (-)

(+)

-

+

j .

;

ompatibility with the Imperative ompatibility with the passive •ossible occurrence with the Infinitive Pontic modals' government

;

(-)

+

+

534

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Cell В of Table 2 refers to the features that specifically make up the focalized meaning (cf. point (i) of section 3.1 for an illustration of this notion). As shown in (23a) above, this reading presupposes that the focalization point be "familiar" to the speaker and the addressee. This may be obtained by means of a punctual temporal adverb (such as at 2 o' clock), or by a temporal clause of the appropriate kind which instantiates the so-called "incidential schema" , as in the example cited. This is a crucial property, for it shows that Foc-PROG has a purely imperfective aspectual value. Indeed, Foc-PROG is only compatible with imperfective tenses, such as the Imperfect in languages which exhibit an overt aspectual opposition among the past tenses. Dur-PROG and ABSV, by contrast, may easily be employed with perfec­ tive tenses, such as the Simple or Compound Past in Romance languages (cf. below for further comments on this). The impossibility of using perfective tenses with the Italian, French or Albanian PROG (the latter, more specifically, in the variant based on the particle po) is a decisive argument for ascribing these tools to the focalized type. It should be noted, in this connection, that the type of syntactic frame to which the incidential schema belongs enables a focalized reading even in languages where PROG is but weakly grammaticalized, like the Baltic Finnic languages (Metslang 1995, Heinämäki 1995, Tommola, this volume) or those, among the Germanic languages, that are less advanced in the grammaticalization process. Another typical textual situation which forces the focalized reading is "reportive" discourse, i.e., the kind of description that radio and television commentators make about live events. In such cases, the focalization point coincides (either explicitly or implicitly) with the speech time. Unsurprisingly, only Foc-PROG is compatible with non-durative verbs, which are instead ruled out in durative contexts such as those implied by Dur-PROG. In particular, achievement verbs (which constitute the bulk of non-durative predicates) may develop an "imminential" reading with Foc-PROG, suggesting that the event will soon occur although is not yet occurring at the focalization point. Consider wellknown examples such as: The grandfather was dying, or The train is leaving. A special class of achievements is constituted by "phasal" verbs such as begin, finish and the like, which again allow only for Foc-PROG (cf. PROGQ: 23-26). Interestingly, with achievement verbs Italian often tends to use an explicit PROG construction, although the Simple (imperfective) tenses may usually express the progressive reading by themselves (Delfitto & Bertinetto 1995). Apparently, in these cases the speaker needs to emphasize that the intended interpretation of the sentence is indeed progressive, despite the non-durative character of the verb. Conversely, not all Germanic languages may freely employ PROG in such contexts, and this is notably the case with the variant based on postural verbs (cf. Table 1). As to the ABSV, nondurative verbs may be employed, although they do not constitute the most common choice. An ABSV situation typically includes "adjacent" actions, besides the one specifically referred to. The structure of the event is in fact as follows: "go to remote

The progressive in Europe

535

location-perform the action-return to source location". Thus, even a momentary verb like cut the ribbon may be used (cf. Italian II sindaco è a tagliare il nastro 'The Mayor is off cutting the ribbon'). Cell С concerns the duration of the event. Several authors have claimed tempo­ ranness to be a relevant feature of the progressive. Indeed, it is even the case that in English some stative verbs (mainly of the postural/locational type) may be used with PROG precisely when the notion of temporariness is involved. Consider cases such as: (27)

(PROGQ: 59) The statue is standing in the garden, [i.e., for a limited period]

Note, however, that this restriction does not concern Foc-PROG, which can also be used in cases such as PROGQ: 60: The earth is turning around the sun, despite the permanent nature of the condition refened to. As to ABSV, temporariness is, of course, one of its defining features, as shown by De Groot, this volume. As a matter of fact, the ABSV implies the idea of a conventionally defined duration, tied to the expected time-window involved in performing the event. As to compatibility with the adverbials indicating "determinate duration", this is excluded with Foc-PROG for obvious reasons, but admitted by Dur-PROG, although the situation varies from language to language, and from adverbial to adverbial. It is a fact, however, that Dur-PROG may admit adverbials such as "from tx to t,", "for X Time", "until tx", "for the whole duration of ...", and the like. For some Germanic languages this is actually the preferred context for the use of the constructions based on postural verbs (Ebert, this volume, Section 3.2.; and this is equally true for the Romance construc­ tions based on motion verbs when perfective tenses are employed (Bertinetto, this volume). As to ABSV, however, the presence of "determinate duration" adverbials seems to be generally (but not necessarily) avoided. Apparently, since the duration of the event is pragmatically defined and easily recoverable, these periphrases tend to leave in the background the explicit indication of the time boundaries. Cell D is, from the point of view of the aspectual characterization, somehow sym­ metric to cell B. Just as a strictly imperfective interpretation is required for FocPROG, Dur-PROG and ABSV are easily accessible to perfective tenses.10 Consider a sentence such as: (28)

Spanish estuvo leyendo todo el dia. was:SP:3SG read:GER all the day 'S/he spent the whole day reading.'

536

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

A special case, among perfective tenses, is represented by those that express the aspectual value of "perfect", which in European languages often assume the form of compound tenses. When this aspectual value is instantiated, Dur-PROG may convey (with activity verbs) an "inclusive" meaning, suggesting that the event has been going on up to the reference time, and may possibly continue beyond that point (as in: I have been reading the whole day, /i.e., until the present moment/).11 Significantly, although the Standard Italian Foe-PROG, based on the auxiliary verb stare, cannot be used with compound tenses, the corresponding Dur-PROG constructions (those based on a motion verb followed by Gerund, or on stare a followed by the Infinitive) can. Cell E deals once more with the possible aspectual interpretation of the three types of constructions under discussion. Foc-PROG is generally unavailable to the habitual reading; it may tolerate it only in correlative structures of the type: whenever PERFECTIVE NON-DURATIVE, then PROGRESSIVE. This is no wonder, for these syntactic frames, besides making explicit the habitual meaning of the sentence, enable us to isolate a series of punctual temporal locations, each of which can work as a focalization point for the progressive event. By contrast, Dur-PROG and ABSV are often available to a habitual interpretation, even without the facilitation of an appropriate syntactic structure. The licensing contexts seem to be the same in both cases. Consider, as an illustration, examples (28-29) in Bertinetto, this volume, from which it also emerges that not all languages have equally easy access to habitual contexts. Somehow related to habituality, although distinct from it, is the so-called "interpretative" use of PROG (König & Lutzeier 1973; König 1995; cf. also Bertinetto this volume, Section 6.3.2). Although this use is quite marginal in most languages, it is fairly frequent, for instance, in English. Here is an example: (29)

If we selected the best described languages, we would also be selecting the languages with the largest number of speakers.

The reason for the analogy with habituality is that this use may be fostered by correlative frames such as: by doing X, you are (implicitly) doing Y, which are somewhat reminiscent of the correlative habitual frames cited above. However, since in interpretative contexts the first clause does not contain a non-durative perfective event, it cannot provide any sort of focalization point. Hence, the interpretative use is only possible with a durative reading. Cell F focuses on agentivity. In the languages where PROG may readily be used with a focalized meaning, this requirement plays no role. Consider: (30)

When I came, the sun was shining gloriously.

The progressive in Europe

537 /

Although this particular sentence may require an appropriate contextualization, it is perfectly acceptable in the relevant situations. By contrast, both ABSV and DurPROG are likely to be sensitive to agentivity, at least to some extent, although (as far as the latter is concerned) the strength of this restriction varies from case to case. It is specially strong in Baltic Finnic languages (Heinämäki 1995), whereas Germanic languages show a differentiated behaviour.1 Obviously, negation is inversely correlated with the notion of agentivity. Thus, it is no wonder that in Baltic Finnic languages and in most Germanic ones, i.e. in the languages where PROG fairly often exhibits a durative interpretation, there are restrictions on the usage of negation in progressive contexts. As to ABSV, the restriction concerning negation is fairly strong for pragmatic reasons but is not absolute. For instance, negation may present a contrastive value (cf. German Rolf ist nicht laufen, er ist einkaufen 'Rolf is away, his purpose being running rather than shopping'). As to stative verbs, they should be ruled out altogether. Indeed, this is often regarded as one of the most important defining features of progressivity. However, it turns out that stative verbs may sometimes combine with PROG, although the situation varies from language to language (Bertinetto 1994). But note that things differ for Foc-PROG and Dur-PROG. With the former, the stative verbs that appear in progressive sentences normally take on a non-stative meaning, with the marginal exception of some postural and locational English verbs (such as the verb stand discussed in (27) above), which preserve their stative character. Consider, for example, the copular predicates of PROGQ: 42-43, where the use of PROG (available to most Ibero-Romance languages as well as English) forces the activity interpretation in these intrinsically ambiguous predicates. To illustrate, John is being clever hints at John's temporary behaviour, rather than to a permanent characterization of his. Consequently, these are not true counterexamples to the regularity stated above. By contrast, Dur-PROG does not necessarily exclude stative verbs. This is shown in particular by languages where the relevant construction is at a very early stage of grammaticalization, such as Estonian (Metslang 1995; Tommola, this volume). And this, as observed above in relation to example (25), seems to be a prominent fact in the diachronic evolution of PROG. Finally, cell G reports data concerning the compatibility of PROG and ABSV constructions with some relevant morphological categories. To start with, consider their compatibility with the Imperative, which is often quite weak. The languages which allow this use tend to be those that present a fully grammaticalized PROG device, like English and Catalan, although this feature may surprisingly appear even in Romanian, a language virtually without PROG (Bertinetto, this volume). Our data suggest that the Imperative is most probable in durative contexts, although it may also appear in focalized ones, such as PROGQ: 73 {Be working when the boss returns!). On the other hand, the Imperative is definitely ruled out with the ABSV. Compatibility with the passive is also a clue to an advanced stage of grammaticalization. Apparently,

538

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

i

this possibility is again excluded with the ABSV.13 Essentially the same applies to the use of the Infinitive with PROG. To be exact, we refer here to contexts where PROG takes the Infinitive as a consequence of being syntactically governed by a predicate governing this mood.14 Note that compatibility with the Infinitive seems to arise before compatibility with the Imperative and the passive. For example, the Infinitive is at least marginally possible in Italian, where the passive and the Imperative are not at all available. Once more it should be noted that this morphological feature is mostly to be expected in durative contexts. One peculiar case of Infinitive PROG is constituted by sentences where this construction is governed by modal verbs, as in PROGQ: 79-80. The interesting fact here is that only Dur-PROG admits the deontic interpretation, although these cases seem to exist only in informal registers. Normally, the only reading available with Foc-PROG is the epistemic one, as in: Anne must be feeding the animals, I guess (Bertinetto, this volume, Section 3.1).

4. A diachronic sketch 4.1. Stages of development Let us now consider the diachronic issue. We shall first examine the problem of Foc-PROG and Dur-PROG, delaying the question of the ABSV (namely of its relationship to PROG constructions) until the end of this section. As to Foc-PROG and Dur-PROG, one could in principle defend the proposal that both types of periphrases can constitute the original nucleus, depending on the language. However, there is some evidence that Dur-PROG represents the original stage. Consider first the case of Italian. We know for sure that this language underwent a dramatic diachronic development. The PROG construction based on the auxiliary stare is now used exclusively (or almost exclusively) in focalized contexts, but in the early stages it could also be used in durative contexts (Bertinetto, this volume, Section 2). This invites the conclusion that Dur-PROG instantiates the original meaning of this construction, and this seems to be true in other cases. The PROG constructions of Ibero-Romance languages and English may cover both the focalized and the durative meaning, but we have good evidence that the forerunners of these periphrases, in Latin and Old English, were readily available in contexts which presented a purely durative, or even stative character (as shown by examples (25) above). It is thus reasonable to suppose that the use in focalized contexts was the result of an expansion of the original durative interpretation. As to Germanic languages other than English, one can say with good approximation that the PROG constructions based on postural verbs seem to be especially appropriate for durative contexts', and even the prepositional constructions, which are more flexible in use, often undergo restrictions in focalized contexts (Ebert, this volume a). Putting this together,

The progressive in Europe

539

it appears that in all these cases the original meaning of PROG constructions seems to have been the durative one. This is not true in every case, though. Among the European languages for which we have reliable data, the only ones, besides Italian, that present a PROG construction exclusively restricted to the focalized meaning are French and Albanian. Now, the French periphrasis "être en train de + INF" started out as an intentional construction, and only towards the beginning of the 19th century was it eventually specialized in its current meaning (Gougenheim 1929). There is no evidence that this periphrasis, in its progressive interpretation, ever played the role of a durative device. The same applies to the Albanian construction based on the particle po, whose meaning is close to German gerade.15 However, it seems rather implausible that the French and Albanian constructions, due to their inherent constitution, will ever expand their use to typically durative contexts. Therefore, the exceptions provided by these two languages only prove that some PROG constructions can directly appear as focalized devices. Until we find clear evidence of a Dur-PROG construction that started out as a purely Foc-PROG device, we are justified in assuming that the diachronic development, when there is one, goes in the direction indicated above. This invites the hypothesis that, putting aside the exceptions represented by French and Albanian, the original meaning of most PROG devices must have been that of a stative construction, expressing the idea of "being (i.e., finding oneself/itself) in a state", as is especially clear with the forerunners of Romance and English PROG constructions, as noted above in relation to examples (25). The purely dynamic (or processual) meaning, which is particularly salient in the focalized type, is in most cases a later development, attained at the end of a rather lengthy grammaticalization process. In other words, it appears that most PROG constructions started out as "actional" periphrases, rather than truly "aspectual" ones. The complete attainment of the latter status corresponds to the stage of full grammaticalization. We may thus hypothesize that, as far as their meaning is concerned, the most typical PROG constructions possibly underwent the stages of development shown in Table 3. Since the historical data are not equally clear for all the languages belonging to the three groups considered in the preceding section (Romance, Germanic and Baltic Finnic), the observations that follow will deal mostly with Romance, with just a few hints to the other groups. The relevant data derive from Bertinetto, this volume. Stage (i) is linked to the inherent meaning of the morphemes employed. At this stage, the non-finite form of the verb accompanying the locative or postural or motion verb presents a purely "con-verbal" meaning.16 Stage (ii) corresponds to the initial stage of grammaticalization, in which the locative (or postural or motion) verb begins to turn into an auxiliary, while the non-finite form (or the verbal noun) is gradually promoted to the status of head of the construction. At this stage, the semantic bleaching of the auxiliary may begin, although this process is completed only at stage (iii). Stage (iv) shows a further development: the context must provide the

540

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

'

Table 3. Diachronie development of progressive constructions in Romance (from actionality to aspect) (i)

pure locativity

=

(ii)

progressivity I

=

(iii)

progressivity II

=

(iv)

progressivity III

=

(v)

pure imperfectivity

=

Stative, durative (ex.: the meaning to be observed in some Latin examples) residually locative, durative, accessible to perfective aspect (ex.: PROG periphrases based on the verb 'come', which preserve some kind of deictic orientation) durative, accessible to perfective aspect (ex.: PROG periphrases based on the verb 'go') focalized, strictly imperfective (ex.: Modern Italian "stare + Gerund") loss of the progressive character (ex.: possibly to be observed on some non-standard varieties of Latin American Spanish)

explicit (or presupposed) indication of a single focalization point, which gives PROG its prototypical character. At this stage, the availability of perfective tenses is totally lost: the transformation from an actional to an aspectual device is completed. Interestingly, French PROG (whose origin is quite different from that of most PROG constructions) entered directly at this level, bypassing all previous stages. Finally, stage (v) is a possible future development, not yet attained by any Romance language. This step, which has been taken by many European and non-European languages, consists in the eventual reduction of PROG constructions to a purely imperfective form; i.e., a form not restricted to progressive contexts, but appearing also in habitual ones, thus behaving like a typically general -purpose imperfective tense such as the Romance Imperfect. To the European examples quoted by Johanson, this volume (section 7.8), one may at least add Igbo and Yoruba (Comrie 1976: 99-101), as well as Punjabi and Urdu (Dahl 1985). Obviously, at this final stage the restriction concerning stative verbs (cf. cell G of Table 2) is dropped altogether. The diachronic reconstruction sketched in Table 3 is not intended to suggest that each PROG device to be found in a given language should correspond, at any given moment, to a single and definite stage. This may be true in particular cases, some of which have been mentioned above (recall e.g. French), but it is easy to point out PROG constructions whose usage covers more than one stage of Table 3. Consider for instance English PROG, or the PROG periphrases based on copular verbs in Ibero-Romance languages, or the prepositional type available in a number of Germanic languages (cf. Table 1). Although the degree of grammaticalization varies (being highest in English and Ibero-Romance languages), all these constructions embrace stages (iii) and (iv) at the same time, i.e. they have steadily reached the focal-

The progHWsive in Europe

541

ized stage, without losing ground as durative devices. This may be seen, for instance, in the following English sentence, which may be read in the durative meaning: (31)

[A scout, pointing to a series of paw prints, says:] A grizzly was walking here!

The reading that is relevant to our present purpose is the one whereby the scout intends to convey the idea that a grizzly trampled the ground in that place for a certain period of time, as indicated by the traces on the ground. According to this interpretation, there is no focalization point whatsoever. The same result may be obtained in Ibero-Romance languages by combining PROG with a perfective Past (e.g., Spanish "estuvo + Gerund"). By contrast, Italian gerundial PROG would be absolutely out of place in a similar context.18 In fact, the span of meaning of each PROG device may be even larger than two stages (cf. again Table 3). This is, for instance, the case with Estonian PROG, which is but weakly grammaticalized (Metslang 1995; Tommola, this volume). On the one hand, this construction may at times suggest the persistence of a locative meaning (stage ii); on the other hand, it seems to appear more and more often in focalized contexts (stage iv), possibly under the external influence of other languages such as Finnish (and perhaps English).

4.2. Progressive vs. absentive The inspection of Table 2 may suggest to us some plausible conclusion concerning the position of ABS V in comparison with PROG. In general, there seems to be little doubt that the ABSV resembles PROG on semantic grounds, as is also proven by the interchangeability of these devices in certain contexts (cf. again example (24)). However, this does not necessarily involve the evolutionary domain. The situation is in fact fairly complex. As may be seen, cell A of Table 2 indicates some kind of solidarity between Dur-PROG and ABSV in contradistinction to Foc-PROG, but cell G presents more of a solidarity between Foc-PROG and ABSV, and the remaining cells exhibit a rather variable situation, where the behaviour of the ABSV approaches one type or the other of PROG, depending on the feature considered. Actually, considering the general characterization of the ABSV as provided in De Groot, this volume, we should expect the relationship to be tighter between Dur-PROG and ABSV than between Foc-PROG and ABSV. However, the fact that Finnish and Faroese are, among the languages known to us, the only ones where PROG and ABSV can be expressed by the same morphosyntactic construction invites the conclusion that an evolutionary link between these grammatical devices may indeed exist, but is un"kely to be the rule.

542

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Let us consider more closely the case of Finnish. As a PROG device, this periphrasis seems to be at a more advanced stage of grammaticalization than its Estonian equivalent, yet it basically spans the same variety of uses. The interesting question is whether this construction started out as ABSV or as PROG. Only a thorough historical investigation could settle the matter. What is clear, though, is that this construction presents some residue of locativity, for it rejects motion verbs such as tulla 'come' or mennä 'go', which have to be replaced by the corresponding nouns (tullo, menno). Now, we observed in the discussion of cell A of Table 2 that locativity is a prominent feature of ABSV constructions. Another prominent feature is agentivity. It is then interesting to note that Finnish PROG, even when used as a true progressive construction, obeys an agentivity constraint, or at least a constraint of intentional control of the event (Heinämäki 1995).19 For example, a sentence such as Kynttilät on palamassa '(the) candles are burning' may only be uttered when one wants to suggest that somebody has lit the candles, with implicit recovering of the hidden agent. Thus, the connections between the two main functions of this periphrasis (PROG and ABSV) appear to be rather evident. But this does not solve the diachronic puzzle. The only thing that may plausibly be said is that the Finnish periphrasis must have entered the developmental path sketched in Table 3 no later than stage (ii). In fact, there is even ground to hypothesize that it actually entered at stage (i), for it may be employed with the prototypical stative verb olla 'be' to convey the meaning of existence, as in: On olemassa toinen mahdolUsuus 'there exists another possibility' (lit.: is be-INF-INESS other possibility). Interestingly, the morphologically identical Estonian periphrasis may be used with an even larger group of statives (Metslang 1995; Tommola, this volume). In any case, one should not forget that the situation of Finnish and Faroese is rather exceptional. In all other languages in which an ABSV construction has been described there is no morphological coincidence of ABSV and PROG. There is even ground to believe that in some languages, like German and Italian, the ABSV has arisen out of the mere deletion of the Participle, as in: Er ist einkaufen gegangen -+ Er ist einkaufen, È andato a comprare —> È a comprare 'he is shopping' (lit.: he is gone (to) buy).20 Thus, in most cases the semantic proximity of ABSV and PROG, as it emerges in Table 2, although not accidental, may be the effect of the overall similarity of the contexts in which these grammatical devices appear, rather than the result of a true evolutionary convergence.

5. A glance towards Eastern Europe: Altaic languages The investigation of progressives, which was started rather late in the Tense-Aspect Group, concentrated on the better known European languages. As progressives have not been much investigated for European languages except English, the inclusion

The progressive in Europe

543

of lesser known languages depended on the availability of some expertise. Besides for the languages treated in the individual articles, we received questionnaires for Basque, Armenian, Kalmyk and Karachai. We have nothing new to say about the first two, but we do want to present some data from Kalmyk and Karachai for the following reasons: Kalmyk uses four different forms (besides a general imperfective and a habitual), where English has one and other Germanic languages have two forms. Karachai is one of the languages for which one form (-ib tur-) has been claimed to have both progressive and perfect meaning, which gave rise to a corresponding question in the theoretical part of the PROGQ (II, lj).

5.1.

Kalmyk

As little is known about the function of Kalmyk TA-forms and as the terms used in the available descriptions are of little help, the questionnaire data were difficult to evaluate. The most useful source was Biasing (1984); without his numerous text examples the interpretation of the questionnaire data would not have been possible. The forms to be briefly discussed here are (for a more detailed analysis see Ebert, to appear): our term IPFV PROG I PROG II

traditional name (1 Präsens I Präs. durativum I2' Präs. durativum I

Continuative Durative

Präsens III Präs. durativum II

-na -Ja- (contraction of PROG II?) -J bää-/ jov- (simultaneous converb + 'be'/'move') -a (bää-) (ipfv. participle + 'be') -ad bää-/ jov-/ suu-/... (anterior converb + 'beV'move'/'sit' ...)

The form used most often in sentences where we expected a progressive form is PROG I (with the Imperfective suffix in present time contexts). It occurs, for instance, in PROGQ: 1, 6-17, 19-27 (cf. Table 4): - with agentive verbs: terkodl-Jana 'she is working' (PROGQ: 1) - with non-agentive verbs usn busl-jana 'the water is boiling' (PROGQ: 37) - with motion verbs: har-cana 'is going out' (PROGQ: 21) ~ with phasal verbs: küce-jänä 'is finishing' (PROGQ: 26) - with temporary states: iiiidn xoornd zogs-jana 'is standing by the door' (PROGQ: 58) « was not used with stative verbs or with the stative component of initio-transformative verbs (med-ne 'knows' - PROGQ: 39), and not with limiting temporal adverbials (PROGQ: 48). -fa combines with finite and nonfinite ТАМ suffixes, e.g.,

544

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

(32)

Kalmyk (PROGQ: 70) ... Martin sad-t naad-ja-la Martin garden-LOC play-PROG-ANT '... Martin was playing in the courtyard' (PROGQ: 30)

a.

b.

... Pjotr colu terz-d haj-/a-sn cag-t P. Peter window-LOC throw-PROG-PART time-LOC '... when Peter was throwing the stone at the window' The form -/ +bää- is seldom used and does not figure in the PROGQ. However, if motion is involved, a PROG II periphrasis with the verby'ov- 'go' is preferred: orad kür-с jovna 'is reaching the top' (PROGQ: 31). (The verb for 'go around, be in motion' is commonly used as a postural verb in periphrastic constructions; cf. also Germanic (Ebert, this volume a)). Combinations with other verbs (ums-j suuna 'sits reading', unt-J kevtna 'lies sleeping') are extremely rare. There is not a single occurrence in the questionnaire or in the text examples of the various sources. Both -Jana and -j jovna can be found in generic or habitual contexts (Biasing 1984: 19). The imperfective participle in -a {-ha after vowels) followed by a form of bää- in non-present contexts, was used in three sentences of the PROGQ:23 nam gerlt-ä 'the sun is shining' (PROGQ: 36), ködl-ä bilä 'was still working' (PROGQ: 3), and Ы xotan ke-hä bääxv 'I will still be cooking' (PROGQ: 83). Two of the questionnaire sentences contained the adverb "still", which signals that the situation has been going on for some time. This is confirmed by other examples in the literature; cf. ter oda cign surhul' sur-a 'er studiert noch' (Benzing 1985: 128), xuur saak kevtän or-a 'rain is/ keeps falling' (Todaeva 1968: 46). The use of the ö-form resembles the 'continuative' in Lezgian (Haspelmath 1993: 145). We tentatively adopt the term here, although an explicit notion of continuativity is not always implied; PROGQ: 83 is not equivalent to '... I will go on cooking'. The Continuative is compatible with focalizing contexts such as PROGQ: 3 and 83, but also with habituality. We therefore do not consider it to be a Foc-PROG. The fourth form to be considered, -ad bää-, specifies actionality. We have little explicit information on the actionality of Kalmyk verbs, but if the Kalmyk-Russian dictionary gives a Perfective-Imperfective pair for a Kalmyk lexeme, we can be quite sure that we are dealing with a two-phase verb;25 e.g., suux 'sit; sit down', ääx 'be afraid; become afraid', satx. 'burn, sparkle; start to burn'. A form like satad bääspecifies the non-transformative meaning component 'burn, be sparkling'. But -ad bää- serves not only to build new lexemes, as Biasing claims. It can indicate various types of durative situations, like continuativity, iterativity, gradually,

The progressive in Europe

545

Table 4. Forms used in the PROGQquestionnaire. The numbers on the left refer to the questionnaire sentences, as specified in the Appendix. Karachay

Kalmyk -JaPROG

-a-ad bää- -a tur-na 1PFV CONT DUR PROG

a) focalized 1 + 6-17 + 44,45,47 + 33-35 + 37-38 + 46 +MOT 31 +MOT b) habitual 2 + 4 c) duration /(iter.) 3,83 36 18 62 d) temporal limits 48 49 50 60 70b + ej"inclusive reading 81-82 f) imperative, negation 73 76 +* 77 g) Stative verbs 39 + 40 + 42 h) postural verbs 29,59 + 58 + 28 + ') telic verbs 54 + 55 +

+ + + + + + (+)

-ib-a-ib turIPFV DUR/RES RES

(+) + + + (+) + + + +

+ +

+

V just now V just now V just now insult, admit, dream boil, rot shop reach top/climb

+MOT

+ work/cook still shine write dissertation bees humming ...

+ +

+sit +MOT

+ + + + + + +

+

+

play for 2 h. talk during (and go on) talk during (and stop) earth turn while . . . play while ...

+ +

j -

+MOT

bake/work all x

+ -n bää

be working! not work not rain

+?

+ (+)

+

+ + + + + +

(+) (+) (+)

know like now be kind now

+ + + +MOT +MOT +MOT

+ +

hang/stand stand (tempor.) sit (tempor.) improve grad. cover grad. die

546

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Kalmyk -ja-

PROG

Karachay -na -a-ad bää- -a turIPFV CONT DUR PROG

j) motion verbs 21 + 22 +MOT 66-67 + k) phasal1 verbs 23/25 + 24/26 +

+MOT

-a-ib tur-ibIPFV DUR/RES RES +MOT +MOT

+

go out just now fly to go tomorrow begin finish

* ködl-j uga bää-J ° öngr-n gi-jä-lä (die-CONV SAY-SIM-ANT) 'was about to die' () forms judged less good by the informant

or simply extended duration. All postural verbs are possible in this form, which often corresponds to the postural verb construction in Germanic languages (d). It is compatible with the Perfective (e) and even with the Progressive (f): (33)

Kalmyk a. b. с d.

e. f.

naad-ad bää-nä 'continues playing' (Todaeva 1976: 154) end-tendänxälä-Mdjov-na 'is looking here and there' (Biasing 1984: 25) ükrmüd rjarrjhd ids-äd jov-la 'the cows were grazing in the meadows' (PROGQ: 62) ter dissertatsan bic-äd suu-па. 'she is [sits] writing her dissertation' (PROGQ: 18) casn narna kücnd xääl-ärf bää-v 'the snow melted slowly in the force of the sun' (Biasing 1984: 25) Ocrig Kitd or-ady'ov-jana 'Ocr was travelling in China ...' (Biasing 1984: 27)

Unlike the Germanic postural verb forms, the Durative was not used in focalized contexts (cf. Table 3). This and the combinability with Progressive show, that it is an actional periphrasis and does not qualify as a Dur-PROG.

5.2.

Karachai

Our material for Karachai is very limited and the interpretation of the data m u s therefore remain rather tentative. The great advantage of the Karachai questionnaire

The progressive in Europe

547

is that the informant gives all possible translations - sometimes up to five different forms - and often comments on their degree of acceptability. The forms to be considered here are: Imperfective Progressive Resultative I Resultative II Durative

-a-(—v-) -a-(~-y-) tur-fbar-ib-ibtur-ib tur-/bar-...

(simultaneous converb) (simultaneous converb + 'be' /'go') (anterior converb) (anterior converb + ' b e ' ) (anterior converb + 'be'/'go' ...)

In the typical progressive contexts (e.g. PROGQ: 6-20), the informant used the PROG marked by -a tur-, usually besides an Imperfective; e.g., with - agentive and nonagentive verbs: suu k'ayna-y tur-a-di (boil-SIM be-IPFV-3:SG) 'the water is boiling' (PROGQ: 37), - motion verbs: cig-a tura-dï / bara-dï 'is just going out' (PROGQ: 21), - phasal verbs: êt-ib bosa-y tura-dï 'is finishing teaching' (PROGQ: 26), - temporary states: bildir-e tura-sïz 'you (PL) are being nice' (PROGQ: 42). It was not used with postural verbs denoting a temporary state (PROGQ: 28, 58), and not with true states (bil-e-di 'knows'). In contrast to Kalmyk -ad bää-, the parallel Karachai form with the anterior converb in -ib was used almost as often as PROG in the questionnaire. This form is sometimes claimed to have resultative, perfect or progressive/actual present function (Nedjalkov & Nedjalkov 1987: 114).26 In most occurrences of the PROGQ, the form in -ib tur- is comparable to Kalmyk -ad bää-, i.e., it serves as an actional specifier. Often it is listed together with the simple form as a lexical unit in the dictionary; e.g., Russ. sidet' 'sit (ipfv)': oltur-urg'a, oltur-ub tur-urg'a; Russ. spat' 'sleep (ipfv)': dzuk'la-rg'a, dzukl'a-b tur-urg'a; Russ. deriat' 'hold (ipfv)': tut-arg'a, tut-ub tur-urg'a. (34)

Karachai (PROGQ: 3) Dzon kel-gen-in-de, Anna alk'i'n dzukla-fr John come-PF-POSS-LOC Anna still sleep-ANT.CONV tur-a-di / dzukl-a-d'i be-IPFV-3SG / sleep-IPFV-3SG 'When John came, Anna was still sleeping.'

Postural verbs in actual present contexts were translated with three different forms: w)

Karachai oltur-a-dï / oltur-ub tur-a-di' / oltur-ub-d'i 'is sitting' (PROGQ: 28) siiel-e-di / siiel-ib tur-a-di / siiel-ib-di 'is standing' (PROGQ: 59) IPFV IPFV/RESII RES I

548

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

The Imperfective is formed from the stative meaning component of the initio-transformative verbs oltur- 'sit down, sit', siiel- 'stand up, stand', or from the composite verbs olturub tur-, süelib tur-. The Resultative is formed from the transformative meaning component. The second form is structurally ambiguous, although the denotative content is identical: it is either a RES II27 of the verbs oltur-, siiel-, or an IPFV of the verbs olturub tur-, süelib tur-. There is no progressive and no perfect meaning involved in this ambiguity. The Durative is used with all types of verbs except momentaneous ones, and exhibits the same range of meanings as the Kalmyk Durative. With accomplishment verbs the informant prefers the auxiliary bar- 'go, move', probably to avoid the ambiguity which -ib tur- can create with these verbs.

(36) a.

b.

Karachai (PROGQ: 57)28 k'art öl-йЬ bar-ъ. èdi. old_man die-ANT.CONV go-SIM.CONV was 'The old man was dying.' (PROGQ: 55) ... k'ar uak'-uak' dzer-ni dzab-ïb snow slowly-slowly ground-ACC cover-ANT.CONV Ьаг-г. édi. go-SIM.CONV was '... snow was gradually covering the ground.'

The Durative is not excluded in habitual contexts; e.g., ol kitab ok'u-y-di (IPFV)/ ok'ub turadï (DUR) 'she reads [every Saturday]' (PROGQ: 2), and it can be combined with a PROG.

(37)

Karachai (PROGQ: 27) 01 xapar-nï ayt-i'fc bar-a tur-a-d'i. s/he story-ACC tell-ANT.CONV go-SIM.CONV be-IPFV-3SG 'She is continuing telling a story.'

In the PFQ the -ib tur- form was used only with telic predicates. Still, not all (nondurative) -ib tur- forms can be interpreted as resultatives. Nedjalkov & Nedjalkov (1987) found that the test adverb alk'ïn 'still' can not be inserted before -ib tur- forms from active transitive verbs,29 i.e., there seem to be no possessive resultatives. (38) has perfect meaning only:

The progressive in Europe (38)

549

Karachai (Nedjalkov & Nedjalkov 1987: 116) Ata esik (*alk"in) ac-ïb tur-a-di. father door still close-CONV be-IPFV-3:SG 'Father has closed the door.'

alk'ïn was also rejected with some intransitive verbs, e.g., ol (*alk"in ) ketib turadï 'he has come'. The resultative -ib tur- form thus goes the common way towards a perfect meaning (the perfect in -gan develops towards a preterite). Which verbs allow a resultative interpretation is not predictable. Nedjalkov & Nedjalkov list 15 verbs (out of 400 they tested) for which they claim actual present meaning with -ib tura-, among them the postural verbs, but also 'sew', 'plough', 'graze (trans.)', 'fish'. We do not see any reason to interprète the -ib tur- form in ol kiyim tigib turadï 'she sews / is sewing dresses' different from the durative forms.

6. Conclusions Of the four forms used in the Kalmyk PROGQ, two (V + -Jana and V + -}bää-/jov-) can be classified as Foc-PROG, although they show some signs of defocalization. Karachai has only one Foc-PROG form {-a tur-). The Kalmyk imperfective participle form in -a is a rare subtype of the imperfective dimension which expresses that a situation holds at a point of reference and for some time before. It could probably be located somewhere between Foc-PROG and Dur-PROG. In both languages the form made up of the anterior converb + postural verb marks various types of durative actionality. The Karachai form in -ib tur- (and no other auxiliary) is also a resultative, often with perfect meaning. There is no evidence in our data that the durative forms can have progressive function. Unlike the Germanic postural verb constructions, which were classified as instances of Dur-PROG, the Kalmyk and Karachai duratives can be combined with a progressive and are not suitable to mark focalization. The situation is similar in other Turkic languages (Schöning 1984: 324, Johanson 1971, 1995). In Tatar the combination of the converb in -p + postural verb is an actional periphrasis specifying the phasal meaning of fini-transformative and initiotransformative verbs, while other auxiliaries specify the transformational meaning component, e.g., (39)

Tatar iil-ep kit-te die-CONV go_away-PT:3 'he died' vs.

550

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot iil-ep yat-tï die-CONV lie-PT:3 'he was dying' kiir-ep al-di see-CONV take-PT:3 'he saw (discovered)' vs. kiir-ep tor-dï see-CONV be/stand-PT:3 'he saw'

;

u. ,;,. ,..„:M ;( f ,

The simultaneous converb in -a together with tor- 'be, stand' has been grammaticalized as a progressive, ruling out combinations of -a with other postural verbs (Schöning 1984: 277). Several authors mention that constructions with -p tor- and -p uti'r('sit') can also mark an actual present and thus come close in meaning to the progressive. The form also expresses perfect meaning: uqïp tordï 'has read'. Kumyk has actional periphrases with actionally ambiguous, but also with activity verbs: yuxlap tur- 'sleep', asap tur- 'eat', oxup- tur- 'read'. The Imperfective from these complex lexemes is formally identical to the Resultative: e.g., oxup tura 'reads', gelip tura 'is/has come' (cf. Johanson 1995 and to appear). The Altaic forms are not unlike those in Western European languages. Simultaneous converbs or participles + auxiliary (< postural verb) yield progressives, anterior converbs or participles + auxiliaries yield resultatives and perfects. Postural verbs are used in durative periphrases. There are, however, some important differences: combinations of converb + auxiliary have been lexicalized to a considerable degree in Altaic languages. In Western European languages such lexicalizations are rare and mainly restricted to non-agentive verbs (e.g., Engl, be tired, Germ, umgeben sein von). The durative periphrases (at least in Kalmyk and Karachai) are not used in focalization and are therefore no instances of Dur-PROG. A further difference results from the weak notion of anteriority associated with the converbs -ad, -ib, which allows them to be used much like simultaneous forms. This leads to a further characteristic of some Turkic languages, namely that one form can have resultative/perfect and durative function.

Notes 1. This work was jointly developed by the three authors. However, CdG bears the main responsibility for Sections 1 and 2, PMB for Sections 3 and 4, and KE for Section 5. 2. An apparent counterexample could be the Albanian construction with the particle po (cf• fn. 3). However, the meaning of this particle is not strictly temporal; thus, it represents a different type.

The progressive in Europe

551

3. For our purposes, the classification proposed by Blansitt turned out to be more insightful than, for instance, the classification by Bybee et al. (1994: 128-129), which on the one hand refers to morphological devices not used by European languages (such as "reduplication"), and on the other hand is mostly oriented towards identifying the possible sources of the progressive (a preoccupation which does not concern us here). 4. The particle po is considered to be an intensifier by Duchet (1995). Indeed, besides forming PROG in combination with the Present or the Imperfect, it may have a purely intensifying function when used with other tenses or other grammatical forms (like pronouns). However, its service as a PROG marker is well established. Albanian has another way of expressing progressivity, namely the construction 'be' + Gerund (Camaj 1984): (i)

Albanian Zogitë janë duke sjellë kandrra ne çerdhe. birds are GER bring insects to nest 'The birds are bringing insects to their nest.'

In Geg, the Gerund-building form is tue rather than duke. Note however that in Northern Geg PROG is preferably expressed by the construction 'be' + kah + inflected verb (where kah is a preposition meaning 'towards'). Apparently, the latter construction is both a PROG marker and a marker of imminentiality (Alexander Murzaku, p.c.). In Arbèresh (a Tosk variety spoken by the Albanian minority in Italy), PROG takes the form of 'be' + e + inflected verb (where e is the conjunction) or 'be' + që + inflected verb (where që is a complementizer analogous to English 'that', its actual realization varying from dialect to dialect) (Francesco Altimari, p.c.). In some Arbèresh varieties one finds also the periphrasis vete ('go') + Gerund (Breu 1982), which apparently has the same functions as the analogous constructions to be found in several Romance languages (cf. the "Motion auxiliary type" in Table 1). 5. It is also possible to use other verbs as auxiliaries, such as egon 'stay', ibili 'walk, function/work', ihardun 'keep doing something', again combined with the Imperfective Participle. However, the use of these constructions is dialectally bound. The inessive case of nominal forms substitutes for the Imperfective Participle in certain cases (PROGQ:l, 35: ametsetan 'dreaming', lit. 'in dreams'). Sometimes, both possibilities (verbal and nominal) coexist: e.g., hitzegiten ari da or hizketan ari da 's/he is talking'; hitzegiten 'talk' (lit. do talk (Participle)), hizketan 'in the talk' (cf. PROGQ:44, 53). Still another possibility (cf. PROGQ:51) is offered by nominal forms with the instrumental suffix combined with ibili 'walk' or joan 'go' as conjugated forms (note that no sense of motion is implied in these cases). A few verbs do not have the analytic construction in ari, but present synthetic forms that are used with the Present and the Preterite: izan 'be', egon 'stay', joan 'go', etorri come', ibili 'walk', eduki 'have', ekarri 'bring', eraman 'take (somewhere)', erabili use', jakin 'know', ihardun 'keep doing something', and some other relatively unproductive ones. These synthetic forms can take the progressive meaning in the appropriate contexts. However, with some of these verbs the analytic construction is gaining ground, as in jakiten ari naiz (lit.: I am knowing) vs. dakit 'I know'. (Data from Miren Lourdes Onederra, p.c.)

552

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

6. Since this is a potential source of misunderstanding, we would like to clarify the conventions that we are going to adopt in this chapter. When referring to the distinction Perfective/Imperfective, as used in the grammatical descriptions of Slavic languages, we shall use capital initials. When this convention is not used, these terms should be intended as they usually are in the grammatical descriptions of Germanic, Romance etc. We refer the reader to Bertinetto & Delfitto (this volume), for a detailed discussion of this topic, and a possible terminological distinction, whereby the terms 'bounded/unbounded' are used in the former sense, and 'terminative/nonterminative' in the latter. However, as the authors make clear, the need for this disambiguation arises only in particular situations. When not strictly necessary (i.e., when there is no real danger of misunderstanding), the traditional terminology can be retained. Cf. also Johanson (this volume), for a thorough clarification of the aspectual matter. Still with respect to terminology, the reader should be aware that we use the word 'actional' as the adjective corresponding to 'actionality' (i.e. Aktionsart). 7. As observed in Bertinetto, this volume, the Italian PROG device based on motion verbs (cf. Table 1) is almost uniquely used in formal styles. 8. Blansitt (1975:3) refers to Chafe (1970) for the notion of "generic progressive" (exemplified by John is playing golf once a week), but also quotes Bee (1973) for the distinction between the two meanings of He is eating, which may indicate either a currently occurring event, or the fact that someone is again able to eat after a severe illness. In addition, Blansitt (p. 4) notes that there is an "intensive durative" construction, somehow related to the progressive, as in He is reading away, which may correspond to Spanish Esta lee y lee. This type resembles the hyperbolic uses of PROG, exemplified by Eng. He is always reading or Spanish Esta siempre leyendo (cf. examples (11) in Bertinetto, this volume). 9. A situation where an event overlaps with a simultaneously ongoing process was called "Inzidenzschema" by Pollak (1960: 129); ci. "a relation between a dynamic situation and a point in time" in Dahl (1985: 91), and Comrie (1976: 3), where this kind of situation is used to illustrate imperfective (progressive) as opposed to perfective aspect in certain languages. 10. In Hungarian, Perfective verbs cannot be used with the ABS V (de Groot 1995), although, as shown in (21b), they appear in progressive contexts (cf. Section 2.6 foi our remarks). But the contradiction is only apparent; one should not confuse Perfective with perfective (cf. again fn. 6 for our notational conventions). The incompatibility of Perfective verbs with the ABSV is an obvious consequence of the fact that this construction depicts a situation as occurring at a particular reference interval, and thus concerns an event that cannot be viewed as completed. 11. As to the contrast between Sp. estuvo leyendo/ha estado leyendo todo el dia on the one side, and Eng. he has been studying the whole day, cf. Bertinetto & Delfitto (this volume, Section 3.2). 12. As shown by Ebert, this volume a, in some Germanic languages this constraint takes the form of an "intentionality" requirement, or in some cases of a "dynamicity" requirement. In particular, prepositional periphrases (cf. Table 1) are preferred with higher dyrtamicity (cf. table 3 in Ebert, this volume a), and are in general more readily available with a focalized interpretation. Among the Germanic languages considered, only Icelandic seems

The progressive in Europe

13.

14.

15.

16.

553

immune from this component; but significantly, the PROG construction of Icelandic is, together with that of English, the most grammaticalized one within the Germanic group. As to the variable strength of the agentive/intentional requirement with the ABSV, cf. De Groot (this volume). An apparent exception is represented by the Swedish ABSV construction, which may be based on the inherently passive auxiliary blir (De Groot, this volume). But one may contend that in this construction the original passive meaning of the auxiliary is bleached. Needless to say, our observations do not refer to those PROG or ABSV periphrases that are normally constructed with an Infinitive governed by the copula with the possible help of a preposition (for the relevant morphological data, see Table 1 here and De Groot, this volume). Whether or not a PROG construction based on the Infinitive in the main verb may also admit the Infinitive in the auxiliary verb depends on the given language. For instance, when PROG is governed by a modal verb, Standard Portuguese presents the Infinitive in both the copula and the main verb. Interestingly, Germ, gerade is also indicated by some scholars as a PROG device. But see the contrary opinion expressed by Ebert, this volume a. In fact, one should note that gerade can be employed in Stative, i.e. clearly non-progressive, contexts (such as: Er ist gerade da), whereas Alb. po is rejected in the same contexts. As to the expression of PROG in Albanian, cf. fn. 4. Note, however, that the permanent stative (or individual-level) interpretation of example (25b) presupposes an equative, rather than a locative meaning of the copula. Thus, there are grounds to believe that there are in fact two possible sources for progressive periphrases, incorporating an existential-locative meaning or an existential-equative meaning respectively, with the latter converging with the former at some later stage. Another hypothesis that one could plausibly put forth is that stage (i) constitutes an entirely independent evolutionary path. Accordingly, one could suppose that the periphrases of stage (ii) exploited an already existing morphosyntactic structure, which had become available for a different usage. However, this hypothesis is weakened by the fact that the same development has made its appearance in two fairly heterogeneous languages, Latin and Old English.

Note finally that PROG constructions based on motion verbs presumably enter the evolutionary path directly at stage (ii), for the purely stative meaning typical of stage (i) is alien to the inherent semantic import of these verbs. 17. Examples of PROG devices implementing the initial phase of stage (ii) may still be found in Modern Romance languages. This is to be observed for instance in: (i)

Italian

(ii)

Che ci stai a fare? What there stand:2SG:PRS at do 'What are you doing here/there?' Spanish Todos los dîas estas una hora escribiendo. all the days be:PRS:2SG one hour writing 'You write one hour every day.'

554

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot The periphrasis in (i) is similar to the Italian infinitival PROG (cf. Table 1 and Bertinetto, this volume), but differs from it in that it includes an explicitly locative morpheme. As to (ii), it may easily be confused with Spanish copular PROG, but its syntactic behaviour is different, as shown by negation (Luis Garcia Fernandez, p.c.): (iii)

Spanish a.

b.

с

d.

i Esta escribiendo. be:3SG:PRS writing 'S/he is writing.' No esta escribiendo. / *Esta sin escribir. ' not be:3SG:PRS writing be:3SG:PRS without write 'S/he is not writing.' Esta una hora escribiendo. be:3SG:PRS one hour writing 'S/he spends one hour writing.' No esta una hora escribiendo. / Esta una hora not be:3SG:PRS one hour writing be:3SG:PRS one hour sin escribir. without write 'S/he does not spend one hour writing / S/he remains for one hour without writing.'

In (d), the presence of the temporal adverbial allows for the negation with sin, which is not admitted in (b) with progressive meaning. What is particularly remarkable is that the auxiliary in (c-d) preserves by and large its original, locative meaning (thus, it is not a true auxiliary). The same applies to the Italian construction "starsene a + Infinitive", which exhibits a verb with an explicit locative meaning preceding the prepositional In­ finitive. In both cases, there is the implication that the event takes place in a specific place. 18. Even in Catalan there is a tendency to avoid PROG in a context like this. But the usage of PROG is in general slightly more restricted in Catalan than in the other Ibero-Romance languages (Bertinetto, this volume). 19. The same remark is put forth for Breton PROG by Hewitt (1985/86). 20. Note that this is not always the case. In Dutch, for instance, when the adverb keen is present, the Participle may not be added: (i)

Dutch Ze zijn heen aardappels rooien (*gegaan). they are away potatoes dig:INF gone 'They are off digging up potatoes.'

21. We thank Igor and Vladimir Nedjalkov, who translated the questionnaire into Russian, as well as B. A. Biceev and A. I. Xasanov, both trained linguists and native speakers of Kalmyk and Karachay respectively, for their collaboration.

The progressive in Europe

555

22. The term "Präsens" applies to forms with the suffix -na: -Jana Präsens durativum I, -ad bäänä Präsens durativum II. The combination of -Ja with the past/anterior marker -la is called Imperfectum III (p. 39), with FUT -ax Präsens II (-/ax does not occur in my data); -ad bää- + PFV -v is called Imperfectum I. 23. This form is rare also in texts; cf. Todaeva (1976: 159), Biasing (1984: 75). 24. Cf. Badm traktorist ködl-ä 'Badm works as a tractor driver' (Benzing 1985: 128). 25. This does not hold vice-versa. For fini-transformatives the dictionary is very careful to give only imperfective translations. Apparently Russian linguists expect telic/perfective verbs to be marked morphologically, and consequently they describe only verbs marked with telecizers like -ck as "perfective". 26. A dual progressive-perfect function has been claimed also for Khalkha Mongolian -aad bai-, but the only example found in articles and texts with a perfect meaning is ir-eed baina 'has come'. Otherwise -aad bai- is a durative periphrasis, like Kalmyk -ad bää(cf. Ebert 1995: 196). 27. There seems to be no big semantic difference between RES I and RES II; whenever the informant used RES I, he also gives a RES II form as an alternative, though not vice versa. RES I seems to be rare. In the PFQ it was given only as one of several possibilities in: ol uyan-ïb-dï (RES I) / uyan-'ib turadi (RES II) 'he woke up already' (PFQ: 30-31). 28. The progressives öle tura èdi and dzaba turn èdi were considered "worse" in those examples. 29. The authors claim that resultatives can be formed from a few transitive active verbs, but the only example given, ol kitab-nï alk'ïn al-ïb turadi 's/he is still holding the book' is most probably an actional periphrasis of the initio-transformative verb al- 'take, hold'.

References

Bee, Darlene L. 1973 Neo-tagmemics: An integrated approach to linguistic analysis and description. Uka rumpa: Summer Institue of Linguistics. Benzing, Johannes 1985 Kalmückische Grammatik zum Nachschlagen. Wiesbaden: Harassowitz. Bertinetto, Pier Marco 1994 "Statives, progressives and habituais: Analogies and divergences", Linguistics 32: 391423. this volume "The progressive in Romance, as compared with English". Bertinetto, Pier Marco, Valentina Bianchi, Osten Dahl & Mario Squartini (eds.) '"95 Temporal reference, aspect and actionality. Typological perspectives. Torino: Rosen berg & Sellier. Biasing, Uwe °"* Diefinitindikativischen Verbalformen im Kalmückischen. Wiesbaden: Steiner. Blansitt, Edward L. Jr. 1975 "Progressive aspect", Working Papers on Language Universals 18: 1-34. Breu, Walter 1982 "Forme verbali perifrastiche arbérisht", in: Guzzetta, Antonino (ed.), Etnia albanese e minoranze linguistiche in Italia. Palermo: Istituto di Lingua e Letteratura Albanese, Université di Palermo, 313-333.

556

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Bybee, Joan L., Revere Perkins & William Pagliuca 1994 The evolution of grammar. Tense, aspect, and modality in the languages of the world. Chicago/London: The University of Chicago Press. Bybee, Joan & Osten Dahl 1989 "The creation of tense and aspect systems in the languages of the world", Studies in Language 13: 51-103. Camaj, Martin 1984 Albanian Grammar with exercises, chrestomathy and glossaries. Wiesbaden: Harrassowitz. Carlson, Gregory 1978 References to kinds in English. Bloomington: Indiana University Linguistics Club. Chafe, William L. 1970 Meaning and the structure of language. Chicago: The University of Chicago Press. Comrie, Bernard 1976 Aspect. An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge University Press. Dahl, Osten 1985 Tense and aspect systems. Oxford: Blackwell. De Groot, Casper 1995 "The absentive in Hungarian", in: I. Kenesei (ed.), Levels and structures (Approaches to Hungarian 5). Szeged: JATE, 45-61. De Groot, Casper this volume "The absentive". Delfitto, Denis & Pier Marco Bertinetto 1995 "A case study in the interaction of aspect and actionality: The Imperfect in Italian", in: Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, Mario Squartini (eds.), Temporal reference, aspect and actionality: Semantic and syntactic perspectives. Torino: Rosenberg & Sellier: 125-142. Dietrich, Wolf 1973 Das periphrastische Verbalaspekt in den romanischen Sprachen. Tübingen: Narr. Duchet, Jean-Louis 1995 "The Albanian tense system", in: Rolf Thieroff (ed.), 253-275. Ebert, Karen H. 1995 "Ambiguous perfect-progressive forms across languages", in: Bertinetto, Pier Marco, Valentina Bianchi, Osten Dahl & Mario Squartini (eds.), 185-203. 1996 Kodqva (Kodagu). Lincom Europa: Languages of the World Materials No. 104. to appear "Focality degrees in Kalmyk imperfectives", in: Leonid Kulikov & Heinz Vater (eds.), Typology of verbal categories. Papers presented to Vladimir Nedjalkov on the occassion of his 70th birthday. (Vol. 2). Tübingen: Niemeyer, this volume a "Progressive markers in Germanic languages", this volume b "Aspect in Maltese". Gougenheim, Georges 1929 Étude sur les périphrases verbales de la langue française. Paris: Nizet. Haase, Martin 1994 'Tense and aspect in Basque", in: Rolf Thieroff & Joachim Ballweg (eds.), 279-292. Haspelmath, Martin 1993 A Grammar of Lezgian. Berlin: Mouton de Gruyter. Heinämäki, Orvokki 1995 "The progressive in Finnish: Pragmatic constraints", in: Pier Marco Bertinetto, Valentina Bianchi, Osten Dahl, Mario Squartini (eds.), Temporal reference, aspect, and actionality. Typological perspectives. Torino: Rosenberg & Sellier: 143-153.

чхщ}

ы'Л »M** '

The progressive in Europe

557

Hetzron, Robert 1982 "Non-applicability as a test for category definitions", in: F. Kiefer (ed.), Hungarian Linguistics. Amsterdam: Benjamins, 131-183. Hewitt, Steve 1985/86 "Le progressif en breton à la lumière du progressif anglais", La Bretagne Linguistique 2. Johanson, Lars 1971 Aspekt im Türkischen. Vorstudien zu einer Beschreibung des türkeitürkischen Aspektsystems. Uppsala: Acta Universitas Upsaliensis. 1995 Mehrdeutigkeit der türkischen Verbalkomposition. In: M. Erdahl & S. Tezcan (Hrsg.), Beläk Bitig. Sprachstudien für Gerhard Doerfer zum 75. Geburtstag (Turcologica 23). Wiesbaden, 81-101. to appear "Grenzbezogenheit in Aspekt und Lexik." Paper read at the workshop "Interaktion zwischen Lexik und Aspekt". May 1995, Universität Konstanz, this volume "Viewpoint operators in European languages". Kiefer, Ferenc 1994 "Aspect and syntactic structure", in: Ferenc Kiefer & Katalin E. Kiss (eds.), The syntactic structure of Hungarian. Syntax and Semantics vol. 27. San Diego: Academic Press, 415^64. King, Alan R. 1994 The Basque Language. A practical introduction. Reno: University of Nevada Press. König, Ekkehard 1995 "He is being obscure: Non-verbal predication and the progressive", in: Pier Marco Bertinetto, Valentina Bianchi, Osten Dahl, Mario Squartini (eds.), Temporal reference, aspect, and actionality. Typological perspectives. Torino: Rosenberg & Sellier, 155— 167. König, Ekkehard & P. Lutzeier 1973 "Bedeutung und Verwendung der Progressivform im heutigen Englisch", Lingua 32: 277-308. Kozintseva, Natalia 1995 "The tense system of Modern Eastern Armenian", in: R. Thieroff (ed.), 277-297. Metslang, Helle 1995 "The progressive in Estonian", in: Pier Marco Bertinetto, Valentina Bianchi, Osten Dahl, Mario Squartini (eds.). Temporal reference, aspect, and actionality. Typological perspectives. Torino: Rosenberg & Sellier, 169-183. Munieva, B.D. 1977 Kalmycko-Russkij slovar /Xal'mg-Ors Toi'. Moskva: Izdatelstvo "Russkij Jazyk". Nedjalkov, Igor V. & Vladimir P. Nedjalkov 1987 "Karacaevo-balkarskaja glagol'naja forma na -b/-p tur-a- со znacenijami nastojaäcego i prosedsego vremeni (v sravnenii s formami na -b tur-a/tur-ib- v uzbekskom jazyke", Funkcional'no-semanticeskie aspekty grammatiki. Moskva: Nauka. 13-21. Poppe, Nicholas 1968 Tatar manual. Descriptive grammar and texts with a Tatar-English glossary. [2d rev. ed.]. (Indiana University Publications. Uralic and Altaic series, v. 25). Bloomington: Indiana University. Olbertz, Hella 1998 Verbal periphrases in a functional grammar of Spanish. Berlin: Mouton de Gruyter. Schönig, Claus 1984 Hilfsverben im Tatarischen. Wiesbaden: Steiner. Tnieroff, Rolf (ed.) ' Tense systems in European languages II. Tübingen: Niemeyer.

558

Pier Marco Bertinetto, Karen H. Ebert, Casper de Groot

Thieroff, Rolf & Joachim Ballweg (eds.) 1994 Tense systems in European languages. Tübingen: Niemeyer. Thieroff, Rolf this volume "On the areal distribution of tense-aspect categories in Europe". Todaeva, Buljaä Chojcievna 1968 "Kalmyckij jazyk", in: Skorik, P. Ja. (ed.), Jazyki Narodov SSSR. Leningrad: Nauka, 35-52. 1976 Opyt lingvisticeskogo issledovanija èposa "Dzangar". Elista: Kalmyckoe kniznoe izdatel'stvo. Tommola, Hannu .•••••. this volume "The progressive in Baltic Finnic".

••mi

pier Marco Bertinetto

The progressive in Romance, as compared with English 1. Prolegomena1 This chapter collapses two apparently disparate entities: the Romance languages and a single language belonging to the Germanic group. This might be considered inappropriate from a strictly genetic point of view. But the typological perspective that we are assuming in this series of contributions justifies such a move. In fact, this solution should not even appear particularly surprising, given the fact that English is, among the Germanic languages, the one that has most dramatically departed from the other languages of the group as a result of the protracted contact with French in a crucial phase of its history. But what matters most is that, regarding the particular phenomenon we are going to discuss here (i.e. the progressive), the languages under consideration present deep affinities. It can easily be shown that Germanic languages other than English exhibit quite different features, as illustrated in the companion chapter by Karen Ebert (this volume). It has even been claimed that the English progressive was shaped by the Romance model. However, the alternative view, according to which this construction represents an autonomous development, is equally défendable (Scheffer 1975). This issue of course needs to be addressed in proper terms, namely through the comparative investigation of ancient texts (to the extent that they provide evidence for the crucial period). Here I shall disregard it, and merely concentrate on the observable synchronic situation, characterized by strong similarities between Romance and English, be it a matter of common origin or of mere convergence. A clarification is in order at the outset. We have to distinguish between "progressive" as a semantic notion and as a formal manifestation (i.e. a morphosyntactic device). To convey the latter sense, I shall use the abbreviation PROG (except when 1 cite the traditional grammatical denomination of a given tense, such as, e.g., the English "Present Pro gressive"). Although in many cases there is a reciprocal implication, this is not always so. Consider Romance languages, where the progressive as Pect is not necessarily conveyed by specialized morphosyntactic devices (namely Periphrases), for the mere usage of imperfective tenses is in most cases perfectly adequate. And this is not the only complication. In fact, on the one hand we find Prototypical contexts where the notion "progressive" is necessarily present indepen-

560

Pier Marco Bertinetto

dent of the particular device employed, be it a general purpose imperfective tense (of the sort we can find in Romance) or a specific periphrasis (like in English, as well as in Romance). On the other hand, specialized PROG devices may also appear in contexts which have little to do with the aspectual notion "progressive". To quote an obvious example, consider English PROG with future-time reference, as in: / am leaving tomorrow. Although the development of this meaning must originally have been licensed by some specific property possessed by what we might call the "prototypical" progressive aspect (as is shown by the fact that English is not the only language showing this particular development; cf. Section 6.3.3), it is clear that this usage of PROG does not convey any progressive meaning, in the proper sense of this term. Thus, the correspondence of form and meaning is not always perfect. Nevertheless, it is fair to say that whenever a specialized device exists, it is quite likely that we find at least some contexts where this is considered by the speakers as a quite natural choice (or maybe as the only option available). In this chapter I shall mainly be concerned with the morphosyntactic device PROG, rather than with the progressive aspect in the strict sense, and its possible manifestations. The data I shall discuss here are drawn in part from the questionnaire that was prepared and distributed by EUROTYP Group 6 (henceforth PROGQ; cf. Appendix 3), but it will also be complemented by a survey of the available literature. The languages for which we collected data, among those relevant here, are the following:2 Catalan: English: French: Italian: Portuguese: Romanian: Spanish:

1 subject (Standard Central Catalan) 1 British subject (no declared dialectal background) 2 subjects (no declared dialectal background) 4 subjects (1 from the North, 1 from the Centre, 2 from Sardinia; all speaking varieties of Standard Italian) 3 European subjects (no declared dialectal background) 3 subjects (no declared dialectal background). 2 European subjects (Standard Spanish).

As to the morphosyntactic devices employed, Table 1 lists the most relevant ones. The label 'St-PROG' indicates periphrases based on auxiliary verbs approximately meaning "be, stand". 'Mot-PROG' stands for periphrases based on auxiliary verbs meaning "go, come". The third type is a miscellaneous category comprising the residual devices. As can be seen, in French and Romanian the third type is virtually the only one existing (considering that "aller + GER" in Modern French is extremely rare). However, even with these languages it is correct to call this type "Marginal because of its relatively infrequent usage, which is quite remarkable in Romanian. This appears also in PROGQ, where PROG is very seldom employed by the Romanian informants. Consequently, in what follows I shall have very little to say about this language, except for some narrowly targeted observations (cf. Section 6.1).

The progressive in Romance, as compared with English

561

Table 1. Survey of the morphological manifestations of PROG devices in Romance and English.

English Catalan French Italian Portuguese Romanian Spanish

State-PROG

Motion-PROG

Marginal types

be + V-ing estar + GER stare + GER stare a + INF estar a + INF estar + GER estar + GER

anar + GER {aller + GER) andarelvenire + GER

? ? être en train de + INF (essere dietro a + INF)

ir/vir + GER

?

irlandarlvenir + GER

afl m curs de a + INF ?

St-PROG divides further into a gerundial and an infinitival type, as shown in Table 1. For clarity, I shall designate them St-PROG-GER and St-PROG-INF respectively. However, I shall avoid providing these additional qualifications when I refer to the general type, or whenever it is sufficiently clear which is the subtype I am referring to. St-PROG-INF is the standard device in European Portuguese, although the gerundial type is equally present, especially in the written language. Curiously, in Brazilian Portuguese the situation is reversed, with St-PROG-GER commonly used, at the expense of its competitor. As shown in Table 1, St-PROG-INF also exists in Italian. This device is now confined to the colloquial usage of the speakers of Central Italy (most typically in Rome, where the rival construction is much less frequent; cf. D'Achille and Giovanardi 1998), but it is also occasionally to be found in literary texts, including early ones. This shows that, besides standard varieties, one should also consider the situation of the local varieties, which in some known cases seem to behave quite differently. However, very little information is available on this, and even less is known on the vernaculars, which in some regions (like Italy and Romania) often present remarkable differences, at all structural levels, relative to the national standards. Thus, the picture I am going to present here is far from exhaustive.

2. On the evolution of PROG in Romance to Romancé languages, the copula of both types of St-PROG is a descendant of Latin STARE, rather than ESSE. However, this was not necessarily the case in the ancient stages of these languages, where both types of copulae could be encountered. As to the origin of these periphrases, cf. in particular Dietrich (1973),5 who reports abundant evidence from Late Latin, particularly from Christian texts, i.e., from a type

562

Pier Marco Bertinetto

of literature relatively close to the spoken language (suggesting that these devices must have been fairly common in the actual usage of the speakers). In Latin, we find four types of construction, which are mirrored almost exactly by the early stages of Italian (Bertinetto 1986: 134-136; the Latin examples are from New Testament versions (Vetus Latina, Vulgata) and late Latin authors; the Italian ones are authors of the 13th and 14th century): (a)

"esse + Imperfective Participle" [...] gemens et tremens eris moaning and trembling be:2SG:FUT '[...] you will be moaning and trembling [...]'

Late Latin [...]

"essere + Imperfective Participle" La Misericordia è parlante [...] the Mercy is speaking 'God's Mercy says [...]' (b)

Italian

"esse + Ablative Gerundive" Late Latin [...] erat Darius vociferando et congregando multitudinem was Darius shouting and gathering crowd:ACC '[...] Darius was shouting and gathering the people.' "essere + Gerundive" Le mani me son lavando the hands I S G J R F L am washing 'I am washing my hands [...]'

Italian [...] [...]

(c)

"stare + Imperfective Participle" Late Latin [...] stabant autem [...] scribae constanter accusantes were:3PL thus scribes constantly accusing eum [...] him 'Thus [...] the scribes were constantly accusing him [...]'

(d)

"stare + Ablative Gerundive"

Late Latin

stetit dux diu cunctando [...] was chief long_time hesitating 'The chief hesitated for a long while [...]' "stare + Gerundive" [...] stetti molt'anni libertà sognando was:lSG many years freedom dreaming '[...] for many years I dreamed of freedom.'

Italian

The progressive in Romance, as compared with English

563

Of these constructions, the first was the most frequently attested in Latin, whereas the last became the standard device in Modern Italian.6 Although this development is also basically shared by Spanish and Catalan (and to some extent Portuguese) PROG, the evolution of Italian St-PROG is almost unique, in the sense that it underwent a radical reinterpretation. The Latin precursors, as well as the early Italian attestations, show that this device could easily be used to indicate a purely durative (static) situation, rather than a true progressive one. In fact, the verbal noun often fulfilled a purely adjectival function, and as such could combine with a habitual (l-2a) or an imperative (2b) meaning, as in the following examples (taken from Dietrich 1973 and Durante 1981, respectively), in which the event is not viewed with respect to a focalized point in time, but rather relative to an unrestricted interval:

(1)

Latin (Vulgata) Erat autem docens in synagoga eorum sabbatis. be:IMPF:3SG thus teaching in synagogue fhey:GEN Saturdays 'Thus he taught in their synagogue on Saturdays.'

(2)

Early Italian (Elegia Giudeo-Cristiana, 12-13th cent. ) La notti e la die sta plorando. the night and the day be:3SG crying 'He cries night and day.' (C. Bascapé, 16th cent.) Perö lascia i piaceri [...] e sta più tosto therefore leave:IMP the pleasures and be:IMP rather pregando. praying 'Therefore, abandon the amusements and rather spend your time in prayer.'

a.

b.

Interestingly, something fairly similar occurred in Old English, where we find sentences such as example (25b) of Bertinetto, Ebert and De Groot, this volume.7 As suggested in Section 4.1 of the quoted chapter, it is likely that this is the preliminary stage in the evolutionary path followed by PROG in most languages. And there is certainly little doubt that Italian St-PROG-GER could be used in the past (up until the beginning of the 19th century, as far as the literary language is concerned) to indicate a purely durative situation, as is proven by examples such as:

564

Pier Marco Bertinetto

(3) a.

b.

Italian (G. Galilei, beginning of 17th cent.) f...] e domani vi starö attendendo amendue per and tomorrow you be:lSG:FUT waiting both to continuare i discorsi cominciati. continue the speech.PL begun '[...] and tomorrow I shall be here, waiting for you both to continue oiif conversation.' ': Ji (A. Manzoni, 19th cent.) Renzo lo stava guardando con un' attenzione estatica/ Renzo him be:IMPF:3SG watching with an attention ecstatic' come un materialone sta sulla piazza guardando al like a guy stands on the square watching at_the giocator di bussolotti. ') player of dice 'Renzo was looking at him with ecstatic attention, like a guy standing on a square and staring a player of dice.'

Example (3a) may be compared to English sentences such as: While you stay here, I'll be going home, which typically present a durative, non-focalized situation. As to (3b), consider in particular the second occurrence of PROG. In Modern Italian, these possibilities were lost (or at least severely restricted in the case of the habitual meaning; cf. Section 6.3.2). This is proven in particular by the fact that the combinability with perfective tenses, such as the Simple Past or any of the Compound tenses, has been entirely lost (Bertinetto 1986). It may be said that Italian St-PROGGER has gone all the way along what could be called "PROG imperfective drift" (cf. Bertinetto, Ebert & De Groot, this volume, Section 4.1). Indeed, it is easy to show that Italian St-PROG-GER may now be employed (with very few exceptions) only in cases of strict focalization, as in the typical 'incidential schema' 8 , where the speaker is only concerned with what is going on at a particular point in time. Italian St-PROG-GER is thus a typical instance of what in the just quoted chapter is called "focalized" PROG. Consider, as an illustration, the following examples taken from PROGQ (cf. also PROGQ:30,76,82): (4)

PROGQ:3: [Last night at 8 o'clock] when John came, Ann still WORK. Cat: ... quan en Joan va venir [SP], l'Anna encara estava treballant. [IMPF:PROG] Fr: ... quand Jean est arrivé [CP], Anne travaillait [IMPF] encore If. ... quando Gianni è arrivato [CP], Anna stava ancora lavorando. [IMPF:PROG]

The progressive in Romance, as compared with English

565

Prt:

... quando о Joâo chegou [SP], a Ana ainda estava a trabalhar. [IMPF:PROG] Rom: ... cînd a venit [CP] Jon, Ana înca lucra. [IMPF] Spn: ... cuando Juan llegö [SP], Ana todavia estaba trabajando. [IMPF:PROG] Eng: ... when John came, Ann was still working. [PST:PROG] (5)

PROGQ:32: /The pardon arrived/just when the captain GIVE the sign /to the firing squad/. Cat: ... justament mentre el capita estava fent [IMPF:PROG] el senyal Fr: It:

... justement au moment où le capitain donnait [IMPF] le signal... ... proprio mentre il capitano stava dando [IMPF:PROG] il segnale

Prt:

... exactamente enquanto о capitäo estava a dar [IMPF:PROG] о sinal... Rom: ... exact în timp ce cäpitanul dädea [IMPF] semnalul... Spn: ... justo cuando el capitan estaba dando [IMPF:PROG] la orden ... Eng: ... just while the captain was giving [PST:PROG] the sign ... As may be seen, PROG is readily used in all the languages considered, with the exception of French and Romanian. However, the situation of these two languages is not identical: in PROGQ:76, i.e., in a context very similar to PROGQ:3, one of our French informants provided an Imperfect Progressive as an alternative to the bare Imperfect, whereas PROG appeared only very marginally in the responses of our Romanian informants and, significantly enough, never in the instances that we might regard as prototypical. This shows that while PROG has a perfectly recognizable grammatical status in French, it barely exists in Romanian. What makes French different with respect to the other languages listed in (4-5) is the relatively low frequency of usage of this device, as compared with its cognates in English, Italian, and the Ibero-Romance languages. But, as I said, English differs from Italian and the Ibero-Romance languages because PROG is the only device available to express a focalized aspectual view at any temporal location (Past, Present or Future), while all Romance languages, even those where PROG appears to be fairly frequently employed, may freely resort, in examples like (4-5), to the Present, the Imperfect or the Simple Future, depending on temporal location. In fact, our informants often provided these responses as a possible option.9 To the extent that PROG is employed in contexts like the ones above, this usage ay rightly be considered prototypical, for this is precisely the type of context that •s most often referred to in the literature. But in several Romance languages, as w ell as in English, PROG is not restricted to a purely focalized interpretation. This c an be observed, most notably, in sentences containing durative adverbials, which m

566

Pier Marco Bertinetto

necessarily rule out the focalized interpretation, for the event must be conceived of as developing over a stretch of time, rather than at a particular instant. Accord­ ingly, in such cases it is appropriate to speak of "durative" PROG, as is done in Bertinetto, Ebert and De Groot, this volume. Consider the following example (cf. also PROGQ-.48-50): (6)

PROGQ:51 : [Moment by moment] the policeman TAKE NOTES of what the speaker said. Cat: ... el policia estava prenent nota [IMPF.PROG] del que deia l'orador. Fr: ... le policier notait [IMPF] ce que l'orateur disait. It: ... il poliziotto prendeva nota [IMPF] di ciö che diceva l'oratore. Pit: ... о policia estava a tomar notas [IMPF:PROG] do que о falante disse. Rom: ... polijistul nota [IMPF] ce spunea vorbitorul. Spn: ... el policia anotaba [IMPF] lo que deci'a el que hablaba. Eng: ... the policeman was taking notes [PST.PROG] of what the speaker said.

In Catalan, English and Portuguese our informants overwhelmingly yielded, at least as an alternative, a PROG response. This could in principle have happened also in Spanish, but the behaviour of our informants suggests that Spanish speakers are more cautious with using PROG in these contexts.10 As to the remaining languages, our informants behaved just as expected, given the presence of the particular adverbial employed (moment by moment), which clearly prevents focalization. Considering that Romanian makes very limited use of PROG devices, the really interesting cases here are Italian and French. Let us consider the situation of Italian (as to French, cf. Section 4 below). We saw in (2) above that in the early stages of the language St-PROG could also appear in contexts that presuppose a purely durative situation (and even, in the most extreme cases, a purely stative situation). Yet in Modern Italian, this possibility is completely lost with St-PROG-GER (as to St-PROG-INF, cf. below). Although it is not possi­ ble to state when exactly this evolution started, there is now good evidence concern­ ing the recent development of St-PROG-GER and Mot-PROG in Italian (Bertinetto 1996). The literary prose of the last two centuries shows that the use of St-PROGGER with perfective tenses, which is only compatible with a durative view, was still possible at the beginning of the 19th century. However, at that time it was already restricted to a very limited set of verbs, such as guardare 'look' or aspettare 'wait (cf. stette guardandolaspettando 'be:SP:3SG looking/waiting'). Obviously, severe lexical specializations normally indicate that the device considered is not yet fully grammaticalized, or is undergoing a process of degrammaticalization (or, possibly.

The progressive in Romance, as compared with English

567

a functional recategorization). There is thus reason to hypothesize that St-PROGGER completed its metamorphosis into a purely focalized device right around that period, and probably somewhat before in the spoken language, considering that the literary language of that time was fairly conservative. Another interesting observation emerging from the same piece of research is that, although the frequency of StPROG-GER and Mot-PROG was never very high (at least in comparison with what we observe in Ibero-Romance texts), it was nevertheless much higher in the first half of the 19th century than in the period immediately following, where the frequency of both periphrases dropped dramatically. However, in recent times (especially after World War II), St-PROG-GER gained ground again, both in written and spoken language. It has been claimed (Durante 1981) that this is due to English influence, although no real evidence has been gathered. As to Mot-PROG, it also recovered a good deal of lost ground, but is still mostly confined to the written language, being seen as a hallmark of formal style. (Andrea Villarini (personal communication) points out to me that in Lessico dell'Italiano Parlato there are 14 occurrences of MotPROG against 640 occurrences of St-PROG-GER.) Finally, there is evidence that the recent recovery of St-PROG-GER was accompanied by a considerable growth in the use of achievement verbs, which tended to be avoided with this device in the previous stages, where a striking prevalence of activities was observed.11 We can thus conclude that, due to the profound transformation that occurred in the semantic interpretation of St-PROG-GER, Italian has radically restructured this subdomain of the grammar, as compared with the Ibero-Romance languages. When durativity, rather than focalized progressivity, is imposed by the context, Modern Italian must revert to Mot-PROG (cf. Section 5 for further comments). Thus, Italian St-PROG-GER is by and large in complementary distribution with Mot-PROG regarding the features 'focalized' vs. 'durative'. Another device available in Italian to express durative progressivity is St-PROG-INF, at least for the varieties having access to it (cf. above). However, although its usage is fairly common in some spoken varieties, its presence in modern literature is negligible (Bertinetto 1996). St-PROGINF is a much more flexible tool compared to Mot-PROG, since it is also available for focalized contexts, whereas Mot-PROG is restricted to durative contexts. But despite this difference, the similarity of St-PROG-INF and Mot-PROG in durative contexts is quite remarkable, especially considering that both periphrases, as opposed to St-PROG-GER, are compatible with perfective tenses. St-PROG-GER, on the other hand, is compatible with achievement verbs, which are rejected or severely constrained by both St-PROG-INF and Mot-PROG.12

568

Pier Marco Bertinetto

3. Durative (non-focalized) PROG 3.1. Durativity in Spanish State-PROG As suggested in the preceding section, Spanish St-PROG has preserved a number of possibilities that are now precluded in its Italian cognate, and this is generally true of St-PROG in Ibero-Romance languages. As a matter of fact, Portuguese would provide an even better illustration of this issue. However, for convenience the following examples will be taken from Spanish, which may be said to occupy an intermediate position between Portuguese and Catalan with respect to the use of PROG. To start with, note that Spanish St-PROG admits perfective tenses. As observed above, this is a clear indication that this device is not restricted to focalized contexts. This fact may typically be seen in sentences where St-PROG is used in conjunction with durative adverbials, such as durante dos horas or desde las très hasta las cinco: (7)

Spanish Pedro estuvo leyendo en la cama durante dos horas / desde Pedro was reading in the bed for two hours / from las très hasta las cinco. the 3 to the 5 'Pedro read in bed for two hours / from 3 to 5.'

In these cases, the event is presented as ongoing during a stretch of time of definite duration. Other types of evidence for durativity in Spanish St-PROG are provided by Squartini (1998), in research based on an extensive survey of the specialized literature and of recent corpora of actual linguistic usage. I owe to this work most of the examples reported in this section. For instance, St-PROG is allowed in conjunction with modal verbs, a possibility that is only marginally observed in Italian, and almost invariably with an epistemic interpretation, while the Spanish equivalents may retain the deontic value: (8)

Italian (PROGQ:80) Anna deve stare facendo lezione adesso (suppongo). Ann must be making lesson now guess:PRS:lSG 'Ann should be teaching now.'

(9)

Spanish (Caracas, spoken corpus) [...] entonces todo individuo esta forzado, debe estar then each individual is forced, must be constantemente produciendo. constantly producing 'Thus, each individual is forced to produce, must constantly produce.'

The progressive in Romance, as compared with English

569

Sentence (8) was provided, as one of two alternatives, by only one of the Italian informants, and is certainly considered marginal by many speakers, who tend to reject PROG in these contexts. Besides, it is clear that this sentence must be taken in the epistemic meaning, according to which the speaker suggests a supposition about the situation developing at the speech time. Consequently, the situation is conceived of as focalized. By contrast, sentence (9) was produced with a clearly deontic intention, and suggests that the event referred to must occur during a certain interval of time, also due to the adverb employed. Futhermore, Spanish St-PROG may also appear in two coordinated sentences depicting durative events: (10)

Spanish (Colombian, spoken corpus) [...] mientras estamos conversando esta escribiendo [...] while be:PRS:lPL talking is:PRS writing 'While we talk, he writes.'

Although none of our Spanish informants exploited this solution in PROGQ:70 (Yesterday, while Ann READ in her room, Martin PLAY in the courtyard), this possibility is frequently exploited in spoken Spanish. It is quite understandable that this occurs in Spanish rather than in Italian, because of the obvious durative character of the situation envisaged. This feature is even emphasized, because of the adverb employed, in hyperbolic contexts such as: (11)

Spanish jSiempre te estas quejandol always you:RFL are complaining 'You are always complaining!'

In these contexts, Italian speakers would rather use, as an alternative to the simple Present, a periphrasis like non fare altro che + INF (lit. 'do nothing but + INF'). 15 Alternatively, some speakers would use St-PROG-INF: ti stai sempre a lamentare 'you are always complaining'. No wonder, then, that Spanish St-PROG may also colloquially appear in habitual contexts, such as: 02)

Spanish a.

No me digas que la echas de menos, porqué la Not me tell that her miss because her estas viendo todos los dias. be:PRS:2SG seeing all the days 'Do not tell me that you miss her, because you see her every day.'

570

Pier Marco Bertinetto b.

Estoy yendo al centro cada très dfas. be:PRS:lSG going to_the centre every three days 'I am going downtown every three days.'

Obviously, there may be focalized habitual contexts, such as: Whenever I arrive, he is writing. In such contexts, St-PROG would be allowed in any language possessing this device (cf. Section 6.3.2). But this is clearly not like sentences (12). It is of special interest, in this connection, to understand the meaning of constructions such as Spn Pedro estuvo dormiendo todo el dia (lit.: 'P. was sleeping [SP-PROG] the whole day'), based on the use of a perfective tense. Since these sentences are grammatical only in conjunction with adverbials expressing a delimited duration, it is clear that St-PROG retains here the perfective meaning inherent to the tense employed. On the other hand, there is good evidence that even in these cases St-PROG detelicizes any basically telic predicate. Consider the following examples: (13)

Spanish En Mérida estuvieron reconstruyendo el puente *en / durante in Merida be:SP:3PL rebuilding the bridge in / for dos afios. two years 'In Merida they kept rebuilding the bridge for two years' (ungrammatical: 'in two years.')

(14)

Spanish a. ??Ayer Pilar llego a su casa, estuvo leyendo la yesterday P. arrived to her home, be:SP:3SG reading the carta, estuvo preparando su ponencia, estuvo comiendo letter be:SP:3SG preparing her paper be:SP:3SG eating y se fue a la cama. and her:RFL went to the bed 'Yesterday, P. came home, spent some time reading the letter, preparing her paper and eating, and finally went to bed.' b. Aquel dia nos lo pasamos muy bien: estuvimos bailando, that day we it spent very well: were:SP:lPL dancing estuvimos charlando, estuvimos comiendo. were:SP:lPL chatting be:SP:lPL eating 'That was a very nice day: we spent some time dancing, chatting, and eating.'

The progressive in Romance, as compared with English

571

From (13) we learn that, with potentially telic predicates, adverbials such as en dos anos (which insist on the completion of the event) cannot be employed, in contrast to adverbials such as durante dos anos, which, although only compatible with perfective tenses, require an atelic situation (Bertinetto and Delfitto, this volume). From (14) we learn that a series of "estuvo + Gerund" forms cannot be used to suggest a strict temporal sequence of events, such that the end of one coincides with the beginning of the next, as in example (a). We can only use them as in example (b) to express a temporally unordered series of events, where each event localization may be interpreted independent of others, allowing for iterations and overlappings. In conclusion, Spanish St-PROG with perfective tenses expresses a perfective, durative and (in most cases) atelic situation. However, some caution is in order concerning telicity, because (as pointed out by Squartini, p.c.) one can find examples such as Estuvimos investigando todos los locales de Valladolid 'We kept searching all the bars of Valladolid', which seems to suggest that the search was conducted till the end. However, I am inclined to think that this is simply a pragmatic implicature, rather than a true semantic effect.

3.2. Durativity in Catalan, Portuguese and English St-PROG Admittedly, some of the uses reported in the preceding section appear to be strictly colloquial, like those illustrated in (9-11), but others are fairly common, not only in Spanish, but also in the remaining Ibero-Romance languages. Support for this also comes from PROGQ, as may be gathered from the following examples, in which I also list the responses of the English, Italian and French subjects for comparison: (15)

PROGQ.48: [Yesterday, during my sleep] Ann PLAY for two hours all by herself. Cat: ... l'Anna va estarjugant [SP:PROG] tota sola durant dues hore. Fr: ... Anne a joué [CP] pendant deux heures toute seule. It: ... Anna ha giocato [CP] per due ore tutta sola. Pit: ... A Ana esteve ajogar [SP:PROG] sozinha durante dos horas. Spn: ... Ana estuvo jugando [SP:PROG] dos horas ella sola. Eng: ... Ann was playing [PST.PROG] for two hours all by herself.

06)

PROGQ:72: [What did Martin do yesterday evening?] He STUDY from . 2 to 6, he READ the paper from 6 to 7, he EAT from 7 to 8, and then he GO to bed. Cat: our informant alternates the Simple Past and the Simple Past Progressive with the first two verbs, while the last two trigger the Simple Past only: va estar estudiant, va estar llegint, va menjar, se'n va anar al Hit.

572

Pier Marco В ertinetto

(Г

Fr: It:

both informants use the Compound Past only. three out of four informants use nothing but the Compound Past; only one (significantly, from Sardinia) alternates this tense, again with the first two verbs, with St-PROG-INF: è stato a studiare, è stato a leggere, ha mangiato, è andato a letto. Prt: all three informants alternate the Simple Past and the Simple Past Progressive with the first two verbs; only one of them extends this usage to the third verb, as in: esteve a estudar, esteve a 1er, esteve a comer I comeu, foi para a cama. Spn: one of the two informants uses the Simple Past Progressive with the first two verbs: estuvo estudiando, estuvo ley endo, cenö, sefue a la cama. Several observations are in order here. First, as can be seen in (16), Spanish is not the most liberal among the Ibero-Romance languages concerning the usage of StPROG with perfective tenses, as is shown by the behaviour of one of the Portuguese speakers.1 Second, the general allergy of Italian speakers to the usage of PROG with perfective tenses does not concern all local varieties of the language. Sardinia is one of the areas (together with large areas in Southern Italy) where St-PROGGER is used most freely. This could be one of the linguistic features left behind by the long period of Spanish domination of the island, but of course this is little more than speculation. Finally, it is interesting to note that, among the four events of (16), the ones which elicit most of the PROG responses are the first two. It will not go unnoted that precisely the first two events correspond to atelic verbs, while the last two correspond to telic ones (with the possible exception of the verb EAT, although it may easily be taken in the sense of 'complete one's meal'). This peculiar distribution of the responses suggests that the durative interpretation of PROG bears clear evidence of its link with the early stages of development of this periphrasis, when it presumably worked like an actional operator much more than an aspectual one, i.e., like a device conveying the idea of the continuous development of the event over a given interval of time (cf. Bertinetto, Ebert and De Groot, this volume, Section 4.1, and Squartini 1998, who presents convincing evidence related to the interaction of actional and aspectual values within the category of progressive). English does not significantly differ from the Ibero-Romance languages. In both cases, we can find St-PROG with perfective tenses, as in the following examples, which exhibit more specifically perfectal tenses: (17) a.

(PROGQ:81) '[I am so tired:] I BAKE all day since I got up this morning.' English I have been baking all day since I got up this morning.

The progressive in Romance, as compared with EngiiÄ b.

с

d.

(18)

a. b.

с

d.

f 73

Catalan He estât fent pa tot el dia des que m' have:lSG been making bread all the day since RFL he llevat aquest matf have:lSG got_up this morning Portuguese Tenho estado a cozinhar desdeque me levantei have:lSG been at cooking since 1SG:RFL got_up Spanish He estado cocinando todo el dia desdeque me have:lSG been cooking all the day since RFL levante. got_up (PROGQ:82) ' [When John came home yesterday,] he was very tired because he WORK hard all week.' English ... he was very tired because he had been working hard all week. Catalan estava molt cansat perqué havia estât treballant be:IMPF:SG very tired because had:3SG been working molt tota la setmana. much all the week Portuguese ... estava muito cansado porque tinha estado a be:IMPF:3SG very tired because had:SG been at trabalhar muito toda a semana working much all the week Spanish ... estaba muy cansado porque habfa trabajado toda la was very tired because had:SG worked all the semana. week

In these contexts, PROG carries an 'inclusive' meaning: it suggests that the event considered has been going on for some time up to (and including) the reference time, which in (17) coincides with the speech time. English has a marked pref­ erence for this solution in contexts of inclusivity: this is indeed the most typical function of PROG with compound tenses in this language. But in other languages this may not be the only way to convey this interpretation, as can also be gathered

574

Pier Marco Bertinetto

from our informants. The non-progressive Compound tenses are a viable alternative in the Ibero-Romance languages, provided the requirements for the usage of these tenses are met. Indeed, this consideration is simply obvious in the case of Portuguese Compound Past, which conveys almost uniquely an inclusive meaning (cf. Squartini and Bertinetto, this volume). Note that a non-progressive tense is the only response given by our Spanish informants in (18) above, although PROG is not altogether excluded in that type of sentence.17 But of course the crucial case is (17), for in some Romance languages there could in principle be competition between Simple and Compound Past. The behaviour of Spanish in this case is particularly instructive, given the broader freedom with which Spanish speakers use the Simple and the Compound Pasts. (Note however that the distribution of these tenses is to some extent, i.e., not for all speakers, regulated by the criterion of proximity, in the sense of "hodiernal/prehodiernal"; cf. Squartini and Bertinetto, this volume). Consider: (19)

Spanish Esta manana, Pedro estuvo /ha estado this morning Pedro be:SP:3SG /have:PRS:3SG been estudiando durante cinco horas. studying for five hours 'This morning, Pedro studied (was busy studying) for five hours.

Here, as opposed to (17-18), there is no inclusive meaning (a possibility that would anyway be precluded to the Simple Past Progressive). The event referred to is entirely confined in the past, i.e., in an interval of time which may even be quite remote from the Speech Time when the Simple Past Progressive is selected. Note, however, that a similar usage is also available to English. Begin ( 1996) cites quite a number of examples like: I've been talking to a clever woman friend of mine this afternoon, or What are you doing up at this hour?-Bertie had a headache ... I have been giving him an aspirin. Here again, the event is clearly confined in the past. Begin explains this usage with the notion of "partly conditioned outcome", whereby the impression of a persistent effect of the event at Reference Time (coinciding here with the Speech Time) is due to some "spin-off or accompanying effect of the activity". For instance, in the first case, the event of talking is presented as though its effects were still active at Reference Time, although the event itself was necessarily completed before. Obviously, this type of English sentences tend to suggest that the event is situated in the recent past, given the well-known restrictions impinging on the English Present Perfect; but this is also true with the Spanish Compound Past, although for different reasons (cf. above). It is perhaps appropriate in these cases to invoque the notion of "aspectual metaphor": the event is presented as though it were still going on, although the context makes clear that this is not the case. Note, in fact, that the (non progressive) Present Perfect is often ruled out completely in these contexts. For

The progressive in Romance, as compared with English

575

comparison, consider that Italian and French informants used the Compound Past in (17) and, preferably, the Pluperfect in (18).18 A good indication of the fact that English PROG may have a durative meaning is provided by the fact that in cases like (10-11), the speakers would normally employ PROG. For further support of this, consider now the following example, in which English and Italian are contrasted: (20) a. b.

с

English Nero was fiddling, when Rome burned, Italian Nerone *stava suonando/ suonava la cetra, mentre Roma Nero was playing/ play:IMPF the lyre while Rome bruciava. burn.IMPF Italian Nerone stava suonando la cetra, quando Roma Nero was:3SG playing the lyre when Rome bruciö. burn:SP:3SG

In (a), the dependent clause may receive two interpretations: a simultaneous reading, in which the fiddling and the burning supposedly have (nearly) the same duration, and an inceptive reading, according to which the beginning of the burning took place at a point in time when Nero was already engaged in his musical activity. These two readings are clearly distinguished in Italian, as shown by (b-c), exhibiting the simul­ taneous and the inceptive reading respectively, due to the imperfective vs. perfective tense employed in the dependent clause. Crucially, PROG is allowed in (c) and is ruled out in (b).

4. A comprehensive picture From what we saw above, we can draw some conclusions as to the present state of evolution of the Romance and English St-PROG devices. Consider again Table 2 in Section 4.1 in Bertinetto, Ebert & De Groot, this volume, repeated here for convenience, which exhibits what may be called "PROG imperfective drift". Ibero-Romance St-PROG and English PROG embrace stages (iii) and (iv), for tn ey may occur both in prototypically focalized contexts (cf. (4-5) above), and in durative contexts such as those exemplified in (6) and in Section 3. This shows that, Hi principle, St-PROG is not restricted to truly imperfective situations, although it shows a striking predilection for such contexts. In fact, as suggested by Squartini v 1998), in the early stages of development the component which is most involved

576

Pier Marco Bertinetto

Table 2. Diachronie development of progressive constructions in Romance (from actionality to aspect) (i)

pure locativity

=

(ii)

progressivity I

=

(iii)

progressivity II

=

(iv)

progressivity III

=

(v)

pure imperfectivity

=

Stative, durative (ex.: the meaning to be observed in some Latin examples) residually locative, durative, accessible to perfective aspect (ex.: PROG periphrases based on the verb 'come', which preserve some kind of deictic orientation) durative, accessible to perfective aspect (ex.: PROG periphrases based on the verb 'go') focalized, strictly imperfective (ex.: Modern Italian "stare + Gerund") loss of the progressive character (ex.: possibly to be observed on some non-standard varieties of Latin American Spanish)

appears to be actionality, rather than aspect proper. This also transpires through the preference for atelic predicates (with the obvious restriction concerning stative verbs) in sentences such as (16) where, in the relevant languages, St-PROG is likely to be used to express a purely durative situation. The extent to which these properties are manifested varies from language to language. As we observed, some varieties of spoken (particularly Latin American) Spanish allow for an extreme behaviour, but on the whole Standard Spanish does not seem to be as liberal as English and Portuguese (or as some colloquial varieties of Italian). Obviously, more research is needed to ascertain this. What is already clear, however, is that Italian St-PROG-GER behaves in a very different way. This periphrasis has steadily reached stage (iv), being now a purely focalized (thus, strictly imperfective) device, although it previously behaved exactly like the English and Ibero-Romance types. On the other hand, Italian St-PROG-INF has retained its ancient status, and is thus strictly comparable to its nearest cognate, namely Portuguese St-PROG-INF, which is the standard PROG device in European Portuguese. As to present day French PROG, its status is fairly similar to that of Italian StPROG-GER, although the story is quite different. The original Old French PROG periphrases were morphologically identical to the ones exhibited by Italian and the Ibero-Romance languages. However, their usage declined in the course of time, so that by the end of the 16th century they had virtually disappeared (Gougenheim 1929; Werner 1980). The "être en train de + INF" periphrasis, which in Table 1 is listed under the label "marginal type", was registered by the grammarians in its current progressive meaning only at the beginning of the 19th century, replacing the original modal (namely intentional) meaning. According to Gougenheim, in the

The progressive in Romance, as compared with English

577

17th and 18th centuries "être en train de + INF" had more or less the sense of: "être en humeur I en disposition de" 'to be in the mood to / to be inclined to'. Thus, apparently, this device entered directly at stage (iv), bypassing all previous stages. If this claim is correct, the French case is interesting both in itself, and for what it tells us about the general evolutionary picture. Although it is easy to reconstruct a locative meaning in French PROG, it is possible that this feature did not play the same role as with the other PROG devices we are considering here. It certainly was not conducive to the purely durative stage (iii). What is particularly remarkable is that, to my knowledge, the only other example in European languages of a PROG device exclusively actualising stage (iv), besides Italian St-PROG-GER and French "être en train de + INF', is the (colloquial) Albanian PROG periphrasis based on the particle po (cf. Section 2.6 of Bertinetto, Ebert and De Groot, this volume). It is thus a fairly rare phenomenon. As to the other, less prominent, Romance languages, it is worth observing that in Galician (Rojo 1974) and Occitan (Schlieben-Lange 1971) PROG is used with both focalized and durative meaning, just like in the Ibero-Romance languages analysed above (Squartini 1998). From the morphological point of view, the Occitan construction ("estre a + INF") is identical to Italian St-PROG-INF, while in Galician there is a larger variety of forms: "estar + GER", "estar a + INF", and even "ser a + INF", the first being the most frequently employed.

5. Mot-PROG As observed in Section 2, Mot-PROG may only carry a durative meaning. But there is more to say about this device. To start with, it is worth observing mat it presents a multifarious morphology, as shown in Table 1. In most languages where it appears (Catalan is a notable exception) there is the choice between 'go' and 'come' as auxiliaries. This was to be observed already in the early stages of Romance languages. It is perhaps daring to make a general statement about the difference between these two auxiliaries. I shall limit myself to noting that in Italian there is clear evidence that 'come' still implies some kind of deictic orientation, possibly at a metaphorical level (Bertinetto 1991).19 What is certainly common to all Romance languages exhibiting Mot-PROG is that the form with 'come' is definitely less frequent than the lorrn with 'go', which has reached a relatively high degree of grammaticalization, as witnessed by the semantic bleaching of the auxiliary verb.20 Although Mot-PROG was characterized above as a device conveying durativity, it interesting to observe that in a language like Spanish, where all the morphological Possibilities are fully exploited, one may contrast St-PROG and Mot-PROG in orer t0 obtain subtle semantic differences. The following example was suggested by Ignacio Bosque: ls

/ 578

Pier Marco Bertinetto

(21) a. b.

Spanish Juan estuvo colocando libros de 3 a 5. Juan was placing books from 3 to 5 Juan fue colocando libros de 3 a 5. Juan went placing books from 3 to 5 'Juan kept putting back books from 3 to 5.'

While (a) simply depicts a durative situation, (b) adds to it the idea that the event be conceived of as a sequence of identical gestures which follow each other, as though the sentence said something like: "Juan kept putting back one book after the other ..." To put it differently, (a) describes a static scenario, while (b) presents a dynamic one. To render this contrast in a sufficiently plastic way, one could legitimately claim that while (a) is a mere case of durativity, (b) is (so to say) an instance of "plurifocalization", i.e., of a situation in which every instant of the given interval is conceived of as a possible vantage point for the evaluation of the event. In the rest of this section I shall concentrate mainly on the actional restrictions that impinge on Mot-PROG. In fact, this device presents striking properties from this point of view, which differ from language to language. As to Italian, there is a notable preference for accomplishment verbs; however, activities may often be accommodated, by means of the appropriate adverbials. Compare the following ex­ amples: (22)

Italian a. *Luca andava ballando la mazurka. Luca go:IMPF:3SG dancing the mazurka 'Luca was busy dancing the mazurka.' b. Luca andava ballando la mazurka con un crescendo Luca go:IMPF:3SG dancing the mazurka with a crescendo di trepidazione e di rapimento. of trepidation and of ecstasy 'Luca was dancing the mazurka with increasing trepidation and ecstasy.' с *Luca andava gradualmente I a poco a poco ballando la Luca go:IMPF:3SG gradually / little by little dancing the mazurka. mazurka d. Luca andava gradualmente I a poco a poco scoprendo Luca go:IMPF:3SG gradually I little by little discovering la verità. the truth. 'Luca was gradually / little by little discovering the truth.'

The progressive in Romance, as compared with English

579

In (a) we have an activity verb, and this creates an unacceptable sentence, whereas in (b) the same verb yields a grammatical sentence, due to the presence of a "manner" adverbial of graduality. Note, however, that in (c) the "temporal" adverb of grad­ ually does not guarantee the desired result. The latter type of adverbial is instead perfectly compatible with the achievement verb of (d), where it plays a reinforcing role, for the sentence would be acceptable even without the adverbial. One might wonder how it is possible to employ achievements, as in (d), given that Mot-PROG necessarily requires durative situations. The answer is that whenever this periphra­ sis may felicitously be applied to an achievement, the verb is inevitably durativized, possibly via an iterative interpretation (but the latter requirement is not even neces­ sary in (d) above). More precisely, when employed with Mot-PROG, achievements take on the reading that is typical of "gradual completion verbs" (such as increase, get fatter etc.; cf. Bertinetto and Squartini 1995); i.e., they suggest that the comple­ tion of the event may be reached at the end of a gradual (and, by implicature, slow) process. Although this is the general trend, one significant exception should be pointed out. There is in fact a class of Italian activity verbs which are easily combinable with MotPROG, namely those that I propose to call "inherently intensified verbs". Among these we may for instance count salterellare 'hop about', scribacchiare 'scribble' and the like, i.e., verbs expressing the idea of an event which is carried out through actions (in most cases frequently iterated ones), similar in nature to the more neuter action to which they are related (e.g., 'jump', 'write'), but specifically connotated regarding the peculiar way in which the action is performed. In this class we may also include verbs such as scrutare 'look very carefully', meditare 'meditate' etc., which once again involve a particular insistence or intensity in the performance of the event. For instance, 'meditate' indicates more than a simple event of thinking, for it means something like 'think over and over'. Interestingly, these activity verbs may easily accept Mot-PROG without the help of any sort of intensifying adverbial, as in: (23)

Italian a.

b.

Lucia andava scribacchiando sul quaderno. Lucia go:IMPF:3SG scribbling on_the exercise book 'Lucia was scribbling on her exercise book.' Teresa andava scrutando Г orizzonte. Teresa go:IMPF:3SG scanning the horizon 'Teresa was scanning the horizon.'

To sum up, in Italian Mot-PROG shows the following actional restrictions (Berti­ netto 1997: ch. 7; cf. also Brianti 1992, Giacalone Ramat 1995a, Squartini 1998). « is preferably combinable with durative telics (although not without exceptions),

"1 580

Pier Marco Bertinetto

and accepts achievements only insofar as they are contextually turned into "gradual completion verbs". Activities are normally rejected, unless (a) they belong to the class of inherently intensified verbs, or (b) they are reinforced by means of the ap­ propriate expressions of intensification (like the "manner" adverbials of graduality exemplified in (22b) above). Note, however, that despite the telic orientation of Mot-PROG, there are clear indications that this periphrasis yields the detelicization of telic verbs, as is evidenced by: (24)

Italian a.

Filippo risolse il problema in due Filippo solve:SP the problem in two 'Filippo solved the problem in two days.' b. *Filippo andö risolvendo il problema Filippo go:SP solving the problem

giorni. days in due giorni. in two days

In (a), the event described fulfills its telic character, due to the aspectual value of the tense employed (a perfective past). The same should happen in (b), for the tense of the auxiliary is the same; yet the sentence is not acceptable. This is clearly an effect of the periphrasis. The ш-adverbial strongly requires perfectivity and telicity (Bertinetto and Delfitto, this volume); since perfectivity is guaranteed by the tense employed, the obvious conclusion to be drawn is that the periphrasis as such has a detelicizing effect, this being the only difference between the two examples pre­ sented. This may look like a paradoxical situation: apparently, one and the same device is telic-oriented, but induces detelicization. However, we should not confuse the passive actional restrictions that Mot-PROG undergoes in the selection of the ap­ propriate predicates with the active consequences that this periphrasis induces in the actional character of the verb employed. Interestingly, Squartini (1998) shows that in Spanish and Portuguese Mot-PROG exhibits different restrictions. Recall that in these languages the auxiliary 'go' can have two translations: ir and andar. The former indicates a goal-directed motion, while the latter preserves by and large its original meaning (present also in Old Ital­ ian) of an undirected motion. This accounts for the divergent meaning of the two constructions. In Spanish, "ir + GER", is preferably used with telic verbs and, when used with activities or statives, suggests an inceptive reading, while "andar + GER prefers activities and may easily be used with statives (Luis todavia anda queriendo comprar el coche 'Luis still wants to buy the car'). As often claimed in grammatical descriptions, the latter form of this periphrasis is often felt a possible alternative to St-PROG, to which it adds something like a hyperbolic meaning, not unlike the se­ mantic nuance added by siempre in ( 11). The only contexts in which the form with tr and the form with andar appear to be basically interchangeable are those expressing

The progressive in Romance, as compared with English

581

iterativity (i.e., with achievements reinterpreted iteratively). As to Portuguese, the situation is essentially like in Spanish, the only relevant difference being that "andar + GER" accepts far more accomplishments and (durativized) achievements than its Spanish counterpart. If one takes this as the decisive factor, one may conceivably claim that Portuguese "andar + GER" is more advanced in the process of grammaticalization than any of its direct competitors because it undergoes virtually no actional restrictions.21 As to the remaining Romance languages, it should be observed that Mot-PROG was originally quite widespread even in languages like French, Catalan and Occitan, where it has now become extremely rare. Although "aller + GER" was still used by French novelists of the last century (cf. the following example, taken from Flaubert's Madame Bovary: elle alla [...] montant et s'en détachant, lit.: 'she rose-Mot-PROG and detached-Mot-PROG herself), and although it may occasionally be met even in contemporary journalistic prose, the decline of this construction began as early as the 17th century. Significantly, in each of the three languages mentioned above the evolution seems to have been the same as in Italian. From the early stage in which activity verbs were frequently employed, the periphrasis has evolved into a condition in which telic predicates tend to be preferred. PROGQ was not specifically devised to investigate all possible semantic subtleties of Mot-PROG. Yet this construction appeared in a number of instances. Specifically, it was used by one Portuguese informant in PROGQ: 18 (anda a escrever), by one Spanish informant in PROGQ:51 (iba tomando notas) and PROGQ:54 (iba mejorando), by one Italian (from Sardinia) in PROGQ:52 (va dimenticando), and by the Catalan informant in PROGQ:55 (anava cobrint). It is notable that Mot-PROG made its appearance in two sentences containing "temporal" adverbials of graduality (PROGQ:54-55). In fact, although quite compatible with St-PROG, these adverbials show a high statistical co-occurrence with Mot-PROG (Squartini 1990; Bertinetto, to appear).

6- Some properties of Romance and English PROG n

the remainder of this chapter, I shall review the most notable morphological, semantic and syntactic restrictions impinging on PROG in Romance and English. Except when explicitly stated, in this section I shall not consider Mot-PROG. Of course, his survey is by no means intended to be exhaustive. Note that in Romanian the usee of PROG is very limited; thus, I shall quote this language only when relevant.

582

Pier Marco Bertinetto

6.1.

Morphological restrictions

One morphological restriction which is often cited in the literature is that concerning the Imperative mood (Hirtle 1967; Scheffer 1975). The relevant input comes here from PROGQ:73 (/For goodness sake,/ WORK when the boss comes back!). Although English Imperative Progressive is reported in the literature, our informant did not use it, thus confirming that this is but a marginal possibility. The Imperative Progressive was however employed by the Catalan informant (estigues treballant), and by two Portuguese ones (voce esteja a trabalhar (SG) / estejam a trabalhar (PL)). Surprisingly enough, two Romanian subjects offered here what seems to be a genuine PROG construction (säfii în curs de a lucra lit. 'be-Subj. in the course of work-INF'). The fact that our Romanian informants employ PROG devices only in this case and (to some extent) in relation to future-time reference (cf. Section 6.3.3), both admittedly non-prototypical circumstances, demonstrates the very low degree of grammaticalization reached by these constructions.22 As to passive PROG (cf. PROGQ:75), the only two languages which appear to use it rather freely are Portuguese (esta a ser servido) and English (is being served). The other informants produced the passive form of the Simple Present or, in order to preserve PROG, transformed the passive sentence into an active one (by means of an impersonal construction in one case, and by means of an unspecified 3PL. subject in another). The latter solutions were adopted by the two Spanish informants. This is notable because passive PROG is reported to appear in Spanish. Squartini (1998) reports the following example from a linguistic textbook: El corpus de los diccionarios espanoles ... esta siendo publicado 'The corpus of Spanish dictionaries ... is being published'. PROG with the Infinitive is also considered to be fairly marginal, although it is possible in English. Indeed, our English informant used it in PROGQ:79-80 (Tom must FEED the animals II guess/; Ann should TEACH now II guess/), and this solution was also adopted by all the French and Portuguese speakers, as well as by one Italian informant (from Central Italy). The Catalan informant employed it only in PROGQ.80. As to Spanish, example (9) attests that this form is allowed in some colloquial varieties. The reason why the speakers tend to avoid PROG with the Infinitive must be due to the relative clumsiness of the construction, rather than to any morphological restriction proper.

6.1.1.

Syntactic restrictions

Negation has no effect on the use of PROG in the languages considered (cf. PROGQj 76-78). The situation looks more varied with causative constructions (PROGQ.1920), for the various languages seem to differ as to the readiness with which they

The progressive in Romance, as compared with English

583

allow PROG in these cases. In fact, our Portuguese and Romanian informants did not use PROG in any of the quoted sentences.24 One of the major syntactic functions of PROG is of course backgrounding. However, when a series of events considered from a progressive viewpoint are juxtaposed in a text, in order to compose a complex situation in which different activities occur simultaneously (as in PROGQ:62),25 not all languages appear to be equally ready to repeat PROG in adjacent sentences. The reason for this may be purely stylistic: it may have to do with the relative heaviness of the construction, which induces the speakers to make use of possible alternatives, whenever they exist. Thus, most Romance languages tend to avoid the repeated use of PROG, resorting to simple tenses or alternating the two options. Obviously, since this possibility is not available in English, it is no wonder that a sequence of several PROG is not uncommon in this language. But it is notable that two Portuguese informants, and one Italian (from Sardinia), offered this as a possible option. This is further evidence that Portuguese, as well as some varieties of other major Romance languages, tends to expand the usage of PROG considerably. This is also confirmed by the use of PROG in two adjacent coordinated clauses (as in PROGQ:70). Once more, English has no choice but to employ PROG in both clauses, while Portuguese proves to be the most flexible among the Romance languages. The remaining languages, in fact, present PROG in only one clause, and a simple tense in the others (a solution also proposed as an alternative by the Portuguese informants). However, this is not an inviolable constraint: one of our Sardinian subjects adopted the same options as the Portuguese informants; and cf. example (10) in Section 3, showing that some Spanish colloquial varieties behave in the same way.

6.2. Semantic restrictions 6-2.1. Actional restrictions m section 5 above, the actional restrictions that impinge on Mot-PROG were described. I will now outline the situation of St-PROG (cf. also fn. 10). The most obvious restriction concerns stative verbs. Indeed, the unavailability °f PROG is often considered to be a diagnostic criterion for the individuation of stative verbs in languages like English or Romance. However, this problem is rather complex, for in quite a number of cases the same lexical entry may or may not be compatible with PROG, depending on interpretation. When this happens, there are grounds to believe that this lexical entry is in fact ambiguous between a stative and a non-stative meaning (Bertinetto 1994). Compare:

•^щящ^

584

Pier Marco Bertinetto

(25) a. b.

John resembles his father. John is resembling his father more and more.

(26) a. b.

The mountains surround the lake. The army was surrounding the enemies.

1

While the (a) sentences depict a static situation (and indeed no informant used PROG in PROGQ:41, repeated here as (26a), the (b) sentences suggest a process of transfor­ mation, to the effect that the predicate of, e.g., (25a) may be paraphrased by means of expressions such as: 'is becoming more and more similar to'. Not all languages are equally prone to exhibiting this kind of duplicity in their lexical storage. This is particularly evident with copular verbs, such as be silly, be kind, be rude, be clever etc. (but also non copular ones, such as have a head­ ache), which in English may instantiate both individual-level and stage-level predi­ cates, to repeat Carlson's (1978) terminology. This is not possible in Italian, French and Catalan. However, this is not a distinctive feature of English. In PROGQ:43 {You BE RUDE this evening), not only the English but also the Portuguese infor­ mants presented PROG as an option; and in PROGQ:42 (you BE very KIND, now!) the Spanish informants also presented this solution in addition to the ones already quoted. In fact, Squartini (1998) cites a number of Spanish examples of this type. Brazilian Portuguese is claimed to be even more flexible from this point of view, for even non-copular stative verbs like saber 'know' or poder 'can' may be treated in this way, suggesting a permanent or a temporary situation, respectively (Schmitz 1982). In fact, it was precisely the verb KNOW that elicited PROG as a possible al­ ternative, with one Portuguese informant in PROGQ:39 (/Now, unexpectedly/ Peter KNOW the answer). The verb LIKE in PROGQ:40 (/Now, unexpectedlylTess LIKE the music) yielded this option not only with all the Portuguese informants, but with one Sardinian subject and (with a question mark added) our English Subject. Thus, in general, English does not seem to be the most liberal language from this point of view.26 English, on the other hand, is unique among the languages considered here, for it exploits this possibility with a small set of (mostly) postural verbs, like stand, lie, sit, wear, and a few others. Indeed, in PROGQ:58-59 (Ann STAND in the doorway /right now/; The statue STAND in the garden /for the summer/), our English informant is the only one who employs PROG. These examples are especially relevant, because with these verbs it is not the case that PROG instantiates a non-stative meaning. Rather, it suggests the idea of the temporary validity of the (inherently stative) situation. In other words, in this particular case English does not exploit this grammatical device in order to destativize the predicate, but only to impose on it a temporal limitation. Connected with stativity is the notion of "non-agentivity", of which the last ex­ ample was an instance. Among the test sentences, PROGQ.36 (The sun SHINE)

The progressive in Romance, as compared with English

585

and PROGQ:37 (The water BOIL) were less effective, possibly for idiosyncratic reasons, in eliciting PROG as compared to PROGQ:35 (He DREAM of his girlfriend) and PROGQ.38 (The apples ROT on the tree). As to the various languages, Portuguese and English presented PROG in each sentence, while some Italian, Spanish and Catalan informants rejected it in a few cases, and only one of the French informants used it in sentences PROGQ:35 and PROGQ:38. Thus, although the single languages differ, non-agentivity as such is no hindrance to the use of PROG. This also applies to non-intentionality (PROGQ: 33-34),28 which did not prevent PROG in any of the relevant languages, although again the only languages where this was the sole solution proposed were Portuguese and English. Non-durative verbs like THROW, REACH, GIVE (PROGQ:30-32), on the other hand, seem to create serious difficulties for French speakers compared to the others, who used it to a greater (Portuguese and English) or lesser extent. This situation is also confirmed by the usage of the verbs LEAVE and DIE in PROGQ:56-57, two sentences devised with the purpose of checking a specific effect of PROG as applied to non-durative verbs, i.e., the possible emergence of an imminential meaning. Further inquiries with native speakers made it clear that the restriction concerning nondurative verbs is not absolute in French, but rather corresponds to a strong tendency. One special case of non-durative verbs is represented by the so-called phasal verbs like begin and FINISH etc. (PROGQ:23-26). Here again, only French informants avoid PROG.30 Finally, it should be noted that in principle no restriction is to be observed concerning verbs of motion like GO OUT and FLY (PROGQ:21-22), which even in French may be associated with PROG in the appropriate context. This is also true of the verbs GO and COME that provide the auxiliaries of Mot-PROG. On the other hand, postural verbs like SIT and HANG in PROGQ:28-29 never elicit PROG in Romance (unlike English), nor do they constitute, as in some Germanic languages, the basis for a PROG construction (cf. Ebert, this volume). 6.2.2. More on aspectual restrictions In Sections 4 and 5,1 observed that St-PROG and Mot-PROG are in principle compatible with both the perfective and the imperfective aspect. However, the situations of the two constructions are not completely identical. The latter appears to be intrinsically available to any kind of aspectual value, whereas the former is liable to be caught in what I named above "PROG imperfective drift", which indeed explains the development of the purely focalized meaning of Italian St-PROG-GER, as well as the transformation of PROG into a general purpose imperfective form in quite a few languages (cf. Section 4.1 of Bertinetto, Ebert and De Groot, this volume). « is appropriate to add here a few more observations about the habitual aspect an d its combinability with PROG devices. As we saw in (1-2) above, in their very

586

Pier Marco Bertinetto

early stages these periphrases had easy access to this aspectual value. However, this possibility was completely lost in the course of time with Italian St-PROG-GER, and severely constrained in the remaining Romance languages and in English. As a matter of fact, none of our informants used PROG in PROGQ.2 (A: What does Ann do every Saturday morning? B: She CLEAN the house) and PROGQ:4 (Last year we usually CLEAN the house on Saturdays). All speakers used a simple tense (in PROGQ:4, our Romance informants used the Imperfect or, as one Italian and one Spanish speaker did, a habitual periphrasis with the Imperfect).31 Nevertheless, the co-existence of PROG with the habitual interpretation can frequently be observed in any of the languages considered, provided correlative connectives such as whenever appear: (27)

Eng: Fr: It: Spn:

Whenever I checked, he was working. Chaque fois que j'ai contrôlé, il était en train de travailler. Ogni volta che ho controllato, stava lavorando. Cada vez que yo contrôlé, él estaba trabajando.

These sentences do not constitute any problem whatsoever, for they even allow a focalized interpretation of PROG (in contrast to example (12) above). The habitual event is viewed as occurring at some particular points in time, which repeat themselves more or less regularly. But if we depart from this kind of correlative contexts, we may find that not all languages are equally ready to accept PROG in truly habitual situations, where the recurring event is not focalized by means of an incidential event which isolates single instants as evaluation points. Indeed, our English informant was the only one to use PROG in PROGQ:63 (At that time, he GO to dance every Saturday). And this is substantially confirmed by the following example (or similar ones, inspired by Hirtle 1967 and Leech 1971), despite the presence of an incidential clause:32 (28)

Eng: Peter is often smoking a cigar when he comes in. Cat: ?En Père esta sovint fumant un cigar quan arriba. Fr: *Pierre est souvent en train de fumer un cigare quand il vient. It: *Pietro sta spesso fumando un sigaro quando viene. Pit: О Pedro esta frequentemente afumar um charuto quando chega. Spn: ?Pedro frecuentemente esta fumando un cigarro cuando llega.

One type of contexts rather close to habituality, although presenting peculiar char­ acters of its own, is that exemplified in (11) (cf. also fn. 15). The hyperbolic meaning of such sentences suggests that the event tends to repeat itself with rather unusual frequency. Besides English and Portuguese, whose behaviour is predictable, Catalan and Spanish also have free access to this usage (at least in the colloquial variety), while in French and Standard Italian this possibility is ruled out altogether, due to the focalization requirement that characterizes PROG in these languages:

The progressive in Romance, as compared with Engtiaft (29)

587

Cat: En Père esta sempre fumant un cigar. Spn: Pedro esta siempre fumando un cigarro. 'Peter is always smoking a cigar.'

However, in colloquial Italian one may hear sentences like the following, which do not seem very far from the hyperbolic contexts illustrated above:33 (30)

Italian E' un mese che me lo sta dicendo. is one month that me:DAT it:ACC is saying 'He's been telling me for a month now.'

Even more common are sentences like: Stapiovendo molto, quest'anno 'It is raining a lot, this year', which express again a meaning of insistence, or (in Blansitt's 1975 terms) a "generic" meaning close to habituality. Also somewhat related to habituality are the "interpretative" uses of PROG, which are not uncommon in English. What is typical of these sentences is that a given action performed by somebody runs parallel to an equivalent action, to which a positive or negative value (possibly a merely explicative value) is assigned (König 1995). Consider PROGQ:64: If you insist in calling me Fred, you INTRUDE in my private life. Given the correlative structure of these sentences, they bear a resemblance to examples such as (27). PROG is consistently used here by the English, Portuguese and Spanish informants, while avoided by the others. However, in the related sentence PROGQ:65 (As soon as you start asking what is the use of education, you ABANDON the basic assumptions of any true culture), two of the Italian informants also offered this alternative, together with the preceding subjects. Once more, we find English and Portuguese in the lead, with French far behind.34 6.2.3. Temporal restrictions There are no restrictions concerning temporal reference in the languages considered. PROG may occur with past, present or future localization of the event. In particular, PROGQ:83 (If you come at Bo' clock, I still COOK) was specifically devised in order to test the availability of PROG with the Future tense, which in some languages is claimed to be subject to restrictions. With the exception of French, Future PROG (or something close to it) was used, or at least offered as an alternative, by virtually a " informants. One surprising finding is represented here by two Romanian informants, who proposed the following quasi-PROG construction: ...sä mä gäsesti încà gätind 'you will find me still cooking'. These data suggest that, except for French ад й to some extent Romanian (two languages known for the limited exploitation of PROG in general), there is no constraint at work, provided the context is felicitous

588

Pier Marco Bertinetto

enough. The observation concerning the restricted use of PROG with future tempo­ ral reference has thus more to do with pragmatics, namely with the relative rarity of such contexts in actual communication, than with morphology or semantics.35 One notable fact is the use of PROG in contexts corresponding to a sort of ex­ tended present (but similar examples may be built with past temporal reference). PROGQ:61 is an example (The boss TYPE his own letters, while the secretary is ill). The only informants who use PROG in this case are, once more, the English and Portuguese ones. This possibility seems to be totally excluded in French, while in the remaining languages, choices may vary depending on register. In the standard varieties, speakers tend to avoid this usage. Perhaps the most striking feature of English PROG, as compared to all the remain­ ing languages considered here, is the possibility of expressing future-time reference, as in sentences like: Ann is leaving tomorrow (cf. PROGQ:66; cf. also PROGQ:6769). Note that in these cases the future-time reference is conveyed by the Present Progressive, rather than by the Future Progressive, as in sentences like: Tomorrow I'll be leaving (cf. also the comment to sentence (3a) above). This property of PROG, although rather rare, may be found in other European (as well as non-European) lan­ guages. This is notably the case in Icelandic (Ebert, this volume) and to some extent also in Finnish (Tommola, this volume), but also in Judeo-Spanish, and in some colloquial varieties of Latin American Spanish (Squartini 1998), as well as in some colloquial varieties of Southern Italian (like in Naples; cf. Gliela sto passando subito 'I am going to pass her [on the phone] right now'). It is not easy to understand how this use may have arisen. The hypothesis that most obviously comes to mind is that it is somehow related to the imminential meaning often expressed by achievement predicates under PROG (cf. Section 6.3.1), which conveys something close to a futural sense. The data gathered from languages other than English are too scanty to allow us to put forth a hypothesis as to the actual meaning of PROG with future-time reference. As to English, Haegeman (1981) suggests, among others, the following facts. First, the Simple Future differs from both the Present and the Present Progressive because it is not oriented towards the speech time: (31) a. ??I will already meet John for lunch and Ann for dinner; I cannot have any other appointment, b. I already meet John for lunch and Ann for dinner; I cannot have any other appointment, с I am already meeting John for lunch and Ann for dinner; I cannot have any other appointment. Here, the relevance of the speech time is ensured by the adverb already, and the con­ trast of (a) with (b-c) is quite evident. Moreover, the Simple Future is characterized

The progressive in Romance, as compared with English

589

by what may be called "subjective speaker-commitment", while the two remaining tenses are rather characterized by some kind of "objective factual commitment". This confirms again the same distribution of grammaticality judgements among the three tenses: (32) a. ??The Queen, who will open Parliament tomorrow, may be detained at the airport. In this case, Prince Charles will take her place. b. The Queen, who opens Parliament tomorrow, may be detained at the air­ port. In this case, Prince Charles will take her place. с The Queen, who is opening Parliament tomorrow, may be detained at the airport. In this case, Prince Charles will take her place. The event designated in (32) may be regarded as fairly objective, inasmuch as it is a scheduled one. The two Present tenses seem to be highly preferable in these contexts. On the other hand, in order to have a perfectly felicitous use of the Present Progressive, as opposed to the Simple Present, the scheduled event should best be conceived of as fairly exceptional, rather than predictable and routine: (33) a. Who is being Santa Claus at the party tomorrow? b. ?Who is being captain of the team tomorrow? (34) a. ?The train is leaving tomorrow at 5. b. The train leaves tomorrow at 5. The difference in (33a-b) is due to the fact that the presence of a Santa Claus must be considered as a fairly uncommon event, while there always ought to be a captain of a team. As to (34a-b), these sentences should be regarded as statements uttered while consulting the timetable, rather than statements concerning a non-customary event. If the train were to leave at 5 for some exceptional reason, then the Present Progressive would be perfectly acceptable.

590

Pier Marco Bertinetto

Appendix. Responses to the Progressive Questionnaire by Romance and English informants Legenda: A slash separates alternative responses provided by the informant. A blank indicates a missing or irrelevant response. Sometimes, the informant provided an irrelevant response in addition to a relevant one; in these cases, only the latter was reported (e.g., in sentence 53 a number of informants provided a perfecta] tense, which obviously is not the intended answer, although it sounds plausible in the given context). + +Inf +Mot +% +? +* 0 \ adv (-caus) cont hab imm imps main mod modep N rec Sard. sb seguir

standard PROG device Infinitival State-PROG construction (cf. Table 1) Motion-PROG construction (cf. Table 1) PROG device used for some, not all verbs in the sentence (mostly depending on actionality) the informant expressed some reservations on her/his own response possibly a (not yet grammaticalized) PROG device no PROG device was used sentence not presented to the informant an appropriate temporal adverb was added the intended causative construction was avoided by the informant continuous periphrasis (It. "non far altro che + Infinitive", Fr. "ne pas s'arrêter de /ne pas cesser de + Infinitive") habitual periphrasis imminential periphrasis impersonal construction PROG device in main clause only modal verb construction modal verb construction with epistemic meaning nominal construction 'recent past' construction (Fr. "venir de + Infinitive") Sardinian informant subjunctive with iussive meaning continuative periphrasis "seguir + Gerund"

The progressive in Romance, as compared with English

591

+®+®«+++++++++++++++++++++++++++++++

+ s + s s + + + + + + + + + + + + + + + + + + + + +s®s>s+< © _

s

о

+ S> + 5s> + + + + + + + + + + + + +

^^ ^S £ 3 ©> © © © © ©

6 © © © © © © © © © © © © © ©


© © © © © © © © © © © © © < © © © © ©

-f- H—Ь + + © Ф © t

++

+ © + © ©

Progressive markers in Germanie languages

1.

Introduction

1.1. The state of the art

Progressive markers have never been treated systematically in any grammar of a Germanic language other than English. (The English progressive was not part of the present investigation, as it is dealt with by Bertinetto (this volume)). The reason for the neglect may lie in the fact that the progressive is nowhere grammaticalized to the same extent as in English. This is reflected on the one hand in the optionality of the markers, on the other in a variety of alternative forms. The progressive constructions are, however, mentioned in various subchapters of grammars, sometimes in a chapter on participles (Koefoed 1969: 188) or on the use of the preterite (Jones & Gade 1981: 119), sometimes only in a footnote (Duden 1984: 94, Erben 1972: 75). In the German tradition they are subsumed under 'durative aspect' or 'aktionsart'; in grammars with an Anglo-Saxon orientation they are often presented as optional possibilities for translating the English progressive. Kress (1982: 159f) has a whole chapter on the Icelandic progressive formed with ad, but refers to the postural verb construction only in a footnote. Hansen (1967: 3031) classifies the Danish postural verb construction as a durative aktionsart/aspect, whereas vœre ved at I vœre ifcerd med is said to indicate non-completed activities. The Dutch grammar by Geerts et al. (1984) offers the most explicit discussion, devoting several pages to both the postural verb construction and aan het V zijn. Both are, together with continuative, subsumed under 'duratief aspect'. Of the few recent articles, Andersson (1989) deserves special mention. Andersson describes the use of the German am-form in the Rhineland dialect, but he also gives a brief summary of the Standard German progressive markers which gives more ^formation than any of the grammars. Boogaart (1991), van der Hauwe (1992) and ßhatt & Schmidt (1993) are attempts to explain the Dutch and German progressives 111 theoretical models. I have earlier published two articles on Fering and on German a nd Dutch progressives (Ebert 1989, 1996), based mainly on native speaker intuition and/or written prose.

606

Karen H. Ebert

1.2. The data The investigation of tense and aspect in EUROTYP was based on questionnaires. Answers to the Progressive Questionnaire (PROGQ, see p. ??? in this volume) were provided for the following languages: Icelandic: Swedish: Danish: North Frisian1: Frysk: Dutch: German: Ziiritüütsch:

1 7 3 6 (4 Fering-Öömrang, 1 Wiidinghiird, 1 Mooring) 1 (based on several native speakers' judgments) 2 8 (5 standard colloquial2, 1 Rhineland, 2 mixed wi influence) 4

•«

We collected no data for Faroese, Norwegian, Low German, and Yiddish, which will be mentioned briefly in part 5. The different forms used in the answers to PROGQ are listed in Table 4 at the end of this article. Eight questionnaires have been neglected in this Table: three Swedish ones which do not contain a single progressive construction; the questionnaire from the Mooring dialect of North Frisian, as practically only lib forms (see Table 1) are used; two German questionnaires which have progressive markers only in the incidential schema3 and in PROGQ: 1. These results are, of course, interesting in themselves, but they need not be listed. Two of the Züritüütsch questionnaires, which did not yield anything new, were neglected mainly for reasons of space. Evaluation of the PROGQs was problematic for several reasons. Some consultants tried to list all possibilities, whereas others gave only the translation that came to mind first. Some checked and discussed with several native speakers, so that their questionnaire already represents a broader consensus (e.g., Frysk). The informants did not always understand what the intended meaning was. If the misunderstanding is obvious, I have left a blank in Table 4. Very few informants answered the theoretical questions. In some of the languages the progressive constructions belong to an informal style. They are frequent in the spoken language, but avoided in writing. In a few cases I had the possibility to conduct an oral interview following the completion of the questionnaire; the results were sometimes rather different, especially for German, where many dialectal and sociolectal differences exist. A systematic comparison of the written and the elicited data on one hand with spoken discourse on the other would be interesting, but was beyond the aims and possibilities of the project.

Progressive markers in Germanic languages

1.3.

0B&

Inventory of progressive markers

t

Most Germanic languages have at least two types of constructions used in the typical progressive contexts: I. Postural verb constructions: (POS) II. Prepositional constructions: (PREP) III. 'hold' constructions: (HOLD)

"sit" + to + INF (Frisian, Dutch) "sit" + and + V (Scandinavian) be + in/at + the + INF (Dutch, Frisian, German) be + at + to + INF (Danish) be + to + INF (Icelandic) hold on/in (Swedish, Norwegian, Yiddish)

The postural verbs include the verbs for 'sit', 'stand', 'lie', 'hang' and 'go (around)'. The verb 'go' in a POS-construction always indicates undirected motion. German and Züritüütsch lack the POS-construction. The PREP- and the HOLD-constructions are used in roughly the same way, and languages have either one or the other. I have distinguished a subgroup PREP lib for Frisian and German forms with the preposition Ы, bei 'at, near', which is restricted to agentive verbs. The expressions listed under IV in Table 1 all emphasize the dynamic, active char­ acter of the event, comparable to French être en train de or English be busy doing. All are marginal; only a few forms were used in the questionnaires (see Appendix).

Table 1. Progressive markers (3rd person sing, forms) I POS

IIa PREP

Ice Swd Dan Wiid FerÖöm Frysk Dut Grm Ziiri

sittur og sitter och sidder og sät to/än sat tu sit te zit te

er аб

Far Nor Yid LowGirm

sit og star og

erog

sit un

is an t

er ved at äs oon't as uun't is oan 't is aan het ist am isch am

IIb 'bei'

III HOLD

IV BUSY

hâlla pâ och/att

er i färd med att er i gang med

äs bai to/än as bi tu/an is oan e gong mei is bezig te ist dabei zu isch draa z'

ist beim

er fâast viö at holde pâ (med) halt in is bii un

608

Karen H. Eben

2. Use of the progressive markers 2.1. Test cases The PREP and POS-constructions were used in typical progressive contexts, e.g., as answers to the question 'What is X doing right now?' (PROGQ: 1, 6-18) and in the incidential schema (see (3)). (As the structure of all Germanic languages is very similar to that of English, I have chosen to provide word-by-word rather than morphemic glosses, which would complicate the reading unnecessarily. I also gloss sentences with parallel constructions together for reasons of space. Icelandic ad corresponds to both 'at' and 'to'). (1)

(PROGQ: Ice Dan Wiid

l) 5 Hun Hun Jü she

Swd

Hon hàller pâ att arbeta / hâller pâ och arbetar. she holds on to work holds on and works

FerÖöm Frysk Dutch Grm Ziiri

(2)

er er äs is

ад at to to

ved bai AT

vinna. arbejde. äprüüten. work / clean_up

Hat as uun r köögin. Hja is oan 't iten-sieden. Ze is aan het koken. Sie ist am Kochen. isch am choche. Si IN/AT she is the cooking6 'She:is working / cleaning up / cooking.'

(PROGQ: 7) Han Swd Han Dan Wiid Hi he

sitter sidder sät sits sat sits

och og an and

uun't in_the

läser lasser löst reads

tidningen. avisen. et bläär. the newsp

FerÖöm

Hi he

bleed newspaper

tu to

Frysk Dut

Ну sit in krante te Hij zit de krant te he sits a/the newspaper to 'He is reading a/the newspaper.'

leezen. read lêzen. lezen. read

Progressive markers in Germanic languages (3) a.

609

(PROGQ: 3) pegar Jon kom, da John kom ... vinna. . var Ice Anna enn ad . var Anne stadigvaek ved at arbejde Dan work:INF . was A. still AT att arbeta . hall Ann fortfarande pâ Swd to work: INF ^ held Ann still on leezen > . wiar Anne noch uun 't FerÖöm lêzen Frysk .wie Anne noch oan 't lezen .was Dut Anne nog aan het Lesen. .war Anna noch am Grm л was A. still IN/AT_the reading (lacking in Icelandic and German) .sad A. og arbejdede. Dan Swd .satt A. och arbetade. sat A. and worked '."Л: FerÖöm .seed A. noch tu leezen. Frysk . siet A noch te lêzen. .zat A. nog te lezen. Dut sat A. still to read '[When John came], Anne was still working / reading.'

As a rule, progressives were not used in the negative test sentences PROGQ: 2, 4, 5, 71 exemplifying habitual (but cf. (33)), repeated or sequential actions. No progres­ sives were used with stative verbs in any of the languages (PROGQ: 39^3). Unlike in English, it is not possible to use a progressive to express temporary states. Thus none of the other Germanic languages can have a progressive marker in the transla­ tions of: She is sitting in the kitchen (PROGQ: 28), The shirt is hanging on the nail (PROGQ: 29), You are being rude this evening (PROGQ: 42). Progressives generally do not combine with adverbs that specify a limited dura­ tion. No PREP- or HOLD-constructions were used in PROGQ: 48. The fact that a number of informants in various languages chose a POS here (cf. (29)) suggests a different status for this construction.

2-2. Transitive verbs and incorporation The North Frisian and Standard German PREP and POS do not combine with a di­ rect object. In order to use a progressive, the verb phrase has to be intransitivized У Way of incorporation. Incorporation is common, also in Frysk and Dutch7, if the combination verb + noun designates a typical activity. This characterization is nee-

610

Karen H. Eben

essarily vague, as no well-defined rales exist and there is much individual variation. Thus peeling potatoes and catching butterflies are obviously 'typical' activities in Germanic societies, whereas catching elephants and writing a thesis are not. If the object cannot be incorporated, North Frisian and German speakers have to resort to some other construction, as in (5b). (4)

(PROGQ: Frysk Dutch FerÖöm Grm

(5) a.

b.

12) Hja is oan 't jirpel-skilen/ is j . Ze is aan het aardappel-schillen/ is â. Hat as uun 't eerdaapler-skelin/ *as e. Sie ist am Kartoffel-schalen/ *ist К. she is IN/AT_the potato(es)-peeling is p. 'She is peeling potatoes.'

(PROGQ: 14) Frysk Hja is trije kilo jirpels she is 3 kilo potatoes Dutch Ze zit drie kilo aardappelen she sits 3 kilo potatoes FerÖöm Hat as diarbi, 6 pünj eerdaapler Grm Sie ist dabei, 6 Pfund Kartoffeln she is there_AT 6 pound potatoes 'She is peeling 6 pounds of potatoes.'

oan 't skilen. aan het schulen. uun r skelin. am Schälen. IN/AT_the peeling

oan t skilen ATjhe peeling te schulen. to peel tu skelin. zu schälen. to peel

There is an alternative construction which transforms the direct object into a prepositional phrase, thereby detelecizing it (cf. also FerÖöm (2)). (6)

(PROGQ: 18) FerÖöm Hat skraft bi sin doktorarbeit. Grm Sie schreibt an ihrer Doktorarbeit. she writes at her dissertation Dutch Ze zit aan haar proefschrift te werken.1 she sits at her dissertation to write/work 'She is writing her dissertation.'

In the Rhineland dialect, am is used with transitive verbs just as naturally as with intransitives (cf. Andersson 1989; Bhatt & Schmidt 1993)." Some Zürirüütsch speakers incorporate definite noun phrases and even locative complements. (The Züritüütsch examples are rendered in the transcription used in the questionnaires.)

Piqjpeeive marfef» in Germanic languages (7)

(PROGQ: 13) GrmRhi Sie ist die Kartoffeln am Schälen. she is the potatoes AT_the peeling Ziiri Si isch (grad) am t' herdöpfel schele. she is just AT the potatoes peel 'She is peeling the potatoes.'

(8)

(PROGQ: 16) GrmRhi Sie she Ziiri Si she 'She

611

ist die Hühner am Raus-jagen. is the chickens AT_the out-chasing isch am t' hiiener us em huus jage. is AT the chickens out the house chase is chasing the chickens out of the house.'

Locative phrases can be incorporated together with objects to some degree also in Standard German and Frisian. Object, locative and verb together are then quasilexicalized. Note, however, that definite objects are totally unacceptable. (9) a.

FerÖöm Grm

b.

Frysk FerÖöm

Jo san uun't (*dön) köölen-deel-uun-kääler-dreegen. Sie sind am (*die) Kohlen-in-den-Keller-tragen. they are AT_the the coals-(down)in-the-cellar-carrying 'They are carrying coals down into the basement.' Hja binne oan't hea yn 'e skuorre bringen. Jo san uun't fooder iin uun skini keeren. they are in_the hay into in the barn taking 'They are taking hay into the barn.'

2.3. Combinability with tense, voice and modal verbs There are in principle no restrictions for the combination of progressives with tenses. As marking of future time reference is not common in any of the languages, it is not surprising that we found only present progressive forms in future contexts (e.g. PROGQ: 83). Only the Icelandic future marker cannot be combined with the progressive (Kress 1982: 159). The Dutch future marker is more acceptable with aan't than with POS (De Groot 1992: 7). In Fering-Öömring, inserting wal or skal with a Progressive marker invariably yields a modal interpretation (cf. Ebert 1994a). (Ю) a.

(FTRQ: 2, Dahl 1992: 64) Swd Han kommer att hâlla pâ att skriva ett brev. he comes to hold on to write a letter

612

Karen H. Ebert Dutch

b.

Grm

Hij zal brieven aan het schrijven zijn Hzüten te schrijven. he shall letters AT the writing be sit to write 'He will be writing a letter/letters.' Wenn du nachhause kommst, werde ich am when you home come FUT I ATjhe Briefeschreiben sein. letter-writing be 'When you come home, I will be writing letters.'

In PROGQ: 81 (=lla) with a present perfect both progressive constructions were used, though POS is somewhat more natural in Dutch and the Frisian languages.13

(11) a.

b.

(PROGQ: 81) Ice ég er buinn ад vera ад baka l allan dag. I PF be AT baking in all day Swd Jag har hällit pâ att baka hela dagen. I have held:PP on to bake whole day:DEF FerÖöm Ik san di hiale daai uun 't baagen weezen. Frysk Ik ben de hiele dei oan 't bakken west. Grm Ich bin den ganzen Tag am Backen gewesen. I am the whole day AT_the baking been FerÖöm Ik haa di hiale daai stänen tu baagen. Frysk Ik ha de hiele dei stien te bakken. Dutch Ik heb de hele dag staan bakken. I have the whole day stood:PP to bake 'I have been baking the whole day.'

Icelandic is the only language where a past perfect was used in PROGQ: 82. Combinations of past perfect with progressive markers seem somewhat odd in the other languages; I have only found a single natural occurrence with a POS. The example translates naturally into Fering with a POS, but not with PREP.

(11) с

Dan

[Pâ politigârden i Malmö bed Per Mânsson tandstikkeren over] som han havde siddet og tygget pâ,... which he had sat:PP and chewed:PP on (Sjöwall &Wahlöö, cited from Haberland 1978: 65)

Progressive markers in GenMBC languages FerÖöm

613

[Üüb a politsei uun Malmö beed Per M. det swaawelstook troch,] huar hi üüb tu kauin seeden hed I where he on to chew sat:PP had ?huar hi uun't kauin üüb weesen wiar '[In the Malmö police station Per Mânsson bit through the match] he had been chewing on.'

The combination of a PREP with a modal verb in PROGQ: 79, 80 yields only an epistemic reading (12a). The two sentences in question do not invite a POS, but this construction combines freely with modal verbs in the deontic meaning. In sentences ( 12b, c) PREP is excluded. (12) a.

b.

с.

(PROGQ: 79) Swd Tom mäste hâlla pà och mata djuren. T. must hold on and feed animals.DEF Dutch Tom moet de dieren aan het voederen zljn. FerÖöm Tom mut uun't fulrin weez, T. must the animals AT_the feeding be 'Tom must be feeding the animals.' FerÖöm Skääl du imer sat tu liaren? must you always sit to learning 'Do you have to sit and learn all the time?' FerÖöm Wi maad äi linger üüb a bus stun tu teewen. we liked not longer on the bus stand to waiting 'We did not want to wait for the bus any longer.'

The progressive is excluded with 'be'-passives and with imperatives except in Icelandic. It is possible with the Scandinavian passives formed with blilblive 'become' and with some passives in -s. (13) a.

b.

с

(PROGQ: 75) Icelandic paö er veriö ад bera matinn a borö. it is become:PP to bring meal:DEF to table 'Dinner is being served.' (PROGQ: 3e) Swd Hur länge har den här bron hâllit pâ att hygga-s? how long has this here bridge:DEF held on to build-PASS 'How long has this bridge been being built?' (PROGQ: 19) Swd H&n hâllerpâoch b\ir klippt. he holds on and becomes cuf.PP

614

Karen H. Ebert Dan

Han er ved at blive kuppet. he is at to become cuf.PP 'He is getting a haircut.'

••'•

2.4. Combinability with types of verbs The progressive constructions are excluded with stative verbs. They are common with dynamic verbs, both atelic (activities) and telic (accomplishments). The restriction of PREP and POS to intransitive constructions in North Frisian and German is syntactic rather than semantic. Telic intransitives do not pose a problem with uun'tlam, although they are somewhat odd with POS (cf. part 3.1). The progressive is even obligatory in some cases, for example in (14a), if the people were later saved from starvation. Replacing the progressive by a simple form in (14b) would describe the situation as much less dramatic. (14) a.

FerÖöm Grm

b.

Grm

c.

FerÖöm Grm

Jo wiar uun't ferhongrin,johedwegenniksrochtstu Sie waren am Verhungern, sie hatten Wochen nichts they were AT_the starving ... iidjen fingen. Rechtes gegessen. 'They were starving, they had not had proper food in weeks.' Alle Varianten des Nordfriesischen sind am Aussterben. all varieties of North Frisian are AT_the out_dying 'All varieties of North Frisian are dying out.' Hi as uun 't promoviirin. Er ist am Promovieren. he is AT_the dissertation_writing 'He is writing his dissertation.'

If a progressive is used with a verb that is normally conceived of as momentaneous, it refers to a short pretransformative phase14, or it has the meaning of 'almost'. No POS is possible here. (15) a.

FerÖöm Grm

Ik wiar uun't/*laai tu tusliapen, iar at telefon Ich war am Einschlafen, als das Telefon I was at_the / lay to falling_asleep when the phone klingert. klingelte, rang 'I was falling asleep when the phone rang.'

Progressive markers in Germanic languages b.

Dan

615

Jeg var ved at falde i s0vn. I was at to fall in sleep 'I almost fell asleep.' (no progressive reading possible)

The progressive is generally excluded with directed motion verbs. Cases with momentaneous motion verbs like (16) are interpreted in the same way as (15a), i.e., referring to a short time span before the transition point. (16) a.

b.

(PROGQ: 21) Grm Sie ist (gerade) am Rausgehen. Züri Si isch (grad) am usegaa. 'She is going out right now.' (PROGQ: 56) Swd Taget hâllerpà att gâ. Dan Toget er ved at k0re. 'The train is about to leave.'

;

'

No progressives were used with phasal verbs in Dutch, standard colloquial German, North Frisian or Frysk. The few cases where a progressive was combined with 'start' or 'finish' (PROGQ: 23-26) have an imminential meaning. The first Swedish informant (see Appendix) uses hâlla pâ att, which generally can express imminential or progressive in these cases, whereas he has hâlla pâ och with progressive meaning.15 Agentivity plays a major role in the choice between PREP and POS. With nonagentive verbs POS is clearly preferred in the languages which have it. The distribution of PREP vs. POS will be dealt with in section 3. The 'busy'-constructions listed under IV are explicit means of indicating that a person is in the process of doing something. They combine only with agentive verbs, as do the PREP-constructions listed under lib. (17)

Dan Dutch FerÖöm Grm

*Han er i gang med at sove. *Hij is bezig te slapen. *Hi as diarbi tu sliapen / *bi tu sliapen. *Er ist dabei zu schlafen / *beim Schlafen. 'He is sleeping.'

2.5. Motion progressive The theoretical part of the PROGQ contained a question about a 'motion progressive' (Part II, le) and a 'locomotive' (Part II, 2c). Of the informants who answered the theoretical part, all but one stated that there were no such constructions in their language. Nevertheless the following sentences were given in the translations of the 4uestionnaire sentences:

616

Karen H. Ebert

(18) a.

b.

с

(19)

Danish (PROGQ: 1) Hun gâr rundt og laver mad. she goes around and makes food 'She is preparing dinner. / She is running around preparing dinner.' (PROGQ: 52) Han gâr rundt og glemmer folks navne. he goes around and forgets people's names 'He keeps forgetting people's names.' (PROGQ: 44) Hun gâr og danser ved siden af. she goes and dances at side:DEF off 'She is dancing in the next room.' Wiidinghiird (PROGQ: 33) Hi lapt steeriwäch an fernärmed sän nääber mä sin he walks continuously and insults his neighbor with his '''• dääsie fraage. stupid questions 'He goes around insulting his neighbor with his stupid questions.'

The Danish consultant explains: One would normally say Hun star og laver mad ('stands and'); with gâr rundt one gets the impression that the person is moving around "like a wild hen". A negative connotation of this construction is also mentioned in the Dutch grammar (Geerts et al.: 538). A negative emotion can, however, also be expressed by (other) postural verbs (cf. (21a), (26)), and a motion verb construction can have a neutral interpretation, parallel to that of POS-constructions, as in: (20) a.

b.

Danish (Jones & Gade 1981: 119) Stephen gik og glaededesig til sin f0dselsdag. S. went and looked_forward to his birthday 'Stephen was looking forward to his birthday.' Dutch Casper was ziek en Hep te hoesten. FerÖöm Kasper wiar kraank an lep tu hoostin. C. was sick and walked to cough 'Casper was ill and was (going around) coughing.'

A negative interpretation seems to be invited by the particle 'around', which has the same effect with 'sit' or 'stand' (21a), but in some languages 'around' does not exclude a neutral reading. Inserting ambi in Fering (21b) would imply a negative

Progressive markers in Germanic languages

617

evaluation of the activity, whereas the Swedish sentence with runt does not necessarily have this reading. (21c) on the other hand, given that the sentence implies a negative judgement, would be odd without 'around' in Fering and Danish, though not in Dutch. It is totally impossible without herum in German, which does not have a POS-construction. (21) a.

b.

с

FerÖöm Grm

Hi sat ambi tu süpen. Er sitzt rum und säuft. he sits around to booze / and boozes 'He sits around boozing.' Swd Hon gär {runt) och samlar pengar for Jugoslavien. she goes around and collects money for J. FerÖöm Hat leept (*ambi) tu saamlin. she walks around to collect:INF 'She is collecting (for Yugoslavia).' (PROGQ: 2) FerÖöm Hi leept ambi tu pochin. Dutch Hij loopt te pochen. Dan Han gar rundt og praler. Grm Er läuft herum und gibt an. he walks around to brag / and brags 'He goes around bragging (all the time).'

There is thus no clear evidence for distinguishing a motion progressive from postural verb constructions, nor for distinguishing a motion progressive from a 'locomotive' ' of the type 'go around'. German is the only language that has no POS, and it does not allow läuft und V. It does have the attitudinal construction läuft herum und V, but also sitzt/steht herum und V. In the other languages there is no clear distinction between constructions with 'go' and 'go around'. The motion progressive can therefore be included in the POS-constructions, 'going (around)' being a type of constant body position.17 A POS-construction with a verb meaning 'go, walk' is common in Danish, Dutch and Frysk, but very restricted in North Frisian. For Icelandic and Swedish more detailed information is necessary.

3. Prepositional vs. postural verb constructions 3.1.

Semantic and pragmatic conditions

What factors determine the choice between PREP/HOLD and POS? Although the "ROGQ was not specially designed to test this opposition, some general rules and tendencies became apparent. In Table 2 I have arranged the forms used in the

618

Karen H. Ebert

Table 2. POS vs PREP with types of vertag Danish Swedish +agentive +telic 0,h* PROGQ: 8 0,h 10 h 14 h 17 -telic 0,h 1 0,POS 7 0,POS 9 0,POS 12 0 44 0 45 -agentive +telic 38? -telic 0,h 35 0,POS 37 77 0 non-intentional 0 33

0,h 0 0 0 0,h 0 0 0,h 0,h

0,h 0 0,h 0

0,h h 0,h 0,h 0,h 0,POS 0 0,h 0,h POS

0,POS,h 0.POS 0 0,h 0 0

0 0 ved ved

(POS),ved POS,ved POS.ved POS.ved

0 0 0 0

POS POS,ved POS POS.ved O.POS 0.POS

«POS

0,h POS

0.POS 0

igang ved igang ved 0,i gang ved POS 0, i gang 0 0

ö,POS

POS

O.POS

POS 0 0 0 0

0,h 0,POS

34

Frysk

Dutch

North Frisian +agentive +telic bai 0 10 bai 14 0 17 -telic bai 1 POS 7 POS 9 POS 12 bai 44 bai 45 -agentive -telic POS 35 0 37 0 38 77 -intentional POS 33 POS 34

uun 0 0,bi

bi 0 0 bi POS,uun POS uun pOS.uun uun uun

uun POS uun uun uun uun

POS 0,POS 0,POS 0

0 0 0

0 0

uun 0 0 0 uun POS uun POS uun, bi POS,uun

POS uun POS uun 0 0

bi 0 0.POS

oan oan,POS oan,POS oan

POS,uun 0 POS,uun bi,POS 0,uun POS.uun

POS oan,POS oan oan,POS oan oan

0 0 0.POS

0.POS 0.POS 0.POS 0 0 0

bezig 0 POS bezig

0 0 0 0 aan 0

POS POS bezig POS

aan 0

POS POS 0 0 POS POS

POS 0 POS 0 0 0

Progressive markers in Germanic languages -(-agentive, +telic PR0GQ:8 10? 14 17

build a shed give a present peel 3 kilos of potatoes chase 2 chickens out of the house

—agentive, +telic 38?

rot (Scandin.)

-intentional 34

admit one's guilt

619

—telic PROGQ: 1 7 9 12 44 45

work read the newspaper sing (a song) peel potatoes dance play cards

38 37 35 77

rot (Frs) boil dream rain

33

insult the neighbor

^

* h stands for hälla pà; the other small letters abbreviate various types or PREP forms, which should speak for themselves. Cf. Table 1.

PROGQ according to verb types. This arrangement reveals certain patterns that are lost in the general list in Table 4 (Appendix). Agentive verbs combine in principle with both PREP/HOLD and POS. Since the postural verbs still carry some of their original meaning, they are in general restricted to events that take place with the protagonist in a constant position. POS was used in PROGQ: 7 (=2) and PROGQ: 70 for 'read' in all the Frisian and in most Scandinavian questionnaires, in PROGQ: 12 for 'peel potatoes' in most Frisian and in some Scandinavian questionnaires, i.e., with activities that are typically carried out while sitting (or standing in the case of Scandinavians peeling potatoes). But POS is always possible if the activity is carried out, or imagined as being carried out, with a certain position of the body. Thus informants have people 'sit' or 'stand' cleaning a gun (PROGQ: 6) or singing a song (PROGQ: 9), 'sit', 'stand' or 'go around' admitting their guilt (PROGQ: 34), 'sit' playing cards (PROGQ: 45) and 'go around' dancing (PROGQ: 44). The choice between PREP, POS and simple form seems to be partly dependent °n personal preferences. The second Danish consultant translated practically all agentive verbs with both POS and PREP; the other two Danes used no or one POS m the examples of Table 2. The POS-constractions of the two Dutch informants are Practically in complementary distribution: one has nearly all his POS forms with agentive verbs, the other has POS only with non-agentive verbs.18 What is most striking in Table 2 is the rarity of PREP forms with non-agentive v erbs. The Danish, Frysk and Dutch informants did not use a single PREP here, a nd only two PREP forms occurred in one of the five North Frisian questionnaires. Whether a postural verb was used or not with non-agentive verbs depends again on "te general condition mentioned above, namely the constant position of the protago-

620

Karen H. Ebert

nist or object. There was a high number of POS in PROGQ: 35 (=22). One usually dreams while lying down, and although it is also possible to sit, stand or go around dreaming only the verb 'lie' was used in the questionnaires. (22)

(PROGQ: 35) Dan Han ligger og dr0mmer om sin pige. Wiid Hi läit an driimt fuon sin friindin. he lies and dreams about/of his girl(friend) FerÖöm Hi läit faan sin foomen tu dremen. Frysk Ну leit fan syn faam te dreamen. Dutch Hij ligt van zijn vriendin te dromen. he lies of his girl(friend) to dream 'He is dreaming of his girlfriend.'

The verb 'dream' is odd with a PREP/HOLD form. Two consultants wrote that this combination would imply that the person is not sleeping, or that he is absorbed in his dreaming; i.e., a more active interpretation results. Table 2 suggests that agentivity is the crucial factor and non-agentive verbs rule out PREP. That this is not generally true was clear already from Fering and German (14a, b). No clear picture emerged for the role of telicity in the choice of the progressive form. PREP is sometimes preferred with telic verbs (cf. (23)), but this did not be­ come clear from the questionnaires. Osten Dahl (PROGQ comments) thinks that "hâlla pâ is more often used with telic verbs, and in general with goal-directed activities. With the postural verbs, it is rather the other way round". This is only partly confirmed in Table 2. There is not a single POS with telic verbs in Swedish, but there are many HOLD forms with atelic verbs. Dahl himself does not make a clear difference between atelic PROGQ: 12 (peel potatoes) and telic PROGQ: 13-15 (peel the / all the / 3 kilos potatoes). Though he uses only POS in the former and only HOLD in PROG 14-15, he gives both a POS and a HOLD form for telic PROGQ: 13 (peel the potatoes). As Table 4 in the Appendix shows, most informants marked all four sentences in the same way. The exceptions are of course North Frisian and German, where incorporation is possible only with the indefinite object. Atelic PROGQ: 16 and telic PROGQ: 17 (chase chickens / two chickens out of the house) were marked identically by all Scandinavian informants, and there are no relevant differences in the other questionnaires. A candidate for a telic non-agentive verb in the questionnaire is 'rot' (PROGQ: 38), but the actionality of this verb in the individual languages is far from clear. Intransitive telic verbs often have ambiguous actionality. For instance apples can (in English) rot in a week or for a week, towels can dry in two hours от for two hoursBut the actionality of the translational equivalents of these verbs need not be the

Progreseive markers in Germanic languages

621

same. Whereas Fering driigi has the same actionality as the English dry there are two different verbs for 'rot'. The atelic röödi combines preferably with POS, the telic ferröödi with PREP (23). With the verb driigi 'dry' both forms are possible (24), but I could not say that uun 't driigin implies the idea that eventually the towels will become dry and hingi tu driigin does not. The Swedish verb ruttna seems to be telic. Dahl finds hâlla pâ "natural since the process is not yet complete". HOLD was used in two of the Swedish questionnaires. Haberland (p.c) excludes a progressive interpretation for er ved at гфапе (23), but not for er ved at t0rre (24). (23)

(PROGQ: 38) FerÖöm A aapler hingi iiiib a buum tu röödin /* tu ferröödin. the apples hang on the tree to rotting A aapler san uun't ferröödin / ?uun't röödin. the apples are IN_the rotting Dan ablerne hœnger og râdner pâ traeet. / ??er ved at r0dne. apples:DEF hang and rot on tree:DEF Swd Äpplena hâller pâ och ruttnar pâ träden. apples:DEF HOLD on and rot on tree:DEF 'The apples are rotting (on the tree).'

(24)

Dan FerÖöm Grm

T0rreklaederne er ved at t0rre. A hoonskiitjen san uun't driigin /hingi noch tu driigin. Die Handtücher sind am Trocknen. the towels are AT_the drying hang still to drying 'The towels are drying.'

Whether there is a different actionality involved or whether the difference is just idiosyncratic needs to be clarified. Generally more research into the actionality of the verbs in the individual languages is necessary. With 'boil' (PROGQ: 37) Swedish HOLD and Danish PREP have only an imminential reading. This suggests that the verb is ingressive-phasal in those languages, meaning 'come to the boil, boil'.19 With HOLD/PREP the ingressive component is selected, with POS the phasal component. In Frisian and Dutch the verb is atelic. (25

)

(PROGQ: 37) Swd Vattnet star och kokar. ф Vattnet holler pâ att koka. Dan Vandet star og koger. ф Vandet er ved at koge. watenDEF stands and boils

622

Karen H. Eben FerÖöm Frysk Dutch

At weeder stäänt tu köögin. = ?At weeder as uun 't köögin. It wetter stiel te sieden. *It wetter is oan ' t sieden Het water Staat te koken. = ?Het water is aan het koken. the water stands to boil 'The water is boiling.'

It is not clear whether the impossibility or oddity of a PREP-construction with 'boil' in Frisian and Dutch is due to non-agentivity or to non-telicity. In Fering telicity seems to matter, as shown in (23) and in (14). According to the Frysk expert oan't is not possible with non-agentive verbs, but sentences like de see wie oan 't bearen 'the sea was roaring' are clearly acceptable. This could be a quasi-agentive use, but there could also be a further factor involved, namely dynamicity.20 Highly dynamic verbs favor PREP, whereas verbs of low dynamicity hardly occur with PREP even if they are agentive; e.g., in Germanic languages people generally 'sit/stand to wait'. Non-intentional events are treated like non-agentive ones. Table 2 shows no PREP forms and only one Swedish HOLD. The POS-constructions chosen esp. by the Wiidinghiird and one Dutch informant seem to imply a negative evaluation, cf.

(26)

(PROGQ: 34) Han star och erkänner skulden oavsiktligt. Swd he stands and admits guilt:DEF unintentionally Soner dat 'r 't wiitj lapt 'r je an sait dat 'r without that he it knows goes he PART and says that he Wiid skili äs. guilty is Hij zit onbewust zijn schuld toe te geven. he sits unconsciously his guilt PREP to give Dutch (toegeven - admit) 'He is (standing/sitting/going around) unintentionally admitting his guilt.'

Two contributors remark that the postural verb in (25) implies duration. Haberland writes (PROGQ): "Vandet star og koger (og har gjort det et stykke tid)" [the water stands and boils (and has done this for a while)]. I think POS always implies some noticeable duration (cf. Ebert 1989), and this explains why it is incompatible with momentaneous verbs, where we only find PREP (cf. (15a)). That the remark is linked to this example is probably due to the fact that one would not normally let the water boil for a long time. In oral interviews Fering speakers sometimes inserted a temporal adverb indicating duration in postural verb sentences:

Progressive m a i n in Germanic languages (27)

FerÖöm -Saaist du: At say you the -Je, at weeder yes the water

weeder water stäänt stands

stäänt stands al already

623

tu köögin? to boil tjiin minüten tu köögin. ten minutes to boil

Another somewhat vague factor has to do with goal-directedness or the importance attributed to the activity (cf. citation from Dahl above). This criterion has been elaborated somewhat in Ebert (1989) for Fering, where uun't is used if an activity belongs to the work that has to be done in a farming society and POS is preferred with leisure activities. Thus one can only say hat as uun't moolkin 'she is milking the cows', although I have never seen anyone milk a Frisian cow other than in a sitting position. The verb is atelic, but the activity is goal-oriented. Mending socks is also goal-oriented, but it is most normal to have a postural verb here: Hat sat tu höözen-stoopin. Mending socks, knitting sweaters, and writing letters counted as a leisure activity in traditional Frisian society, done when the serious work was finished. All activities in the living quarters of the house allow POS; cf. also stun tu baagen (1 lb), stun tu köögin, stun tu eerdbäären-iinmaagin (stand to bake, to cook, to strawberries-preserving), sat tu eerten-ütjpülin, sat tu höözenpreglin (sit to peashelling, to stockings-knitting). The choice between PREP and POS here has to do with the degree of importance or urgency attributed to the activity. This factor seems to be relevant only in Fering. De Groot (p.c.) confirms that the factors a)-e) of Table 3 are also crucial for the choice between PREP and POS in Dutch, where one can, however, say hij zit te melken (he sits to milk). Apart from some variation between languages or idiolects, the use of POS is remarkably stable across languages and apparently also across time.21 Compare the following combinations with POS from Dutch and Swedish22 prose with examples from English sources.

Table 3. Factors determining the choice between PREP and POS

a) b) c) d) e)

PREP

POS

rbconstant position +agentive +telic high dynamicity shorter duration goal-directed, important

+constant position ±agentive —telic low dynamicity longer duration not goal-directed, less important

624

Karen H. Eben

(28)

Dutch ZATTE SAT TO

STOND TE STOOD TO

LAG TE LAY TO LIEP TE WENTTO Swd SATT OCH SAT AND STOD OCH STOOD AND MiEngl SAT AND STOD AND LAY AND ModEngl SAT AND

lezen, schrijven, kijken, breien, bedenken, read, write, look, knit, think, schommelen, appels schulen rock to and fro, peel apples wachten, dromen, kijken, bekijken, verteilen, wait, dream, look, look at, tell, pulken, stampen picking (one's nose), stamp (one's foot) wachten, dromen, slapen, rommelen wait, dream, sleep, mess around/fuck leuren peddle/hawk laste, pratade, vaktade, stirrade read, talked, waited, stared vred pâ..., höll ögat intill..., vacklade rubbed, held the eyes against, shook lokede, thoughte, playd, behelde, song, dinede, loghe,... lokede, thoughte, beheld, logh, wondrede, prechede,... lokede, plaid, thoughte,...

talked, gazed, looked, smiled, listened, thought, mused, worked,... STOOD AND talked, gazed, looked, smiled, watched, stared,... LAY AND gazed, watched,...

(and similar for Old English; see van der Gaaf 1934, Visser 1973: 1401-1408,19021916). Earlier periods of English also had the constructions sit V-ande, sit V-ing and - with decreasing frequency - sit to V. Note that all the text examples have human subjects and none has a telic verb.

3.2. Temporal conditions The POS-construction occurs in some contexts that are usually incompatible with progressives, e.g. with adverbs that indicate a temporal limit. In PROGQ: 48 (=29) no PREP, but various POS forms were used.

Propessive ^ * м in Germanic languages ( 29)

625

(PROGQ: 48) Dan Anne sad to timer og legede helt alene. A. sat 2 hours and played all alone FerÖöm Anne seed tau stünj alian tu spelin. Dutch Anne zat twee uur alleen te speien. A. sat 2 hours alone to play 'Anne played (?was playing) for two hours by herself.'

L

Although in some languages limiting adverbs can be combined with progressives (cf. Bertinetto & Delfitto, this volume), they are excluded with PREP-constructions in most Germanic languages. Dutch and the Rhineland dialect are an exception. (30) a.

b.

Dan FerÖöm Grm

Anne Anne Anna A. Dutch (?)Anne GrmRhi Anne A.

var (*2 timer/*fra 2 til 4) vedat lege. wiar (*2 stiinj/*faan 2 tu 4) uun 't spelin. war (*2 Stunden /*von 2 bis 4) am Spielen. was (2 hours/from 2 to 4) AT_the play:INF/VN was twee uur alleen aan het spelen. < war zwei Stunden alleine am Spielen. was 2 hours alone AT_the playing

The PREP-constructions usually imply a focussing perspective: at a certain moment such-and-such event is in progress. This explains why they do not combine with adverbs that indicate a temporal limit. No such perspective is involved with POS. Here the event can be located in relation to a time interval with a specified duration or temporal endpoint. The temporal structure for the contexts in question can be symbolized as follows: (A)

(B)

El E2

• R

El E2 T

•

•

i

T

J

(A) symbolizes the focussing perspective, which is usually taken as a test case for progressives (cf. examples in section 2.1): El holds at reference point R, which can be a second event (E2) (B) symbolizes the durative constellation, i.e., El holds over a time interval: El while E2 / from T ; to Tj23

626

Karen H. Ebert

In situation A all Germanie languages use a progressive construction, PREP and POS often being equally adequate; in situation В most languages allow only POS. In the focussing perspective the termination of El may roughly coincide with an E2 serving as R. Sentences like (3) 'When John came, Ann was still working' with a PREP-form in the subordinate clause can in all languages continue: 'but she got up immediately to mix a drink'. PREP is also compatible with adverbials which delimit a stretch of time including R, like 'since' (31a). The crucial factor is that the situation holds at R and leaves the endpoint unspecified. PREP is not possible if there is no temporal overlap, e.g., with adverbials like 'after 5 / after John left', 'before 5 / before John left'. (31) a.

FerÖöm Grm

b.

FerÖöm

Anne as sant 2 stiinj / santklook4 uun't spelin. Anna ist seit 2 Stunden / seit 4 Uhr am Spielen. A. is since 2 hours/ since 4 o'clock AiyiN_the playing 'Ann has been playing for (since) 2 hours / since 4 o'clock.' [Efter a klook 7 / biföör John kaam,] haa wi seeden tu koordin. / *wiar wi uun't have we sat to card_playing were we IN_the koordin card_playing '[After 7 o'clock / before John came] we were playing cards.'

For the POS-constraction no limitation and no temporal overlap are necessary. Ad­ verbials with somewhat vague limits, like 'during' and 'while' clauses, are bad with PREP (though apparently not with Swedish HOLD). Most informants used a POS in PROGQ: 49 and 50. Only one Swedish informant made a difference and used pro­ gressive markers in the unlimited context of PROGQ: 49 (=32a), but not in PROGQ: 50, which suggested that the talking stopped when the class was finished. Most Scandinavian and Frisian informants used POS in the 'while' clause of PROGQ: 70 (=32b), but not in the main clause. The clauses can easily be reversed, as in (32c), with the POS remaining with the verb 'sit'. If temporal limits are pragmatically not relevant, the use of POS is conditioned mainly by the type of activity. If temporal limits do matter, both POS and PREP are excluded with co-extensive events, as is shown by the oddity a POS or PREP would create with (32d). (32) a.

Swedish (PROGQ: 49) ... hall Ann pâ och pratade / satt Ann och pratade med held A. on and talked sat A. and talked with sin granne. her neighbor '[During the whole class] Ann was talking to her neighbor.' [and she carried on even afterwards].

Pnagmrave makers in Germanic languages b.

с

d.

627'

Danish (PROGQ: 70) Mens Ann sad og laeste pâ sit vaerelse, legede Martin i while A. sat and read on her room played M. in garden. garden:DEF 'While Anne was reading in her room, Martin was playing in the garden.' Danish Mens Martin legede i garden, sad Ann og laeste pâ sit vaerelse. 'While Martin was playing in the garden, Ann was reading in her room.' Dutch (Boogaart 1991: 6) [Oom hurkte, nam de arm van de man en telde diens polsslagen,] terwijl hij op zijn horloge keek/*zat te kijken/*aan het kijken was. while he on his watch looked '[Uncle squatted down, took the arm of the man and counted his pulse] while he looked /*was looking at his watch.'

Due to the lack of a focussing perspective the POS-constructions combine easily with habitual contexts ((33a) and analogous for the other languages). But then PREP seems to be compatible with a habitual interpretation, too (33b). (33) a.

b.

FerÖöm At lidj sat an injem bluat noch tu fernseen. the people sit in_the evenings only PART to TV_watching Iar seed 's tu koordin of tu leesen. earlier sat_they to card_playing or to reading 'Nowadays people only watch TV in the evenings. Earlier they used to play cards or read.' FerÖöm Hat as imer uun't werkin. Dut Zij is altidj aan het werken. Grm Sie ist immer am Arbeiten, she is always AT_the working 'She is always working.'

To the question 'What does Ann do on Saturdays' (PROGQ: 2) one can answer with (c), but hardly with (d).24 Ш) c.

d.

FerÖöm Grm

An saninjem as hat imer uun't renskin. Samstags ist sie immer am Putzen. on-Saturdays is she always IN/AT_the cleaning 'On Saturdays she is always cleaning.' FerÖöm *An saninjem as hat uun't renskin. Grm *Samstags ist sie am Putzen.

628

Karen H. Ebert

The adverb 'always' is crucial in those sentences, as it invokes plurifocalizatkm: El holds at every R.

3.3. Grammaticization and stylistic variation The Germanic progressives correspond to typologically frequent patterns; locative expressions and postural verbs are among the most frequent sources for progressive markers (Bybee & Dahl 1989, Bybee et al. 1994: 127ff., Heine et al. 1991: 117f.). Grammaticization has proceeded to different degrees. The PREP-constructions are all desemanticized. Syntactically they do not behave like 'be' + locative complement any longer. Only the German and North Frisian progressives do not allow a direct object, the PREP-construction still being analogous to a locational phrase; cf. Grm. sie ist am Schreibtisch' 'she is at the desk', sie ist am Schreiben 'she is writing'. But the Rhineland forms are spreading into other areas. Most German speakers accept a sentence like Das sind wir noch am Diskutieren 'We are still discussing that' with a pronominalized and preposed object, even if they otherwise reject objects with am. The POS-constructions of North Frisian underlie the same syntactic restrictions as the PREP forms. Syntactically the postural verbs behave like auxiliaries. In FeringÖömrang the perfect auxiliary with motion verbs is 'be', but in a POS-constraction the perfect takes the auxiliary 'have'. In Dutch auxiliaries appear in a past participle position without the prefix ge-, and te is or can be omitted after a non-finite POS (cf. Dutch (12a)). The meaning of the postural auxiliaries is bleached, as mentioned by several authors (Geerts et al. (1984: 538), Hansen (1967: 30)). Desemantisation has proceeded furthest in Danish (cf. (18b), (20a)). Generally the postural verbs can not replace each other, but in emotive use they are desemanticized in all languages (cf. also (19), (26)). (34)

Dutch Loop I lig I zit niet te zeuren. walk / lie / sit not to whine 'Stop whining.'

Postural verb periphrases indicating durative actionality are found in many languagesThey may develop into progressives, but then it is usually only one verb that is geß" eralized as a grammatical marker.27 Often the actional and the aspectual functions exist side by side, and this seems to be the case in the Germanic languages. There are some loose semantic restrictions on both progressive constructions, as described in part 3.1. These restrictions are also indicators of incomplete grammaticization. As a corollary of grammaticization a construction becomes obligatory, often ruling out alternative markers. This happened with English at V-ing, which replace"

Progressive markers in Germanic languages

620

the postural verb constructions that were common in earlier periods. In most Germanic languages marking of progressive aspect is not obligatory. For Frysk, Hoekstra (PROGQ) writes that "Progressive marking is nearly obligatory in West Frisian. The use of a simple verb instead of a progressive construction is perhaps not fully ungrammatical, but marked to say the least [...] There is no relevant difference with respect to style and register. Both constructions are very frequent both in the spoken language and the written (standard) language." In traditional Fering progressive