No title

ISSN 0167-5133 Volume 5, Number 4, 1986/87 JOURNAL OF SEMANTICS AN INTERNATIONAL JOURNAL FOR THE INTERDISCIPLINARY ST...

Author: Oxford University Press

11 downloads 393 Views 5MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

ISSN 0167-5133

Volume 5, Number 4, 1986/87

JOURNAL OF SEMANTICS AN INTERNATIONAL JOURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE MANAGING EDITOR:

Peter Bosch

EDITORIAL BOARD:

Simon C. Garrod Leo G.M. Noordman Pieter A.M. Seuren

REVIEW EDITOR:

Rob A. van der Sandt

ASSISTANT EDITOR:

Bart Geurts

CONSULTING EDITORS: W. Levelt (Max Planck Inst. Nijmegen), J. Allwood (Univ. GOteborg), J. Lyons (Trinity Hall, Cambridge), R. Bartsch (Amsterdam Univ.), W. Marslen-Wilson (Max Planck Inst. J. van Benthem (Amsterdam Univ.), Nijmegen), D.S. Bree (Erasmus Univ. J. McCawley (Univ. Chicago), Rotterdam), 6 . Dahl (Stockholm Univ.), H.E. Brekle (Univ. Regensburg), B. Richards (Edinburgh Univ.), G. Brown (Univ. of Essex), R. Rommetveit (Oslo Univ.), H.H. Clark (Stanford Univ.), H. Schnelle (Ruhr Univ. Bochum), H.-J. Eikmeyer (Univ. Bielefeld), J. Searle (Univ. Cal. Berkeley), G. Fauconnier (Univ. de Vincennes), A. von Stechow (Univ. Konstanz), J. Hintikka (Florida State Univ.), M. Steedman (Edinburgh Univ.), J. Hobbs (SRI, Menlo Park), Ch. Travis (McGill Univ.), St. hard (Sussex Univ.), Z. Vendler (UCSD), D. Israel (SRI, Stanford), P.N. Johnson-Laird (MRC Appl. Psych. Y. Wilks (New Mexico State Univ.), D. Wilson (UCL). Unit, Cambridge), E. Keenan (UCLA), EDITORIAL ADDRESS: Journal of Semantics, P.O. Box 1454, 6501 BL Nijmegen, The Netherlands

Published by Foris Publications, P.O. Box 509,3300 AM Dordrecht, The Netherlands. * N.I.S. Foundation ISSN 0167-5133 Printed in the Netherlands by ICG Printing

I

I Natural Language & Linguistic Theory Editors Joan Maling, Dept. of Psychology, Brandeis College, Waltham, USA Associate Editors Judith Aissen, Cowell College, University of California, Santa Cruz, CA, USA; Michael Kenstowicz, Dept. of Linguistics, University of Illinois, Urbana, IL, USA; Frederick J. Newmeyer, Dept. of Linguistics, University of Washington, Seattle, WA, USA and David Pesetsky, Dept. of Linguistics, University of Massachusetts, Amherst, MA, USA Natural Language & Linguistic Theory provides a forum for discussion of theoretical research that pays close attention to natural language data, so as to provide a channel of communication between researchers of diverse points of view. The journal actively seeks to bridge the gap between descriptive work and work of a highly theoretical, less empirically oriented nature. In attempting to strike this balance, a primary goal of the journal is to encourage work which makes complex language data accessible to those unfamiliar with the language area being studied, and to publish papers presenting theoretical discussions in such a way that they are accessible to those working outside the framework in question. 1987 contributors include: Leonard Babby, Heles Contreras, Lyn Frazier, James Harris, Larry Hyman, Christer Platzack, Ken Safir, Mark Steedman, Edwin Williams, Moira Yip. A feature of the journal is the "Topic . . . Comment" column that appears in every issue Subscription Information ISSN 0167-806X 1988, Volume 6 (4 issues) Institutional rate: Dfl. 225.00/US $106.00 incl. postage/handling Private rate: Dfl. 90.00/US $32.50 incl. postage/handling Special rate for students: Dfl. 82.00/US$28.50 incl. postage/handling Private subscriptions should be sent direct to the publishers

Back Volume(s) Available Volumes 1-5 (1983-1987)

Price per Volume excl. postage Dfl. 194.00/US$81.00

Kluwer Academic Publishers P.O. Box 989. 3300 AZ Dordrechl. The Netherlands 101 Philip Drive. Norwell. MA 02061. U S A

SCOPE OF THIS JOURNAL The JOURNAL OF SEMANTICS publishes articles, notes, discussions, and book reviews in the area of natural language semantics. It is explicitly interdisciplinary, in that it aims at an integration of philosophical, psychological, and linguistic semantics as well as semantic work done in artificial intelligence and anthropology. Contributions must be of good quality (to be judged by at least two referees) and should relate to questions of comprehension and interpretation of sentences or texts in natural language. The editors welcome not only papers that cross traditional discipline boundaries, but also more specialized contributions, provided they are accessible to and interesting for a wider readership. Empirical relevance and formal correctness are paramount among the criteria of acceptance for publication. INFORMATION FOR AUTHORS Typescripts for publication should be sent in 3 copies to the managing editor. They should be typed on A4 (or similar format), one-sided, double spaced, and with a wide margin and must be accompanied by an approx. 200 words summary. Footnotes and bibliographical references must appear at the end of the typescript. Diagrams must be submitted camera-ready. All papers submitted are subject to anonymous refereeing. Authors receive 20 offprints of their published articles and 10 offprints of their published reviews, free of charge. Larger numbers can be supplied at cost price by advance arrangement. Unless special arrangements have been made, copyright rests with the NIS Foundation. PRICES AND CONDITIONS OF SUBSCRIPTION The JOURNAL OF SEMANTICS appears in four issues per year of approx. 100 pages each. Subscriptions for private use are available at the reduced rate of Dfl. 64,— per year; the institution rate is Dfl. 160,—. The price for single issues is Dfl. 38,—. Postage Dfl. 13,— per volume. Airmail- and SAL-rates are available on request. We regret that no delivery can take place before payment has been received. Subscriptions not cancelled before October 1st automatically extend to the following year. Placement of orders implies the consent of the subscriber to these conditions. All orders should be sent to Foris Publications, P.O. Box 509, 3300 AM Dordrecht, The Netherlands.

JOURNAL OF SEMANTICS Volume 5, Number 4

CONTENTS

EDITORIAL NOTE

i

DIRK GEERAERTS:

RENATE

On necessary and sufficient Conditions . . .

BARTSCH:

The Construction

of Properties under

Perspectives

293

WOLFGANG WILDGEN:

D.S. BREE and R.A. SMIT:

B1RGIT WESCHE:

275

Processual Semantics of the Verb . . .

Temporal Relations

At Ease with "At"

321

345

385

EDITORIAL NOTE

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The papers in this issue were originally presented at a symposium on "Word Meaning and Representation" at the 1987 Annual Conference of the Deutsche Gesellschaft fur Sprachwissenschaft, held on March 4th-6th at the University of Augsburg.

Journal of Semantics 5: 275-291

ON NECESSARY AND SUFFICIENT CONDITIONS DIRK GEERAERTS

It is often said that prototypically structured concepts are a counterexample with regard to the classical view of categorization, viz. that categories always have a single definition in terms of necessary and sufficient conditions. However, such statements about the nature ofprototypicality do not take into account the distinction between vagueness and ambiguity. In particular, they are never accompanied by a specification of the criteria according to which the allegedly prototypical concepts are indeed a single semantic category. An exploration of the distinction between analytic and introspective critena for ambiguity shows that prototypicality is part of a larger class of phenomena exhibiting discrepancies between both approaches. Specifically, prototypicality and autohyponymy together make clear that analytic distinctness of meaning is neither a necessary nor a sufficient condition for introspectively recognized ambiguity, such as it determines the truth conditions of utterances.

1. THE COGNITIVE AND THE CLASSICAL VIEW OF CATEGORIZATION

The prototypical conception of natural language categories1 is usually presented in contrast with the classical view of categorization, which implies that conceptual categories can be defined by means of a set of criterial (or 'defining') attributes: "Many traditions of thought in philosophy, psychology, linguistics, and anthropology imply that categories are Aristotelian in nature - that is, that categories are logical, clearly bounded entities, whose membership is defined by an item's possession of a simple set of criterial features" (Rosch 1975:193). The criteriality of these features is indicated by the fact that they are "features common to all members of the category which distinguish those members from all others" (Rosch and Mervis 1975:580). The characteristics that together make up the criteriality of features are traditionally identified as necessity and sufficiency.2 On the one hand, a definition of a category specifies a sufficient set of conditions for any one thing to belong to that category, if the joint presence of the listed conditions guarantees that the item in question3 is a member of the extension of the category. On the other hand, each of the conditions mentioned in the definition is necessary to the extent that its absence guarantees that any one thing is not a member of the extension of the category. In this sense, the prototypical view of categorization seems to imply that some categories may not be defined by means of a set of necessary and


ABSTRACT

276

2 THE STRUCTURE OF THE PROTOTYPICAL ARGUMENT

Not every concept is a lexical meaning. This is trivially so because some concepts (such as "glancing at the last page of a novel to see how the story ends") simply are not lexicalized. Less trivially, if we call any concept that can be expressed by a single lexical item in a particular linguistic and extralinguistic context a conceptual specification of that item, it is also the case that not every concept is a lexical meaning because not every conceptual specification is a lexical meaning. Traditionally at least, conceptual specifications can be illustrations of either vagueness or ambiguity, and only in the latter case will they be considered lexical meanings. Consider (1), which can be paraphrased as either (2) or (3).5 (1)

They passed the port at midnight.

(2)

They passed the harbour at midnight.

(3)

They passed the port wine at midnight.

The conceptual difference exhibited by port, which can refer to a type of wine as well as to a harbour, is known as ambiguity: the lexical form port is said to represent two lexical meanings, and more precisely, since the example involves homonymy, it is said that there are two lexical items port I and port2, each with a different meaning.6 On the other hand, port in (3) may refer to various types of wine; it can be white, tawny or ruby, and it can be vintage or blended. These conceptual differences are not considered to represent different lexical meanings, but they are taken to illustrate the vagueness (or 'generality', or 'lack of specification') of port! "sweet fortified Portuguese


sufficient conditions. 4 Taking into account that sufficiency is a requirement with regard to the set of conditions as a whole, whereas necessity is required of each of the conditions separately, it can be stated that the prototypical view implies that some categories may not be defined by means of a sufficient set of necessary conditions for category membership. In this paper, I will challenge this view, though not with the intention to reinstate the classical definition of categorization. I will try to make clear that the definition of prototypicality just mentioned is imprecise because it does not take into account the distinction between vagueness and ambiguity. Moreover, a revised definition will be shown to entail that prototypicality is merely one example of a larger class of phenomena in which the traditional definitional approach to lexical-semantic structure is at odds with the intuitive facts of linguistic usage.

277


wine". In this sense, conceptual differences need not be reflected by semantic differences; that is to say, distinct conceptual specifications may represent the same lexical meaning.7 From now on, the notion 'conceptual specification' shall be restricted to examples of vagueness. Thus, "white sweet fortified Portuguese wine" and "ruby sweet fortified Portuguese wine" are conceptual specifications of port, but only "sweet fortified Portuguese wine" is a lexical meaning. Ambiguity is then defined as the existence of two or more lexical meanings for a particular lexical form, vagueness as the existence of two or more conceptual specifications. This is, of course, a purely terminological and not an operational definition: it does not tell you how to distinguish vagueness from ambiguity. That is a point to be taken up further on; first, the relationship between the conditions on definitions mentioned in the first paragraph and the distinction between vagueness and ambiguity has to be determined. Notice that the requirement that definitions take the form of sufficient sets of necessary attributes applies to conceptual specifications as well as to lexical meanings. It is a general condition on definitions, not a distinctive condition that applies only to one of the two types of concepts. For instance, necessary and sufficient conditions can in principle be given for something being a white port wine just as well as for something being a port wine tout court. As such, a failure to define a lexical item by means of a single set of necessary and sufficient attributes cannot as such, i.e. without further stipulations, be taken to mean that the item in question has more than one lexical meaning. The requirement of necessity-cum-sufficiency does not tell you what you are defining, meanings or conceptual specifications, and therefore, having to define a lexical item by means of more than one set of definitional conditions does not tell you whether that item is ambiguous or merely vague. To illustrate this, take the item bird as a prototypical example of prototypicality. On the one hand, there are not many attributes that are common to all birds and that could therefore be listed as necessary: ostriches and penguins cannot fly, kiwis do not have wings, penguins and kiwis do not have clearly distinguishable feathers, and so on. On the other hand, those characteristics that are universal among birds are not sufficient to distinguish birds from other species (in particular, reptiles lay eggs, and the duck-billed platypus has a bill). Assuming that this analysis, schematically represented in Figure 1, is correct,1 we can state in neutral terms that the extension of birdthat is at stake here9 cannot be covered by a single set of criterial features. The proponents of prototype theory conclude from this and similar examples, that some lexical categories may not be definable by means of necessary and sufficient attributes, but of course, this conclusion is only warranted if it can be shown on independent grounds that bird in the "animal species" sense is a single lexical meaning. Conversely, proponents of the classical conception might maintain that lexical meanings are always definable by means of necessary and suffi-

278

6

5-

4

1 kiwi

robin (and others)

chicken, ostrich

Figure 1. The extension of various attributes that are commonly used in the definition of English "bird". The numbered boxes represent the extension classes of the corresponding attributes: 1 being able to fly; 2 having feathers; 3 being typically £ -shaped; 4 having wings; 5 being born from eggs; 6 having a beak or bilL

cient characteristics, but then, of course, they will have to admit that bird is ambiguous, i.e. that it represents several meanings. Logically speaking, there are two mutually incompatible assumptions involved in the discussion, viz. (a) a single lexical meaning can always be defined by means of a single sufficient set of necessary conditions, and (b) the concepts analyzed by prototype theorists in favour of their approach represent a single lexical meaning. The cognitive view rejects (a) on the basis of (b), but the classical view might safeguard (a) by denying (b). The argument can be settled by determining whether concepts such as bird in the "animal" sense are ambiguous or merely vague. Intuitively, prototype theory seems to be correct in treating the "animal" sense of bird as a single meaning, but are there any arguments or tests to support this intuition? Since the failure to provide a necessary and sufficient definition for bird is equivocal in this respect, independent evidence for vagueness or ambiguity has to be sought.10

3. DELIMITATIONS OF AMBIGUITY

There are basically two approaches to the question 'What is a distinct


penguin

279

I mean, for example, supposing that we require a definition of high-mindedness (megalopsychid), we must consider individual high-minded persons whom we know, and see what one characteristic they all have qua high-minded. E.g., if Alcibiades and Achilles and Ajax were high-minded, what was their common characteristic? Intolerance of dishonour; for this made the first go to war, roused the wrath of the second, and drove the third to commit suicide. Then we must apply the same process to anulhcr group, e.g., Lysander and Socrates. Suppose that here the common characteristic is being unaffected by good and bad fortune. Now I take these two and consider what there is in common between indifference to fortune and intolerance of dishonour; and if there is nothing, there must be two kinds of high-mindedness (Posterior Analytics Il.xiii).


meaning?': an analytic, and an intuitive approach. The analytic definition of what it is to be a different lexical meaning involves the relationship between the extension of an item as a whole, and the subset of that extension covered by any definition of that item. It is then required that definitions of lexical meanings be maximally general and minimally specific. The definition of a lexical meaning should be maximally general in the sense that it should cover as large a subset of the extension of a lexical item as possible. Thus, separate definitions for "red sweet fortified wine from Portugal" and "white sweet fortified wine from Portugal" could not be considered definitions of lexical meanings, since they can be brought together under the definition "sweet fortified wine from Portugal". On the other hand, definitions of lexical meanings should be minimally specific in the sense that their extension should not exceed the bounds of the extension of the lexical item. Thus, a 'maximally general' definition covering both port "harbour" and port "kind of wine" under the definition "thing, entity" is excluded because the extension of entity is much larger than that of port. The requirements of maximal generality and minimal specificity define an upper and a lower bound to the specificity of the definition of lexical meanings. If a definition exceeds tiie upper bound, it defines a conceptual specification instead of a lexical meaning; if it exceeds the lower bound, it defines neither a lexical meaning nor a conceptual specification of the item under consideration. Against the background of these requirements, we can now define when an item will have more than one lexical meaning: an item has more than one lexical meaning if there is no minimally specific definition covering the extension of the item as a whole, and it has no more lexical meanings than there are maximally general definitions needed to describe its extension as a whole. This definition of ambiguity in function of the definitional analysis of an item is related to the traditional requirement that definitions be neither too broad nor too narrow," but it changes the component 'not too narrow' into 'as broad as possible'. As such, its traditional basis lies in Aristotle's remarks on the generality of definitions:

280


This definitional procedure (attempting to cover the entire extension of an item under a single definition, and postulating ambiguity only when this is not possible) presupposes a particular view of categorization. Concept formation is seen as a process of generalization, individual differences among the members of a category being neutralized by their common identity in terms of the point of view determined by the category. Generalization is a process of abstraction: concrete cases are subsumed under an abstract essence, e.g., the distinction between red and white port is neutralized by their essential identity as instances of the category port "sweet fortified wine from Portugal". But Antwerp (a harbour) and Noval (a brand of port wine) do not have a common essence generalizing over their differences as instances of port, and hence,port must represent two semantic categories. From this point of view, conceptual specifications can only be concretizations of lexical meanings; definitionally, they consist of a lexical meaning plus one or more additional attributes. (Thus, the lexical meaning "sweet fortified Portuguese wine" is concretized into a conceptual specification by adding features such as "tawny", "white", or "vintage".) For obvious reasons, the failure to provide a single necessary-and-sufficient definition for an item automatically entails its ambiguity according to the analytic criterium. On the basis of the requirement that lexical meanings be maximally general, ambiguity in the analytic sense simply equals the absence of a single extensionally maximal (and minimally specific) definition of an item. Thus, bird will be considered ambiguous because there is no minimally specific definition covering its extension as a whole. (The characterization "animal of which the female lay eggs and that has a bill or a beak" is maximally general but not minimally specific, since it applies to the duck-billed platypus). Consequently, each of the maximally defined subsets of bird (represented by the areas 2, 3, and 4 - but not by 1 - in Figure 1) will be considered a lexical meaning. In short, the analytic view of ambiguity links up with the classical view of categorization. It equates ambiguity with the absence of a single necessary-and-sufficient definition for all and only the members of the extension of an item,12 and it does so on the basis of a particular conception of the process of category formation, specified by the assumption that lexical meanings always represent maximally general subsets of the extension of an item for which a common essence can be defined. It is precisely this conception of category formation that is rejected by prototype theory. Categorization, it is claimed, is not based on essential identity, but on analogy (Rosch & Mervis 1975:574), i.e. partial identity, or resemblance.13 Curiously enough, studies in prototypical semantics always tacitly assume that the items whose semantic structure they unravel represent a single category; at least, the possibility that the absence of a single definition for the item as a whole might simply be an indication of ambiguity in the classical sense is hardly ever taken into consideration14. Let us therefore turn

281

(4)

Noval is a port (in a bottle), not a port (with ships).

(5)

*A penguin is a bird, not a bird."

(6)

At midnight the ship passed the port, and so did the bartender.

(7)

Vintage Noval is a port, and so is blended Sandeman.

(8)

*Valparaiso is a port, and so is vintage Noval.

(9)

A kiwi is a bird, and so is a penguin.

Second, the identity test described by Zwicky & Sadock (1975) involves transformations that require the semantic identity of particular constituents, such as conjunction reduction or 50-reduction. Thus, (6) is awkward if the two lexical meanings ofport are involved; disregarding puns, it can only mean that the ship and the bartender alike passed the harbour (or, perhaps, that both moved a particular kind of wine from one place to another). A 'crossed' reading in which the first occurrence of port refers to a harbour, and the second to wine, is normally excluded. This is taken to be evidence for the fact that, intuitively, portl and port! are recognized as scmantically different. Conversely, the fact that the notions "vintage sweet wine from Portugal" and "blended sweet wine from Portugal" can be crossed in (7) indicates that they are conceptual specifications rather than lexical meanings. As the only possible interpretation of (8) involves crossed readings, the sentence is intuitively disqualified. When applied to bird, the identity test shows the item to be


to the intuitive approach to ambiguity, and see whether it can furnish a criterion to prove that a prototypically organized category such as bird is not ambiguous. The intuitive approach implies, roughly, that one meaning is distinct from another if it is intuitively recognized as such by the language user. Analogously, an item will not be considered ambiguous if it is not judged to be so by the intuitions of the language user. As ambiguity is a property of lexical items as linguistic units, the intuitions in question will be linguistic intuitions; therefore, it should be possible to reveal ambiguities by using a variety of the most common test for determining the nature of linguistic intuitions, viz. a language user's spontaneous appreciation of sentences exhibiting a particular property that is related to the linguistic property under consideration. As far as ambiguity is concerned, two such tests stand out. First, according to the criterion described by Quine (1960:129), a lexical item is ambiguous if it can simultaneously be clearly true and clearly false of the same object. As (4) and (5) show, the test yields positive results for port but not for bird.

282

vague rather than ambiguous. Although kiwis and penguins do not belong to one and the same subset of bird that would be considered a lexical meaning in the analytic sense, (9) is not dismissed as awkward (cf. Figure 1): the semantic distinctions predicted by the analytic criterium are not revealed by the intuitive tests of ambiguity.

4. DELIMITATIONS OF PROTOTYPICALITY

The two revised definitions of prototypicality are not in opposition in a case such as bird. Examples can be found, however, where they are not in accordance with each other. Consider the Dutch adjective vers.11 It has three basic meanings.18 The most frequent one, exemplified by (10), refers to fruit, meat, bread, milk, and other foodstuffs, and can be paraphrased as "recently produced, and as such, optimal for use". This meaning can be seen as the conjunction of two components, viz. the notion "recent, new, not old", and


Against the background of these observations, the definition of prototypicality may be revised in two different ways. On the one hand, the difference between the analytic and the intuitive approach to semasiological structure may be accepted as a basic characteristic of prototypicality. Constitutive of prototypicality is then the fact that what is a distinct meaning from an analytic point of view need not be so from an introspective point of view. In particular, what is intuitively recognized as a single meaning may analytically be a cluster of overlapping meanings (i.e. of maximal subsets of the extension of a lexical item). On the other hand, one may hesitate to combine the analytic and the intuitive criterion of ambiguity, and try to find a definition of prototypicality that does not rely on introspective evidence (just as the original definition, mentioned in Paragraph 1, was couched in a purely analytic, definitional approach). In this case, prototypicality will be defined in terms of the structural relationship among the analytically defined meanings of the item. Notice, in fact, that the classical approach to semantic structure seems to imply that the meanings of an item are clearly separated: the referential subsets corresponding to the meanings of "high-mindedness" do not overlap. This is also the dictionary model of ambiguity: the semasiological range of an item is made up of numerically distinct, clearly separated senses.16 At least, even in the classical approach, overlapping cannot be excluded entirely. It is not difficult to imagine, for instance, a super-hero (say, Ulysses) being high-minded in both senses intended by Aristotle. Such cases are, however, incidental: overlapping will be rare, and will not occupy a firm position in the structure of the item. In prototypical categories, however, overlapping of meanings is maximal and structural. In fact, the prototypical centre of the category is formed by the maximally overlapped area of the clustered meanings.

283 the notion "optimal, in an optimal condition (and in particular, not contaminated)". As illustrated by (11) and (12), each of these two components can occur separately. On the one hand, there are expressions such as een verse wonde "a fresh, recent wound", in which the aspect of optimality is obviously absent. On the other hand, the component "optimal, not contaminated" is isolated in an expression such as verse lucht "fresh, pure, untainted air". Wie eet er de beste kost, wie koopt er verse vis en wild? "Who can eat the finest food, who can buy fresh fish and game?"

(11)

Voor een goede verpakking dienen de verse sigaren tot zolang in vochtige, weke toestand te blijven. "In order to be packed properly, the newly produced cigars have to remain soft and moist until packing".

(12)

Nu meende zij, dat zij aan alles zou kunnen wennen, behalve aan gemis van beweging en verse lucht. "Now she thought that she would be able to get used to everything, except lack of excercise and fresh air".

The semasiological structure of vers, schematically represented in Figure 2, is basically the same as that of bird: the item as a whole is a set of analytically distinct meanings, overlapping in a central application. The difference with bird lies not on the level of analytic structure, but on the intuitive level. Whereas the peripheral instances of bird are not recognized as semantically distinct, the peripheral senses of vers clearly are. Consider the following quote from the Dutch translation of Thomas Mann's Der Zauberberg.

01)

(12)

Figure 2. Dutch "vers": (10) New and therefore in an optimal condition (in particular, optimal for use); (11) New, recent, recently produced; (12) In an optimal condition (in particular, optimal for use).


(10)

284 De oude plekken, waar ik vroeger al eens ziek geweest ben, zonder het te weten, heb ik zelf gehoord, toen hij me beklopte, en nu moet er ook ergens een verse te horen zijn haha, 'vers' is anders wel een eigenaardig woord in dit verband (Mann 1979:254). "I have heard myself, when the doctor examined me, the old (tubercular) spots, where I had been ill without knowing it, and now it seems a fresh one can be heard somewhere round there - well, 'fresh' is a rather peculair word in this context".


The pun in German can be preserved in the Dutch translation, not only because Dutch vers exhibits roughly the same multiplicity of applications as German frisch,19 but also because language users are able to recognize the multiplicity. The senses "recent" and "in an optimal condition (in particular, optimal for use)" are recognized as being different, whereas there is no awareness of internal differences in the case of bird Of course, an example of literary usage is not the best kind of evidence one can imagine, particularly not in the case of punning. However, bona fide linguistic tests of ambiguity such as those mentioned in Paragraph 3 confirm the picture presented by the quotation given above. Against the background of the Quinean test, for instance, it would be perfectly acceptable to say that the cigars in (11) are vers in one sense but not in another. In fact, it is precisely because they are vers "newly produced", and hence, still moist, that they are not vers "optimal for use", since they would burn with difficulty. In short, then, vers shows that prototypicality in the analytic sense (maximal structural overlapping of maximally defined subsets) need not coincide with prototypicality in the other sense (introspective monosemy coupled with analytic polysemy). Vers is prototypical according to one, but not according to the other criterium. At this point, it is tempting to ask which of both revised definitions specifies the true nature of prototypicality, i.e. to ask what prototypicality really is. Should we take the analytic criterium as basic, and consider vers and bird indiscriminately as prototypical categories? Or should we take a more restrictive stance and speak of prototypicality only if analytic polysemy goes hand in hand with introspectively established monosemy? Such a discussion would seem, however, to be falling metatheoretically into the trap that is object-theoretically exposed by the prototypical research itself. Bird shows that things that are intuitively classified together need not be definable by means of a single definition; likewise, we should not be misled into thinking that everything that has been called 'prototypical' need exhibit one common essence, to be captured in a necessary and sufficient definition. Whether the notion 'prototypicality' is restricted to one of those differing things is a terminological decision that is less important than seeing the difference as such. In fact, there is at least one other phenomenon that has been considered prototypical and that is covered by neither the structural, analytic definition, nor by the introspective definition. As Armstrong, Gleitman & Gleitman (1983) have shown, some odd numbers are more salient than others. Now,

285 "odd number" is a concept that can be defined straightforwardly by means of a necessary-and-sufficient definition; also, the concept is not revealed to be ambiguous according to the intuitive tests mentioned before. As such, if "odd number" is called a prototypical concept, it is a form of prototypicality that is distinct from either the analytic or the introspective conception of prototypicality.20 Again, the important thing is not to know what prototypicality 'really' is, but rather to see what different kinds of things are intuitively grouped together as prototypical phenomena: categorization on the basis of similarity rather than identity, structural clustering round a central sense, and differences in membership salience. 2122

Next to questions about prototypicality, the problem made apparent by vers also raises more general questions about the domain of lexical semantics. Would it be acceptable to restrict lexical-semantic research to the analytic point of view, or should the introspective point of view be taken into account as well? There are two reasons for going beyond a purely analytic approach. First, if specifying the truth conditions of natural language sentences is a minimum programme for semantics, it is reasonable to expect lexical semantics to specify the contribution made by specific lexical items to the truth conditions of sentences. But the truth-conditional properties of natural language sentences will come out wrong if the description is restricted to the analytic level. For instance, if it is the task of lexical semantics to explain why (5) is contradictory, the analytic approach to ambiguity will make the wrong predictions, because a penguin is indeed a bird according to one analytically established meaning, but not according to another. Second, broadening the scope of semantics to incorporate the introspective approach to ambiguity allows a unified treatment of two hitherto unconnected phenomena, which can then be seen to form an elegantly symmetrical pattern. In fact, if bird shows that what is intuitively recognized as a distinct meaning may referentially be larger than an analytically defined meaning, it should be noticed that the reverse situation obtains with autohyponymous terms.23 For instance, dog has the two applications "male Canis familiaris", and "Canis familiaris in general, regardless of sex". As (13) makes clear, these are distinct meanings according to the intuitive Quinean criterium.24 According to the analytic criterium, however, they cannot be distinct meanings, since "male Canis familiaris" is always properly included in "Canis familiaris in general": only the latter is a maximal subset of the item. Thus, what is intuitively recognized as a distinct meaning (and what determines the truth conditions of sentences such as (13)), is referentially smaller than a distinct meaning according to the analytic criterium.


5. DELIMITATIONS OF SEMANTICS

286 (13)

Lady is a dog, but not a d6g.

(14)

He ate three biscuits.

(15)

He didn't eat three biscuits: he ate four.

A sentence such as (14) is intuitively ambiguous between a reading "exactly three" and "at least three" of three, since (15) shows that (14) can be simultaneously true and not true of the same situation. (This is the sentential


Bird and dog alike show that what is intuitively recognized as a separate meaning may be distinct from a 'meaning' according to the analytic criterium. More particularly, the differences between analytically and intuitively distinguished meanings fall neatly into two complementary kinds: as far as lexical meanings are concerned, bird involves analytic multiplicity combined with introspective, truth-conditional singularity, whereas dog involves analytic singularity combined with introspective, truth-conditonal manifoldness. In short, prototypicality a la bird and autohyponymy a la dog are sides of the same coin, viz. of the distinction between introspectively perceived ambiguity and analytically established ambiguity. The discrepancy between analytic and intuitive categorization provides a starting-point for broadening prototype theory into an encompassing theory of lexical semantics. The cognitive orientation of prototype theory25 favours such an extension: as it tries to determine the principles of natural language categorization, a cognitively orientated semantics should systematically explore the heteromorphic relationship between analytical distinctness and intuitive truth-conditional distinctness. To round off, it should be mentioned that there is an interesting similarity between the foregoing remarks on lexical ambiguity and some of the claims made by Kempson (1986) with regard to sentential ambiguity. Notice, to begin with, that bird and dog show that analytic distinctness of meaning is neither a necessary nor a sufficient condition for intuitively recognized ambiguity. As the latter involves the truth values attributed to particular propositions (such as the contradictoriness of (5)), distinctness on the analytic level appears to be neither necessary (dog) nor sufficient (bird) to determine truth-conditional distinctness. Likewise, Kempson tries to show that distinctness on what she calls the level of 'linguistic meaning' is neither necessary nor sufficient for truth-conditional distinctness on the level of propositional content. Typically, such truth-conditional distinctions may be revealed (among other things) by a sentential version of Quine's lexical test. An example of difference in 'linguistic meaning' not being a necessary condition for truth-conditional distinctness on the propositional level is provided by numerals.

287 version of theQuinean test. Notice that (14) applies to the situation since it is an implication of he ate four, but it is at the same time explicitly denied in the first part of (15)). This ambiguity on the propositional level (which is the level where truth conditions are involved) is not mirrored on the level of linguistic meaning: the ambiguity need not be specified in the specifics of any grammar for a particular language, since it can be described on the basis of general, language-independent pragmatic principles. In particular, Kempson explains the ambiguity on the basis of Sperber & Wilson's theory of relevance (1986). Conversely, conventional implicatures are said to show that being part of the semantics of a natural language is not a sufficient condition for being part of the truth-conditional content of propositions. He is an Englishman; he is, therefore, brave.26

The use of therefore in (16) conventionally implies that the speaker commits himself to it being the case that being brave is a consequence of being an Englishman, but an utterance of (16) would not necessarily be false, strictly speaking, should the consequence in question fail to hold. In short, "The concept of truth conditions does not provide a basis for determining what is part of semantics and what part of pragmatics if we take semantics to be that aspect of content specified within the grammar of the language" (Kempson 1986:88). Analogously, dog and bird show that the concept of truth conditions does not provide a basis for determining what belongs to a distinct category on the semantic level if lexical semantics is defined in terms of a referentially orientated definitional analysis of an item. This similarity should not, on the other hand, obscure the differences between Kempson's approach and the points made here. Although Kempson's definition of the propositional level in terms of introspectively tested truth conditions corresponds closely with our own, her conception of the 'semantic' level is not based on a definitionally analytic approach, but seems rather to involve the distinction between language-specific grammatical conventions and language-independent pragmatic principles. In the same vein, the general orientation of her paper is pragmatic whereas our own has been mainly cognitive.27 A detailed comparison of both approaches would take us too far beyond the proper domain of this article, but one major conclusion may be retained as a topic for further research. Both the cognitive approach (involving, e.g., the observation that truth-conditionally relevant categorization may be based on similarity rather than complete identity) and the pragmatic approach (involving e.g., Sperber & Wilson's principle of relevance) seem to entail that the truth conditions spontaneously attributed to natural language utterances do not constitute the only level of content analysis, but should themselves be analyzed as an outcome of the interplay between different principles and different levels of analysis.


(16)

288 6. CONCLUSION

Department of Linguistics University of Leuven Blijde-Inkomststr. 21 B-3000 Leuven Belgium NOTES 1. The prototypical view was developed in psycholinguistics by Eleanor Rosen. Apart from the articles mentioned in the text, Rosen (1978) should be singled out for its introductory value. Prototype theory is one of the central aspects of the school of cognitive semantics exemplified by the recent work of, among others, Lakoff, Langacker, Lindner, Hawkins, and Talmy: see Geeraerts (1986:187-243) for an introduction to the recent developments in lexical semantics. Rudzka (1987) contains a good sample of the work done in the framework of cognitive semantics. 2. See, for instance, Smith & Medin (1981:23).


Summarizing, then, I have argued for the following claims. First, the frequently made assertion that prototypicality refutes the classical view of categorization, viz. that categories always have a single definition in terms of necessary and sufficient conditions, is imprecise and incomplete since it does not specify the criteria according to which alleged prototypical concepts are indeed a single semantic category. Second, a revised definition of prototypicality may be based on either an analytic or an introspective criterium of ambiguity. In the former case, it is typical for prototypicality that analytically distinct meanings exhibit maximal and structural overlapping round a core application. In the latter case, it is typical for prototypical concepts to be monosemous according to the introspective criterium, but polysemous according to the analytic criterium. Third, cases in which both revised definitions are not simultaneously applicable should not lead to an essentialistic attempt to define the 'true nature' of prototypicality, but rather to the recognition (in the spirit of prototype theory itself) that what has intuitively been categorized as a single notion "prototypicality" covers a number of phenomena that frequently co-occur but that prove to be distinct upon closer analysis. Fourth, a systematic exploration of the relationship between the analytic and the introspective approach to semasiological structure reveals that prototypicality is part of a larger class of phenomena exhibiting discrepancies between both approaches. In particular, autohyponymy and prototypicality show that analytic distinctness of meaning is, respectively, neither a necessary nor a sufficient condition for introspectively recognized ambiguity. Finally, an investigation into the relationship between the analytic and the introspective conception of semasiological structure links up with other kinds of research showing that content analysis should not stop short at the level where the truth conditions of utterances are envisaged.

289


3. The term item is used here in a neutral sense, to indicate any aspect of reality: it may refer to processes or relations just as well as to objects or attributes. 4. An explicit statement may, e.g., be found in Labov (1973), Coleman & Kay (1981), Lakoff (1982). As for my own adherence to this definition of prototypicality, see Geeraerts (1985a: 142). 5. The example is taken from Hurford & Heasley (1983:121). 6. If two or more distinct meanings occur within the same lexical item, polysemy rather than homonymy obtains. The long-standing issue how to distinguish between one or more lexical items, i.e. between polysemy or homonymy, can remain unsolved for the purposes of this article: when talking about ambiguity, we shall mean semantic multiplicity in general, disregarding the distinction between homonymy and polysemy. The relationship between the latter issue and the subject-matter of the present paper is treated in Geeraerts (1988) against the background of the views of Kempson (1980). 7. The distinction between vagueness and ambiguity prevents items from being infinitely polysemous. If any contextual reading of port would automatically be a lexical meaning, the item would have at least as many meanings as there are harbours and bottles combined. The literature on vagueness and ambiguity is vast; for a bibliographical overview, see Fries (1980). 8. A possible scientific definition offtirrffulfilling the requirement of necessity-cum-sufficiency (say, in terms of a particular DNA-structure) would not necessarily be a counterexample, since prototype theory explicitly deals with concept formation in natural language and in common sense thinking, not in technical contexts such as that of specialized sciences. 9. Notice that the discussion is restricted to the literal, "animal species" sense of bird. Figurative extensions are not taken into account, basically because they are intuitively and traditionally recognized as distinct meanings. The question of vagueness or ambiguity is automatically answered in favour of ambiguity in the case of the non-literal meanings of bird. 10. An alternative way of putting the problem is this: given the fact that prototype theory takes the form of a conjecture about the definitional structure of semantic categories, what proof is there that the traditional examples of prototypicality do in fact constitute single categories? 11. Cf. Copi (1972:138). 12. Notice that the requirements of minimal specificity and maximal generality correspond rather with the 'all and only' part of this expression than with the 'necessary-and-sufficient' part. Maximal generality implies attempting to define all the members of a category, minimal specificity implies defining only these. Sufficiency and necessity are requirements with regard to the conditions listed in definitions, specificity and generality with regard to the referential scope of the definitions. 13. At this point, prototype theory links up with the family resemblance theory of Wittgenstein (1953:§67). 14. For a concept such as bird, this seems plausible; it is, however, less intuitively clear for an item such as lie (Coleman & Kay 1981), or, for that matter, my own vergrijpen (Geeraerts 1983). But even for the plausible cases, specific criteria for the absence of ambiguity should be given. 15. Of course, this sentence is perfectly acceptable if figurative meanings of bird art taken into account. However, as mentioned in note 9, only the "animal species" applications are considered. 16. This way of putting things is not entirely fair with regard to lexicography. Large-scale dictionaries with a scholarly rather than a commercial or a pedagogical purpose do indeed exhibit a model of semasiological structure that corresponds more closely with the prototypical conception than with the classical conception (see Geeraerts 1985b). The fact that the latter conception seems paramount in popular, smaller dictionaries is less a matter of theoretical principle than a matter of practical requirements with regard to the organization of the microstructural level in dictionaries. 17. Vers has roughly the same range of application as English fresh. Both words go back to Proto-Germanic */n^-> though along different routes: vers is a direct cognate, whereas/rwA goes back to Old French freis,fresche, which is itself a loan of the original Germanic form.

290

REFERENCES Armstrong, S., L. Gleitman &H. Gleitman 1983: "What some concepts might not be'. Cognition 13, 263-308. Bosch, P. 1983: Agreement and Anaphora. New York: Academic Press. Coleman, L. & P. Kay 1981: 'Prototype semantics: the English verb lie'. Language 57, 26-44. Copi, I. 1972: Introduction to Logic. New York/London: MacMillan/Collier. Fries, N. 1980: Ambiguita't und Vagheit. Tilbingen: Niemeyer. Geeraerts, D. 1983:'Prototype theory and diachronic semantics. A case study'. Indogermanische Forschungen 88, 1-32. Geeraerts, D. 1985a: Paradigm and Paradox. Explorations into a Paradigmatic Theory of Meaning and its Epistemological Background. Leuven: University Press.


18. This is only a rough sketch; a more refined analysis would show that there arc not three, but some thirty different semantic applications to be considered: see my article in the Woordenbock der Nederlandsche Taal XX, 2136-2171. As these applications do fall together into the three major groups illustrated in (10H12), a restriction of the discussion to the latter does not create any relevant distortions. The quotations in (10H12) are taken from the corpus of the Woordenboek der Nederlandsche Taal. 19. Dutch vers lacks the specification "cool" which is marked in the structure of German/rwc/i. 20. Armstrong, Gleitman & Gleitman themselves rely on an 'essentialistic' stance when they take the view that "odd number" is not a prototypical concept, precisely because it lacks a necessary-and-sufficient definition. See the discussion in Lakoff (1982:39-43). 21. See also Wicrzbicka(1985) for the "many senses" of the notion prototype. 22. Another 'prototypical' phenomenon that might be mentioned here and that is covered by neither of the revised definitions is the absence of clear referential boundaries for a concept. Cf. Pulman (1983) on the differences between membership salience and membership as such. 23. See Horn (1984), Rohdenburg (1985) for the phenomenon of autohyponymy and its complexities. 24. Zwicky &Sadock (1975) have pointed out that autohyponymous terms are not recognized as ambiguous by the identity test. In fact, any time the subordinate interpretation is combined with an instance of the superordinate use of the item, the latter is applicable to the former as well; this is to say, ungrammaticalities due to a 'crossed' combination of the subordinate and the superordinate interpretation of an autohyponymous item are filtered out by the fact that in the same circumstances, a uniform superordinate reading is available. This means that the introspective tests mentioned here do not always yield the same result. Further research will have to examine why the introspective test a la Zwicky & Sadock need not always correspond with the Quinean test. Another difficulty with regard to the identity test of ambiguity is pointed out by Bosch (1983:80-83). His observation that the result of the identity test may vary according to the context is another indication of the variability of intuitively perceived identity; the distinct results of the Quinean criterium and the Zwicky & Sadock test point towards the same conclusion. 25. Cf. n o t e l . 26. This example is not given by Kempson herself, but is taken from Grice (1975), to whom she refers for the phenomenon of conventional implicature. As Kempson's remarks with regard to this second type of distinction between the semantic and the pragmatic level are rather scant, the analysis of (16) given here closely follows Grice's. The analysis is expository only; no attempts at evaluation are made. 27. The pragmatic approach deals with questions of the sort "How do utterance interpretations come about?", whereas the cognitive approach asks questions such as "How does categorization work?".

291


Geeraerts, D. 1985b: 'Les donnees prototypiques, stereotypiques et encyclopediques dans le dictionnaire'. Cahiers de Lexicologie 46, 27-43. Geeraerts, D. 1986: Woordbetekenis. Een Overzicht van de Lexicale Semanlick. Leuven/ Amersfoort: Acco. Geeraerts, D. 1988: 'On hypertrophic homonymy'. To be published in F. Heyvaert & F. Steurs, eds., Worlds behind Words. Leuven: University Press. Gnce, P. 1975: 'Logic and conversation'. In P. Cole & J. Morgan, eds., Syntax and Semantics 3. New York: Academic Press. Horn, L. 1984: 'Ambiguity, negation, and the London School of Parsimony'. NELS 14,108-131. Hurford, J. & B. Heasley 1983: Semantics. Cambridge: University Press. Kempson, R. 1980: 'Ambiguity and word meaning'. In S. Greenbaum a.o., eds., Studies in English Linguistics for Randolph Quirk. London: Longman. Kempson, R. 1986: 'Ambiguity and the semantics-pragmatics distinction'. In C. Travis, ed., Meaning and Interpretation. Oxford: Blackwell. Labov, W. 1973: 'The boundaries of words and their meanings'. In C. Bailey & R. Shuy, eds., New Ways of Analyzing Variation in English. Washington: Georgetown University Press. Lakoff, G. 1982: Categories and Cognitive models. Trier: Linguistic Agency of the University of Trier. Mann, T. 1979: De Toverberg (transl. P. Hawinkels). Amsterdam: De Arbeiderspers. Pulman, S.G. 1983: Word Meaning and Belief. London: Croom Helm. Quine, W.V.O. 1960: Word and Object. Cambridge, Mass.: MIT Press. Rohdenburg, G. 1985: 'Dogs, bitches, and other creatures'. Journal of Semantics 4, 117-135. Rosch, E. 1975: 'Cognitive representations of semantic categories'. Journal of Experimental Psychology 104, 192-233. Rosch, E. 1978: 'Principles of categorization'. In E. Rosch & B. Lloyd, eds., Cognition and Categorization. New York: Academic Press. Rosch, E. & C.B. Mervis 1975: 'Family resemblances. Studies in the internal structure of categories'. Cognitive Psychology 7, 573-605. Rudzka, B., ed. 1987: Advances in Cognitive Grammar. Amsterdam: John Benjamins. Smith, E. & D. Medin 1981: Categories and Concepts. Cambridge, Mass.: Harvard University Press. Sperber, D. & D. Wilson 1986: Relevance. Oxford: Blackwell. Wierzbicka, A. 1985: Lexicography and Conceptual Analysis. Ann Arbor: Karoma. Wittgenstein, L. 1953: Philosophical Investigations. Oxford: Blackwell. Zwicky, A. & J.Sadock 1975: 'Ambiguity tests and how to fail them*. In J. Kimball.ed., Syntax and Semantics 4. New York: Academic Press.


THE CONSTRUCTION OF PROPERTIES UNDER PERSPECTIVES RENATE BARTSCH

ABSTRACT

1. THEMATIC DIMENSIONS AND CONTEXT DEPENDENT INTERPRETATION OF ADJECTIVES

Not all adjectives express properties but there is a common semantic characterization of adjectives, namely as functions which map thematic dimensions onto properties. Thematic dimensions are aspects of the context in which predications of adjectives have to be understood. Adjectives indicate specifications within these dimensions. Thematic dimensions can be more or less precise in providing the perspectives which are relevant for the interpretation of adjectives. A dimension can be understood extensionally as a set of properties. Those properties that are situated within one dimension have the dimension as a common presupposed semantic marker or, in other words, they contain the dimension as a presupposed perspective which makes the predication of these properties possible. But this containment of a dimension in a property is different from the way a superordinated concept is contained in a subordinated one, though also these thematic dimensions are given by explicit or implicit questions which a text is supposed to answer. Let us first consider a case of superordination. A question like "What is this grey thing over there?" restricts the realm of answers to those referring to grey things: the concept GREY THING will be presupposed in all the answers, except if a correction is made, like " it is not grey, it is brown"; but otherwise answers like "a cat" or "a handbag", or "it is not a handbag, but a cat" all presuppose the property GREY THING as being


Adjectives express properties, is the traditional opinion, be it properties an individual has itself (absolute properties) or properties that it has only in relation to others (relative properties or relations). I shall show in this paper that most properties are not expressed by adjectives, rather they are denoted by them in what I call "thematic dimensions". Properties are expressed by thematic dimensions and adjectives together. Adjectives can be used in predicative, adnominal, adverbial, or adsentential position and function. Besides elaborating the notion of 'thematic dimension' and explaining the relationship between properties and these dimensions, an aim of this paper is to assign suitable semantic types to adjectives and the 'thematic dimensions' they are used in. These serve for forming conjunctions and other combinations of thematic dimensions, and for forming conjunctions of expressions in several categories. These operations form the basis for the construction of properties under perspectives, i.e. in thematic dimensions.

294 true of the things referred to in the answer. These may be grey cats, grey handbags, or whatever grey thing may be at the place indicated. In the context of this question, the concept GREY THING is superordinated to the concepts GREY CAT, GREY HANDBAG, etc., which are expressed by just the nouns cat, handbag, etc. These nouns are interpreted with respect to the domain of grey things in the answers "a cat", "a handbag", or whatever. Likewise, when we ask "How is John's health?", the answer "bad" will be interpreted as John being in bad health. Dimensions, like HEALTH are domains of interpretation for adjectives. The answer "bad" does not simply express a property BAD, but rather the property BAD WITH RESPECT TO HEALTH, which is expressed by the adjective sick. The relationship between

and narrower concepts (which are all of the same order), but one of hierarchy in terms of a different order: SICK is a health-property, like RED is a colourproperty. This means that the dimension HEALTH and the dimension COLOUR are the set of all health-properties, and the set of all colours, respectively. In this paper, I shall treat all properties on a type-level identical to the type-level of thematic dimensions. Within these dimensions, they will be treated as subsets consisting of just a single property, and then the relationship between a thematic dimension and such a lifted property contained in it is formally exactly like in the first example, namely one of superordination in a hierarchy between semantic 'entities' of the same order, though the whole relationship takes place on a higher level. Not all properties are directly expressed by adjectives, but those adjectives which express properties, like sick, 3 miles long, 5 feet tall, red, are marked semantically for their dimension, HEALTH, LENGTH, HEIGHT, COLOUR, and are thus interpretable in a context-independent way. They can only be used in one single dimension which means that they are stable characters in the sense of Kaplan (1978), who formulated a theory of indexical interpretation for demonstratives. This approach can be extended to other kinds of words. Stable characters cannot be used in another dimension than the one for which they are semantically marked, except in a metaphoric way. If metaphoric use happens repeatedly for a property-expressing adjective and extends over several dimensions, we get a metaphorically initiated broadening of its meaning, whereby the specific dimension is elimited from its content; it thus becomes unstable, i.e. becomes a function from dimensions to properties. Different from concept broadening, metaphorical extension of the use of an adjective is an extension over several dimensions, which means that it becomes polysemic: for each dimension it indicates another concept (or, as I shall point out later, another equivalence class of concepts if the dimension is not fully precise with respect to the meaning of the adjective). Those adjectives that can be used in several dimensions do not express properties but characters that, together with a dimension, make up a property. Therefore they express a property only in a context, but not by themselves. These


the thematic dimension HEALTH and the property BAD WITH RESPECT TO HEALTH is not one of superordination in the sense of a hierarchy of broader

295

FAR AS SPEED IS CONCERNED, Or GOOD AT SWIMMING AS FAR AS STYLE IS CONCERNED, Or GOOD AT SWIMMING AS FAR AS ENDURANCE IS CONCERNED, Or it

can mean some of these or all of these together. We can interpret good at swimming as a property, namely as GOOD AT SWIMMING UNDER AT LEAST ONE OF THE RELEVANT ASPECTS, or if speed generally is taken to be presupposed as the most important aspect, it can mean GOOD AT SWIMMING AS FAR AS SPEED IS CONCERNED. But we can also take good at swimming to be a pre-property, which is more specific than the pre-property GOOD, but which has still to be applied to another thematic dimension in order to yield the property predicated of John. Thus, a pre-property can be a function from a thematic dimension to a more specific pre-property, which in a special case can be stable, i.e. can be a property, which is a pre-property in which a maximally precise dimension is included. A pre-property can be understood extensionally as an equivalence class of properties, namely as the set of all the properties it assigns in all the different dimensions. The notion of context is broad; in one way because it comprises linguistic and situational context. I refer to both. It is also broad, because it comprises different context properties. Here I shall restrict myself to one kind of context properties which I call 'thematic dimensions', extending the notion of'dimension' which we know from spatial and temporal specifications of entities. Thematic dimensions can be differentiated into finer grained dimensions by additional perspectives and aspects, some of which are contained in the lexical semantics of nouns that are used to describe the entities of which the


adjectives, by themselves, are dimensionally not determined or they are underdetermined. I also call them dimensionally weakly determined adjectives. Examples are good, strong, much, satisfactory, big, great and their antonyms, and especially the Dutch adjective flink, which was treated extensively in Bartsch (1986). To be able to interpret these adjectives one has to know under which perspective they are used in the respective situations: good, well, strong, weak, satisfactory, great in which respect? Some of them have, if no dimension is indicated, a (linguistically) 'unmarked' meaning, such as good as morally good, strong as physically strong, big as big in length and width; and this is the case as long as nothing in the context is contrary to this assumption. In these situations of use they express properties relative to the 'unmarkedly' presupposed dimension. In all other cases they are functions from contexts to different contents, here properties, and are thus generally 'characters' in the sense of Kaplan (1978).' I shall use the name 'pre-property' for functions from contexts to properties. A dimension is fully precise with respect to a pre-property, if and only if the pre-property assigns in the dimension exactly one property. Otherwise, if it assigns a class containing more than one property, further specification is possible in order for the pre-property to assign a property. Sometimes several steps of specification are necessary to get from a preproperty to a property. In the sentence John is good at swimming, the predicate good at swimming can mean different things GOOD AT SWIMMING AS

296

(WELL DONE, MEDIUM DONE), WET, DRY.

Dimensions can be of lesser or greater extension: the dimensions of colour or shape are very large and open. They are not ordered, though they permit several partial orderings from various points of view. Other dimensions are small; they exist merely of one property and its antonym, as e.g. the dimension of genuineness, which consists only of GENUINE and FAKE, or the dimension of truth, which is {TRUE, FALSE}; the dimension of verification is broader, namely {VERIFIED, CONFIRMED, SUPPORTED, UNCERTAIN, NON-VERIFIED, DISCONFIRMED, FALSIFIED}. For small dimensions there is mostly no separate name. A nominalization of the adjective, or the adjective itself, sometimes in disjunction with its antonym, or the adjective followed by the question "yes or no?" serve to name the dimension, for example: "human, yes or no?" The general pre-properties in these cases are only the two poles of a dimension, expressed by positive and negative, sometimes with a third, neutral, value. In linguistics, for example, adjectives are used as names for dimensions in


adjectives are predicated. In this way we have more general and more specific dimensions in which properties can be located. Other than space and time dimensions, thematic dimensions are intentional dimensions. An object or event can be determined with regard to its location in space and time: it can be projected onto space and time coordinates, i.e. it has its space and time projections. Likewise it can also be determined with regard to thematic dimensions: what is John's situation, for example, from the point of view of health, financially, socially, what is his way of dressing, living, thinking, etc? The projections of John onto these dimensions are John's health, John's financial situation, etc. The nominal terms John's health, John's financial situation refer to subsets of the set of pre-properties of John, for example to the set of those pre-properties that designate health properties John has. If we look at an individual from different perspectives, we see it differently. The projections onto different thematic dimensions are the ways in which we single out the relevant properties an individual has, relative to the thematic interest given. These can be absolute properties (e.g. YELLOW) versus relative properties (e.g. LONG), or essential properties (e.g. HUMAN) versus accidental ones (e.g. THICK, BLOND, HEALTHY). Before treating the grammar of adjectives and dimension indicating expressions I shall make the notion of dimension more concrete. Thereby I extend the traditional notion of 'dimension' to all kinds of sets of properties that are grouped together and ordered under some perspective. Dimensions in the narrower sense contain measures and values, quantitative as well as qualitative. Next to the evaluation and measure adjectives, which permit grading, there are non-gradable adjectives which express substantial properties as FELINE, HUMAN, MALE, properties of shape or form as CIRCULAR, ELLIPTIC, non-gradable value properties like FAKE, GENUINE, TRUE, non-gradable state properties like RAW, COOKED, BOILED, PREGNANT, and disposition properties like BREAKABLE, SOLVABLE. Some dispositional properties are also gradable, as well as some state properties, like RIPE, DONE

297


semantic or phonetic marker analysis, together with'+' a n d ' - ' as the names for the two antonymic pre-properties (human: + or -, round: + or -), or in medical analysis, where positive or negative are used as names of pre-properties, which together with different dimensions x, defined by questions of the form "Are there antibodies of the virus x in the sample?", express a medical property of the sample (or of the person from whom the sample is taken). Words expressing pre-properties are answers to questions for a suitable specification of an object in the dimension which is 'opened' by the question. A yes-no-question divides a dimension into two parts, such that only one of the two pre-adjectives (+ or -) is a suitable answer. Those adjectives, which also name the dimension, in an unmarked case express the positive value within the dimension. The other value is expressed by negation. Antonyms can serve to name the same dimension. For example, "Wet, yes or no?" and "Dry, yes or no?" point to the same dimension, though in an opposite orientation: in the first case, WET is the positive, and in the second DRY is the positive value. Note that the common colour words express stable characters, which have the dimension COLOUR encoded within their semantics, but a word like soft can also be used with respect to the colour dimension. It is not a property, since it does not denote a colour by itself, but a pre-property which, when applied to the colour dimension, selects a certain class of colours from that dimension. Some of these are the properties that are predicated when we say that the colours of a certain dress are soft. The most important dimensions of physical properties are SHAPE (round, angular, triangular, ..., pearshaped, eggshaped, ...), CONSISTENCY (pappy, liquid, viscous, solid), HUMIDITY (wet, damp, dry), SPATIAL EXTENSION (specified in three dimensions HEIGHT, LENGTH, WIDTH), TEMPORAL EXTENSION, and SPEED. The first two do not have an ordering structure by grading, the other dimensions have. Such an order can be found in quantitative as well as in qualitative dimensions: quantity can be ordered according to measures from big to small quantity, and quality from high to low quality. In Fig. 1 the vertical lines represent different dimensions and the horizontal lines represent different characters (i.e. pre-properties) whereby a character indicates a property at each dimension, namely its intersection with that dimension. Note that a pre-property and a dimension are both sets of properties of which, extensionally speaking, the intersection is formed. The pre-properties are ordered, i.e. they are points on a 'pro-dimension', which is projected on all the other dimensions such that an ordered set of properties is determined which corresponds to the ordered set of the pre-properties. The dimensions in Fig. 1 are taken to be fully precise with respect to the pre-properties under consideration. This means that the intersection between a pre-property and a dimension contains exactly one element.

298

much, high

Figure 1: Pre-propcrties (characters) and dimensions

The dimensions d,, d-, d k are ordered sets of properties, the pre-properties or characters K; are functions which have for each dimension the intersection point as a value (cf. Fig. 1). The Kj represent the points in the general ordering structure which is projected on the various dimensions: by applying a norm of comparison to a dimension it is mapped onto a measure scale or a scale of valuation, as represented in Fig. 2. Measure dimensions are scaled by absolute dimensional adjectives, which are the (numeral) measures of comparison with a standard unit, for example l CM. Valuation dimensions contain properties denoted by the relative adjectives, which are the valuations. A measure dimension, for example LENGTH IN CENTIMETRES, is compatible with a valuation dimension, for example LENGTH OF FIVE YEAR OLD CHILDREN, because the ordering structure is kept intact in the mapping from the basic dimension d ; (e.g. LENGTH) to d(J (e.g. LENGTH IN CENTIMETRES) or dik (e.g. LENGTH OF FIVE YEAR OLD CHILDREN). If we apply a number just to the dimension LENGTH we get something like 3 LONG, or 5 LONG, which are not

properties but pre-properties, namely the equivalence classes like, for example, {3 CM LONG, 3 M LONG, 3 INCHES LONG, . . . } . In order to yield a property, a measure specification is needed, for example 3 LONG IN CM, or 5 LONG IN CM. These are properties in the dimension LENGTH IN CM. Likewise, LONG and SHORT are pre-properties in the dimension LENGTH VALUATION which, in order to designate a property, need the specification of a norm or value for comparison, like AVERAGE OR EXPECTED LENGTH OF FIVE YEAR OLD CHILDREN. Measure and valuation is possible in quantitative as well as in qualitative respects. The pair is a dimension specified with


little, low

299 dimensions

valuations with respect to A, B, C

real numbers

much compared to A

compared to B

: projections

Figure 2: Measures and valuations

respect to measure, in which an object can be determined, for example, according to its length in inches. The number 30 in the property 30 CM LONG is a character, i.e. a pre-property. It is located on a pro-dimension, namely on the ordered set of real numbers. The number indicates a property on each measure-specific dimension, which can be predicated of an object, as, for example 30 CM LONG.

Likewise the pair is a dimension specified with respect to perspective, on which, for example, properties are situated which are expressed by relative adjectives like tall, long, short, narrow in the dimensions of space extension. The pragmatic perspectives, which I have taken over from Ehrich (1975), differentiate dimensions into dimensions which incorporate a norm of valuation. This means that a norm of comparison or middle-value has been specified on the dimension. These norms or middle-values of comparison can be based on: 1. objects of comparison of the same kind, e.g. 'This mouse is big compared to other mice'. 2. objects of comparison in the surrounding, e.g. 'This baby-elephant is big compared to the children standing around'. 3. objects of comparison of the same kind in the surrounding, e.g. 'This


compared to C little

300

kinds of measures

measure standards

cm

inches

space dimensions \ height

length

width

Figure 3: Kinds of measures and perspectives of valuation

4.

house is big here in Muiden; in Amsterdam the same house would be small'. the speaker or observer as the object of comparison, e.g. the small child thinks "Daddy is big", and for children in general, adults are the big people.

In this way, a dimension, e.g. the dimension of body-size, is differentiated into various dimensions dtJ = , where p^stands for pragmatic perspectives. These perspectives are context-properties. They are functions p= which map dimensions dj on dimensions specified with respect to perspective djj, &{ = &(&), which come about by installing on the respective dimension a norm of comparison which is defined by the perspective p r This means that one of the properties on the dimension is fixed as the norm, or that a group of properties form the centre of the dimension with which all the others are compared as being more or less, for example, the size of a small child in the surrounding or the average size of small children generally, whereby a small child can be big. With respect to absolute dimensional adjectives, we thus have as their dimensions pairs consisting of dimension and measure d- = . Hereby we can take the measure to be a function which maps the respective dimension on a measure-determined dimension rn^dj) by marking properties on the dimension according to the measure, e.g. a certain length property as being the property 30 INCHES LONG. And with respect to relative dimensional


Pragmatic perspectives: objects of comparison of 1. the same kind objects of comparison from the surrounding objects of comparison of the same kind in the surrounding the observer/speaker as object of comparison

301

WITH RESPECT TO WHATEVER IDEAL, AIM OR GOAL, and it Consists of the

pre-properties GOOD, BAD, SATISFACTORY, and grades in between. As it happens, people also try to develop measurements of qualities by singling out a cluster of properties that are relevant for reaching the respective ideals or goals, and then devising a mapping into a set of numbers, for example according to the number of times each of the properties has been shown in a standard test, and possibly according to the weight that has been fixed for each property to begin with, in order to represent importance of a property for the goal involved. However, the use of numbers in these contexts is quite different from the use of numbers in measuring quantities: with respect to


adjectives, we have as their dimensions pairs consisting of dimension and pragmatic perspective dij = , or the value-determined dimension Pj(dj), which comes about by marking one of a cluster of properties on the dimension as the norm and marking the other properties on the dimension according to their position with respect to the norm. If the dimension is SIZE and the pragmatic perspective of comparison is the average size of small children in the respective surrounding, then every small child whose size is a property way out on the positive end of the dimension is a small child that is big when seen under the given perspective. Therefore, "This small child is big" is not a contradiction if small child is interpreted with respect to older children, and big is interpreted with respect to its own younger age group: this [small child] d in is [big] d. im. The distinction between measure-determined and value-determined dimensions is not identical with the one between quantitative and qualitative dimension. A quantitative dimension can be determined according to measure as well as according to valuation. This is, for example, the case with respect to the dimension of body hight (180 cm tall versus tall) or the dimension of temperature (JO degrees Celsius versus hot). This means that there are acts of quantitative measurement and acts of quantitative evaluation. The same is partly true for qualitative dimensions: intelligence can be measured (according to psychologists), and it can also be evaluated as high or low. The measurement in these cases is done in order to establish a scale which can be used in evaluation relative to the established average or norm. The difference between qualitative and quantitative dimensions is the following: quantities are determined by how many times they contain a standard quantity, for example an inch in length, which is the measure. Thus we get the quantity properties of length on a rod, ordered according to the ordering structure of real numbers, which is the general pro-dimension of measurement. Or quantities are determined by how they are situated in terms of more or less with respect to a quantity which is the standard or norm. The general ordering structure ( < o n an ordered set) is the pro-dimension of evaluation by comparison. Qualities, on the other hand, are determined with respect to ideals, purposes, aims and goals, which people have with respect to other beings, tools, events, and actions. The general pro-dimension here is APPROPRIATENESS

302

PERSIST IN TRYING TO OVERCOME ADVERSITIES.

For most properties natural languages have no names; most properties are constructed from a small set of pre-properties and a set of thematic dimensions, and accordingly they are not expressed by a single word but by a text. Thematic dimensions together with pre-properties establish in a text the properties predicated of individuals, as I shall show in 2.4. The introduction of thematic dimensions in a conversation or text provides a basis for text coherence. Next to being introduced by questions they are also introduced by expectations the speaker has about the thematic interest the hearer might have.


qualities, numbers are attached to certain characteristics (i.e. the relevant properties) in a purely conventional manner; likewise the use of an algebraic operation on these numbers is merely conventional, and there is no way back from the final result of measurement to the distribution of occurences of properties on which it was based. Certain arithmetric operations make no sense on such qualitative measurements: what does it mean to divide the quality measure of an apple or of the intelligence of a person by half? Which property is the result of such an operation? We cannot tell. But note that the same question can always be answered with respect to quantities. Which quality dimensions are relevant for judging an object depends on the kind of object. There are descriptive ones like SHAPE, COLOUR, POSITION, FINANCIAL STATE, SOCIAL STATE, and normative ones like HEALTH, SUCCESS, MORAL CONDUCT, and others, where the descriptive ones are often also marked normatively. We can make a basic distinction between kinds of dimensions of valuation from the points of view of norms of ideal for objects and courses of events, including persons and actions, on the one hand, and of norms of purpose and aim concerning objects, and of norms of result concerning telic processes and especially actions. The norms of ideal with respect to objects and courses of events concern the esthetic aspects of 'gestalt' (form, substance, colour, movement), and for humans and living beings generally also aspects of well-being (health, situational conditions), and especially with respect to people there are aspects of their social, economical, occupational situation, as well as aspects of moral obligation and emotional states and dispositions. The norms of purpose and result are differentiated by all kinds of relevant goals and results. All these norms with their respective specifications provide perspectives under which qualitative dimensions are defined, in which the respective qualitative properties are located, as there are, for example, the various properties of a person's character. The pre-properties are then functions which indicate properties within these dimensions. These kinds of dimensions are the themes of conversations with respect to those objects that are topic, and the properties predicated about a topic under such a thematic dimension are in focus. For example, the pre-property STRONG, indicates the property ENDURING if it is used with respect to a person as the topic, and with respect to the fairly specific thematic dimension CAPACITY, YES OR NO, TO

303 2. THE GRAMMAR OF DIMENSION INDICATING EXPRESSIONS

Dimensions can be indicated by adjectives, e.g. by long in 3 inches long or by nominalizations of adjectives, e.g. length. In order to express that a predication is restricted to a certain dimension, dimension indicators can be used in the form of prepositional phrases, morphologically modified forms of the respective adjectives, or adjectives without any special modification. The dimension indicators appear syntactically as ad-sentential, adverbial, or ad-terminal phrases: 1.

3.

In these different ways they are, from a grammatical point of view, functions which restrict a predication to a specific dimension or perspective.2 The sentence John is fine by itself is vague as long as no thematic dimension is specified. In everyday conversation, some relevant perspectives are presupposed, and the sentence is understood to be meant with respect to all or most of them. Often, hearers are not interested in which exactly are the dimensions referred to. In this way, sentences can be vague.3 Expressions which indicate dimensions can be iterated and can thus be used in different categories: John exceeds Paul in swimming (, but not in running). With respect to speed (, but not with respect to style), John exceeds Paul in swimming. In swimming (, but not in running), John's speed exceeds Paul's. This iteration amounts to a stepwise precisification of the dimension such that the relational pre-property EXCEED first maps the dimension IN SWIMMING on the pre-property EXCEED IN SWIMMING, and then, by applying this to the dimension SPEED, finally results in the stable relation EXCEED IN SWIMMING WITH RESPECT TO SPEED. This can be done in one step by constructing a fully precise thematic dimension SPEED OF SWIMMING and interpret exceed with respect to it:

With respect to speed of swimming (, but not with respect to style), John exceeds Paul. The 'intentional structure' is different, i.e. the way of getting at a fact differs,


2.

as ad-sententials combinable with sentences, where the result is again a sentence; example: With respect to health John is fine. as adverbials combinable with adjectives or verbs, where the result is again an adjective or verb; example: John is fine healthwise. as ad-terminals combinable with terms (i.e. quantified noun phrases, also called 'generalized quantifiers'), where the result is again a term; example: John's health is fine.

304

but the fact is the same. The way of getting at a fact is a matter of focussing. Note that although the order of specifying the dimension is in principle arbitrary, it makes a difference in the way a text can be continued: With respect to swimming, John's speed exceeds Paul's, and so does his style. With respect to speed John's swimming exceeds Paul's, *and so does his style. With respect to speed of swimming, John exceeds Paul, *andso does his style.

1.

2. 3.

+Truth evaluation = Relative Truth evaluation, i.e. the sentence is claimed not simply to be true, but true with respect to a certain dimension and under a certain perspective. This means that truth evaluation is restricted to a dimension and/or perspective. + Pre-property = (Pre-)property. +(intentional) Object = Partial Object.

In what follows, I am going to describe how this limitation of the interpretation of a sentence is performed in the ad-sentential, adverbial, and ad-terminal manner. With each of these constructions I shall investigate the combinability of dimension indicators. First some remarks about semantic types are necessary: In a Fregean semantics, the interpretation of an indicative sentence has as its result a truth value, its denotation. Here the interpretation is dependent on several parameters, called indices: world (:w), time (:z), and context (:c).4 A sentence-concept then is a function that maps a triple onto a truth value (:t) 'true' or 'false'. The semantic type of an indicative sentence is accordingly , its denotation type with respect to the triple is t. Such a sentential concept is the context-, time- and world-undetermined content of the sentence. This means that these parameters are not specified, i.e. the interpretation of a clause that expresses such a sentential concept still depends on information about the context, the time, and the world, which have to be provided separately. A merely world-undetermined sentential content is called a 'proposition', which is a function from possible worlds to


In the first sentence, the focus frame is the thematic dimension of swimming, and within this dimension two facts are in focus by referring to speed and to style and by comparing John and Paul in these respects. In the second sentence the focus frame is speed, and within this thematic dimension John's and Paul's swimming can be considered, but style does not fit in this dimension: there is a speed of swimming but no 'speed of style'. The following semantic operations correspond with the three syntactic structures listed above:

305

Let us compare how the boys do at swimming. With respect to style, John is best, with respect to speed, Paul. The second sentence is of semantic type ; it has to be interpreted with respect to the contextually given thematic dimension AT SWIMMING. The ad-sententials with respect to style and with respect to speed must be of semantic type < c , « c , t > , t » , and of denotation type « c , t > , t > at the contextually provided thematic dimension. In Montague Grammar, (modal) ad-sententials have the syntactic category t/t, and in the translation into intensional logic the translation of the sentence d within the scope of the ad-sententials have the syntactic category t/t, and in the translation into intensional logic the translation of the sentence d within the scope of the


truth values, i.e. of type <w,t>: A sentence expresses a proposition, i.e. its semantic type is <w,t>, and it denotes (with respect to a world) a truth value, i.e. its denotation type there is t. Likewise we can speak of time-undetermined or context-undetermined sentential contents, which are of type and , respectively. Only when context- and time-specifications are provided, the truth of sentential contents can be considered and can be judged with respect to the thematic interest involved, and with respect to time. The relationship between the triple and the truth value can be viewed in different functional ways, for example as < c , < z , < w , t > » . This type determines that the interpretation has to be fixed first according to the context, then according to the time, and finally according to the world to which the sentence is related. In what follows I shall ignore time and world in order to avoid denotation types that are too long for easy writing and reading. We can suppose that z and w have been fixed once and for all; then only the context c is of interest. And among the context properties, I merely focus on thematic dimensions. A context-undetermined sentential content is then a function of type,i.e. it maps contexts on truth values. Its denotation with respect to a fully precise context is of type t. A context is fully precise with respect to a sentence, if and only if it gives all the information necessary for determining the reference of pronouns and for determining properties as the denotations of the predicates of the sentence. Here, we only deal with full precision of a context with respect to the assignment of properties to predicates, i.e. with full precision of thematic dimensions with respect to pre-properties. A context-specifying ad-sentential denotes a function of type « c , t > , t > , i.e. it maps, at a certain context index, a context-undetermined sentence content on a contextually (here: dimensionally) determined one, i.e. a stable sentence content, which amounts, in our simplified semantics (without time and world indices), just to the truth value at this context index. Many thematic ad-sententials are not fully precise by themselves. Either a thematic dimension from the context or thematic modifiers within the clause contribute to precision. Consider the following text.

306

2.1. The adsentential use of thematic dimension indicators Examples of the ad-sentential use of dimension-indicating expressions are: As far as health is concerned, John is doing well. Healthwise and financially it holds: John is doing well. As far as length is concerned, this board is sufficient; but not with respect to width. With respect to length, John exceeds Peter. Lengthwise, John exceeds Peter. The adsententials have the semantic type < c , « c , t > , t » and the denotation type « c , t > , t > , and the embedded sentence has its semantic type as denotation type. The predications is doing well, is sufficient, exceeds Peter have the semantic type < c , < e , t » and the denotation type <e,t>. In the above examples, the proper names can be of type e ('entity1), but since they can be replaced by noun phrases like a man, every man, no man, etc. we use the type of generalized quantifiers for all (quantified) noun phrases or terms, as is done in Montague grammar: nominal terms denote functions that map properties onto truth values.6 They are thus two type levels higher than the basic entity level. But in the above examples, they are not sets of full-blown properties, but rather sets of pre-properties, which, when predicated of an object, do not make up a full sentence content, but a contextually undeter-


translation of the adsentential g appears under the intensionality sign: g'(Ad'). This means that the corresponding semantic type of g is < s , « s , t > , t » and the denotation type at a world index is « s , t > , t > . This corresponds to the denotation type assignment « c , t > , t > for thematic ad-sententials. The difference between possible worlds and thematic dimensions (and contexts generally) is that the latter are constructed and introduced by the text under consideration, while possible worlds are, once and for all fixed, in the model. The semantic type of all thematic ad-sententials is taken to be < c , « c , t > , t » , and the fully precise ones are special cases among these, namely those that, when applied to a sentential content provide a stable sentential content. A thematic ad-sentential is fully precise relative to a clause if it maps an unstable sentential content (which is the meaning of the clause) on a stable one, i.e. if the sentential content merely lacks that extra thematic information in order to become stable. In this case a contextually provided thematic dimension (i.e. the context index) has no influence on the interpretation. A thematic adsentential is precise relative to a clause if and only if it maps the meaning of the clause on a function that yields a stable result (i.e. a truth value) when applied to the contextually provided thematic dimension In this paper I shall skip the whole component of translation into intensional logic.5 I assign semantic types directly to natural language expressions, depending on their syntactic surroundings in the respective sentence.

307

F' &< G> = ,

An example is: As far as health and financial affairs are concerned: John is doing well. This sentence is of the form: (healthwise &. financially)(John is doing well) or more explicitly: ((healthAS8t financialAS)AS(John is doing well)s)s According to the above definition of conjunction, this has the same interpretation as: Healthwise, John is doing well, and financially, John is doing well. Formally, i.e. qua type, the conjunction of two dimensions or perspectives is always possible, but contentwise there are restrictions: Two dimensions or perspectives/and g can be conjoined if and only if there is a pre-property P such that P(f) and P(g) are two compatible properties, i.e. if there is an entity of which both are possibly true.


mined sentence content, which acquires a truth value only when it becomes related to a thematic dimension and/ or a perspective. The semantic type for nominal terms therefore is in these sentences < c , « c , < e , t » , t » , which means that their denotation type is « c , < e , t » , t > . The basic type for thematic dimensions/, g, etc. is c. Thematic ad-sententials F1 and G1 express thematic dimensions on a level that is two steps higher7 than the basic type, namely of denotation type « c , t > , t > . A conjunction of thematic ad-sententials is formally possible: if a function is of a type <x,t>, where x may be any type, a conjunction can be defined on the basis of the conjunction defined for truth-value denoting expressions. If p is a sentence of type and F^ and G* are thematic ad-sententials, then F'(p) and Gl(p) are sentences, and also their conjunction is a sentence of denotation type t. On the basis of such a conjunction we can define the conjunction of thematic ad-sententials by abstraction from the specific sentence p:

308 2.2. The adverbial use of dimension indicators Examples are: John is fine healthwise. John does well in his business and financially. The beam is 2 meters long. John exceeds Paul in length, not in weight. John exceeds Paul lengthwise and in width.


The combination of a dimension adverbial with a dimensionally weakly determined or undetermined adjective results into a dimensionally stronger determined, or even fully determined, adjective. On the semantic level a pre-property is applied as a function to a dimension, where the result is a (pre-)property. A pre-property is of semantic type < c , < e , t » . We can take the dimension indicator long in "The beam is 3 meters long" as of type c, the basic type for dimensions, and construct the types of the dimension-indicating expressions in length or with respect to length as of a type that is two levels higher, a type we need anyhow for forming the conjunction, like "in length and with respect to weight" or "in length but not in weight". The semantic type for adverbial dimension indicators is < c , « c , < e , t » , < e , t > » , and its denotation type at a certain thematic dimension is « c , < e , t » , < e , t » . This expresses that, with respect to a contextually given thematic dimension, they are functions from pre-properties to (pre-)properties. Thematic adverbials, as well as thematic ad-sententials are thematically opaque operators, i.e. an expression under their scope gets as denotation type its semantic type, i.e. if an expression with denotation type x is used under such an operator, it is lifted to denotation type . Recent work about type shifting8 investigates when and under what conditions it is necessary to assume different semantic types for a syntactic construction or phrase in different grammatical combinations and contexts. The simplest example is that we can use the type e for a simple nominal term like John in a sentence like John drinks, but every man or John and Paul need a type that is two levels higher, namely « e , t > , t > , as in Montague Grammar. Montague assumed this higher type for all nominal terms, in order to have a one-one correspondence between syntactic categories and semantic types, i.e. in order to have a mapping from syntactic categories into types. In the literature on type shifting and flexible categorial grammar in general, it has been shown, that we cannot assign one single semantic type, namely the highest necessary, to a single syntactic category in all its contexts. Some inferences are only available on a lower type level, and others are only available on a higher level. And this may apply even to one and the same grammatical category (cf. Groenendijk and Stokhof (1986)). These various type levels which may correspond to a single syntactic category are necessary

309 if we want to take into account inference properties between questions and not simply between indicative sentences. In this paper we restrict ourselves to indicative sentences, and here the lifted type « c , < e , t » , < e , t » is sufficient for thematic adverbials, next to the basic type c, which we need for building up the types, but which we do not need as a type for any linguistic expression. The conjunction between two thematic dimension adverbials is defined via the conjunction between property expressions: Let Fand G be two property expressions, i.e. of denotation type <e,t>. Their conjunction is then defined as F&2G=de/kx(F(x)&G(x)J.

F2 &3 G2 = def\F(F2(F°) &2 G2(F°)).

Here, F° is of denotation type < c , < e , t » , i.e. a pre-property, and F2 a.nd G2 are of denotation type « c , < e , t » , < e , t » . 2.3. The adterminal use of dimension indicators Examples are: John, with respect to health, is fine. John, from the financial and economical point of view, is doing well. John, as far as intelligence goes, exceeds Paul. Every man, from a moral point of view, is responsible, even if, from a psychological point of view, he is not in controle. John, as a teacher, is excellent. In Montague grammar and in the theory of generalized quantifiers, the denotation of a nominal term is a set of properties. The combination of such a nominal term with a predicate means that the predicated property belongs to the set of properties which is the denotation of the nominal term. In our examples, no property is predicated, but a pre-property. Accordingly, the whole nominal term, including the ad-terminal dimension indicator, has as its denotation at a certain context index, i.e. at a certain contextually provided thematic dimension, a set of pre-properties, to which the predicated pre-property belongs in case the sentence is true at that context index. Therefore, as expressed in the formula below, the denotation of the ad-terminal dimension indicator is a function (F3) which maps a set of pre-properties (A), including stable ones, i.e. properties, on a set of pre-properties (A"), namely precisely those pre-properties (P°) which, when applied to the indicated dimension or perspective (f) have (pre-)properties (P) as the result which,


And the conjunction between two thematic adverbials F1 and G1 is:

310 when applied to the indicated dimension or perspective (f), have (pre-)properties (P) as a result which are elements of the denotation (A) of the nominal term without the dimension indicator. Let A be the set of (pre-)properties, which is the denotation of a nominal term a. Then A is of denotation type « c , < e , t » , t > . Let F3 be the denotation of an ad-terminal dimension indicator, which is of denotation type « c , « c , < e , t » , t » , « c , < e , t » , t » , i.e. a function of sets of preproperties on sets of pre-properties. The denotation A'of the whole nominal term, let us call it 'perspective- or dimension-term', is a set of pre-properties, i.e. of type « c , < e , t » , t > which is a selection from the first. According to the dimension indicator, this set A' contains merely those pre-properties which belong to the indicated dimension or perspective. This is expressed by: Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

F> (A) = A\ and A' = {P'| there is P with P'(f)=Pand A(P)=1}

Such a set of pre-properties is a partial object, namely an object viewed under a certain perspective, or in other words, under consideration as far as it is determinable within a certain dimension. This is comparable with the partial objects Evening Star and Morning Star, which are Venus as it appears in the light of its Morning Star properties, and Venus as it appears in the light of its Evening Star properties, respectively. There is a whole class of nouns that can have such partial objects as denotations. These are the 'nomina agentis', like eater, drinker, runner, speaker, teacher etc. A dimension or perspective of possible predications is included in these denotations. This means the nominal terms constructed from these nouns, like a teacher, every drinker, etc. are essentially of type « c , < e , t » , t > . I call these nouns 'perspective- or dimension nouns'. The typical predications are pre-properties which provide specifications within the perspectives and dimensions which are given by the respective verbs teach, drink etc. This means that the pre-properties select (pre-)properties from these dimensions, as in "This teacher is good" or "This runner is slow" or "John, as a teacher, is good but, but as a runner, bad".9 It is possible that stable pre-properties, i.e. properties, are selected (or constructed) in more than one step by interation of thematic dimensions, as e.g. in the sentence "As far as keeping order in class is concerned, this teacher is good". Also complex nouns like size of John, width of this car, health of Fred belong to the class of perspective- or dimension nouns. This means that the dimension indicators size, width, health, etc. are relative dimension- and perspective-nouns. These nouns themselves can be modified ad-nominally, like in good health, extreme size, etc., and likewise good teacher, slow runner, moderate drinker, bad dancer, inspiring speaker. The 'nomina agentis' are basically of denotation type <e,t>, as in "John is a teacher'. But there is also a type < e , « c , < e , t » , t » , which is the lowest denotation type of phrases like as a teacher. Their denotations map individuals (or two levels higher: sets of (pre-)properties) on sets of pre-properties, as in John, as a teacher. Such a

311

For all X: X(GOOD TEACHER) = 1 if and only if X (TEACHER) = 1 AS A TEACHER(GOOD) = 1. Hereby, X AS A TEACHER = {P' I there

and X are P

with P'(d TEACH ER)=P and X(P)= 1}. The property GOOD TEACHER thus is a conjunction of the property TEACHER and the property GOOD AS A TEACHER, which is expressed by saying that the pre-property GOOD belongs to the set of pre-properties which have as values properties of x under the teacher perspective, x AS A TEACHER is, according to this semantics, a partial object, like the denotation of "John, from a financial point of view" and others listed above. The dimension and perspective indicator as a teacher can also be used adverbially, as in "John is good as a teacher". This does not mean, that John is a teacher; it merely indicates the dimension or perspective under which the pre-property GOOD is predicated. This dimension is designated by the index d TEACHER in the semantic model, whereby the adverbial is, as explained above, of a denotational type which is two levels higher. It is evident10 that in this way the sentence "John, as a teacher, is good" and "John is good as a teacher" (and also "Under the perspective of teaching, John is good"), though of different syntactic form, receive the same interpretation. The conjunction of thematic dimension ad-terminals takes place on the type level « c < c , < e , t > , t » , « c , < e , t » , t » . For the sentences John, from the financial and economical point of view, is successful. John, as a teacher and director, acts responsibly.


set consists of all pre-properties which assign, in the perspective or dimension expressed by the nomen (agentis), properties of an individual. For the noun teacher these are all the teacher properties of the individual, i.e. all his properties which are comprised by the dimensions and perspectives of an individual's teaching. The ad-nominal adjectives are of the type « c , < e , t » , < e , t » , as in blond teacher, or in good teacher. They map the dimension-noun onto a noun of property-type, namely <e,t> at a context-index. When we consider iteration of thematic dimensions in a sentence, further specification is possible as in "good teacher as far as keeping order in class is concerned": here good teacher has its semantic type < c , < e , t » as its denotation type under the scope of the ad-terminal operator "as far as keeping order in class is concerned." The semantics of ad-nominals which are based on pre-properties is derivable from their predicative use in accordance with the following meaning postulate: somebody is a good teacher if and only if he is a teacher, and as a teacher he is good. We can represent the conjunction by treating individuals on the generalized quantifier level: to an individual variable x of denotation type e there corresponds, two levels higher, a variable X which has sets of properties as its denotation type.

312

we have the following forms: (from the financial and economical point of view (John)) (is successfill) (as a teacher and as a director (John)) (acts responsibly) The conjunction of ad-terminal dimension indicators is defined on the basis of the conjunction of the dimension and perspective terms. This conjunction is for two such terms A" and Y\ X- &4 Y'=defkP°(X' (P°) & Y' (P0)), where P° is a variable for expressions of pre-properties.

The length of Fred and the width of Fred is enormous. The conjunction of two ad-terminal dimension indicators F3 and G3 is then: F3&.5G3=defkX

(F3(X) &4 G3(X)).

Here, A'is a variable for terms and the complex expressions F3(X)and G3(X) are dimension or perspective terms. In the above form ((as a teacher and as a director (John)) (acts responsibly)), the predication acts responsibly is a pre-property. In the grammar presented here, this sentence has the same interpretation as the sentence John as a teacher and John as a director acts responsibly. The property which is assigned by the pre-property in the dimension dTeacher Bnd Dinaor IS t n e property TO ACT RESPONSIBLY AS A TEACHER AND AS A DIRECTOR. In the same way the sentence "The length and the width of John are enormous" can be treated. If a conjunction of pre-properties and properties is predicated, we have to assign the higher type to the conjunction, i.e. the one of pre-properties, namely < c , < e , t » , which means that all adjectives have to be treated as pre-properties, the properties just being stable pre-properties. From this it follows that the term that is constructed from the nomen agentis, e.g. the runner, has to be treated as a set of pre-properties. Although it is possible to define a different type for runner in blond runner and in fast runner, and likewise for the nominal term in the runner is blond, and in the runner is fast, we take the highest type that takes care of all combinations and interpret blond]ust as stable with respect to the thematic dimension IN RUNNING: the stable pre-property BLOND just yields the property, i.e. itself, when applied to the dimension of 'run'-properties. It might count against this option that sentences like "This runner is fast and blond" appear somewhat strange" because a dimensional specification seems to be forced upon a property that has nothing to do with it. But this dimensional specification has no effect,


An example of such a conjunction is:

313 because the property BLOND is stable. Note that "This runner is fast and weighs 250 pounds" is not strange at all, probably because weight is relevant to the possibility of runners being fast, and here the high weight makes it surprising that the runner is fast. The sentence would even be better with an adversative conjunction: This runner is fast, but weighs 250 pounds. We can also form the conjunction between John and this runner as in sentences like "John and this runner are both slow". The two nominal terms are of the same type. This is equivalent to "John is slow and this runner is slow", where the first conjunct is not yet stable but needs a specification of the thematic dimension and the second conjunct is stable because the thematic dimension is specified in the meaning of the nominal term.

The iteration of the application of pre-properties to dimensions, where the result of one application is again a pre-property that can be applied to another dimension to yield a (stable) pre-property, is the basis of propertyconstruction. If such a process of repeated application results in a stable pre-property, we have constructed a property. In texts this iteration is reflected in the repeated application of dimension indicators. Examples of iterations of thematic dimension indicators are: With respect to speed John is excellent at swimming. As far as swimming is concerned, John is excellent with regard to speed. Regarding the speed of swimming, John is excellent. As far as swimming is concerned John's style is excellent. As a teacher John is excellent, as far as mathematics is concerned. The type assignment discussed in the earlier sections allows for these iterations: if x is a denotation type of an expression then a thematic dimension modifier on this expression is of denotation type « c , x > , x > , i.e. the semantic type of an expression is taken as its denotation type under a thematic dimension modifier. In principle there is no limit on the number of these iterations, but once the relevant pre-property becomes stable, iterations of these kinds of modifiers are semantically vacuous. Therefore the sentence As far as style is concerned, John's speed of swimming is excellent. just amounts to John's speed of swimming is excellent. The meaning of this sentence is stable with respect to any dimension. But the first sentence is unacceptable because it does not make sense to indicate


2.4. Iteration of thematic dimension indicators

314 explicitly that a stable expression should be interpreted relative to a thematic dimension. With respect to style, John is excellent in swimming. is acceptable, because excellent in swimming has as its meaning an unstable pre-property which can be applied with respect to the thematic dimension STYLE resulting in the stable (pre-)property EXCELLENT AT SWIMMING WITH RESPECT TO STYLE.


Let us consider what happens in iteration semantically. Thematic dimensions are sets of (pre-)properties, including stable ones, i.e. properties. This means that they are basically second order properties, like COLOUR or GEOMETRIC SHAPE. In iteration, the second order properties are put in conjunctions as conditions on the properties that can be the interpretation of an expression of a pre-property, e.g. an adjective. The property P assigned by the pre-property P° with respect to the iteration XY of dimensions X and Y has to be an X-property and a Y-property. Thus, excellent in the above sentence assigns a property to John which is a property of swimming, i.e. a swiM-property, and a property of style, i.e. a STYLE-property. EXCELLENT IN STYLE is a pre-property, which can be applied to all or some activities, and so "John is excellent with regard to style" means vaguely that his style is excellent in several kinds of activities, if it is not interpreted relative to a contextually given kind of activity. But in order to make a precise statement it must be applied to specific thematic dimensions, for example to swimming, or to a particular selection of activities. Let us consider the above examples in detail: within the set of swiM-properties a person can have proporties concerning the style of swimming, the speed of swimming, the endurance in swimming, etc. Likewise, within the set of STYLE-properties a person can have properties of style of clothing, style of talking, style of walking, style of swimming, etc. The pre-property EXCELLENT designates a class of properties on the swiM-dimension which all contain EXCELLENT as a pre-property. It is an equivalence class within the class of swiM-properties. It consists of all those properties that express excellence in the respective subdimensions STYLE OF SWIMMING, SPEED OF SWIMMING, ENDURANCE OF SWIMMING. A pre-property applied to a dimension designates and equivalence class in this dimension. This equivalence class here is the pre-property EXCELLENT AT SWIMMING. Likewise, we form the equivalence class EXCELLENT IN STYLE on the style-dimension. This equivalence class consists of the properties that express excellence in the respective subdimensions STYLE OF WALKING, STYLE OF TALKING, STYLE OF SWIMMING, etc. If we now apply the pre-property EXCELLENT AT SWIMMING to the thematic dimension STYLE, it picks out a property of excellence from the subdimension STYLE OF SWIMMING. What actually happens is that we form the intersection between the equivalence class EXCELLENT AT SWIMMING from the swim-dimension and the equivalence class EXCELLENT IN STYLE from the STYLE-dimen-

315 sion. This intersection is a class which is the pre-property EXCELLENT IN STYLE OF SWIMMING. It contains just one element which means that it contains the property EXCELLENT IN STYLE OF SWIMMING. In Fig. 4 we see that the dimension SWIMMING is the union of properties from the subdimensions STYLE OF SWIMMING, SPEED OF SWIMMING, etc. For a

swimming

style

walking

speed

talking

endurance

Figure 4: Iteration of thematic dimensions

swimming


dimension dj to be determined further by a perspective or dimension dj means that d; is mapped on a subset (subdimension) dy. The same is possible the other way around, i.e. that d= is determined by dj. We can also think of the pre-property EXCELLENT IN STYLE OF SWIMMING as containing a whole set of properties ordered in subdimensions according to different kinds of swimming. Then we can have a third iteration, for example EXCELLENT IN STYLE OF SWIMMING AS FAR AS BUTTERFLY STROKE is CONCERNED. Considering this third step of refinement of precision of the thematic dimension, we must say that the pre-property expression excellent in style of swimming designates the property EXCELLENT IN STYLE OF SWIMMING WITH REGARD TO SEVERAL KINDS OF SWIM STROKES, if it is not used with respect to a particular kind of swim stroke referred to in the context or in the situation. In terms of the composition of a property, the procedure just described amounts to a composition of properties from pre-properties:

316 Let d;: = SWIMMING, d2: = STYLE, P°: = EXCELLENT. We then have two constructed pre-properties P ' = P 0 (dj)= EXCELLENT (SWIMMING), and P" = P°(d2) = EXCELLENT (sTYLE).There is a combination '@' of two pre-properties into a single pre-property, P ' " = P ' @ P", which is the intersection of both. If this intersection is a singleton, i.e. PT = {P}, we say that the property P is combined out of the pre-properties P' and P". In our example, this property is P = EXCELLENT (STYLE + SWIMMING), which is expressed by either excellent in style of swimming or by excellent in swimming as far as style is concerned. This can be expressed by the following Schema of Property Construction:

If this construction results in a set with exactly one element, we have constructed a stable pre-property, namely a property. In our example this set (i.e. the resulting pre-property) just contains a single P (i.e. the property) where {P} = P"(d,) = P'(di)- In this case P is constructed from the pre-properties P' and P", which are constructed from the two dimensions d{ and dj and the pre-property P°. In the light of this construction of (pre-)properties from pre-properties and thematic dimensions, with or without iteration it follows that the operators in iteration are commutative, as is the combination ' + ' between dimensions themselves: If P° is a pre-property, then the dimension operators D'and D"and their semantic bases, the dimensions d' and d", satisfy the formulas: D'(D"(P°)) = D"(D'(P°)) on the higher type level, and P°(d' + d") = P°(d" + d') on the basic type level. Because of this, we can apply thematic dimensions in different, and as far as semantics is concerned, in arbitrary order. Order is, relevant of course, in the course of text construction, which requires a build-up of dimensions such that the thematic dimension given in the context is presupposed and not expressed again in the sentence that stands in this context.

3. CONCLUSIONS

This paper has investigated the context dependent type assignment for adjectives on the one hand, and dimension and perspective indicating expressions on the other. The type assignments to these kinds of expressions had consequences in the area of nominal terms, where a new type was introduced,


P°(di) P«(dj) = P(d, + dp {P| there are P \ P" with P ' = P°(d,) & P " = P°(d2) & Pe P"(d,) & P« P'(d2)}

317


the type of the dimension-and perspective-terms. Special attention was paid to combinations between thematic dimension expressions, those where a combination results in two properties (i.e. conjunctions of thematic dimensions) and those in which the combination results in one single property (i.e. iterations of thematic dimensions). For the treatment of these combinations, operations had to be performed on higher type levels, especially on the level of second order properties. Pre-properties are sets of properties. For type-shifting the difference between properties and pre-properties has complicating consequences in the area of type-assignment to other categories of expressions. Therefore, I have treated properties as stable pre-properties. This amounts to lifting a property to the level of a singleton, i.e. a set consisting of just this one property. Properties and pre-properties are expressed by the same class of expressions, namely adjectives (and some verbs of comparison), which can be used in the predicative and in the ad nominal category. Without this general lifting of properties to pre-properties, we would have two categories and four semantic types for one class of expressions, namely two predicative and two adnominal types, which are based either on the type for properties or on the type for pre-properties. At the same time there would be type differences within the category of nouns: the 'nomina agentis' require a more complex type than other nouns that basically denote sets of individuals. Likewise, there are type differences between nominal terms (sets of properties) and dimension and perspective terms (sets of pre-properties). Here we would have, for one category, two different types which are not distinguished by a two level difference, but by their kind of context dependence. This is avoided by lifting properties to pre-properties. The dimension and perspective indicators appear each in three categories of use, in the ad-sentential, adverbial, and ad-terminal category. The relationship between these became apparent on the semantic level, where the alternative constructions result in the same interpretations. Another interesting aspect is that pre-properties, like GOOD, SUCCESSFUL, ENORMOUS are of a relational basic type, namely <e,c,t > . Pre-properties are, apparently, functions from pairs consisting of an entity and a dimension (and/or perspective) onto truth values. This means they are relationships between entities and thematic dimensions (including: perspectives, themes). This reflects the fact that properties of individuals get constituted in combining perspectival and thematic information in the light of which the individual is viewed when very general pre-properties are predicated of it, which are the result of comparing individuals to each other under the perspectives given, and then abstracting from these dimensions. This construction of properties is different from the kind of construction which consists in forming conjunctions of properties on the first order level. There, intersections between sets consisting of zero-order entities are formed, i.e. operations are formed on first-order sets. But the construction considered here involves second-order conjunction, as has been spelled out in detail in the section on

318

University of Amsterdam Department of Philosophy Grimburgwal 10. geb. 13 1012 GA Amsterdam The Netherlands


iteration. Here, intersections between sets of first-order entities are formed, i.e. the operations are performed on second-order sets. Both constructions are formally of the same kind, but they take place on different levels. The construction treated in this paper has not been taken into account formally in lexical decomposition up to now. But this can be done along the lines of construction spelled out in Section 2.4. It is very important to be aware of this kind of construction of concepts in order to understand how predication and the construction of properties works, and hence how 'interpretation' is achieved in texts. 'Parts' of a property can be spread over a text as a distribution of thematic dimension indicators and a pre-property expression, or it can be spread partly over a conceptualized, i.e. 'interpreted' situation, if the described situation is present at the speech situation. It is obvious that the dependence of the interpretation of pre-property expressions (especially adjectives) on thematic information is one of the important connections that establish text coherence. Texts express properties and provide for their construction by specifying thematic dimensions and pre-properties. Generally, properties are expressed by texts rather than simply by adjectives, except for a small class of dimensionally fully determined adjectives that express properties directly. Pre-properties are basically relations which are predicated of pairs of entities and thematic dimensions. This is what it all amounts to in the end. But assuming this as a grammatically adequate view, keeps out of sight why there are grammatical differences between ad-sententials, adverbials, and ad-terminals. To these grammatical distinctions correspond different routes of interpretation which yield a difference in the result of the interpretation once we deal with scope relevant phenomena, like quantifiers, time, modality, and negation. Scope phenomena require a functional construction instead of a relational one, at least if always the same basic subject-predicate sentence pattern is used, as in the cases from English we have considered: a relation is split up into steps of successive applications of functional expressions and these steps build up complexes the meaning of which can be referred to anaphorically, can be in the scope of several modal, time, and negation operators, and can play a role in the topic-focus division of a sentence.

319 NOTES

REFERENCES Bartsch, R. 1972: Adverbialsemanttk. Frankfurt a. M.: AthenSum Verlag. English version: The Grammar of Adverbials. North-Holland, Amsterdam, 1976. Bartsch, R. 1986: Context-dependent Interpretations of Lexical Items. In: J. Groenendijk, D. de Jongh, and M. Stokhof (eds.), Foundations of Pragmatics and Lexical Semantics (Papers of the 5th Amsterdam Colloquium 1984) GRASS-Series 7. Foris, Dordrecht. Pp. 1-26. Ehnch, V. 1975: Pragmatische Restnktionen der Bedeutung von relativcn Adjektiven und Vergleichssatzen. In: V. Ehrich and P. Finke (eds.), Beura'ge zur Grammatik und Pragmattk. Scriptor, Kronberg. Pp. 141-154. Groenendijk, J. and M. Stokhof. 1986: Type-Shifting and the Semantics of lnterrogatives. ITLI/Department of Philosophy, University of Amsterdam. Kaplan, D. 1978: The Logic of Demonstratives. Journal of Philosophical Logic 8: 81-98. Montague, R. 1974: Formal Philosophy. Selected Papers of Richard Montague. Edited by R. Thomason. Yale University Press, New Haven.


1. Kaplan mainly deals with pronouns and other indexical expressions, which express functions from contexts to entities, but he introduces the notion 'character' in a general way as function from contexts to contents (i.e. from contexts to intensions). Those characters that give the same value for all contexts, i e. are independent from any particular context, are called stable. 2. Compare the class of predicate-limiting adverbials in Bartsch (1972/76) and the treatment of dimension indicators in a Montague-Kaplan grammar in Bartsch (1986). 3. Pinkal( 1981) has treated the semantics of this kind of vagueness and introduced the notion of precisification, which amounts to specifying the dimension until it is fully precise. The vague predicate interpreted with respect to the fully precise dimension denotes a property, i.e. the sentence is true or false with respect to this precisification. With this notion in mind, we can say that the sentence is weakly true if there is at least one relevant precisification under which the sentence is true, and the sentence is strongly true if it is true under all relevant precisifications. And there are degrees to which a sentence can be true according to under how many relevant precisifications the sentence is true. 4. 1 disregard place in this paper. 5. It has been worked out for the kind of constructions we are concerned with in Bartsch (1986), where the syntactic rules, the translation into (intensional) logic, and the interpretation in a model are formulated. 6. Another way of expressing this is: terms designate sets of properties, i.e. are of type « e , t > , t > , or, with intensionality included, « s , « s , e > , t » , t > . 7. 1 use superscripts to distinguish expressions of different type, but the number used as a superscript does not really correspond with the height of the level referred to. Properties have no superscript, and pre-properties have superscript 0. Correspondingly the conjunction is of different types, which is indicated by a superscript. 8. Partee and Rooth (1983), Partee (1986), and Groendijk and Stokhof (1986). 9. These nouns are treated in Bartsch (1986) in a Montague-Kaplan grammar. 10 See for details Bartsch (1986), where also terms like the length of John are treated as dimension- or perspective-terms, as in sentences like The length of John is enormous, where enormous denotes a pre-property. The interpretation of this sentence has the same result as the one of the sentence John is enormously long or John is enormous in length. 11. This sentence with but is not strange at all in the following context: We are looking for somebody who is a fast runner and who is dark. Then somebody points at a runner, but concludes: This runner is fast, but blond.

320 Partee, B. 1986: Noun Phrase Interpretation and Type-Shifting Principles. In: J. Groenendijk, D. de Jongh. and M. Stokhof (eds.). Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers. GRASS-Series 8. Foris, Dordrecht. Pp. 272-299. Partee, B. and M. Rooth. 1983: Generalized Conjunction and Type Ambiguity. In: R. BSuerle, Ch. Schwarze, and A. von Stechow (eds.), Meaning, Use, and Interpretation of Language. De Gruyter, Berlin. Pp. 361-383. Pinkal, M. 1981: Semantische Vagheit. Phanomene und Theorien. Linguistische Benchte 70: 1-25, and 72: 1-26.



PROCESSUAL SEMANTICS OF THE VERB WOLFGANG WILDGEN

ABSTRACT

a) b) c)

Verbs referring to bodily motions Verbs referring to actions (in space and time) controlled by one agent Verbs referring to interactions (in space and time) between different agents.

Making use of major results of ecological psychology, we show that the dynamics of basic motions and actions (including their perception and control) can essentially be described with reference to fundamental physical systems (the simple and the double pendulum). Our main point is that those processual schemata relevant for the control and perception of bodily motions and actions are also cognitively and semantically basic. We claim that these schemata can describe and explain the geometrico-dynamic component in the meaning of verbs of motion and bodily action. These hypotheses are tested in the analysis of several groups of verbs, which are prototypical for the domain in question. When we add further classificatory schemes (e.g. auditory ones) and the internal evaluation of the process (e.g. using Osgood's semantic space), we can approximate the mental organization of meaning. The neurolinguistic aspects of our issue are discussed in the last section of the paper.

1. THE PROCESSUAL POINT OF VIEW

The theoretical perspective chosen in this article can be called processual or dynamic (with reference to dynamic systems). The scientific strategy in this perspective states that we must look for fundamental principles or natural laws in those natural domains underlying the phenomena of reference and meaning in order to get a semantic model which is not descriptively ad hoc and thus superficial. In a theory of reference the basic domain is that of "real" processes (the mechanism of the world in a Cartesian view); in a theory of meaning (intensions) it is the brain (functionally embedded in the


Processual semantics opens a new field of research by looking at basic physical, organismic, and cognitive processes involved in meaning. Its strategy of model-building is directed towards the application of the theory of dynamic systems. We consider three rather specific domains of the lexicon of verbs (using the classification of Ballmer and Brennenstuhl as an approximation of our descriptive goal):

322

2. LEXICAL SEMANTICS: INTRODUCTORY REMARKS

The linguistic sign has in terms of Peirce's semiotics three distinct levels of consideration: the form of the sign itself (the representamen), the object to which the sign refers, and the interpretant, which can be the system in which the triadic relation is realized. Simplifying the relational view of Peirce we may consider: - a linguistic form, i.e. a word (a morpheme), a sentence, a text in its appearance (e.g. as a sequence of letters in written communication, or as a quasilinear sequence of phonemes or phonetic feature-clusters), - an ontological substratum, e.g. the physico-chemical nature of an object, a process in the dimensions of space-time, - a cognitive (organismic) structure in the sign user. The cognitive structure (our specification of Peirce's interpretant) is a complex in itself, which has a place for correlates of the sign-form and of the object. A semantic theory must relate the sign-form to the object via their representations in the interpretant. This means that in this perspective we may consider three functional subdomains in the cognitive structure: a) The representation of the sign-form (or in Saussure's terms the internal image of the word), which has two aspects:


human organism). These two basic domains are linked by the periphery of the body (our senses and the limbs, perception and motor control). Although the basic processes in human semiotic activity are more complex than physical and chemical laws they are intrinsically related to these, and every semantic model which aims to be more than an ad hoc taxonomy must connect to these laws. In other words, it must be formulated in terms of the language of the theory of dynamic systems (in the tradition of Galilei, Kepler, and Newton). Secondarily, every society and every languageuser invents more specific forms and rules exploiting the domains of indeterminacy left by the underlying (natural) laws. As the traditional "models" in semantics are rather concerned with the conventional aspects of the mapping: word/sentence —• meaning, the processual point of view is rather complementary to them. However, a successful (natural) model of semantics will change the frame for the investigation of conventional "rules" and can thus be considered as a challenge to existing models in semantics.1

323

The first type of coordination plays a role in the constitution of polarities in a qualitative space and will be considered in the following. The second type leads to a specific segmentation in the field of verbs of emotion and of categories of modality. - The coordination of central processes of organization and storage of information (as a kind of cognitive meaning of stimuli). These processes are important in the domain of syntactic and narrative structure and for the analysis of lexical and encyclopedic knowledge. They will not be discussed in detail although they are of crucial importance.


- the perception of the sign-form via perceptual categories (categorical perception) - the motoric programs which can realize the sign-form. The different modes of a sign as a written sign, a spoken sign, or a gestural sign (cf. deaf-mutes) define perceptual/motoric schemes; therefore we must ask for a more general representation independent of the specific mode in the periphery of perceptual/motoric processes. In our terminology a "representation" is a thing or entity which is localized in space and time; it is a rather specific functional relation of the different kinds of phenomena in perception and motoric production. In this sense we do not subscribe to the paradigm of mentalistic constructions. b) The object, which is perceived and motorically controlled in the same manner as the sign-form. The notion of "object" refers in general to the way we deal with the "external" world; thus the object is an abstract ontological category containing the following elements: processes in space-time, phases of external systems, agents in their context of action and as a very specific case, and solid objects with a high stability of form and position. Thus the object is a static entity only under highly specific conditions. This view has the effect that our treatment is highly "processual" compared to logical worlds which are apriori stable set-theoretical constructs. c) The cognitive structure as a complex organizational scheme relating the input and the output of (a) and (b), which can be functionally conceived of as containing a mechanism which relates the processes of signing and perceiving/acting. It is clear that several perceptual modes and motoric/agentive modes have to be coordinated at a higher level. We may consider the following: - The coordination of perceptual/motoric structures with those parts of the sign-system which are directly connected with this field (which is our main concern in the following paragraphs). - The coordination of emotional processes with perceptual/motoric schemes and sign-forms.

324

A.

Verbs referring to bodily motions occurring in the immediate field of the body, i.e. in the relation of body-parts and limbs.

B.

Verbs referring to motions, actions controlled by a single agent. The difference between motion and action emerges at this level depending on the intentionality of the process. We shall try to give a first approximation of a naturalistic concept of intentionality.

C.

Verbs referring to the interaction between agents. This "interaction" can be a purely coordinated action (i.e. the actions of type B in coordination) or it can presuppose very specific scenarios of social and communicative interaction, such as speaking/listening.

As semantics relates the domains of perception/control and the domain of signs, the interactive scheme of sign-using is a general presupposition, a dynamic apriori of all semantic schemes at the levels of A, B and C. It is a higher control scheme serving as an orientation to all lower schemes. The strategy of the following analysis will contain three stages: a)

b) c)

Seek a very fundamental spatio-temporal system which could underlie the motion or action. We consider simple mechanical models as good hypotheses for such basic schemes. However, our analysis is not reductionist. On the contrary, the physical model is a kind of semiotic approximation to the basic phenomena underlying our world, it constitutes a basic semiotic system. Specify the basic perceptual and motor schemes underlying a class of events and actions. Describe the content of a class of verbs using the schemes found in (a) and (b). Such a semantic description of subclasses in the lexicon of verbs aims at


In general our concern will rather be the basic lexicon. As the lexicon of concrete nouns (cf. Labov, 1978) and of adjectives (e.g. the basic lexicon of color-terms; cf. Rosch e.a., 1978) has already received some attention, we shall concentrate on the basic lexicon of verbs (verbs with a concrete meaning, which can be explained by concrete processes observable in spacetime). As Ballmer & Brennenstuhl (1986) have analyzed the lexicon of German verbs, we shall take their results as a preliminary hypothesis and elaborate further on the theoretical background which can explain such an ontologico-dynamic classification of the verb lexicon. Based on the dimension called "Eingriffsgrad", i.e. degree of control (by a human agent), we propose a first rather coarse subdivision into three domains:

325

3. THE PROCESSUAL SEMANTICS OF THREE GROUPS OF VERBS

3.1. Processes underlying the verbs of bodily motion The psychophysics of the perception and control of simple motion is a central topic in psychology. In the field of concurrent theories we chose those which have a clear psychophysical orientation. They stand in the tradition of the work of N.A. Bernstein (1967) and Gibson (1968). We used the more recent results of Turvey (1977), Kugler, Kelso and Turvey (1980), and Kelso, Holt, Kugler and Turvey (1980). The central notion for the following discussions is the notion of "control" and (later) of "coordination". Movements of living bodies and body-parts are subject to two types of control: (a)

(b)

The nonlinear control of movements, which is largely independent of specific contextual factors and defines the general teleonomic nature of a movement. Non-linear controls involve catastrophes, i.e. sudden changes in the evolution of a process (cf. Casti, 1985, Ch. 1 for the concept of non-linearity). The linear control adapts the movement in its metrical detail to specific contextual features, it "tunes" the qualitative motion-scheme.

If we consider simple movements with one or two limbs and look for analogies in physical mechanics, we find the elastic pendulum and the double pendulum. Fig. 1 shows the analogy between the motion of an elastic pendulum (left) and the movement of a limb (right) between its position of


establishing invariants relating the ontology (as a basis of perception and motion), the categories of perception and the schemes of motion, locomotion, and action, and finally the intrinsic structure of the lexicon. Only such descriptions containing invariants of the three domains (sign-form, object, interpretant) can be potential elements of a realistic semantic model. This basic position defines a new paradigm in semantics which we call "processuaP'. This paradigm is opposed to "internal semantics" (exploiting only language-specific features such as distributions, sequential correlations, commutations and replacements) and to set-theoretical semantics, which interprets sign-forms in relation to set-theoretical constructs called "world" or "possible world" (cf. Montague, 1970; Cresswell, 1973). Processual semantics takes up ideas developed mainly by Rend Thorn in the sixties (cf. Thorn 1974) and Wildgen (1982, 1985). Similar approaches are Fillmore's (1977) frames-and-scenes semantics and Ballmer's (1982) ontologicdynamic semantics.

326 flexion

mass-soring system

muscular system of a limb

Fig. 1 (cf. Kelso e.a. 1980)

rest and its extension versus inflection (cf. Kelso e.a. 1980). The peripheral mechanism of a muscular system controlling the movement of the limb is a damped oscillator of the kind given by the elastic pendulum. This means that higher (e.g. cerebral) controls only specify this peripheral quasi-autonomous system and do not govern it in detail. Thus the brain has to "acknowledge" the physical principles built into the peripheral systems. Fig. 2 shows the analogy between a double pendulum, whose motion is very complex (it contains chaotic phases as Richter (1985:13f) has shown). The dynamic system of the human leg is comparable to a double pendulum (strongly damped and with restricted domains of freedom). The right-hand side of Fig. 2 shows phases in the movement of the human leg while the person is walking (experimental results of Johansson, 1976:386). If a person performs a locomotion which is composed of a number of limb-motions we can distinguish two levels: (a)

(b)

The rhythm of the composed movements, which is a code for the categorical perception of moving agents as Kruse e.a. (1983) and Wehner & Stadler (1982) have shown. The overall "Gestalt" of the movement. In the case of a simple locomotion we have the initial phase which starts the locomotion. It destabilizes the system in its position of rest and creates a steady evolution until the system is at rest again.

The rhythmical differentiation is the source of a diversified lexical differentiation (the items include further information concerning the agent in locomotion). In German we find the following list of corresponding verbs, all


extension

327 ////////y

/ /

double pendulum

/

/

/ /

/

1 sec

movement of a human leg paths of the hip, knee, ankle

of which denote ways of walking (translations follow with the more detailed treatment below): bummeln, hinken, hopsen, hiipfen, humpeln, laufen, latschen, pirschen, preschen, schlurfen, schwanzeln, staken, stapfen, stelzen, stiefeln, tappen, tapsen, tigern, tippeln, trampeln, trippeln, trotten, wieseln, zockeln, zuckeln. The coarse topology of the locomotion has three phases: - loss of the position of rest, beginning of the motion, - steady motion, - gain of a new position of rest, end of the locomotion. Higher controls act like a cause of destabilisation, i.e. the initial impulse, and as a cause of stabilisation, i.e. the final stopping. If the system seeks by itself the position of rest, the control defines a new position of rest rl' relative to the position of rest rO. In process models we distinguish two spaces: the control space and the state space. The latter describes the internal stability function of the system (seeking the position of rest), the former reflects an external process which operates on the parameters of the state space. In Fig. 3 we show (on the left side) the model of Kugler e.a. (1980) for the simple movement as a re-parametrization of a system with minima (the bottom of the "funnel"). The point T is first in the minimum. As the "funnel" is re-parametrized relative to yO (the parameter of rest positions) it moves on the slope of the potential (P) curve until it reaches the new position of rest. The basic scheme of processes of locomotion is thus given by a process without catastrophe. The controls impose a beginning and an end but these singularities do not belong to the topology of the process as a type but to its instantiations. This basic scheme underlies the semantics of simple verbs of locomotion like go, run, or drive. Traditionally these have been characterized as durative.


Fig. 2

328

Y7X T -*—T" change of the control XQ to x,

mode) of Kugler e.a. (1980);

catastrophe-theoretic model based on the zerounfolding Al and a versal parameter p

We can add instabilities of a simple type to the basic scheme using two different types of information: a)

Intrinsic information contained in the background scheme: "A speaks to B", where A = speaker, and B = hearer. This scheme divides the space into the fields of A and of B, which serve as implicit frontiers. The continuous locomotion can enter the field of A or leave it. Thus we obtain: C comes (towards A = speaker) C goes (away) (away from A = speaker)

b)

Extrinsic information given in the utterance or by the context of the utterance, as in: John enters (the house) John leaves (the house)

In both cases the underlying topological scheme contains an instability of the type called birth/death in catastrophe theory. The basic dynamics is shown in Fig. 4. The specified domain (intrinsic: the domain of the speaker; extrinsic: any given domain) is a position of rest which is only reached if the boundary of the given zone is crossed. The existence of some larger domain of critical or final rest is given as a background of the processual scheme, (lower plane). The specified domain is modelled by the surface of stable minima (represented by the striped area above in Fig. 4); each point is of the "funnel" type as shown in Fig. 3. The boundary is given as a suspension of A2-points (fold). Thus the process of locomotion of a body is topologically either of the continous kind or it involves an implicit or explicit boundary, an orientation of the process. As soon as the locomotion has crossed the boundary, it can


Fig. 3

329

seek its position of rest. The introduction of this orientation gives the scheme a teleonomic value; i.e. if x^ is the position of departure, x, is not an accidental point of rest (e.g. if the organism gets tired); it is defined as being already present in the initial state and is the reason of the locomotion. Thus this feature of teleonomic movement introduces a kind of intentionality. The path from XQ to x, is not only goal-oriented (x, being the goal), it can be complicated by the introduction of intermediate agents. We find two fundamental types of intermediate agents in linguistic scenarios. (1)

Instrumental "mediators". They modify the reach of our locomotion and its energy consumption. The overall scheme remains qualitatively the same, as is the case with the traveller who wants to get to Rome and who can reach it by foot, bicycle, car, train, or plane. The type of intermediate action, however, is either of the type B: actions controlled by an agent (or a non-agent), or of type C: interaction.

(2)

Causation. Causation is a broader class of mediation which includes the control of general processes, such as physical laws and techniques exploiting this knowledge. Causality, however, is linked to spatio-temporal mediations and correlations, as the results of attributional psychology show (cf. Heider 1958 and Michotte 1952).

The cognitive schemata of the process that we have classified are not only relevant to categorial structure in the lexicon of the verb, they also form the


Fig- 4

330

cognitive basis for the grammatical schemata such as source/goal, instrumental case and path case (as discussed in case theory), and for causative constructions. The notions of durative and resultative aspects in verbsemantics make use of these schemata. The verb classes B and C (cf. Introduction), which will be treated now, presuppose the basic schemes of the movement of the body and of locomotion and add schemata of coordination and the control of these processes initiated by an agent in a large field of forces (B) and mutual (cooperative or confrontational) actions (C). 3.2. Processes underlying the verbs of action by one agent

(a)

(b)

The configurational aspect. This aspect is only concerned with the spatio-temporal relationship, the topologico-dynamic "connectivity" in a scene. This aspect naturally occurs with a detached observer of a scene (a strict behaviorist). The energetic or intentional aspect. Here the forces controlling the process, the irreversible path of an effect by an agent, are considered. A natural realization of this aspect is the self-referential observation the agent makes of himself. In a wider sense one can attribute similar intentions to other agents or even to natural forces (as in natural religions).

In our treatment we can exemplify both aspects. As psychophysical analyses of complex processes are not available we shall resort to the dynamic schemes proposed in catastrophe theoretic semantics on the one hand, and to the inductive classification of the German lexicon of verbs by Ballmer & Brennenstuhl (1986) on the other. The convergence of these two lines of research allows us first hypotheses on the natural basis of semantic schematizations in the lexicon. Ballmer & Brennenstuhl (1986) distinguish two main groups of verbs at this level of control (their axis being called "Eingriffsgrad"): (1) (2)

verbs referring to the creation, destruction, and regeneration of entities (pieces of the environment), verbs referring to the effect of an agent on the state of entities in his environment.


In the situation where one agent (a higher organized body with its periphery) acts on an entity which has a lesser degree of agency (matter, solid objects, living beings dominated by the agent) and which is not under its central control per se, i.e. which has normally a certain degree of dynamic and topological independence of the agent, we must distinguish two aspects under which the process can be considered:

331

Fig. 5

We shall now consider each of these two classes in turn.

Alan tells a story

Charles eats the soup

M,

M,

M2

M2

Erna tailors (e.g. a dress).

Fritz drinks (e.g . beer).

M,

M,

The incorporation of features pertaining to the created or observed object into the verb system is a very general procedure, in other cases a semantically poor verb like "make" is combined with a noun specifying the product. For similar processes in the domain of nominal composition (cf. Wildgen 1987). The details of the linguistic realization of basic processual schemes are language-specific. They depend on historical processes of lexicalization and grammaticalization (cf. Lehmann 1978; Wildgen & Mottron 1987). These historical processes have the effect that the configurational structure on the


ad 1: The creation, destruction, and regeneration of entities The first group clearly mirrors the fundamental schemes of emission and capture derived in catastrophe theoretic semantics from the cusp (A+ 3) (cf. Wildgen 1982:42-45 and 1985:118-136). As the process of emission and capture is energetically/intentionally poor, we can consider rather poor agents (semantically enriched with attributional processes by the human observer). Fig. 5 shows the basic process-types: emission and capture. As the examples given by Ballmer & Brennenstuhl (1986:377-393) show, we can distinguish two sub-types of emission: (a) something appears (against a background); this subtype is exemplified by verbs like "splash" or "spray" (but cf. Ballmer and Brennenstuhl 1986 for an analysis of the whole lexical field). The other subtype (b) of emission concerns situations where an agent creates something. In the latter case the verb frames can take either one or two nominal roles, depending on whether or not the understood second nominal role is incorporated into the verb. Cf. the following examples:

332

level of processual schemes is not isomorphic with structures of the lexicon or of syntax. This relativizes strongly the claims made by Rene Thorn in his original proposals for a universal grammar. In Wildgen (1982, 1985) we modified the applications of catastrophe theory to linguistics for precisely this reason. The subtype called "regeneration" by Ballmer & Brennenstuhl (1986) forces us to introduce a space of qualities (qualia) as a basis of the process. The objects considered have different qualitative "phases". In many cases the quality space has two (or in rarer cases three) stable phases. The symmetry between the two modes (or the two dominant modes, if more than two phases coexist) is normally broken and one pole is marked. The process called "regeneration" is given by a path XQ - Xj in the control-space with two conflicting states in the state-space. The general model for this situation is the catastrophe called cusp (V = x3 + ux2 + vx). Some typical verbs are the following: (a)

to distort, bend (German: verbiegeri) qualitative scale: © Q straight twisted, crooked

(b)

to clean (German: reinigen) qualitative scale: 0 © dirty clean, neat, . . .

The labelling of the scales is only provisional; it would be useful to take a more general space of qualities; for example the three-dimensional space


Fig. 6

(a)

(b)

Fig. 7

ad 2: The effect of an agent on the state of entities in his environment The asymmetry between agent and non-agent becomes more prominent in this group of verbs and the energetic or intentional aspect is foregrounded. Following our basic strategy we start with a mechanical analogy. In Fig. 7 we see two systems of a pendulum: (a) (b)

Two pendulums A and B. A gives its impulse to B via a series of punctual effects. Two pendulums dynamically coupled. The coupling may be either rigid or elastic.

The following sentences exemplify processes of type (a) in linguistic categorization: The player kicks the ball The man pushes a chair The girl throws the ball The general configurational scheme is a modified scheme of emission. Two systems (the agent and the object) are co-present. The agent is in motion and the object begins to move under the control of the agent (emission of movement). In the complementary case the movement of an object is absorbed by the agent. In all cases the existence and stability of the agent and the object are not changed by the action. Fig. 8 shows the modified scheme. Due to the fact that the agent (M,) is energetically superior to the object the


established on the basis of Osgood's semantic differential. If we assume a space with two poles we may describe the process contained in the two verbs above as in Fig. 6. This model shows that we must consider regeneration and degeneration/deterioration as two distinct subtypes. The invariant scheme is the archetype of (bipolar) change.

334 M 2 moves

M 2 moves

M moves beginning

end

Fig. 8

3.3. Process underlying the verbs of interaction The level of interaction between human agents cannot be strictly separated from the level of movement or manipulation of objects. It is rather a higher level of organization expanding these domains and requiring specific control of the coordination in a system of several (free) agents. The kinematic principles and the energetic or intentional orientation discussed in Sections 1 and 2 remain the same. What calls for explanation is the almost "unlikely" stability and constancy of patterns of interaction in a domain which has such a high degree of freedom. A first hint at the basis for such patterns can be found in animal ethology. Fentress (1982) and Golani (1982) have shown that there are very specific paths for the contact behaviour of mammals. The paths and their attractors can be lines of contact (between the tip of the mouth and the body of the partner) or lines followed in the bodily orientation of one animal (the direction of its head and its eyes). These lines follow very stable paths and stabilize in very specific regions. Thus a very small part of the surface of possible contacts is selected. Furthermore in the course of repeated contact very


energy of M, is not absorbed totally by M2 (this would correspond to the physical example (a) in Fig. 7). In Fig. 8 every point on the lines of the diagram is an entity in movement. Energetically the transfer of movement necessitates a higher archetype. The transfer of energy of intentional direction from M, to M2 (via some mediator) can be isolated (as in "kick"), repeated, or continuous (as a sum of rhythmic muscular actions, the basic frequency of the muscle). In the case of movement, the object changes from rest to movement under the influence of the agent. In a similar way the object can change its shape and even its qualities. Thus if we introduce a quality space we obtain a very rich field of actions on objects, which can be labelled and organized in the verb lexicon. The reader is referred to the classification of Ballmer & Brennenstuhl for more examples. An empirical re-categorization or an experimental foundation of verb categorizations is a goal for further applications of our theoretical sketch.

335

TO

Tl

Fig. 9

specific symmetries and asymmetries in the relative behaviour appear in the interaction and a highly ritualized pattern is created (cf. the analysis of the behaviour of wolves by Fentress, 1982:480, Fig. 10a,b). The punctual attraction in the relative movements of two agents play a similar role as the body-joints in Section 1. Thus even the patterns of interaction reproduce basic mechanical processes. Different types of social contact make use of different "joints": - the eyes of the mother are an attractor for the baby and are basic for the first contacts with the mother (humans attract humans specifically by the white parts of the eyeball and the movements of the eyes); - the bodily contact zone (at small distance) using the lips (compare the sucking activity of the baby); in the same line the breast of the mother is an attractor for the baby; - the contact at a certain distance using the hands (grasping, petting); - the contact of exchange (using the hands, controlled by the.limbs and the eyes); - the communicative contact (using the mouth and the ears as a kind


T5

336 of secondary instrument, which have a double function as they control primarily (in phylogenesis and ontogenesis) other processes. The coordination of interactive processes exploits these kinetic and energetic sources and elaborates them. Linguistic activity itself is such an interactive process; but if it refers to interaction linguistic classification takes up the basic schemata involved in these processes. We shall analyse one specific process in this field: the process of giving (receiving or exchanging).


(1) The configuration^ structure of "giving" We can describe the basic schema or prototype of giving configurationally, using a model proposed by Zeeman (1965). The scenario is given by a sequence of snap-shots: Tl to T5; each snap-shot describes an instantaneous three-dimensional configuration, in which the third dimension is a density function. The density is a correlate of the subjective focus in the perception of the scene or the concentration of motor control on certain regions. At the beginning and at the end of the series we observe two maxima of density (two attractors); in between the two maxima we witness the appearance of the object, its position of prominence, position of symmetry, its loss of prominence, and the disappearance of the object (i.e. the exchanged entity). The central, symmetric scene is the most unstable one. As the control of the two objects concentrates on one target (T3), the coordination of both controls must secure the smooth exchange (if A releases his control before B takes the object, or if A holds on to the object, although B seizes it, the character of the schema is dramatically changed and degenerates to "A loses or drops the object" or "A and B fight over the object C"). Thus the most unstable state of exchange is the "junction" of the process, the point of maximal coordination of the controls. It may be a meta-stable state, if the object gains some autonomy, for example if it is lying on a table between A and B and is in the reach of both but is not strictly controlled by either of them. This configuration is captured by the archetype of transfer (cf. Wildgen 1985:185). The higher archetype of the prominence of the object or instrument describes the difference between an unstable and a metastable intermediate phase of the process (ibid. 186ff). The process of exchange, transfer or change of possession is highly differentiated in the lexicon of German verbs. Ballmer and Brennenstuhl distinguish 137 specific "models" in the large group called HABEN ("have", possession). The phases which are empirically identified in their description map perfectly into the set of distinct phases one finds by a topologico-dynamic analysis of the archetype of transfer. Fig. 10 shows the correspondence. The single phases are the following (labels by Ballmer and Brennenstuhl):

337

AIV

I

HAVE2

Fig. 10

V

=

R

=

B

=

A

=

V

=

[

=

D = E =

VORSPIELPHASE RUHEPHASE BEGINNPHASE ABLAUFPHASE VERLAUFSPHASE 1NTENSIV1ERUNG DEZELERATIONSPHASE ENDPHASE

(preparatory phase) (phase of rest) (starting phase) (phase of acceleration) (phase of progression) (phase of intensification) (phase of deceleration) (final phase)

(2) The dynamic (intentional) structure of "giving" The two concepts of control and intentionality can be distinguished basically in two ways: (a)

(b)

By the range or level of control. A simple control is a function inside the agent and its immediate parts (limbs); a second type of control is created if the agent attends to an entity that is not part of himself and has its own dynamics (its own forces, goals, or aims). The control is not only "magnified" in order to cover this "domain", but necessitates a recognition of the envisaged action by the patient. This specific form of reflexive control of our environment will be called intentionality in the narrower sense. Intentionality integrates a complex scenario with objects and other agents into a higher "Gestalt" and defines quasi-intersubjective


GIVE /TAKE

HAVE,

338 person 2

person 2

object 1

y

V

person 1

object 2

person 1

phase I

phase II

Fig. 11: The "energetic" cycle.

As a first consequence the archetype of giving is intentionally in disequilibrium, as agent A finishes "poorer", and agent B "richer". An equilibrated schema is obtained by the schema of mutual exchange i.e. by a closed loop in the control space of the catastrophe-theoretic model. Fig. 11 shows this structure. In phase I, person 2 gets object 1 and "wins", thus creating an asymmetry of possession; in phase II, person 1 gets object 2 and "wins". The general figure represents two basic moves in a simple game. In a sense intentionality leads to a kind of game-theoretic constellation. Thus equilibria of games become the basis of complex interaction and of communication (compare the theory of game-theoretic equilibria and the notion of convention in Lewis, 1969). At the level of the basic archetype of transfer we can elaborate the dynamics using the concept of dominance in a dynamic system. In a (negative) gradient system the deepest minima are dominant (and structurally stable). Thus changes of dominance give the basic dynamic (energetic) characteristics of the archetype. We can say that with respect to the cycle the archetype of transfer is one perspective. Further perspectives are the following: - the archetype of emission (person 1 emits object 1; person 2 emits object 2) - the archetype of capture (person 2 takes object 1 etc.) - the archetype of metastable existence (prominence of objects 1 and 2) - the archetype of birth/death (appearance of the objects, focussing on the object in the intermediate phase of the process)


processes, social states and instabilities. It creates a trans-subjective schematization which is the presupposition for the evolution of communicative routines and of language.

339 The attractors of the dynamic system are correlated with: Ml :control of the initial phases M2 intermediate self-control of the object M3 :control of the final phases and in the semantic model (cf. Fig. 10) we can interpret the attractors as: Ml :source/sender/ergative or nominative case M2 :object/intermediate phase/path or absolutive or accusative case M3 :goal/receiver/dative case

- geben, schenken (give, donate) - erhalten, nehmen, abnehmen, rauben, stehlen (receive, take, take off, rob, steal) - austauschen (exchange) - kaufen, abkaufen, ankaufen, einkaufen (buy, buy of, acquire, shop) - verkaufen, ausleihen, zuriickgeben (sell, lend/borrow, return)

4. SOME ASPECTS OF THE NEURODYNAMICS OF VERBAL MEANINGS

The first sections of this article sketched the processual correlation between peripheral bodily processes (in their contact with the external world) and the processual semantics of verbs. Although the last few years have witnessed advances in the field of neurolinguistics, the link between bodily processes and symbolic processes, i.e. the cognitive mediation between action and perception on the one hand and language on the other cannot yet be modelled in exact terms. I shall try to give an idea of these mediating processes as they can explain the finer semantical structures of verbs, which were left unanalysed in the previous sections. The external bodily dynamics described in Sections 1 to 3 are directly related to the "action circuit" in the human brain. Its processual structure is summarized by Arnold (1984:2180 as follows: "Motor memory is both registered and reactivated by the action circuit. Motor engrams are registered in the prefrontal cortex during movement and reactivated when the movement is repeated. The action circuit runs from limbic areas via the hippocam-


The different dynamic perspectives are the basis for sub-classes of verbs. In German we may consider the following examples of verbs of "giving" (compare the more exhaustive classification in Ballmer and Brennenstuhl, 1986):

340 pus and post-commissural fornlx to the midbrain reticular formation and cerebellum and projects via cerebellar nuclei, reticular formation and ventral thalamus to prefrontaJ, premotor and motor cortex. Hence, motor memory is automatically activated with every relevant action and guides its further progress."


The action circuit is one of several circuits one may functionally and physiologically assume in the activity of the brain: the affective memory circuit (related to limbic 'appraisal' areas; cf. Arnold, 1984:219) and the imagination circuit (for which the amigdaloid complex is a kind of switching station). As this paper is not the place for an elaboration of neurophysiological details, I shall summarize recent results of brain research by hypothesizing that the peripheral dynamical systems descibed in the first sections are mapped into the neurodynamical systems of motor memory, motor imagination and appraisal. Thus the peripheral dynamic systems which match structurally with external dynamic systems (the mechanics of motion) are replaced by the (biochemical and electrophysiological) dynamic systems of the action circuit, the imaginative circuit and the evaluative circuit. These neurosystems are pre-linguistic in the sense that they are evolutionarily prior to the appearance of linguistic behavior. As soon as stable processual schemata in these neurodynamic systems are correlated with special subfields of (spoken and written) sign-production and sign-reception, we reach a new level of memory organization and of conscious control of action and imagination. However, this new level of complexity, which is introduced by the use of language is not something totally different, rather it emerges from the neurodynamic substrate processes (the basic mechanism could be that of selective stabilization advocated by Changeux, 1984). In our context, motor memory and motor imagination display a large range of patterns, only parts of which are selected for conceptual categorization. This rather small set is naturally discontinuous by the effect of selective stabilization and it is further stabilized by the coordination with linguistic patterns. This 'dual coding' in the memory (in motor memory and in auditory memory for spoken language) is reinforced in linguistic use. Our short elaboration of the processual semantics of verbs by the introduction of some neurophysiological processes that mediate between linguistic signs (e.g. verbs) and bodily processes (as controlled and perceived by the speaker) has consequences with respect to multi-modality and evaluation. The classification of external processes in one modality (such as vision or audition) is elaborated by the integration of multimodal features into complex schemata. These schemata are further differentiated with respect to evaluative parameters. Thus the input to the human conceptual memory underlying the lexicon of verbs is:

341

- action and perception based, - complex in the sense of multi-modularity, - reorganized by evaluative parameters. If we consider multi-modularity and evaluation in our description of German verbs we can achieve a more specific account. A. Multi-modularity If we re-analyse the list of motion verbs in Section 1 we notice the classificatory force of auditory patterns (in addition to motor patterns).

hopsen humpeln latschen schlurfen staken stapfen

(to (to (to (to (to (to

hop) hobble) shuffle) scuffle) stake) plod)

tappen tapsen tippeln trampeln trippeln zockeln, zuckeln

(to (to (to (to (to (to

grope about) walk clumsily) tramp) trample) trip) move slowly)

The meaning of these verbs refers not only to motor schemata (and corresponding visual controls), it also mirrors an auditory classification which becomes functionally prominent when the visual cues are absent. In some examples visual and auditory classification are closely related; thus the specific rhythmic features of hopsen, humpeln, tippeln, trampeln, trippeln can be perceived visually or auditorily. Other verbs like latschen, schlurfen, staken, stapfen, tappen, tapsen refer more directly to auditory impressions. The visual memory and the visual imagination of larger settings can be utilized to achieve a richer content of verbal meanings. Examples: pirschen preschen stelzen stiefeln tigern

(to (to (to (to (to

stalk; cf. the huntsman) race; cf. the racing car) walk (as) on stilts) march (as) in (heavy) boots) walk like a tiger)

B. Evaluation Many verbs have an evaluative dimension. If we take as a starting point the connotative space of Osgood (1976) with the axes Evaluation (E), Potency (P), and Activity (A) we can assign loadings (+ vs. - ) to the following list of verbs:


Examples (with an approximate translation):

342 Examples: bummeln hinken humpeln sen wdnzeln trampeln zockeln

(-A) (-P) (-P) (-E) (-E) (-P)

laufen pirschen preschen stapfen tigern wieseln

( + A) ( + P) ( + P) ( + P) ( + P) ( + A)

Examples: trippeln, htipfen, tippeln, wieseln (the vowels [i], [y] have high values of the formant F2)

quick rhythm of the movement referred to by the verb

schlurfen, zuckeln, latschen (the vowels [u], [a] have lower values of the formant F2)

slow rhythm of the movement

The processes of evaluative restructuring and of multimodal integration (aided by secondary links between conceptual content and linguistic form in iconicity) result in a very rich and very stable inventory of cognitive' 'entities" and create the basis for language. As our analysis has shown, meanings are organized at least on three levels which are intrinsically interrelated: (a)

(b)

(c)

The deepest (and evolutionary oldest) level is constituted by basic processes in the control of motion and action (including basic evaluative schemes). At a higher level different modalities are integrated into complex and stable entities. Memory-aided imaginative processes take up these results so that the planning of motion and action can be achieved independent of specific contexts. At the last (specifically human) level stable products of the processes in (a) and (b) emerge and a stable subclass is selected. Social coordination (conventions) can create a quasi-objective structure on this basis, which is called "language".


C. Onomatopoetic "meanings" of verbs In the case of auditory classification underlying the meaning of verbs we can observe a kind of dynamic resonance between the conceptual level and the acoustic form of the word. This kind of phonetic iconism in the lexicon is called onomatopoeia. In some of the examples listed above under A we can observe a similarity between phonetic features (front vs. back vowels, long vs. short vowels) and auditory features of the real world process.

343 5. CONCLUSION: THE MEANING OF "MEANING" IN THE DYNAMIC PARADIGM

University of Bremen FB 10 P.O. Box 330 440 D-2800 Bremen Fed. Rep. of Germany

ACKNOWLEDGEMENT I acknowledge the valuable comments by two anonymous referees and the editor. The final version of this paper has been written during my stay at the Institut des Hautes Etudes Scientifiques at Bures-sur-Yvette (Paris). I thank the Institute, especially Prof. Reni Thorn, for the hospitality.

BIBLIOGRAPHY Arnold, M.B. 1984: Memory and the Brain. Lawrence Erlbaum, London. Ballmer, T.T. 1982: Biological Foundations of Linguistic Communication. Towards a Biocybernetics of Language. Benjamins, Amsterdam. Ballmer, T.T. & Brennenstuhl, W. 1986: Deutsche Verben. Eine sprachanalytische Untersuchung des deutschen Verbwortschatzes. Narr, Tubingen. Ballmer, T.T. & Wildgen, W. 1987: Process Linguistics. Exploring the processual aspects of language and language use, and the methods of their description. Niemeyer, TObingen. Bernstein, N.A. 1974: The Co-ordination and Regulation of Movements. Oxford. Casti, J.L. 1985: Nonlinear System Theory. Academic Press, Orlando.


The term "meaning" gains a new dimension in the new morphological and dynamic paradigm. It is far more than a generalization about distributions (cf. Harris 1968), more than internal oppositions in a closed semiotic system (cf. structuralism) and even more than a mapping into set-theoretical "worlds" (cf. Montague 1970). It seriously considers the underlying processes in our body (including the brain) and the perceivable and controllable "reality" of our environment. In this respect it could be called ecological, but the notion of "morphology" in its original sense includes the morphologies in our environment, as well as our own organismic and mental morphologies. Thus the concept of "morphology" relates human communication and the world around us. In the new paradigm the structure of meaning must refer to basic laws of nature and of the mind (including ' 'laws of the speaking community"). It is not sufficient to build a corpus of "legal" rules to be followed by speakers; communication is far more than a game with conventional rules. As Wittgenstein recognized already, it is rather a form of life and intimately connected to the world which is accessible to our knowledge.

344

(Mass.).


Changeux, J.-P. 1984: Der neuronale Mensch. Rowohlt, Reinbeck. Cresswell, M.I. 1973: Logics and Languages. Methuen, London. Fentress, J.C. 1982: Ordnung in der Ontogenese: Dynamik von Beziehungen. In: K. Immelmann e.a. (eds.), Verhaltensentwkklung bei Mensch und Tier. Das Bielefelder Projekt. Parey, Berlin. Pp. 454-487. Fillmore, Ch. 1977: Scenes-and-frames semantics. In: A. Zampolli (ed.), Linguistic Structure Processing. North Holland, Amsterdam. Pp. 55-81. Gibson, J.J. 1968: What gives rise to the perception of motion. Psychological Review 75: 335-346. Golani, I. 1982: AufderSuchenachlnvarianteninderMotorik. In:K. Immelmanne.a. (eds.), Verhaltensentwkklung bei Mensch und Tier. Das Bielefelder Projekt. Parey, Berlin. Pp. 488-506. Harris, Z.S. 1968: Mathematical Structures of Language. Interscience Tracts in Mathematics 21; New York. Heider, F. 1958: The Psychology of Interpersonal Relations. Wiley, New York. Johansson, G. 1976: Spatio-temporal differentiation and integration in visual motion perception. An experimental and theoretical analysis of calculus-like functions in visual data processing. Psychological Research 38: 379-393. Kelso, J.A.S., Holt, K.G., Kugler P.N. & Turvey, M.T. 1980: On the concept of coordinative structures as dissipative structures: II Empirical lines of convergence. In: G.E. Stelmach and I. Requin (eds.), Tutorials in Motor Behavior. North Holland, Amsterdam. Pp. 48-70. Kugler, P.N., Kelso, I.A.S. & Turvey, M.T. 1980: On the concept of coordinative structures as dissipative structures. I. Theoretical lines of convergence. In: G.E. Stelmach and I. Requin (eds.), Tutorials in Motor Behavior. North Holland, Amsterdam. Pp. 3-47. Kruse, P., Stadler, M., Vogt, S. & Wehner, T. 1983: Raum-zeitliche Integration wahrgenommener Bewegung durch Frequenz-analyse. Gestalt Theory 5: 83-113. Labov, W. 1978: Denotational structure. In: D. Farkase.a. (eds.): Papersfrom theParasession on the Lexicon. Chicago Linguistic Society, Chicago. Pp. 220-260. Lehmann, Chr. 1981: Thoughts on Grammaticalization. A Programmatic Sketch, Bd. 1 (Arbeiten des Kolner Universalienprojekt (Akup) 48), Cologne. Lewis, D. 1969 : Convention: A Philosophical Study. Harvard University Press, Cambridge

Michotte, A. 1954: La Perception de la Causaliti. Publications de Louvain, Louvain. Montague, R. 1970: Universal Grammar. Theoria 36: 373-398. Osgood, Ch. 1976: Focus on Meaning. Mouton, The Hague. Richter, P. 1985: Das Doppelpendel oder: Wiestabilist unser Sonnensystem? In: H.-O. Peitgen und P.H. Richter (eds.), Harmoniein Chaos undKomos. Bilderausder Theorie dynamischer Systeme. Bremen. Pp. 11-15. Thorn, R. 1974: Modiles Mathimatiques de la Morphoginise. Recueil de textes sur la theorie des catastrophes et ses applications. Union Generale d'editions. Paris. Turvey, M.T. 1977: Preliminaries to a theory of action with reference to vision. In: R. Shaw and J. Bransford (eds.), Perceiving, Activity and Knowing: Towardsan Ecological Psychology. Erlbaum, Hillsdale. Pp. 221-267. Wehner, T. & Stadler, M. 1982: Frequenzanalytische Untersuchungen zur kognitiven Steuerung von Mehrfachhandlungen durch Rhythmisierung. ZeitschriftfurPsychologie 190: 183-201. Wildgen, W. 1982: Catastrophe Theoretic Semantics. An Elaboration and Application of Rent Thorn's Theory. Benjamins, Amsterdam. Wildgen, W. 1985: Archetypensemantik. Grundlagen fur eine dynamische Semantik aufder Basis der Katastrophentheone. Narr, Tubingen. Wildgen, W. 1987: Dynamic aspects of nominal composition. In: Ballmer and Wildgen (eds.), pp. 128-162. Wildgen, W. & Mottron, L. 1987: Dynamische Sprachtheorie. Brockmeyer, Bochum. Zeeman, Ch. 1975: Catastrophe Theory: A Reply to R. Thorn. Reprinted in: Zeeman, Ch., 1977: Catastrophe Theory. Selected Papers, 1972-1977. Addison-Wesley. Reading (Mass.). Pp. 622-632.


TEMPORAL RELATIONS D.S. BREE and R.A. SMIT

ABSTRACT

1. INTRODUCTION

Our aim is a representation that will capture the temporal properties that people place on events and states. We will proceed as follows: 1. Establish an "ideal" representation; 2. Use this to represent the meaning of the temporal prepositions and subordinating conjunctions in English, extending the representation where necessary; 3. Use the revised representation to capture the meaning of the temporal prepositions and SCs in a second language (Dutch); 4. Establish a mapping of the meanings of the temporal prepositions and SCs in the two languages. The semantic analysis of the use of temporal subordinating conjunctions (SCs) and prepositions will not use sample sentences from a corpus. This analysis is a preliminary theoretical step to empirical analysis.1 The aim is


An event or state can be located on the time dimension or given an extent of time. This can be done by adding a temporal prepositional phrase or subordinate clause to a sentence. We give here an analysis of the different temporal prepositions and subordinate conjunctions that are found in English. This analysis has two parts. The first takes the form of determining a set of rules for distinguishing the temporal use of such function words from other uses. The second gives rules for distinguishing the different semantic meanings of these words. These rules are then drawn together to produce a decision tree for selecting the appropriate function word, the function words appearing as the leaves of the tree. The English words are replaced by Dutch temporal function words, and the small differences between the two are noted. The criteria that are used to construct the selection trees provide the set of temporal Universal Linking Dimensions.

346 to provide the following types of rules for the temporal SCs and prepositions, which we will call temporal "function words" for short: - interpretation rules for distinguishing the temporal from other uses of some of the function words; - distinction rules for describing the difference between closely related temporal function words; - choice rules for deciding which temporal function word to use. 1.1. An ideal representation for temporal relations

- A is earlier than B - A is later than B - A and B are at the same time. Stevents are considered as being either punctual or durative: - punctual: the length of time during which the stevent is taking place is so short that no other stevent can occur within a shorter time. - durative: a stevent lasts sufficiently long so that another stevent can occur within its time span. This distinction holds only up to the granularity of time for the types of event under discussion. For example when the topic of discussion is human reaction times, then the granularity is around the 10 millisecond mark; but when talking about epochs in the development of species, then the granularity may be as large as a million years. The distinction is important only for stevents that occur at equivalent times. Formally, the temporal relationship is made using an operator, which relates the time of the stevent to one or more t-values. These t-values may be either taken from the time dimension or they may be the time of other stevents. We need to be able to specify that the time of a stevent is in some


We begin with some general comments about the nature of the ascription of time to an event or state. Formal specifications will follow. The time of a stevent, a word which we coin to mean "either a state or an event", can be ascribed in two ways: by relating it to the time dimension, e.g. 'at 5 p.m.'; by relating it to the time of another stevent, e.g. 'during dinner'. The time of any particular stevent does not have to be anchored in the time dimension, even through a chain of other stevents to which it is related. Relating the time of a stevent, A, to that of another stevent, B, should be quite simple. Either:

347

1.2. The English temporal prepositions and SCs If we attempt to place the English prepositions and subordinating conjunctions (SCs) that are used to indicate a temporal property of the main proposition, then we arrive at Table 1, in which the sign, '0', indicates the absence of any preposition, e.g. I'll see you Saturday. The prepositions and SCs shown have been drawn from various sources (Quirk et al., 1972; Bennett, 1975). Archaic forms, such as betwixt have not been included. There are some cells under the SC column which are empty; to express the desired relationship to a subordinate clause in such cases, use is made of the corresponding preposition, augmented with the time (that) or the moment (that), as in: I'll be at your disposalfrom the time that you get here to the moment you leave.


way related to a single t-value or that it falls within a range of t-values. The operators that we propose are: @ T signifying that the time of the stevent is the same as T; T signifying that the time of the stevent falls within the range T; < t signifying that the time of the stevent is earlier than t; > t signifying that the time of the stevent is later than t. where: t is a t-value, T is a T-value, a T-value being a range of t-values, T is either a t-value or a T-value. There are two ways of specifying the 7-values: - directly: giving the values on the time dimension: t as a t-value on the time dimension, T using two extreme t-values, tl and t2, which bound the range of the durative stevent; - indirectly, referring to another stevent which is either: - punctual and has a t-value, - durative and has a T-value. We first try to place the English temporal relationship words into this ideal framework. We find that the framework has to be extended. We then place the Dutch temporal relationship words into this extended framework.

348 Table 1. English temporal relations: first representation Code

Prepositions

Subordinating conjunctions

@t @T @[tl,t2]

at/around/by t for/throughout T tl through t2

when/as t as /o/ig as T

~ T

on/upon/0 T in/during T between tl and t2 before/until/to t after/since/from t

when/while T

' [tl.t2] < t > t

2. EQUATING TIMES OF STEVENTS USING SUBORDINATING CONJUNCTIONS

The time of one stevent can be equated with that of another stevent using three SCs: when, as and while. Each can be used for other than a temporal function, and each, when used temporally, fulfils a different temporal function. We first show that these SCs are closely related when used temporally. Their dictionary definitions are:2 When, 1. adverb interrogative. At what time?, on what occasion?, how soon?, how long ago? 2. adverb relative. at the or any time that, on the or any occasion that, at whatever time, as soon as; although, considering that.


In some of the cells there is more than one entry. We will examine these to see how the simple representation should be extended in order to differentiate between them. Some other distinctions that are made in the above table will also be discussed. The aim will be to determine what is the distinction between the entries and to capture this distinction by extending the set of permissible temporal operators. We will look first at the many different possibilities for equating the time of two stevents. In this category we deal first with the subordinating conjunctions and then with the prepositions. Then we examine the function words used to place two stevents at a distance from each other in time. Here the distinction between subordinating conjunctions and prepositions is not so relevant, most of the function words being used in either syntactic role.

349 As, 2.

1.

relative adverb or conjunction in subordinate clause. expressing manner, degree, etc., of the principal sentence: (degree, manner, time, reason, result, illustration). while, when; at the time that. (Collins)

While, (conjunction). 1. During the time that, for as long as, at the same time as. 2. In contrast more or less marked with the fact that simultaneously, although, whereas, and.

(e.g. leave):3

Sub-clause verb with durative aspect: i As the dog slept, Judy played with her dolls. While the dog slept, Judy played with her dolls. When the dog slept, Judy played with her dolls. Sub-clause verb with punctual aspect: i ? George was shot while he left the bank. George was shot as he left the bank. George was shot when he left the bank. Before examining these semantic differences, we will look at the circumstances under which as and while are given temporal interpretations rather than manner, inferential or contrastive ones.


As we can see from these dictionary definitions the meanings of when and as and while in their temporal use are closely related: Judy was making her way home when the fireworks were exploding. Judy was making her way home as the fireworks were exploding. Judy was making her way home while fireworks were exploding. Is there any difference between their meanings and, if so, what is this difference? The simplest thesis concerning the relationship between their meanings is: as indicates a time instant, while indicates a time period, when can be used for either case and so is ambiguous between as and when. This is demonstrated by the following examples, in which the verb in the sub-clause has either a durative aspect (e.g. play with) or a punctual aspect

350 2.1. Distinguishing the temporalfrom the non-temporal uses of as and while If the verb in an as clause has a durative aspect, then as may be given either a manner or an inferential interpretation: Manner: Inferential:

Judy played with her dolls as Daniel looked after his hamsters, with care. Judy played with her dolls, as the dog was asleep.

Temporal: George was shot while he left the bank. (= was leaving) Contrastive: Susan was only wounded, while George was killed. So it appears that as is interpreted temporally when it introduces a clause with a punctual verb, but while is interpreted temporally when it introduces a clause with a durative verb, or one that can be given a durative interpretation in relation to the punctual verb in the main clause, whereas when can be used in either case. However, this is too simple. There are occasions in which when may be used, but neither as nor while can be given a temporal interpretation. For instance, if the sub-clause has a perfect aspect: With the sub-clause verb in the present: i As George has left the bank, Arthur will go in. i While George has left the bank, Arthur will go in. When George has left the bank, Arthur will go in. With the sub-clause verb in the past: 1 As George had left the bank, Arthur went in. i While George had left the bank, Arthur went in. When George had left the bank, Arthur went in. Here as is given an inferential interpretation and while is given a contrastive one. In both cases a temporal interpretation is not permissible. The temporal interpretations can be restored by also having a perfect aspect in the main clause:


It is beyond the scope of this paper to provide rules to detect the difference between a manner and a inferential interpretation of as. If the verb in a while clause has a punctual aspect, then either the duration of the activity of the sub-clause verb is considered to be long enough for the main event to occur within its time span, or white is given a contrastive interpretation:

351 With the verbs in the present perfect, referring to previous events which have occurred repeatedly: When George has left the bank, Arthur has gone in. As George has left the bank, Arthur has gone in. While George has left the bank, Arthur has gone in. With the verbs in the past perfect, referring to a single previous event:

However, as may be given a causal or a manner interpretation, rather than a temporal one, even with a punctual aspect verb in the sub-clause, when the mode is not progressive: Punctual aspect, progressive mode: Judy cried as the ballons were bursting. Punctual aspect, no progressive mode: i ? Judy cried as the balloon burst. So the choice of which interpretation to place upon an as clause with a unctual verb depends partly on its mode. If the progressive mode is used, as is given a temporal interpretation. If the progressive mode is not used, then we see that as may also be given a causal or manner interpretation. What determines the interpretation that is to be given to as with a nonprogressive punctual verb in the sub-clause? Comparing: Temporal: Judy laughed as her mother left the room. Inferential: i ? Judy laughed as her mother pulled a funny face. i He was executed as he had killed his king. Manner: i Judy laughed as her mother laughed. i He killed his father as he had killed his king. we propose, as a first suggestion, that the interpretation given to as in such


When George had left the bank, Arthur had gone in. As George had left the bank, Arthur had gone in. While George had left the bank, Arthur had gone in.

352 Table 2. As/while interpretation rules: For a temporal interpretation of as and while it is necessary that the tense and aspect of the main and sub-clauses be the same. Further: as

while

is interpreted temporally when it introduces a clause with a durative verb, or one that can be given a durative interpretation in relation to the punctual verb in the main clause; otherwise it is given a contrastive interpretation.

circumstances depends on the relationship between the verbs in the main and the sub-clause. If the sub-clause verb describes an action that typically occupies a shorter period of time than the action described in the main verb, then as is given a temporal interpretation. This suggestion needs to be tested against data from a corpus. Summarizing these results gives the revised rule for the interpretation of as and while shown in Table 2. Now that we have roughly defined the domain in which as and while are to be given temporal interpretations, we look at the semantic differences between them and when. In the process we will need to refine our as/while interpretation rule. We first look at the distinction between when and as and then between when and while. 2.2. When and the temporal use of as We initially postulated that the semantic difference, if any, between when and as is that as is more precise in its specification of the time of the main stevent than when. Is this actually so? A dictionary definition of as gives little help in the answer to this question, beyond pointing out that the uses of as are legion. This difference is similar to that between at and around, as we see below (in 3.1). However, the difference between as and when does not correspond to that between at and around, as the following examples illustrate:


If the verb in the sub-clause has: a. a durative aspect, then no temporal interpretation is possible; b. a punctual aspect with the progressive mode, then as is given a temporal interpretation; no progressive mode, then as will be given a temporal interpretation if the verb describes an action that typically takes place a in period of time shorter than the period required for the action described by the main verb.

353

I'll meet you II'U meet you I'll meet you I'll meet you ?/'// meet you ll'll meet you ?/'// meet you I'll meet you

at 2 minutes past 5 precisely. around 2 minutes past 5 precisely. precisely as you leave the bank. precisely when you leave the bank. roughly at 5 o 'clock, roughly around 5 o 'clock. roughly as you leave the bank. roughly when you leave the bank.

Punctual aspect, progressive mode: Judy cried as the balloons were bursting. Judy cried when the balloons were bursting. Punctual aspect, no progressive mode: i ? Judy cried as the balloon burst. Judy cried when the balloon burst. Durative aspect: i Judy cried as her mother was away. Judy cried when her mother was away. The major difference that we have noted so far between as and when is that as must refer to a progressive action. This suggests that if when refers to a progressive action, there should be no difference between the two SCs. This is apparently not the case, e.g. compare: The gangster was shot when he came out of the bank. The gangster was shot as he came out of the bank. With 05 the shooting must take place specifically at the moment that the gangster is leaving the bank. With when the timing of the shooting is not so precise; it may have taken place just after the gangster left the bank. The reason is that come is not given a progressive interpretation with when, as opposed to its use with as. If an explicit progressive is used, then there is no difference between the two SCs:

=

The gangster was shot when he was coming out of the bank. The gangster was shot as he was coming out of the bank.


So when is not restricted to an approximate time, but as is restricted to a precise time. We noted in 2.1 that a durative verb in the sub-clause prevents a temporal interpretation of as. When does not have this restriction. We illustrate this with three examples in which the aspect of the verb in the sub-clause and its mode are varied:

354 Table 3. As/when distinction rule: When may refer to a point in time, to a progressive action or to a period of time, but temporal as must refer to a progressive action; The temporal use of as must have in its sub-clause a verb which has a punctual aspect and either: is in the progressive mode, or must be capable of being given a progressive interpretation; If the sub-clause contains a punctual verb with no progressive mode, but the verb can be given a progressive interpretation, then it is so interpreted with as but not with when.

2.3. When and the temporal use of while The difference in the meaning of when and while is easier to capture. When points to the time(s) or occasion(s) at which the main clause stevent occurs; while, in its temporal use, points to the time duration in which the main stevent occurs. However, we have seen that when and while can be used more or less interchangeably in some sentences, a point which is captured in the dictionary definition of while by 'at the same time as'. If we compare: Judy was making her way home when the fireworks were exploding. Judy was making her way home while the fireworks were exploding. we find that there is a difference in nuance. The when sentence simply indicates the time of 'Judy's making her way home', whereas the while sentence adds a sense that 'Judy is making her way home specifically in the time period that the fireworks are exploding' and not some time around about the time of the exploding fireworks. This difference leads to other causal attributes being given to the while sentence, such as: Judy is going home under cover of the fireworks. She wants to get there before the fireworks finish. But these added connotations are something that the reader brings to bear in trying to make a causally coherent story out of the stevents. They are not part of the meaning of while. When can also be interpreted as indicating a period of time rather than a point in time, e.g. the difference in meaning between:


These differences between when and the temporal use of as are summarized in Table 3.

355 Table 4. When/while distinction rule: -

When indicates the point in time that the main stevent occurs, whereas while indicates the time period in which the main stevent occurs; - If the clause contains a verb having a durative aspect compared to the main clause verb, then: - with when the time of the main stevent may be interpreted as falling within the time duration indicated by the sub-clause verb. - with while the time period is explicit and so is taken up in making causal sense of the sentence.

is small. The when but not the while version may be given a whenever interpretation. The difference is not because when has an alternative durative interpretation, but because of the durative aspect of the sub-clause verb, sleep. Our conclusion of this analysis, in Table 4, is simple. 2.4. A pragmatic difference So far we have sought, found and specified semantic differences between when, as and while. But these differences may be a result of pragmatic distinctions between when on the one hand and as and while on the other. Consider this example: ? George went to the bank this morning. As he left he was shot. George went to the bank this morning. When he left he was shot. George went to the bank this morning. At 5 o'clock he was ready to go home. As he left he was shot. 1George went to the bank this morning. While he was leaving he was shot. George went to the bank this morning. When he was leaving he was shot. George went to the bank this morning. At 5 o'clock he was ready to go home. While he was leaving he was shot. The insertion of the at sentence serves to update the Time Of Reference (TOR) of the story to the time appropriate for 'leaving the bank', and so permits the following ay or while sentences to be acceptable. For when this is not necessary. This suggests that when itself has the function of updating


While the dog slept, Judy played with her dolls. When the dog slept, Judy played with her dolls.

356 the TOR (cf. the analysis of then by Kamp, 1981), but that as and while do not. This same difference accounts for why when can introduce interrogative sentences but as and while cannot: *As was George shot? * While was George shot? When was George shot?

Table 5. When versus temporal as/while distinction rule: - When sets the Time of Reference (TOR) to the time of the sub-clause event, whereas temporal as and while equate the times of the sub and main events, without affecting the TOR.

2.5. Summary

The various strands have been pulled together in Table 6. Table 6. When/as/while interpretation rules: When - is always given a temporal interpretation. It sets the Time of Reference (TOR) to the time of the sub-clause stevent; - if the when clause contains a verb having a durative aspect compared to the main clause verb, then the time of the main stevent may be interpreted as falling within the time duration indicated by the sub-clause verb. As and while - in their temporal interpretations, equate the times of the sub and main stevents, without affecting the TOR; - so for a temporal interpretation of as and while it is necessary that the tense and aspect of the main and sub-clauses be the same.


The answer to a when question may well be an as or a while clause, indicating that the time of the main event is equal or within the time of the subevent now introduced. When is used to query the TOR of the main stevent. The equality of the times of the main and sub-stevents cannot be queried by using as or while as interrogates, because the sub-clauses are not yet available. In Table 5 we present the rule or distinguishing the use of when from temporal as and while.

357 Table 6. Continued

While - is interpreted temporally when it introduces a clause with a durative verb, or one that can be given a durative interpretation in relation to the punctual verb in the main clause; otherwise it is given a contrastive interpretation; - interpreted temporally, indicates the time of the main stevent that falls within the duration indicated by the sub-stevent.

3. FIXING THE TIME OF A STEVENT USING PREPOSITIONS

Whereas the SCs when, as and while are used temporally to equate the time of two stevents, the prepositions at, on and in are used to fix the time of a stevent on the time dimension or to be the same as a simple stevent, e.g. an arrival. We look first at these three prepositions in turn, comparing them with other prepositions having a similar temporal function. Then we make a comparison between them to determine which of the three should be chosen. 3.1. At, around and by

At, preposition 1. Expressing exact, approximate, or vague position, literally and 2.

figuratively, as . . . at work, at dinner, at midday. Expressing motion towards, literally and figuratively, as . . . at.

Around,

2.

preposition. On, along, the circuit of; about, enveloping.

arrive


As - the interpretation given to an as clause is governed by the verb in the sub-clause. If it has: a durative aspect, then no temporal interpretation is possible; a punctual aspect, then if the mode of the sub-clause verb is: progressive, then as is given a temporal interpretation; not progressive, then: if the verb can be given a progressive interpretation, then it is so interpreted (but not with when). if the verb describes an action that typically takes place in period of time that is shorter than the period typically required for the action described by the main verb, then ay will be given a temporal interpretation; otherwise it will be given a manner or inferential interpretation; - interpreted temporally, indicates that the time point of the main stevent is the same as that of the sub-stevent.

358 By, preposition and adverb. 3. During, in the circumstances of: by day. 5. As soon as, not later than. 3.1.1. At We first take at, and note two distinctions: at to express position versus motion and a temporal versus locative interpretation. At can be used to express position or motion. It is the main verb that distinguishes between these two uses. For example, with a locative interpretation of at: The ice-cream man is selling at the street corner. The ice-cream man is arriving at the street corner. The ice-cream man is looking at the street corner.

A verb of motion "forces" the at PP to supply the value of the direction. For instance, the 'looking' is not being done on 'the street corner', but is being done towards 'the street corner'. An at PP may be given a locative, temporal or figurative interpretation. When an at PP is used for a position then it is the nature of the NP that governs the interpretation: space: time: money:

The ice-cream man sells marijuana at the street corner. The ice-cream man sells marijuana at noon. The ice-cream man sells marijuana at a high price.

If the NP combines locative and temporal features, then the at PP also combines these, e.g. in: There were many students at the professor's funeral. the many students were present at the time and at the place of the funeral. It is not possible to have a temporal motion interpretation. Either a temporal position interpretation is given, even though a motion interpretation would be given with a locative NP: space:

The ice-cream man is arriving at noon.

or, when this is not possible, a weird interpretation is evoked: i The ice-cream man is looking at noon. Besides these literal uses of at PPs, there are figurative uses:


position: motion:

359 figurative:

The ice-cream man is good at repartee.

An analysis of how the temporal uses of at PPs are to be distinguished from figurative uses is beyond the scope of this paper. 3.1.2. Around versus at Around can also be used both with location and time: space: The ice-cream man sells marijuana around the street corner. time: The ice-cream man sells marijuana around noon.

I'll meet you at 2 minutes past 5 precisely. *I'll meet you around 2 minutes past 5 precisely. Note that the converse does not hold; at can be used with an approximation: I'll meet you roughly at midday. I'll meet you roughly around midday. So around will be distinguished from at by adding a specification of approximation. Informally, around is equivalent to at about. 3.1.3. By versus at A by PP, like an at PP, but unlike specifically temporal prepositions as before, only receives a temporal interpretation if its NP is temporal: There was an uncanny silence before the ceremony. i There was an uncanny silence by the ceremony. The use of by for during, in the phrases by day and by night (see O.E.D. definition above), will be excluded from the analysis.4 The semantic difference between at and by is clear: whereas at specifies the time of occurrence of the main stevent, by specifies the latest time by which the main stevent will be true/occur:


Again the distinction between the two uses is on the bases of the NP in the PP. Turning now to the difference between at and around, it is obvious that the temporal use of around introduces a degree of approximation concerning the location of the time of the main stevent with respect to the time given by the NP. For around, this approximation cannot be negated:

360

Claire will be home at 6 o 'clock. Claire will be home by 6 o'clock. Temporal by is semantically close to before. The difference is that by allows for the main stevent to occur at a time that includes the reference stevent, whereas before does not: Claire will be home by 6 o 'clock. Claire will be home before 6 o 'clock.

3.2. In/during/within In, prepositional expression. Inclusion or position within limits of space, time, circumstance, etc., as: (of time) in (during) the day, in (within the space of) three months, in (at the end of) five minutes: During, preposition. Throughout, at some point in, the continuance of. Within, preposition. in a time no longer than: shall have it within an hour; before expiration: within a year of his death; since beginning of: have seen him within these three days. Both in, within, and during locate the main stevent within a time period. However, they differ on several points. In and within can be used for spatial location whereas during is always


Below we compare by with until. Indeed, a case could be made for classifying by more closely with the non-contemporaneous prepositions, such as before and until, rather than with the contemporary prepositions. The reason that we have not done so is mainly that before and until are both essentially temporal, as well as being used both as prepositions and SCs, whereas by is a general purpose preposition. We turn now to the other categories of contemporary prepositions. We will then compare these with at.

361 temporal (Bennett, 1974, p. 126). In and within only receive a temporal interpretation if their NP cannot have a spatial interpretation: The boy disappeared during the fireworks. i The boy disappeared in the fireworks, (spatial) i The boy disappeared within the fireworks, (spatial) We first examine the different interpretations that can be given to in. Then we turn to the distinction between in and during, finally to the finer distinction between in and within.

In introduces a phrase that confines the main stevent to occur in a reduced space. This space may be locational, temporal or circumstantial. Here we are interested only in temporal confinement. We assume that the temporal use of in can be distinguished from the locational and circumstantial by the nature of the NP governed by in. If this NP specifies a time, then the in is given a temporal interpretation, i.e. the nature of the NP selects the temporal interpretation for the whole PP. Just as with at and around, there is no apart temporal meaning of in. As the dictionary definition of in indicates, there are three interpretations that can be given to a temporal in PP: P during a specific named Period on the time dimension; E a time Extent; ET at the end of a time Extent, commencing at the TOR. For example: P Barry shaved in the morning. E Barry shaves in 5 minutes. ET Barry will shave in an hour. The question that we now address is: how do we make the choice between these three interpretations of the temporal use of inl We begin by distinguishing the specific period and the two varieties of time extent. A period interpretation is given when the NP contains a specific time period rather than a general extent of time, i.e. the time is definite rather than indefinite, compare: P Concorde flies to Singapore in the morning. E Concorde flies to Singapore in a morning. ET Concorde flies to Singapore in a minute.


3.2.1. In

362 Note that quantifiers introduce indefinite time periods: E(T)Concorde flies to Singapore in 5 hours. Next we distinguish between the 'Extent' versus 'Extent TOR' interpretations. As the last example illustrates, some in PPs are ambiguous in this respect; Concorde could take 5 hours to fly to Singapore or it could be leaving 5 hours after the TOR. It appears that the tense of the main verb might enable the E versus ET interpretation to be made. Comparing:

we see that having a past tense main verb favours an 'Extent' interpretation, whereas a future modal favours an 'Extent TOR' interpretation. However, this latter can be overruled by a context in which it is not possible for the event to occur at the end of the specified time period, e.g. in a context where plans are being made for the development of a European space-plane, the sentence E

The new space-plane will fly to New York in 30 minutes.

must be given an 'Extent' interpretation, since there is no chance of a spaceplane being in existence within 30 minutes from now. When the main verb is in the present, then the interpretation depends on the durative versus punctual aspect of this verb. Compare: E Concorde conveys passengers to Singapore in 5 hours. ET Concorde departs for Singapore in 5 hours. Conveying is a durative activity and departures are punctual, in contrast to the time of 5 hours. Sometimes the time extent is such that both a durative and a punctual interpretation of the main verb are possible, as in E? Concorde flies to Singapore in 5 hours. where flying is ambiguous between conveying (through the air) and taking off.

Seeing that: - a main verb in the past tense favours an 'Extent' interpretation; - a punctual main verb in the present gives an 'Extent TOR' interpretation; what happens when a punctual verb is given a past tense?


E Concorde flew to Singapore in 5 hours. ET Concorde will fly to Singapore in 5 hours.

363 Table 7. Temporal in PP interpretation rules: If the NP in an in PP is: -

? Concorde departed for Singapore in 5 hours. is acceptable in a 'dramatic past' story telling style, as in: What was I to do now? Concorde departed for Singapore in 5 hours and I had to find Bailey before then. Similarly seeing that: - a main verb in the future gives an 'Extent TOR' interpretation; - and a durative main verb in the present gives an 'Extent' interpretation; what will happen when a durative verb is used in the future? The space-plane will cruise to Singapore in 2 hours? is also acceptable if the main clause, without the in PP is acceptable, i.e. that there is some reason that it cannot be true at the present time but will be true in the future. We conclude that it is not the tense of the main verb that is directly relevant for deciding the interpretation of the in PP, but its aspect (durative versus punctual). We capture all these differences in Table 7. 3.2.2. During versus in During in contrast to in, is always given a temporal interpretation (Bennett, 1975): / met a friend of yours during the vacation. I met a friend of yours in the vacation.


definite then the PP is given a specific time period interpretation, i.e. the main stevent occurs at the specified period on the time dimension. - indefinite (which includes quantified time periods), then it is given an 'extent of time' interpretation, i.e. the main stevent is related to an extent of time. If the aspect of the main verb is: durative, then the in PP is given an 'extent' interpretation, i.e. the stevent lasts for this length of time; punctual, then the in PP is given an 'Extent TOR' interpretation, i.e. the stevent takes place at a moment on the time dimension later than the TOR by the extent of the NP.

364 This can be most clearly seen when the PP has a non-temporal noun: Locative noun in the PP: * We'll have a chance to chat during the dining room. We'll have a chance to chat in the dining room. Activity noun in the PP: • We'll have a chance to chat in dinner. We'll have a chance to chat during dinner.

P * !

Barry shaved during the morning. Barry shaves during 5 minutes. Barry will shave during an hour.

We turn now to the difference between during and temporal in PPs when both can be used, i.e. with a definite time (e.g. an hour) as the NP. Comparing: We'll have a chance to chat in the evening. We'll have a chance to chat during the evening. suggests that in PPs are used to indicate that the main stevent occurs once at a time within the range indicated by the NP, whereas during is used to indicate that it occurs for the whole of the PP. This can be checked by using a punctual main verb, which should not be acceptable with a during PP. But this is not the case, e.g.: Wendy left during the morning Wendy left in the morning are both acceptable. Another possible difference is that one of the two PPs updates the TOR whereas the other does not (cf. the difference between when and as/while). We turn to some sketches for evidence:


If the noun is definitely a locative, then the in PP is interpreted as a locative expression whereas the during PP is not acceptable. If the noun is an activity that can take place during a time period, then the in PP is unacceptable whereas the during PP indicates the time period in which the activity is taking place. During has no ambiguity. A during PP is always given a temporal reading with a pointer to a specific period of time on the time dimension. Replacing in by during in the examples used above confirms this, e.g.:

365 Table 8. During versus temporal in rules: A during PP: - specifies the definite time period in which the main stevent occurs. During versus In - the NP with during may be a time or an activity, whereas for an in to receive a temporal interpretation, its NP must explicitly be temporal. - it is also possible that a temporal in PP updates the TOR, whereas a during PP does not.

Departures always affected our family. We had a rousing farewell dinner that night. Wendy left in the morning. Lunch was a subdued affair. The in PP updates the TOR with no problem. Is the same true for the during PP? The evidence here is not clear cut. The possibility that an in PP updates the TOR whereas a during PP does not, remains open. So aside from the fact that a temporal interpretation of an in PP, as opposed to a during PP, requires a temporal NP, the semantic differences between the two have not been resolved. These results are summarized in Table

3.2.3. Within versus in

Within specifies that an event takes place in a time period that is less than, or equal to, the time extent indicated by the NP. It cannot be used to specify a definite time period on the time dimension, such as 'the morning': *P Barry shaved within the morning. There appears to be an exception to this rule; the definite time period just before the TOR can be addressed using the last: P

Barry hasn 't shaved within the last 5 days.

But the NP only appears to be definite; it is actually a quantified NP. Dropping the quantifier is unacceptable:


? Departures always affected our family. We had a rousing farewell dinner that night. Wendy left during the morning. Lunch was a subdued affair.

366

*Barry hasn 't shaved within the last day. Barry hasn't shaved since yesterday. Just as with in PPs, the time extent can be free or attached to the TOR. When the TOR is involved, then the main event must occur in the period between the TOR and the time of the NP (cf. with in the stevent occurred at the end of this time period). E Barry shaves within 5 minutes. ET Barry will shave within an hour.

ET Barry shaved within a minute. The within PP specifies the length of time after the TOR (which is in the past) in which the main stevent takes place.5 We turn to the rules for choosing the interpretation (E or ET) to be given to a within PP. The choice depends on the durative v. punctual aspect of the main verb in the same way as an in PP. If it is durative then it refers to a pure extent of time, but if it is punctual then the time period is linked to the TOR: E Concorde conveys its passengers to Singapore within 5 hours. ET Concorde departs for Singapore within 5 hours. Finally note that the difference with the corresponding in sentences is that with within the length of time is specified as being less than (or equal to) the time given by the NP, whereas with an in PP it is equal to this time. This distinction is not the same as that made by the O.E.D. The three distinctions that the O.E.D. do make for temporal within would all be classified as ET interpretations by us. The differences have more to do with the nature of the NP than with within: - "in a time no longer than" is equivalent to the ET interpretation; - "before expiration" is similar to the ET interpretation, but the extent of time does not start with the TOR but with a point in the qualifier in the NP. In within a year of his death


Note that a main verb in the past tense and an indefinite NP is acceptable with no special style indication (cf. with in, see 3.2.1 above):

367 the time extent is anchored to the time of death rather than the TOR. We see that the TOR is a default value when no beginning time point is specified (using of...). Note that the same phemomenon is not permissible with an in PP: 6 *in a year of his death - "since beginning of" captures the period ending at the TOR: within the last 5 days; In Table 9 we propose a rule for the correct interpretation of within.

A within PP: - with a specific time noun in its NP specifies the extent of time in which the main event takes/can take place. - when the aspect of the main verb is: E durative, then the main event occupies (most of) the time 'Extent' specified by the NP; ET punctual, then the main event occurs somewhere in extent of time specified by the NP and either: - by default, beginning at the TOR; - if a time point is used to qualify the NP (using of ...), then the beginning of the time extent is measured from this point rather than the TOR; - if the indefinite quantifier is governed by these/the last, then the time extent ends at the TOR.

3.2.4. Summary We bring together all the differences between in, during and within PPs in Table 10. Table 10. In/during/within

PP differences:

In versus during/within PPs - an in PP updates the TOR whereas a during/within PP does not. - An in PP can specify the definite time period or an indefinite time extent of the main stevent, whereas: - a during PP can only specify a definite time extent; - a within PP can only specify an indefinite time extent, which may begin (or end) at the TOR (or some other time specified by a qualifying of in the NP); In/within versus during PPs - The NP governed by during does not have to be explicitly temporal but may indicate any activity, whereas for a temporal interpretation of an in/within PP the NP must be explicitly temporal.


Table 9. Within PP interpretation rule:

368 The representations to capture more formally the interpretation rules are given in Table 11. Table 11. Formal representation of temporal in, during and within time location

time extent

in the T during the S

durative + in a T durative + within a T

T7 S , 0 0

= within a year of his death

7. The TOR is updated to this time.


The selection trees that we have constructed are a step in providing the temporal aspects of a Universal Set of linking Dimensions. This set should be sufficient to describe the temporal function words in any language. If the present selection tree can be used for categorizing the temporal function words of other non-European languages, or, failing this, if the same criteria can be used in another sequence for this task, then we will have a theory about either the way our minds are capable of grasping temporal relations, or a theory about how temporal relations exist in the world, depending on one's point of view concerning the origin of such words in language (Bree & Smit, 1986). It will also provide a basis for the machine translation of such words.

383 8. Note that the temporal on PP is not ambiguous in the way the locative on PP is, i.e. between attachment to the verb or to the object NP. 9. Night is considered an instant in time. 10. Of course major chunks of stories are often told in other than a chronological sequence, see Waterland, by Swift (1983), for a good example. But within each chunk a chronological sequence is the norm. 11. Or some other time event specifically specified. 12. This criterion is dubious, see the discussion in section 3.2.2.

REFERENCES


Bennett, D.C. 1975: Spatial and temporal uses of English prepositions. Longman, London. Bree, D.S. 1985: The durative temporal subordinating conjunctions: since and until. Journal of Semantics, 1985,4: 1-46. Bree, D.S. & Smit, R.A. 1986: Linking propositions. Proceedings of the Eleventh International Conference on Computational Linguistics. Association for Computational Linguistics, Bonn; 1986, 177-180. Bree, D.S., Smit, R.A. & Schotel, H.P. 1984: Generation and comprehension of Dutch subordinating conjunctions by computer. In T. O'Shea, ed., Proceedings of the 6th European Conference on Artificial Intelligence. Elsevier, Amsterdam, 205-208. Heinamaki, O. 1978: Semantics of English temporal connectives. Reproduced by the Indiana University Linguistics Club. Kamp, H. 1981: A theory of truth and semantic representation. In J.A.G. Groenendijk, T.M.V. Janssen & M.B.J. Stokhof, eds., Formal methods in the study of language, vol.1. Mathematisch Centrum, Amsterdam, 277-322. Quirk, R., Greenbaum, S., Leech, G. & Svartvik, J. 1972: A grammar of contemporary English. Longman, London. Rohrer, C. 1977: How to define temporal conjunctions. Linguistische Berichte, 1977, 51: 1-11. Swift, G. 1983: Waterland. Poseidon Press, New York.


AT EASE WITH "AT' BIRGIT WESCHE

The preposition at has a range of rather diverse meanings - locative, temporal, causal, etc. which would not seem to be captured by any common denominator that would still be strong enough to distinguish at from other prepositions. In order to clarify the relationship of the various meanings of at to each other, this paper assumes the notion of a semantic prototype and shows how other senses of at develop from a basic locative sense of at, and how these derived senses are motivated by the fact that their domain of application is conceptualized in analogy to the domain of the basic spatial sense, or in analogy to another sense of a; that is directly or indirectly related to the original locative sense along the same lines.

1. INTRODUCTION

The English preposition at belongs to the set of fundamental locative prepositions at, on, and in. While an NP to which at is attached should ideally refer to a (zero-dimensional) topological point, an NP that goes with on refers to a (one-dimensional) line or (two-dimensional) surface, and an NP with in refers to a (two-dimensional) area or a (three-dimensional) space. But these, in a sense ideal, uses of at, on, or in are not their only uses. At, for instance, is also found in applications such as at school, at someone's discretion, at random, arrive at, look at, etc. How are these senses related to the basic topological point sense and how are they related to each other? In the following we will use the notion of a prototype in order to characterize the structure of the overall concept of the preposition at. Concepts with prototype structure "are composed of a 'core meaning' which consists of the 'clearest cases' (best examples) of the category, 'surrounded' by other category members of decreasing similarity to that core meaning" (Rosch 1973:12). The core meaning, or prototype, is the one displaying the most characteristic set of features associated with the general concept. In this sense the locative meaning will be assumed as the prototype of the general concept denoted by the preposition at. Depending on the context, variations of this core meaning successively extend the protytype to further concepts, which, in their turn, can serve as the basis for new concepts. One can thus distinguish between "centrally-motivated senses", which emerge directly from of the core mea-


ABSTRACT

386 ning and "distantly-motivated senses" (cf. Lakoff 1982:83), which develop from a derived concept and are less representative of the overall concept. Even these distantly-motivated senses, however, will still bear enough of the characteristic features found in the prototype to be clearly distinguished from concepts of other prepositions such as on or in. We shall now consider the principles that motivate the extensions of the prototypical meaning and that yield the various derived meanings of at. The two basic operations that are at work here are -

successive change in the set of features and projection onto a new domain.

(a) a set of features that characterize the kind of relationship between the located object and the reference object (such as [±locative], [idirectional], [itemporal], [±causal], [iintentional], etc.), (b) a set of features that characterize the reference object (e.g. [±static], [±dimensional], [±shape], [±concrete], [ianimate], [±goal], [±target], [±cause]), and (c) a set of features very much like those under (b) that characterize the located object, as required by the specific relation expressed. The most general ones of the features in the first of these sets define what we call a domain of application: [locative], [directional], [temporal], [causal], and [intentional]. The second of the above operations, projection, accounts for the transition from one domain to another. The process of projection would seem to rest on the principle that the new is conceptualized in terms of the familiar, and thus, in general, that more abstract domains tend to be conceptualized in analogy to more concrete ones. This is illustrated by the transition from reference objects conceptualized as points of location to reference objects that are conceptualized as points in time.


The first applies to concepts within the same domain. Thus a locative concept is extended to another locative one, a temporal concept is extended to another temporal one, etc. The transition from one concept to the next thus takes place gradually. Features are successively changed so that at can be applied to successively further removed kinds of relationships within the same domain. Let us, in order to describe the relationship expressed by at in a phrase like "x is at y", call x the located object and y the reference object*. A feature characterization for a particular sense of at may then be formulated as a triple of

387

2. THE CONCEPT OF AT

2.1 The prototype: simple point of location The prototypical locative sense of the preposition at is illustrated by the following sentence: (1)

You are just at the point where three welfare districts join. 2

Here the preposition at expresses a two-place relation between the located object (you) and the reference object (the point where three welfare districts join). The two central characteristics of this core sense of at are that -

the reference object is a point, and that the located object is in the immediate vicinity of this point.

The application in sentence (1) clearly displays these two properties: the reference object is a (zero-dimensional) topological point with the located object in its immediate vicinity. If reference object and located object were to coincide the preposition on would have to be used instead, as in (la): (1)

a. You are just on the point where three welfare districts join.


In the following the motivation for, and the relationship between, the various senses of the preposition at will be discussed and illustrated. The analysis will not, however, contain the full feature specifications. It will only describe the factors that motivate the non-prototypical applications of at. In our description of the overall concept of at, we will take a synchronic perspective. This implies that some applications will not be accounted for within our paradigm. At the hands of, for example, is one of the few remnants displaying the once productive conceptualization of the reference object containing the feature [+source]. While some such phrases have currently lost their motivation entirely and have become idiomatic, others receive an interpretation different from the one that motivated those uses historically. In the adverbial at stake, at has lost its original locative sense and may be seen as synchronically motivated by the concept of "state of affairs or mind", which we can also find e.g. in at ease, at one's leasure, or at work. Meaning differences between the concept oi at and those of other prepositions will be discussed at points where the domains of application overlap and where the choice of one rather than the other preposition determines the specific interpretation of the sentence.

388

(2)

Maria was standing at the window.

In such applications, however, the actual size and shape of the reference object are irrelevant, as is shown by the following paraphrase, where the relation to be expressed is still the same: (2)

a. Maria was standing at a point very close to the window.

Thus the reference object can take any shape here, as long as one can still conceptualize the located object as being at a point in the direct vicinity of the reference object. In the following we will discuss only the prototypes of each concept, without going into further extensions within once concept, which eventually lead to the development of new concepts. It should be born in mind, however, that - concerning extensions within the same domain - transitions from one concept to the next take place gradually. Boundaries between the single concepts will thus always be fuzzy and not as clear cut as it might appear from merely considering the respective prototypes.


This characterization of the core sense or prototype of at contrasts with a proposal rhade by Herskovits (1986), who assumes for the ideal meaning of at "for a point to coincide with another" (1986:50). But this notion would seem to be incorrect. When reference object and located object coincide rather than being in each other's immediate vicinity, we lose the distinction between at and on. A similar problem occurs for derived senses of at, especially in the domain of directionality, which, as will be argued below, are equally impossible to distinguish from the meaning of other prepositions without reference to the notion of immediate vicinity. It should be noted that the determination of a prototype is not guided by the quantity of actual occurences of the relevant expression in that specific sense. It is rather its specific feature structure which constitutes the prototypicality of a particular word-sense. And this also holds for the preposition at we may in fact not find very many instances in which the reference object actually is a topological point. Nevertheless we experience this realization as the best example of the overall concept of at. The application of at in sentence (1) actually should be considered a prototypical instantiation of the prototypical concept of "simple point of location". Just as the general concept of at is structured around the prototypical core meaning (as characterized by the two properties given above), also each single concept consists of a prototypical sense and less representative members. Further, non-prototypical instances of the prototype of "simple point of location" are relations where the located object is still found in the immediate vicinity of the reference object, but where the latter is not a zero-dimensional point but a physical object that has a particular shape, cf. e.g. (2):

389 2.1.1. Locative extensions Various extensions now emerge out of the prototypical concept "simple point of location". Along one chain the notion of a point is sucessively extended to applications in which the reference object denotes more complex entities. This comprises relations in which the shape of the reference object does not any more permit a paraphrase by "...at a point very close to...", but would rather require the application of the preposition on (because the reference object is conceptualized as a surface, cf. (3), or in (because the reference object is conceptualized as a three-dimensional space, cf. (4)): My friends had a party at the beach yesterday.

(4)

The meeting at Manchester was illegal.

The extension to the concept of "extended location" is motivated by a conceptualization of the relevant reference object "simply as a place on the map" (Leech and Svartvik 1975:85). There are, however, limitations to this type of extension. With certain verbs that specify the relation between located object and reference object a reduction to a simple point of location in the conceptualization of a larger area is not possible: (5)

*Peter has lived at London for some time.

That live, for example, blocks the application of at when the reference object is a city would seem to be due to the fact that here the more dominant conceptualization of the city as a three-dimensional space is triggered, which requires the use of the preposition in. 2.2. Intentional concepts A further variation in the features of the relation results, at least for appropriate reference objects, in the concept of at that expresses an "intentional location". This extension can be seen as a projection of the above two concepts of "simple point of location" and "extended location", in the sense that the speaker refers to exactly the same point of location, either simple or extended, but emphasises the respective regular intention that goes with the relation of being located at the relevant reference object. Cf. (6)

Peter was waiting at the bus stop.

(7)

The children are at school.

(8)

Tom and Jerry are at the zoo.


(3)

390 In (6) the idea of waiting for the bus prevales over the mere indication of the place where Peter was waiting. In fact, the sentence (6)

a. Peter was waiting by the bus stop.

(9)

Maine and Spender are hard at work right now.

(10)

As it is, there are very few places left now where one can be at ease.

The af-phrases in these sentences do not any more answer a wAere-question, but refer to a certain "state of affairs or mind". 2.3. Scale concepts The concept of a scale forms the basis fora number of extensions. The notion of a point is applied here in the specific sense of a point on a scale. Also Herskovits notes that "the range of uses of at indicates that at is a graded concept" (1986:51). Evidence for this view is provided by the supple-


places Peter at exactly the same point of location as (6), but does not trigger the desired intentional interpretation. Sentence (7) expresses the idea that the children are learning. If one wanted to specify merely the point of location without this intentional reading a phrase such as, "... inside the school building" would be appropriate. In sentence (8) the use of at indicates that Tom and Jerry are visiting the zoo. Using the preposition in here would make it more likely that they are zoo animals. This difference nicely illustrates the distinction between, as it were, a bare locative relation on the one hand, and a locative relation that is viewed from a particular intentional aspect on the other. The abstract "intentional location" concept of at is an instance of a doubly motivated concept. Both the topological point prototype and its locative extension motivate a mapping onto this abstract domain, where the intentional aspect dominates over the actual dimensions of the reference object. Tne extension from one concept to another does not always occur linearly, but often the development of a new concept is guided by extending the characteristic features of a specific concept plus incorporating features of another. It is this fact in particular that supports the idea of taking a prototype approach to the problem of representing wordmeaning. With a set-theoretical classification a definite description, which must hold for each single instance, will decide over the membership in a certain category. But no such classification could account for the fact that in some applications we find more and in others we find less of the characteristic features which determine the meaning of a specific word, as in our case of the preposition at. In an extension of the concept "intentional location" the idea of denoting a landmark is completely abandoned:

391 mentary use of modifiers such as exactly, precisely, etc. Her understanding of the graded concept, however, comprises even those realizations which we classified under the prototypical concept of "simple point of location" (see (11)) or its locative extension to "extended location" (see (12)): (11)

The car is at the corner of 6th and Broadway.

(12)

There is a Film of oil at the surface of the water. (Herskovits 1986:51)

(13)

He could pass us by at a thousand yards, and we still would not miss him.

While in this sentence the underlying unit, yards, is of a concrete nature, an abstract scale concept on the basis of speed-units is realized in the following application: (14)

Why are those ambulances driving around at breakneck speed?

This "abstract scale" concept, a projection from the corresponding concrete concept, furthermore comprises gradings on the basis of price-units (buy something at a certain price), moods (keep the morale at fever pitch), etc. For the concrete as well as for the abstract scale concept it holds that the scale itself can be horizontally or vertically oriented. The conceptualizations of the end-point of an arbitrary scale reflect these types of orientation accordingly - e.g. at the beginning, at the bottom. Furthermore phrases such as at best, at most, at least, or at all also belong to the abstract scale concept. 2.3.1. Temporal Concepts Also the temporal concepts of at could be viewed as conceptualizing a reference point on a scale, consisting of varying time units. The reason for


The reason for this is, as mentioned above, that she considers any application of at to express a relation in which the located and the reference object coincide more or less. The actual distance then is determined by the nature of the objects in question. Since she only considers the spatial senses of a/this generalization may be justifiable to some extent, although it leads into problems already in the locative domain, with the distinction between at and on, as we saw above. Things get worse, however, when we also want to account for the abstract senses of at. It will be very hard for example to justify a graded concept as a basis for the abstract concept of "intentional location". In the present analysis the idea of a graded concept will therefore only be assumed for those cases which - because of the kind of object involved suggest an explicit scale, divided into particular units. The scale concept employed in sentence (13) is an instance of the locative "distance" concept, a centrally-motivated sense:

392 discussing them separately lies in the observation that they additionally display a structure analogous to all of the three locative concepts. The correspondence can be summarized as follows: - point of location versus point of time, - local distance versus temporal distance, - expanded location versus period of time. Examples of these temporal concepts of at are found in (13)—(15) in corresponding order: It was Sunday, August 25th at 3:32 standard British summer time.

(16)

Well, this is the best I could do at short notice.

(17)

He couldn't study at night because he was afraid of the dark.

With these temporal concepts we find again the phenomenon of multiple motivation. On the one hand they can be seen as further instances of the concept of an abstract scale, just as the domain of price, velocity, etc. On the other hand, they may also be viewed as projections from the corresponding locative relations, which clearly sets them apart from other abstract scale realizations. A further reason for treating them separately is that they motivate further applications. 2.3.2. Causal concepts As an extension of the abstract concept "point of time" one can view the development of the causal concept of at. Consider the following opposition: (18)

Peter left the house at 6 o'clock.

(19)

Peter left the house at his father's signal.

In (19) the aspect of denoting a point in time recedes in favour of the more prominent causal component. The latter can trigger any kind of action as in (19), or an emotion as in (20): (20)

My husband and I confess ourselves extremely surprised at such unrefined behaviour.

2.4. Directionality Another concept of at, which requires a reference object that is conceptualized as a goal, again emerges directly out of the prototype "simple point of location". When the reference object is a point of location, its interpretation,


(15)

393 however, is restricted to a position reached after a certain movement, journey, etc.: (21)

The train arrived at Victoria Station.

While the movement itself is not included in the a/-relation tied to this "goal" concept, it does form part of the relation in its extension to reference objects conceptualized as a "target": (22)

Bella snatched at the lipstick.

(23)

Have you looked at the newspaper headlines today?

Again the reference object, within this chain of extension the "focussed object" (cf. Radden 1981:154), is not affected. This is well demonstrated by the opposition between (24) and (24a); cf.: (24)

John has been talking to his neighbour for hours.

(24)

a. John has been talking at his neighbour for hours.

While with the use of the preposition to the neighbour is seen as an interlocutor, the use of at indicates that only John has been talking. The application in (24a), in fact, lies on the borderline to the concept of "abstract intentional target" (see below), since the speaker could also - instead of merely being carried away in talking - have had the intention of hurting or humiliating the other person. Within the abstract target concept the reference object is conceptualized as a point of focus for the relevant action or event. In an extension of this concept of abstract target we find again the principle which we already saw in an extension of the locative concept of at, i.e. the extension from a reference object that is a point of location to a reference object that is a two- or


Within this directionality chain of extension the criterion of immediate vicinity plays again an important role: the reference object remains unaffected by the relevant af-relations. The conceptualization stops, as it were, immediately before the reference object. Thus sentence (22) is to be interpreted in the sense that Bella reached out for the lipstick, but most likely did not get it. And although (for technical reasons) a train arriving at Victoria Station will usually enter (and in this sense affect) the station, this is not a matter of the af-relation. A bus arriving at the station makes this point clear. Modification of the kind of movement in the "target" notion of at yields a further extension of the concept. While in (22) we are concerned with actual, physical movement, sentence (23) - the prototypical realization of an "abstract target" notion - is concerned with an abstract kind of movement:

394 three-dimensional area. Just as for the locative uses, the actual dimensions of the reference object are irrelevant for this concept of "focussed domain", since the reference object only serves as a point of reference. Examples of the focussed domain concept often refer to a capability of some sort, as in (25): (25)

Bill is quite clever at organizing things, but not much good at anything else.

The same distinction as in (24) and (24a) can be found in the concept of "intentional target", the second extension of the concrete target concept; cf. (26) vs. (26a): Peter threw the ball to Mary.

(26)

a. Peter threw the ball at Mary.

Again, the difference between these two applications lies in the fact that in one case the reference object is seen as a partner, while in the other it is excluded from the action and is seen only as the target the respective action is aimed at. Thus in sentence (26) the ball was thrown for Mary to catch, whereas (26a) suggests a negative intention accompanying the throwing of the ball - it was meant to hit Mary. By removing the component of actual physical movement towards the reference object while maintaining the intentionality feature, the intentional target concept is extended to the concept of "abstract intentional target"; a concept that could equally well be construed as resulting from the abstract target concept by adding the intentionality feature. The concept of at with an abstract intentional target as a reference object is illustrated in sentence (27) (which also shows that the intention need not always be negative). (27)

She kept smiling at him all evening long.

To what extent the abstract intentional meaning of at is productive cannot better be demonstrated than by the following sentence: (28)

On no account allow a Vogon to read poetry at you. (Adams 1982:45)

How could one have told the reader any better than by the use at, which stands in direct opposition to the preposition to here, that Vogon poetry was unbearable and that it was one of the most terrible tortures when a Vogon read it at someone? In fact, it is by these kinds of extensions that the whole concept of at is subject to constant change. Certain applications may gradually acquire different interpretations, or, as mentioned in the beginning, whole sense chains may even become obsolete, as was the case with its meaning component "source".


(26)

395 2.5. Target and cause The last concept to be discussed here establishes the link between the target concepts and the scale concepts. This extension, doubly motivated by the abstract intentional target concept as well as by the concept of cause, is mainly found in the context of emotion predicates: (29)

I'm not sick at you, just sick with fear.

(30)

*Susan is mad at the earthquake.

(30)

a. Susan is mad at her car.

The overall concept of the preposition at is presented in the diagram, illustrating how the various extensions are chained to one another. It shows the basic concept of "simple point of location" as the best and most central case which displays the most characteristic features of at: the notion of a (dimensionless) point as reference object, and the location of the located object in its immediate vicinity. All of the extensions display a gradual decrease in prototypicality. Nevertheless, each single concept of at still bears enough of the basic characteristics to be clearly distinguished from other concepts such as on, in, or to, as has been shown with selected examples.

CONCLUSION

By assuming the notion of a prototype in order to represent the overall concept of the preposition at, we have been able to give a plausible model for how its various senses are related to each other. An important observation that emerges from this analysis is that the extensions from one concept to another may take place gradually, via intermediate concepts.


The conceptualization here shows a complex structure: The reference object, here the addressee, is at the same time the cause and the target of the speaker's emotion. Within this concept the reference object carries the feature [+ animate], which it inherits from the motivating abstract intentional target concept. The reason for this is that it is senseless to direct ones emotions at an inanimate object, which can neither be held responsible for its annoying behaviour, nor show any kind of reaction to the aroused emotions (cf. Radden 1985:183), unless, of course, there is a possibility for personification. Accordingly, (30) - with no way of personifying an earthquake - is unacceptable, whereas (30a) is fine. Cf.

POIHT IH TIME

at 5O mph

ABSTRACT SCALE

at school

STATE OF AFFAIRS/HIHD at ease

IBTEHTIOHAL LOCATIOH

EXTENDED LOCATIOH

night.

atHanchester

at

PERIOP OK TIME

throw something at

IHTEHTIOHAL TARGET

o-

abstract

- concrete

clever at

FOCUSSED DOMAIH

snatch at

TAROET

look at

ABSTRACT TARGET

smile at, shout at

ABSTRACT IHTEHTIOHAL TARC3ET

o—o

at a distance

notice/~~~\

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

Recommend Documents