JOURNAL OF SEMANTICS AN INTERNATIONAL JOURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE MANA...
26 downloads
413 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
JOURNAL OF SEMANTICS AN INTERNATIONAL JOURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE MANAGING EDITOR: PETER BOSCH (IBM Germany) R E V 1 E W E D I T O R : BART GEURTS (IBM Germany)
EDITORIAL BOARD: PETER BOSCH (IBM Germany) SIMON C. GARROD (Univ. of Glasgow) BART GEURTS (IBM Germany)
PAUL HOPPER (SUNY Binghampton) LAURENCE R. HORN (Yale University) STEPHEN ISARD (Univ. of Edinburgh) HANS KAMP (Univ. of Stuttgart)
LEO G. M. NOORDMANN (Univ. of Tilburg) ROB A. VAN DER SANDT (Univ. of Nijmegen) PIETER A. M. SEUREN (Univ. of Nijmegen)
CONSULTING EDITORS: R. BARTSCH (Univ. of Amsterdam) D. S. BREE (Erasmus Univ., Rotterdam) G. BROWN (Univ. of Cambridge) 0 . DAHL (Univ. of Stockholm) G. FAUCONNIER (Univ. of California, San Diego) P. N.JOHNSON-LAIRD (MRC, Cambridge) SIR JOHN LYONS (Univ. of Cambridge)
J. D. MCCAWLEY (Univ. of Chicago) B. RICHARDS (Imperial College, London) H. SCHNEIXE (Ruhr Univ., Bochum) M. STEEDMAN (Univ. of Pennsylvania) Z. VENDLER (Univ. of California, San Diego) Y. WILKS (New Mexico State Univ., Las Cruces) J. VAN BENTHEM (Univ. of Amsterdam)
H. E. BREKLE (Univ. of Rcgcnsburg) H. H. CLARK (Stanford University) H.-J. EIKMEYER (Univ. of Bielefeld) J. HOBBS (SRI, Menlo Park) D. ISRAEL (SRI, Menlo Park) E. L. KEENAN (Univ. of California, Los Angeles) W. MARSLEN-WILSON (MRC, Cambridge)
H. REICHGELT (Univ. of Nottingham) A. J. SANFORD (Univ. of Glasgow) A. VON STECHOW (Univ. of Konstanz) D. VANDERVEKEN (Univ. of Quebec) B. L. WEBBER (Univ. of Pennyslvania) D. WILSON (Univ. College, London).
EDITORIAL ADDRESS: Journal of Semantics, IBM Germany Scientific Center, IWBS 7000-75, Postfach 800880, D-7000 Stuttgart 80, W . Germany. Phone: (49-711-) 6695-559. Telefax: (49-711) 6695-500. BITNET: boschQdsolilog. New Subscribers to the Journal of Semantics should apply to the Journals Subscription Department, Oxford University Press, Pinkhill House, Southfield Road, Eynsham, OX8 iJJ. For further information see the inside back cover. Volumes 1-6 are available from Foris Publications Holland, PO Box 509, 3300 Am Dordrecht, The Netherlands.
Published by Oxford University Press
Copyright by NIS Foundation
ISSN 0167-5133
JOURNAL OF SEMANTICS Volume 7 Number i SPECIAL ISSUE O N PRACTICAL ASPECTS OF SEMANTICS Guest Editors: Gillian Brown, Simon C. Garrod
CONTENTS D. S. BREE, R. A. SMIT and J. P. VAN WERKHOVEN
Translating Temporal Prepositions between Dutch and English
i
JENNIFER COATES
Modal Meaning: The Semantic-Pragmatic Interface
$3
PETER WRIGHT
Using Constraints and Making Reference in Task-Oriented Dialogue
65
S. B. BARTON and A J. SANFORD
The Control of Attributional Patterns by the Focusing Properties of Quantifying Expressions
81
GERARD HEYER
Semantics and Knowledge Representation in the Analysis of Generic Descriptions Book Reviews
93 111
Editors' Preface This issue draws together four papers on what we have called practical aspects of semantics. In contrast to more traditional approaches to meaning, we see practical semantics as attempting to take into account the richness of language when used in real communicative contexts. Perhaps the most significant feature of these papers is that they deal directly with real as opposed to idealised language and so avoid the restrictions on the kinds of language structures and features which have tended to dominate linguistic theorizing. Such restrictions have taken several forms: the use of artificial or invented utterances to illustrate semantic phenomena, the written language bias in research and the low profile of work on interactional aspects of meaning in natural dialogue. The papers in this selection set out their various ways to go beyond these restrictions, and in doing so extend the scope of semantic enquiry. Quite apart from this theoretical benefit, such an approach may also yield results of real practical significance, for instance in the design of effective computer-based language processing systems which must by their very nature confront practical problems of language and its use in communication. Although the four articles share this common practical basis, they employ very different methodologies, and look at different forms of language use. The first paper by Bree, Smit and Van Werkhoven investigates the semantics of Dutch and English temporal prepositions. Its practical basis comes from two sources, first the overall motivation to produce an effective computational algorithm for translating between the prepositions and secondly from an extensive analysis of two actual texts and their translations. The second paper by Barton and Sanford is also concerned with written language, but in this case with the specific rhetorical effects of using different quantifiers and frequency adverbials. The experiments reported in the paper demonstrate that readers will draw different attributional inferences about who or what is responsible for an event as a function of the choice of semantically equivalent terms in its description. This is explained in terms of the way the different expressions focus the reader's attention onto different sets of individuals. The practical significance of such studies lies in extending considerations of the meaning of quantifiers into the rhetorical domain. The remaining two papers by Wright and Coates depend upon the analysis of natural dialogue and are both concerned with interactional contributions to meaning and interpretation. Wright discusses the extent to which conversants actually follow the linguistic strictures on the use of definiteness to signal shared knowledge of a referent His results indicate that there is still some way to go in relating established theories
iv Editors'Preface
of reference to actual conversational usage. The final paper by Coates addresses the very different area of modality in conversation. Her main conclusion is that consideration of how the two speakers interact is essential to a full analysis of modality. So this group of papers considers a wide spectrum of fundamental problems in semantics but all from a practical point of view. Looking to the future, such work raises the challenge of how to extend current semantic theory and formalisms so as to accommodate the richness of language in everyday use and the various ways communicators manage their interactions. We hope that practitioners of both this practical approach and the more traditional formal approach to semantics will be able to recognise the mutual significance of each others contributions. G. BROWN S. GARROD
Journal ofSemantics 7: 1-51
© N.I.S. Foundation (1990)
Translating Temporal Prepositions between Dutch and English1 D. S. BREE, R. A. SMIT and J. P. VAN WERKHOVEN
Abstract
1. I N T R O D U C T I O N Learning a second language presents many challenges. One of the most exacting is mastering the use of prepositions and subordinating conjunctions. Some are trivially easy, e.g. the Dutch translation of AFTER is almost always NA or, when it is used as a subordinating conjunction, NADAT. For others, there are simple rules, e.g. the Dutch translation of WHEN is TOEN for unique events in the past, and ALS otherwise. However, sometimes the choice between two such function words (preposition or subordinating conjunction) seems to be arbitrary, e.g. the English translation for ALS used temporarily (ALS has many other uses) may be WHEN or AS, but no clear-cut rule is available for choosing between the two. Sometimes both seem appropriate: And I did look like a stranger when/as I walked up their street.
but on other occasions one or other is infelicitous: When/as* I returned she was desperate As/when(?) I departed with my box, he told me. . .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The decision trees for selecting the appropriate temporal conjunction or preposition in English and Dutch, developed in an earlier paper, are tested. Data are from the translations of parts of two books, one English and one Dutch, into Dutch and English respectively. The analysis of the data has led to the complete recasting of the original selection trees. The new trees are based on about a dozen different attributes that are needed to classify the normal, non-idiomatic use of almost all the temporal conjunctions and prepositions in both languages. Some of these attributes are: time point v. period; simultaneity v. order; relative order of the matrix and sub events; whether or not the Time of Discourse is used to mark the end of a period, etc. Among other details we have been able to show how the conjunction AS, in its temporal use, is not ambiguous between WHEN and WHILE as was originally thought. There is a large overlap between the selection trees for both languages, as well as some interesting differences. These trees have been specified in sufficient detail to be readily incorporated in natural language computer programs.
2 Translating Temporal Prepositions between Dutch and English
For other prepositions with multiple translations, native speakers have definite judgments about correctness, e.g. the Dutch translation for temporal AT depends on the noun phrase that follows: at eleven o'clock ->• otn elfuur at other times ofthe day -* op andere tijden van de dag at the end of the day —• aan het eind van de dag
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
While such alternatives are offered in dictionaries, no rules for choosing between them are given—neither in dictionaries nor in grammar books. The general consensus is that no rules are to be found; non-native speakers have to go through a process of trial and error learning the uses of each such preposition as it is encountered. We believe, however, that rules can be found, although we expect them to be less simple than those for choosing, for instance, between TOEN and ALS. We propose to use decision trees, with choice criteria at their nodes and prepositions at their leaves, to represent the rules for selecting the function words in one language. The aim of our research programme is to establish such selection trees for both subordinating conjunctions and prepositions in any language, beginning with Dutch and English. Such trees can be tested against corpora, and in particular against the translations made of such prepositions by professional translators. Once established, the selection trees can be incorporated into any natural language translation program that is based on the principle of first 'comprehending' the original text before producing the translation. This restriction is necessary, as the criteria occurring at the nodes of the selection trees may be different for each language. Semantic information from the context of the function word may be required to determine which one to use in the translation. This information may well be different for the pair of languages under consideration, as we have just seen is the case for the translation of AT into Dutch. In this paper we propose selection trees for English and Dutch temporal positions and subordinating conjunctions. We begin by recalling an earlier model for the selection of temporal conjunctions and prepositions. Then we present some data on translations between Dutch and English with which we test this model. Lastly, we propose a revised version of our model.
Bree, Stnit and van Werkhoven 3
2. THE ORIGINAL SELECTION TREES
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Our aim in carrying out this research is to establish a description of the meaning of the temporal function words in such a way that these meanings can be included as part of the semantics of any natural language computer program. We needed a formalism that was precise but not one that was too restrictive, otherwise aspects could not be represented that would be needed for a particular application program. Furthermore, we looked for a representation that was aesthetically satisfying, and one that would provide an overview from which an explanation of why some of the possible temporal relationships were conveyed by words and others were not. The model that we previously presented (Bree & Smit 1987) was based on what we called selection trees, a variation on conventional decision trees. A selection tree is a tree with nodes and branches. There is one root node, which is the starting point for the processing required to select the function word appropriate for the task at hand. The leaf nodes represent the function words; all other nodes are choice nodes representing tests on what the speaker/writer wishes to convey about the time of the event or state described in the matrix clause. The links between nodes are labelled with the answers to the tests that are posed in the nodes. We have chosen selection trees rather than two other possible means of representing the meaning of temporal function words—semantic features and the first order predicate logic. Selection trees provide sufficient constraint on the representation to give structure to the process of finding an interesting pattern in the data. We ourselves (Bree, Smit & Schotel 1984) as well as others (Bennet 1975) have used semantic features for representing temporal function words. The use of semantic features brings with it one major problem: there are no words for some of the combinations of the features and there is no satisfactory reason why this is the case. The other serious alternative representation isfirstorder predicate logic, e.g. Rohrer (1977), Herweg (1987). One can translate most things into first order predicate logic. However, it does not give a very satisfactory overview of why some meanings are given names in the form of temporal words, and other possible meanings are not. In short, the first order predicate logic does not constrain the search for pattern at all. Furthermore, formalisms tend to require concentration on the precise details of the onset of the various events, but this aspect is not so carefully specified by the usual temporal function words (Heinamaki 1978). Schemes based on an a priori ideal of the different functions that temporal prepositions need to fulfil, as the one given in Quirk et al. (1985: 528), require
4 Translating Temporal Prepositions between Dutch and English
that the prepositions are used in more than one category, depending on the context in which they are found. For example, Quirk et al. propose the following categories: time position, forward span, backward span, duration, frequency and relationship. Many temporal prepositions fall into more than one such functional category, e.g.: — IN: position, frequency, — UNTIL: forward and backward span; — FOR: duration and forward span.
— extent: — dimension:
The Concordfliesto New York in three hours. The Concorde is departingfor New York at 5 o'clock.
This is a very simple distinction.
a. Extent oftime Working first along the extent branch, the first (uninteresting) distinction is whether the extent of time of the matrix stevent is the same or less than that of the extent of the sub stevent. The second distinction hinges on the nature of the matrix verb in the construction: — with accomplishment2 verbs use IN, e.g.: The Concorde flies to New York in 3 hours. — with activity verbs use FOR, e.g.: I amflyingto New Yorkfor 3 hours. In Dutch the same distinctions hold. However, there is also an extra distinction with accomplishment verbs: ones that indicate the accomplishment will be completed in a certain time from the here and now. In such cases, Dutch uses OVER instead of IJV.3
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We regard the classification of any preposition in more than one category as a mark of weakness for the categorisation method being used, although we ourselves have not always been able to avoid this (see Dutch VOOR in Tables i and 2). So we keep on using selection trees. Figure 1 shows the proposal made for English temporal function words, and Figure 2 shows the equivalent for Dutch (Bree & Smit 1987). These trees form the starting point of our analysis. After describing them in some detail, we will turn to our translation data to test their adequacy. We now go through these figures in some detail, using the English tree as base and pointing out the differences for the Dutch tree. The first distinction that we made was between extent of time and location on the time dimension. Either one wants to say how long an event or state (which we will call a stevent) lasts, or you want to put the stevent on the time axis, to assign it a place on the time dimension, e.g.
M ?S
-in a activity -for
-within event
s is
S gives
point
period dimension \
^ M ?S
^ 35
at around
oaded from jos.oxfordjournals.org by guest on January 1, 2011
achievement Main verb
by states throughout y^
M ?S
point I
Sis -after order
border
-before
on/0
TOR update?
•^—during - in the
endt$\ both ends?
M S TOR
Main stevent time
——between... and ... other end ~ (TOR) | +
- from • since
Sub stevent time Time Of Reference
3
order other end (OTR)
6 -until
o
Figure I Selection tree for English temporal function words
I
S from
TOR
activity
+
oaded from jos.oxfordjournals.org by guest on January 1, 2011
achievement Main verb
I 1 terwijl
5.
1 X 2a O 13 fn
both ends? M S TOR
Main stevent time Sub slevent time Time Of Reference
other endl ~
N, N
i
(TOR) 1 +
van (aQ sinds
order tot fdat)
Figjure 2 Selection tree for Dutch temporal function words
Bree, Smit and van Werkhoven 7
b. Point oftime Returning to the root of the tree, we now explore the dimension branch. Along this branch we find all the temporal function words which place the time of the matrix stevent on the time axis. The first choice that is to be made is whether the sub stevent gives: — a point on the time dimension, or — a simple ordering in time between the matrix and sub stevents, or — the ends of a range of time in which the matrix stevent must fall.
— the point is some kind of event, e.g. As theJones's were leaving ..., or — the point is a time point (sic), or — the point has some kind of extension, i.e. is a period. If it really is a time point, then it only remains to decide whether: — the matrix stevent occurs precisely at the same time as the sub-stevent (=), e.g. AT NOON, or
— roughly at the time time (-) as the sub-stevent, e.g. AROUND NOON, or — just before (.). The translations are indented. To take the first example from First love, last rites and its translation:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The Dutch selection tree differs slightly at this juncture. Dutch does not make the distinction between TO and UNTIL, using ror for both. Note that using the Time of Reference (TOR) rather than the time of utterance avoids classifying temporal range prepositions as being both 'forward span' and 'backward span' (Quirk etal. 1985: 533-40). We will retain this in our further analysis but will refer to it as Time of Discourse (TOD) since the analysis has been carried out on stories. The TOD is the time that has been reached in the discourse. In the story, references can be made to events occurring both before and after the TOD, without altering the TOD: e.g.:
Bree, Smit and van Werkhoven 9
(9.04) .. . Connie. . .sitting on the edge ofthe bath weeping, while I filled the sink with warm water. .. (7.03) . .. Connie.. .opde rand van het bat zat te huilen, terwijl ik degootsteen Het vollopen met warm water. . .
VAN.
The results of this analysis are shown in Tables 1 and 2. In total there are more than 900 prepositions used temporally. Table 1 shows the number of rimes that an English temporal function word in First love, last rites, shown in the column headings, is translated into a particular Dutch word, shown by the row labels. Table 2 shows the number of times that one of the Dutch temporal function words in Turks Fruit, shown in the row labels, is given a particular English translation, shown by the column headings. What do these data say about the selection trees? First of all, many (58%) of the translations were as predicted. The translations of AS and WHEN in English were not given any Dutch equivalents. They should be translated by TOEN, ALS or WANNEER. More importandy, where predictions failed, then the alternative chosen was generally along the same sub-branch as the predicted word. But there were exceptions. These have led us to rearrange the order of the function words, for both Dutch and English, in a different way from the one that would be a logical consequence of the selection trees shown in Figures 1 and 2. By logically, we mean that words that appear close in the trees, appear close in the tables. We have chosen the order in the tables to indicate how the selection trees should be revised. This order is such as to group most of the nonzero entries around the main diagonal in both tables. Boxes have been drawn around these main diagonal clusters.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Note that the temporal conjunctions are in bold. We looked at all the temporal prepositions and conjunctions in those two sample texts to see how they were translated. If our selection trees were perfect predictors then, for example, every time SINCE occurs in English, used temporally, SINDS should occur in the Dutch translation, and vice versa. However, we did not expect this degree of perfection in our predictions. The correct use of prepositions is notoriously difficult for non-native language speakers. What we did expect was that where the predictions failed, the alternative translation would be a word close to the correct word in the selection tree, i.e. if SINCE was not translated by SINDS, it would be translated by
io Translating Temporal Prepositions between Dutch and English
Table I Number of occurrences of different translations from English temporal prepositions into Dutch, in First Love, Last Rites. From the English: A S
w
w
H E N
H I L E
3 25 77 2 65 5 20 2 33 1
ZODRA TOEN ALS WANNEER TERWIJL ZOLANG
3 1
2
0 N
I N
D U R I N G
F 0 R
A F T E R
5
s
F R 0 M
i N C E
1
1 1 16
1
TlJDENS
Bu
4 4 1 3
1
1 1
GEDURESDE VoOR* ...LANG AL
31 10 2
2
1
3 6 9 4
2
NA(DAT) DMRNA SlNDS SEDERT VAN VANAF
1
1 1
1 31 7
34 7 7
1 7 1
1
i
i.
1 1
2 4
TOTAL
>>
1 2 24 2 3 2 23
VOOR(DAT) EER(DAT) TOT(DAT)
Other* Nonec
39 15 7 4 5 6 9 6
3 1
6 3
2 4
1 8
1 1 29
45 16 24
57
8 55
16 2
58 160 37
2 1
In the duralive rather than order sense. Other word, not a preposition. Preposition not translated.
3 114 68 6 55 1 4 10 6 22 1
1
1
's
c
T 0 T A L
T 0
N T I L
3
1 1 5
3
IN
a
u
B E F O R E
1
3 7
OM AAN TEGEN OP OVER
B Y
1 3 45 li
14 3 2
1
45 59
1
47 26
26 3 26
2
593
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
To the Dutch:
A T
Bree, Smit and van Werkhoven 11
Table 2 Number of occurrences of different translations from Dutch temporal prepositions into English, in Turks Fruit. To the English: A S
From
w
w
H E N
H I L E
ZODRA TOEN ALS WANNEER TERWIJL
1 7 80 2 23 1 7
0 N
3
1
7
1
1 N
D U R I N G
F 0 R
s
A F T E R
B E F O R E
1 N C E
2
N
t h
0
e r
n e
4
5 6 11
5
's BlNNEN TlJDENS ONDER BIJ VoORt ... LANG AL
1 1
3 2 17 4
1
1
1
1
9
5
5 3
i
1 1 1 2 2
1 1
NA NADAT DAARNA SlNDS VAN
1
1
1 2
1
1
1 15 3
4
3 1 2
23 5 6 1 1
1
1
23 12
23 10 1 25 54
337
21 1 15
7
Other word, usually not a preposition. Preposition not translated. In the durative rather than order sense.
28
5
6
26
1
35 9 0 3 2 6 1 2 18
3
1
19 106 11
2 0 0 22 2
7
1
TOTAL
1 103 33 1 26
7 1
VOOR(DAT) TOT(DAT)
T O T A L
b
3 2
7 2
IN
* >>
0
2
OM AAN TECEN OF OVER
c
U T N O T I L
16 i
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
the Dutch:
A T
12 Translating Temporal Prepositions between Dutch and English
a. Simultaneous events The top left boxes in Tables i and 2 show the translations between temporal subordinating conjunctions that indicate the time of the matrix event is approximately the same as that of a sub event. The predictions were that: — WHILE is equivalent to TERWIJL; — WHEN with the past tense will be translated by TOEN; — WHEN with the present or future tense will be translated by ALS or WANNEER.
— WHILE was indeed translated by TERWIJL but TERWIJL was only sometimes translated by WHILE; — WHEN was indeed translated by TOEN, ALS or WANNEER and vice versa. The difference between TOEN and ALS is that TOEN is used for single events in the past and ALS and WANNEER are used for everything else which includes repetitive events in the past. Because of the low frequency of the occurrences of WANNEER, no analysis will be made of the difference with ALS and it will not be mentioned in the rest of this paper; — AS was translated not only by TERWIJL but also by TOEN. SO we need to determine the conditions for choosing between TERWIJL and TOEN; — AS SOON AS was always translated by ZODRA and vice versa. Looking at this global pattern we could conclude that:6 — AS in English is ambiguous between WHEN and WHILE;
— WHEN is equivalent to TOEN, ALS or WANNEER, depending on the tense of the matrix clause; — WHILE is equivalent to TERWIJL Problems in translation will come when choosing between: — TOEN/ALS or TERWIJL for the English AS; — AS or WHEN for the Dutch TOEN/ALS; — AS and WHILE for the Dutch TERWIJL. So we first examine the difference between WHEN and WHILE. (a) WHEN and WHILE Initially, the difference appears obvious; WHEN introduces a point in time at which the matrix stevent occurs, whereas WHILE introduces a period in time in which the matrix stevent occurs. As Quirk etal. (1985: 1083) point out, WHILE
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
No prediction was made for the translation of AS, although the dictionary gave TERWIJL,4 and As SOON AS is clearly equivalent to ZODRA. If we look at the data we see that mainly:5
Bree, Smit and van Werkhoven 13
must govern a durative clause. While this difference is not obvious in all examples, e.g. the difference between / was robbed while I was on vacation. I was robbed when I was on vacation.
is marginal, there are clear-cut examples of the difference, e.g. in: (58.14) Jasmin came down the steps, and when he reached the stage he spoke softly. "Jasmin came down the steps, and while he reached the stage he spoke softly.
interpretation, unless the progressive form is used: When the tomatoesripened,they were plucked. PWhile the tomatoesripened,they were plucked. While the tomatoes ripened, the plums did not. (non-temporal) While the tomatoes were ripening, they were sprayed, (progressive)
Examples of the use of WHILE to refer to a non-conclusive durative situation (an activity) and state are, respectively: (10.38) While I stuffed into my many pockets a selection ofslim volumes of prestigious verse, Raymond was concealing on his person . . . (17.36) . . . I was pledged to look after my sister that eveningwhile my parents were at the Walthamstow dog track,. . .
In contrast to WHILE, WHEN is used for conclusive durative verbs and punctual verbs, e.g.: (91.22) . . . and when I could hear him no more I said.. . (conclusive durative) (28.08) But when I opened the door she was standing right in my way with a shoe in her
hand, (punctual) However, like WHILE, WHEN can also be used for states: (30.23) You didn't have nightmares whenyou were working. You didn't have nightmares while you were working.
Sometimes substituting WHILE for WHEN in clauses with stative verbs leads to a non-temporal interpretation, in particular with ages: (76.36) When I was seventeen my mother was just thirty-eight. ?While I was seventeen my mother wasjust thirty-eight.
The rule for choosing between WHEN or WHILE with statives is not clear.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
WHILE cannot be used, as reaching something is a punctual not a durative verb, i.e. WHILE cannot be given a temporal interpretation when it refers to a punctual situation. For conclusive durative verbs, it is difficult to give WHILE a temporal
14 Translating Temporal Prepositions between Dutch and English Our present hypothesis about the difference between WHEN and WHILE is
that: — WHEN is used to refer to a time point, which may be indicated by either a punctual or conclusive durative or a state; — WHILE is used to refer to a time period, which may be indicated by a nonconclusive durative or a state.
(b) Ambiguous AS
Table 3 Substituting WHEN and WHILE for AS, and vice versa
Original
Translation
Substitutable by: None
Total
WHEN
24
0
1
25
0
13
7
20
77
AS
TOEN
—
AS
TERWIJL
-
WHEN
TOEN
WHILE
TERWIJL
Total
WHILE
AS
18
3
—
12
77 33
18
27
13
97
153
0
—
The most striking fact about these substitution data is that whereas AS can always be replaced by WHEN or WHILE, WHEN and WHILE cannot always be
replaced by AS, contrary to expectations, e.g.: (27.34) Iwas sitting in the bathroom one evening writing . . . when suddenly she was outside, rapping on the door. . . *I was sitting in the bathroom one evening writing . . . as suddenly she was outside, rapping on the door. . . (68.03) I made her wait outside while I went in the shop and bought her what she wanted, *I made her wait outside as I went in the shop and bought her what she wanted,
One of the reasons is that AS cannot be used temporally to refer to a state. As with a state is given a causal interpretation, e.g. replacing WHILE by AS in 17.36:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We now introduce AS into this scheme. As is said to denote 'merely simultaneity' (Quirk et al. 1985: 1083), so we begin with the hypothesis that AS is the super-ordinate of both WHEN and WHILE. T O test this, we have taken all the AS, WHEN and WHILE sentences from First love, last rites and made some substitutions. It should be the case that AS should be substitutable by WHEN or WHILE and that WHEN and WHILE should be substitutable by AS. The results of these substitutions are given in Table 3.
Bree, Smit and van Werkhoven 15 ~ .. .1 was pledged to look after my sister that evening as my parents were at the Walthamstow dog track,. . .
More importantly, AS is used in those cases where the sub stevent is an event that logically follows, or is very close to, the previous Time of Discourse (TOD), e
-g-
(28.16) Maisie was in the bathroom about ten minutes and as she came out I caught her neatly and squarely on the top of her head.
By contrast WHEN is used to move the TOD to a new time point as in 27.34 above. Our hypothesis is:
There is no corresponding choice in Dutch. As is almost always translated by TOEN (when) or TERWIJL (while). Summary The rules for choosing between these temporal conjunctions can now be summarised. First for English: — if the TOD is not to be shifted to a new point, use AS; — to refer to a point of time, use WHEN, with a punctual, conclusive durative or stative verb; — to refer to a period of time, use WHILE, with a non-conclusive durative or stative verb. And for Dutch: — to refer to a point of time that: — is a single occurrence in the past, use TOEN, — is in the present or is repeated in the past, use ALS/WANNEER, in both cases with a punctual, conclusive durative or stative verb; — to refer to a period of time, use TERWIJL, with a non-conclusive durative or stative verb.
(c) Refinements and exceptions The translation data show that there are certain refinements and exceptions to these general rules. Here we will list these and briefly account for the more important ones. As -» ZODRA, always as a translation of AS SOON AS. —• ALS, for repetitive past events.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
— use AS if the TOD is only to be advanced to the next event rather than be moved.
16 Translating Temporal Prepositions between Dutch and English — OP HET OGENBLIK DAT, literally ON THE EYE-BLINK (MOMENT) THAT, for JUST AS.
-» Other, idiomatic: (15.31) . • .as soon as possible. -* zospoedig mogelijk. (as speedily as possible). (91.10) As the weeks passed. .. -* Naarmate (in the measure that). — Not translated, because the sub clause becomes either a matrix or a relative clause.
— Other, by DAT (that) in expressions as THE DAYS WHEN; by EN DAN (and then),
i.e. setting up a second matrix clause. —• not translated, because the temporal clause is transformed into a noun phrase, e.g. (79.13) There were days when ...-*•
Sommige dagen . . . (Some days).
While
-*
(as long as), for a conditional use of WHILE: (63.80) . . . and while he touched me he had the power. —• TIJDENS (during), when the sub-clause is replaced by a phrase so that a preposition is required rather than a conjunction: (34.07) It had stood in his study while he lived. (39.22) Het had tijdens zijn leven in zijn studeerkamergestaan. —• Not translated, by using the gerundive, WHILE I SUPPORTED -• SUPPORTING ; the whole sentence dropped. ZOLANG
Toen7
-* WHILE, in all cases with stative verbs in the sub-clause (cf. WHEN -» TERWIJL). — AT, again replacing the sub-clause with a phrase: (120.30) .. . toen de vacantie om was. (95.10) .. .at the end of the vacation. —• AFTER, again dropping the perfect aspect in the original (see WHEN — NADAT above) — Other, mainly with -ING verb form: (147.24) Of toen hij een keer terugkwam . . ., (or when he one time came back...) (117.22)
Or once coming
back...
—• Not translated, either by making a gerund, or making the clause into a noun phrase or a matrix clause.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
When — TERWIJL, both cases referring to a stative stance, e.g. SITTING, LOOKING, for which both TOEN and TERWIJL may be used -• NA(DAT), for a perfect aspect in the WHEN clause, and dropping this in the NADAT (after) clause. This is an option that occurs in both directions, but an explanation would take us beyond the scope of this paper.
Bree, Smit and van Werkhoven 17
Als — AS, this being a story, ALS was most frequently used temporally for repetitive events in the past rather than with the present or future tense, and AS is not normally appropriate in these circumstances. However, when it was used it was appropriate: (167.18) . . . ah ze's ochtends kwam dat ze haar op degang tegemoet stormden . .. (... when she in the morning came that they her in the passage up to stormed) (134.27) As she arrived in the morning they would storm up to her. . .
— AFTER, again with the dropping of the perfect aspect (see WHEN — NADAT above).
— not translated, instead using an -ING verb form or making the temporal clause into a matrix clause or a noun phrase. The refinements which we see are: AS SOON AS
*±
JUST AS
Ji
ALL THE WHILE J±
ZODRA OP HET OGENBLIK DAT INTUSSEN
Some general rules which emerge from these exceptions are: — dropping the perfect aspect from the sub-clause requires a different temporal preposition, e.g. WHEN + perfect -» AFTER; — a sub-clause may be replaced by a prepositional phrase, requiring that the subordinating conjunction be replaced by a preposition; — the preposition may be dropped altogether when the temporal clause is made into either: — an -ING construction; — a separate matrix clause; — a relative clause within a noun phrase in the matrix clause.
b. Simultaneous points in time We now turn to the prepositions that set the time of the matrix stevent. All these prepositions, with one exception—DURING—are also used as locatives. In order for them to be given a temporal meaning the phrase that follows them must be clearly temporal in itself. Note that this was not the case with temporal subordinating conjunctions, as we have just seen and as we will see again in the section on temporal order conjunctions.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Terwijl — AT A TIME WHEN,
18 Translating Temporal Prepositions between Dutch and English
First we look at prepositions that indicate the point in time at which the stevent occurs. The predictions that we had made here were that — AT is equivalent to OAT, — AROUND is equivalent to BOND; — BY is equivalent to TEGEN. Neither AROUND nor ROND occurred in the original texts. But our translation data play havoc with the other predictions. (a)
AT
?±
OM
Table 4 Number of occurrences of different translations of AT From: at with clock time at the time (. . .) at that time/moment at any time/moment at the same time at one time at the beginning/end of at the end a lfirst at last at high/low tide at night at meals
To: with preposition 3 3 3
om toen (. ..) op dat tijd/ogenblijk
no preposition
2 2
4 1
7 1
4
aan het begin/eind van ah het klaar was in het begin
6 1
3 4 1
29
toendertijd elke/ieder ogenblik tegelijk(ertijd)/meteen vroeger
eerst/aanvankelijk eindelijk
bij hoog/laag water 's nachts/'s avonds tijdens maaltijden 16
From this sample it is not possible to give more than an indication of a possible system. Clearly OM is only used with specific clock times in Dutch. If the word TIME is specifically mentioned in the AT phrase, then there are several options in Dutch: — if the rime has already been or is going to be specified, use TOEN (then); — if the rime is open, use OGENBLIK (moment) governed by either OP DAT (at that) or ELK/IEDER (any). OP (on) is also used for days of the week—see below. For border rimes, i.e. beginnings and ends, Dutch uses AAN, which is also used for locative edges, e.g. AAN ZEE (at the seaside). The use of AT in AT NIGHT is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
For the predicted equivalence between AT and OM we find that, while the only two occurrences of OM are translated by AT, only 3 of the 4 5 ATS are translated by OM. We will now examine the different translations of AT.
Bree, Smit and van Werkhoven 19
peculiar to English; the Dutch 's NACHTS is in line with the use of's with other periods of the day, e.g. 's AVONDS for IN THE EVENING. The '5 represents the word DES, genitive form of the article, a usage which is now archaic, but still found in modern German. A similar pattern emerges from the translations into AT of various Dutch prepositions, as can be seen in Table 2. (b) B Y *± TEGEN
Table 5 Number of occurrences of different translations of BY8 From
To: with preposition
by without time by + year by the end of. . .
5
tegen omstreeks (about) aan het eind van . . .
by the time. . . . . .,by which time. . . by this time by now
5
1 1
1
no preposition
toen. . . wanneer. . .
13
1
MM
1
al
2
We see again: — the use of AAN for beginnings/ends. — the incorporation of the word TIME, to give in Dutch either: — a subordinating conjunction: TOEN, WANNEER (when), or — an adverb: AL (already), NU (now).
(c) Conclusion The general conclusion that we can cautiously draw from this analysis of the translations of the prepositions indicating simultaneous points in time is: — for a clock time: at ^ om; — for beginnings/ends of an event: at/by the beginning/end of — before a time: by
n aan het begin/eind van I* tegen;
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
TEGEN did not occur in the Dutch original text, so no analysis is possible. Only 5 of the 16 occurrences of BY are translated by TEGEN. Most of the remainder had the word TIME in the phrase. See Table 5 for details.
20 Translating Temporal Prepositions between Dutch and English
— from English THE TIME clauses: at/by the time •* toen;9 — from English AT . . . TIME phrases, use the corresponding adjective + TIJD/ OGENBLIK (moment); e.g.: at any time -~ elk ogenblik at the same time -* tegelijk (ertijd) at that time — op dat ogenblik.
— the same as (AT) or — up to and including (BY) the point on the time axis indicated by sub-event. In Dutch the time line is two dimensional. Clock time is one dimension. But on this dimension are mounted events. These events can be referred to in the prepositional phrase in two ways, either by taking an event cross-section against time or by taking a two dimensional time-event chunk. So in Dutch we get the time of the matrix event being located either:10 — around a clock time point using OM, which means AROUND in its locative use, — on an event time slice using OP, which means ON in its locative use, — supported by an event-time box using AAN, which in its locative use means functionally next to, e.g. for ON in 'a painting hanging on the wall' one uses AAN in Dutch, cf. the functional interpretation of AT in English, or — up to and including an event-time slice using TEGEN, which means in its locative use oppositionally against, e.g. for ON in 'John is leaning on the wall' one uses TEGEN in Dutch. Following this schema, in Dutch: — clock times are zero-dimension points on the time axis; — moments, as in THAT TIME, are one dimensional event slices out of the eventtime space; — beginnings and ends of events are two dimensional event-time objects.
c. Simultaneous time selectedfrom a series In our original proposal (Figures i and 2), we postulated that a period of time could be a state, border or container. Here we examine only the time border.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
While the details still remain to be worked out, it is clear that the English and Dutch simultaneous point of time prepositions do not have a one-to-one correspondence. We suggest that English and Dutch have different ways of viewing the time line in this respect. In English the time line is a single axis. To locate an event at point on that axis there are only two possibilities: either the time the event occurs is:
Bree, Smit and van Werkhoven 21
The idea of a rime border is borrowed from a space border, e.g. customs posts are on borders, ships are on the surface of the sea. In English, days are the only rime borders, e.g. ON SUNDAY, ON MAY 5TH. In Dutch spatial borders and days both require OP. SO the simple prediction was made that ON and OP were equivalent. In both languages the preposition is optional, i.e. the ON/OP can be dropped. So the ON-OP equivalence should turn out to be only partially the case.
Otlfi- op
THIRTEENTH BIRTHDAY.
Only some of the translations of OP were to ON. However, for most cases of OP followed by a specific day ON was the chosen translation." This leads to the revised hypothesis that ON and OP are both used for particular days or parts of days from a series, e.g. days of the week/month, the Nth morning/afternoon/ evening/night. ON is not obligatory for days, as in: John's coming Sunday Thefactory will be ready 5 May. (NB not fifth May)
In these examples OP is not generally used. And neither ON nor OP are permitted when the (part of the) day is governed by a quantifier, as in: ~ We hoped to see Veronica on one/next/last Monday.
Indeed, for the translations of ON + a day of the week there was no preposition. O N -• something other than OP
The other cases in which ON with a day was not translated by OP all refer to repeated events, e.g.: (15.06) . .. rest up on S u n d a y . . .
-• . . . zondags uitrusten . . . (on Sunday resting out)
There were also two cases in which ON was used without a specific (part of a) day in a series. In the firsc (34 -15)
• on lonely n i g h t s . . .
— . . .gedurende (during)
eenzame nachten . . .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Most of the translations of ON were to OP. In some cases (5) the time referred to was a day of the month; in other cases (6) to a particular day or part of a day in a series, e.g. ON THE THIRD MORNING. The use of ON with a part of a day is regarded by Quirk et al. (1985:688) as exceptional. We would rather see it as generic. ON picks out one day or part of a day from a series, which is frequently the days of the week or the days of the month, but may also be birthdays, e.g. ON HIS
22 Translating Temporal Prepositions between Dutch and English
it is possible that in English a LONELY NIGHT is a night in a series of nights, whereas in Dutch OP requires an ordinal number of nights. In the other: (15-15)
• • on special occasions...
— . . . bij
bijzonderegelengheden
it is not clear whether this is a temporal use of ON. There is a whole category of such expressions in which ON is translated by BIJ, namely for types of weather condition, e.g.: On clear mornings At low water
->• Bij heldere ochtenden. -* Bij laag water.
OP -* something other than ON
— AT used for (153.09)
OP ..
. MOMENT where
MOMENT
Op hetzelfde moment...
is translated by TIME/MINUTE (2):
-