Tense, Mood and Aspect Theoretical and Descriptive Issues AHIERS CHRONOS 17
Collection dirigée par Carl Vetters (Université du Littoral – Côte d’Opale)
Directeur adjoint: Patrick Caudal (CNRS – Université Paris 7)
Comité de lecture: Anne-Marie Berthonneau (Université de Lille 3) Andrée Borillo (Université de Toulouse-Le Mirail) Anne Carlier (Université de Valenciennes) Renaat Declerck (KULAK-Courtrai) Walter De Mulder (Université d’Anvers) Patrick Dendale (Université d’Anvers) Ilse Depraetere (KUB - Bruxelles) Dulcie Engel (University of Swansea) Laurent Gosselin (Université de Rouen) Emmanuelle Labeau (Aston University) Véronique Lagae (Université de Valenciennes) Sylvie Mellet (CNRS - Université de Nice) Jacques Moeschler (Université de Genève) Arie Molendijk (Université de Groningue) Louis de Saussure (Université de Neuchâtel) Catherine Schnedecker (Université de Metz) Marleen Van Peteghem (Université de Lille 3) Genoveva Puskas (Université de Genève) Co Vet (Université de Groningue) Carl Vetters (Université du Littoral - Côte d’Opale) Svetlana Vogeleer (Institut Libre Marie Haps - Bruxelles) Marcel Vuillaume (Université de Nice)
Ce volume est une réalisation de l’équipe de recherche “HLLI” - EA 4030 de l’Université du Littoral - Côte d’Opale, de la Société Académique de l’Université de Genève (Fonds Charles Bally) et du Groupe de recherche en sémantique et pragmatique de l’Université de Neuchâtel. Tense, Mood and Aspect Theoretical and Descriptive Issues
Edited by
Louis de Saussure, Jacques Moeschler and Genoveva Puskas
Amsterdam - New York, NY 2007 Cover design: Pier Post
Le papier sur lequel le présent ouvrage est imprimé remplit les prescriptions de “ISO 9706:1994, Information et documentation - Papier pour documents - Prescriptions pour la permanence”.
The paper on which this book is printed meets the requirements of “ISO 9706:1994, Information and documentation - Paper for documents - Requirements for permanence”.
ISBN-13: 978-90-420-2208-9 ©Editions Rodopi B.V., Amsterdam - New York, NY 2007 Printed in The Netherlands
Acknowledgements
The Editors are grateful to all the colleagues who spent time helping them in the complicated process of the selection of papers. In particular, all of them who were involved in peer-reviewings deserve a special mention here. We warmly thank the institutions who supported the organization of the 6th Chronos colloquium in Geneva: the Swiss National Science Foundation, the Swiss Academy of Social Sciences, the Academic Society of the University of Geneva, the Faculty of Arts of the University of Geneva. We also thank the Chronos board for giving this opportunity to us, in particular Carl Vetters and Co Vet. This book would not have been published withough the support of the ‘Charles Bally’ fund of the University of Geneva (Academic Society). Last but not least, we are very thankful to Patrick Morency, PhD student at the University of Neuchâtel, for the time and energy he spent with and without Louis de Saussure in proof-readings, manuscript layout checking and editing, bad and missing references tracking, etc.
This page intentionally left blank
Table of contents Louis de Saussure Introduction 1-5 Co Vet The descriptive inadequacy of 7-26 Reichenbach’s tense system: A new proposal Hans Smessaert The evaluation of aspectual distance, 27-45 speed and progress Steve Nicolle The grammaticalization of tense 47-65 markers: A pragmatic reanalysis Maria Asnes Aspectual interactions between 67-80 predicates and their external arguments in French Anne Le Draoulec Alors as a possible temporal connective 81-94 Myriam Bras in discourse Tijana Asic The power of prepositions: Is he 95-110 sleeping now or usually? Hortènsia Curell On the dual nature of the Catalan 111-127 Mercè Coll present perfect Andrea Rocci Epistemic modality and questions in 129-153 dialogue. The case of Italian interrogative constructions in the subjunctive mood André Meinunger In the mood of desire and hope: remarks 155-176 on the German subjunctive, the verb second phenomenon, the nature of volitional predicates and speculations on illocution Linde Roels Dutch equivalents of the German past 177-196 Tanja Mortelmans conjunctive: zou +infinitive and the Johan van der Auwera modal preterit Boban Arsenijević Slavic verb prefixes are resultative 197-213 Sophia Delidaki The acquisition of aspect in child 215-227 Greek: A production experiment Jiranthara Srioutai The Thai cla: a marker of tense or 229-239 modality?
. This page intentionally left blank
Introduction
Louis de SAUSSURE University of Neuchâtel
This volume is a selection of papers presented at the 6th Chronos colloquium dealing with both theortical issues in the study of tense, mood and aspect, and specific semantic and syntactic analysis of linguistic expressions dedicated to these domains across a variety of languages. The contributions presented here address specific linguistic features in order to describe the expression of this threefold domain and provide explanations for them across languages and with a significant theoretical import. Through these papers, strong variations are explored, but also crosslinguistic convergences are investigated. Nume- rous phenomenas so far often left aside in linguistics are described and enlightened by different scientific standpoints, which they serve to illustrate. The languages investigated in this volume include Germanic languages (Dutch, English, German), Romance (French, Catalan, Italian), Slavic (Serbo-Croatian, Czech, Russian), Greek, and non-indoeuropean languages such as Thai, Digo and Kikuyu. Related topics such as grammaticalization, presuppositions, questions in dialogue, illocutionary acts and acquisition are incidentally called upon in order to shed light from the outside onto tense, mood (and modality) and aspect. The volume is opened by a theoretical concern expressed about the classical symbolic formalism proposed by Reichenbach. In his paper, Co Vet takes issue with this formalism; through a number of criticisms, he proposes a new way to organise temporal representations, still with coordinates, but without the defects of the classical system. Although there is an abundant literature trying to solve the poverty, or, as Vet puts it, the “descriptive inadequacy” of Reichenbach’s formalism, Vet’s paper is a really new proposal. Focusing on French tenses, he convincingly argues that the reference point R cannot be placed in the future in the same way as it can be placed at the present time or at a past time. Vet therefore ends up with a twofold system of tenses for French, that he calls a neo-reichenbachian system, and where two possible positions are available for saturation by a perspective point. With regard to an aspectual notion (phasal aspect), he explains in much details how to solve in such a system classical problems such as the polysemy of the Composed Past and of the Periphrastic Future in French, before addressing overcomposed forms. Hans Smessaert’s paper addresses the Dutch paradigm of adverbial expressions that are about the internal structure of events, which thus have to
© Cahiers Chronos 17 (2007) : 1-5. 2 Louis de Saussure do with imperfectivity. His investigation aims at charting “the various constraints on the combination of objective aspectual information and subjective evaluative information” through i) specific aspectual features beared by these adverbs, ii) an evaluative parameter of ‘distance’ (with the boundaries of the considered eventuality), and iii) an evaluative parameter of ‘speed’ and ‘progress’ of “the actual course of events”. He shows how these properties interact, moreover, with basic aspectual information. Steve Nicolle presents a tight analysis of the structural and semantic changes that occur when a movement verb becomes a tense marker through grammaticalization. He takes notably the example of go in cases like go and V. On the syntactic side, he argues that the movement verb can become more closely linked to the main verb that it modifies, loosing its proper inflexional properties and becoming a verbal affix. At the semantic level, he takes into account that the meaning of the considered verb, which relates to the physical movement of an entity relatively to the deictic centre, through this kind of grammaticalization, takes a meaning that is about events and reference time. Exploring such constructions in Digo and English, he argues that it is rather subjectivisation than semantic change that occurs in these cases. Maria Asnes’ paper a She proposes to extend Krifka’s definition of homomorphism to the domain of external arguments, as a result from the observations from Dowty and Jackendoff that not only internal arguments can function as incremental themes. She then formulates her extended version of homomorphism. She takes a systematic look at various situations where external arguments play a role in aspectual determination. With interesting data, she claims for instance that iterativity creates a sort of homomorphism between the denotations of the external arguments and of the verbal predicate, a type of homomorphism which is different, she says, from the unique-event type triggered by the incremental theme relation. In the general idea that homogeneous / heterogeneous reference applies both to nominal and verbal constituents and thus, provides a common ground for the treatment of aspectual interactions, she identifies eight combinations of features in ditransitive constructions, which she discusses with regard to her corpus. Anne Le Draoulec and Myriam Bras present their investigation of French alors in the framework of DRT (Discourse Representation Theory), an approach they already used in their former works on a variety of connectives in French. The main question they address is the temporal property of the adverbial alors: they attempt at providing evidence for – or against – the fact that alors can be classified as a temporal connective. They discuss the hypothesis that the syntactic position of alors is a crucial parameter for this item to acquire the function of connecting temporally the utterance it belongs to with a former element in the considered discourse. Introduction 3
They show that when occupying an initial position (or at least preverbal), alors always implies a dependency relation thus has the connective function, while when in internal and final positions, alors has a softer connective constraint (although the connective value is not blocked), because it depends on the semantic-pragmatic context. However they assume that when an expression depends on context to trigger a discursive connection, the expression can’t be considered a connective by itself. Their paper gives further wrinkles about possibilities to develop more fine-grained analyses of alors when internal or final. Tijana Asic’s paper addresses an interesting issue about the temporal interpretation of spatial prepositions in Serbian and Kikuyu. The semantic distribution of the spatial prepositions po/na (on/over) in Serbian is the following: po is used with continuous substances, either static (there is water everywhere ‘po’ the floor) or in motion (Koljia is walking ‘po’ the tightrope), wheras na is limited to static discrete objects (Dusan is sitting ‘na’ a chair). Interestingly, the distribution of these prepositions is not restricted to spatial relations: they occur also to implicate temporal relations, that is, a temporal interval interpretation (with state or activity nouns like darkness and rain). In these cases, the interpretation of the temporal relation is the external contact one, with no physical contact, whereas the na NP construction implicates a physical contact (‘no’ rain means ‘during the rain’, ‘na’ rain ‘under the rain’). This extension to the temporal domain is parallel in Kikuyu, a Bantu language, which reveals a very similar pragmatic behaviour. The final description of these prepositions uses a minimal semantic analysis of po/na based on Casati & Varzi’s topological notion of ‘external connection’, a pragmatic device of triggering implicatures and the dot objects property of nouns like rain and darkness (physical substance x event). Hortènsia Curell and Mercè Coll present an empirical study aiming at better understanding a classical problem in most Romance languages, that is, the double value of the present perfect (temporal and aspectual). Namely, the present perfect communicates a resulting state, which Curell and Coll call ‘present relevance’, with focus either on past remoteness (existential use) or on resulting state itself (resultative use). The intriguing thing with the present perfect in Catalan is that it still has an hodiernal temporal value (i.e. it marks reference to an eventuality obtained within the day), a value lost long ago in most of other Romance languages. Curell and Coll made a survey of a corpus of 9 theatre plays and present the distribution of the different semantic values they obtain by crossing the categories they have. Their study shows a strong correlation between predicate type (telic or atelic) and the type of reading obtained (resultative or existential). They show that telic predicates trigger a resultative reading (except in some particular cases) while atelic predicates trigger an existential reading, except when a strong cause-effect contextual constraint applies. Their next step is to bridge the gap of these different 4 Louis de Saussure readings, in order to avoid an ambiguous description of the tense; they anchor their account on a series of parameters, notably contextual, but also semantic, that apply constraints on the interpretation of the tense. Andrea Rocci’s paper focus on interrogative constructions in the subjunctive mood in Italian – with some cross-linguistic investigation –, in particular on the constraints imposed on interpretation by the complementizer che in such constructions. One of his first strong points is that che needs to be analyzed along the lines of Fillmore and Kay’s assumptions on the grammar- pragmatics interaction. First, che+subjunctive is associated with a meaning that is not derivable compositionally; second, it inherits properties of generic constructions such as yes/no question and subjunctive construction; Third the meaning of these constructions is associated with both specific contextual restrictions, discursive relations and constraints on the illocutionary force. His account is elaborated in the framework of Congruity theory, in particular with recourse to the notion of pragmatic predicates developed within this approach. Andre Meinunger tackles the German subjunctive through a number of problems raised by the analysis of this form. In relation with the ongoing debate on the ‘Verb second’ problem (i.e. constructions where embedded CPs exhibit phenomena which are expected to occur in main clauses only) in Germanic, Meinunger focuses on verbs of propositional attitude and volitional predicates in order to discuss the relation between propositional attitudes and ‘Verb second’ patterns. After discussing some features affecting the licensing of these patterns, he suggests that Verb second blockers form a homogeneous class of verbs making reference to factivity, and gives a detailed discussion of the relation between volitional (in particular counter- factive volitional) predicates and the ‘Verb second’ phenomenon in German. In their article, Linde Roels, Tanja Mortelmans and Johan van der Auwera compare the German and Dutch ways of expressing conjunctive; these languages differ insofar as Dutch, contrarily to German, doesn’t have a morphologically realized conjunctive mood (although Middle Dutch had one, still recognizable today in some formulaic remnants). In particular, the authors go through a detailed semantic analysis of the Dutch zou, which marks counterfactuality and hypotheticality; they confirm their assumptions through a survey of zou in the Dutch Spoken Corpus. Boban Arsenjievic addresse Slavic verb prefixes. In Slavic languages (here Serbo-Croatian and Czech are discussed), verb prefixes have a complex twofold function that combine with, or result from, their status of semantic predicate modifiers. These prefixes narrow or even switch the meaning of the imperfective verb while forcing perfective reading. Arsenjevic suggests that in fact these event-modifying prefixes in Slavic languages are all resultative. On the syntactic level, quite logically, Arsenjevic assumes that these prefixes are to be analyzed as agreement markers. More precisely, he shows that they Introduction 5 can be considered as concerning the agreement between the values of a specific property of both the initiating and the resulting sub-events of the VP. The article by Sophia Delidaki presents an experimental study about the acquisition of aspect in child Greek. She recalls the main options taken by the literature on aspect acquisition, notably that a certain type of morphology (tense or grammatical aspect) is used mistakenly by children in order to mark lexical aspect. Delidaki’s study applies to Greek the methodology Bronckart and Sinclair’s used for French. She enlights the contrast between children and adult choice in performance for tense and aspect through six events that the subjects of the experiment have to report; the events presented vary as far as the aspectual situation is concerned so that the results imply conclusions on the relationship between lexical and grammatical aspect features. More precisely, her paper investigates the role of four semantic features (telicity, duration, frequency and completion), and shows that lexical aspect influences the choice of tense and grammatical aspect, that duration influences choices mainly when an atelic verb is encountered, with the nuance that adults show however some influence of duration with telics. Interestingly, she shows that there is a duration limit of ‘3 seconds’ attributed to events below which atelic verbs will be used by children with a past perfective and above which present is selected. Other relevant conclusions regarding frequency and completion are also presented. Jiranthara Sriouai analyzes a Thai marker, cla, which has a similar kind of underspecification as the English will – to which it is generally compared. This marker behaves in effect like the English will regarding its epistemic and temporal values. Through a detailed analysis of cla, she concludes however that the core semantics of this item is one of a modal marker rather than a temporal one. In a shorter section at the end of her paper, she shows how a formal model of discourse can show useful for the description of the semantics of such an item; the framework she uses is DRT, however she shows that cla is better accounted for when following Jaszczolt’s claim that Grice’s acceptability concept – which covers epistemic modality – should be introduced as a scalar modal operator, so to speak. With such a modal operator, a full account for the various types of uses cla can have is proposed as a promising hypothesis that must be further explored and validated.
This page intentionally left blank
The descriptive inadequacy of Reichenbach’s tense system: A new proposal
Co VET Groningen University
0. Introduction
The aim of this paper is to show that Reichenbach’s (1966 : 287-298) tense system is descriptively inadequate. The problems do not appear if the description is limited to one or two tenses. However, if one tries to assign Reichenbachian configurations to all the tenses of a language the shortcomings of the system become apparent. In section 1, I will examine the main principles underlying Reichenbach’s tense system and discuss the main problems of his proposal. One of these problems concerns the division of the time axis into past, present and future. It will be shown that a first division into past and present gives a descriptively more satisfactory result. It will be no longer necessary to assign more than one configuration to some tense forms (to the Future Perfect, for example). A more general problem in Reichenbach’s system is that meanings (configurations of speech point S, reference point R and eventuality E) are directly linked to specific tense forms. In our approach, a specific tense function can be assumed by more than one tense form. In section 2, a tense system will be proposed which possesses only two perspective points, one in the past the other in the present. It will be shown that in this system some of Reichenbach’s problems can easily be solved. A problem of a different nature is that Reichenbach characterizes the meaning of all the tense forms in terms of temporal relations. However, some of the French tenses have to be interpreted in some contexts as providing aspectual information (resultative or prospective), whereas in other contexts they behave as tenses (expressing anteriority or posteriority). In section 3 I will introduce the notion of ‘Phasal Aspect’ and in section 4 this notion will be used to describe the meanings of the Composed Past, the Pluperfect and of the Periphrastic Future (aller ‘go’ + infinitive) of French. Finally, I will describe the so-called Overcomposed forms of French (section 5), in which the aspectual and temporal meanings of the Composed Past and the Pluperfect are combined. The mere existence of these forms in French proves that the sequence avoir ‘have’ + Past Participle has to be given two different interpretations. In the conclusion I will enumerate all the tense forms of French together with the tense functions they can fulfil.
© Cahiers Chronos 17 (2007) : 7-26. 8 Co Vet
1. Reichenbach’s system
In Reichenbach’s system, R can have three positions relatively to S:
S R R R
S: speech point, R: reference point
Figure 1
Eventualities, in turn, can be placed in the same way with respect to R: E can be anterior, simultaneous or posterior to R. This repeated tripartition yields a tense system with nine tense positions as is shown in Figure 2. In this figure I have tried to indicate which French tenses might be used to express the different configurations (see table 1 below for symbols).
S R R R ------o------o------o------E E E E E E E E E PP IP SFP CP PR PF FP SF Ø AP SP PFP PF OP OPP
Figure 2
In Table 1 below I represent the complete list of French tenses, with an example and the Reichenbachian formulae that could express them (note that in some cases we have to improvise, which is indicated by a question mark).
The descriptive inadequacy of Reichenbach’s tense system 9
Reichenbach’s tense system applied to French :
Tense form Example Reichenbachian Formula Present (PR) Mange (‘is eating’) S,R,E Imperfective Past (IP) mangeait (‘was eating’) R,E – S Composed Past (CP) a mangé (‘has eaten / ate’) E – S,R Pluperfect (PP) avait mangé (‘had eaten’) E – R – S Overcomposed Past (OP) a eu mangé (‘has had ?E – R – S eaten’) Overcomposed Pluperfect avait eu mangé (‘had had ?E – R – S (OPP) eaten’) Periphrastic Future (PF) 1 va manger (‘is going to S,R – E or S – R, E eat’) Periphrastic Future of the allait manger (‘was going ?R – E – S Past (PFP) to eat’) ?R – S,E ?R – S – E Simple Future (SF) mangera (‘will eat’) S – R,E Simple Future of the mangerait (‘would eat’) 2 R – E – S Past (SFP) R – S,E R – S – E Future Perfect (FP) aura mangé (‘will have S – E – R eaten’) S,E – R E – S – R Future Perfect of the aurait mangé (‘would impossible without a Past (FPP) have eaten’) second R (R’) : (impossible in ?R – E – R’ – S Reichenbach’s system) ?R,E – R’ – S ?E – R – R’ - S Future of the Future (does ---- S – R – E not exist) Simple Past (SP) mangea (‘ate’) E,R - S Anterior Past (AP) eut mangé (‘had eaten’) E – R – S
Table 1 : The tense forms of French
1 Reichenbach (1966: 296, note 1) suggests that the Periphrastic Future expresses the order S,R – E and the Simple Future the order S – R, E. It will be shown in section 4 that the Periphrastic Future can have both meanings. 2 The Simple Future of the Past and the Future Perfect of the Past are often regarded by French grammars as forms pertaining to the category of mood (cf. Grevisse 1970: 565, 680) and called ‘Conditionnel Présent’ et ‘Conditionnel Passé’ respectively. In this paper I only consider the temporal meanings of these forms. See Vet and Kampers-Manhe 2001 for a critical view on the question.
10 Co Vet
Note that the difference between the Pluperfect and the Anterior Past consists in the fact that the auxiliary avoir ‘have’ of the Pluperfect is in the Imperfective Past (avait) while that of the Anterior Past is in the Simple Past (eut). Figure 2 and Table 1 already show the weak points of Reichenbach’s system. One of them is the lack of balance between the past and the future. The other inadequacies are listed below (See also Verkuyl, Vet, et al. 2004 : 249-252). 1. The Future of the Future is not attested in natural languages. Reichenbach (1966 : 207) mentions the form I shall be going to see him, but remarks that it does not refer directly to the eventuality itself, but to the preparatory state leading to it. The Latin form abiturus ero, translated by Reichenbach as ‘I shall be one of those who will leave’, does not refer directly to the eventuality either. 2. Reichenbach (1966 : 297) represents the Future Perfect (aura mangé ‘will have eaten’) by three formulas : S – E – R ; S, E – R or E – S – R. This would imply that this tense form is threefold ambiguous, which is not the case. Two of Reichenbach’s meanings do not occur in natural language. The sentence Il aura mangé maintenant (S, E – R) ‘He will have eaten now’ only has a modal meaning (‘He has eaten now, I think’) and the only (rather forced) interpretation of Il aura mangé hier ‘He will have eaten yesterday’ can also only be interpreted as a modalised Present Perfect. We will see that the configuration S – E – R does not correctly express the meaning of aura mangé ‘will have eaten’ either, because it has an aspectual meaning (resulting state) and not a temporal one as in Reichenbach’s proposal. 3. The past counterpart of the Future Perfect (aurait mangé ‘would have eaten’) cannot be represented in Reichenbach’s system. Compare the following examples :
(1) a. Jean est convaincu que Pierre réparera la voiture dans deux heures. ‘ Jean is convinced that Pierre will repair the car in two hours.’ b. Jean est convaincu que Pierre aura réparé la voiture dans deux heures. ‘Jean is convinced that Pierre will have repaired the car in two hours.’ (2) a. Jean était convaincu que Pierre réparerait la voiture dans deux heures. ‘Jean was convinced that Pierre would repair the car in two hours.’ b. Jean était convaincu que Pierre aurait réparé la voiture dans deux heures. ‘Jean was convinced that Pierre would have repaired the car in two hours.’ There is a complete parallelism between (1a, 1b) and (2a, 2b). However Reichenbach cannot represent the meaning of the Future Perfect of the Past of (2b) without introducing an additional reference point as in (3) :
The descriptive inadequacy of Reichenbach’s tense system 11
(3) Representation of aurait mangé ‘would have eaten’ in a Reichenbachian notation enriched with an auxiliary reference point (R’) : R – E – R’ – S or R, E – R’ – S or E – R – R’ – S.
4. Reichenbach has to represent the temporal part of the meanings of the Imperfective Past and the Simple Past in the same way, that is, R,E – S . He notes however (1966:291) that there is an aspectual difference between the two tenses: in his view the Imperfective Past expresses duration whereas the Simple Past is said to refer to a punctual eventuality. Reichenbach gives the following representation of this difference:
Simple Past: -----|------|---- Imperfective Past: ----|======|------|--- R,E S R,E S
Figure 3: The difference between the IP and the SP (Reichenbach 1966:291).
This analysis is erroneous because it does not explain why it is precisely the Imperfective Past that is incompatible with a duration adverbial while the Simple Past combines without any problem with this kind of adverbial. Compare (4) and (5).
(4) Pierre parla pendant trente minutes. ‘Pierre spoke (Simple Past) for 30 minutes.’ (5) *Pierre parlait pendant trente minutes. (*except in a habitual reading). ‘Pierre spoke / was speaking (Imperfective Past) for 30 minutes.’ In Reichenbach’s system no explanation can be given for the fact that the Imperfective Past has temporal and modal meanings and the Simple Past only a temporal one.
(6) S’il était riche, il partirait aux États-Unis. ‘If he were (Imperfective Past) rich, he would leave for the USA.’ (7) S’il fut riche, il … (only a factive reading: ‘Although he was (Simple Past) rich, he…’)
I will not discuss here the difference between Imperfective Past and Simple Past, and refer to Molendijk (1990), Vet (1994), Molendijk, de Swart et al. (2004) for a complete description. 5. Reichenbach’s system cannot provide an explanation for the fact that the Periphrastic Future does not exist in the future:
(8) Jean va partir. (R, S – E) ‘Jean is going to leave.’
12 Co Vet
(9) Jean allait partir. (R – E – S; R – S, E; R – S – E) ‘Jean was going to leave.’ (10) *Jean ira partir. (*S – R – E) ‘Jean will be going to leave.’
(10) should be possible in Reichenbach’s system because his reference point can be in the past, the present and the future. The examples (8) – (10) indicate that the existence of a future reference point is problematic in a tense system of French (and possibly for any natural language, see also point 1 above: in fact (10) is a Future of the Future). 6. In Reichenbach’s system there is no possibility to differentiate the formulas for the Pluperfect, the Anterior Past, the Overcomposed Past and the Overcomposed Pluperfect. All these tenses have to be represented by the configuration E – R – S. This is rather counterintuitive. Compare the following examples:
(11) Dès que Chantal avait eu terminé sa thèse, elle était partie pour les États- Unis. (E – R – S) ‘As soon as Chantal had had finished her thesis, she had left for the US.’ (12) Dès que Chantal a eu terminé sa thèse, elle est partie pour les États-Unis. (E – R – S) ‘As soon as Chantal has had finished her thesis, she has left (i.e. she left) for the US.’ (13) Dès que Chantal eut terminé sa thèse elle partit pour les États-Unis. (E – R – S) ‘As soon as Chantal had finished (Anterior Past) her thesis, she left (Simple Past) for the US.’ (14) *Dès que Chantal avait terminé (Pluperfect) sa thèse, elle partait (Imperfective Past) pour les États-unis (* in the non habitual reading) (E – R – S ) ‘As soon as Chantal had finished (Pluperfect) her thesis, she left / was leaving (= ‘used to leave’) (Imperfective Past) for the US.’
It is obvious that (11) does not have the same meaning as (12): the eventuality is viewed from a different perspective point. The same is true for (11) and (13). (13) and (14) may be regarded as stylistic variants. (14) cannot be interpreted in a non habitual way. Reichenbach’s notation obscures here important semantic distinctions. I will show in section 5 below how these differences can be accounted for. But in order to do so some fundamental principles of Reichenbach’s approach will have to be abandoned as we will see in the next section.
The descriptive inadequacy of Reichenbach’s tense system 13
2. Towards a binary tense system
If Reichenbach had examined more closely the morphological regularities of the verbal systems of the Germanic and Romance languages, he would have noticed that there are in fact two series of tenses : one based on the past tense, the other on the present tense. In French there is a series of tenses that have the verb ending –ait and another one that does not. The absence or presence of this morpheme yields the following pairs of tenses :
Series I Series II Present mange ‘is eating’ Imperfective Past mangeait ‘was eating’ Composed Past a mangé ‘has eaten’ Pluperfect avait mangé ‘had eaten’ Overcomposed Past a eu mangé ‘has Overcomposed Pluperfect avait eu had eaten’ mange ‘had had eaten’ Simple Future mangera ‘will eat’ Simple Future of the Past mangerait ‘would eat’ Future Perfect aura mangé ‘will have Future Perfect of the Past aurait mangé eaten’ ‘would have eaten’ Periphrastic Future va manger ‘is going Periphrastic Future of the Past allait to eat’ manger ‘was going to eat’ Simple Past mangea ‘ate’ ---- Anterior Past eut mangé ‘had-SP eaten’ ----
Table 2 : The two series of French tense forms
The tenses of the two series have exactly the same meaning. The Imperfective Past may be regarded as a Present in the past, the Pluperfect as a Composed Past in the past and so on. The only difference is that in Series I the eventuality is viewed from a perspective point in the present and in Series II from some point in the past. This suggests that the tense system of French has only two main perspective points and not three as in Reichenbach’s proposal 3. From both perspective points there is a tripartition into past, present and future as in Figure 4 and in Table 3:
3 This is not the only argument in favor of the bipartition of the tense system. Modality can only be used in the past and the present. This is true for French and is, from a morphological point of view, even more evident in English, where the modal auxiliaries (can/could, shall/should, etc.) have only a present and a past form.
14 Co Vet
Eo R P R R P R ------ANT PAST POST ANT PRES POST E E E E E E
Eo : Speech Act P : Perspective Point ANT : Anterior POST : Posterior PRES : Present
Figure 4 : A tense system with two perspective points.
Subsystem P, Eo Subsystem P – Eo Anterior Present (R, E – P) Anterior Past (R, E – P) Present (P, E) Past (P, E) Posterior Present (P – R, E) Posterior Past (P – R, E)
(,: simultaneous to; –: anterior / posterior to, Eo: Speech Act; P: Perspective Point; R: Reference Point)
Table 3 : Neo-Reichenbachian binary tense system
The system possesses two perspective points (P), one simultaneous to the speech act (P, Eo) the other anterior to it (P – Eo). The reference points (R) can be anterior or posterior to P. The tense system represented in Figure 4 contains only six tense functions or meanings, which are defined as in Table 3. Some of the problems we mentioned with respect to Reichenbach’s proposal can be solved now. For example, in our system there is no room for a Future of the Future. The Imperfective Past and the Simple Past no longer express the same functions: the Simple Past expresses an Anterior Present and the Imperfective Past a Past (that is a Present in the past). Finally the absence of an example like (10) (*Jean ira partir ‘Jean will be going to leave’) can also be explained by the fact that with this form it is possible to adopt only two perspective points, one in the present the other in the past, as our system predicts. Example (10) presents the eventuality of the sentence from a future perspective point, which is excluded in our system. As we will see, the main problem in the description of the French tenses is that a given tense function can be expressed by more than one tense form and that some tense forms do not express a tense function but an aspectual one. This is the case of the Composed Past (and the Pluperfect), the
The descriptive inadequacy of Reichenbach’s tense system 15
Periphrastic Futures and the Overcomposed forms. We will examine these tense forms and their meanings in the sections 4 and 5 respectively. In order to describe the meaning of some of the tenses we need the notion of ‘Phasal Aspect’, which we will define in the following section.
3. Phasal Aspect
Phasal Aspect is a grammatical device whose main function is to indicate that the sentence does not refer to the eventuality itself, but to one of its phases. In Figure 5 we give the representation of an eventuality E and the two phases that are of importance for us here 4: the preparatory phase (pre-state) and the resultative phase (post-state).
E’ E E” ______|||||||||||||||||||||||||||______preparatory phase eventuality resultative phase (pre-state) (post-state)
Figure 5 : The phasal structure of an eventuality.
In (15) below, the sentence refers to the eventuality itself. In example (16) the Periphrastic Future refers to the preparatory state of the eventuality type described in the Predication of the sentence. The Composed Past of (17) indicates that the sentence refers to the resulting state of eventuality E:
(15) Jean abat le vieux chêne. ‘Jean is cutting down the old oak.’ (P, Eo, E) (16) Jean va abattre le vieux chêne. ‘Jean is going to cut down the old oak.’ (P, Eo, E’, where E’ stands for the preparatory phase of eventuality E) (17) Jean a abattu le vieux chêne. ‘Jean has cut down the old oak.’ (P, Eo, E”, where E” stands for the resultative phase of eventuality E) In (16) the form aller + infinitive (go + infinitive) expresses ‘Prospective Aspect’. A sentence containing Prospective Aspect refers to the preparatory state of an eventuality. In (17) the auxiliary avoir ‘have’ + Past Participle expresses Resultative Aspect. If a sentence has Resultative Aspect it refers to
4 Dik (1997: 236 - 243) also regards inchoative and egressive aspect as belonging to the category of phasal aspect. Since these aspects do not play a role in the following discussion I do not include them here.
16 Co Vet the Resulting Phase of an eventuality, not to the eventuality itself. Note that the examples (15) – (17) are all in the Present tense. The eventuality, the preparatory state and the resulting state coincide with the speech act Eo and with P. This analysis is confirmed by the fact that the sentences of (15) – (17) can be embedded under a perception verb.
(18) Je vois que Jean abat le vieux chêne. (Present) ‘I see that Jean is cutting down the old oak.’ (19) Je vois que Jean va abattre le vieux chêne. (Prospective Present) ‘I see that Jean is going to cut down the old oak.’ (20) Je vois que Jean a abattu le vieux chêne. (Retrospective Present) ‘I see that Jean has cut down the old oak’.
In these examples, the eventuality, the preparatory state and the resulting state are simultaneous to the speaker’s perception. Sentences (21) and (22) show that one cannot perceive a future or a past eventuality.
(21) *Je vois que Jean abattra le vieux chêne. (*with voir ‘see’ with the meaning of ‘visual perception’) ‘I see that Jean will cut (Simple Future) down the old oak.’ (22) *Je vois que Jean abattit le vieux chêne. ‘I see that Jean cut (Simple Past) down the old oak.’
The states of affairs of (18) – (20) can also be presented from a past perspective point:
(23) Je vis que Jean abattait le vieux chêne. (Past) (P, E – Eo) ‘I saw that Jean was cutting down the old oak.’
(24) Je vis que Jean allait abattre le vieux chêne. (Prospective Past) (P, E’ – Eo) ‘I saw that Jean was going to cut down the old oak.’
(25) Je vis que Jean avait abattu le vieux chêne. (Resultative Past) (P, E” – Eo) ‘I saw that Jean had cut down the old oak.’
4. The polysemy of the Composed Past and the Periphrastic Future
We have seen in section 3 that the term ‘Composed Past’ is not very adequate because in the examples we have examined so far this tense form had the meaning of a (Resultative) Present rather than that of a Past. However this tense form can also function as an Anterior Present. Compare the examples (26) and (27).
(26) Ce jour-là, Paul a pris la hache et a abattu le vieux chêne. ‘That day Paul took (literally: has taken) the axe and cut (literally: has cut) down the old oak.’
The descriptive inadequacy of Reichenbach’s tense system 17
In (26) the Composed Past a pris ‘took’ (literally: has taken) and a abattu ‘cut’ (has cut) can be replaced by the Simple Past without changing the meaning of this fragment5.
(27) Ce jour-là, Paul prit la hache et abattit le vieux chêne. ‘That day Paul took (Simple Past) the axe and cut down (Simple Past) the old oak.’
(26) and (27) are stylistic variants. (26) can no longer be embedded under a verb of perception:
(28) *Je vois que ce jour-là Paul a pris une hache et a abattu le vieux chêne. ‘I see that that day Paul took (has taken) the axe and cut (has cut) down the old oak.’
I propose for the Composed Past the following interpretation rules, in which its context sensitivity is taken into account.
(29) Interpretation rules for the Composed Past
(i) Introduce P and Eo and the relation P, Eo. A. Default reading (Resultative Present): Introduce E” (the resulting state of eventuality E) and the relation E”, P, Eo. B. Anterior Present reading : If the context contains a time adverbial such as ce jour-là ‘that day’ referring to an interval or point t anterior to Eo, introduce the eventuality E 6 and the relations ce jour-là (t) and R, E, t – P, Eo . The Present Perfect of English cannot have the B reading (*John has arrived yesterday). As for the Periphrastic Future, the problems are analogous to that of the Composed Past. In a present context (that is when P coincides with Eo) this tense form has to be interpreted as a Prospective Present.
5 There are differences between the Simple Past and the Composed Past in their function of Anterior Present. The Simple Past does not easily combine with deictic adverbials : ??Jean arriva hier ‘Jean arrived (SP) yesterday. vs Jean est arrivé hier ‘Jan has arrived yesterday’. The Simple Past may be regarded as a remote past tense in the sense that it places the eventuality in a stretch of time that does not contain the speech moment. The Composed Past is neutral in this respect. In their grammar Riegel et al. (1994 : 301 – 303) also distinguish for the Composed Past the same meanings as I do. Cf. also de Saussure (2003 : 232- 237). 6 It is evident that this rule should be completed especially as far as reading B is concerned. The Anterior Present reading can also be triggered by a temporal clause or by a preceding sentence in the Imperfective Past.
18 Co Vet
(30) (Je vois que) Paul va abattre le vieux chêne. ‘(I see that) Paul is going to cut down the old oak.’
In (30) there is no indication that a reference point R posterior to Eo should be introduced. In that case the Periphrastic Future gives the instruction to introduce into the discourse representation E’, the pre-state preceding an eventuality ‘cut down the old oak’ and the relation E’, P. Eo (the pre-state E’ coincides with P and Eo), in which case the sentence can be embedded under a perception verb. If the sentence contains a time adverbial as in (31) :
(31) Demain Paul va abattre le vieux chêne. ‘Tomorrow Paul will cut (literally: is going to cut) down the old oak.’ (32) ??Je vois que Paul va abattre le vieux chêne demain. ‘I see that Paul will (is going to) cut down the old oak tomorrow.’
The adverbial demain ‘tomorrow’ creates a future context (P, Eo – R, R is posterior to P and Eo). In that case the eventuality overlaps with R (R, E) and this blocks the Prospective Present reading. In this context the Periphrastic Future has to be interpreted in the same way as the Simple Future. Both the Composed Past and the Periphrastic Future are tenses whose interpretation is context sensitive, a fact that is not recognized in most grammars of French. The Grammaire méthodique by Riegel et al. (1994 : 253 and 312-316) gives a rather inconsistent description of the facts. It states that the Periphrastic Future has only the meaning of a Prospective Present (and as such refers to the phase preceding the eventuality) and at the same time the grammar repeats that it can replace the Simple Future, which has a different meaning. This replacement should not be possible if these tense forms have different meanings. My analysis is more precise: the Periphrastic Future and the Simple Future only compete in future contexts. In present contexts the Periphrastic Future has to be interpreted as a Prospective Present and cannot be replaced by a Simple Future and vice versa. We propose the following interpretation rules for the Periphrastic Future :
(33) Interpretation rules for the Periphrastic Future
(i) Introduce P and the relation P, Eo A. Default interpretation (Prospective Present): Introduce E’ (the pre-state of eventuality E) and la relation P, E’ B. Simple Future reading: If the sentence contains a time adverbial referring to an interval or time point t posterior to P, Eo, such as demain ‘tomorrow’: introduce R and the relations demain (t) and P – R, E, t.
The descriptive inadequacy of Reichenbach’s tense system 19
As for the Pluperfect of French, our approach can explain why in sentences such as (34) this tense form can be interpreted in two ways:
(34) A huit heures Marie-Cécile avait visité le musée. ‘At eight o’clock Marie-Cécile had visited the museum.’
This example can be interpreted as referring to the resulting state of the eventuality visiter le muse (‘to visit the museum’), which is true at perspective point P (P – Eo). In that case à huit heures ‘at eight o’clock’ localizes P. The other possibility is to understand (34) as referring to the eventuality E (= visiter le musée), which overlaps with the reference point R, E – P – Eo (E coincides with R, R is anterior to perspective point P, and P is anterior to the speech act Eo). In that case à huit heures ‘at eight o’clock’ localizes R. The interpretation rules for the Future Perfect are very simple since this form only refers to the resulting state E” which coincides with R, where P, Eo – R, E” (E” coincides with R). So we predict that example (35) has only one interpretation :
(35) A huit heures, Marie-Cécile aura visité le musée. ‘At eight o’clock Marie-Cécile will have visited the museum.’
In (35), à huit heures ‘at eight o’clock’ localizes R posterior to P,Eo and E” coincides with this R. Consequently this example only refers to a resulting state at some point in the future. The same is true for the Future Perfect of the Past (aurait visité ‘would have visited’). The Composed Past, the Pluperfect, the Future Perfect and the Future Perfect of the Past all contain the sequence avoir + PP ‘have + Past Participle’. This sequence can only have two interpretations if the sentence contains a Composed Past or a Pluperfect. I represent the two meanings of the Composed Past in Figures 6 and 7 below.
20 Co Vet
SENTENCE
TENSE ASPECT Predication | | x manger PRESENT RESULT | | P, Eo E”
Figure 6 : Resultative reading of the Composed Past (a mangé ‘has eaten’)
SENTENCE
TENSE ASPECT Predication
x manger PRESENT ANTERIOR ∅ | | | P, Eo R – P E
Figure 7: Anterior reading of the Composed Past (a mangé ‘ate’)
The Anterior reading of the Composed Past has developed from its resultative reading. In fact, the Anterior Present reading is presupposed by the resultative reading so that the meaning shift is quite easy. Figures 6 and 7 show a shift from aspect (Resultative) to tense (Anterior). In Figure 7 the aspect of the sentence is neutral (∅). Note that I adopt the representation of Figure 7 also for the Simple Past of French. The two possible readings of the Pluperfect are obtained by replacing PRESENT by PAST (P – Eo) in both figures. The Future Perfect and the Future Perfect of the Past have only a resultative reading. Figure 8 gives the representation for the Future Perfect (aura mangé ‘will have eaten’):
The descriptive inadequacy of Reichenbach’s tense system 21
SENTENCE
TENSE ASPECT Predication | x manger PRESENT POSTERIOR RESULT | | | P, Eo P – R E” Figure 8 : Future Perfect (aura mangé ‘will have eaten’) As for the Periphrastic Future we propose analogous representations for the two readings of this tense form:
SENTENCE
TENSE ASPECT Predication | | x manger PRESENT PROSPECTIVE | |
P, Eo E’ Figure 9 : Prospective reading of the Periphrastic Future (va manger ‘is going to eat’)
SENTENCE
TENSE ASPECT Predication | x manger
∅ PRESENT POSTERIOR | | | P, E P – R E o Figure 10 : Futurate reading of the Periphrastic Future (va manger ‘will eat’)
22 Co Vet
The representation of the Simple Future (mangera ‘will eat’) is the same as that of Figure 10. The readings of the Periphrastic Future of the Past (allait manger ‘was going to eat’ / ‘would eat’) are obtained by replacing the node PRESENT by PAST (P – Eo).
5. The overcomposed forms of French
As we have seen in section 2, French possesses an Overcomposed Past (a eu mangé ‘has had eaten’) and an Overcomposed Pluperfect (avait eu mangé ‘had-Imperfective-Past had eaten’). Both forms are regarded as substandard. In standard French the Anterior Past (eut mangé ‘had-Simple-Past eaten’ has the same meaning as the Overcomposed Past. In recent analyses of the Composed Past (cf. Molendijk, de Swart et al. 2004), it is assumed that this form has one meaning (the same as in Reichenbach, i.e. E – R,S). The Composed Past of English, French and Dutch are said to differ only with respect to the elements with which the eventuality introduced by the Composed Past can combine in discourse. In this one-meaning approach of the Composed forms, the existence of the Overcomposed forms remains quite mysterious since it is not clear what could be the contribution of the two identical sequences avoir + PP to the meaning of the form. One of the two would be superfluous since they would have the same semantic effect. In my approach, the two occurrences of avoir ‘have’ + PP in one sentence can be easily explained. Let us examine more closely example (36).
(36) Dès que Chantal a eu terminé sa thèse, elle est partie pour les États-unis. ‘As soon as Chantal had finished (literally : has had finished) her thesis, she left (literally : has left) for the United States.’
In this example, the first of the two avoir + PP has to be interpreted as ANTERIOR and the second one as RESULT :
SENTENCE
TENSE ASPECT Predication | PRESENT ANTERIOR RESULT∅ | | | Chantal terminer sa thèse P, Eo R – P E” avoir + PP avoir + PP Figure 11: Reading of the Overcomposed Past (a eu terminé ‘has had finished’)
The descriptive inadequacy of Reichenbach’s tense system 23
So in (36) the Overcomposed Past indicates that at some reference point R anterior to P (ANTERIOR) the result (E”) of the eventuality ‘Chantal terminer sa thèse’ (Finish her thesis, Chantal) was true. The representation of the Anterior Past, eut terminé ‘had-Simple-Past finished’, is the same as in Figure 11. In this form, the Simple Past of the auxiliary avoir ‘have’ (eut ‘had’) introduces the node ANTERIOR into the representation and avoir + PP the aspectual node RESULT. As for the Overcomposed Pluperfect the representation of sentences with this tense form can be obtained by changing PRESENT (P, Eo) into PAST (P – Eo). In my approach, it can also be explained why native speakers of French do not easily accept an Overcomposed Future (which would have the form aura eu terminé ‘will have had finished’). The explanation is that the Future Perfect (aura terminé ‘will have finished’) has not developed an ANTERIOR reading (it has only a resulting state reading). The same is true for the Future Perfect of the Past (aurait terminé ‘would have finished’). Note that the Periphrastic Future can also be combined with avoir + PP :
(37) Je vais avoir terminé. ‘I am going to have finished.’
In this kind of sentences, the only reading available for avoir + PP is the resultative one. (37) indicates that the speaker finds himself in the state preceding the resulting state of ‘je terminer’ ‘I finish’. This means that the speaker is finishing his or her job. An ANTERIOR reading is not available for (37). The difference with (38) is that in this example it is said that the speaker is involved in the state preceding the eventuality ‘je+ terminer’, ‘I finish’ itself.
(38) Je vais terminer. ‘I am going to finish.’
So the difference between (37) and (38) can be accounted for in my proposal. The past variant of (38):
(39) J’allais terminer (quand Marie est entrée dans mon bureau). ‘I was-Imperfective-Past going to finish (when Marie entered-Composed- Past my office).’ is often regarded by the grammars of French as a modal form. In my approach, it is not necessary to do so. In (39), it is said that Marie’s entering the office coincides with the preparatory state that normally precedes a finishing event (if E1: Mary enter my office and E’ the pre-state of ‘I finish’, then (41) means that E1 coincides with E’ (or perhaps better that E1 is included in E’)). The fact that from (39) it cannot be concluded that the speaker actually finished what he was doing does not mean that this sentence
24 Co Vet has a modal meaning. The same is true for (38), which is never regarded as modal by the grammars. (38) and (39) actually confirm my analysis in which the Periphrastic Futures can refer to the preparatory phase of an eventuality.
6. Conclusion
In this paper I have shown that Reichenbach’s tense system is descriptively inadequate. Some tense forms cannot be described in his framework, whereas other tenses whose existence it predicts do not occur in natural language. The main reason for this inadequacy is that Reichenbach divides the time axis into past, present and future and then situates the eventualities anterior, simultaneous and posterior to his three reference points. I have given arguments in favour of the existence of two perspective points (P), one in the past and the other in the present and that a tripartition into past, present and future should be made from these two perspective points. The tense system that results from these operations possesses six tense functions (meanings) and not nine as in Reichenbach’s system. Most grammars of French distinguish twelve or even fourteen tenses. It is evident that neither Reichenbach’s system nor mine have room for so many tenses. But a closer inspection of the meaning of these tenses reveals that they should not all be regarded as pure tenses. In fact all the composed forms have or may have an aspectual reading (RESULT). Some of the composed forms (the Composed Past and the Pluperfect) can be interpreted in two ways: as an aspectual form (RESULT) or as a temporal one (ANTERIOR). The Periphrastic Futures also have an aspectual interpretation (PROSPECTIVE) and a temporal one (POSTERIOR). The results of our analyses are listed in Table 4 below. Note that some tense functions can be expressed by more than one tense form.
TENSE FUNCTION ASPECT TENSE FORM Present je mange PRESENT ∅ ‘I am eating’ Composed Past j’ai mangé PRESENT RESULT ‘I have eaten’ Periphrastic Future je vais manger PRESENT PROSPECTIVE ‘I am going to eat’ Simple Past je mangeai PRESENT + ‘I ate’ ∅ ANTERIOR Composed Past j’ai mangé ‘I ate’ (literally ‘have eaten’) Anterior Past eut mangé ‘had-Simple-Past eaten’ PRESENT + RESULT Overcomposed Past a eu mangé ANTERIOR (literally ‘has had eaten’, same meaning as the Anterior Past)
The descriptive inadequacy of Reichenbach’s tense system 25
Simple Future je mangerai PRESENT + ‘I will eat’ ∅ POSTERIOR Periphrastic Future je vais manger ‘I will eat’ PRESENT + Future Perfect j’aurai mangé RESULT POSTERIOR ‘I will have eaten’ Imperfective Past je mangeais PAST ∅ ‘I was eating’ Pluperfect j’avais mangé PAST RESULT ‘I had-Imperfective-Past eaten’ Periphrastic Future of the PAST PROSPECTIVE Past j’allais manger ‘I was-Imperfective-Past going to eat’ Pluperfect j’avais mangé PAST + ANTERIOR ∅ ‘I had-Imperfective-Past eaten’ Overcomposed Pluperfect j’avais eu mangé PAST + ANTERIOR RESULT (literally ‘had-Imperfective-Past had eaten’) Simple Future of the Past je mangerais ‘I would eat’ PAST + POSTERIOR ∅ Periphrastic Future of the Past j’allais manger ‘I would eat’ Future Perfect j’aurais mange PAST + POSTERIOR RESULT ‘I would have eaten’ To obtain the meaning of the different forms the following rules have to be applied: PRESENT: P, Eo PAST: P – Eo ANTERIOR: R – P POSTERIOR: P – R The features RESULT and PROSPECTIVE introduce the resulting state (E”) and the preparatory state (E’) of the eventuality type described in the predication of the sentence and not the eventuality itself.
Table 4 : The tenses of French.
References
Corblin, F. ; de Swart, H., (eds), (2004). Handbook of French Semantics, Stanford : CSLI Publications. Dik, S. C. (1997). The Theory of Functional Grammar, Vol.1 : The Structure of the Clause, Berlin – New York : Mouton de Gruyter. Grevisse, M. (1970). Le bon usage : Grammaire française, Gembloux : Duculot.
26 Co Vet
Kamp, H. ; Reyle, U. (1993). From Discourse to Logic, 2 vol., Dordrecht: Kluwer. Molendijk, A. (1990). Le passé simple et l’imparfait: une approche reichenbachienne, Amsterdam – Atlanta : Rodopi. Molendijk, A. ; de Swart, H. et al. (2004). Meaning and use of the past tenses in discourse, in : F.Corblin and H. de Swart, (eds), 271 – 308. Reichenbach, H. (1966). Elements of Symbolic Logic, New York : The Free Press (originally 1947). Riegel, M. et al. (1994). Grammaire méthodique du français, Paris : Presses Universitaires de France. de Saussure, L. (2003). Temps et pertinence: Eléments de pragmatique cognitive du temps, Brussels : Duculot. de Swart, H. (1998). Aspect shift and coercion, Natural Language and Linguistic Theory 16 : 347 – 385. Verkuyl, H. ; Vet, C. et al. (2004). Tense and aspect in sentences, in: F.Corblin ; H. de Swart, (eds), 233 – 270. Vet, C. (1980). Temps, aspects et adverbes de temps en français contemporain, Geneva : Droz. Vet, C. (1994). Petite grammaire de l’Aktionsart et de l’aspect, Cahiers de Grammaire 19 : 1 – 17. Vet, C. ; Kampers-Manhe, B. (2001). Futur simple et futur du passé: leurs emplois temporels et modaux, in: P. Dendale ; L.Tasmowski, (eds), Le conditionnel en français, Recherches Linguistiques 25, Université de Metz, Metz, 89 – 104.
The evaluation of aspectual distance, speed and progress
Hans SMESSAERT K.U. Leuven
1. Introduction
The adverbial expressions to be discussed in this paper 1 belong to the general realm of imperfective aspect in that they concern the internal structure of events and states, i.e. their beginning, continuation or ending. The four basic elements in English, namely not yet, already, still and no longer in (1a), as well as their counterparts in various other European languages in (1b-d), have been discussed at length in the literature, and have been given various labels, as listed in (2) :
(1) a. (English) not yet / already / still / no longer b. (French) ne ... pas encore / déjà / encore / ne ... plus c. (German) noch nicht / schon / noch / nicht mehr d. (Dutch) nog niet / al / nog / niet meer
(2) a. perspectivity particles (Van Baar 1990 ; Vandeweghe 1992 ; Declerck 1994) b. particles of change and continuation (Van Baar 1991 ; van der Auwera 1993) c. phasal adverbials (Löbner 1990 ; 1999 ; van der Auwera 1998) d. phasal polarity items (Van Baar 1997).
In the case of Dutch, which will be the language focussed on, the basic aspectual meaning of the adverbials in (1d) may be augmented with various kinds of evaluative information. The aim of this paper is to chart the various constraints on the combination of objective aspectual information and subjective evaluative information. The investigation proceeds in four steps. First of all, the basic aspectual analysis of the adverbials in (1) is presented (section 2). Secondly, evaluative information about the distance towards the beginning or the ending of an event is added (section 3). Afterwards, evaluative information concerning the speed and progress of the actual course of events is considered (section 4). Finally, the three different types of
1 A preliminary version of this paper was presented at the Sixth Chronos Conference (Université de Genève, September 22-24, 2004). We thank the audience for their critical questions and comments.
© Cahiers Chronos 17 (2007) : 27-45. 28 Hans Smessaert evaluative information – i.e. distance, speed and progress – are shown to interact, not only among the three of them, but also with the basic aspectual information (section 5).
2. The basic aspectual analysis
The figure 1 below illustrates that the meaning of the four basic elements in Dutch can be defined in terms of the three binary parametres of ACTUAL POLARITY (AP), POLARITY TRANSITION (PT), and PERSPECTIVE (PS) in (3) :
nog niet al nog niet meer [B-] B [B+] [E+] E [E-] ------|++++++++++++++++++++++++|------negative positive positive negative beginning beginning ending ending prospective retrospective prospective retrospective
Figure 1 : a definition in terms of three binary parametres
(3) a. ACTUAL POLARITY AP = 1 (posit. polarity) AP = 0 (negat. polarity) b. POLARITY TRANSITION PT = 1 (begin = -/+) PT = 0 (end = +/-) c. PERSPECTIVE PS = 1 (retrospective) PS = 0 (prospective)
With the parametre of actual polarity in (3a) a distinction is drawn between the positive adverbials which locate the temporal reference point ‘inside’ the event, i.e. between the beginning B and the ending E in fig. 1, and the negative ones locating the reference time ‘outside’ the event. The parametre of polarity transition in (3b) distinguishes the beginning, which ‘switches on the event’ (from negative to positive polarity) from the ending, which ‘switches off the event’ (from positive to negative polarity). And finally, the parametre of perspective in (3c) encodes the difference between adverbials (retrospectively) ‘looking back upon an actual polarity transition’ and those which are (prospectively) ‘looking forward to an anticipated polarity transition’. Table 1 below defines the adverbials in (1-3) in terms of the three binary parametres of (3). This analysis allows one to represent the semantics of the basic aspectual adverbials as a three-digit bit-string, where each position encodes the value for one of the binary AP-PT-PS parametres. Assigning nog niet in table 1 (a) the bit-string ‘010’ for instance, thus serves as a compact way of specifying that the temporal reference point is located in an area of negative
The evaluation of aspectual distance, speed and progress 29 polarity [AP=0], is related to the polarity transition of beginning [PT=1] and is prospectively looking towards that transition [PS=0].
BASIC ASPECT AP PT PS
a. nog niet (‘not yet’) 0 1 0
b. al (‘already’) 1 1 1
c. nog (‘still’) 1 0 0
d. niet meer (‘no longer’) 0 0 1
Table 1 : Basic aspect Although the three binary parametres in (3) are strictly speaking logically independent in that each yields its own bipartition of the group of four adverbials in fig. 1, they are nevertheless mutually interrelated. As has been pointed out by Löbner (1990 : 134-135), Vandeweghe (1992 : 101) and van der Auwera (1998 : 42), only four out of the eight logically possible combinations of binary AP-PT-PS values are actually available 2. In other words, the four bit-strings in (6) are excluded, in principle :
(4) a. * 011 b. * 110 c. * 101 d. * 000
The particular predictions are stated in the first aspectual constraint in (5) :
(5) ASPECTUAL CONSTRAINT 1 (AC1) : a. If two parametres are 1, then the third parametre is 1 b. If two parametres are 0, then the third parametre is 1 c. If one parametre is 1 and the second is 0, then the third parametre is 0
Applying (5c), for instance, an AP-PT combination of ‘01’ can only yield the full ‘010’ string of nog niet in table 1 (a) but not the ‘011’ string of (4a). The conceptual motivation for AC1 is to be found in the mutually interdependent definitions of the parametres in (3) : ‘if one knows the values for any two of the binary parametres, one can predict the value of the third parametre’. The fact that the temporal reference point is located in a negative polarity area
2 See Smessaert & ter Meulen (2004 : 222-223) for a more detailed discussion, both of the relationship between the bit-string analysis developed here and the Löbner-Vandeweghe-van der Auwera approaches, and of the differences among the latter three.
30 Hans Smessaert
[i.e. AP=0] of the beginning transition [i.e. PT=1], e.g., necessarily implies a prospective perspective [PS=0]. Similarly, a prospective perspective [PS=0] with regards to a beginning transition [i.e. PT=1] is only possible when the temporal reference point is located in a negative polarity area [i.e. AP=0].
3. The evaluation of aspectual distance
In addition to the objective (or extensional) information encoded in the AP- PT-PS parametres of (3), many aspectual adverbials, especially the more complex ones, also convey additional subjective or evaluative (i.e. intensional) information. Consider the scenario of a ferry sailing into the harbour in (6a) or sailing out of the harbour in (6b) where faces on the deck and the pier start or stop being recognizable respectively :
(6) a. De gezichten waren toen NOG LANG NIET te herkennen. ‘The faces were then not recognizable yet by far.’ b. De gezichten waren toen BIJNA NIET MEER te herkennen. ‘The faces were then almost no longer recognizable.’ In the case of (6), the evaluative dimension concerns the distance between the temporal reference point – indicated by the time adverb toen (‘then’) – and the polarity transitions of beginning (B) or ending (E) : nog lang niet in (6a) locates the reference time at a long distance before the beginning, whereas bijna niet meer in (6b) locates it very close before the ending. Although language has many ways to quantify over distances, this analysis focuses on the most elementary opposition between long and short distances towards the polarity transition, which can thus (by simplifying) be encoded in the binary parametre for distance in (7) :
(7) EVALUATION OF DISTANCE D = 1 (long dist.) D = 0 (short distance)
Table 2 below reflects the straightforward combination of this evaluative D parametre (in the rightmost column) with the objective AP-PT-PS parametres of the previous section (in the three columns to the left) : each of the four basic adverbials in table 1 gives rise to two more complex adverbials specifying either a short or a long distance towards the PT 3 :
3 Notice, first of all, that Dutch tends to encode the notion of aspectual distance in a much more compositional way than English by stacking modifiers into adverbial clusters (with French presumably occupying an intermediate position). Furthermore, the adverbials in table 2 (a’), (b) or (c’) reveal that Dutch often has various ways of lexicalising a particular aspectual constellation. These different lexicalisation strategies and the crucial question of the scope relations between modifiers and negative particles are discussed in full detail in Smessaert 2006.
The evaluation of aspectual distance, speed and progress 31
EVALUATION OF ASPECTUAL DISTANCE AP PT PS D
a. nog lang niet (‘not yet by far’) 0 1 0 1
a’. nog net niet / bijna (‘just not yet / almost’) 0 1 0 0
b. net / nog maar pas (‘only just’) 1 1 1 0
b’. al lang (‘already for a long time’) 1 1 1 1
c. nog lang (‘still for a long time’) 1 0 0 1 nog net / bijna niet meer (‘still just / almost c’. 1 0 0 0 no longer’) d. net niet meer (‘just no longer’) 0 0 1 0
d’. lang niet meer (‘no longer for a long time’) 0 0 1 1
Table 2 : Evaluation of aspectual distance Whereas, logically speaking, the combination of four binary parametres should yield 16 different four-digit bit-strings, only eight of them actually show up in Table 2. No extra constraints need to be introduced here, however, since Table 2 simply inherits the ‘reduction by half’ in Table 1 / (4), which was accounted for in terms of the aspectual constraint AC1 in (5).
4. The evaluation of aspectual speed and progress
Apart from the aspectual evaluation of the distance between the reference time and the polarity transition discussed above, the subjective component of evaluation may also be due to an (implicit) relation of comparison between the actual course of events and possible alternatives. Consider for instance :
(8) a. De gezichten waren toen NOG ALTIJD NIET te herkennen. The faces were then still not recognizable (French : ne ... toujours pas) b. De gezichten waren toen EINDELIJK te herkennen. The faces were then finally recognizable (French : enfin) c. De gezichten waren toen al niet meer te herkennen. ?? The faces were then already no longer recognizable (French : ne ... déjà plus) The precise nature of the meaning difference between the basic adverbials in table 1 and the more complex ones in (8) has been a matter of considerable debate in the literature. The present proposal can be considered to occupy an
32 Hans Smessaert intermediate position in between the ‘minimalist’ approach of van der Auwera (1998) in terms of emphatic counterparts and the ‘maximalist’ approach of Löbner (1999) in terms of presuppositions 4. The idea is that the expressions in (8) combine two types of evaluation, namely that of speed and that of progress. In the case of nog altijd niet in (8a) the negative evaluation of the course of events as not making any progress entails its negative evaluation as lacking speed 5. At the other extreme, al niet meer in (8c) combines the positive evaluation of the course of events as having speed with that of making progress. The intermediate position of eindelijk in 8b) is then due to its peculiar constellation of mixed evaluation : on the one hand eindelijk resembles the negative nog altijd niet in (8a) in evaluating the course of events as lacking speed (as opposed to positive al niet meer) but on the other hand it resembles that positive al niet meer in (8c) in evaluating the course of events as making progress (in contrast to negative nog altijd niet). In order to be able to account for the mixed evaluation in (8b), the two evaluative notions of aspectual speed and progress are independently encoded in their respective binary parametres in (9) :
(9) a. EVALUATION OF SPEED S = 1 (speed) S = 0 (lack of speed) b. EVALUATION OF PROGRESS P = 1 (progress) P = 0 (lack of progress)
Furthermore, the resulting table 3 below strongly resembles that of Table 2 in the previous section in that the three objective AP-PT-PS parametres occur in the three leftmost columns, whereas the two evaluative S-P parametres occupy the two rightmost columns :
EVALUATION OF ASPECTUAL SPEED & PROGRESS AP PT PS S P
a. nog altijd niet (‘? still always not’) 0 1 0 0 0
b. eindelijk (‘finally’) 1 1 1 0 1
b’. al (‘already’) 1 1 1 1 1
4 Smessaert & ter Meulen (2004), which offers a detailed DRT analysis both for the non-evaluative aspectual adverbials in Table 1 and for the evaluative ones in (8), provides a much more thorough account of this so-called ‘intermediate’ position. 5 Notice that in Dutch the difference between the basic nog niet in Table 1 (a) and the evaluative nog altijd niet in (8a) is again a matter of compositionally inserting the altijd element, whereas in English the addition of the evaluative dimension turns not yet in Table 1 (a) into still not in (8a).
The evaluation of aspectual distance, speed and progress 33
c. nog altijd (‘? still always’) 1 0 0 0 0 eindelijk niet meer (‘? finally no d. 0 0 1 0 1 longer’) d’. al niet meer (‘? already no longer’) 0 0 1 1 1
Table 3 : Evaluation of aspectual speed and progress Restricting the attention to the evaluative component of the S and P parametres, one expects the two binary parametres to yield four logical combinations of values. As is observed in Smessaert & ter Meulen (2004 : 236-238), however, one of these combinations is excluded in principle, which is captured in the second aspectual constraint in (9) :
(9) ASPECTUAL CONSTRAINT 2 (AC2) : [ speed S = 1 ] => ¬[ progress P = 0 ]
According to this constraint one ‘cannot have speed without progress’. In other words, the presence of speed (value ‘1’ for S in column four) automatically entails the presence of progress (value ‘1’ for P in column five) 6. As a result of aspectual constraint AC2, the original four-way distinction generated by combining the two bipartitions is replaced by one tripartition ranging from ‘no progress’ [SP=‘00’], over ‘slow progress’ [SP=‘01’], to ‘fast progress’ [SP=‘11’]. Both the AC1 in (5) and the AC2 in (9) can be considered ‘local’ constraints in that with the former the interaction is internal to the objective non-evaluative component of the AP-PT-PS parametres, whereas with the latter the interaction is restricted to the evaluation component of the S-P parametres. Notice that the evaluative table in Table 3 fundamentally differs from the one in Table 2 since it reveals an asymmetry between the polarity areas which is absent from Table 2 : in the prospective areas of Table 3 (a) and (c) only one evaluative expression shows up (i.e. the negative one) while the retrospective areas in Table 3 (b-b’) and (d-d’) contain expressions with both mixed and positive evaluation. The purpose of the next section is then to investigate such ‘global’ constraints between the evaluative and the non- evaluative components of the aspectual system in more detail.
6 By virtue of the Law of Contraposition, the AC2 in (9) also comes in an alternative version : (i) ASPECTUAL CONSTRAINT 2' (AC2') : [ progress P = 0 ] => ¬[ speed S = 1 ] according to which the absence of progress (value ‘0' for P in column five) logically entails the absence of speed (value ‘0' for S in column four).
34 Hans Smessaert
5. The evaluation of aspectual distance, speed and progress
Although, for Dutch at least, the two particular subclasses of evaluative adverbials illustrated in Tables 2 and 3 have been studied in quite some detail (Smessaert & ter Meulen 2004 ; Smessaert 2006), the descriptions of the two groups have basically been independent of one another : the expressions in Table 2 do not evaluate for speed or progress, whereas those in Table 3 do not evaluate for distance. Nevertheless, this mutual exclusion is not a necessary one. And indeed, the focus of the present section will precisely be on expressions such as the ones in (10) which do combine all three evaluative dimensions :
(10) a. De gezichten waren toen NOG ALTIJD LANG NIET te herkennen. ‘The faces were then still not recognizable by far.’ b. De gezichten waren toen AL NET NIET MEER te herkennen. ?? ‘The faces were then already just no longer recognizable.’ With nog altijd lang niet in (10a) the reference time is far before the beginning (the contribution of lang), and the course of events lacks both speed and progress (the contribution of altijd). For al net niet meer in (10b) the addition of net indicates that the reference time is just beyond the ending and the addition of al signals a course of events with fast progress. The table in Table 4 below summarizes the values for the different parametres, and reflects the growing semantic complexity ranging from three to six parametres. The basic adverbials in Table (a/e) are characterised in terms of the three non-evaluative parametres only, whereas the most complex cases in (10a / Table 4 d) and (10b / Table 4 h) combine the three non-evaluative parametres with all three of the evaluative ones – i.e. distance, speed and progress – into a six-digit bit-string. As demonstrated in sections 3 and 4 respectively, two independent intermediate steps lead up to this full complexity : either the exclusive evaluation of aspectual distance yields the four valued system in Table 4 (b) and (f) or else the exclusive evaluation of aspectual speed and progress gives rise to the five valued system of Table 4 (c) and (g) :
EVALUATING ASPECTUAL AP PT PS D S P DISTANCE, SPEED & PROGRESS a. nog niet (‘not yet’) 0 1 0 - - -
b. nog lang niet (‘not yet by far’) 0 1 0 1 - -
c. nog altijd niet (‘still not’) 0 1 0 - 0 0
The evaluation of aspectual distance, speed and progress 35
nog altijd lang niet (‘still not by far’) d. 0 1 0 1 0 0 = (15a) e. niet meer (‘no longer’) 0 0 1 - - -
f. net niet meer (‘? just no longer’) 0 0 1 0 - -
g. al niet meer (‘? already no longer’) 0 0 1 - 1 1 Al net niet meer (‘? already just no longer’) h. 0 0 1 0 1 1 = (15b) Table 4 : Evaluating aspectual distance, speed and progress Even though (semantic) compositionality may not be straightforward for the assignment of the non-evaluative parametres to the basic adverbials nog niet and niet meer in Table 4, the table manifestly reveals the way in which both distance indicators, such as lang and net, and speed / progress indicators, such as altijd and al, independently contribute to the overall semantics of the complex adverbials. In the remainder of this section the various constraints on combinations of values in the full-fledged six valued system will be discussed for each of the four polarity areas in (3) from left to right : the [B-] area of nog niet (section 5.1), the [B+] area of al (section 5.2), the [E+] area of nog (section 5.3) and the [E-] area of niet meer (section 5.4).
5.1. The ‘not yet’ area [B-]
In Table 5 below the left-hand side of the bit-strings is kept constant since the not yet area is characterised as the negative polarity side [AP=0] of the beginning [PT=1] from which one looks forward into the future [PS=0]. The eight different bit-strings in Table 5 thus constitute the full set of logical combi-nations of the evaluative D-S-P parametres in the right-hand half of the table. Two distinct types of constraints are revealed in Table 5. First of all, the combinations in Table 5 (c-c’) are excluded for reasons internal to the evaluative component : by virtue of AC2 in (14) the presence of speed [S=1] is logically incompatible with the absence of progress [P=0].
EVALUATING ASPECTUAL AP PT PS D S P DISTANCE, SPEED & PROGRESS a. nog altijd net niet (‘still always just not’) 0 1 0 0 0 0
a’. nog altijd lang niet (‘still always by far not’) 0 1 0 1 0 0
36 Hans Smessaert
b. eindelijk bijna (‘finally almost’) 0 1 0 0 0 1
b’. AC3a 0 1 0 1 0 1
c. AC2 0 1 0 0 1 0
c’. AC2 0 1 0 1 1 0
d. al bijna (‘already almost’) 0 1 0 0 1 1
d’. AC3b 0 1 0 1 1 1
Table 5 : The ‘not yet’ area A more complex constraint arises in Table 5 (b’) and (d’), however, demonstrating the interaction between the evaluation component and the basic aspectual analysis :
(11) ASPECTUAL CONSTRAINT 3 (AC3) : [ PS = 0 ] & [ P = 1 ] => ¬[ D = 1 ] a. [prospective PS=0] & [progress P=1] & [speed S=0] => ¬[distance D=1] b. [prospective PS=0] & [progress P=1] & [speed S=1] => ¬[distance D=1]
In its general formulation AC3 states that making progress [P=1] towards a future polarity transition [PS=0] is incompatible with still being a long distance away from that transition [D=1] 7. As such, this constraint does not take into account the speed of that progress. Hence, it gives rise to two concrete implementations : (11a) excludes a constellation of slow progress which is still far before the transition of beginning in Table 5 (b’), whereas (11b) renders a combination of fast progress with such a long distance impossible in Table 5 (d’). In other words, only when there is no progress in Table 5 (a-a’) it makes sense to distinguish between short and long distances in looking ahead : neither slow progress with eindelijk bijna in Table 5 (b) nor fast progress with al bijna in Table 5 (d) has a primed counterpart for long distance. Obviously, the semantics of bijna (‘almost’) in the latter two cases contributes to the location of the temporal reference point close to the polarity transition.
7 As demonstrated for AC2 in (9), the Law of Contraposition may generate an equivalent formulation. However, this alternative will not be dealt with explicitly here for AC3 nor for any of the subsequent constraints.
The evaluation of aspectual distance, speed and progress 37
5.2. The ‘already’ area [B+]
Moving from the not yet area in Table 5 to the already area in Table 6 still concerns the beginning [PT=1], but the actual polarity is switched from negative to positive [AP=1] and the perspective is switched from prospective to retrospective [PS=1]. As was the case in Table 5, the first three values of the bit-strings (i.e. the values for the AP-PT-PS parametres) are kept constant, in this case as ‘111’. Once again, the eight different value assignments to the evaluative D-S-P parametres in the three columns on the right exhaust the logically possible combinations :
EVALUATING ASPECTUAL AP PT PS D S P DISTANCE, SPEED & PROGRESS ?? nog altijd bijna niet a. 1 1 1 0 0 0 (‘still always almost not’) AC4b a’. AC4a 1 1 1 1 0 0
b. eindelijk (‘finally’) 1 1 1 0 0 1
b’. AC5 1 1 1 1 0 1
c. AC2 1 1 1 0 1 0
c’. AC2 1 1 1 1 1 0
d. al net (‘already just’) 1 1 1 0 1 1
d’. al lang (‘already for a long time’) 1 1 1 1 1 1
Table 6 : The ‘already’ area Different types of constraints again account for the missing expressions in Table 6. First of all, the combinations in Table 6 (c-c’) correspond to those in Table 5 (c-c’) in that the AC2 of (9) prohibits the combination of presence of speed [S=1] with absence of progress [P=0]. The constraint underlying Table 6 (a) and (a’), by contrast, stems from the more complex interaction between the evaluative and the non-evaluative components of the aspectual analysis :
(12) ASPECTUAL CONSTRAINT 4 (AC4) : [ PS = 1 ] => ¬?? [ S = 0 ] & [ P = 0 ] a. [retrospective PS=1] & [distance D=1] => ¬[speed S=0] & [progress P=0] b. [retrospective PS=1] & [distance D=0] => ?? [speed S=0] & [progress P=0]
The basic reading of AC4 posits that ‘a retrospective perspective [PS=1] is very hard to combine with total absence of speed [S=0] and progress [P=0]’,
38 Hans Smessaert the idea being that the crossing of a polarity transition – which is inherent in retrospectivity – implies at least minimal progress. Similar to the under- specification for speed with AC3 in (11), AC4 in (12) does not take into account the distance beyond the polarity transition and hence gives rise to two more specific formulations. According to (12), having arrived far beyond the beginning [D=1] is altogether incompatible with lack of speed and progress in Table 6 (a’). In the case of (12b), however, where the polarity transition has only just been crossed [D=0], this incompatibility may be less radical, whence the question marks ?? instead of the regular negation symbol ¬. The form nog altijd bijna niet in Table 6 (a) does occasionally occur, but then it does so in contexts where the aspectual adverbial interacts with modal operators expressing (lack of) ability such as kunnen (‘be able to’) or lukken (‘manage to’). The final constraint in Table 6 (b’) is related to the AC2 – in that the interaction is again internal to the evaluative component only – but differs from it since it involves all three of the evaluative parametres :
(13) ASPECTUAL CONSTRAINT 5 (AC5) : [speed S = 0] & [progress P = 1] => ¬[distance D = 1]
According to AC5 a course of events which exhibits slow progress cannot result in a temporal reference point being located far beyond the polarity transition. In other words, the sense of relief accompanying the achievement of the transition with eindelijk (‘finally’) necessarily locates the reference time close to that transition.
5.3. The ‘still’ area [E+]
Within the area of positive polarity [AP=1], the shift from already to still involves switching to the polarity transition of ending [PT=0] as well as switching to a prospective perspective [PS=0]. As in Tables 5 and 6, the values for the AP-PT-PS parametres are kept constant in Table 7 below – this time as ‘100’ – and the eight combinations of values for the evaluative D-S-P parametres fully cover the logical possibilities. In two major respects the distributional pattern of prospective nog (‘still’) in Table 7 corresponds to that of prospective nog niet (‘not yet’) in Table 5. First of all, as was the case in Table 5 (c-c’) and Table 6 (c-c’), the evaluative incompatibility of the presence of speed [S=1] with the absence of progress [P=0], which is captured in the AC2 of (9), accounts for Table 7 (c- c’). Secondly, according to AC3 in (11), making progress [P=1] towards a future polarity transition [PS=0] runs into conflict with still being a long distance away from that transition [D=1] : (11a) excludes a constellation of slow progress which is still far before the transition of ending in Table 7 (b’),
The evaluation of aspectual distance, speed and progress 39 whereas (11b) renders a combination of fast progress with such a long distance impossible in Table 7 (d’). Both in the slow progress scenario with eindelijk bijna niet meer in Table 7 (b) and in the fast progress one with al bijna niet meer in Table 7 (d), the bijna (‘almost’) element locates the temporal reference point close to the polarity transition and neither of them has a primed counterpart for long distance :
EVALUATING ASPECTUAL AP PT PS D S P DISTANCE, SPEED & PROGRESS ?? nog altijd net a. 1 0 0 0 0 0 (‘still always just’) AC6a ?? nog altijd lang a’. 1 0 0 1 0 0 (‘still always for a long time’) AC6b eindelijk bijna niet meer b. 1 0 0 0 0 1 (‘finally almost no longer’) b’. AC3a 1 0 0 1 0 1
c. AC2 1 0 0 0 1 0
c’. AC2 1 0 0 1 1 0 al bijna niet meer d. 1 0 0 0 1 1 (‘already almost no longer’) d’. AC3b 1 0 0 1 1 1
Table 7 : The ‘still’ area One crucial difference remains, however, between negative prospective nog niet (‘not yet’) in Table 7 and positive prospective nog (‘still’) in Table 7. In AC6 in (14) below – which once more illustrates the complex interaction between the evaluative component and the basic aspectual analysis – the use of question marks again indicates that the aspectual constraint is weaker than most of the previous ones :
(14) ASPECTUAL CONSTRAINT 6 (AC6) : a. [end PT=0] & [prosp. PS=0] & [speed S=0] & [progr. P=0] => ?? [dist. D=0] b. [end PT=0] & [prosp. PS=0] & [speed S=0] & [progr. P=0] => ?? [dist. D=1]
The evaluative constellation in Table 7 (a-a’) concerns a possible tension between looking forward towards the ending on the one hand and the absence of speed and progress on the other hand. More in particular, if the anticipated end of the course of events [PT-PS=‘00’] is not approaching at all [S- P=‘00’], it seems rather irrelevant to qualify the distance between the temporal reference point and that finish as being either short in (14a) or long
40 Hans Smessaert in (14b). In a sense, the lack of progress can be reinterpreted as retrospect- tivity, which may thus neutralise or overrule the original prospective perspec- tive. Notice, however, that this type of constraint does not show up with nog niet (‘not yet’) in Table 5. In other words, in the absence of speed and pro- gress, predicting the start of an event (by measuring the distance towards the transition) in Table 5 is considered to be easier than predicting the ending of an event in Table 7. Obviously, further research is needed to account for this peculiar discrepancy between (negative polarity) prospectivity towards the beginning and (positive polarity) prospectivity towards the ending.
5.4. The ‘no longer’ area [E-]
The final area to be discussed, namely that of no longer, concerns the same polarity transition of ending [PT=0] as that of still in the previous section. Having crossed the polarity transition, however, the actual polarity is switched off to negative [AP=0] and the perspective changes from prospec- tive to retrospective [PS=1]. Hence the AP-PT-PS values are kept constant as ‘001’ for all eight logical combinations of D-S-P values in Table 8 :
EVALUATING ASPECTUAL AP PT PS D S P DISTANCE, SPEED & PROGRESS ?? nog altijd bijna a. 0 0 1 0 0 0 (‘still always almost’) AC4b a’. AC4a 0 0 1 1 0 0 b. eindelijk niet meer (‘finally no longer’) 0 0 1 0 0 1
b’. AC5 0 0 1 1 0 1
c. AC2 0 0 1 0 1 0
c’. AC2 0 0 1 1 1 0 al net niet meer d. 0 0 1 0 1 1 (‘already just no longer’) al lang niet meer d’. 0 0 1 1 1 1 (‘already no longer for a long time’) Table 8 : The ‘no longer’ area Unlike the discrepancy observed with the prospective adverbials between the negative nog niet (‘not yet’) cases in Table 5 and the positive nog (‘still’) cases in Table 7, the constellation of problematic combinations in the retrospective realm runs perfectly parallel with the negative niet meer (‘no
The evaluation of aspectual distance, speed and progress 41 longer’) expressions in Table 8 and the positive al (‘already’) expressions in Table 6. First of all, the combinations in Table 8 (c-c’) follow the pattern of all three of the previous polarity areas where the AC2 of (9) excludes the combination of presence of speed [S=1] with absence of progress [P=0]. Secondly, according to AC4 in (12), the crossing of a polarity transition, which yields a retrospective perspective [PS=1], requires at least minimal progress and is therefore hardly compatible with total absence of speed [S=0] and progress [P=0]. This constraint is stronger in (24a’), which reflects (20a), since one cannot have arrived far beyond the ending [D=1] without at least some progress. As was observed in connection with Table 6 (a) and (12b), however, where the polarity transition has only just been crossed [D=0], the incompatibility may be less radical. And indeed, the adverbial nog altijd bijna in Table 8 (a) may occasionally show up, especially when modifying quantificational expressions such as de helft (‘half of the’) or honderd (‘a hundred’). And thirdly, the excluded combination in Table 8 (b’) reflects the observation, captured in the AC5 of (13), that the efforts involved in achieving the transition with eindelijk (‘finally’) require the reference time to be located close to that transition. In other words, if a course of events is evaluated as making slow progress, its temporal reference point cannot end up far beyond the polarity transition.
6. Summary and conclusions
The aim of this paper was to chart the constraints on the ‘logical space” which is generated by six binary parametres used in the semantic analysis of aspectual adverbials, i.e. the three basic (non-evaluative) parametres of ACTUAL POLARITY (AP), POLARITY TRANSITION (PT) and PERSPECTIVE (PS) on the one hand, and the three evaluative parametres of DISTANCE (D), SPEED (S) and PROGRESS (P) on the other hand. Although, theoretically speaking, sixty-four logical combinations of six values are available (two to the power six) quite a few of them are excluded in principle. Previous research had already revealed two constraints, namely AC1 and AC2 :
(15) a. AC1 [64 - 32 = 32] Tables 5 + 6 + 7 + 8 Löbner (1990 : 134-5), Vandeweghe (1992 : 101), v.d. Auwera (1998 : 42) b. AC2 [32 - 8 = 24] Tables 5(c-c’) 6(c-c’) 7(c-c’) 8(c-c’) Smessaert & ter Meulen (2004 : 236-238)
Aspectual Constraint AC1, which concerns the mutual interdefinability of the three basic parametres in (5/15a), is clearly the most powerful constraint in reducing the initial number of possibilities by half. By virtue of AC1, the total number of combinations covered by the four tables in section five thus
42 Hans Smessaert adds up to thirty-two. Aspectual Constraint AC2 in (9/15b) further reduces that number by eight, since it excludes courses of events with speed but without progress, eliminating rows five and six in each of the four tables. In the present analysis, four new constraints have been formulated :
(16) a. AC3 [24 - 4 = 20] Table 5 (b’-d’) Table 7 (b’-d’) b. AC4 [20 - 4 = 16] Table 6 (a-a’) Table 8 (a-a’) c. AC5 [16 - 2 = 14] Table 6 (b’) Table 8 (b’) d. AC6 [14 - 2 = 12] Table 7 (a-a’)
Observe, first of all, that none of these constraints has a power comparable to that of AC1 or AC2 in (15). Secondly, two subgroups can be distinguished on the basis of their effect. The constraints AC3 in (16a) and AC4 in (16b) both exclude four extra combinations. According to AC3, which cancels rows four and eight in Tables 5 and 7, making progress towards a future polarity transition is incompatible with still being a long distance away from that transition, whereas AC4, which cancels rows one and two in Table 6 and 8, captures the observation that the crossing of a polarity transition requires at least minimal progress. The aspectual constraints AC5 in (16c) and AC6 in (16d), by contrast, are both responsible for the elimination of only two constellations, namely the fourth rows in Tables 6/8 and rows one and two in Table 7 respectively. In the former case, slow progress necessarily locates the temporal reference point close to the polarity transition, while in the latter case, the absence of any speed and progress renders the measuring of the distance towards an end point irrelevant. Table 9 gives an overview of which of the six binary parametres are involved in the formulation of the six Aspectual Constraints :
DISTANCE, SPEED & PROGRESS AP PT PS D S P [64 - 32 = 32] Aspectual Constraint 1 x x x [32 - 8 = 24] Aspectual Constraint 2 x x [24 - 4 = 20] Aspectual Constraint 3 x x x [20 - 4 = 16] Aspectual Constraint 4 x x x [16 - 2 = 14] Aspectual Constraint 5 x x x [14 - 2 = 12] Aspectual Constraint 6 x x x x Table 9 : Aspectual Constraints The two constraints from the literature in (15) can be considered local in that AC1 only involves the three basic, non-evaluative parametres whereas AC2 only concerns the two evaluative parametres of speed and progress. Among the four constraints introduced in section five above, only AC5 is local in this
The evaluation of aspectual distance, speed and progress 43 sense. Although being restricted to the evaluative component, AC5 is more complex than AC2 since it refers to all three of the evaluative parametres. In this respect, it is the result of integrating two dimensions of aspectual evalua- tion which had hitherto been analysed independently, namely the evaluation of distance on the one hand, and that of speed and progress on the other (witness sections three and four respectively). The remaining three constraints, by contrast, namely AC3, AC4 and AC6, demonstrate the more complex interaction between two evaluative pa- rametres and one or more non-evaluative basic parametres. Notice that the latter invariably include the parametre of perspective PS, thus confirming the central role which was assigned to this notion in the dynamic treatment of temporal reasoning with aspectual adverbials in Smessaert & ter Meulen (2004). By way of overall summary, Table 10 integrates the constellations for each of the four polarity areas which were analysed independently in Tables 5, 6, 7 and 8. Rows five and six of these original tables, which reflected the excluded combination ‘10’ for the S-P parametres, have been omitted altoge- ther. In other words, the eight by three matrix below yields the twenty-four logical combinations of six values that remain after the application of constraints AC1 and AC2 in (15) : S-P = 00 S-P = 01 S-P = 11
AP-PT-PS D=0 nog altijd net niet eindelijk bijna al bijna = 010 nog altijd lang D=1 AC3a AC3b niet ?? nog altijd bijna AP-PT-PS D=0 eindelijk al net niet AC4b = 111 D=1 AC4a AC5 al lang ?? nog altijd net eindelijk bijna al bijna D=0 AP-PT-PS AC6a niet meer niet meer = 100 ?? nog altijd lang D=1 AC3a AC3b AC6b ?? nog altijd bijna eindelijk niet al net niet D=0 AP-PT-PS AC4b meer meer = 001 al lang D=1 AC4a AC5 niet meer Table 10 : Constellation of the four polarity areas analyzed in Tables 5, 6, 7 and 8 The ‘horizontal’ distribution of the adverbial expressions across the three types of progress can be read in a broad and a narrow way. On the broad reading, the expressions with the question marks in the light shaded cells are
44 Hans Smessaert taken into account as well, yielding a symmetric pattern with six expressions for ‘no progress’ in the first column, as well as six expressions for ‘fast progress’ in the third column, but only four of them in the second column for ‘slow progress’. On the narrow reading, by contrast, the problematic cases are disregarded, giving rise to an increasing number of white cells from two on the left, over four in the middle to six on the right. A similar distinction between a broad and a narrow reading can be made when considering the ‘vertical’ distribution of the aspectual adverbials in Table 10 across the four polarity areas. The broad interpretation again yields a balanced pattern containing four adverbials in each polarity area, whereas the narrow one reveals an asymmetric distribution with three adverbials in both retrospective polarity areas, but a discrepancy between four white cells in the negative prospective ‘010’ area of nog niet (‘not yet’) as opposed to two such cells in the positive prospective ‘100’ area of nog (‘still’). Finally, the distribution of the four aspectual constraints across the four polarity areas in Table 10 is by and large determined by the third basic parametre of perspective PS, which – as observed in connection with Table 9 – is crucially involved in the formulation of four out of the six constraints. On the one hand, AC3 holds for the two prospective areas, on the other hand AC4 and AC5 both concern the two retrospective areas. The vertical asymmetry in Table 10 is due to AC6, which only applies to one prospective area, i.e. the ‘100’ area of nog (‘still’). One important question for further research thus remains why a retrospective perspective does not seem to make any difference between looking back upon a transition of beginning and looking back upon that of ending, whereas in the prospective case of anticipating polarity transitions, looking forward towards the beginning seems more obvious than looking forward towards the end.
References
Declerck, R. (1994). The only/already puzzle : A question of perspective, Cognitive Linguistics 6(4) : 307-350. Löbner, S. (1990). Wahr neben Falsch. Duale Operatoren als die Quantoren natürlicher Sprache, Tübingen : Max Niemeyer Verlag. Löbner, S. (1999). Why German schon and noch are still duals : a reply to van der Auwera, Linguistics and Philosophy 22 : 45-107. Michaelis, L. A. (1996). On the use and meaning of already, Linguistics and Philosophy 19 : 477-502. Mittwoch, A. (1993). The relationship between schon/already and noch/still : a reply to Löbner, Natural Language Semantics 2 : 71-82. Smessaert, H. (2006). Elementary imperfective quantification and the scope of negation, Ms.
The evaluation of aspectual distance, speed and progress 45
Smessaert, H. ; ter Meulen, A.G.B. (2004). Temporal reasoning with aspectual adverbials, Linguistics and Philosophy 27/2 : 209-262. Van Baar, T. (1990). The Dutch Perspectivity Particles in FG (Working Papers in Functional Grammar 36), Amsterdam : Universiteit van Amsterdam. Van Baar, T. (1991). Adverbs and Particles of Change and Continuation outside Europe, in : J. van der Auwera, (ed), Adverbs and Particles of Change and Continuation (EUROTYP Working Papers, Volume 2), Strasbourg : European Science Foundation, 117-130. Van Baar, T. (1997). Phasal Polarity. Studies in Language and Language Use 30, Amsterdam : IFOTT. van der Auwera, J. (1993). ‘Already’ and ‘still’ : beyond duality, Linguistics and Philosophy 16 : 613-653. van der Auwera, J. (1998). Phasal adverbials in the languages of Europe, in : J. van der Auwera ; D.P. Ó Baoill, (eds), Adverbial constructions in the languages of Europe, Berlin : Mouton de Gruyter, 25-145 Vandeweghe, W. (1992). Perspectivische evaluatie in het Nederlands : de partikels van de AL/NOG/PAS-groep, Gent : Koninklijke Academie voor Nederlandse Taal- en Letterkunde.
This page intentionally left blank
The grammaticalization of tense markers : A pragmatic reanalysis
Steve NICOLLE BTL Kenya and SIL International
1. Introduction
Grammaticalization is the process whereby lexical categories and constructions containing lexical material develop, in specific morphosyntactic contexts, into GRAMS, that is, members of functional categories, including tense and aspect markers 1. It is generally agreed that grammaticalization involves both structural and semantic change, and that semantic change either precedes structural change (Givón, 1991 ; Nicolle, 1998) or occurs in parallel with it (Bybee et al. 1994 : 106 ; Emanatian 1992 : 19) 2. To begin with, certain inferences frequently associated with the use of the grammaticalizing expression become conventionalized, and the expression undergoes ‘semantic bleaching’ (or ‘generalization’), in which components of meaning relating to perceivable events and relations between entities is lost. The expression typically also develops increased functional dependence on associated lexical material. Following these semantic changes, structural changes occur which result in increasingly restricted morphosyntactic distribution, and possibly affixation, together with corresponding phonetic reduction. This widely accepted model of grammaticalization is represented in (1) :
(1) change of use change of meaning change of form (pragmatic) (semantic) (syntactic + phonological)
It is this model that I wish to question through an analysis of GRAMMATICAL- IZATION CHAINS involving source constructions containing verbs of movement 3.
1 This is a definition of what Traugott (2002) refers to as “primary grammatical- ization” ; I will not be concerned with ‘secondary grammaticalization’ which concerns an increase in the formal correlates of grammaticalization such as morphological fusion and phonetic erosion. 2 For an alternative view, in which autonomous syntactic change precedes semantic change, see Lightfoot (1991), Warner (1993), and discussion in Hopper & Traugott (1993 : 68) and Nicolle (1998 : 15-16). 3 Traugott (2002) argues that ‘pragmatic polysemy’ in the form of Generalized Invited Inferences, rather than coded (semantic) polysemy, is the necessary pre-
© Cahiers Chronos 17 (2007) : 47-65. 48 Steve Nicolle
A grammaticalization chain (or PATH) is a series of developments leading from a lexical construction to a gram (Heine et al. 1991 : 222 ; Heine 1992). It has been observed that many very similar grammaticalization chains occur in typologically, geographically and genetically distinct languages (Bybee et al. 1994 ; Heine et al. 1993 ; Koch 1999 ; Kuteva 2001). One of the most frequently cited grammaticalization chains is the development of constructions containing movement verbs into tense and aspect markers (Bybee et al. 1991 ; Emanatian 1992 ; Traugott 1978 ; Ultan 1978). Before considering this kind of grammaticalization chain let us define what is meant by the term MOVEMENT VERB. A movement verb is defined here as a verb which expresses change of location on the part of the subject (as opposed to motion in a single location or movement of the object of the verb). All movement events are perceived as involving the semantic components path, manner, figure and ground (Talmy 1985), and most movement verbs incorporate at least one of these semantic components along with the notion of change of location ; this is termed CONFLATION. Path is most frequently conflated in movement verbs, followed by manner and then (rarely) figure (Thornell 1997 : 165). Conflation of path is expressed by the verbs arrive, leave, enter, pass, descend and return, conflation of manner by walk, fly and swim, and conflation of figure by flow (the figure must be a liquid). Finally, movement verbs may be deictic or non-deictic. In the case of deictic verbs such as come and go, the default source or goal (starting point or end point) of the motion event is the location of the speaker or (less often) the hearer, whereas this is not the case with non-deictic movement verbs such as move, travel, leave and approach (Radden 1996). The most common sources of tense and aspect markers are deictic movement verbs which involve conflation of path ; typically these are verbs corresponding to come (to) and go (to) in English 4. The development of movement verbs into tense and aspect markers typically has the following characteristics (based on Hopper & Traugott 1993 : 1-4) : 1. Constructions (containing certain lexemes as components) undergo grammaticalization rather than lexemes per se. 2. Lexical sources of grams are general in meaning. 3. Grammaticalization involves morphosyntactic reanalysis ; in particular, there is an increase in C-command scope.
requisite to grammaticalization ; in the light of the data presented below this claim also is too strong. 4 Verbs which indicate movement of the figure towards the ground are common sources for future tenses and prospective aspects, and verbs which indicate that the figure moves from the ground are common sources of pasts and perfects (Bybee et al. 1994 ; Demirdache & Uribe-Etxebarria 2000 : 181-2).
The grammaticalization of tense markers 49
4. Grammaticalization involves phonological reduction, sometimes as a result of morphosyntactic reanalysis e.g. going to > gonna. 5. Various stages of grammaticalization may coexist ; that is, grams may exhibit structural and phonological allomorphy. (The lexical source construction may also continue to be used concurrently with the gram that derives from it.) 6. Concrete meaning such as physical motion and directionality is lost when grammaticalization occurs. Using verb-to-tense / aspect grammaticalization chains as a norm, I will evaluate other instances of linguistic change from English and Digo (a Bantu language spoken in East Africa) which take constructions involving verbs of movement (in particular the verbs come and go) as their starting points. I will demonstrate that the constructions in question behave structurally like tense / aspect markers but refer semantically to physical movement. Rather than being intermediate stages in grammaticalization chains leading ultimately to purely temporal markers, these constructions show no sign of developing temporal, aspectual or modal meanings. I will argue that the structural changes that have occurred in these constructions are the result of discourse-pragmatic factors rather than semantic change, which in turn suggests that the structural and semantic changes that typify grammatical- ization are epiphenomena : symptoms rather than causes.
2. Grammaticalization without semantic change in English
The construction which I will discuss in this section has been referred to as a kind of ‘double-verb construction’ (Carden & Pesetsky 1977), the ‘quasi- serial verb construction’ (Pullum 1990) and ‘aspectual come and go’ (Jaeggli & Hyams 1993). In order to avoid the theoretical bias attached to these terms, I will follow Pullum (1990) and refer to the construction as go get. The go get construction has the form Vmovement Vmain and is illustrated below.
(2) Let’s go find the paragraph marker. (Said whilst ‘moving’ through a document displayed on a computer screen.) (3) Come wave goodbye.
As stated in section 1, the aim of this paper is to demonstrate that certain constructions, including the movement verb in go get, behave syntactically as tenses whilst encoding non-temporal meaning which is identical to that of their source constructions (thereby demonstrating that grammaticalization need not involve semantic change). It is therefore necessary to show that go in go get behaves syntactically as a tense marker, and that it shares with its source construction the semantic component of physical movement. I will begin by arguing that go get derives diachronically
50 Steve Nicolle from fake coordination and shares with fake coordination the semantic component of physical movement. I will then argue that in finite clauses, go in go get is best analyzed syntactically as the head of TP, that is, as a tense marker.
2.1. Fake coordination as the source of go get
Fake coordination has the form Vmovement and Vmain as in (4) and (5), and occurs in all moods and tenses and with a number of movement verbs, all of which express conflation of path, except for run which expresses conflation of manner.
(4) Professor Bright has agreed to come and speak to us this weekend. (5) Would you run and get some more milk ?
The go get construction on the other hand is restricted to the verbs come and go and does not occur with overt inflection, that is, in the past tense or third person singular non-past tense without auxiliary do.
(6) a. Let’s go watch the match. b. We go watch a match every week. (7) a. *We went watched a match. (OK : We went and watched a match.) b. Did they go watch a match ? (8) a. *She goes watches a match every week. (OK : She goes and watches a match every week.) b. Does she go watch a match every week ?
Visser (1969), Zwicky (1969) and Carden & Pesetsky (1977) suggest that go get derives from fake coordination by a process of and-elision, but this suggestion is rejected by Shopen (1971), Pullum (1990) and by Jaeggli & Hyams (1993) 5. I will discuss the arguments against a source in fake coordination and show that none are convincing. Shopen (1971), Pullum (1990) and Jaeggli & Hyams (1993) observe that verbs in fake coordination cannot stack as they can in go get ; thus (9b) is only grammatical if interpreted as ordinary coordination. Stacking in go get is also unacceptable in finite clauses, as (9c) illustrates.
5 Jaeggli & Hyams (1990 : 317 fn. 4) do suggest that the construction in (i) in which the second verb is a participle which is phonologically identical with its base form may be derived from a coordinate structure. However, since (i) is only acceptable for some speakers, Jaeggli & Hyams do not pursue this analysis. (i) Bill has come put a copy of his new paper on my desk.
The grammaticalization of tense markers 51
(9) a. Come go eat with us. b. *Come and go and eat with us. c. *They come go eat with us every week.
This observation simply shows that go get and fake coordination have different syntactic structures, which is obviously true (see (7a) and (8a) above). It is irrelevant as far as the claim that go get derives from fake coordination is concerned since this claim does not entail that (9a) is derived from (9b). Stacking is a synchronic property of go get which need not be a property of its source construction. The authors mentioned above fail to mention (and therefore explain) that although stacking is acceptable in imperatives, it cannot occur in finite clauses (see (9c)). Pullum (1990 : 226) suggests that another syntactic distinction is that the first verb in fake coordination can take a complement whereas go in go get cannot. The extraction in (10c) is intended to demonstrate that (10b) is a case of fake rather than ordinary coordination.
(10) a. Go and read something. b. Go away and read something. c. What do you want me to go away and read ? d. What do you want me to go away and read ? e. *What do you want me to go away read ?
Leaving aside the fact that grammaticalization occurs in specific morphosyntactic contexts (e.g. complement-less ones) and so one need not expect go get to take complements, and the fact that go away differs semantically from go just as throw away differs from throw, Pullum’s extraction argument is not sound. (11a) is not a case of fake coordination (the ordinary coordination reading should be ignored) and neither is (11b), and yet (11c) is acceptable.
(11) a. *Walk and read something. b. Walk away and read something. c. What do you want me to walk away and read ?
Jaeggli & Hyams (1993 : 320) note that coordination allows various verbs in V1 position whereas go get is restricted to go and come (although they fail to distinguish fake from ordinary coordination, and the example they provide is one in which three verbs are stacked and so is ungrammatical with go get for independent reasons). This is also irrelevant, since grammatical- ization often involves certain lexemes but not others. There is no more reason why all of the verbs which occur in V1 position in fake coordination should grammaticalize in the same way that go and come have than that all the verbs
52 Steve Nicolle that occur with the progressive aspect and allative to should grammaticalize as go has in the be going to future construction. Visser (1969 : 1399) states that fake coordination but not go get “already occurs in (late) Old English, which seems to indicate that ‘go see’ developed from ‘go and see’ by elision of the conjunction.” Pullum (226, fn. 9) claims that this is implausible “since go get did not take over from go & get but rather proceeded to coexist with it for a clear six or seven centuries.” This is irrelevant ; structural and phonological allomorphy are common features of grammaticalizing constructions, as noted in section 1 above (point 5 in the list based on Hopper & Traugott 1993). The authors who argue against a source in fake coordination all note that non-agentive (or non-volitional) subjects occur in fake coordination but not in go get constructions. What they fail to mention is that when a non- agentive subject occurs with fake coordination the utterance receives either an unexpected event reading with go as in (12b), or it should be interpreted as a case of genuine coordination (13b, c). The unexpected event reading is also possible (but not obligatory) with go (but not come) and an agentive subject ; in (12c, d) a sense of disappointment or disapproval of the subject’s actions is conveyed as well as the unexpected nature of the event. Note, however, that with an unexpected event reading, physical movement need not be conveyed. For this reason I view fake coordination and go get with an unexpected event reading as distinct constructions from those denoting physical movement.
(12) a. *The bottle may go break. b. The bottle went and broke. c. John went and broke the bottle. d. Did John go break the bottle ? (13) a. *Pieces of driftwood come wash up on the shore. b. Pieces of driftwood come and wash up on the shore. c. Pieces of driftwood come in on the tide and wash up on the shore.
Jaeggli & Hyams (1993 : 321) propose that (13b) is acceptable because come is a main verb and therefore does not impose selectional restrictions on its subject, whereas come in (13a) selects an agentive subject ; that is, it assigns a ‘secondary’ θ-role to the subject NP (ibid. 325). This argument fails to rule out fake coordination as the source of go get for two reasons. First, if fake coordination is the source of go get it does not necessarily follow that the two constructions must assign identical θ-roles. Grammaticalization occurs in specific morphosyntactic contexts but the resulting construction may subsequently generalize to other contexts (for example, the be going to future originated in clauses containing subjects capable of physical movement, but subsequently generalized to contexts with other kinds of
The grammaticalization of tense markers 53 subject). I would tentatively suggest that go get is derived from fake coordination involving agentive subjects (as for example in imperative clauses) and has retained this selectional restriction. Second, main verbs are predicates, but go and come in fake coordination do not behave like typical predicates. This can be seen by comparing the behaviour of go / come in fake coordination with the go get construction and also with go / come before infinitival to. In the examples below, fake coordination patterns the same as go get (where, according to Jaeggli & Hyams, go is NOT a main verb) and differently from go with infinitival to (where go is a main verb). Only (14c) can be felicitously answered by (15) since the main event in (14a) and (14b) is the act of buying, whereas in (14c) it is the act of going.
(14) a. Did she go buy apples ? b. Did she go and buy apples ? c. Did she go to buy apples ? (15) Yes, but there weren’t any.
Unfortunately no alternative source for the go get construction is proposed in any of the accounts that reject fake coordination. Although Pullum (1990 : 227) notes the similarity between serial verb constructions and coordination, he rejects the idea that there is any diachronic connection between them in English. This means that the symmetry between V1 and V2 as regards overt morphological marking must be stipulated in his analysis rather than independently motivated. Having looked at the evidence against treating fake coordination as the source of go get, let us look at the evidence in favour. Circumstantial evidence comes from the fact that try and get and be sure / certain and get pattern syntactically exactly like go get, as noted by Carden & Pesetsky (1979) 6. This suggests that go get may also have had a source in go and get but whilst and-elision occurred with go get, the corresponding constructions with try and be sure / certain developed without and-elision. However, the strongest evidence in favour of viewing fake coordination as the source of go get is simply that (with the exception of the unexpected event reading (12b)) the two constructions are virtually synonymous, as (14a) and (14b) demonstrate. If we accept that go get derives from fake coordination, this develop- ment exhibits almost all the typical characteristics of grammaticalization. It involves a limited set of general movement verbs (typical source constructions for tense markers) in a particular syntactic frame ; go get exhibits phonological reduction in comparison with fake coordination ;
6 Contrary to what is claimed by Pullum (1990 : 224), Carden & Pesetsky do not equate the go and get and try and get constructions.
54 Steve Nicolle allomorphy applies because we see go get constructions, fake coordination and genuine coordination co-occurring in contemporary English ; and go get constructions exhibit syntactic reanalysis (lack of overt inflection, ability to stack). The only typical characteristic of grammaticalization that is missing is the loss of concrete semantic content, specifically physical motion and directionality. Thus, despite exhibiting all the structural characteristics of a gram, go get encodes the same semantic content as its source construction suggesting that semantic change is not a necessary prerequisite for formal grammaticalization to occur.
2.2. A syntactic analysis of go get
In section 2.1 I suggested that go get derives from fake coordination involving agentive subjects (as for example in imperative clauses). In this section I will elaborate this proposal and argue for an analysis of go get with a single semantic content and phonological realization but two underlying syntactic structures, dependant on the morphosyntactic context in which it occurs. My hypothesis is that go get originated in some kind of infinitival fake coordination : imperatives, subjunctives, or infinitival complements (bare infinitives with modals or to-infinitives following verbs such as remember). Given the subjective nature of both go get and fake coordination (see section 2.3 below) and the fact that go get requires agentive subjects, imperatives are the most likely source 7. Imperative go get can be derived from go and get by and-elision, and can be described synchronically as a serial verb construction (Baker 1989 ; Pullum 1990) 8. I hypothesize that the resultant non-finite serial verb construction was extended to other non-finite clauses without the need for any further syntactic reanalysis. However, it is not obvious how (or indeed whether) a serial verb analysis can explain the use of go get in simple declarative clauses in which the ‘no overt morphology’ rule applies. Pullum (1990 : 235) admits as much when he concludes, “I am not yet ready to
7 Jaeggli & Hyam (1993 : 322) note that go get does not occur with stative verbs, since the subject of a stative clause cannot be agentive. Unfortunately their examples involve stative verbs in imperative clauses, which are infelicitous in any case, and the same restriction applies to fake coordination : (i) ? Did she come know the answer ? (ii) ? She came and knew the answer. 8 Jaeggli & Hyams (1993) analyze go and come as auxiliary verbs. However, they behave differently from other auxiliary verbs in various ways and Jaeggli & Hyams require a number of stipulations in order to maintain their account, so I will not discuss their analysis further.
The grammaticalization of tense markers 55 provide a formal account of the phonological, morphological, and syntactic aspects of the inflection condition” (that is, the ‘no overt morphology’ rule). Simple declarative clauses with go get are not tenseless (the semantic interpretation is of non-past tense) and so they must contain a T(ense) projection. The question is why no overt morphology can occur under TP. If we adopt the minimalist account of tense and aspect proposed in Giorgi & Pianesi (1997), fake coordination can be represented with the structure in (16). The functional head AGR/T 9 checks agreement and assigns a T-role within a VP, and since both verbs must express the same person and tense morphology, the conjoined phrase is governed by a single AGR/T node (illustrated by the arrows from AGR/T).
(16) AGR/T-P
AGR/T VP
go and VP
V
In the go get construction however, no overt morphological marking is possible on either verb, as examples (7) and (8) illustrated. In Nicolle (2002) I accounted for this by suggesting that the movement verb go or come functions as the head of the AGR/T projection. Since the AGR/T node has been filled by go or come, there can be no phonologically realized person or tense morphology on the following verb. This is illustrated in (17).
9 Giorgi & Pianesi (1997) argue for the composite category AGR/T in English (in contrast to the Split-Infl hypothesis which posits separate AGR and T heads) because either tense or agreement, but not both, can be morphologically marked on a verb in English : (i) He loves/loved/*loveds
56 Steve Nicolle
(17) AGR/T-P
go VP
V In more recent versions of minimalism (Butler 2004 ; Stowell in press) AGR is no longer used and T alone occurs as a head. Following Butler (2004) for example, the structure in (17) would be represented as follows :
(18) CP
C TP
go CP
C vP
v VP
V
The analysis of go get as head of TP is not applicable when go get occurs in non-finite clauses such as (19) and (20) below. According to Butler (2004), when a modal takes T within its scope only epistemic readings are allowed, and when a modal occurs within the scope of T only deontic readings are allowed. This same claim is made in Role and Reference Grammar (Van Valin & La Polla 1997) where epistemic modality is termed ‘status’ and takes tense operators within its scope, whilst (deontic) modality operators must occur within the scope of tense. (19) is ambiguous between an epistemic reading (‘it is possible that the subject will buy some apples’) and a deontic reading (‘the subject is permitted to buy some apples’), which suggests that may does not occupy a fixed position relative to T. This can be explained by assuming that there is no T node. If may is replaced by must,
The grammaticalization of tense markers 57
(19) can only receive a deontic interpretation (‘the subject is required to buy some apples’), therefore go could not be the head of TP in this clause even if TP existed. Similarly, in (20) the verb remember takes a non-finite complement which means that go cannot be the head of TP, either because there is no T-node in infinitives (pace Butler 2004) or because the position of head of TP is occupied by infinitival to (Roberts 1993 ; Giorgi & Pianesi 1997 : 82).
(19) She may go buy some apples. (20) She remembered to go buy apples.
There are therefore two syntactic structures underlying a single phonological realization of go get. In non-finite clauses, go get can be analyzed as a serial verb construction, but in finite clauses go is the head of TP (= AGR/T-P or tense operator in RRG). This analysis explains why stacking of movement verbs is acceptable in imperatives but not in finite clauses, as illustrated in (9) above. It is also consistent with saying that morphosyntactically go get is fully grammaticalized in finite clauses, as tense is clearly a grammatical category. This can be confirmed by applying the C- command scope test (Tabor & Traugott 1998), which states that the C- command scope of a gram (in this case go get) must be greater than the C- command scope of its source (in this case fake coordination) in the same syntactic context. However, go get still encodes physical movement as part of its semantic content rather than the temporal meanings characteristic of tense and aspect markers. In short, the structural changes associated with grammaticalization have occurred in the go get construction without semantic change.
2.3. A pragmatic analysis of go get
I have proposed that go get developed from fake coordination involving the verbs come and go in the context of imperative clauses (hence with agentive subjects). In this section, I will briefly describe what I take to be the motivation behind the grammaticalization of go get. Whenever go get or fake coordination involving come or go is used in an utterance, it contributes to what in cognitive grammar (Langacker 1990) is termed ‘subjectification’. Subjectification occurs when the perspective of the ‘viewer’ or ‘conceptualizer’ of an event (typically the speaker) is incorporated into the description of that event. Now, all deictic movement verbs have the effect of “providing the deictic anchoring of a situation with respect to the speaker or the hearer” (Radden 1996 : 431), hence whenever a deictic movement verb is used, the event being described is subjectively construed. Similarly, when a deictic movement verb is linked with another
58 Steve Nicolle verb by coordination or juxtaposition, the result is a subjectified construal of both the action of moving and the other event. As frequency of use diminishes the force of the deictic movement verb, the perspective of the conceptualizer becomes incorporated into the description of the event described by the main verb, whilst less prominence is given to the act of physical movement. So, in an utterance of ‘Come and eat’ or ‘Come eat’ the pragmatically most salient verb is eat, and come expresses the speaker’s perspective (the eating is to be done where the speaker is situated – physically or conceptually – and not somewhere else). Thus the function of the deictic movement verb has changed whilst its semantic content remains unchanged10. Frequency of use and pragmatic weakening of the movement verb would also contribute to phonological erosion (Bybee 2003) leading to and- elision, thereby beginning the process of formal grammaticalization which eventually resulted in syntactic reanalysis of go get in finite clauses (as described in section 2.2).
3. Grammaticalization without semantic change in Digo 11
In this section I will discuss grammaticalized constructions in Digo which share some of the characteristics of go get and fake coordination in English. Like go (and) get, these constructions express physical movement in relation to a deictic centre, and hence contribute to a subjectified construal of associated events. Some of them behave morphosyntactically exactly like tenses (similar to go get in finite clauses) while others exhibit behaviour which is both verb-like and tense-like, (similar to fake coordination and go get in non-finite clauses). Grammaticalized constructions which encode physical movement and direction as their semantic content (termed ‘movement grams’ in Nicolle 2002) seem to be widespread in Bantu languages. Movement grams are distinct from itive and ventive markers 12, which express direction towards
10 This is in contrast to other subjectified expressions such as He came round slowly, I have come to appreciate modern jazz and They came to blows, in which physical movement is no longer part of the intended meaning (see Radden 1996). 11 Digo is a Bantu language spoken near the coast in Kenya and Tanzania and classified as E.73 (Guthrie 1967-71) or North-East Coast (Nurse 1999). Research into Digo was conducted in Kenya under Research Permit No. OP.13/001/17 C 180/20 issued by the Government of Kenya. For an overview of Digo grammar see Nicolle (2004). 12 These are the terms employed in, for example, Dimmendaal (1983) and Bourdin (2000). The terms ‘andative’ and ‘allative’ are sometimes used for ‘itive’, and
The grammaticalization of tense markers 59 and away from a point of reference respectively. Although itive and ventive markers are often derived from verbs meaning ‘go’ and ‘come’, they must be accompanied by a motion verb such as ‘jump’, ‘move’, ‘throw’ or ‘drive’ (Heine et al. 1993 : 103-108), whereas movement grams encode both direction and movement and can co-occur with most verbs. Of the five movement grams in Digo, I will discuss just three : edza (‘movement towards the deictic centre’), enda (‘movement away from the deictic centre’) and cha (‘action occurs at a distance from the deictic centre’).
3.1. Movement towards and away from the deictic centre : edza and enda
Edza (and its variant kpwedza) means ‘come (to)’ and can function both as a main verb and as a movement gram. When it functions a main verb, it is preceded by a tense / aspect prefix or infinitive prefix, and occurs word finally. This is illustrated in example (21), in which the two occurrences of edza used as a main verb are underlined. In the first occurrence, edza is preceded by a locative noun class prefix (ku-) and a past tense prefix (a-) which combine to form kpwa- ; its function here is to introduce new participants (atu ‘people’) into the story. In the second occurrence, the arrival of the men-folk is treated as a separate event which is encoded by the predicate edza preceded by a third person plural noun class prefix (a-) and a perfective aspect prefix (ka-, which reduces to k- before a vowel).
(21) Kukacha, KPWA-KPWEDZA atu a-na-taka ku-heka madzi, Early.morning there.PST-come people 3p.PROG-want INF-draw water a-ri-pho-tsungurira a-ona mutu. A-chi-lungb-wa alume, 3p.PST-REL-peep 3p.PST-see person 3p-NAR-fetch-PAS men hinyo alume A-K-EDZA, a-amb-wa, these men 3p-PF-come 3p.PST-tell-PAS “Sino hu-ka-ona mutu a-im-ire mo chisima-ni.” we 1p-PF-see person 3s-stand-PF inside well-LOC ‘Early in the morning there came people wanting to draw water. When they looked (into the well) they saw a person. The men-folk were called, and when they had come they were told, “We have seen someone standing in the well.”’
The same form, with the same basic meaning, also occurs in constructions in which it is preceded by a tense / aspect marker or infinitive prefix (therefore behaving like a main verb) but immediately precedes a different main verb (therefore behaving like a tense / aspect marker, most of which occur as
‘venitive’ is sometimes used for for ‘ventive’ (see Bourdin 1992 for a discussion).
60 Steve Nicolle verbal prefixes). In this construction, edza functions as a movement gram, as illustrated in example (22) from the same story as (21) above. The women again come to the well to draw water, but this time their arrival is treated as incidental and edza functions as an auxiliary preceding the main verb heka ‘draw (water)’ 13.
(22) Ligundzu achetu a-k-EDZA-heka madzi. morning women 3p-PF-COME-draw water ‘In the morning the women came and drew water.’
Digo also has a movement gram enda (with a variant kpwenda) meaning ‘go (to)’. Although enda does not occur as a main verb in Digo (the lexical verb ‘go’ is phiya), it is almost certain that it did in the past. There are two main reasons for this assumption : First, a verb enda is found in related languages with the same meaning. Second, (mono-syllabic) preverbal tense / aspect markers in Bantu languages can almost always be traced to (bi-syllabic) main verbs, and the future tense in Digo is nda- ; since movement verbs are typical sources of future tenses, it is highly probable that nda- derives from enda. From a semantic perspective, some uses of enda and edza as movement grams (as opposed to lexical verbs) are redundant, as they accompany other explicit indicators of physical movement and direction. This kind of redundancy is also characteristic of tense and aspect markers ; for example, a future tense can co-occur with the temporal adverbial tomorrow. In the following example enda is preceded by the verb phiya (‘go’) 14.
(23) atu osi kala a-ka-PHIYA mzuka-ni people all be.PST 3p-PF-go shrine-LOC kpw-ENDA-voya mikoma ili a-jali-w-e. INF-GO-pray ancestral.spirits so.that 3p-bless-PAS-SUB ‘all the people had gone to the shrine to (go and) pray to the ancestral spirits so as to be blessed.’
Although there is no change of meaning when enda and edza occur as movement grams rather than as main verbs, their function in discourse is to express a subjectified construal of the event described by the main verb, in a
13 To distinguish edza functioning as a movement gram and as a predicate, I have glossed it as ‘COME’ and ‘come’ respectively. 14 The following abbreviations are used : 1s = 1st person singular ; 3p = 3rd person rd plural ; 3s = 3 person singular ; CON = Conditional ; DIS = Distal marker ; INF = Infinitive ; LOC = Locative ; NAR = Narrative (sequential) tense ; PAS = Passive ; PF = Perfective aspect ; POS = Possessive ; PROG = Progressive ; PST = Past tense ; REL = Relative marker ; SUB = Subjunctive.
The grammaticalization of tense markers 61 similar way to go (and) get in English. In (23) the narrator did not go to the shrine, so the praying took place away from the narrator’s vantage point, which is the deictic centre. Emphasizing this through the use of enda has the added effect of reinforcing the narrator’s disassociation of himself from this event. In (24) below, the subject is described as having gone to his brother (using the main verb phiya) to beg for food. In the final clause he is described as eating the food at home with his family ; this is the deictic centre of the narrative, that is, the notional location of the narrator (or conceptualizer). The use of edza (‘come’) makes it clear that the subject has moved back to the deictic centre (where his family are located) rather than being joined by his family at his brother’s place or somewhere else. However, the most salient action is the eating rather than the moving which preceded it.
(24) Sambi yuya mchiya kala a-chi-phiya kpwa ndugu-ye Now that poor.man be.PST 3s-NAR-go to brother-3s.POS a-CHA-voya chakurya ; a-ka-he-wa na a-k-EDZA-rya 3s-DIS-beg food 3s-PF-give-PAS and 3s-PF-COME-eat na mche-we na ana-e. with wife-3s.POS and children-3s.POS ‘Now that poor man had gone to his brother and begged for food (there) ; he was given some and (came and) ate with his wife and children.’
3.2. Action at a distance from the deictic centre : cha
Example (24) also illustrates the use of the prefix cha which indicates that movement away from the deictic centre has occurred prior to the action described in the main verb. Unlike enda and edza, cha occurs in the tense / aspect ‘slot’ in the verbal complex (immediately prior to the verb or incorporated argument if there is one) and cannot itself be preceded by another tense marker. (Unlike some other Bantu languages, Digo only allows one tense or aspect prefix per verb.) It therefore behaves morphosyntactically like a typical tense / aspect marker. Like the narrative tense marker chi in the first line of (24), cha is temporally dependent, in that temporal reference must be established by a preceding finite verb. The past tense copular kala establishes the temporal reference for both the act of going (marked with chi) and the act of begging (marked with cha).
3.3. Summary
In this section I have briefly described three constructions in Digo which exhibit typical morphosyntactic characteristics of grams without having undergone corresponding semantic change. They give rise to a subjectified
62 Steve Nicolle construal of the event described by the associated verb phrase, in so far as the notional location of the conceptualizer (that is, the deictic centre of the narrative) is made explicit. In the case of enda and edza, movement away from or towards the deictic centre respectively is incorporated into the description of the event described, and in the case of cha, the event is viewed as occurring at a distance from the deictic centre. Movement grams differ from their lexical counterparts in that physical movement is a less salient aspect of the meaning of the clause as a whole than the action described by the main verb.
4. Conclusion
When a movement verb becomes a tense marker through the process known as grammaticalization it undergoes structural and semantic change. Structurally, it may become more closely linked to the main verb which it modifies, even losing its own inflections and becoming a verbal affix, and it may undergo phonological reduction. Semantically, meaning relating to the physical movement of an entity relative to the deictic centre develops into meaning relating to temporal relations between events and reference times. It has generally been assumed that structural and semantic changes necessarily co-occur during grammaticalization. In this paper, however, I have described constructions in English and Digo which derive from movement verbs and behave syntactically like tense markers in these languages, but which describe physical movement (or more rarely location) rather than temporal relations. This suggests that the structural changes that characterize grammaticalization need not result from (or be accompanied by) semantic change. This is not to say that the constructions I have described exhibit the same discourse-pragmatic functions as their lexical sources. All deictic movement verbs help anchor a situation with respect to the deictic centre (that is, they make the speaker’s perspective explicit), but in the constructions I have described, this secondary function has become primary. This is an example of subjectification, and it is subjectification rather than semantic change which underlies grammaticalization in these cases.
References
Baker, M. (1989). Object sharing and projection in serial verb constructions, Linguistic Inquiry 20 : 513-553. Bourdin, P. (1992). Constance et inconstances de la déicticité : la resémantisation des marqueurs andatifs et ventifs, in : M-A. Morel ; L. Danon-Boileau, (éds), La Deixis (Colloque en Sorbonne, 8-9 juin 1990), Paris : Presses Universitaires de France, 287-307.
The grammaticalization of tense markers 63
Bourdin, P. (2000). À propos des ‘futures’ ventifs et itifs : remarques sur un paradoxe, Verbum 22 : 293-311. Butler, J. (2004). Non-finite temporality and modality, Paper presented at Chronos 6, Geneva, 22-24 September 2004. Bybee, J. L. (2003). Cognitive processes in grammaticalization, in : M. Tomasello, (ed), The New Psychology of Language, Vol. II, New Jersey : Lawrence Erlbaum. Bybee, J. L. ; Pagliuca, W. ; Perkins, R. D. (1991). Back to the future, in : E. C. Traugott ; B. Heine, (eds), Approaches to grammaticalization (volume 2), Amsterdam : John Benjamins, 17-58. Bybee, J. L. ; Perkins, R. D. ; Pagliuca, W. (1994). The Evolution of Grammar : Tense, Aspect and Modality in the Languages of the World, Chicago : University of Chicago Press. Carden, G. ; Pesetsky, D. (1977). Double-verb constructions, markedness, and a fake co-ordination, Papers from the 13th Regional Meeting of the Chicago Linguistic Society, Chicago : Chicago Linguistic Society, 82- 92. Demirdache, H. ; Uribe-Etxebarria M. (2000). The primitives of temporal relations, in : R. Martin ; D. Michaels ; J. Uriagereka, (eds), Step by Step : Essays in Minimalist Syntax in honor of Howard Lasnik, Cambridge MA : MIT Press, 157-186. Dimmendaal, G. J. (1983). The Turkana Language, Dordrecht : Foris. Emanatian, M. (1992). Chagga ‘come’ and ‘go’ : metaphor and the development of tense-aspect, Studies in Language 16 : 1-33. Giorgi, A. ; Pianesi, F. (1997). Tense and aspect. From Semantics to Morphosyntax, Amsterdam : John Benjamins. Givón, T. (1991). Serial verbs and the mental reality of ‘event’ : grammatical versus cognitive packaging, in : E. C. Traugott ; B. Heine, (eds), Approaches to grammaticalization (volume 1), Amsterdam : John Benjamins, 81-127. Guthrie, M. (1967-71). Comparative Bantu. An introduction to the comparative linguistics and prehistory of the Bantu languages, Farnborough : Gregg International. Heine, B. (1992). Grammaticalization chains, Studies in Language 16 : 335- 368. Heine, B. ; Claudi, U. ; Hünnemeyer, F. (1991). Grammaticalization. A conceptual framework, Chicago : University of Chicago Press. Heine, B. ; Güldemann, T. ; Kilian-Hatz, C. ; Lessau, D. A. ; Roberg, H. ; Schladt, M. ; Stolz, T. (1993). Conceptual Shift. A lexicon of grammaticalization processes in African languages, Köln : Afrikanistische Arbeitspapiere, 34/35. Hopper, P. J. ; Traugott, E. C. (1993). Grammaticalization, Cambridge : Cambridge University Press.
64 Steve Nicolle
Jaeggli, O. A. ; Hyams, N. M. (1993). On the independence and inter- dependence of syntactic and morphological properties : English aspectual come and go, Natural Language and Linguistic Theory 11 : 313-346. Koch, P. (1999). Cognitive aspects of semantic change and polysemy : the semantic space HAVE/BE, in : A. Blank ; P. Koch, (eds), Historical Semantics and Cognition (Cognitive Linguistics Research 13), Berlin : Mouton de Gruyter, 279-305. Kuteva, T. (2001). Auxiliation : An Enquiry into the Nature of Grammaticalization, Oxford : Oxford University Press. Langacker, R. W. (1990). Subjectification, Cognitive Linguistics 1 : 5-38. Lightfoot, D. (1991). How to Set Parametres : Arguments from Language Change, Cambridge MA : MIT Press. Nicolle, S. (1998). A relevance theory perspective on grammaticalization, Cognitive Linguistics 9 : 1-35. Nicolle, S. (2002). The grammaticalization of movement verbs in Digo and English, Révue de Semantique et Pragmatique 11 : 47-68. Nicolle, S. (2004). Concise grammar of the Digo language, in : J. Mwalonya ; A. Nicolle ; S. Nicolle ; J. Zimbu, Mgombato : Digo–Swahili–English Dictionary, Nairobi : BTL, 204-214. Nurse, D. (1999). Towards a historical classification of East African Bantu languages, in : J-M. Hombert ; L. M. Hyman, (eds), Bantu Historical Linguistics. Theoretical and Empirical Perspectives, Stanford CA : CSLI, 1-41. Pullum, G. K. (1990). Constraints on intransitive quasi-serial verb constructions in modern colloquial English, in : B. D. Joseph ; A. M. Zwicky, (eds), When Verbs Collide : Papers from the Ohio State mini- conference on serial verbs (Working Papers in Linguistics), Ohio : Ohio State University, Department of Linguistics, 218-239. Radden, G. (1996). Motion metaphorized : the case of coming and going, in : E. H. Casad (ed), Cognitive Linguistics in the Redwoods : The expansion of a new paradigm in linguistics, Berlin : Mouton de Gruyter, 423-458. Roberts, I. (1993). Verbs and Diachronic Syntax, Kluwer : Dordrecht. Shopen, T. (1971). Caught in the act : an intermediate stage in a would-be historical process providing syntactic evidence for the psychological reality of paradigms, Papers from the 7th Regional Meeting of the Chicago Linguistic Society, Chicago : Chicago Linguistic Society, 254- 263. Stowell, T. (in press). Sequence of perfect, in: L. de Saussure ; J. Moeschler ; G. Puskas, (eds), Recent Advances in the Syntax and Semantics of Tense, Modality and Aspect, Belin - New York : Mouton De Gruyter.
The grammaticalization of tense markers 65
Talmy, L. (1985). Lexicalization patterns : semantic structure in lexical forms, in : T. Shopen (ed), Language Typology and Syntactic Description (volume 3), Cambridge : Cambridge University Press. Tabor, W. ; Traugott, E. C. (1998). Structural scope expansion and grammati- calization, in : A. G. Ramat ; P. J. Hopper, (eds), The Limits of Grammaticalization, Amsterdam : John Benjamins, 229-272. Thornell, C. (1997). The Sango Language and its Lexicon, (Travaux de l’Institut de Linguistique de Lund 32), Lund : Lund University Press. Traugott, E. C. (1978). On the expression of spatio-temporal relations in language, in : J. H. Greenberg (ed), Universals of Human Language. Vol. 3, Word structure, Stanford : Stanford University Press, 369-400. Traugott, E. C. (2002). From etymology to historical pragmatics, in : D. Minkova ; R. Stockwell, (eds), Studying the History of the English Language : Millennial perspectives, Berlin : Mouton de Gruyter. Ultan, R. (1978). The nature of future tenses, in : J. H. Greenberg (ed), Universals of human language. Vol. 3, Word structure, Stanford : Stanford University Press, 83-123. Van Valin, R. D. Jr. ; La Polla, R. J. (1997). Syntax : Structure, Meaning and Function, Cambridge : Cambridge University Press. Visser, F. Th. (1969). An Historical Syntax of the English Language. Part three, first half : Syntactical units with two verbs, Leiden : E. J. Brill. Warner, A. (1993). English Auxiliaries : Structure and History, Cambridge : Cambridge University Press. Zwicky, A. M. (1969). Phonological constraints in syntactic descriptions, Papers in Linguistics 1 : 411-463.
This page intentionally left blank
Aspectual interactions between predicates and their external arguments in French
Maria ASNES Bar Ilan University
0. Introduction : Aspectual and non-aspectual external arguments
Numerous works in the area of aspectual composition (Krifka 1989, 1998, Verkuyl 1972, 1993, Tenny 1994, etc) have treated at length the interactions between the verbal predicates and their internal arguments. However, as far as external arguments are concerned, their treatment is rather deficient (cf. van Voorst 1988, Dowty 1991, Jackendoff 1996, Ramchand 1997, Verkuyl 2000) in the current literature on aspect. It is generally assumed that external arguments as opposed to the internal ones do not interact aspectually with their predicates (The Non-Measuring Constraint on External Arguments by Tenny 1994 : 83). It has been often noticed (cf. Tenny 1994) that external arguments do not participate in the aspectual composition, since unlike internal arguments they are not involved in the path construction, but are rather associated with it. Once the path is constructed, the external argument is projected to it by means of a participation function (cf. Verkuyl 2000). This description matches well the hypothesis of asymmetry : while internal arguments are directly involved in the path construction, the external arguments are related to the entire path constructed by the verbal predicate and its internal arguments. Two types of external arguments are sometimes distinguished (cf. van 1 Voorst 1988, Ramchand 1997) : causative or agentive arguments (Θext) marked by the feature [+involved] (van Voorst 1988) that participate in the entire process and stative arguments (Locatum) that are told to possess a property described by the predicate and are marked by the feature [-involved].
(1) Stative argument : Paul se trouve à l’étranger. ‘Paul is abroad.’ (2) Agentive (causative) argument : Paul a construit une maison. ‘Paul built a house.’
1 Ramchand (1997) and van Voorst (1988) talk about subjects but it is clear that they identify the notion of subject with that of external argument.
© Cahiers Chronos 17 (2007) : 67-80. 68 Maria Asnes
According to this approach, only causative or agentive arguments that participate in the events, namely Θext, are aspectual and combine on the syntactic level with AspP (aspect projection). In case of stative external arguments, the AspP projection is absent from the syntactic configuration. But are all the ‘aspectual’ arguments of Ramchand (1997) capable of having an aspectual effect ? Is it enough to have a thematic role of Cause or Agent in order to participate in the aspectual composition ? It will be shown here that we need to apply an additional set of semantic distinctions in order to determine if an external argument is likely to have an impact on the VP aspect. As opposed to the approaches that denied the aspectual force of external arguments, it will be claimed here that these arguments do interact with their predicates, though in a manner which is different from that of the internal ones. This might happen, however, only if one important condition is met : there should be an iterative homomorphism between the external argument and the complex predicate. It will be demonstrated that the unique event homomorphism (Krifka 1992) cannot trigger that kind of interactions.
1. External arguments, event- object homomorphism and participation in the path
First, we will examine if the external arguments can maintain the same type of relations with the complex predicate as the internal arguments do with the verbal predicate. More precisely, it will be checked if the external arguments may function as incremental themes and stand in a homomorphic relation with the denotation of the complex predicate. Let’s examine the following example inspired by Dowty (1991) and Jackendoff (1996) who where the first to notice that the internal arguments are not the only ones that can function as incremental themes :
(3) L’eau a rempli la cave. ‘Water filled the cellar.’
In (3), there is a homomorphism between the sub-intervals of the process remplir (fill), the parts of the denotation of the internal argument la cave and the parts of the denotation of the external argument l’eau (water). The external argument as well as the internal one is an incremental theme of the process denoted by the verbal predicate. We can observe here a localizing relation between the external and internal arguments such as at each instant i a part of the denotation of the external argument is localized in some position of the denotation of the internal argument. Krifka (1989, 1992) defined the concept of homomorphism using three sense postulates :
Aspectual interactions between predicates and their external arguments 69
(4) Unicity of objects ∀R[UNI-O (R) ↔ ∀e, y, y’ [R(e,y) ∧R (e, y’) → y = y’]] (5) Mapping on objects ∀R[MAP-O (R) ↔ ∀e, e’, y [R(e,y) ∧ e’⊆E e → ∃y’ [y’⊆O y ∧ R(e’, y’)]] (6) Mapping on events ∀R[MAP-E (R) ↔ ∀e, y, y’ [R(e,y) ∧ y’⊆O y → ∃e’ [e’⊆E e ∧ R(e’, y’)]] Since we saw that some external arguments can also stand in a homomorphic relation with events, we suggest here to extend Krifka’s definition of homomorphism to the domain of external arguments. Let x be an external argument, and P the thematic relation between the verbal predicate and this external argument :
(7) Unicity of objects ∀P[UNI-O (P) ↔ ∀e, x, x’ [P(e,x) ∧ P (e, x’) → x = x’]]
A thematic relation P between an event e and an object x satisfies the unicity of objects property if there is no other object x’ that stands in the same thematic relation P with e.
(8) Mapping on objects ∀P[MAP-O (P) ↔ ∀e, e’, x [P(e,x) ∧ e’⊆E e → ∃x’ [x’⊆O x ∧ P(e’, x’)]] Every part e’ of an event e corresponds to a part x’ of the object x.
(9) Mapping on events ∀P [MAP-E (P) ↔ ∀e, x, x’ [P(e,x) ∧ x’⊆O x → ∃e’ [e’⊆E e ∧ P(e’, x’)]] Every part x’ of the object x corresponds to a part e’ of the event e.
Jackendoff 1996 noticed that if the subject is incremental, the object is also necessarily so (this is the case in (1)). We believe that it is possible to further elaborate Krifka’s postulates in order to account for the incrementality of external and internal arguments. In constructions such as (1) :
(10) Mapping on objects ∀e, e’, x , y , P, R [P (e,x) ∧ R (e, y) ∧ e’⊆E e → ∃x’, y’ [x’⊆O x ∧ y’⊆O y ∧ P(e’, x’) ∧ R (e’, y’)]]
Every part e’ of an event e corresponds to a part x’ of the object x (referent of the external argument) and to a part y’ of the object y (referent of the internal argument).
70 Maria Asnes
(11) Mapping on events ∀e, x, x’ y, y’ [P (e,x) ∧ R (e, y) ∧ x’⊆O x ∧ y’⊆O y → ∃e’ [e’⊆E e ∧ P (e’, x’), ∧ R (e’, y’)]]
Every part x’ of the object x (referent of the external argument) and every part y’ of the object y (referent of the internal argument) correspond to a part e’ of the event e.
We know that incrementality of internal arguments means that they can modify the event aspect. But is it so as far as incremental external arguments are concerned ? Let’s examine the following paradigm :
(12) a. L’eau / une grande quantité d’eau a rempli le réservoir pendant / en une heure. ‘Water / a large quantity of water filled the reservoir during / in one hour.’ b. De l’eau a rempli le réservoir pendant / en une heure. ‘Some water filled the reservoir during / in one hour.’ (13) a. La fumée a envahi la chambre ? ?pendant / en quelques instants. ‘Smoke invaded the room during / in several instants.’ b. De la fumée a envahi la chambre *pendant / en quelques instants ‘Some smoke invaded the room during / in several instants.’
In example (12) the predicate remplir (‘fill’) can be interpreted as an accomplishment or as a state. The atelic state reading is compatible with the durative adverbials whereas the telic accomplishment reading is compatible with time-span adverbials. However, we see the homogeneous (a) or heterogeneous (b) reading of the external argument does not alter the VP aspect. Similarly, the VP in example (13) with a heterogeneous verbal head envahir (invade), is heterogeneous invariably and independently of the nature of the external argument. It has to be noticed that the external arguments in (12) and (13) are mass nouns. It seems thus that mass nouns in the external argument position cannot modify the VP aspect in spite of them being incremental themes. But let’s examine the following example :
(14) a. Paul a traversé la rue en / *pendant cinq minutes. ‘Paul crossed the street in / during five minutes.’ b. Des piétons ont traversé la rue en / pendant cinq minutes. ‘Pedestrians crossed the street in / during five minutes.’
The external arguments of the telic predicate traverser (‘cross’) are incremental themes. In (14a) where the external argument is a count noun (heterogeneous), the entire VP is heterogeneous (telic). At this stage, we do not take into consideration the possibility of the iterative reading which
Aspectual interactions between predicates and their external arguments 71 would be compatible with durative adverbials. We can observe, however, that with a plural indefinite in the position of external argument, like in (14b), the VP can be homogeneous in case of a distributive interpretation. This is a type of iterative reading which is different from that that we disregarded for (14a). Generally speaking, we distinguish two kinds of iterativity. In (14a) the entire process [Paul traverser la rue] can be repeated. Iterativity functions here as an external operator with a proposition in its scope :
(15) ITER [Traverser (Paul, la rue)], where ITER is an iterativity operator
As opposed to (14a), in (14b), it is not the entire process but its sub-parts that are iterated. More precisely, every part of the denotation of the external argument des piétons (‘walkers’) corresponds to a path denoted by traverser la rue (‘cross the street’). This is an iterativity triggered by the distributive reading of the external argument and it is thus internal to the process. It could be represented as :
(16) ∃X (Piéton (X)→∀X’ : X’⊆ X [Traverser (X’, la rue)], where X is a group variable (a plural individual).
If we return to the example (14b), we can state that the process denoted by the VP is homogeneous under iterativity effect. The iterativity creates a sort of homomorphism between the denotations of the external arguments and of the verbal predicate. This type of homomorphism is different from the unique-event type triggered by the incremental theme relation (the one that exists between the external or internal mass arguments and the verbal predicate). It appears thus that the homomorphism caused by iterative reading is the only one that can give rise to aspectual interactions between the external arguments and the verbal predicates. A similar hypothesis was suggested by Tonne 1997 for the Norwegian data. In what will follow, we will present a corpus of data that will allow us to draw conclusions about the way in which external arguments participate in aspectual composition.
2. Corpus construction
There are three kinds of constructions in which the aspectual behaviour of external arguments can be examined : intransitive, transitive and ditransitive ones. In this paper, we will limit ourselves to ditransitive constructions which are the more complex ones and where the interactions between the verb and its external and both internal arguments can be examined
72 Maria Asnes
2.1 Feature configurations
We will introduce a binary referential feature [±hom] 2, which can mark both nominal and verbal constituents. This feature takes its origin from the criteria of cumulative (Quine 1960) and distributive (Cheng 1973) reference which describe the mereological properties of mass and count nouns. Later, these notions were unified in a more general principle of homogeneous reference (Bunt 1976, ter Meulen 1981) which applies to mass nouns and according to which all parts of a quantity x that are themselves quantities of x, can become parts of another quantity x.
(17) (∀ x ) (∀ y) [ x ∈ y → (P(x) ↔ P(y))], where P is a mass noun (ter Meulen 1981, 123)
Count nouns are then said to possess the property of non-homogeneous, or heterogeneous reference since they do not satisfy the criteria of cumulative and distributive reference. Independently from these analyses, Vendler 1997 showed that in the verbal domain, activities and states are homogeneous predicates whose subintervals satisfy the criteria of cumulative and distributive reading. A great number of works has been dedicated to the referential parallels between nouns and verbs, saying that their referents can have a similar mereological structure (Mourelatos 1978, Bach 1986, Borillo 1988a, 1988b, Franckel et alii 1988, etc). Thus mass nouns and durative, atelic predicates refer homogeneously whereas count nouns and telic predicates refer heterogeneously. Our analysis is a further development of this general idea according to which homogeneous / heterogeneous reference applies both to nominal and verbal constituents and thus, provides a common ground for the treatment of aspectual interactions. Thus, in ditransitive constructions we can have the following combination of features (notice that the verbs are normally telic) 3 :
(18) I. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] II. DP[-hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] III. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[+hom] IV. DP[-hom] + V[-hom] + DP [+hom]+ Prep + DP[+hom]
2 [+hom] = homogeneous reference, [-hom] = heterogeneous reference 3 In this paper, we will treat the ditransitive constructions formed by dative predicates which are normally telic since they select an internal argument with the thematic-role of Terminus bounding the process. Therefore, these predicates are telic and hence heterogeneous.
Aspectual interactions between predicates and their external arguments 73
V. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] VI. DP[+hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] VII. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[+hom] VIII. DP [+hom]+ V[-hom] + DP[+hom] + Prep + DP[+hom]
We will be particularly interested in feature configurations where an external argument can trigger an iterative homomorphism and thus participate in aspectual composition.
2.2. Corpus
The examples presented here will contain dative constructions. The same treatment can be also applied to locative constructions which we will not treat here. Ditransitive constructions present a special interest since the external argument (EA) as well as the internal dative argument (DA) participate at the same time in the process (in the sense of Verkuyl 2000), whereas generally the external argument is the only one that is thought to be assigned a participation function. From a semantic point of view, there exists a transition or exchange relation where the external and dative arguments are both participants. According to Tenny (1994) and Verkuyl (2000), internal direct argument is the only one among other arguments that constitutes a path. It follows from Verkuyl’s (2000) representation, according to which a single path is a set of pairs
4 (19) π(y) = Term (lx)
4 Term is a θ-role of dative arguments which mark the terminal point of the process.
74 Maria Asnes where π is a participation function, y is dative argument, Term is a θ-role of dative arguments which mark the final point of a process, l is a path and x is an external argument. The participation of a dative argument y is such that it functions as a term of a path associated with external argument x. As such, it is a passive participant. Let’s examine path constructions in the examples of (20) :
(20) a. Paul a envoyé une lettre à Max. ‘Paul sent a letter to Max.’ b. Trois filles ont envoyé une lettre à Max. ‘Three girls sent a letter to Max.’ c. Paul a envoyé une lettre à trois filles. ‘Paul sent a letter to three girls.’
In (20a), if we exclude the case of iterative interpretation, there is one single path which is associated by means of participation function with the referent of the external argument Paul and the referent of the dative argument Max :
(20’a) EA : π (Paul) = lPaul ; DA : π(Max) = Term (lPaul) In (20b) the denotation of the external argument contains three individuals. This sentence can be interpreted collectively or distributively, and it is the denotation of the external argument that will determine the number of paths. In case of collective interpretation, the referent of the DP trois filles (‘three girls’) participates as one group in one single path. On the contrary, in case of distributive interpretation, each element of the denotation of the external argument participates separately from other elements in its path :
(20’b) COLLECTIVE READING : EA : π (trois filles) = ltrois_filles where ltrois_filles = lfille1 + lfille2 + lfille3 ∩ lfille1= lfille2= lfille3 The participation of the EA trois filles is such that it constructs collectively one single path which in its turn is a sum of three overlapping paths.
DA : π(Max) = Term (ltrois filles) The participation of DA Max is such that it is a term of the path constructed by trois filles.
DISTRIBUTIVE READING EA: π (trois filles) = lfille1 + lfille2 + lfille3 ∩ lfille1≠ lfille2 ≠lfille3
Aspectual interactions between predicates and their external arguments 75
The participation of the EA trois filles is such that it constructs three different paths, each corresponding to a different participant.
DA : π(Max) = Term (ltrois filles) The DA is a single term of these three paths. A similar situation can be observed in (20c) where this time it’s the dative argument trois filles whose collective or distributive interpretation determines the number of paths. If this argument is interpreted collectively, there will be one path, whereas if it is interpreted distributively, the number of paths will be equal to the number of individuals that compose the denotation of the dative argument. In other terms, in our case you need to follow three paths in order to attain three terms.
(20’c) COLLECTIVE READING : EA : π (Paul) = lPaul = l1Paul + l2Paul + l3Paul where l1Paul = l2Paul = l3Paul The EA Paul participates in one path.
DA : π(trois filles) = Term (lPaul) = Term1(l1Paul) + Term2 (l2Paul) + Term3 (l3Paul) where Term1(l1Paul) = fille1 , Term2 (l2Paul)= fille2 et Term3(l3Paul) = fille3 ∩ Term1(l1Paul) = Term2 (l2Paul)= Term3(l3Paul) The DA trois filles forms one single term of a path performed by Paul.
D ISTRIBUTIVE READING EA : π (Paul) = lPaul = l1Paul + l2Paul + l3Paul où l1Paul ≠ l2Paul ≠ l3Paul Paul participates in three different paths.
DA : π (trois filles) = Term1(l1Paul) + Term2 (l2Paul) + Term3 (l3Paul) où Term1(l1Paul) = fille1 , Term2 (l2Paul) = fille2 et Term3 (l3Paul) = fille3 ∩ Term1(l1Paul) ≠ Term2 (l2Paul)≠ Term3(l3Paul) Three girls function as three different terms, each corresponding to a different path performed by Paul. After we have outlined some functional analogies between the external and dative arguments, we will now analyze the empirical data of our corpus.
I. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] (21) Paul a envoyé 1000 Euros à Max *pendant / en 5 minutes. ‘Paul sent 1000 Euros to Max during / in 5 minutes.’
All the elements of this configuration have a heterogeneous reference and consequently the resulting VP will also be heterogeneous (telic). Thus, the
76 Maria Asnes sentence in (21) is compatible with time-span adverbials formed by en (‘in’) and does not accept the durative adverbials introduced by pendant (‘during’).
II. DP[-hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] (22) Paul a envoyé de l’argent à Max pendant 2 ans / en 5 minutes. ‘Paul sent (some) money to Max during 2 years / in 5 minutes.’
When only the direct internal argument is homogeneous and all the other constituents are heterogeneous, the sentence can have either homogeneous or heterogeneous interpretation. But, the only possible kind of homogeneity is an iterative one where the whole process is iterated, while the participants remain the same. Thus, Paul a envoyé de l’argent à Max pendant 2 ans means that Paul regularly sent money to Max during a period of 2 years. The heterogeneous interpretation is that of an achievement. We have here one path with an external argument Paul as a participant and one term Max.
III. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[+hom] (23) Paul a envoyé 1000 Euros à des proches pendant 2 ans / en 5 minutes. ‘Paul sent 1000 Euros to relatives during 2 years / 5 minutes.’
Let’s see what happens when the dative argument is the only one that is marked by a [+hom] feature. As we have seen earlier, an indirect argument, just like an external one, can affect the number of paths. The sentence in (23) is ambiguous between homogeneous and heterogeneous readings. This ambiguity can be explained in terms of difference between a collective and a distributive reading of the indirect argument. If this argument is interpreted collectively, the sentence will have a heterogeneous interpretation since the path will have a well-delimited precise term. The collective interpretation corresponds to a case where Paul sent a total of 1000 Euro to relatives, namely his relatives received altogether a sum of 1000 Euro. The DP des proches (‘relatives’) functions here as a plural individual (a group). On the contrary, if an indirect argument has a distributive reading, the VP will have a homogeneous reference. In (23) des proches (‘relatives’) is interpreted no longer as a plural individual but rather as a sum of individuals. The sentence can mean that Paul sent several times 1000 Euro, and each time the money was destined to a different relative.
IV. DP[-hom] + V[-hom] + DP [+hom]+ Prep + DP[+hom] (24) Paul a envoyé de l’argent à des proches pendant 2 ans / en 10 minutes. ‘Paul sent (some) money to relatives during 2 years / in 10 minutes.’
In this configuration the two internal arguments, direct and indirect, are homogeneous whereas the external argument and the verb are heterogeneous.
Aspectual interactions between predicates and their external arguments 77
The aspect of this configuration is calculated on the basis of the same criteria that we suggested for configurations II and III. The homogeneous reading corresponds either to a case where the indirect argument has a distributive reading or to the case where it is interpreted collectively but the direct internal argument contributes to an iterative reading. In other terms, the VP in (24) will be homogeneous if Paul sent money each time to a different relative or relatives (distributive reading of des proches) or if Paul sent money several times to the same members of his family. We can also have a heterogeneous interpretation for (24) if the indirect argument is interpreted collectively and if an iterative interpretation is excluded (unique event interpretation).
V. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] (25) Des touristes ont envoyé une carte postale à leur voyagiste pendant / en une heure. ‘Tourists sent a postcard to their travel agent during / in one hour.’
The sentence in the configuration V where the external argument is marked by a [+hom] feature can be interpreted homogeneously or heterogeneously as a result of a distributive or collective reading of the external argument. The VP will be heterogeneous if the external argument is interpreted collectively. Thus, if the tourists sent one card for all to their travel agent, the sentence will be interpreted as an achievement (heterogeneously). However, if the external argument is interpreted distributively, it can trigger an iterative reading of the sentence and thus impose a homogeneous interpretation. This is the case in (25) where each of the tourists or groups of tourists sent a different postal card to their travel agent.
VI. DP[+hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] (26) Des touristes ont envoyé des cartes postales à leur voyagiste pendant / *en 1 mois. ‘Tourists sent postcards to their travel agent during / in 1 month.’
When external and dative arguments are both homogeneous, the entire VP becomes also homogeneous. It appears that if only the external argument is homogeneous like in configuration V, the heterogeneous reading is still possible, but if at least one other element is also homogeneous, the entire VP is necessarily so. Even if the external argument in (26) is interpreted collectively, it’s the direct argument that introduces the effect of iterativity.
VII. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[+hom]
78 Maria Asnes
(27) Des touristes ont envoyé une carte postale à des proches pendant / *en 1 mois. ‘Tourists sent a postcard to relatives during / in 1 month.’
The VP becomes homogeneous if the external and the indirect arguments are homogeneous. The two poles of the path, the source and the destination, are denoted here by homogeneous DPs (plural indefinites). In this particular case and in general, it would be difficult to imagine a situation where the sentence of this configuration will be interpreted as marking one path (but a theoretical possibility exists). We will rather have an iteration of a path triggered either by an external argument or a dative one.
VIII. DP [+hom]+ V[-hom] + DP[+hom] + Prep + DP[+hom] (28) Des touristes ont envoyé des cartes postales à des proches pendant / *en 1 mois. ‘Tourists sent postcards to relatives during / in 1 month.’
In configuration VIII the verbal predicate is the only heterogeneous constituent. In this case the VP will inherit the feature of the arguments. The most plausible interpretation is that of an iterated event.
3. Conclusion
As opposed to analyses like the one defended by Tenny (1994), according to which the external arguments do not participate in the VP aspect construction, we have claimed that certain types of external arguments are capable of altering the VP aspect. These interactions are possible if there is a homomorphism of the iterative type between the denotation of the external argument and that of the complex predicate (composed of a verb only or of a verb with its internal arguments). We examined the behaviour of the external arguments in the ditransitive constructions and we arrived at the following conclusions concerning the aspect feature of VP :
I. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] =VP[-hom]
II. DP[-hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] = VP[± hom]
III. DP[-hom] + V[-hom] + DP[-hom] + Prep + DP[+hom] = VP[±hom]
IV. DP[-hom] + V[-hom] + DP [+hom]+ Prep + DP[+hom] = VP[± hom]
V. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[-hom] = VP[± hom]
VI. DP[+hom] + V[-hom] + DP[+hom] + Prep + DP[-hom] = VP[+hom]
Aspectual interactions between predicates and their external arguments 79
VII. DP[+hom] + V[-hom] + DP[-hom] + Prep + DP[+hom] =VP[+hom]
VIII. DP [+hom]+ V[-hom] + DP[+hom] + Prep + DP[+hom] = VP[+hom] Several important generalizations can be formulated on the basis of our analysis of aspectual interactions with the external argument. they exclude a distributive interpretation (we didn’t say a word about the distributive heterogeneous DPs like les étudiants, but it seems that they cannot induce a homogeneous reading of a VP since the contextually delimited number of participants can participate in a bounded number of paths, hence the heterogeneity of a VP). 2. If a homogeneous external argument has a distributive reading, the entire VP will be homogeneous. 3. An external argument interpreted collectively functions as a heterogeneous DP.
References
Bach, E. (1986). The algebra of Events, Linguistics and Philosophy 9 : 5-16. Bunt, H.C. (1985). Mass Terms and Model-Theoretic Semantics, Cambridge : Cambridge UP. Borillo, A. (1988a). Notions de ‘masse’ et de ‘comptable’ dans la mesure temporelle, in : J. David ; G. Kleiber, (eds), Termes massifs et termes comptables, Paris : Klincksieck, 215-234. Borillo, A. (1988b). L’expression de la durée : construction des noms et des verbes de mesure temporelle, Lingvisticae Investigationes 12(2) : 363- 396. Cheng, C. (1973). Response to Moravcsik, in : J. Hintikka et al., (eds), Approaches to Natural Language, Dordrecht : Reidel. Dowty, D. (1991). Thematic Proto-Roles and Argument Selection, Language 67(3) : 547-619. Franckel, J-J. ; Paillard, D. ; de Voguë, S. (1988). Extension de la distinction discret, dense, compact au domaine verbal, in : J. David ; G. Kleiber, (eds), Termes massifs et termes comptables, Paris, Klincksieck, 239- 247. Grimshaw, J. (1990). Argument Structure. Cambridge (Mass.) : MIT Press. Jackendoff, R. (1996). The Proper Treatment of Measuring Out, Telicity and Perhaps Event Quantification in English, Natural Language and Linguistic Theory 14 : 305-354.
80 Maria Asnes
Krifka, M. (1989). Nominal Reference, Temporal Constitution, and Quantification in Event Semantics, in : R. Bartsch et al., (eds), Semantics and Contextual Expression, Dordrecht : Providence, Foris Publications, 75-115. Krifka, M. (1992). Thematic Relations as Links between Nominal Reference and Temporal Constitution, in : I. Sag ; A. Szabolsci, (eds), Lexical Matters. Stanford : CSLI. Krifka, M. (1998). The Origins of Telicity, in : S. Rothstein, (ed), Events and Grammar, Dordrecht : Kluwer, 197-236. Meulen, A. ter (1980). Substances, Quantifiers and Individuals, Oxford : Oxford University Press. Mourelatos, A. (1978). “Events, Process and Status”, Linguistics and Philosophy 2(3) : 415-434. Ramchand, G. C. (1997). Aspect and Predication : The Semantics of Argument Structure, Oxford : Clarendon Press/Oxford University Press. Tenny, C. (1994). Aspectual Roles and the Syntax-Semantics Interface, Dordrecht-Boston-London : Kluwer Academic Publishers. Tonne, I. (1997). “Homomorphism through (Partly) Iterative Events”, Paper presented at ESSLLI 1997. Quine, W.V. (1960). Word and Object, Cambridge : M.I.T. Press. Vendler, Z. (1967). Linguistics in Philosophy, Ithaca, NewYork, Cornell UP. Verkuyl, H. (1972). On the Compositional Nature of the Aspects, Dordrecht : Reidel. Verkuyl, H. (1993). A Theory of Aspectuality : The Interaction between Temporal and Atemporal Structure, Cambridge : Cambridge UP. Verkuyl, H. (2000). Events as Dividuals : Aspectual Composition and Event Semantics, in: J. Higginbotham ; F. Pianesi ; A. C. Varzi, (eds), Speaking of events, New York-Oxford : Oxford University Press, 169- 205. Voorst, J. van (1988). Event Structure, Amsterdam/Philadelphia : J. Benjamins Publishing Company.
Alors as a possible temporal connective in discourse
Anne LE DRAOULEC Myriam BRAS ERSS, UMR 5610, CNRS and Université de Toulouse - Le Mirail
1. Introduction
This work is part of a larger study on temporal connectives. It seeks to establish the status of the French temporal adverbial alors by answering the question of whether it can be considered a temporal connective, and under which conditions. We consider a temporal connective any adverb or adverbial phrase that expresses a temporal (or aspectuo-temporal) relation between two eventualities and, in addition, implies a logico-pragmatic relation between the two utterances in which the eventualities are described. In other words, temporal connectives are adverbs that play a role at the discourse level in introducing discourse relations. This definition has already been applied to puis (‘then’, ‘afterwards’), as opposed to the adverbial un peu plus tard (‘a little later’) or deux heures plus tard (‘two hours later’), in the formal framework of Segmented Discourse Representation Theory (SDRT, Asher 1993 ; Asher & Lascarides 2003). In Bras et alii (2001, 2003) and Borillo et alii (2004), we showed that only puis plays a role at the level of discourse relations. The same definition was used in Le Draoulec (2005) to reveal that the aspectuo-temporal adverbs aussitôt (‘immediately’, ‘straight away’) and soudain (‘suddenly’, ‘all of a sudden’) can play the role of discourse connective. Much work has been done on alors or its (approximate) English equivalent then (Franckel 1987, Gerecht 1987, Glasbey 1993, Hybertie 1996, Reyle 1998, Gosselin in press, and others). The work of Hybertie (1996) will be our starting point, since it considerably helped in the development of our own analysis. Hybertie distinguishes temporal uses of alors (with or without a consequential value), merely consequential uses (close to donc ‘therefore’), and other uses where alors is a kind of ‘structuration’ marker1. We will focus our attention on the temporal uses of alors, possibly associated with a consequential value – but this consequential value will be of interest to us only inasmuch as it involves the temporal value.
1 These other uses are in particular associated with intonation markers as in as alors, ça s’est bien passé ? (“so, how did it go ?”) or alors ça, c’est incroyable ! (“now, that is incredible !”), etc.
© Cahiers Chronos 17 (2007) : 81-94. 82 Anne Le Draoulec & Myriam Bras
Among the parametres that will be taken into account concerning the status of alors as a possible Temporal Connective, its position in the sentence (initial vs. internal and final) proves to be crucial. In particular the importance of the initial (or at least preverbal) position will be put to the test, as it is generally assumed to favour the connective function. For puis the position parametre was not relevant because this adverb has gone through a grammaticalization process from old and middle French to modern French that ended in a fixed position at the beginning of the sentence (cf. Hansen 1995 : 33). The position parametre should have been examined for aussitôt and soudain, but it was left aside in Le Draoulec (2005) where the only studied position was the initial one ; however Le Draoulec (2005) suggested that these adverbials function differently when they are in internal or final position. A second parametre, the aspectuo-temporal one, will be partially considered here. We will mainly examine relations between utterances describing eventualities with a perfective aspect, namely event descriptions, leaving out state descriptions associated with imperfective aspect sentences 2. We will first introduce the two analyses we started from : the one by Creissels, as concerns the position alors (in §2), and the one by Hybertie (§3.1). We will then discuss Hybertie’s analysis (§3.2) in order to propose new hypotheses to account for the role of alors (§4). We will conclude on the connective status of alors (§5).
2. Syntactic status and position
Our first starting point is an analysis of Creissels (1995), who clearly formulates the problem of the relationship between adverbs and connectives. He notes that (we translate below) :
“[...] un nombre important d’‘adverbes’ ont en fait un statut syntaxique qui n'est pas différent de celui [des] ‘conjonctions de coordination’ : comme elles, ils impliquent une relation entre la structure phrastique où ils figurent et une autre structure phrastique [...].” (Creissels 1995 : 151) [...] a great number of ‘adverbs’ have a syntactic status which is in fact not different from that of ‘coordinating conjunctions’ : like these they imply a relationship between the sentence structure where they appear and another sentence structure [...] (translation ours).
2 The ontological opposition between events and states stems from logical and representational theories of discourse semantics such a Discourse Representation Theory (Kamp and Reyle 1993), where sentences with perfective aspect contribute an event discourse referent to the Discourse Representation Structure, whereas sentences with an imperfective aspect contribute a state referent.
Alors as a possible temporal connective in discourse 83
Creissels (1995 : 151-153) takes the example of the adverb aussi (‘also’, ‘too’) and notes that, although it is similar to coordinating conjunctions by its ability to imply a relationship between two sentence structures 3, it differs from them by its ability to occupy various positions in the sentence that it links with another sentence. Among these various positions, there is the initial position, and, according to Creissels, it is difficult, in that case, not to acknowledge that aussi has the status of a connective. But what about its status when it occupies other positions ? Creissels identifies the difficulty in this way (we translate below) :
“L’exemple de aussi est révélateur d'une lacune importante de la grammaire traditionnelle en ce qui concerne les unités dont la fonction est de mettre en relation deux unités phrastiques : parmi ces unités, seules sont correctement identifiées celles qui se placent invariablement à la périphérie de l'unité phrastique. Mais il est vrai qu’à partir du moment où un connecteur peut se placer en différents points de la structure phrastique qu’il contribue à insérer dans son contexte, la distinction avec des unités intégrées à la structure prédicat-arguments n’est pas toujours évidente.” (p.153) The example of aussi reveals an important deficiency of traditional grammar concerning units whose function is to relate two sentence units : among these units, only those appearing invariably at the periphery of the sentence unit are properly identified. But it is true that, as soon as a connective can take up different positions in the sentence structure which it contributes to integrate in its context, its distinction from units integrated to the argument-predicate structure is not always easy (translation ours).
The units lying in this deficient part of grammar are precisely the ones we want to study. For temporal adverbs which are not placed in front of the verb, the question of distinguishing between modifiers (i.e. integrated units) and connectives (i.e. non integrated units) arises crucially : the distinction is far from easy when the temporal connection with what precedes coincides with the temporal location of the eventuality described by the sentence in which the adverb appears. The goal of these preliminary remarks was to place the problem in its syntactic context. In this study however, we will not take a syntactic perspective, since we are going to examine discursive aspects, in terms of discourse relations, paying special attention to the position of alors. Nevertheless, it is important to emphasize that this kind of problem arises as soon as one tries to apply syntactic criterion to elements that go beyond the sentence scope.
3 Hansen (1995) established the same connection about puis.
84 Anne Le Draoulec & Myriam Bras
3. Hybertie’s analysis
Our second and main starting point is the study of Hybertie (1996), which we are going to present briefly.
3.1. Presentation
Hybertie studies the configuration S1 alors S2 in which alors appears somewhere in S2, S2 is separated from S1 by a comma or a full stop, and alors has a temporal value. She distinguishes two cases. The first one, where a temporal referent is introduced by a locating adverbial in S1, is illustrated in (1) :
(1) J’ai rencontré Pierre EN 1987. J’étais alors une jeune étudiante. ‘I met Pierre in 1987. I was a young student THEN / AT THAT TIME.’
In (1), alors is anaphoric, referring to the temporal referent introduced in S1 by in 1987. According to Hybertie, alors gives the temporal reference of S2, via this anaphoric link. In that case, alors has a temporal value of strict concomitance, which is its first value from a historical point of view, as Hybertie points out. It can be paraphrased by à cette époque-là, à ce moment- là (‘at that time’). The second case, where S1 does not introduce any temporal referent, is illustrated by the following example of Hybertie (p.24) :
(2) Je suis allée jusqu’à la place du village, ALORS je l’ai vu arriver. ‘I walked up to the village square, THEN I saw him arrive.’
In (2), it is still possible to paraphrase alors with à ce moment-là, but alors no longer indicates a strict temporal concomitance or coincidence as we saw for (1). Instead, it ‘drifts’ towards the expression of a temporal succession, although it is not a simple succession. Because according to Hybertie (p.25) :
“[...] alors construit une séquence d'événements temporellement ordonnés. Il indique que les états de choses exprimés respectivement dans P1 et P2 sont ordonnés selon un ordre de succession temporelle qui est lié à un ordre logique de déroulement des faits, faisant apparaître le premier comme la condition de réalisation du second. Il introduit entre les faits une relation qui ne relève pas de la successivité, mais les présente comme dépendants l’un de l’autre […].” (Emphasis ours) [...] alors builds a sequence of temporally ordered events. It indicates that the states of affairs expressed in S1 and S2 respectively are ordered along a temporal succession order which is linked to a logical order of the sequence of events, showing the first one as the preliminary condition for the second one to happen. It introduces a relationship between the facts which does not
Alors as a possible temporal connective in discourse 85
fall within the field of successivity, but which presents them as dependent from each other […] (translation ours, emphasis ours).
It is this relation of dependency between the eventualities – understood as a ‘preliminary condition’ with a “logical order of the sequence of events” – that remains to be further explored. It can coincide with a cause-consequence relation, as in (3), where S1 results in S2 :
(3) Elle s’est mise en colère, ALORS il est parti. She got angry, THEN he left.
But there are other cases of dependency where the result relation is not present, as we saw for example (2), where the temporal succession is linked to a logical sequence of events : the subject had to go up to the square, so that she was able to see the other person arrive. Hybertie takes a strong position, according to which both temporal order and logical order are necessary if we want to use alors felicitously. When there is only temporal order, as in example (4) below, the use of alors is not felicitous 4 :
(4) Nous avons déjeuné. *ALORS nous sommes allés au cinéma. ‘We had lunch. *ALORS we went to the movies.’
3.2. Discussing Hybertie’s proposal
We will not take into account the criterion according to which it is the presence or absence of a temporal referent introduced by a locating adverbial in S1 that determines the value of alors (namely case 1, illustrated by (1), vs. case 2, illustrated by (2) above). In our view, in (1) repeated below,
(1) J’ai rencontré Pierre EN 1987. J’étais ALORS une jeune étudiante. the temporal value of concomitance is not due to the presence of en 1987, but to the imperfective situation in S2. In classical analyses following the anaphoric hypothesis for a tense such as the imparfait in French (see for example Kamp and Rohrer 1983), we would say that the reference time of S1 (r1) is included in the state described by S2 (s2), the temporal relationship between them being an inclusion relation : r1 ⊆ s2. This reference time can be given by the temporal locating adverbial of S1, as in (1). But it can also be
4 Of course, the translation with then is good, but it does not correspond with alors anymore. The correct equivalent in French would be puis. In order to avoid misunderstandings, we will not translate alors anymore in the English glosses.
86 Anne Le Draoulec & Myriam Bras given by the event in S1, in case S1 does not include any temporal adverbial, as in (1’), with the same temporal interpretation r1 ⊆ s2, here equivalent to e1 ⊆ s2 :
(1’) J’ai rencontré Pierre en fac de lettres. J’étais ALORS une jeune étudiante. ‘I met Pierre at the faculty of arts. I was a young student THEN/AT THAT TIME.’
This question remains to be explored further, but we won’t go into further detail here, since from now on we will only consider ‘events’ (and not states). The second point in Hybertie’s analysis – the hypothesis of a dependency relation necessarily associated with alors when it expresses a temporal succession – seems to be extremely relevant and interesting. We are going to explore it further, starting from the analysis of examples (5) to (7) as proposed by Hybertie (pp.27-28).
(5) Nous sommes sortis du cinéma. Il pleuvait ALORS sur Nantes. ‘We got out of the cinema. It was ALORS raining on Nantes.’ (6) Mme de Staël s'installe dans un fauteuil. La femme de chambre frappe ALORS à la porte. ‘Mme de Staël sits down in an armchair. The maid knocks ALORS at the door.’ (7) Paul partit, ALORS l'orage éclata. ‘Paul left, ALORS the thunder broke out.’
These examples are given by Hybertie as possible counterexamples to her proposal, according to which alors, when not preceded by a temporal referent introduction, implies both temporal succession and dependency. These examples are chosen as describing simple narrations. They could, for this reason, seem to convey no dependency relation : for instance, in (6), between the event of sitting-down in an armchair, on the one hand, and the event of knocking at the door, on the other hand. In order to re-establish the validity of her analysis, Hybertie puts forward the notion of ‘point of view’ of an omniscient narrator : in the framework of a narration, alors builds the point of view of the omniscient narrator, according to which S1 has to occur so that S2 can occur, and can be a supplementary step in the narration.
4. Towards a new hypothesis
Interesting as it may seem, Hybertie’s hypothesis does not prove to be entirely appropriate for the examples that she wants to explain, at least not for (5) and (6). We can analyse (5) along the same lines as (1) : we have a
Alors as a possible temporal connective in discourse 87 concomitance relationship between two eventualities, since the second utterance, including an imparfait, describes a state. So there is no need to try and build any dependency link. Besides, example (5) was not particularly commented on by Hybertie. Hybertie applies in a more explicit way her general analysis to (6). According to her, the presence of alors in (6) indicates that the narrator positions herself within a clear and coherent sequence of events in which the sitting-down event has to take place first, and then the knocking event has to take place, in order to enable a third event to take place. Yet we don’t think it necessary for the reader to place herself in the narrative sequence to understand this example. It seems to us that in (6), alors just refers to the time of the event in S1, namely the sitting down event. Now, we must ask whether this position is in contradiction to the one we assumed before, namely that alors requires a dependency link between the two utterances (about examples (2), (3) and (4)). As a matter of fact, there is no contradiction, and example (6) does not have to be regarded as a real counterexample to Hybertie’s hypothesis, as long as one notes that alors is not initial (at the beginning of S2), but internal to S2, in contrast to the examples (2), (3), (4), where alors is initial. We thus claim that this dependency link is present when alors is initial, but not when it is internal. Therefore, the analysis of Hybertie would be all right for (6’), but not for (6) :
(6’) Mme de Staël s'installe dans un fauteuil. ALORS la femme de chambre frappe à la porte. ‘Mme de Staël sits down in an armchair. ALORS the maid knocks at the door.’
On the other hand, an example such as (7), with an initial alors, clearly fits Hyberties’s proposal. It is necessary to include the larger narrative sequence, otherwise (7) is hard to interpret, as Hyberties herself points out. If we move alors in final position as in (7’), the example is perfectly acceptable :
(7’) Paul partit. Un orage éclata ALORS. ‘Paul left, a storm broke ALORS .’
(7’) seems quite acceptable without resorting to the logic of the narrative sequence adopting the point of view of an omniscient narrator. Besides, without the omniscient narrator’s point of view, we move from the definite to the indefinite determiner (i.e. de l’orage à un orage). Let us underline the (obvious) fact that it is thanks to the examples of Hybertie and to the proposal she made that we have been able to elaborate our own hypothesis. In this new hypothesis, the need to establish a distinction according to the position of alors is stated – initial position on the one hand, internal or final position, on the other :
88 Anne Le Draoulec & Myriam Bras
Hypothesis 1 • Initial alors necessarily expresses a dependence link between S1 and S2, • Internal or final alors doesn’t. However, this does not mean that no dependency link can be established when alors is internal or final. Let us compare (8) and (8’) :
(8) Il m'a rejointe. ALORS je me suis souvenue que j'avais oublié mes clés. ‘He joined me. ALORS I remembered that I had forgotten my keys.’ (8’) Il m'a rejointe. Je me suis ALORS souvenue que j'avais oublié mes clés. 5 ‘He joined me. I remembered ALORS that I had forgotten my keys.’
In (8), we have to interpret alors as introducing a dependency relation between the two events, which could be glossed by “it is when he joined me that I remembered that I had forgotten my keys.” In (8’), this interpretation is possible, but only possible. Here, alors only refers to the time of the event described by S1, e1, and could be replaced by à ce moment-là (‘at that time’). If the dependency interpretation is present, it seems that it is not fundamentally part of the meaning of internal alors, as it is for initial alors. In other words, the role of initial alors requires a dependency link between the eventualities described in S1 and S2, whereas the role of internal alors is basically temporal : it may reveal a pre-existing dependency link which is determined by the semantic content of the eventualities (eventuality types) and pragmatic knowledge 6. If the semantic content of the two eventualities linked together is such that it is possible to see a dependency relation between them, the difference between the two cases (initial alors vs. internal alors) is very subtle. It only stems from the way the dependency relation is implied : direct implication, or indirect, secondary effect. And it is true that the distinction can, under these conditions, appear as very intuitive. Nevertheless, if the eventualities appear as obviously independent, it is much easier to show the difference. With an initial alors we have to construe a link (i.e. to put the two eventualities linked by alors in the logical order of a narrative sequence) in order to make the use of alors felicitous. With an internal alors it is not necessary. We already
5 Note that the English translation here is in the simple past, whereas the French version uses the “passé compose” and locates alors between the auxiliary and the past participle. 6 The tendency to adopt, when possible, the interpretation of a strong link in the sequence of events is a well known phenomenon. The presence of alors strengthens this tendency – which nevertheless would exist if alors was not there.
Alors as a possible temporal connective in discourse 89 mentioned that point with the analysis of examples (7) and (7’). We can check it again by comparing (9) and (9’) :
(9) Il s'assit dans son fauteuil. ALORS le téléphone se mit à sonner. ‘He sat down in his armchair. ALORS the phone started to ring.’ (9’) Il s'assit dans son fauteuil. Le téléphone se mit ALORS à sonner. ‘He sat down in his armchair. The phone started ALORS to ring.’
Only the initial alors of example (9) makes us build a dependency relation between “He sat down in his armchair” and “the phone started to ring”. In (9’), the relationship is understood as a properly temporal one. Now, as a second stage of our argument, we are going to re-examine hypothesis 1, in order to see if we can account for the acceptability or non- acceptability of the examples examined up to now. Thanks to this hypothesis we can account for the intuitive difference between initial alors and internal alors. This difference has been observed many times : the temporal value assigned to internal alors is described for example by Franckel (1987). Hybertie herself points out this difference, claiming that the internal position is the favourite place for temporal alors (p.25). Nevertheless, according to Hybertie, this ‘more temporal’ feature does not exclude the requirement for a dependency link between the two eventualities described in S1 and S2. It is on this point that we disagree. We saw above that thanks to her hypothesis, Hybertie could account for the unacceptability of (4) :
(4) Nous avons déjeuné. *ALORS nous sommes allés au cinéma. ‘We had lunch. *ALORS we went to the movies.’
We can also account for this unacceptability with our Hypothesis 1, in accordance with the conflict between initial alors and the absence of dependency relation. Let us now examine example (10), which derives from example (4), where we have moved alors in internal position :
(10) Nous avons déjeuné. *Nous sommes ALORS allés au cinéma. ‘We had lunch. *We went ALORS to the movies.’
We cannot account for (10) with our Hypothesis 1, whereas Hybertie can, as she does for (4), because the dependency link is required, whatever the position of alors. If we want to maintain our Hypothesis 1, we have to explore other means to explain why (10) is not good. This explanation will also lead us to underline another important element for the distinction between initial or internal alors. Let us now observe the differences in acceptability in the pairs of examples below :
90 Anne Le Draoulec & Myriam Bras
(11) Il m'a fait un sale coup. ALORS je me suis vengée, des années plus tard. ‘He played a dirty trick on me. ‘ALORS’ I took my revenge, some years later.’ (11’) Il m'a fait un sale coup. *Je me suis ALORS vengée, des années plus tard. ‘He played a dirty trick on me. *I ALORS took my revenge, some years later.’ (12) Il m'a dit que j'avais l'air fatigué, ALORS je suis partie en vacances. ‘He told me I looked tired, ALORS I set off on holiday.’ (12’) Il m'a dit que j'avais l'air fatigué, *je suis ALORS partie en vacances. ‘He told me I looked tired, *I set off ALORS on holiday.’
These four examples describe successions of events with a temporal gap between e1 and e2. This gap could not be expressed by à ce moment-là (‘at that time’). It is particularly obvious in (11) and (11’), with the adverbial des années plus tard (‘some years later’), which indicates the gap explicitly. We can observe that only initial alors in (11) can combine with this temporal gap. In (11’) we see that internal alors conflicts with the temporal succession and the gap conveyed by des années plus tard. The temporal gap may or may not be explicitly mentioned. In (12) and (12’), it is not explicit. Nevertheless, alors in internal position yields an awkward sentence in (12’). So, if (11’) and (12’) are more difficult to accept than (11) and (12), this means that when alors is not at the front of S2, it cannot express a temporal gap with the eventuality of S1. Thus we have pointed out another major difference 7 between alors in front of the sentence and alors inside the sentence. This prompts us to formulate our second hypothesis as follows : Hypothesis 2 • Only initial alors, implying a dependency link between the utterances, licences the relation of temporal succession with a temporal gap between the events described. • internal alors, whose value is primarily temporal, (which may co- occur with a dependency link, but does not need to) keeps the temporal value of concomitance originally conveyed by alors (cf. à ce moment-là). Let us investigate a little further the case of internal alors. We have just seen that if there is a dependency relation, this dependency is associated with a relation of concomitance : we would make examples (11) and (12) much better if we could modify S2 so that it is understood that S2 describes an immediate reaction to S1 – i.e., that e2 occurs just after e1. We can get this interpretation by replacing je me suis vengée, des années plus tard (‘I took
7 This distinction conflicts with Hybertie’s claim, who talks of shifting from concomitance to temporal succession the same way in both cases.
Alors as a possible temporal connective in discourse 91 my revenge’) in (11’), by je me suis promis de me venger (‘I vowed to take my revenge’) in (13), since se promettre (‘to vow’) can be an immediate response :
(13) Il m'a fait un sale coup. Je me suis ALORS promis de me venger. ‘He played a dirty trick on me. I vowed ALORS to take my revenge.’
We can do the same for (12’), replacing partir en vacances (‘set off on holiday’) by décider de partir en vacances (‘decide to set off on holiday’), or ranger ses affaires (‘put one’s things away’) :
(14) Il m'a dit que j'avais l'air fatigué ; j'ai alors décidé de partir en vacances / j'ai ALORS rangé mes affaires et je suis partie. ‘He told me I looked tired ; I decided ‘alors’ to set off on holiday / I put ALORS my things away and left.’
To make things more precise, we should say that it is possible to have a succession relation with internal alors, but a relation of immediate succession, which is a quasi-coincidence that still can be paraphrased in French by à ce moment là (‘at that time’). Now, taking for granted that this coincidence or immediate succession relation is the only one possible with internal alors, we can account again for the unacceptability of (10) repeated below :
(10) Nous avons déjeuné. *Nous sommes ALORS allés au cinéma. ‘We had lunch. *We went ALORS to the movies.’ because there must be a temporal gap between having lunch and going to the movies, which is not possible with internal alors.
5. Conclusion regarding the connective status of alors
This study of alors in association with event descriptions allowed us to show that : • alors in initial position has a steady role at the discourse level because it always implies a dependency relation. This is a strong constraint, which does not depend on semantic or pragmatic conditions. In accordance with the definition of connectives we gave in the introduction, we can conclude that alors plays a true role of connective when it is in this initial position, • alors in internal / final position shows, in some cases, traces of his connective role. But it is a softer constraint because there is no
92 Anne Le Draoulec & Myriam Bras
steady discourse relation, and because the semantic effect of dependence is determined by the semantico-pragmatic context. In that case, and according to our definition, alors cannot be considered a real connective. We should now investigate more deeply the internal / final distinction. Apart from (7’), we have only considered examples with internal alors, putting internal and final positions on the same level, as opposed to the initial position. Yet a continuum could no doubt be established between the various possible positions. Our intuition is the following : even when the two eventualities linked by alors lend themselves to a consecution relation – in the sense that the second one can be interpreted as a consequence of the first one – it seems that the consecution interpretation becomes less prominent as alors moves away from the beginning of the sentence. Thus, if we compare the examples in (15), the piece of information conveyed by alors seems closer to a purely temporal indication in (c) than in (b) and even more so than in (a), where the temporal information goes with the consecution relation.
(15) (a) Il a continué à insister. ALORS je lui ai dit qu’il se trompait. ‘He kept on insisting. ALORS I told him he was mistaken.’ (b) Il a continué à insister. Je lui ai ALORS dit qu’il se trompait. (c) Il a continué à insister. Je lui ai dit ALORS qu’il se trompait.
Of course this intuition needs to be checked and confirmed through a deeper study of alors in various internal positions as in (b) and (c) above, or in final position. Another important element for the description of alors is the interaction with aspectual and temporal structure. We have restricted ourselves here to event descriptions, but we should also study alors when it appears in sentences describing states. An example such as (16) shows that in such a sentence, even if it is in initial position, alors may not be a connective : its role is only temporal here.
(16) Jours lointains… ALORS tous les espoirs étaient encore permis. (Le destin de Mr Crump) ‘Far-off days… ALORS all the hopes were still permitted.’
Last, we intend to integrate these analyses in SDRT. This will put SDRT to the test in an interesting fashion. SDRT is a theory of the semantics / pragmatics interface, grounded on compositional semantic representations (DRSs) for the propositional content of sentences, built on their syntactic structures. To describe the contribution of alors to discourse representation (SDRS), we need to take into account syntactic factors (cf. the question of position), as well as semantic factors (more particularly,
Alors as a possible temporal connective in discourse 93 aspectuo-temporal parametres). This will be necessary to distinguish between the two different roles that we pointed out here : • the temporal adverbial role when alors introduces a temporal relation of coincidence or immediate succession inside the DRS representing the propositional content of the sentence including alors ; • the connective role when alors triggers the inference of a discourse relation whose temporal effects will have to be compatible with succession. This discourse relation remains to be better specified. We have described it as a ‘dependency’ relation. One of the main questions will be how to express this relation within a discourse relation system including Narration and Result as discourse relations : dependency seems to be situated between Narration and Result, but will it be necessary to introduce it as a new discourse relation, or will we be able to account for it via the scalarity of Narration or Result8 ? We hypothesize that initial alors triggers a weak form of the Result relation, which could be defined by a weak causality relation between the eventualities. The definition of a ‘weak-Result’ relation should imply the redefinition of Result as a scalar relation (with a strong form and a weak one).
References
Asher, N. (1993). Reference to abstract Objects in Discourse, Dordrecht : Kluwer. Asher, N. ; Lascarides, A. (2003). Logics of Conversation, Cambridge : Cambridge University Press. Borillo, A. ; Bras, M. ; Le Draoulec, A. ; Vieu, L. ; Molendijk, A. ; De Swart, H. ; Verkuyl, H. ; Vet, C. ; Vetters, C. (2004). Tense, Connectives and Discourse Structure, in : H. De Swart ; F. Corblin, (eds), Handbook of French Semantics, Standford : CSLI publications, 309-348. Bras, M. ; Le Draoulec, A. ; Vieu, L. (2001). French Adverbial Puis between Temporal Structure and Discourse Structure, in : M. Bras ; L. Vieu, (eds), Semantic and Pragmatic Issues in Discourse and Dialogue : Experimenting with Current Theories, CRiSPI series, vol. 9, Amsterdam : Elsevier, 109-146. Bras, M. ; Le Draoulec, A. ; Vieu, L. (2003). Connecteurs et temps verbaux dans l'interprétation temporelle du discours : le cas de « puis » en interaction avec l'imparfait et le passé simple, Cahiers Chronos 11 : 71- 97.
8 The same kind of question was left open at the end of the study of aussitôt (Le Draoulec 2005).
94 Anne Le Draoulec & Myriam Bras
Creissels, D. (1995). Eléments de syntaxe générale, Paris : PUF. Franckel, J.-J. (1987). Alors, alors que, Bulag 13 : 17-49. Gerecht, M. J. (1987). Alors : opérateur temporel, connecteur argumentatif et marqueur de discours, Cahiers de linguistique française 8 : 69-79. Glasbey, R. S. (1993). Distinguishing between events and times : some evidence from the semantics of “then”, Natural Language Semantics 1 : 285-312. Gosselin, L. (in press). Contraintes pragmatico-cognitives sur l'ordre des constituants. Le cas de séquences de connecteurs exprimant la consécution temporelle, in : Ph. Lane, (éd.), Linguistique du texte et du discours, Rouen : Presses Universitaires de Rouen. Hansen, M-B.M. (1995). “Puis” in spoken French : from time adjunct to additive conjunct ?, Journal of French Language Studies 5 : 31-56. Hybertie, Ch. (1996). La conséquence en français, Paris : Ophrys. Kamp, H. ; Rohrer, C. (1983). Tense in Texts, in : R. Bauerle ; C. Schwarze ; A. von Stechow, (eds), Meaning, use and interpretation of Language, Berlin : De Gruyter, 250-269. Kamp, H. ; Reyle, U. (1993). From Discourse to Logic, Dordrecht : Kluwer. Le Draoulec, A. (2005). Connecteurs temporels d’immédiateté : le cas de aussitôt et soudain, Cahiers Chronos 12 : 19-34. Reyle, U. (1998). A note on enumerations and the semantics of “puis” and “alors”, Cahiers de Grammaire 23 : 67-79.
The power of prepositions: Is he sleeping now or usually?
Tijana ASIC University of Belgrade
1. Introduction 1
What is a temporal reading of a sentence? Does it affect the interpretation of the present tense? How do rain drops change into temporal intervals? What are some possible means of achieving this semantic transformation? Is it possible to conceptually neglect the boundaries of some temporal entities? What is static and dynamic in the spatial and temporal sense? In this paper I will address these questions through an analysis of the po-na opposition in Serbian and its role in the interpretation of the imperfective present. In addition, I will compare it with a similar phenomenon in Kikuyu.
2. About the present tense in Serbian
The imperfective present is the central tense in the Serbian system of verbal tenses. It can be defined (using the taxonomy of Reichenbach 2, 1947) as: S,R,E (S= speech point; R= reference point: E= event point). It should be understood that imperfective in this case actually means that it is a present of imperfective verbs. This is due to the fact that the verbal aspect in Serbian is encoded morphologically 3 – by prefixes and infixes. Hence, we have the following pairs of verbs: citati (to read – imperfective) – PRO-citati (to read – perfective); ot-vA-rAti (open imperfective) – otvO-rI-ti (open perfective). There are two main functions of the imperfective present: First of all, it refers to the eventualities taking place at the moment of speech:
(1) Dusan sada jede svoj ručak Dusan now eats his lunch ‘Dusan is now eating his lunch.’
1 This paper is dedicated to my son Dusan and to my Mali muz, who do not like sleeping but love prepositions. 2 According to Reichenbach (1947) every tense can be defined using three points on the temporal axis: S, R and E. The only difference between tenses is in the relation of these three points: whether they are concomitant (,) or not (-). 3 It used to be coded grammatically (by means of different tenses) but this way to encode aspect is almost completely abandoned today (Asic, 2000).
© Cahiers Chronos 17 (2007) : 95-110. 96 Tijana Asic
This type of present is called referential in the grammars of Serbian (Stanojcic & Popovic, 1997), but I shall call it existential. The same imperfective present can also be used non-referentially (generically – so I will call it here generic present), in which case it denotes eventualities that are not actual in S but that are generally true in the present in a large sense (they can also repeat themselves in some intervals):
(2) Kolja rado jede maline. Kolja eagerly eats raspberries ‘Kolja likes eating raspberries.’ (3) Autobus 45 prolazi svakih deset minuta pored moje kuće. Bus 45 passes every ten minutes by my house ‘Bus number 45 passes by my house every ten minutes.’
Hence, the generic interpretation is triggered either by some adverbs (such as eagerly, every ten minutes) or by the context 4 in its pragmatic sense (Sperber and Wilson, 1986). The imperfective present has also other usages; it can be used interpretatively 5 as well (present for future 6 and narrative present 7) but this is not important for my analysis here (see Asic, 2000). As, for the perfective present (i.e. the present of perfective verbs), it has a great variety of usages in Serbian (it is commonly used instead of the aorist) but can never denote an activity taking place at the moment of speech (Asic, 2000). Its basic semantics is minimalist and its interpretation totally depends on the context or the adverbial in the sentence 8 :
(4) Kolja u tom casu zatvori oci i poljubi vilu. Kolja in that moment closes eyes and kisses fairy. ‘At that moment Kolja closed his eyes and kissed the fairy.’
4 The context is composed of information coming from three different sources: the interpretation of the previous utterances, the situation of the communication act and from the encyclopedic knowledge that we access through concepts. 5 Sperber and Wilson in their book on Relevance make a distinction between descriptive and interpretative usages of utterances. An utterance is used descriptively if it denotes a state of affairs and interpretatively if it denotes a thought about a state of affairs (Sperber and Wilson, 1986). 6 Sutra putujem u Pariz. Tomorrow travel in Paris ‘Tomorrow, I am traveling to Paris’ 7 Petog okotobra 2000 zavrsava se vladavina Milosevica u Srbiji. Fifth October 2000 finishes poss reign Milosevic’s in Serbia ‘On the fifth October 2000 Milosevic’s reign in Serbia finishes’ 8 A sentence in which the perfective present is used (without adverbials), taken in isolation, cannot be situated on the temporal axis.
The power of prepositions 97
In the above-presented example, thanks to the presence of the adverbial locution (at that moment) the present tense of the perfective verbs (zatvoriti (‘to close’) and poljubiti (‘to kiss’)) is interpreted as denoting eventualities in the past. However, in the following sentence, the perfective present refers to an iterative activity taking place in the present in the broad sense:
(5) Svako jutro Dusan poljubi tri puta svoju pra baku. Every morning Dusan kisses three times his great-granny ‘Every morning Dusan kisses his great-granny three times.’
3. The role of some specific adjuncts in the interpretation of the present tense 3.1. The opposition spatially dynamic/spatially static and the prepositions na and po
In previous work (Asic, 2003, 2004, in print), I have shown that Serbian has two different prepositions representing a basic spatial relation contact- support, the prepositions na and po. For both prepositions I suggested the following basic semantics:
xna/po y = dfECxy (x na/po y is equal by definition to x is in external connection with y)
Evidently, the above quoted definition is based on the topological notion “external connection”, defined in the following way by Casati and Varzi (1999):
ECxy =df Cxy ∧ ¬Oxy (x is externally connected to y is equal by definition to x is connected to y and x does not overlap y).
In my opinion this definition suits better the prepositions po and na than does the definition based on the relation of weak contact, (probably the prototype of contact, see Aurnague, Vieu & Borillo, 1997) defined in the following way by Casati and Varzi (1999):
9 10 WCxy =df ¬Cxy ∧ Cx (c (n y)) (x is weak contact with y is equal by definition to x is not connected to y and x is connected to the closure of the neighborhood of y 11)
9 c is closure (cx =df ~(ex)) (Casati & Varzi, 1999). 10 ny =df ιw(Pyw ∧ Opw ∧ ∀z(Pyz ∧ Opz → Pwz)) (Neighborhood operator defined by Asher & Vieu, 1995).
98 Tijana Asic
I have to emphasize here that my definitions of prepositions are minimalist (see Reboul & Asic, 2004) and respond to the Gricean principle of modified Occam’s razor 12 – a minimalist definition of a preposition must satisfy all the varieties of its usages (standard and non-standard) and also leave some space for the contextual inferences (Grice, 1975, Sperber & Wilson, 1986). In the case of na and po it means that the relation of suppor / carrier-carried (that according to some linguists (see Vandeloise, 1986; Herskovits, 1986) should be a part of the basic semantics of such prepositions) is, since it is not present in every usage of these prepositions, purely a matter of contextual inferences. It is clear that in this definition I have avoided the notion of verticality (x on y means that x is higher than y) and that po and na are purely topological prepositions. It seems to me that the relation of external connection implicates better than does the relation of weak contact the notion of support. As for the difference between these two prepositions, it resides in the fact that they give different constraints on the ontological nature and actual state of the figure 13. Let me illustrate this with some examples:
(6) Dusan sedi na stolici. Dusan sits on chair ‘Dusan is sitting on a chair.’ (7) Voda je svuda po podu. Water is everywhere over floor ‘There is water everywhere on the floor.’ (8) Kolja hoda po zici. Kolja walks over tightrope ‘Kolja is walking on the tightrope.’
As shown above, the choice of po or na depends on the ontological nature of a figure and its actual state (movement or rest). If it is continuous or seen as such (a group of objects can be seen as a mass) the preposition po (similar to
11 There is also another way to present the formula of the weak contact (Aurnague, Vieu & Borillo, 1999, 8): WCxy =df ¬C(cx, cy) ∧ ∀z ((P(x,z) ∧ OP (z)) → C(cz,y)) 12 Grice’s principle of modified Occam’s razor (see Grice, 1978) enjoins semanticists not to multiply the senses beyond necessity. 13 I use here the terminology of Talmy: The figure is a moving or concept movable entity whose site, path or orientation is conceived as a variable, the particular value of which is the relevant issue. The ground is a reference entity, one that has a stationary setting relative to a reference frame. The figure’s site, path or orientation is characterized with respect to this reference entity (Talmy, 2000, 184).
The power of prepositions 99 the English preposition ‘over’) is used. If it is discrete (one or many individuated objects), the preposition na (‘on’) is used. Again, if it is static (used with a spatially static verb 14) the preposition na (‘on’) is used, and if it is dynamic (used with a spatially dynamic verb 15) the preposition po (‘over’) is used. The following figure summarizes the above- mentioned rules: FIGURE STATIC IN MOVEMENT Continuous substance PO PO Discrete object(s) NA PO Figure 1: Static and movement constraints All in all, it can be said that na represents a static, discreet contact and po a dynamic, continuous contact.
3.2. Po and na and the non-spatial domain
Thanks to their simple basic semantics the prepositions po and na are also used in more abstract, non-spatial domains, such as the domain of effective natural phenomena (sun, rain, wind). They are called effective (Ivic, 1995) because they act on the subject and she can feel their effects – the heat of the sun, the wetness of the rain drops or the force of the wind:
(9) Dusan sedi na suncu. Dusan sits on sun ‘Dusan is sitting in the sun.’ (10) Kolja trci po kisi. Kolja runs over rain ‘Kolja is running in the rain.’
Note that here, just like in the spatial domain, with spatially static verbs the preposition na is used and with spatially dynamic verbs the preposition po is used. However, it is possible to violate this rule (po with spatially dynamic activities, na with spatially static activities) as in the following sentence:
(11) Moja beba spava po kisi. My baby sleeps over rain ‘My baby sleeps as / when it rains.’
14 Spatially static verbs refer to eventualities that do not involve a linear movement and a horizontal displacement. The examples of this are: standing, singing, running in place. 15 Spatially dynamic verbs describe eventualities involving a linear movement – so there is a horizontal displacement of the subject. The examples of this are: walking, running, crawling.
100 Tijana Asic
This sentence is acceptable in Serbian but it has a specific interpretation, that I call a temporal reading (Asic, 2005). Example (11) means is: what usually happens is that the speaker’s baby is asleep when it is raining. But this does not mean that the baby is exposed to the rain: the contextual implication of the sentence is that she is most probably indoors. Thus, there is no physical contact between the body of the child and the rain drops. Interestingly, a similar sentence where the preposition na is mandatorily used has a physical (spatial) reading:
(12) Kuca spava na kisi. Doggy sleeps on rain ‘The doggy is sleeping in the rain.’
This means that there is a contact between the body of the dog and the rain drops. But it clearly is not a prototypical contact where the ground serves as a support for a figure (as in: ‘The doggy is sleeping on the sofa’.). What kind of contact do we have here? Actually, the dog is under the rain and hence some of the rain drops (during a certain period of time) are touching its fur. But even this kind of relation can be expressed with na, since its semantics is based on the relation of the external connection. It should be emphasized that both sentences (11 and 12) have exactly the same syntactic structure, where the na/po prepositional phrase has the role of a VP-adjunct 16. Yet, the semantic difference between them is obvious. Thus, the meaning compositionality is here partly divorced from the syntax. This case could be seen as a possible argument against the strict form of the principle of compositionality 17.
3.3. One more difference between the spatial and the temporal reading
Native speakers of Serbian have a strong intuition that in sentences where po creates a temporal reading the present tense is commonly understood as generic, while in the sentences with na and the spatial (physical) reading the present is existential. Thus, in (11) the speaker describes what is generally true for her baby and in (12) he is stating what is happening hic et nunc. Let me represent the difference between the two reading formally:
16 In VP-adjunct position a temporal or spatial adverb modifies the verb phrase, which, semantically speaking, means that it specifies the location of the eventuality described by the VP. 17 The principe of compositionality states that the meaning of a complex expression is fully determined by its structure and the meanings of its constituents (see Partee, 1995).
The power of prepositions 101
a. Temporal sentences (with po) – generic reading: ∀ e (rain(e) > ∃ e’ (my baby sleeps e’ ∧ e’ ∈ e)
b. Spatial sentences (with na) – existential reading: ∃ e (rain (e) ∧ ∃ e’ (my baby sleeps e’ ∧ e’ ∈ e) Important: e’ overlaps S (the moment of Speech).
However, these formulae are not completely satisfactory because they do not specify whether e’ is a temporal or a spatial part of e. In fact, in the first case e’ is only a temporal part of e, while in the second one e’ is a spatial and temporal part of e. Is there any possible explanation for this fact? My opinion is that the default reading is e’ is both a spatial and a temporal part of e, but that in the case of the temporal reading we pragmatically infer (using Gricean implicatures (Grice, 1978) or Relevance Theory’s contextual implications (Sperber & Wilson, 1986)) that there is no contact between a sleeping baby and rain drops. Namely our world knowledge tells us that babies normally do not sleep under the rain. However, this inference is defeasible as shown in the following example:
(13) Moja beba spava po mraku. My baby sleeps over darkness ‘My baby sleeps when the night comes.’
What this sentence means is that there is e’ is obligatory a temporal part of e but also that (though this is facultative) e’ is a spatial part of e - the baby is sleeping in the dark. However the main difference between the spatial and the temporal reading lies in the fact that in the first case spatial contact is obligatory, while in the second it is not:
(14) Moja bebe spava po mraku ali uvek u njenoj sobi gori svetlo My baby sleeps over darkness but always in her room burn light ‘My baby sleeps when the night comes but there has to be some light in her room.’
After this digression I will return to my main question: Is the generic interpretation of the present the consequence of the use of po? In order to answer it I shall examine a sentence in which we have some “real” temporal complements:
(15) Moj beba spava posle kiše / popodne / u pet. My baby sleeps after rain /afternoon / in five ‘My baby sleeps after the rain/in the afternoon/at five o’clock.’
102 Tijana Asic
My view is that the present in this kind of sentences is generic, although the interpretation hic et nunc is not excluded. But if we want to insist on it, it is better to say explicitly now and to accentuate it:
(16) Moja beba spava sada, posle kiše / popodne / u pet. My baby sleeps now, after rain / afternoon / in five ‘My baby sleeps now, after the rain/in the afternoon/at five o’clock.’
Moreover, with some kind of adverbs the hic et nunc interpretation can really be annulated:
(17) Moja beba često / retko / uvek / ponekad spava u svom krevetu. My baby often / rarely/ always/ sometimes sleeps in his bed. ‘My baby often/rarely/always/sometimes sleeps in his bed.’
This sounds logical, since these adverbs communicate that the eventuality is repeating itself at certain intervals (regular or irregular) and this is incompatible with the referential present. All in all, the interpretation of the present in sentences with temporal po is not due to its specific nature but to the general tendency to consider the utterances with temporal complements (except deictic adverbs like, yesterday, tomorrow, two weeks ago, in ten days, and other complements referring precisely to a point in past or future) as expressing what is happening in general. In other words, it is because the sentences with po¬- adjuncts are understood as temporal that the interpretation of the present changes.
4. Po-adjunct and the temporal reading 4.1. Introduction
In section 3.2 I have shown that when the rule “spatially static verbs demand na-adjunct” is violated (when the preposition po is used) we get a temporal reading. However I haven’t explained how this happens. In other words, I want to clarify how the preposition po changes the nature of entities like ‘rain’ and ‘darkness’, transforming them into temporal intervals? In order to answer that question I need to introduce the notions of dot objects and coercion.
4.2. About qualia, dot objects and coercion
According to Pustejovsky (1995) the lexicon of any human language is generative and can be viewed as a system of at least four levels of semantic representation:
The power of prepositions 103
a. Argument structure. b. Event structure. c. Qualia structure (different modes of predication possible with a lexical item). d. Lexical inheritance structure.
The notion of “qualia structure” is of a particular relevance for this work for it actually refers to the analytic knowledge associated with words and hence gives us different modes of descriptions of an object (Pustejovsky, 1998). The qualia structure is composed of four components: its constitutive, formal, agentive and telic roles. What they precisely refer to is shown bellow:
CONST – what x is made of FORMAL – what x is TELIC – function of x AGENTIVE – how x came into being
According to Pustejovsky (idem, 10) the FORMAL quale is treated as 18 reference to the super-type and CONST to the material structure of the object, AGENTIVE and TELIC make reference to events, but in very different ways. AGENTIVE identifies a set of individual events associated with the object, TELIC refers to an event description, namely that which under all appropriate circumstances will by default be associated with that object as its function. So, the TELIC quale is typically a partial function from individuals to event descriptions. Thanks to all these components of qualia we can for instance understand why ‘food’ has properties of a physical substance but also refers to edibility (its FORMAL quale is substance and its TELIC quale is ‘x eats y’) and why ‘wine’ is understood as particular liquid made for the purpose for drinking (its FORMAL quale is liquid, its AGENTIVE quale is ‘x makes y’, its TELIC quale is ‘x drinks y’). However, this technique of typing objects becomes unsatisfactory when it comes to words like ‘books’, ‘newspaper’, ‘sonata’, ‘lunch’, ‘appointment’ etc. The reason for this is that, when trying to define their FORMAL qualia (super-types), we get contradictory information from different sentences (or even from a single one). For example:
(18) Your book is on the table. (19) This book is really interesting. (20) This book is heavy and boring.
The book in (18) in a physical object, while in (19) it refers to the information contained in the book. Finally, in (20) we have both meanings of the book (for more examples see Pustejovsky, 1998 and Asher &
18 It says to which broad category the item belongs.
104 Tijana Asic
Pustejovsky, 2004). However it is important to emphasize that two senses of the book are related to one another in a specific way. What unifies them is a deeper structure relating these two senses, known as a dot object. The dot object construction (the type product) allows otherwise contradictory types to be combined into a single type. So, words like ‘books’, ‘newspaper’, ‘sonata’, ‘lunch’, ‘appointment’ are known as complex types (unlike for example ‘food’, ‘wine’, ‘skirt’, ‘table’, which are unified types) or dot objects. What more could be said about the relation that relates the apparently contradictory senses? It has to be a part of their semantic definition: for example, for nouns such as ‘book’, ‘disc’, ‘record’ the relation in question is one of the containment – it is present in their physical and abstract senses (Pustejovsky, 1998). The four levels of semantic representation of words (given in 4.2.1) are connected by a set of generative devices, providing for the compositional interpretation of words in context. Included in these generative operations are the following semantic transformations, all involving well-formedness conditions on type combinations:
1) Type coercion: a semantic operation that transforms the argument into the type imposed by its syntactic function. 2) Selective binding: a lexical item or phrase operates specifically on the substructure of a phrase, without changing the overall type in the composition 3) Co-composition: multiple elements within a phrase behave as functors, generating new non-lexicalized senses for the words in the composition.
The mechanism of coercion is particularly relevant here, for it allows different kind of constructions with verbs and prepositions:
(21) Kolja wants a glass of champagne, a plate of oysters, another cigarette.
In (21), the NPs (a glass of champagne, a plate of oysters, another cigarette) are elevated to the status of VPs (‘to drink a glass of champagne’, ‘to eat a plate of oysters’, ‘to smoke another cigarette’), because it is the function demanded by the predicate (‘to want’). The same thing happens with temporal prepositions and conjunctions like ‘before’ and ‘after’ (Pustejovsky, 1995, 230):
(22) Let’s leave before / after coffee.
The example (22) actually means ‘Let’s leave before / after they serve’ coffee and the NP ‘dessert’ is elevated to the status of a clause. Of course the semantic function of this clause is clear: it is a temporal clause. Let me just
The power of prepositions 105 underline that in all these cases words that are coerced are unified objects and that coercion is possible thanks to their TELIC roles.
4.3. ‘Rain’ and ‘darkness’ as dot objects
It is quite evident that entities like kisa (‘rain’) and mrak (‘darkness’) are a kind of dot objects because they have some physical characteristics (we can feel rain drops on our skin and darkness prevents us from seeing) but they also have duration. One proof of this is that it is possible to say posle kise (‘after the rain’), pre mraka (‘before the darkness’). So, they are composed of the pair of types: event•physical object (substance). What happens when we combine them with spatially static verbs and the preposition po – the marker of continuity and dynamicity? There is a conflict between the semantics of po (it imposes some constraints on the figure; if it is not mass it has to be in movement) and the semantics of these verbs. Namely, these verbs are spatially static but the eventualities they denote are unfolding in time. The only relation conceivable here is a relation between two temporal entities: an activity (like ‘sleeping’) and a quasi-event (like ‘rain’). So, what we get is a dynamic relation between an activity progressing in time and a temporal “interval” during which it is true that it is raining or it is dark. One thing has to be highlighted here: the temporal usage of po is only possible with some specific kinds of dot objects but not with prototypic events (such as ‘storm’, ‘war’, ‘wedding’), as shown bellow:
(23) *Po jučerašnjoj oluji pala su tri drveta. Over yesterday’s storm fallen are three trees ‘During yesterday’s storm three trees fell.’
With real events the preposition tokom (during) has to be used:
(24) Tokom jučerašnje oluje pala su tri drveta. During yesterday’s storm fallen are three trees ‘During yesterday’s storm three trees fell.’
When we use tokom, it means that the figure-eventuality is wholly situated inside the boundaries of the ground (‘storm’). However the preposition tokom cannot be used in Serbian with words like ‘rain’ and ‘darkness’:
(25) *Moj beba spava tokom kise. My baby sleeps during rain ‘My baby sleeps during the rain.’
106 Tijana Asic
(26) *Nedeljama ga vidjam uglavnom tokom mraka. Weeks him see mainly during darkness ‘I have been seeing him mainly at night for weeks.’
This is because real events have prominent boundaries and internal structure (they are heterogeneous, see Casati & Varzi, 1996), while quasi- events do not. They can be regarded as homogeneous and unbounded. This suggests that the temporal relation denoted by po is not an inclusion but merely an overcrossing between two homogeneous entities whose boundaries are neglected (Casati & Varzi, 1999):
OXxy = df Oxy ∧¬Pxy. (x overcrosses y is equal by definition to x overlaps y and x is not a part of y)
In the relation of overcrossing the boundaries of x and y can be disregarded. This relation matches with the semantics of po, based on the spatial (mereo- topological) predicate “external connection”. The final question in this section is: Can the temporal reading of ‘rain’ and ‘darkness’ be seen as a case of coercion? My view is that with this kind of dot-objects the temporal reading does not need to rely on the mechanism of coercion. It is because these dot-objects are composed of events (and some physical substance) so we do not really need to coerce their complements into clauses (as it was a case in my example 22). Moreover, the preposition po in Serbian (unlike prepositions ‘before’ and ‘after’) cannot even have a clause for its argument. In fact, the temporal component of the quoted dot-objects is triggered by the preposition po but at the same time the meaning of po depends on them. Thus, maybe this case of semantic transformation is closer to the phenomenon of co-composition (see 4.2.3).
5. The equivalent of this phenomenon in Kikuyu 5.1. Why Kikuyu?
Kikuyu is a Bantu language spoken in Kenya. It is an agglutinative language with different markers for the noun classes (know as classifiers, see, Kang’ethe, 2003) and a rich system of verbal tenses. Among the languages I have investigated in my work (both Indo-European and non-Indo-European, such as Swahili, Luo, Arab, Hebrew and Japanese) Kikuyu is the only one that, just like Serbian, marks (though not with the same grammatical means) the difference between spatially static and spatially dynamic activities (see, Asic, 2004).
The power of prepositions 107
5.2. Ini and na
In Kikuyu, the general marker of location (both in space and time) is a post- position –ini:
(27) Ibuku rimwe ri metha –ini. Book one is table – at ‘A book is on the table.’
Na is both a preposition and a connector and means ‘with’ or ‘and’. Interestingly, both -ini and na can be used to represent a situation in which spatially dynamic verbs occur under effective natural phenomena:
(28) Ted arahyuka na riua / riua-ini. Ted runs with sun / sun-at ‘Ted is running in the sun.’ (29) Marvin arahyuka na mbura / mbura-ini. Marvin runs with rain / rain-at ‘Marvin is running in the sun.’
However, with spatially static verbs, -ini is the only possibility:
(30) Ted akomete riua-INI / *na riua Ted sleeps sun– at/ *with sun ‘Ted is sleeping in the sun.’
The fact that -ini can be used only with spatially dynamic verbs is not particularly surprising, for it is easy to imagine a kind of concomitance between two processes (‘running’ and ‘raining’) that progress (in a parallel way) in space and time. It seems that spatially static verbs are seen as atemporal (their unfolding in time is conceptually neglected in this case).
5.3. The temporal reading
When I began investigating the data from Kikuyu my hypothesis was that, since in that language the opposition spatially static – spatially dynamic verb is expressed, it might be possible to find something similar to the temporal reading in Serbian. The sentences bellow show that these expectations are realized:
(31) Dusan akomete na mbura Dusan sleeps with rain ‘Dusan sleeps when / as it rains.’
108 Tijana Asic
(32) Dusan akomete na nduna. Dusan sleeps with darkness ‘Dusan sleeps at night (when it is dark).’
In (31) and (32) the above mentioned rule “with spatially static verbs na (‘with’, ‘and’) cannot be used” is violated and the result is a temporal reading for the sentences. The analogy with Serbian is perfect here, for, if the rule is obeyed (-ini is used) the reading is physical:
(33) Ngui akomete mbura-ni. Dog sleeps rain-at ‘The dog is sleeping in the rain.’
It means that in Kikuyu the usage of na (‘with’, ‘and’) highlights the temporal characteristics of entities like rain and darkness and at the same time insists on the temporal progression of the eventuality (‘sleeping’).
5.4. Two types of present in Kikuyu
I have shown (and explained why) in Serbian the temporal reading is associated with the generic interpretation of the present tense, while the physical reading is associated with the existential interpretation of the present tense. The same is valid for Kikuyu: (31) and (32) refer to what is happening in general, while (33) refers to what is happening at the moment of speech. This is quite amazing given that Kikuyu possesses two different present tenses, one for the existential and the other for the generic present:
(34) Marvin a-ra- ria irio. Marvin he-tense-eat food ‘Marvin is eating food.’ (35) Ted a-ria-ga nyama. Ted he-eat-tense meat ‘Ted eats meat.’
The present tense used in examples of the temporal reading ((31) and (32)) is the same as the present tense used in the examples of the spatial (physical) reading and it is the existential present. In other words Dusan anakomete, alone (without any adjunct) means ‘Dusan is sleeping’. But, in spite of that, the sentences (31) and (32) have a generic interpretation. It means that in Kikuyu temporal adverbs can change the default interpretation of tenses.
The power of prepositions 109
6. Conclusion
I hope that it is now possible to give at least a partial answer to the question raised in the introduction. A temporal reading is the interpretation of a sentence in which the adjunct that is not by default an eventuality is understood as a kind of temporal interval. It affects the interpretation of the present tense by making it refer to what is true (or valid) in general, rather than referring to what is happening at the moment of speech. This is also possible in Kikuyu in which there is a special existential present. Moreover, I have shown that if some arguments of po (in sentences violating the po-na rule) are understood as temporal intervals it is because they are cases of complex dot objects composed of the pair of type: physical substance•event. Although these entities have durations, it is possible to disregard their boundaries and to treat them as open intervals.
References
Asher, N. ; Vieu, L. (1995). Towards a Commonsense Geometry. In Proceedings of the International Joint Conference on Artificial Intelligence, San Francisco: Morgan Kaufman Publishers, 846-852. Asher, N. ; Pustejovsky, J. (2004). Metaphysics of words in context, ms., Boston : Brandeis University (http://www.utexas.edu/cola/depts/- philosophy/faculty/asher/papers/MWC.ps) Asic, T. (2000). Le présent perfectif en serbe: temps, mode ou puzzle? Cahiers de Linguistique Française 22 : 275-294. Asic, T. (2003). The po-na-u opposition in Serbian and its equivalents in some Slavic languages and Kikuyu, in : Investigation into Formal Slavic Linguistics, P. Kosta ; J. Blaszczak ; J. Frasek ; Lj. Geist ; M. Zygis, (eds), Frankfurt : Peter Lang, 783-797. Asic, T. (2004). La représentation cognitive du temps et de l’espace; étude pragmatique des données linguistiques en français et dans d’autres langues, Ph.D. diss., Institut des Science Cognitives, University of Lyon-2 and University of Geneva (ms.). Asic, T. (2005). The po-na-u opposition in Serbian and its equivalent in Bulgarian, Balkanistica 18 : 1-30. Aurnague, M. ; Bras, M. ; Vieu, L. ; Asher, N. (2001). The Syntax and Sema- ntics of Locating Adverbials, Cahiers de Grammaire 26 : 11-35. Casati, R. ; Varzi, A., (eds), (1996). Events, Aldershot, Dartmouth Publishing Company. Casati, R. ; Varzi, A. (eds), (1999). Parts and places : The structure of Spatial Representation, Cambridge: MIT Press. Grice, H. P. (1975). Logic and Conversation, in : P. Cole ; J. L. Morgan, (eds), Syntax and Semantics 3: 41-58.
110 Tijana Asic
Grice, H. P. (1978). Further notes on Logic and Conversation, in : P. Cole ; J. L. Morgan, (eds), Syntax and Semantics 9 : 113-127. Herskovits, A. (1986). Language and Spatial Cognition, Cambridge : Cambridge University Press. Kang’ethe, F. I. (2003). Lecture pragmatique des morphèmes temporels du swahili. Ph. D. diss., University of Geneva (ms.). Reichenbach, H. (1947). Elements of Symbolic Logic. New York : Free Press. Partee, B. (1995). Lexical semantics and compositionality, in : D. Osherson (ed), An Invitation to Cognitive Science, Vol. 1, Cambridge : MIT Press, 311-360. Pustejovsky, J. (1995). The Generative Lexicon. Cambridge : MIT Press. Sperber, D. ; Wilson, D. (1986). Relevance. Communication and Cognition. Oxford : Blackwell. Stanojcic, Z. ; Popovic, Lj. (1997). Gramatika srpskog jezika. Beograd : ZUNS. Talmy, L. (2000). Toward a cognitive Semantics. Cambridge : MIT Press. Vandeloise, C. (1986). L’espace en français. Paris : Minuit.
On the dual nature of the Catalan present perfect
Hortènsia CURELL Mercè COLL Universitat Autònoma de Barcelona
1. Introduction
In Catalan, the present perfect has two very clearly differentiated values, as established in Curell (2002, 2003), and Pérez Saldanya (2002) : temporal and aspectual. Temporally, the present perfect is a hodiernal perfective past, i.e., it is used to make reference to eventualities that have obtained within the day. Aspectually, it refers to past situations with present relevance, and, traditionally, two main uses (existential and resultative) and a very marginal one (continuative) have been established. The aim of this paper is to study the aspectual nature of the Catalan present perfect. On the one hand, we have compiled and analysed a corpus, and, on the other hand, we provide a theoretical account of the aspectual use of the present perfect in Catalan. The corpus, manually gathered from 9 contemporary plays 1 written in Catalan between 1974 and 1999, consists of 1020 tokens. The time period was selected to ensure that they would represent present-day Catalan. Theatre plays were chosen because it is the written genre that best represents real language. We are aware, though, that a corpus of spontaneous speech would be necessary for further verification of the proposal presented here. Unfortunately, such a corpus of oral language does not exist for Catalan at the moment. In this paper no attempt is made to provide a review of the numerous papers on the present perfect 2 which have been published, dealing with Catalan and with other languages. The analysis we propose for the Catalan present perfect is based on recent neo-Reichenbachian approaches, such as Kiparsky (2002). When used hodiernally, the present perfect is a temporal operator, and, as such, it establishes a temporal relation of precedence between Reference time and Speech time (R,E – S). In all its other uses, it is an aspectual operator : it establishes a temporal relation of precedence between Event time and Reference time (E – R,S). This aspectual operator focuses on the present resulting state of a past eventuality.
1 For a full list of the plays, see the Appendix. 2 For a good review of existing views on the perfect, the reader can refer to Portner (2003).
© Cahiers Chronos 17 (2007) : 111-127. 112 Hortènsia Curell & Mercè Coll
2. The data
As stated in the Introduction, the data were collected from 9 contemporary Catalan plays, observing the following procedure : each play was read from beginning to end, and the sentences containing the present perfect verb phrase were selected, ensuring that the minimum context for the correct interpretation of the perfect was included. As far as the analysis is concerned, the first step was to separate the hodiernal perfects from the rest. This, clearly, had to be done keeping in mind the whole of the play, since it is the only way to know whether the eventuality referred to took place within the same day or earlier. Having read the whole play, it was easy to determine which perfects were hodiernal and which were not, that is, whether the eventually denoted by the perfect occurred during the day or not. We have excluded from our analysis of the aspectual use of the present perfect hodiernal perfects, although it is the case that sometimes the two readings would be possible for the same utterance. 714 out of the 1020 tokens turned out to be hodiernal pasts, which represents exactly 70%. The remaining 306 instances of present perfect (30%) correspond to the non-hodiernal or aspectual use of the present perfect, which constitutes the core of this paper. The analysis of these uses was carried out independently by the two researchers, and then the results were contrasted and agreed upon. In the first place, we classified the instances into two types : resultative (1) and existential (2) 3. A resultative perfect expresses that a past situation has produced a present result ; and an existential perfect conveys that a given situation has obtained at least once in a period of time that started somewhere in the past and finishes at the present moment.
(1) D’on els han robats ? ‘where have they stolen them from ?’ (2) Ha sentit a parlar de Georges Singles ? ‘have you heard of Georges Singles ?’
In Catalan, contrary to English, the continuative reading of the present perfect is highly marginal, the default form to express that a situation started in the past and still continues at Speech time being the simple present (as in many other European languages, such as Spanish, French, Italian or German). Thus, the English sentence in (3a) would be translated as (3b).
3 Unless otherwise specified, all the examples are from the corpus.
On the dual nature of the Catalan present perfect 113
(3) a. He has studied French for 10 years. b. Estudia francès des de fa deu anys. ‘he studies French since 10 years ago’
In fact, we claim that continuative perfects in Catalan are a subtype of existential readings, a topic which will not be dealt with in this paper. Second, and independently from perfect type, we determined the aspectual class of the predicate 4 in the present perfect : telic or atelic. That is, there was no access to the classification of the perfect into types when the determination of predicate type was made. Third, we noted the instances which contained time adverbials (frequency, time span, ja ‘already’, encara ‘yet’).
2.1. Non-hodiernal present perfect and predicate type
We present here the distribution of the different predicate types in the non- hodiernal perfects of our corpus :
41% telic 59% atelic
Figure 1 Non-hodiernal present perfects and predicate type Clearly, telic predicates are more common than atelic ones. We do not know how relevant this is, since we do not know how often this predicate type appears in other uses and tenses. However, this information will be used as a backdrop to compare it to the distribution of predicate types with the different readings of the present perfect.
2.2. Non-hodiernal present perfect and reading
Here the same data are presented, but from the point of view of the type of perfect. Out of the 306 instances, 186 are resultative and 120 are existential,
4 By predicate type we mean the verb together with the selected arguments.
114 Hortènsia Curell & Mercè Coll which shows that, in our corpus, the resultative reading is the most common one:
39% result
61% exist
Figure 2 Non-hodiernal present perfect readings In the following table, we present the relationship between predicate type and perfect reading :
Predicate type Reading Atelic Telic Resultative 17% 83% Existential 77% 23% Table 1 Non-hodiernal perfects : predicate type and reading It can be clearly seen, from the results presented in Table 1, that there is a strong correlation between predicate type and perfect reading : 83% of the resultative perfect phrases contain a telic predicate, and 78% of the existential perfects appear with an atelic one. We will develop this issue further in the following sections.
2.3. Reading, predicate type and adverb
In this section we deal with each reading of the present perfect, taking into account the predicate type and the presence of adverbials.
2.3.1. Resultative reading
As stated above, a resultative perfect expresses that a past eventuality has a present result. There are 186 sentences with resultative perfects, out of which 155 contain a telic predicate, and 31 contain an atelic one. The results are presented graphically in Figure 3 :
On the dual nature of the Catalan present perfect 115
17%
telic atelic
83%
Figure 3 Resultative reading and predicate type The results shown in this figure support our claim that there is a very strong correlation between the type of perfect and the type of predicate : 83% of the resultative perfects are telic predicates 5, as in the following examples :
(4) Hem firmat l'informe… i el tenen ells ! ‘we have signed the report… and they have it !’ (5) I aquell parell s'han escapat. ‘and those two have run away’
As for adverbs, the data presented in the following graph show that the vast majority of resultative instances occur without adverbial modification. The most frequent adverbs are ja ‘already’ and encara ‘yet’.
13%
adverb no adverb
87%
Figure 4 Resultative reading and adverbs As for the 31 (17%) cases of atelic predicates with a resultative reading, in all instances the context provides a clear resulting state, a consequence of the
5 Recall the data presented in Figure 1 : 59% of aspectual perfects contain a telic predicate.
116 Hortènsia Curell & Mercè Coll event denoted by the verb. The resulting state is, thus, not associated with any adverbial modification at the sentence level.
(6) He jugat amb la bola del món i el senyal m’ha dut fins aquí. ‘I’ve played with the globe and its message has brought me here’ (7) Fas coses estranyes, t’has deixat influir massa per ell. ‘you do strange things, you’ve allowed him to influence you too much’
In (6), the specific NP el senyal ‘its message’ from the second sentence makes reference to the consequence (the result) of having played with the globe. In (7) the first sentence clearly states the consequence of the fact that you have allowed him to influence you excessively, namely that your behaviour is strange.
2.3.2. Existential reading
As stated above, existential perfects express that a situation has obtained at least once in a period of time which started in the past and finishes at the time of speech. Out of the 120 instances of existential reading, 93 are atelic, and 27 are telic, as shown in the following graph.
23%
telic atelic
77%
Figure 5 Existential reading and predicate type Here we find more confirming evidence of the very strong correlation of the perfect reading and the type of predicate. If with resultative perfects, 83% of the predicates were telic, here 77% of the sentences contain an atelic predicate.
(8) Almenys així ho he vist als meus somnis. ‘at least, that’s how I’ve seen it in my dreams’ (9) Quanta paciència que heu tingut amb mi ! ‘how patient you’ve been with me !’
On the dual nature of the Catalan present perfect 117
In relation to adverb presence, the following graphs show that the majority of existential perfects occur with an adverbial, 61% (time span or frequency) :
39% adverb 61% no adverb
Figure 6 Existential reading and adverb Three different factors explain the 27 predicates (23%) which are telic. First, the sentence contains a frequency adverbial : 20 instances, which represent 74% of existential perfects with telic predicates, opposed to the 61% of all existential perfects (example 10). Second, the co-occurrence of a quantified argument : three tokens (example 11). And third, the sentence is negative, 8 examples 6, so that the telicity is negated (example 12).
(10) T’ho he dit mil vegades : és perillós (…) ‘I’ve told you one thousand times : it’s dangerous (…)’ (11) He apunyalat a sang freda setanta criatures innocents (…) ‘I’ve stabbed to death seventy innocent children (…)’ (12) Jo, venjar-me d'un pobre desgraciat que no m'ha fet res ? ‘me, taking revenge of a poor sod who has not done anything to me ?’
Frequency adverbs and quantified NPs bound the resulting state. This bounding ultimately accounts for the iterative reading that these sentences get, which is what gives them the existential reading. Negation renders the resulting state inaccessible by negating the very existence of a resulting state. Thus, these telic predicate are not really telic, but atelic 7.
6 There are four cases in which there is both negation and adverbial. 7 In this sense, our proposal in compatible with a compositional approach to aspect.
118 Hortènsia Curell & Mercè Coll
2.4. Summary
In the corpus analysed a strong correlation has been found between predicate type (telic or atelic) and perfect reading (resultative or existential). The findings can be summarized in the following statements 8 : • Telic predicates have a resultative reading, unless they co-occur with a frequency adverb or a quantified NP, in which case they end up with an iterative existential reading. Negated telic predicates can also have an existential reading, since negation removes the intrinsic goal of the telic predicate, thus rendering it atelic. • Atelic predicates have an existential reading, unless the context in which they are used establishes a clear, unmistakable cause-effect relationship between the event they denote and the resulting state that the present perfect communicates.
3. Towards a unified account of the aspectual present perfect in Catalan
As said before, the present perfect in Catalan has two different uses : temporal and aspectual. As a temporal operator it is a past perfective tense which anchors situations which take place within the day. As an aspectual operator it refers to an event that took place in the past and that has some current relevance. This definition says nothing new about this tense. It is half way between the Indefinite Past (Giorgio and Pianesi 1998) and the Result State (Smith 1991) theories 9. Our analysis will account for the two different attributes these theories consider to be the distinctive feature of this form : indefinite past and result state. Thus, we will not be focussing on either the existential reading or the resultative reading. Rather, we will put forward a unified account of both readings. We aim at giving a single analysis of the present perfect which explains the different readings and how these readings are related to the predicate type of the sentence. Our departure point is, thus, the generalizations that were established in the previous section (2.4). Our proposal is based on Kiparsky (2002) and uses some ideas from Portner (2003). Specifically, we assume, following Kiparsky (2002), that atelic predicates (including states and activities) denote simple events and
8 Cf. Brinton (1988 : 43 & ff.). 9 We leave aside extended now theories, such as Portner (2003) or Iatridou et al. (2001), since they are mainly based on the continuative reading of the present perfect. As mentioned in the Introduction, in Catalan the continuative reading of the present perfect is marginal, and can be considered a subtype of existential reading. For a different approach to the continuative aspect and the present perfect, see García Fernández (2004).
On the dual nature of the Catalan present perfect 119 that telic predicates denote complex events (an activity which leads to a change of state), and that all verb predicates take an event argument which is assigned to the temporal parametres E, R and P. E corresponds to event time, R to reference time, and P to perspective time, the ‘now’ of temporal deixis. If the sentence contains a complex event (a telic predicate), its subevents can be associated to the different temporal parametres. Additionally, we also assume, following Portner (2003), that a perfect tense presupposes the existence of a current relevant state which is related (to be further qualified) to the fact that the event has taken / took place. The unified analysis for the different readings of the aspectual present perfect can be schematically represented as in (13). (13) E R,P
e r
εεε
e stands for the event that took place in a time prior to reference time. r stands for the resulting state of the fact that e took place. εεε stands for the argument event. e corresponds to (part of) εεε. The first two tiers of the structure correspond to the meaning of the perfect. Event time (E) precedes reference time (R), which is included in P. The event (e) is associated to E, whereas the resulting state (r) is associated to R. The relationship between e and r, on the one hand, and that between them and the event argument εεε on the other is eventually going to account for the different perfect readings. In the next subsections we are going to see in detail how this analysis can explain the resultative and existential readings and how they are correlated with predicate type, that is, how it can explain the data presented in Table 1, repeated here for convenience :
120 Hortènsia Curell & Mercè Coll
Predicate type Reading Atelic Telic Resultative 17% 83% Existential 77% 23% Table 1 Non-hodiernal perfects : verb type and reading We will deal first with the resultative reading with telic predicates and the existential reading with atelic predicates, which we consider to be unmarked, default cases. We will then proceed to explain the resultative readings with atelic predicates and existential reading with telic predicates.
3.1. R-reading with telic predicates
The data analysed allows us to claim that there is a strong correlation between telicity and resultative reading and that this correlation works both ways : 83% of telic predicates have a resultative reading, and 83% of the sentences with a resultative reading contain a telic predicate. In Figure 1 we saw that the percentage of telic predicates in the present perfect in general is 59%, which is significantly lower that in the resultative reading. Our proposal can easily explain this correlation. Recall that what characterises the aspectual present perfect is the occurrence of a past event and the existence of a present result state. The complex argument structure of telic predicates provides the two sub-events that are directly associated with the temporal parametres. In other words, telic predicates naturally fulfil the requirements of the present perfect
(14)
E R,P
e r
εεε
e corresponds to the first sub-event (e1), the process part of the telic event. r corresponds to the second sub-event (e2), the result state or change state of the telic event.
On the dual nature of the Catalan present perfect 121
Given the lexical semantic structure of telic verbs, the relationship between e and r is one of cause and effect, determined at the lexical level. Let’s consider now the following example.
(15) Quan arribi, dirà que li ha cridat l’atenció un ocell, o un temple, o una muntanya, que per això s'ha aturat. ‘when he gets there, he will say that his attention was caught by a bird, or a temple, or a mountain, and that’s why he has stopped’
In this particular, case e corresponds to the slowing down process (prior to the stopping itself). And r is the result state : he is not moving. What this sentence asserts is that he is not moving now, he is still because he has slowed down and stopped. A caveat is in order here. Note that Catalan resultative present perfects differ from English in relation to what is asserted and what is presupposed. In Catalan both the process event and the resulting state are asserted, whereas in English only the resulting state is asserted ; the process seems to be presupposed since it is not accessible to questions. For instance, when a speaker sees that the addressee is wearing a new sweater, in Catalan it is possible to ask a question in the present perfect, whereas in English the simple past has to be used, the present perfect being ungrammatical 10 :
(16) Quin jersei més elegant ! On te l’has comprat ? ‘What a smart sweater ! Where have you bought it ?’ (17) a. What a smart sweater ! Where did you buy it ? b. What a smart sweater ! * Where have you bought it ?
3.2. E-reading with atelic verbs
There is also a strong correlation between atelic predicates and existential reading : 77% of atelic predicates have an existential reading and 76% of the sentences with an existential reading contain an atelic predicate – recall that the average of atelic predicates in the present perfect sentences is 41%. In these cases the fact that, as our analysis claims, there is a resulting state anchored to reference time is not as obvious as in the case of resultative readings 11. In an R-reading present perfect the resulting state corresponds, as we have seen, to the second sub-event of the telic predicate. That is, it is part of what is said in the sentence.
10 The following examples are not from the corpus. 11 The difficulty in determining its exact nature has been one of the biggest criticism to resultative theories (see Depraetere 1998).
122 Hortènsia Curell & Mercè Coll
In an E-reading, although it is not so obvious, the present perfect implies that there is a resulting state from the past event. Otherwise, the present perfect and the perfective past would overlap completely. To illustrate this point, let’s consider the following examples (not from the corpus) :
(18) Va ploure molt ‘It rained a lot’ (19) Ha plogut molt ‘It has rained a lot’
Both sentences share the fact that they assert that in a past time the eventuality of raining a lot took place, in other words, their true conditions are the same. They differ, though, in as much as sentence (19), in its aspectual value also refers to a current state, for example, that it is all muddy or that the river is full to overflowing. (19) with an aspectual value can make reference to rain fallen at any time before today, provided that the results of the heavy rain still hold at the time of speech. (19) with a temporal (hodiernal) value can only be used if the rain has fallen within the day, regardless of whether or not the results of the rain are still visible. Following Portner (2003), and departing from Kiparsky (2002), we claim that any sentence with the present perfect, not only the ones with a resultative reading, has to answer the topic of the discourse. That is, the present perfect presupposes the existence of such an answer. Note that by saying that the present perfect presupposes the existence of such an answer we are not referring to the answer’s content, the proposition, but to the fact that it must exist 12. More specifically, we claim that this answer is a state.
(20) E R,P
e r
εεε
12 Portner (2003 : 449) “A sentence S of the form PERFECT (φ) presupposes : ∃q [ANS (q) & P (p, q)]”.
On the dual nature of the Catalan present perfect 123 e corresponds to the event described in the sentence. The existence of r is part of the meaning of the present perfect. The precise, particular, concrete nature of this resulting state is discourse-determined, that is, it depends on the particular discourse in which the sentence is uttered, and thus it may be different when the sentence is part of another discourse. The simple argument structure of atelic predicates does not provide directly the two elements that characterise the present perfect. There is no result state in its structure to be mapped as is the case with telic predicates. That is why atelic predicates need the context in order to be felicitous when used in the present perfect.
(21) He tingut la grip 13 ‘I have had influenza’ e = suffering from influenza r can have different meanings. The first, unmarked meaning, which would be the least discourse-dependent, is that now I am not ill anymore (r = ¬¬¬e). Or it may mean that now I am weaker than usually, or that now my immune system is stronger. It all depends on the specific discourse. Once we consider cases under the particular circumstances in which (21) has actually been uttered, the content of r is not undetermined anymore, but pragmatically fixed, as we can see in the following example from the corpus :
(22) Pensa que per aquí ha desfilat mig país amb la il·lusió de fer carrera. ‘take into account that half of this country have paraded here hoping to make headway’
What we have to consider is not the fact that some people have come to this place, but its peculiarity. The fact that some people have come here has made this place more important and relevant. So far we have explained the majority of the examples from our corpus, that is, telic predicates with resultative reading and atelic predicates with existential reading, which together constitute 82% of the sentences. The main difference between these two types of perfect relies on the properties of r, the resulting state. In the first case it is part of the event argument, specifically the second sub-event of the complex predicate, which is a consequence of the first sub-event, the activity, and is thus asserted ; whereas in the second case the resulting state is determined by the context.
13 Example not from the corpus.
124 Hortènsia Curell & Mercè Coll
3.3. R-reading with atelic verbs
As mentioned in section 2.3.1, in all the instances of present perfect with resultative reading and atelic verbs the result state is specified by the linguistic context, as can be seen in examples 6 and 7, repeated here :
(23) He jugat amb la bola del món i el senyal m’ha dut fins aquí. ‘I’ve played with the globe and its message has brought me here’ (24) Fas coses estranyes, t’has deixat influir massa per ell. ‘you do strange things, you’ve allowed him to influence you too much’
In both examples the content of the resulting state is determined by the linguistic context. The causal relationship between e and r could be considered to be the result of aspectual shift, coercion Che. According to de Swart (2000), the present perfect is an aspectual operator which matches sets of telic events onto sets of states, that is, the present perfect is a stativizer. This tense has input restrictions : it requires an event, a telic predicate. Whenever there is a mismatch between the verb and the input requirements of an operator (be it aspectual or temporal) there is an aspectual shift, a coercion. The coercion, as this author defines it, is both morphologically and syntactically invisible, that is, it is a covert operation, whose effects can be seen at the semantic level. We diverge from de Swart in the specific use of this aspectual shift and atelic predicates. She claims that whenever we have a present perfect with an atelic predicate, there has been a coercion, which is eventually responsible for the existential reading. Instead, we propose that only in those cases in which the atelic predicate gets a resultative reading does coercion take place. It gives us the causal relationship between e and r. This aspectual shift is semantically visual in the causal relationship between the event that took place in a past time and the current resulting state. The analysis of (24) under our proposal is :
(25) E R,P
e r
εεε
On the dual nature of the Catalan present perfect 125
e= allow yourself to be influenced. r= your behaviour is funny, you are quirky. r is the consequence of e, Che.
3.4. E-reading with telic verbs
Finally, the last case to be considered is the existential reading with telic predicates. What characterizes these cases is the fact that the resulting state is not related to the second sub-event of the telic predicate.
(26) T'ho he dit mil vegades : és perillós (...) ‘I’ve told you a thousand times : it is dangerous (…)’
(27) E R,P
e r
εεε
e = tell it (a thousand times). r= you should already be aware of the fact that it is dangerous. Here e corresponds to the whole complex event, both the process sub-event and the result state sub-event. r does not coincide with the second sub-event, as in the case of resultative reading with telic predicates. As was mentioned in section 2.3.2, sentences with telic predicates and existential readings generally contain a frequency adverb – in our corpus 74%. What frequency adverbs do, we claim, is bound the second sub-event, the result state and thus render it not accessible to the result state which is associated with R. Similar roles are played by negation and quantified direct objects. The content of r is not lexically determined but it is to be pragmatically filled up.
3.4. Conclusions
The following tables summarise the analysis of our data according to our proposal :
126 Hortènsia Curell & Mercè Coll
R-reading : e causes r telic predicates atelic predicates lexical cause cause is due to Che E = e1 e = εεε r = e2 r’s content depends on linguistic context Table 2 Resultative Perfect
E-reading : e doesn’t cause r atelic predicates telic predicates E = εεε e = εεε r = ¬e r ≠ e2 r’s content is pragmatically determined Table 3 Existential Perfect We have provided a unified account of the different readings of the present perfect. What tells the resultative reading apart from the existential reading is the nature of the relationship between the event that takes place in a time previous to reference time and the resulting state which holds at reference time. The causal relationship can be lexically determined (telic predicates) or due to an aspectual shift (atelic predicates). When it is not causal, the present perfect gets an existential reading. In the latter cases, the content of r is pragmatically determined.
Appendix : Catalan theatre plays
Abellan, J. (1986). Eclipsi, Barcelona : Llibres del Mall. Altés i Campà, J. (1998). Montserrat, Marta i Maria, Barcelona : Editorial Millà. Batlle, C. (1999). Les veus de Iambu, Barcelona : Edicions 62. Baulenas, L.-A. (1992). No hi ha illes meravelloses, Barcelona : Institut del Teatre. Belbel, S. (1998). La sang, Barcelona : Edicions 62. Benet i Jornet, J. M. (1997). El gos del tinent, Barcelona : Edicions 62. Bergoñó, J. (1974). Por, Barcelona : Edicions 62. Bras, P. (1995). “City bang blues”. Sang i fetge a la ciutat, Barcelona : Editorial Millà. Coquard, L. (1995). Ens ha caigut la sogra, Barcelona : Editorial Millà.
On the dual nature of the Catalan present perfect 127
References
Brinton, L. J. (1988). The Development of English Aspectual Systems, Cambridge : Cambridge University Press. Curell, H. (2002). The present perfect in English and in Catalan, in : K.M. Jaszczolt ; K. Turner, (eds), Meaning through Language in Contrast, Amsterdam/Philadelphia : John Benjamins, 101-115. Curell, H. (2003). El perfet en català, in : C.D. Pusch ; A. Wesch, (eds), Verbalperiphrasen in den (ibero-)romanischen Sprachen, Hamburg : Helmut Buske Verlag, 35-46. Depraetere, I. (1998). On the resultative character of present perfect sentences, Journal of Pragmatics, 29 : 597-613. García Fernández, L. (2004). Aspecto y estructura subeventiva en las formas compuestas del verbo, Cuadernos de Lingüística Instituto Ortega y Gasset 11 : 43-59. Giorgio, A ; Pianesi, F. (1998). Tense and Aspect, Oxford : Oxford University Press. Iatridou S. ; Anagnostopoulou E. ; Izvorsky R. (2001). Observations about the Form and Meaning of the Perfect, in : M. Kenstowicz (ed.), Ken Hale. A Life in Language, Cambridge (Mass.) : The MIT Press, 189-238. Kiparsky, P. (2002). Event structure and the perfect, in : D.I. Beaver ; L.D. Casillas Martínez ; B.Z. Clark ; S. Kaufmann, (eds), The Construction of Meaning, Stanford : CSLI Publications, 113-133. Pérez Saldanya, M. (2002). Les relacions temporals i aspectuals, in : J. Solà ; M.-R. Lloret ; J. Mascaró ; M. Pérez Saldanya, (eds), Gramàtica del català contemporani, Vol. 3, Barcelona : Empúries, 2567-2662. Portner, P. (2003). The (temporal) semantics and (modal) pragmatics of the perfect, Linguistics and Philosophy 26, 459–510. Smith, C. S. (1991). The Parametre of Aspect, Dordrecht : Kluwer Academic Publishers. Swart, H. de (1998). Aspect shift and coercion, Natural Language and Linguistic Theory 16 : 347-385. Swart, H. de (2000). Tense, aspect and coercion in a cross-linguistic perspective, in : M. Butt ; T.H. King, (eds), Proceedings of the Berkeley Formal Grammar Conference, University of California, Berkeley : CSLI Publications. http://csli-publications.stanford.edu.
This page intentionally left blank
Epistemic modality and questions in dialogue. The case of Italian interrogative constructions in the subjunctive mood
Andrea ROCCI University of Lugano
1. Introduction
In this paper I present a semantic analysis of the Italian interrogative construction in the subjunctive mood introduced by the complementizer che exemplified in (1).
(1) [La macchina di Giovanni non è nel parcheggio.] Che sia andato a casa ? That go-perf.-subj. to home Roughly 1 : [John’s car is not in the parking lot.] Has he possibly gone home ?
According to Fava (1995) the construction introduces in the question “a nuance of doubt and/or desire”. For this reason the construction is defined ‘optative / dubitative’. Portner (1997 : 193) sees this kind of sentences as expressing “doubt and astonishment”. As we will see below, these glosses are far too generic to account for the rich, very specific constraints that the construction imposes on interpretation. The purpose of our analysis is threefold. Firstly, it will serve as a ‘case study’ in a broader investigation of the interaction of questions and epistemic modality, and more specifically of the presence in interrogative contexts of the kind of non-propositional epistemic modality and inferential evidentiality 2 that is manifested in assertive contexts by markers such as the
1 Since in this paper I will be discussing modal markers which have no direct equivalent in English, all the translations provided should be considered just as an aid to follow the argumentation and not as an accurate rendition of the actual constraints on interpretation imposed by the Italian or French constructions considered. Most “natural” English translations of these examples, while contextually appropriate, do not encode the same meaning at all of the Italian original. 2 Here we define evidentiality as the indication in the utterance of the source of the information presented in the propositional content (i.e. the type of evidence the speaker has for the proposition expressed). See Faller (2002) and Rocci (2005a)
© Cahiers Chronos 17 (2007) : 129-153. 130 Andrea Rocci epistemic uses of the modal verb dovere (‘must’) (2.a) and the future tense of the indicative mood (2.b).
(2) La macchina di Giovanni non è nel parcheggio. a. Dev’essere andato a casa. b. Sarà andato a casa. John’s car is not in the parking lot. He must (?will) have gone home.
Utterances such as (2.a) and (2.b) present the asserted proposition as inferred by the speaker at the moment of utterance from available evidence with a medium or high degree of confidence. In contexts such as (2.a-b) dovere and the ‘epistemic’ future help establishing inferential discourse relations between utterances where the second utterance is regarded as a conclusion derived from the first 3. Interestingly similar discourse relations arise in the context of question-type speech acts with the interrogative constructions with the subjunctive (1), as well as with the epistemic future (3) :
(3) La macchina di Giovanni non è nel parcheggio. Sarà andato a casa ? Roughly : John’s car is not in the parking lot. Has he possibly gone home ?
In French, similar discourse relations are established by a particular ‘inferential’ use of the conditional mood 4, which is found only in questions (cf. Diller 1977) :
(4) La voiture de Jean n’est pas dans le parking. Serait-il rentré à la maison ? Roughly : John’s car is not in the parking lot. Has he possibly gone home ?
It is not easy to figure out, however, how exactly the epistemic modality and the inferential evidentiality combine with the illocutionary force of the non- assertive speech acts associated to yes/no questions in examples such as (1), (3) and (4). Firstly, it is not clear whether the modality should be considered as part of the questioned propositional content, as a sort of concomitant expressed attitude, or as a presupposed evaluation. Secondly, one has to figure out who exactly is the subject of the epistemic evaluation and of the act of inferring, as these interrogative constructions seem to express not only the speaker’s evaluation of a state of affairs, but also the speaker's evaluation of hearer’s evaluation of the state of affairs.
3 See Rocci 2000, 2005a and 2005b for an in depth analysis of these constructions. 4 I do not enter here the vexata quaestio concerning the nature of tense or mood of the French conditional morphemes. Here we will be concerned only with “modal” uses of the conditional. For the sake of simplicity, I chose to call it a mood.
Epistemic modality and questions in dialogue 131
Apart from opening a window on the interactions between epistemic modality and questions, an analysis of the ‘epistemic’ interrogative construction in the subjunctive introduced by che (henceforth ‘che+subjunctive’ construction) is interesting for two other reasons. The analysis will show that the ‘che+subjunctive’ construction is a striking example of a conventional, internally non compositional, indicator of a series of abstract constraints on the common ground (CG) of the dialogue and on the pragmatic effect of the utterance. In this sense the analysis of such a construction brings further support to the claims made by researchers in Construction Grammar (Kay 2003) on the great variety of “illocutionary forces” encoded by “minor sentence-types” and grammatical constructions in general 5. Finally, it will be shown how congruity theory (Rigotti & Rocci 2001 ; Rigotti, 2005 and Rocci 2005c), an integrated approach to speech acts and discourse relations, can be used to represent straightforwardly in terms of a predicate-argument frame the constraints imposed by such a construction on the structure of the dialogue. In the following pages I will first provide a first sketch of a general picture of the compatibility of markers of epistemic modality in Italian with questions. Then I will move to consider the ‘che+ subjunctive’ interrogative construction as a specific marker of epistemic modality and inferential evidentiality in questions. In order to characterize its semantics and to argue for its status of a non-compositional construction I will compare and contrast it with other markers of epistemic modality in questions such as the use of the epistemic predicate credere (to believe), the epistemic-inferential future in Italian, and the French inferential conditional.
2. Epistemic modality and questions : a first sketch of the big picture
If we consider different lexical and grammatical markers of epistemic modality in Italian such as mental state predicates (credere ‘believe’), epistemic adjectives and adverbs (probabile ‘probable’, probabilmente ‘probably’), the epistemic readings of the modal verbs (dovere ‘must’ and potere ‘can / may’), or the epistemic use of the future tense, we find that their behavior in yes/no questions varies along various dimensions : whether they can occur in questions at all or not, whether the epistemic attitude expressed is part of the questioned propositional content or not, the cognitive subject to which the epistemic attitude is attributed, which can be the speaker, the hearer (as construed by the speaker) or a larger collective epistemic
5 On the grammatical encoding of “conversational move types” see also Ginzburg, Sag & Purver (2003).
132 Andrea Rocci community, which might include the speaker, the hearer, as well as other subjects. Variation along these dimensions seems to reflect certain features of the semantics of the marker. It is therefore interesting to observe how this variation correlates with the strength of the epistemic qualification expressed, with the presence of an inferential evidential meaning in addition of the epistemic qualification, the truth-conditional or non truth-conditional status of the marker, the type of modal quantification (e.g. necessity vs. possibility), and more generally the strength of the commitment to the truth of the proposition. Mental state predicates such as credere (‘believe’) and pensare (‘think’) are unmarked as far as the strength of the epistemic qualification is concerned. Credere, in particular, ranges from the strong philosophical meaning of take to be true to the conversationally more frequent weak meaning of take to be probable at least to some specific degree 6. Quite naturally with these predicates the cognitive subject to which the epistemic attitude belongs coincides with the syntactic subject of the verb. When occurring in the first person in sentence initial position, as in (5),
(5) Credo / penso che sia andato a casa ‘I think he’s gone home.’ these verbs can be taken either as part of the propositional content or as semantically ‘parenthetical’ non-propositional markers modifying the strenght of the assertive commitment 7. This interpretation is invariably associated with the weak reading of the verb. In yes/no interrogatives these mental state verbs can appear as part of the questioned content and – again quite obviously – continue to attribute the attitude to their syntactic subject. However, their use in the first person in interrogatives is pragmatically constrained – in most situations it does not make sense to question our own beliefs – and their use in the second person gives rise to an interesting implicature. A sentence such as (6),
(6) Credi che Giovanni sia andato a casa ? ‘Do you think John went home ?’ can be used either to question the addressee about her own beliefs – especially when the speaker thinks they are false – or to ask indirectly a question about the truth of the embedded proposition. This second use of (6) differs from the direct question (7)
6 Cf. Pinkal (1983) for similar observations on French croire and German glauben. 7 Cf. Urmson (1969), Hooper (1975) and Venier (1991) for Italian.
Epistemic modality and questions in dialogue 133
(7) Giovanni è andato a casa ? ‘Did John go home ?’ because (6) conveys the additional implicature that the speakers believes that the hearer cannot fully commit herself to the truth or falsity of the proposition because she lacks conclusive evidence. As we will see, this reading of the questions with credere in the second person – a verb which always takes clausal complements in the subjunctive mood – is particularly interesting for its closeness to the meaning of the ‘che+subjunctive’ construction. The behaviour of the epistemic readings of the modal verbs potere (‘can’ / ’may’) and dovere (‘must’) in questions diverges in interesting ways that are predictable given the semantic analysis of these verbs proposed in Rocci (2005a and 2005b) :
(8) a. Giovanni può essere andato a casa ? b. *Giovanni dev’essere andato a casa ?
The non-propositional evidential dovere cannot occur in questions neither as an element of the questioned propositional content or as a non-propositional modifier of the speech act. This behaviour can be explained as follows. Epistemic dovere, like English epistemic must, points deictically to an on-line act of inference on the part of the speaker at the moment of utterance. Nuyts (2001) calls this property – which is shared by other non-propositional markers of epistemic modality, such as the epistemic adverbials, and ‘parenthetical’ first person mental state predicates – performativity. Faller (2002) calls it M- performativity in order to distinguish it from the performativity stricto sensu exhibited by predicates referring to speech acts rather than to mental acts.8 Let us consider the following dialogue :
(9) A : Giovanni non deve essere uscito (I infer that Giovanni hasn't left) B : Non è vero ! (That’s not true !)
In (9-B) the denial (That's not true) cannot take the epistemic modality expressed by dovere as part of its (anaphoric) argument and takes scope only over the embedded proposition, as the impossibility of a continuation like *Non deve esserlo (literally He must not have) shows.
8 Note that speech acts are actions which are accomplished by uttering certain linguistic signs, while mental attitudes and acts of inference simply happen to be concomitant with the utterance and have an independent (mental) existence. This makes a substantial ontological difference.
134 Andrea Rocci
The property of M-performativity explains directly why dovere cannot be part of the questioned propositional content as questioning one’s own current inferential process or belief state does not make sense. Uttering *Giovanni dev’essere andato a casa ? would count for the speaker as asking whether or not she is inferring at the very moment of utterance that John is at home. Moreover, dovere cannot appear as a non-propositional evaluation of the questioned content concomitant with act of questioning, because the high degree of confidence it expresses is incompatible with the ‘open’ epistemic evaluation (cf. Gobber 1999) presupposed by sincere questions. This semantic-pragmatic explanation of the impossibility of dovere in questions is confirmed by two apparent exceptions, where epistemic dovere does appear in interrogative contexts. The first exception is represented by iconic echo questions such as (10),
(10) A- Giovanni dev’essere a casa a quest’ora. ‘John must be at home by now’
(11) B - Dev’essere a casa ? ‘He must be at home ?’
Here the act of inference is not attributed to the speaker at the moment of utterance, and consequently there is no incompatibility with questioning. The second apparent counterexample is represented by requests of confirmation with tag questions such as
(12) Dev’essere andato a casa, vero ? / no ?/ nevvero ? / non è vero ? ‘He’s gone home, hasn’t he ?’
Given the behaviour of dovere, it is not surprising that the non truthconditional sentence adverbs probabilmente ‘probably’ and certamente ‘certainly’, which share the same status of dovere as regards the deictic reference to the current epistemic attitude of the speaker, have similar restrictions as regards their occurrence in questions.
(13) *Certamente è andato a casa ? ‘Did he certainly go home ?’ (14) *Probabilmente è andato a casa ? ‘Did he probably go home ?
Both are impossible with a normal interrogative intonation pattern, while they are marginally acceptable as requests of confirmation with a very different intonation pattern. Their unacceptability seems to derive from the combination of their indexical properties with the high degree of certainty
Epistemic modality and questions in dialogue 135 they convey. Their adjectival counterparts certo and probabile signal the same degree of certainty, but differ markedly as indexicality is concerned. They are clearly propositional and can occur in questions, where they are clearly interpreted as part of the questioned propositional content :
(15) E’ probabile che sia andato a casa ? ‘Is it probable that he went home ?’ (16) E’ certo che sia andato a casa ? ‘Is it certain that he went home ?’
In classic accounts of modality, epistemic adjectives are said to convey an objective epistemic modality, that is a mathematically quantifiable probability (see, for instance, Lyons 1977), or a probability which is based on known facts rather than on subjective appreciation (see Kratzer 1981). This objectivity is supposed to account for their propositionality. In my view, the tendency towards objectivity in modal adjectives and their behaviour in questions can be better explained if we consider them as context dependent operators as opposed to true indexicals such as the epistemic adverbs and the modal dovere. The propositions in their conversational background (cf. Kratzer 1981) 9 – or, in other words, the set of accessible worlds over which they quantify – is determined as the set of the collective knowledge available in a certain contextually determined epistemic community : possible instantiations of this conversational background range from what is known to humankind and what is known to biologists to what is known by a very restricted contextually relevant set of people. Similarly, the modal verb potere, which expresses a propositional kind of epistemic modality can enter the propositional content of questions, where it seems to refer to an attitude of the hearer. On the contrary, as we have seen above in (3) the epistemic-evidential reading of the Italian future tense – which is largely synonymous with the epistemic reading of dovere (cf. Rocci 2000) – can occur in interrogative sentences that realize certain subtypes of question acts. What is the precise meaning of the epistemic future in questions such as (3) ? How does its
9 The notion of modal conversational background introduced by Kratzer (1981) is the heart of most modern formal treatments of the semantics of the modals. According to Kratzer’s theory the force of modal operators in natural language is not absolute but is relative to a contextually determined set of propositions (the conversational background). For instance if someone in a train says to a fellow passenger You can’t smoke in here this is a non smoking compartment she refers to a conversational background consisting of the laws and the railways regulations. In section 6.2 below we will provide a little more detail on the technical functioning of Kratzer’s approach.
136 Andrea Rocci epistemic-inferential meaning interact with the conversational force of the question act ? More generally, we are faced with a problem concerning the proper representation of non-propositionally expressed epistemic attitudes in the context of questions, that is of epistemic modalities that seem to lie ‘outside’ the questioned propositional content and of signal an attitude of the speaker which is concomitant with the act of questioning and has the propositional content of the question – or some other related proposition – as its object. With a few exceptions (cf. Doherty 1987), non-propositional epistemic modalities have been mainly examined in the context of assertions, where they have often been regarded as signalling the strength of the speaker’s commitment to the truth of the asserted proposition (cf. Lyons 1977). Authors adopting more or less orthodox versions of Searlian Speech Act Theory as the framework for the description of epistemic modality and evidentiality (cf. Faller 2002) consider these markers, as illocutionary force modifiers which add to the felicity conditions of the illocutionary act or modify them. Within assertions the concomitant attitudes are seen as modifying or adding to the attitudes which are part of the sincerity conditions of the illocutionary force. Our analysis of the ‘che+subjunctive’ construction will show that this account cannot always be extended to the epistemic-evidential markers appearing in speech acts of the question type. For instance, an analysis of these markers cannot be carried out without referring on the one hand to presupposed attitudes – which in Searlian terms would belong to the preparatory conditions of the speech act – and, on the other hand, the deictic reference to the speaker’s evaluation of the hearer’s evaluation of the state of affairs 10.
3. The interrogative construction che + subjunctive as a marker of epistemic modality and evidentiality
Let us examine more closely the ‘che + subjunctive’ interrogative construction that we find in sentences such as (1). Starting from a few basic remarks on its morphosyntax, we will then move to consider the semantics and pragmatics attached to it. First basic remark : as Fava (1995 : 117) observes, here che is not the interrogative pronoun (= ‘what’), like in :
10 Note that both presupposed attitudes and the speaker’s evaluation of the hearer’s evaluation were found to be relevant for the semantic analysis of German modal particles in Doherty (1987).
Epistemic modality and questions in dialogue 137
(17) Che ha detto Giovanni ? (What did Giovanni say ?) but the homophonous complementizer (= ‘that’). The question in (1) is interpreted as a yes/no (decision) question and a not as a wh- (content) question. This feature already sets apart the ‘che+subjunctive’ construction from the epistemic future, which can also occur in content questions, as in (18).
(18) Che cosa avrà detto Giovanni ? (Very roughly : What did Giovanni say ? I’m wondering.)
It should not be difficult to advocate for ‘che+subjunctive’ interrogatives the status of a construction, in the technical sense used by Construction Grammars, that is of ‘a form-meaning pair such that some aspect of its form or some aspect of its meaning is not strictly predictable from its component parts or from other constructions’ (cf. Goldberg 1995) 11. The detailed constraints that ‘che+subjunctive’ imposes on interpretation do not seem to be plausibly derivable compositionally from the semantics of its components : that is the subjunctive mood, the complementizer che, and the fall-rise intonation pattern that characterizes yes-no questions in Italian. In fact, if the contribution of the interrogative intonation pattern to the whole seems reasonably clear, those of the complementizer and the subjunctive mood are not. Attempts at providing a unitary semantic explanation to the uses of the Italian subjunctive, both in traditional grammatical studies (see Schneider 1999) and in formal semantics (cf. Portner 1997) have singled out, at best, very general meanings for this inflectional category, such as ‘non-assertion’ (cf. also Hooper 1975 : 123) which are largely underdetermined with respect to the semantics of the ‘che+subjunctive’ construction 12. Similar remarks can be done for the
11 For a sketch of the overall conception language as a sign system in which I ideally place the notion of construction I mutuate from Construction Grammar see Rigotti & Rocci (In press). 12 Portner (1997) in his formal account based on a version of situation semantics see the contrast between the Italian indicative and subjunctive moods as one between clauses having the features of (doxastic) necessity and factivity (indicative), and clauses deviating from these two features for various reasons (subjunctive). The two features in which Portner decomposes the “informal notion of certainty” associated with the indicative are given a rather sophisticated formal specification in terms of situation semantics, which I cannot adequately discuss here. What has to be remarked, however, is that in Portner’s semantics there are many ways of deviating from necessity and factivity, which is quite understandable, as Portner is trying to provide a basic semantics that accounts for the occurrence of the subjunctive in many different
138 Andrea Rocci complementizer che : provided we want to assign a semantics proper to this functional word, this semantics cannot be but very generic and, perhaps also very close to the non-assertiveness associated with the subjunctive. Comparing the interrogative ‘che+subjunctive’ construction with superficially similar optative and imperative subjunctive constructions prefaced by che does not yield particularly illuminating results as far as compositionality is concerned. Here the complementizer che occurs optionally in 3rd person imperative constructions :
(19) (Che) se ne vada pure ! ‘Let him go !’ and alternates more or less freely with subject inversion in certain optative constructions :
(20) Che tu possa essere felice ! / Possa tu essere felice ! ‘May you be happy !’
What this comparison suggests is a non-compositional explanation similar to the generative semantic one proposed in R. Lakoff (1968) for the Latin subjunctive, which postulated a phonologically null ‘abstract verb’ of desire or belief as the hidden matrix clause of the independent subjunctives. Applied to our examples, this approach would lead to hypothesize an abstract desire predicate for (19) and (20), and some sort of abstract belief predicate for (1). Such an explanation – which may have a grain of truth diachronically – does not explain, though, the specific constraints of the interrogative ‘che+subjunctive’ construction. Let us see why. Credere (‘believe’), or pensare (‘think’) and many other verbs expressing mental states in Italian obligatorily take subjunctive clauses as objects (cf. Schneider 1999 for a comprehensive list) 13. As we have seen in section 2, these verbs can be used in the 2nd person in a yes/no question to obtain an effect which is close to that of the ‘che+subjunctive’. A question like
contexts. As for the occurrence of the subjunctive in the ‘che+subjunctive’ questions, Portner’s semantic rule is certainly compatible with it, but the rule remains enormously underspecified with respect of the actual epistemic attitudes regularly associated with this type of sentence which still have to be explained by other factors, be they conversational (inferential) or linguistic (conventional) factors. 13 These clauses are either introduced by the complementizer che (Credo che sia andato) or by no complementizer at all (Credo sia andato). In Italian, The ‘null complementizer’ option is impossible with complement clauses in the indicative mood (So che è andato / *So è andato).
Epistemic modality and questions in dialogue 139
(21) Credi che Giovanni sia andato a casa ? ‘Do you believe that Giovanni went home ?’ can be interpreted either as asking a question about the belief state of the hearer – maybe by someone who knows about the truth of the state of affairs embedded in the belief predicate, like in (22), or, like in (23), as asking a question about the truth of the embedded state of affairs without expecting that the hearer has full knowledge about that, and more importantly without binding her to the strict respect of the Gricean maxim of quality that a simple question would have implied.
(22) A : Credi che Giovanni sia andato a casa ? ‘Do you believe that Giovanni went home ?’ B : Sì, lo credo. ‘Yes, I do’ A : Beh, sei veramente ingenuo ! ‘Well, you’re really naïve !’
(23) A : Credi che Giovanni sia andato a casa ? ‘Do you think that Giovanni has gone home ?’ B : Credo / penso di sì. ‘I think so’
Manfred Pinkal (1983) has offered a detailed account along Gricean lines of the way in which this second use of belief questions comes about. Here the speaker is asking an educated guess from the hearer, in a way which is reminiscent of our construction. Clearly, our construction is not compatible with the first interpretation of the belief question : a question about beliefs, as shown in (24.b).
(24a) Credi che si diverta ? Do you think he’s having fun ? > He’s not having fun at all. (24b) *Che si diverta ?
Can we think of ‘che+subjunctive’ as just a conventionalised elliptical version of the second reading of 2nd person belief questions ? In fact, the pragmatic effects associated with the second reading of the belief questions are not equivalent to those conventionally associated with the ‘che+subjunctive’ construction, which imposes tighter restrictions on the context of utterance. Let us compare the following two utterances :
140 Andrea Rocci
(25a) Credi che Giovanni sia andato a casa ? Io, non ne ho alcuna idea. ‘Do you think that Giovanni went home ? I have no idea / I doubt it.’ (25b) *? Che Giovanni sia andato a casa ? Io, non ne ho alcuna idea.
While the discourse in (25.a) is perfectly acceptable, in (25.b) the incompatibility of the the ‘che+subjunctive’ construction with a state of complete ignorance of the speaker renders the discourse somewhat incoherent. Another difference between our construction and questions with belief verbs in the 2nd person emerges if we examine the following discourses :
(26a) La macchina di Giovanni non è nel parcheggio. Che sia andato a casa ? (26b) La macchina di Giovanni non è nel parcheggio. Credi che sia andato a casa ? ‘Giovanni’s car is not in the parking lot. Do you believe / think that he went home ?’
While in (26.a), the ‘che+subjunctive’ construction makes us immediately establish an inferential discourse relation between the first utterance presenting the evidence and the second utterance, the discourse in (26.b) sounds rather clumsy and does not lend itself immediately to such an interpretation. The above evidence supports the hypothesis that these aspects of the interpretation of ‘che+subjunctive’ interrogatives are not conversational inferences and that they are conventionally associated with the construction and have to be represented in its semantics. If we now compare our construction with the uses of the epistemic future in questions we get an interesting picture of the similarities and the differences between the two constructions :
(27) Giovanni sarà andato a casa ? Io, non ne ho alcuna idea. Roughly : ‘I wonder if Giovanni has gone home. I have no idea, though.’
Surprisingly, the epistemic-inferential future, which in the assertive conveys a rather strong degree of probability (cf. Rocci 2000 and 2005a), is compatible with an epistemic state of complete ignorance on the part of the speaker in interrogative constructions. Another test setting apart the epistemic future from the ‘che+subjunctive’ construction is a variant of the one applied by Diller (1977) to the French inferential conditional. Diller remarked that the impossibility of adding ‘oui ou non’ (‘yes or no’) to the question was accounted for by the hypothesis that the conditional question presupposes “a positive orientation .”
Epistemic modality and questions in dialogue 141
(28a) *Est-ce qu’il ferait beau, oui ou non ? (Diller, 1977) ‘Is there possibly sunshine, yes or no ?’ (28b) Sarà andato a casa o no ? / *sì o no ? ‘Has he gone home, I wonder, or not ?’ (28c) *Che sia andato a casa o no ? / *sì o no ? ‘Has he possibly gone home, or not ?’
I cannot delve here in questions of intonation, but it is also interesting to remark that in Italian the intonation pattern required for questions ending with ‘o no ?’ is squarely incompatible with the fall –rise intonation that always accompanies our construction. As for the relationship with the evidence given in the previous context we have already seen that the epistemic future promptly establishes a discourse relation of inference both in assertive and in interrogative contexts :
(29) La macchina di Giovanni non è nel parcheggio. Sarà andato a casa ? Very roughly : John’s car is not in the parking lot. Has he possibly gone home ?
If the above observations are correct, ‘che+subjunctive’ ends up as having a series of properties, which are quite typical of the constructions singled out by Charles Fillmore and Paul Kay in their works on the interaction between grammar and pragmatics :
– It specifies a morphosyntactic arrangement and a precise intonation pattern, which are conventionally associated with a meaning which is not derivable compositionally from the constituents of the construction ; – Both at the formal and at the semantic level it inherits properties of more generic template constructions, such as the yes/no interrogative construction, and the subjunctive complement construction which however do not suffice to account for the whole form and meaning of the construction. The construction, although non-compositional is not completely arbitrary, as its similarity with the complements of belief verbs motivates in part the meaning of the construction 14 ;
14 On the use of the formal mechanism of inheritance to represent motivation in non fully compositional linguistic construction see for instance Goldberg (1995 : 72-81) : “Inheritance allows us to capture the fact that two constructions may be in some ways the same and in other way distinct”. Here, however, I do not pursue the formalization of the ‘che+subjunctive’ construction in a construction grammar of Italian based on inheritance. Many different formal implementations of the basic ideas of construction grammar approaches are possible.
142 Andrea Rocci
– The meaning associated with the construction is pragmatic and discursive, as it involves restrictions on the contexts of utterance, information that enables the establishment of discourse relations, and, perhaps, the conventional specification of a particular illocutionary force, which is distinct from that of basic yes/no questions.
Our construction offers a particularly clear illustration of the fact that, as Paul Kay (2003 : 676) puts it, “the interpretation of linguistic utterances can involve an interaction of grammar and context which vastly exceeds in complexity, formal structure and wealth of interpretive content the data discussed in the standard literature on indexicals”. We will see how the construction places restrictions on the common ground of the conversation (cf. Clark 1996). These restrictions are, in fact, a sort of very abstract, schematic, presuppositions – which function as instructions for the hearer to examine the common ground in order to flesh out the interpretation by specifying the illocutionary force of the utterance, establishing discourse and dialogue relations, etc.
4. Interaction of the construction with the tense-aspect system
It is interesting to observe that, although internally non-compositional, ‘che+subjunctive’ differs from other types of idiomatic units such as synthemes (cf. Martinet 19964), because it only specifies some grammatical material and is productive giving rise to a number of grammatically and lexically saturated sentences. Once we consider the construction as a unique sign, we can see that it combines compositionally with other grammatical and lexical signs. For instance, our construction admits all tenses of the subjunctive mood :
(30) (a) Che vada a casa ? (present subjunctive) (b) Che sia andato a casa ? (perfect subjunctive) (c) Che andasse a casa ? (imperfect subjunctive) (d) Che fosse andato a casa ? (pluperfect subjunctive)
The various subjunctive tenses shown in (30) encode differences at the level of temporal reference and aspect which interact with the inferential evidential component of the construction. All the inferences signalled are drawn at the moment of utterance, but the state of affairs which is hypothesized as a result can be a concomitant ongoing process (30.a), an accomplished event (or a present resulting state of a previous event) (30.b), a process seen as unachieved in the past (30.c), an accomplished event anterior to a reference point in the past. These different temporal and aspectual combinations lend themselves more naturally to different types of inference process. The perfect is easily
Epistemic modality and questions in dialogue 143 associated with inferring an event in the past from evidence available at the moment of utterance. Sometimes this coincides with an inference from effect to cause :
(31) La sua macchina non è nel parcheggio. Che sia andato a casa ? (perfect subjunctive) ‘His car is not in the parking lot. Has he possibly gone home ?’
While the present and, in particular, the imperfect associate more easily with evidence which is concomitant with the process inferred. This often coincides with inference from the perceptual appearances to some non directly perceivable fact :
(32) Giovanni era molto nervoso. Che avesse paura di qualcosa ? (imperfect subjunctive) ‘John was very nervous. Was he possibly afraid of something ?’
If we use the subjunctive imperfect in an inference from a piece of evidence available at the moment of speech to a past event, like in (33) we generate the additional presupposition of a point of observation in the past in which the event was perceived as an ongoing process :
(33) La sua macchina non è nel parcheggio. Che andasse a casa ? (imperfect subjunctive) ‘His car is not in the parking lot. Was he possibly going home ?’
In (33) it is presupposed that there was a moment in the past, in which the speaker or hearer, or both, witnessed a segment of the action of going home, arguably without recognizing it as such (e.g. they saw him going somewhere, but, at the moment, they did not realize that he was going home). We can see this presupposition as descending from the internal reference point (R ⊆ E) encoded by the basic semantics of the imperfect (Saussure 2003 : 242-243). Similar observations can be made on other compositional interactions between the semantics of the construction and the tense-aspect system.
5. Pragmatic predicates in Congruity Theory
In the final section I provide a semantics for the ‘che+subjunctive’ construction specifying the constraints it imposes on the epistemic attitudes of the speaker and the hearer, on the common ground of the dialogue, and on the nature of the speech act performed. In order to capture this semantics I will use the notion of a pragmatic predicate (or connective predicate) from Congruity Theory (Rigotti & Rocci 2001).
144 Andrea Rocci
An extended notion of predicate plays a kernel role in Congruity Theory : a predicate is conceived ontologically as a possible mode of being, a general notion that subsumes more specific distinctions such as those between properties and relations, states and events, actions and non-actions. All predicates are analysed in terms of the presuppositions they impose on their argument places (or n-tuples of argument places) and of their semantic entailments (cf. also Seuren 1988). For instance the two-place lexical predicate to read (x1, x2) can be analysed roughly as follows :
to read (x1, x2 ) = [presuppositions : x1 exists, x1 is a human being, x1is literate, x2 exists, x2 is a written text⏐entailments : x1 reconstructs the phonetic structure and retrieves the message of x2] For the predicate to be congruous with respect to its arguments, the semantic traits associated to the actual arguments have to be both compatible with and more informative than (that is to say hyponymous of) the traits featured in the presuppositions associated to the argument place (cf. Rigotti & Rocci 2001 for a more developed discussion). When a predicate appears as the rheme of a speech-act of the assertive type, its presuppositions have to be satisfied (or accommodated) 15 in the common ground of the dialogue, while its entailments are added to the common ground. For this reason we can also call them update conditions. In Congruity Theory the pragmatic and discourse level of meaning is addressed through a notion of pragmatic predicate, which make it possible to treat discourse and dialogue coherence in terms of predicate-argument congruity. The notion of a pragmatic predicate addresses the areas of meaning covered by notions such as illocutionary act, or dialogue act, as well as those which are usually treated in terms of discourse relations rhetorical relation, or rhetorical predicates. While the conflation of these two kind of notions might seem surprising from the viewpoints of traditional speech act theory and studies of the semantic coherence of monologue, it appears quite a natural step to take in view of the analysis of dialogue, and is increasingly adopted by researchers in the field (cf. for instance Asher & Lascarides 2003 : 305) Simply put : a pragmatic predicate is an action predicate saying what the speaker does to the addressee with the utterance. Like ordinary predicates, pragmatic predicates are analysed in term of the presuppositions they impose on their argument frame and of the conditions with which they update the common ground. Since pragmatic predicates not only represent but also realize social actions, their update conditions are pragmatic effects, involving the creation of commitments of the participants which become part of the
15 For the notion of accommodation see Lewis (1979).
Epistemic modality and questions in dialogue 145 common ground 16. A pragmatic predicate is a relational predicate which minimally takes at its arguments the speaker (Spk), the hearer (Hr) and an 17 utterance (U0) :
P (Spk, Hr, U0)
Since many pragmatic predicates (such as answer, object, conclude, etc.) are relational also in the sense that they are logically dependent from the content of another utterance the argument frame of the pragmatic predicate can also include other anaphorically recovered utterances (U-n) or inferred contextual propositions (X) as arguments :
P (Spk, Hr, X, U-n, U0)
In this framework the felicity conditions imposed by Searlian illocutions (Searle 1969) – which typically involve the speaker and the hearer – are reinterpreted either as presuppositions imposed by the pragmatic predicate on n-tuples of argument places including for instance the speaker (Spk), the 18 hearer (Hr) and the utterance U0 or as pragmatic effects of the predicate . Rhetorical relations defined, as in Mann & Thompson (1987) in terms of conditions on the utterances involved, conditions on their combination, and effects of the combination are even more straightforwardly rendered in terms of presuppositions and effects of the pragmatic predicate. Often pragmatic predicates receive little or no linguistic manifestation, and hearer have to infer them, in order to make sense of an utterance or discourse. There are however a number of linguistic items whose business is to impose quite detailed constraints on the pragmatic predicates to be established. We can say that these items conventionally express pragmatic predicate templates, which the hearer has to flesh out in the dialogue context (cf. also Ginzburg, Sag & Purver 2003 on this point).
16 On the ontology and the dynamics of commitment in dialogue see Colombetti, Fornara & Verdicchio (2003). 17 Note that the argument is identified with “the utterance” and not directly with the propositional content of the utterance in order to account for the pragmatic predicates associated with utterances such as Hi! which do not have a semantic propositional content proper, but only a pragmatic frame (e.g. Spk greets Hr by uttering U0). On the treatment these proposition-less speech acts see also (Ginzburg, Sag & Purver 2003) 18 Consider, for instance, the preparatory conditions of the assertion as formulated by Searle (1969). One of them states that 'It is not obvious to the Speaker that the Hearer knows (does not need to be reminded of) p'. This type of condition is treated as a (relational) presupposition imposed by any 'assertive' connective predicate on the argument places characterized by the roles of the Speaker, the Hearer and the asserted proposition.
146 Andrea Rocci
6. A non-propositional semantics for the ‘che+subjunctive ?’ construction
In this final section I provide a full specification of the semantics of the ‘che+subjunctive’construction in terms of a pragmatic predicate template in a way which accounts for the phenomena we have examined in section 3. I will first present the full semantic analysis, I will then proceed to justify the various components in view of the data examined in the previous sections. The pragmatic predicate template conventionally associated with ‘che+subjunctive’ questions specifies the following argument places : the speaker (Spk), the hearer (Hr), a contextually available set of propositions E, the utterance U0 of the ‘che+subjunctive’ sentence, and a projected utterance U+1, to which the predicate refers with a sort of cataphora. The presuppositions that the predicate imposes on these argument places, and the pragmatic effects that are added to the common ground are listed below.
!"#$%&'(&)"*+,$-. . /
. 0 1
. ∉ 23¬∉ 2 ¬ 2 $ 3
¬
6.1. Justification of the preconditions (a) and (b)
The precondition (a)
(a) p ∉ CG ; ¬ p ∉ CG, and CG can be updated by either p or ¬ p is a basic requirement of sincere yes/no questions as it ensures that the answer is not already part of the common ground CG, and that its presuppositions are not known to be false in CG. For instance the answer to the question in (34),
(34) Are you still living in Geneva ?
Epistemic modality and questions in dialogue 147 cannot update a CG where it is known that Hr never lived in Geneva. It is well known that uttering ‘rhetorical’ questions, whose answer is known to be false or true in the CG is a common way to present a premise in an argumentation in a particularly emotional, ‘interactive’ way 19 conveying the implicature that the hearer is urged to take into account that p or ¬ p are in the CG and not to overlook their relevance in view of a certain conclusion q, as in the famous French example in (35) (cf. Anscombre & Ducrot 1983 : 128) :
(35) Est-ce que je te l’ai rendu ton livre ? Bon alors laisse-moi tranquille ! ‘Did I give you back your book ? Well then, leave me alone !’
‘Che+subjunctive’ interrogatives cannot be used as ‘rhetorical’ invitations to the speaker to consider the presupposed truth or falsity of a proposition, and precondition (a) has also the task of accounting for this impossibility.
(b) There is no sign that either Spk or Hr knows whether p or ¬ p
Presupposition (b) is different and stronger than the Searle’s (1969) preparatory condition of questions that says that Hr is not going to tell the answer as a matter of course. This condition amounts to a limiting condition on the epistemic attitudes of both Spk and Hr formulated in terms of CG. This precondition greatly deviates, in its second part from the semantics of prototypical sincere questions. In questions the Hr is often expected to know the answer, but ‘Che+subjunctive’ cannot be used in situation where Hr is supposed to know the answer. Consider the weirdness of (36.a) and (36.b)
(36a) *?Che tu sappia l’ora ? ‘Do you possibly know the time ?’ (36b) ?*Che tu abbia già mangiato ? ‘Have you possibly already eaten ?’
6.2. The presuppositional status of the ‘modal base’ E
Let us examine the precondition (c) which plays a crucial role in the semantics of our construction :
(c) There is some set of facts E in CG, so that E represents non conclusive evidence in favour of p
19 On the emotional and interactive value of the “appeal to the listener to respond” see Gobber (1999 : 93-100), who also cites Quintilan’s remark that a rhetorical question makes argumentation “sharper and more forceful” (“acrior ac vehementior fit probatio” Quintiliani, Institutio Oratoria, IX, 2, 6).
148 Andrea Rocci
According to this condition, it is presupposed that there is a set of propositions E which represents the evidence in favour of p available to the participants in the dialogue. With respect to this analysis one legitimate question arises : why is it necessary to suppose that the set of evidence E is shared by both participants, and hence is part of the common ground ? Couldn’t it be that the speaker forms his/her positive attitude towards the proposition on the basis of evidence which is available only to him/herself ? This is a general question which concerns not only the semantics of this construction, but, in general, the semantics of epistemic modals and inferential evidentials. In the semantics of the construction this set of propositions plays a role similar to the role played by the modal base or modal conversational background (cf. Kratzer 1981) in the semantics of epistemic modals such as Italian dovere or English must. In Kratzer's approach, necessity modals are taken to indicate that the argument proposition is necessarily entailed by (that is logically follows from) a set of propositions, called modal conversational background 20 :
Must / Necessarily (CB, p) ⇔ (CB → p)
In the epistemic interpretation the CB provides the evidence for inferring the proposition p. In formal semantics, and in particular in dynamic approaches, such as DRT, various continuators of Kratzer’s approach have maintained that roughly the CB in the epistemic interpretation is to identify with a subset of the common ground CG 21. Other linguistic analyses of epistemic modals such as English must (Papafragou 2001) and Italian dovere (Rocci 2005a) postulate that the CB is to identify deictically with the beliefs of the speaker regarded as such, rather than with the common ground. 4 " 6 # 4 2 8890:8
20 Kratzer (1981 : 43) defines the semantics of relative modal operators directly in terms of the possible world semantics of the relations of logical consequence and logical compatibility. Here the semantics is given by translation into expressions where the absolute modal operator appears, for which I assume a standard unrestricted possible worlds semantics. So logical consequence is expressed in terms of absolute necessity () of a material conditional (→). 21 See Roberts (1989 : 687) : “the common ground will play the role of the propositions given by Kratzer’s modal base”. See also Portner (2003 : 480) : “with an epistemic modal [...] we use a different conversational background, one representing the set of propositions mutually accepted by the speaker and the hearer).”
Epistemic modality and questions in dialogue 149 ; <=
(37) Ann must be at Heathrow by now. % 4 # ? 4 & 2 8890 :8 A B B 0
(38) Che io abbia dimenticato le chiavi al ristorante ? Very roughly : ‘Have I possibly left my keys at the restaurant ? C # # 0
(39) I must have forgotten my keys at the restaurant. So, I’m going back to check. 2 # 4 150 Andrea Rocci " # " # 6.3. The pragmatic effects If we now examine the pragmatic effects (d) and (e), we find that (d) Spk puts forth p, as a cautious hypothesis in view of E (which entails the expression of a weak positive belief attitude of towards p) # " B # A (e) Spk proposes to Hr to contribute with a utterance U+1 to determine whether p or ¬ p by considering all the evidence available to her/him (including E). ! EE 88: # " B # B " # # B E< B G B " B # 2 8880=86 D B 6 B " # " B # <:6 H.0 (40) Sei così gentile stamattina. Che tu debba chiedermi qualche favore ? ‘You are so kind this morning ! Are you possibly going to ask me some favor ?’ 22 On the use of the notion of pre-commitment in the analysis of directives see Colombetti, Fornara & Verdicchio (2003 : 79 ff.). 23 The notion of uptake of a speech act originates from Austin (1962). Epistemic modality and questions in dialogue 151 H. " H. B G " B # 7. Concluding remarks " " B # 6 6 " B # # References Asher, N. ; Lascarides, A. (2003). Logics of conversation. Cambridge : Cambridge University Press. Anscombre, J.-C. ; Ducrot, O. (1983). L’argumentation dans la langue. Bruxelles : Mardaga. Bertinetto, P.M. (1979). Alcune ipotesi sul nostro futuro (con osservazioni su potere e dovere), Rivista di Grammatica Generativa 4(1-2) : 77-138. Colombetti, M. ; Fornara, N. ; Verdicchio, M (2003). Speech acts in artificial agent communication, in : A. Giacalone Ramat ; E. Rigotti ; A. Rocci, (eds), Linguistica e nuove professioni, Milano : Franco Angeli, 70-92. Diller, A.-M. (1977). Le conditionnel, marqueur de dérivation illocutoire, Semantikos 2(1) : 1-17. Doherty, M. (1987). Epistemic Meaning, Berlin : Springer. Faller, M.T. (2002). Semantics and pragmatics of evidentials in Cuzco Quechua, PhD Thesis, Stanford : Stanford University. 152 Andrea Rocci Fava, E. (1995). Il tipo interrogative, in : L. Renzi ; G. Salvi ; A. Cardinaletti, (eds), Grande grammatica italiana di consultazione, vol. III, Bologna : il Mulino. Fillmore, C. J. ; Kay, P. ; O'Connor, M.C. (1988). Regularity and idiomaticity in grammatical constructions : the case of “let alone”, Language 64 : 501-538. Ginzburg, J. ; Sag, I.A. ; Purver, M. (2003). Integrating conversational move types in the grammar of conversation, in : P. Kühnlein ; H. Rieser ; H. Zeevat, (eds), Perspectives on dialogue in the New Millennium, Amsterdam/Philadelphia : Benjamins, 75-96. Gobber, G. (1999). Pragmatica delle frasi interrogative, Milano : Pubblicazioni I.S.U. Università Cattolica. Goldberg, A. (1995). Constructions. A Construction Grammar Approach to Argument Structure, Chicago/London : University of Chicago Press. Groefsema, M. (1995). “Can, may, must and should” : a relevance theoretic account, Journal of Linguistics 31 : 53-79. Hooper, J. B. (1975). On assertive predicates, in : J. Kimball, (ed.), Syntax and Semantics 4, New York : Academic Press, 91-124. Kay, P. (2003). Pragmatic aspects of grammatical constructions, in : L. Horn ; G. Ward, (eds), Handbook of Pragmatics. Oxford : Blackwell. Kratzer, A. (1981) The Notional Category of Modality in : H.J. Eikmeyer ; H. Rieser, (eds), Words, Worlds and Contexts, De Gruyter : Berlin, 38-74. Lakoff, R. (1968). Abstract syntax and Latin complementation, Cambridge : MIT Press Lambrecht, K. (1994). Information structure and sentence form. Topic, focus and the mental representation of discourse referents, Cambridge : C.U.P. Lewis, D. (1979). Scorekeeping in a language game, Journal of Philosophical Language, 8 : 339-359 (reprinted in S. Davis, (ed.) Pragmatics : a Reader, Oxford : Oxford U.P., 1991 : 416-427). Mann, W.C. ; Thompson, S.A. (1987). Rhetorical Structure Theory : a theory of text organization, Technical Report RS-87-190 : USC Information Science Institute. Martinet, A. (1996). Éléments de linguistique générale, Paris : Armand Colin. Nuyts, J. (2000). Epistemic Modality, Language and Conceptualization, Amsterdam : John Benjamins. Papafragou, A. (2001). Modality : Issues in the Semantics-Pragmatics interface, Amsterdam : Elsevier. Pinkal, M. (1983). Questions of believing, in : F. Kiefer, (ed.), Questions and Answers, Dordrecht : Reidel, 241-256. Portner, P. (1997). The semantics of mood, complementation and conversational force, Natural Language Semantics 5 : 167-212. Epistemic modality and questions in dialogue 153 Portner, P. (2003). The (temporal) semantics and (modal) pragmatics of the perfect, Linguistics and Philosophy 26 : 459-510. Rigotti, E. (2005). Congruity theory and argumentation, to appear in : M. Dascal ; F.H. van Eemeren ; E. Rigotti ; S. Stati ; A. Rocci, (eds), Argumentation in Dialogic Interaction. Special Issue of Studies in Communication Sciences, 75-96. Rigotti, E. ; Rocci, A. (2001). “Sens — non-sens — contresens”. Studies in Communication Sciences 1 : 45-80. Rigotti, E. ; Rocci, A. (2006). Le signe linguistique comme structure intermédiaire, in : L. de Saussure (ed.), Nouveaux regards sur Saussure. Mélanges offerts à René Amacker, Publications du Cercle Ferdinand de Saussure, Genève : Droz. Roberts, C. (1989). Modal Subordination and Pronominal Anaphora in Discourse, Linguistics and Philosophy 12 : 683-721. Rocci, A. (2000). L'interprétation épistémique du futur en italien et en français : une analyse procédurale, in : J. Moeschler (ed.), Inférences directionnelles, représentations mentales et subjectivité, Cahiers de Linguistique Française 22 : 241-274. Rocci, A. (2005a). La modalità epistemica tra semantica e argomentazione, Milano : Pubblicazioni I.S.U.-Università Cattolica. Rocci, A. (2005b). Epistemic Readings of Modal Verbs in Italian : the relationship between propositionality, theme-rheme articulation an inferential discourse relations, in : B. Hollebrandse ; A. van Hout ; C. Vet, (eds), Crosslinguistic Views on Tense, Aspect and Modality, Amsterdam/New York : Rodopi. 229-246. Rocci, A. (2005c). Connective predicates in dialogic and monologic argumentation, in : M. Dascal ; F.H. van Eemeren ; E. Rigotti ; S. Stati ; A. Rocci, (eds), Argumentation in Dialogic Interaction. Special Issue of Studies in Communication Sciences. 97-118. Saussure, L. de (2003). Temps et pertinence, Bruxelles : de Boeck-Duculot Schneider, S. (1999). Il congiuntivo tra modalità e subordinazione. Uno studiosull’italiano parlato, Roma : Carocci. Searle, J.R. (1969). Speech Acts. An Essay in the Philosophy of Language, London : Cambridge University Press. Searle, J.R. & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge : Cambridge University Press. Urmson, J. O. (1969). Parenthetical verbs, in A. Flew (ed.), Essays in Conceptual Analysis, London : Macmillan. Venier F. (1991). La modalizzazione assertiva, Milano : Franco Angeli. This page intentionally left blank In the mood of desire and hope : remarks on the German subjunctive, the verb second phenomenon, the nature of volitional predicates, and speculations on illocution André MEINUNGER ZAS Berlin 1. About predicates that disallow V2 - Do they form a natural class ? 1.1. Predicate classes – the big divide : V2 licensors and V2 blockers Much work has been spent on constructions in which a predicate or whatever linguistic entity licences a clause that behaves similarly to a matrix-sentence. Only to mention a few examples for the sake of illustration, Hooper and Thompson (1973), in response to Emonds (1969), list predicates that are able to embed CPs that exhibit phenomena which are expected to occur in main clauses only. For German or Germanic in general, there is an ongoing debate concerning embedded or dependent verb second (V2) constructions. Almost all researchers that have worked on dependent V2 in German present a classification of predicates that potentially allows for main clause word order in argument (realizing) clauses (‘Argument realisierende Sätze’ in Reis’ notation). Classifications can be found in Reis (1977, 1997) ; Helbig and Kempter (1974), Butulussi (1991), Romberg (1999), and to a lesser degree in Dunbar (1979), Oppenrieder (1987, 1991), and Meinunger (2004, 2006), and even standard grammars like Eisenberg (1994) or Duden (e.g., 1995) illustrate the facts and list predicates. The proposed classifications differ rather minimally and are listed below. Class (v) will be the main object of investigation in the present study, and its placement among the V2 licensors will be critically discussed. Class (i) verbs of saying : sagen, antworten, behaupten, bemerken, berichten,... (say, reply/respond, claim, remark, report,...) Class (ii) evidential verbs : hören, merken, spüren, bemerken, sehen, auffallen, ... (learn/hear, notice, feel, realize, see, strike…) Class (iii) verbs of thinking : annehmen, denken, einsehen, fürchten, glauben, meinen,... (assume, think, see, be afraid, think, believe, mean...) Class (iv) semifactive verbs : wissen, begreifen, beweisen, herausfinden, herausbekommen... (know, realize, prove, find out (both)...) © Cahiers Chronos 17 (2007) : 155-176. 156 André Meinunger Class (v) volitional predicates : wollen, wünschen, hoffen 1, empfehlen, überreden, das beste/besser/lieber sein, lieber haben (hätte...), vorziehen, bitten, verlangen... (want, wish, hope, recommend, convince, be better, prefer, ask, demand/require…) The predicates that do not allow for embedded or dependent V2 have been of much less interest. The reason seems obvious : the non-V2 realization, i.e., a verb-final construction introduced by a complementizer is the canonical case for an embedded clause, and insofar no extra story has to be told about the unmarked use. To my knowledge, a notable exception is Romberg (1999) inspired by classifications in Reis (1977 : 202). Romberg lists ‘Berücksichtigungs¬prädikate’ (predicates of consideration) = class (a), semantically complex, inherently negative predicates (b) and emotive verbs (c). Class (a) vernachlässigen, ignorieren, bedenken, beachten... (neglect, ignore, consider, bear in mind...) Class (b) verdrängen, vergessen, verheimlichen... (repress/suppress, forget, hide/conceal...) Class (c) bedauern, bereuen, übelnehmen, beklagen,... (regret/feel remorse, take offense, deplore...) (1) Ich bereue, dass ich es nicht sofort gekauft habe. I regret that I it not immediately bought have (2) *Ich bereue, ich habe es nicht sofort gekauft. I regret I have it not immediately bought both : ‘I regret that I did not buy it right away.’ One can easily see that the given classification is rather tentative than completely satisfying. Cross-classification, hence ambiguity, is inevitable. Class (a) and (b) are not very distinct, the first two verbs of class (a) could as well be argued to belong into class (b) 2. Furthermore, Romberg mentions causative verbs and gives the following example (Romberg 1999 : 25) : (3) Hans hat verursacht/bewirkt, dass Peter nach Hause geht. Hans has tried /caused, that Peter to home goes 1 This predicate (i.e. ‘hoffen’) is very delicate; see the discussion in paragraph 3. 2 Few people, e.g., Eisenberg (1994) find semi-factives like herausfinden, entdecken, beweisen etc. (find out, discover, prove) and claimed-to-be factive ‚wissen’ relatively unacceptable and argue for an unintegrated reading with the so-called ‘Doppelpunktlesart’. In the mood of desire and hope 157 (4) *Hans hat verursacht/bewirkt, Peter geht nach Hause. Hans has tried /caused, Peter goes to home both : ‘Hans caused Peter to go home.’ Romberg cites these sentences to show that only verbs that potentially report a propositional attitude can realize their complements in a V2 pattern. Meinunger (2004), considering work by Quer (1998, 2001) argues that Romberg’s observation can be stretched to cover more than only pure causatives. It seems that generally implicative predicates do not allow for independent, integrated V2 clauses (causatives being only weakly implicative). Note also that most of Reis’ negative predicates are considered to be negatively implicative (see Bußmann (1990), also the pioneer study Karttunen (1971), after Kiparsky and Kiparsky (1970)). Furthermore Meinunger (2004, 2006) argues that volitional predicates (volitive predicates, preferential expressions, desideratives, Reis’ so-called ‘Präferenzprädikate’ etc.) are misclassified if they are claimed to belong to the V2-licensors (as for V2 and volitional predicates see also the work of Frank (1998)). Nevertheless all researchers have classified volitional predicates as V2 licensors. However, examples like (5) vs. (6) show that under normal circumstances, volitional verbs do not allow for V2 complements. For details and more examples see below. (5) Hans will, dass du ihm sein Hemd mitbringst. Hans wants that you him his shirt with-bring (6) *Hans will, du bringst ihm sein Hemd mit. Hans wants you bring him his shirt with(=prt) ‘Hans wants you to bring (along) his shirt.’ 1.2. Valency and discourse properties This concludes our detour concerning verb classes. Another important observation is that the option of V2 also depends on factors different from just the nature of the sole verb (al predicate). Pinkal (1981) and Vogel (1998) observe that 3-place ‘glauben’ is also not construable with a V2-argument. (7) Hans glaubt, Peter geht nach Hause. Hans believes, Peter goes to home. (8) *Hans glaubt seinem Bruder, Peter geht nach Hause Hans believes his brother, Peter goes to home. ‘Hans believes (his brother) that Peter is going home.’ 158 André Meinunger Both authors offer a similar explanation according to which the 3-place variant implicates a discourse-old reading for the dependent clause. I.e., they both argue for a reading where the embedded proposition is known to hearer and speaker. 1.3. More factors to render V2 impossible Other linguistic triggers for the inapplicability of V2 in dependent clauses other than the nature of the matrix predicates have also been discussed in literature. There is the famous negation restriction (already Blümel (1914), see however Butulussi (1991) and Meinunger (2004)), i.e., (9) and (11) vs. (10). (9) Er hat gesagt / geglaubt, sie ist schwanger. He has said / believed she is pregnant (10) *Er hat nicht gesagt/ geglaubt, sie ist schwanger. He has not said / believed she is pregnant (11) Er hat nicht gesagt/ geglaubt, dass sie schwanger ist. He has not said / believed that she pregnant is ‘He did not say /believe that she was/is pregnant’ Apart from negation other focus (sensitive) operators like nur, lediglich, auch (only, just, too…) and the like render V2 close to impossible (Romberg 1999, for an overview). Less clear are the facts for a non-assertive mood in the matrix, but the observed tendency is that V2 becomes worse if the root clause is a question or a command (Meinunger 2004, 2006). Last but not least it has been observed that old information, i.e. if the proposition expressed in a CP is known or can easily be inferred from what is known, V2 is not appropriate either (Mikame (1986), Meinunger (2006)). Meinunger states : “A further important observation is that prominence in discourse renders V2 awkward albeit under a licensing predicate.” (Meinunger 2006 : 465) That means that if (the proposition of) a V2 utterance is to be repeated, the subordinated shape sounds much more appropriate (12Ba) vs. (12Bb). (12) A : Bernd ist endlich gekommen ! ‘Bernd has arrived – finally.’ B : (a) Ja, ja – ich weiß/habe schon gehört, dass Bernd endlich gekommen ist. (b) # Ja, ja – ich weiß/habe schon gehört, Bernd ist endlich gekommen. ‘O yeah, I know/have heard #(dass) Bernd has finally arrived.’ But not only if the sentence is repeated with the same lexical material – even if the propositional content can be inferred (13), V2 sounds inappropriate. In the mood of desire and hope 159 (13) A : Bernd ist endlich gekommen ! ‘Bernd has arrived – finally.’ B : (a) Ja, ja – ich weiß/habe schon gehört, dass Bernd hier ist. (b) # Ja, ja – ich weiß/habe schon gehört, Bernd ist hier. ‘O yeah, I know/have heard #(dass) Bernd is here.’ This also holds if an obvious fact is being uttered. The scenario is such that a speaker enters a room where a specific person he had been looking for is present. Then speaker makes a more natural statement if he utters (a) instead of (b). (14) Hans hat gesagt, dass du hier bist. Hans has said that you here are (15) Hans hat gesagt, du bist hier. Hans has said you are here Both : ‘Hans said you’re here.’ Thus the conclusion is that discourse-linked propositions (whether explicitly introduced into the discourse by a speaker’s statement, or by mere accommodation) cannot be uttered in the shape of V2. Note that this observation reminds a lot of the condition on the licensing of definite anaphoric noun phrases. So much for the non-lexical, i.e. grammatical restrictions. 1.4. Back to V2 blocking predicates At this point, I want to come back to the categorical V2 blockers. In 1.2 and 1.3, I discussed the contexts in which potential V2 licensors cannot embed a V2 clause. It seems that some form of givenness of the subordinate proposition disallows the V2 realization. Now, reconsidering again the verbal classes, the following question arises. Is there something common to the relevant predicates ? To come to a proposal, let me present the predicate classes in a single list : (16) (i) Emotive verbs (ii) Predicates of consideration (iii) Inherently negative predicates (iv) Implicative verbs (including causatives, i.e. weakly implicative verbs) (v) Volitional predicates My claim will be that all of these predicate classes make reference to factivity. (i) Emotive verbs are the prototypical representatives of factive 160 André Meinunger verbs. However, even they display one intriguing factor. It is an old observation that emotive verbs can be coerced to act as verba dicendi, i.e. they can be used as verbs of saying. For a brand new treatment of this phenomenon in German see Fabricius-Hansen and Sæbø (2004). The authors show that with Konjunktiv 1 (a special German form of subjunctive), many of these verbs lose their factive character and acquire a reading in which the embedded proposition is the object of an utterance report (=reportive subjunctive). (17) Der Chef bedauerte, dass er ein Drittel der Belegschaft entlassen müsse. The boss regretted that he a third the employees fire must-subj (18) Der Chef bedauerte, er müsse ein Drittel der Belegschaft entlassen. The boss regretted he must-subj one third the employees fire ‘The boss said in a sad way that he was forced to dismiss a third of the employees.’ This use, however, must be disregarded here. This is not so difficult since Konjunktiv I (a special German subjunctive) signals this non-factive use. And indeed the non-indicative verbal mood is obligatory in this construction 3. Furthermore, Reis (1977) shows convincingly that true factives can also be found among the non-emotives and lists the German examples from above (next to some English predicates), which she differentiates into consideratives (ii) and negatives (iii) 4. However – just to mention it : it seems to me that these (cognitives) are still somewhat less factive than emotive, they appear to be marginally possible with if or ‘ob’ for that matter - a use which does not trigger a factivity presupposition, and which is not available to emotives. (19) Ich habe vergessen, ob er nach München fährt. I have forgotten, if he to Munich goes. ‘I forgot whether he is going to Munich.’ (20) Du musst bedenken, ob du das wirklich willst. You must think-about if you that truely want ‘You have to think if you really want it.’ Implicative verbs (iv) are also relatively easily classified as factives in the broad sense. Whereas factives are defined as predicates that presuppose the 3 There are rare cases of Konjunktiv II under emotive verbs that are due to tense phenomena. It might again complicate matters if spoken registers are taken into consideration, so this should not concern us here. 4 The English terminology is my own. In the mood of desire and hope 161 truth of the argument proposition regardless of negation, implicatives are predicates that trigger a less rigid validity of the argument proposition. These predicates trigger (a claim about) the truth of the embedded proposition only if the relevant predicate is used positively, and they mark its falsity (i.e., non- truth) if the predicate is used under negation. Such an example would be ‘zustande bringen’ (to manage, to bring about), ‘sich die Zeit nehmen’ (to take one’s time). Weakly implicative verbs comprise causatives. These verbs only trigger a presuppositional reading in the positive use. Romberg’s examples belong in this category ; see (3) and (4) above. 2. The counter-factivity of volitionals 2.1. Counter-factive predicates This leaves us with volitional predicates (v). These verbal and adjectival predicates can hardly be argued to be factive in the canonical sense. However, there seems to be something to them which transforms them into factives. In order to see what this is exactly, we have to undertake one more detour. In some work on factives, one finds vague mentioning of counter- or anti-factivity triggered by the use of predicates (e.g. Manning 1995). The English examples that are normally given – if at all – are pretend and wish (or sometimes imagine, which I think is not a good candidate). The German counterpart to pretend is ‘vorgeben’, and indeed a sentence like (21) seems to carry the presupposition that the negation of the complement proposition is true, i.e. if (22) is true (see Meibauer 1999). (21) Egon gibt vor, dass seine Frau Nastassja ist. Egon pretends prt that his wife Nastassja is. ‘Egon pretends Nastassja to be his wife.’ (22) Nastassja ist nicht seine Frau. ‘Nastassja is not his wife.’ However, if there were a complete mirror image in the behaviour of the presupposition, we would expect the falsity of the complement proposition also under negation. This is arguably not the case. From a negated sentence containing vorgeben, nothing about the validity of the argument proposition can be concluded. (23) Egon gibt nicht vor / hat nie vorgegeben, dass seine Frau eine Adlige war. Egon gives not prt / has never pretended, that his wife a nobility was ‘Egon never pretended / does not pretend his wife to belong to the nobility.’ In this respect, the verb vorgeben is rather something like a mirror image of a semi-factive or a weak implicative. There is yet one more curiosity found 162 André Meinunger with this verb. It seems that it is possible to construe a grammatical sentence with it (i.e. Vorgeben), which can embed a verb second clause. In this case, however, the embedded verb must show subjunctive mood. (24) Egon gibt/gab vor, seine zukünftige Frau sei/wäre Millionärin. Egon pretends/pretended his future woman be-subj millionaire. ‘Egon pretended/pretends that his future wife is a millionaire.’ With indicative mood the complex sentence sounds rather bizarre. (25) ?/*Egon gibt/gab vor, seine zukünftige Frau ist/war Millionärin Egon pretends/pretended his future woman be-ind millionaire. ‘Egon pretended/pretends that his future wife is/was a millionaire.’ 2.2. Volitional predicates With respect to the verbal mood in the subordinate, ‘pretend’ behaves similarly to the other counter-factive verb, ‘wish’. This verb is (mainly) used for counter-factual wishes with the subordinate sentence surfacing in subjunctive mood. (26) I wish you were here. Or less idiomatic : (27) They wish you had spent more energy on this…. ‘Wish’ sounds slightly marked with a canonical subordinate sentence exhibiting a complementizer (26) and ungrammatical with indicative mood (27). (28) ?I wish that you came once more. (29) *I wish (that) you are taller 5. Despite the sentences in footnote 5, under regular circumstances wish cannot be negated (30), (31). (30) ??/*I don’t wish that you be tall. (31) ??/*They do not wish that he come/came once more. 5 Under specific conditions (negation of the very predicate in a dialogue) the ban on indicative mood under that can be dispensed with. (i) Do you really wish to get the job? (ii) Well… I do not really wish that I get the job (…but…) In the mood of desire and hope 163 In these respects it resembles German constructions with ‘wünschte’. This verb is morphologically defective in the sense that it seems to have no regular indicative present (Präsens Indikativ), and depending on the point of view it could be argued that it does not have an infinitive either. It is also difficult to imagine past and true future tense(s). Be it as it may, sentences like (32) are relatively easy to be found. (32) Ich wünschte, du hättest mehr Zeit für mich. I wish you had-subj more time for me ‘I wish you had more time for me.’ (33) Seine Eltern wünschten sehr wohl, er hätte sie nie kennen gelernt. His parents wished very well he has-subj her never know learned ‘Certainly, his parents wish(ed) he had never met her.’ The phenomena that we observed with ‘wish’ also occur here. That is, the proposition expressed in the complement clause can be inferred to be false, i.e. it seems the whole sentence presupposes that [you have more time for me] or [he never met her] does not hold in the actual world with (30) and (31) respectively. In German, even more than with English wish, the verb ‘wünschte’ cannot be negated (34). (34) *Ich wünschte nicht, sie könnte/würde sich dabei verletzen. I wish not she could/would herself thereby hurt ‘I don’t wish that she hurt herself doing this.’ The subordination shape with ‘dass’ is also slightly marked, though fully grammatical – but only on the condition that the verb carries subjunctive mood (35) vs. (36), (37). Subordination under ‘wenn’ (=if) is completely out also (38). (35) (?)Ich wünschte, dass du mehr Zeit für mich hättest. I wished that you more time for me have-subj ‘I wish you had more time for me.’ (36) *Ich wünschte, du hast mehr Zeit für mich I wished you have-ind more time for me (37) *Ich wünschte, dass du mehr Zeit für mich hast I wished that you more time for me have-ind (38) *Ich wünschte, wenn du mehr Zeit für mich hast I wished if you more time for me have-ind Other volitional predicates show a behaviour which overlaps with ‘wünschte’ only partly. The first division concerning volitional predicates I 164 André Meinunger would like to make at this time is also one that is inspired by Reis (1997). Firstly, there are the rather regular volitives like mögen, (möchten ?), wünschen, sich wünschen, wollen, bitten, fordern (the English counterparts I give here are tentative translation proposals, the relevant verbs behave differently and have to be modified, however : like, want, (wish ?), ask, beg, demand). Among them ‘möchte’ – a word used quite frequently - is a similarly defective case as ‘wünschte’. The other examples are rather regular, and as I have shown, they never allow for V2, see (5) vs. (6) above (and also Helbig and Kempter 1974 6). Secondly we have the so-called ‘Präferenzausdrücke’ – preferential expressions like vorziehen, lieber haben / mögen, besser sein, günstiger sein etc (prefer, be better, be more appropriate...). These predicates show an interesting behaviour (see also Meinunger 2004). Similarly to ‘wünschte’, the subordinate sentence is mostly (though not necessarily in this case) in subjunctive mood. But this is not enough : for the relevant constructions to be grammatical the matrix itself must occur in ‘Konjunktiv’, i.e. subjunctive mood, or, if the predicate is an adjective, the adjectival predicate must not be simple positive, but appear in comparative or superlative form. (39) Es ist/wäre besser, du gehst nicht hin. It is/were better you go-ind not there Almost all below something like : ‘It would be better for you not to go (there).’ (40) Es ist/wäre besser, du gingest nicht hin / du würdest nicht hin gehen. It is/were better you go-subj not there / you would not there go (41) Es ist/wäre das beste, du gehst nicht hin. It is/were the best you go-ind not there (42) Es ist/wäre das beste, du gingest nicht hin / du würdest nicht hin gehen. It is/were the best you go-sub not there / you would not there go (43) ??Es wäre gut, du gehst nicht hin. It is/were good you go-ind not there (44) Es wäre gut, du gingest nicht hin / du würdest nicht hin gehen. It is/were good you go-sub not there /It is/were good you go-ind not there 6 The only possibility to get a relatively integrated V2 clause under a volitional verb like bitten, fordern, verlangen (ask, demand, require) etc. is to construe it with ‘sollen’/’mögen’ : (i) Peter bittet, du mögest an seine Tasche denken. Peter asks, you should of his bag think ‘Peter is asking you not to forget his bag.’ In the mood of desire and hope 165 (45) *Es ist gut, du gehst nicht hin. It is good you go-ind not there (46) *Es ist gut, du gingest nicht hin / du würdest nicht hin gehen. It is good you go-ind not there /It is/were good you not there go-ind Furthermore, the corresponding verb final variant of the embedded clause must be introduced by the complementizer ‘wenn’, not by ‘dass’ – except for the realization with the simple positive. (47) Es ist/wäre besser, wenn du nicht hingehst. It is/were better if you not there- go-ind Almost all below something like : ‘It would be better if you don’t go there.’ (48) Es ist/wäre besser, wenn du nicht hingingst / hingehen würdest. It is/were better if you not there- go-subj / go-would (49) Es ist/wäre das beste, wenn du nicht hingehst. It is/were the best if you not there- go-ind (50) Es ist/wäre das beste, wenn du nicht hingingst / hingehen würdest. It is/were the best if you not there- go-subj / go-would (51) Es wäre gut, wenn du nicht hingehst. It were good if you not there- go-ind (52) Es wäre gut, wenn du nicht hingingst / hingehen würdest. It were good if you not there- go-subj / go-would (53) *Es wäre besser, dass du nicht hingehst 7. It were better that you not there- go-ind (54) *Es ist/wäre besser, dass du nicht hingingst / hingehen würdest. It is/were better that you not there- go-subj / go-would (55) ?/*Es ist/wäre das beste, dass du nicht hingehst. It is/were the best that you not there- go-ind (56) *Es ist/wäre das beste, dass du nicht hingingst / hingehen würdest. It is/were the best that you not there- go-subj / go-would (57) *Es wäre gut, dass du nicht hingehst. It were good that you not there- go-ind 7 ‘Es ist besser, dass du nicht hingehst’ – i.e. the non-subjunctive variant is fully grammatical, but here we have indicative in the subordinate as well and the reading is factive and means something different. It means ‘It is better that you are not going there’, thus we have a different interpretation here – a factive one – thus one that is crucially distinct from the V2 variant – and from all the other as well, of course. 166 André Meinunger (58) *Es wäre gut, dass du nicht hingingst / hingehen würdest. It were good that you not there- go-subj / go-would All this shows clearly that these verbs are to be kept apart from the other V2 licensors. 2.3. The nature of volitional predicates The claim that I would like to put forward is that volitional predicates are factive in some sense (i.e. counter-factive). It seems possible to assume (as one reads occasionally in the literature) that wish and ‘wünschte’ are really counter-factive in the sense that they presuppose the falsity of the complement proposition. The same seems to hold for the somewhat quirky preferential predicates. Here even the negation test seems to be applicable to some degree ; which means that negation leaves the subordinate proposition unaffected, i.e. the presupposition of the embedded proposition’s falsity holds. (59) Es wäre besser, sie trüge ihr Haar offen / sie würde ihr Haar offen tragen. It were better, she wear-sub her hair open / she would her hair open wear (60) Es wäre besser, wenn sie ihr Haar offen trägt / trüge / tragen würde. It were better, if she her hair open wear (all forms, i.e., ind & sub) ‘It would be better she wore her hair down.’ Interestingly also : (61) Es wäre besser, trüge sie ihr Haar offen / sie würde ihr Haar offen tragen. It were better, wear-sub she her hair open From (59), (60) and (61) one can infer : (62) She wears her hair down. i.e. not open The same (i.e. (62)) is the presupposition under negation. (63) Es wäre nicht besser, wenn sie ihr Haar offen trägt / trüge / tragen würde. (64) Es wäre nicht besser, trüge sie ihr Haar offen / sie würde ihr Haar offen tragen. (65) ??/* Es wäre nicht besser, sie trüge ihr Haar offen / sie würde ihr Haar offen tragen (the same presupposition as for (59) to (61) – although just under negation) The near-ungrammaticality of (65) seems interesting. V1 (64) and ‘wenn’ in combination with verb-final order (64) is fine. V2 is not (65). However, V2 gets much better if embedded in a question : In the mood of desire and hope 167 (66) Wäre es nicht besser, sie trüge ihr Haar offen ? However, the negation here is not a regular one. It is not used to negate inside and in combination with a question. Something similar is to be observed in connection with negation and canonical V2 licensors. (Here we are dealing with so-called meta-linguistic negation.) To recapitulate : Verbs of saying allow for V2 in complement clauses, negation usually takes this option away, see above (9) – (11). However, when putting quirky (non- canonical) negation inside a non-assertive sentence, things become possible all of a sudden : (67) Hat er nicht gesagt, sie ist schwanger ? Has he not said she is pregnant ‘Didn’t he say she was pregnant ?’ (68) Glaube ja nicht, sie ist schwanger ! Believe prt not, she is pregnant ‘Don’t be so stupid and think she’s pregnant.’ Be it as it may, in most cases the counter-factuality of the embedded proposition can be argued to be presupposed. If a present tense volitional verb is used eventively and the embedded clause is a finite CP which is not the projection of a stative or habitual predicate, then it can be concluded that the resultative or change component of the predicate has not been reached yet, or does not hold (yet). (69) Ich wünsche (mir), dass du in den Garten gehst. I wish me that you in the garden go ‘I wish you to go to the garden.’ (70) Ich will, dass er ein Auto kauft. I want that he a car buy ‘I want him to buy a car.’ Similarly, if a non-present tense is used, it can (always) be concluded that if there is an event at all to which reference is possible, the event time must be after the event time of the matrix. This means again that at a relevant point the proposition described by the embedded clause must not be true or is presupposed not to hold. (71) Ich wollte, dass er ein Auto kauft. I wanted that he a car buy ‘I wanted him to buy a car.’ 168 André Meinunger (72) Sie wollte, dass der Vogel stirbt. She wanted that the bird die ‘She wanted the bird to die.’ (73) Hans bat Maria, dass sie ihm eine Goldmünze mitbringt. Hans asked Maria that she him a gold -coin with-bring ‘Hans asked Maria to bring him a gold coin.’ Thus, the implicature of (69) is that the addressee is not in the garden at speech time, also for (71) at the time of wish-holding, the referent of ‘er’ did not have a car. These data, however, should not lead one to conclude that volitionals are robustly anti- or counter-factive. There are uses from which nothing about the validity of the embedded proposition can be concluded. This is so if the volitional predicate is to be understood as stative, i.e. the volition is carried over a long period, almost like predicates of an individual level. (74) Viele Männer wollen, dass ihre Frauen arbeiten. Many men want that their wifes work ‘Many men want their wives to have a job.’ (75) Hans wünscht sich, dass seine Frau ihr Haar offen trägt. Hans wishes himself that his wife her hair open wears ‘Hans wants his wife to wear her hair down.’ (76) Eine Mutter möchte, dass ihr Kind glücklich ist. A mother wants that her child happy is ‘A true mother wants her children to be happy.’ This use of volitional predicates remains mysterious. It is definitely an interesting topic for future research. Uli Sauerland (personal communication) suggests that there might be a hidden counter-factuality nevertheless. The wish-holders in (74) to (76) all seem to consider the possibility that in some accessible, but non-preferred world the embedded proposition does not hold (throughout). Be it as it may, putting aside this (non-eventive) use illustrated in (74) – (76) – which is available to a subgroup of volitional predicates of the first class only – volitional predicates can be considered to be counter-factive. Considering the observations about stative verbs of wanting, the claim from Meinunger (2006 : 471) can still be made for most volitive constructions : (77) Volitional predicates in a broad sense are anti-factive (or counter-factive). Similar to counterfactual constructions, they refer to eventualities that are not given. In the mood of desire and hope 169 The claim thus is that true volitional predicates presuppose the non-givenness of the proposition contained in their complement clause. Thus these predicates trigger the implicature that the proposition in the complement clause does not hold (in the actual world at the utterance or reference time). The relatedness of anti- or counter-factivity seems to be supported by the complementizer choice (if-like C°) and non-indicative verbal mood, see also Adger and Quer (2001). 3. ‘Hoffen’ as special case within the special case The authors who care about volitional predicates (especially Helbig and Kempter (1974), but also Reis (1977, 1997)) do not only classify them as V2 licensors, they also seem to be little interested in the differences that they show class-internally. However, there are crucial differences. An interesting predicate for further research seems to be the respective linguistic variants of the verb ‘to hope’. The German corresponding verb ‘hoffen’ is - traditionally and unsurprisingly – listed as V2 licensor among the volitional predicates 8. However, having analyzed these predicates as V2 blockers, I am urged to say something about this verb. As a matter of fact, ‘hoffen’ is a good V2 embedder (78). (78) Ich hoffe, du schaffst es. I hope you get it ‘I hope you’ll manage to do it.’ In all its uses, this verb imposes no commitment of the speaker to the (non-) validity of the complement proposition. For this reason, ‘hoffen’ is not future orientated as canonical volitional predicates are claimed to be, and hence in contrast to volitionals, hoping can be directed toward the past. (79) Ich hoffe, du hast es geschafft / du warst pünktlich. I hope you have it gotten / you were punctual. ‘I hope you did it / were on time.’ This is impossible with all the other canonical volitional predicates 9 : 8 Similar things must be said about ‘fürchten’ (be afraid, fear) – which is some sort of not-hoping. 9 The only thing which can be observed is some sort of aspectual anteriority in the peculiar stative use (see above). (i) Zahnärzte wünschen, dass sich ihre Patienten die Zähne geputzt haben (bevor sie zu ihnen kommen) Dentists wish that (refl) their patients the teeth cleaned have (before they come to them) ‘Dentists prefer for their patients to arrive with brushed teeth.’ 170 André Meinunger (80) *Ich will/ wünsche, du hast es geschafft / du warst pünktlich. I want/ wish you have it gotten / you were punctual. ‘I hope you did it / were on time.’ In earlier work of mine (i.e. Meinunger 2004, 2006), I have compared the verb second phenomenon in German(ic) with the verbal mood selection in the Romance languages. The claim I make is given here in (81) (Meinunger 2004 : 323; 2006 : 467) : (81) Correspondence alignment : Those predicates and grammatical phenomena that block V2 in German(ic) subordinate clauses trigger subjunctive mood in Romance. This is correct for the predication for French ‘espérer’. German ‘hoffen’ allows for V2, thus ‘espérer’ should not select for subjunctive, and indeed it selects for indicative 10 11. (82) J’espère que tu es /*sois satisfait. (French) I hope that you be-ind / *be-subj satisfied ‘I hope you are satisfied.’ The other canonical volitional predicates like (bien) vouloir, désirer, préférer, demander, exiger etc. (want/wish, desire, prefer, demand/ask, (ii) Manche Schulen fordern, dass die Erstklässler im Kindergarten waren. Some schools require that the first-class-pupils in-the kindergarten were ‘Some schools want to accept only children that have been to a kindergarten.’ However, there is no sequence of tense observable in the sense that the embedded event time precedes the matrix event time, the reason being that in this use, there is no (matrix) event. The anteriority of the embedded CP exhibiting past tense is between the embedded temporal reference and another silent event, which is realized in (i) within brackets. The wanting or wishing itself is not directed toward the past. These sentences show, however, that there is no formal requirement : no +past in the scope of volitional predicates. 10 The specific behavior of ‘espérer’ in connection with mood selection is also discussed in Schlenker (2004). His approach, although considering presupposition issues, is different however. 11 Interestingly, the French counterpart to ‘fürchten’ - which licenses V2 without any problems as well as ‘hoffen’ – i.e., ‘craindre’ is more delicate. It rather selects for subjunctive. The reason is unclear. On the one hand it contains a negative semem in it (meaning not-hope) and hence it is expected to pattern like Reis’ inherently negatives (class b). On the other hand – according to the correspondence alignememnt in (82), it is expected to patterns like its positive counterpart ‘espérer’ as well as German ‘fürchten’. Closer scrutiny will have to explain this behavior. In the mood of desire and hope 171 require etc.) obligatorily select for subjunctive. However, most Romance languages do not only allow for, but strongly favour subjunctive under their respective verb for ‘to hope’. The crucial difference is only that indicative is possible under certain conditions, which is not so under other volitionals 12. Future research shall bring interesting results. Preliminary inquiries and descriptive characterizations in traditional grammars and text books seem to point into the direction that the mood choice under the ‘hope’-predicate goes together with a different expectation as to the likeliness of the validity of the proposition. The suspicion is that the verb indeed carries different attitudes (towards a potential factivity of the embedded proposition). In the Slavic languages, like Russian for example, volitional verbs require a specific complementizer. This C-element is a complex formative that consists of the regular C-element ‘čto’ (=that) and a particle that is found in the formation of irrealis or subjunctive mood ‘by’ – resulting in ‘čtoby’, which must co-occur with past morphology on the verb. Thus, the presence of this specific complementizer is related to the use of subjunctive mood under volitional predicates in Romance. Considering this, Russian (and Slavic) in general behaves more according to the expectations. The complementizer that the verb for ‘to hope’ (=nadevat’sya) selects for is not the one that all the other verbs of wanting and demanding subcategorize for (i.e., ‘čtoby’ in Russian), but it is the neutral C-element ‘čto’. (83) Ya nadeyus’ čto on spit / vyspal. (Russian) I hope indicative-C° he sleps / slept (84) *Ya nadeyus ‘čtoby’ on spit / spal / vyspal. I hope subjunctive-C° he sleps / slept Both : ‘I hope that he’s sleeping / he slept.’ A closer look at this predicate might also reveal why ‘hope’ does not allow for neg-raising in English. (85) I hope (that) he won’t come. -/-> 12 Farkas (1992) (also) discusses the mood choice under factive predicates in Romance. This fact is completely ignored here, if not even challenged or neglected. However, this point in the discussion shall be used to refer to Farkas’ excellent work on the matter, i.e., indicative vs. subjunctive selection in Romance – also in subsequent work of hers. Furthermore my speculation is that there are subtle differences in the meaning of the respective language-specific verb for ‘hope’. A quite comparable case seems to me to be the difference between English ‘to know’ and German ‘wissen’ (‘ignorance reading’, see Reis 1977 : 142). These lexical entries slightly differ in their semantics, which has important impact on their factivity implication. 172 André Meinunger (86) ?/*I don’t hope that he will come. This is different from German. Hopefully future research will bring (some) clarification. 4. Conclusions and Speculations Now, what is the impact of V2 ? In earlier works, there was the proposal that some version of assertion(ality) or assertivity plays the crucial role (e.g. see above or Wechsler for V2 in Swedish (1991), and in connection with the correlation expressed in the correspondence alignment (81), compare Panzeri (2003) for Romance). To a certain degree and for a subset of the cases I adopted this view for my QR-analysis (e.g., Meinunger 2004). Reis (1997) and Gärtner (2001a, b) also make reference to the notion of assertion, but they explicitly refrain from the standard notion as illocutionary force and speak of ‘vermittelte Assertion’ (something like conveyed assertion) and ‘proto-assertion’ respectively. However, there are many occurrences of dependent V2 clauses where an assertive speech act is hard or even impossible to argue for. The most convincing examples for non-assertive use are those with ‘Konjunktiv’ (subjunctive mood) : (87) Ihm wäre lieber, du würdest mit dem Rauchen aufhören. Him were dearer, you would with the smoking stop ‘He’d prefer if you quit smoking.’ (88) (Du bist hier.) Ich dachte, du wärst diese Woche in München. (you are here) I though you were this week in Munich. ‘(So you’re here.) I thought you’re in Munich this week.’ But also with the indicative, these sentences do not convey a statement to whose truth the speaker would be committed. Similar things hold for non- assertive sentences like (89) and (90). (89) Glaubst du, er hat das Auto gekauft 13 ? Think you, he has the car bought ? ‘Do you think he bought the car ?’ 13 The embedded V2 clause can pronounced with raising intonation. In this case, the clause could and would not be a canonical assertion anyway. Then it either expresses ‘force identity’ or ‘illocutionary agreement’ with the matrix – or in the spirit of Gunlogsen (2003) or Asher (2005) it expresses some sort of hearer commitment – at any rate definitely no speaker assertion/commitment. In the mood of desire and hope 173 (90) Sag bloß, er hat das Auto gekauft ! Say just he has the car bought ‘Don’t tell me he bought the car !’ And similarly inside several forms of conditional sentences (91) - (94), where V2 clauses are to be found systematically (see Gärtner and Schwager, in preparation) : (91) Wenn du dann nach hause kommst und der Gerichtsvollzieher steht vor der Tür... if you then to home come and the marshal stands before the door ‘If you come home with the marshal standing in front of your door…’ (92) Kommt der heute abend, (dann) gehe ich. Comes him today evening then go I ‘Should he show up tonight, then I’ll leave….’ (93) Ich gehe sofort wieder, sehe ich, dass der auch nur ein Glas Wein trinkt. I go right-away again, see I, that he also only one glass wine drinks ‘Even if I see that he has one glass of wine, I will leave right away.’ (94) Du trinkst noch ein Bier und ich gehe. You drink yet one beer and I go ‘You have one more beer and I go…’ In such cases, no statement is made about the truth of the antecedent (or about the consequent in isolation). Rather there is an implicature that the state of affairs described in the clause does not hold (in the actual world at the utterance time) – at least in the given sentences. The tentative proposal thus is that something weaker than assertion is of concern. However, I want to maintain that the use of V2 has to do with an attitude of the speaker and not of a third individual (usually expressed as the subject of the matrix clause). In the many articles and books that have appeared after the classical writings of Austin and Searl, some researchers tried to refine and newly define the five to six canonical speech acts or illocutions (such as assertives, directives, commissives, expressives / expositives, declaratives, narratives). Some of those propose hypothesizing or supposing/speculating. Something like this might run in the mind of a theoretician like McGilfrey (1991). One can imagine that a speaker uttering a declarative sentence does not always make a statement about the (actual) world, thereby claiming the truth of what he is saying and committing himself to this. Often the speaker’s intension is much less strong. He may present a case and invite the hearer to accept this just for a given context. Thus the use of a verbum dicendi with a third person subject followed by a 174 André Meinunger sentence that exhibits main clause features is similar to what McGilfrey calls ‘mock saying’. Here, the speakers gives away something about the commitment comparable to what is going on when using evidential modifiers like ‘according to the news’, ‘as per’…, laut meiner Mutter, Berichten (ausländischer Beobachter) zufolge (‘as my mom says’, ‘according to reports of foreign observers’). Still, the speaker (himself) offers the following proposition and takes responsibility for it (to a certain degree). I will go on claiming that the role of V2 is to introduce new information (see Meinunger 2006). This claim seems pretty indisputable for V2 in adjunct clauses ; see the important work on V2 relatives by Gärtner (2001a, b). Although the facts seem less clear for argument - especially for complement sentences - I keep arguing that V2 is impossible with discourse-old propositions. Thus, in order to capture the different uses of V2 mentioned in this paper, a possible term would be to offer or to dispose a proposition. Ben Shaer (p.c.) proposes the term ‘to entertain’ a proposition. He is drawing on work of his own (Shaer 1996) and is inspired by work of McGilvray, who claims that in certain constructions – especially in indirect speech – sort-of-assertions are made by the speaker (see above ‘mock saying’). At any rate, a speaker who uses present tense under a past verb of saying somehow expresses his own point of view concerning the embedded proposition. Shaer is also aware of Banfield’s work (1982). Banfield construes acceptable complex sentences like (94) which prove that the speaker might not commit himself completely and once and for all times to a mock-utterance. (95) John said that his roommate has green eyes, but I know that they are blue. Thus it seems to me that a weaker notion of assertion is needed to capture a main clause phenomenon like V2 in German. References Adger, D. ; Quer, J. (2001). The syntax and semantics of unselected embedded questions, Language 79(4). Asher, N. (2005). What's going on with final rises ? talk delivered at ZAS research seminar. Banfield, A. (1982). Unspeakable Sentences, London : Routledge & Kegan Paul. Blümel, R. (1914). Einführung in die Syntax, Heidelberg : Impresum. Bußmann, H. (1990). Lexikon der Sprachwissenschaft. 2nd ed. Stuttgart : Kröner. Butulussi, E. (1991). Studien zur Valenz kognitiver Verben im Deutschen und Neugriechischen, Linguistische Arbeiten 262. In the mood of desire and hope 175 Duden [Band 4] (1995). Grammatik der deutschen Gegenwartssprache. Mannheim / Leipzig / Vienna / Zurich : Dudenverlag. Dunbar, R.W. (1979). Discourse Pragmatics and Subordinate Clause Word Order in German : An Explanation of Related Main Clause Phenomena in German and English Clauses. Ph.D. dissertation, Ann Arbor : University of Wisconsin-Madison. Eisenberg, P. (1994). Grundriß der deutschen Grammatik (2., überarb. u. erw. Auflage), Stuttgart : Metzler. Emonds, J. E. (1969). Root and Structure Preserving Transformations, Doctoral dissertation, MIT (unpublished ms., printed in 1970). Fabricius-Hansen, C. ; K. J. Sæbø (2004). In a meditative mood : The semantics of the German reportative, Natural Language Semantics 12 : 213-257. Farkas, D. (1992). On the semantics of subjunctive complements, in : P. Hirschbühler ; K. Koerner, (eds), Romance Languages and Modern Linguistic Theory. Amsterdam/ Philadelphia : Benjamins, 69-104. Frank, N. (1998). Präferenzprädikate und abhängige Verbzweitsätze. Arbeitspapiere des SFB 340, Linguistic Theory & the Foundations of Computational Linguistics 128. Gärtner, H. M. (2001a). Bound focus and assertionality, http://www2.hu- berlin.de/asg/blutner/dialog [Online paper only]. Gärtner, H. M. (2001b). Are there V2 Relative Clauses in German ? Journal of Comparative Germanic Linguistics 3 : 97-141. Gärtner, H. M. ; Schwager, M. (in preparation), Pseudo-Coordination Meets the Type/Force Square, Ms, Berlin&Frankfurt/M. Gunlogson, Ch. (2003). Rising and Falling Declaratives as Questions in English, Outstanding Dissertations in Linguistics New York : Routledge (Taylor and Francis). Helbig, G. ; Kempter, F. (1974). Die uneingeleiteten Nebensätze im Deutschen und ihre Vermittlung im Fremdsprachenunterricht, DAF 11 : 75-86. Hooper, J. B. ; Thompson, S.A. (1973). On the Applicability of Root Transformations, Linguistic Inquiry 4 : 465-479. Karttunen, L. (1971) Implicative verbs. Language 47 : 340-358. Kiparsky, P. ; Kiparsky, C. (1971). Fact, in : D. Steinberg ; L. Jakobovitz, (eds), Semantics. Cambridge : CUP, 345-369. Manning, C. (1995). Presents embedded under pasts [Online paper], http:/nlp.Stanford.edu/~manning/papers/tense.ps McGilfrey, J. A. (1991). Tense, Reference, and Worldmaking, Montréal/Kingston/London/Buffalo : McGill–Queen’s University Press. Meibauer, J. (1999). Pragmatik. Eine Einführung, Tübingen : Stauffenburg Verlag. 176 André Meinunger Meinunger, A. (2004). Verb position, verbal mood and the anchoring (potential) of sentences, in : H. Lohnstein ; S. Trissler, (eds), Syntax and Semantics of the Left Periphery, Berlin : Mouton de Gruyter, 459-487. Meinunger, A. (2006). The discourse status of subordinate sentences and some implications for syntax and pragmatics, in : V. Molnár ; S. Winkler, (eds), The Architecture of Focus, (Studies in Generative Grammar 82.) Berlin/New York : Mouton de Gruyter, 459-488. Mikame, H. (1986). Die Einstellung des Sprechers zur Komplementsatzpropo sition und diesbezügliche syntaktische Phänomene bei Komplementsätzen mit‚ dass, Deutsche Sprache 14 : 232-337. Romberg, J. (1999). Magisterarbeit : Verbzweitstellung in Komplemen- tsätzen, Masters thesis, TU Berlin. Oppenrieder, W. (1987). Aussagesätze im Deutschen, Linguistische Arbeiten 180 : 161-189. Oppenrieder, W. (1991). Von Subjekten, Sätzen und Subjektsätzen, Tubingen : Niemeyer. Panzeri, F. (2003). In the (indicative) or subjunctive) mood, in : M. Weisgerber, (ed.), Proceedings of the Conference “sub7 – Sinn und Bedeutung”, Arbeitspapier Nr. 114, FB Sprachwissenschaft Universität Konstanz. Pinkal, M. (1981). Some Semantic and Pragmatic Properties of German glauben, in : H-J. Eikmeyer ; H. Rieser, (eds), Words, Worlds and Contexts, Berlin/New York : de Gruyter, 469-484. Quer, J. (1998). Mood at the Interface, LOT dissertation, Holland Academic Graphics. Quer, J. (2001). Interpreting mood, Probvs 13(1) : 81-112. Reis, M. (1977). Präsuppositionen und Syntax, Tubingen : Niemeyer. Reis, M. (1997). Zum syntaktischen Status unselbständiger Verbzweit-Sätze, Sprache im Fokus, 121-144. Schlenker, P. (2004). The Lazy (French)man's Approach to the Subjunctive (Reference to Worlds, Presuppositions and Semantics Defaults in the Analysis of Mood : Some Preliminary Remarks), ms., University of California at Los Angeles. Shaer, B. (1996). Making sense of tense : Tense, time reference, and linking theory, Ph.D. thesis, McGill, ms. Vogel, R. (1998). Polyvalent verbs, Dissertation, Berlin : University of Berlin : http://dochost.rz.hu-berlin.de/dissertationen/vogel-ralf-1998-07- 13/PDF/Vogel.pdf. Wechsler, S. (1991). Verb Second and Illocutionary Force, in K. Leffel ; D. Bouchard, (eds), Views on Phrase Structure, Dordrecht : Kluwer, 177- 191. Dutch equivalents of the German past conjunctive : zou + infinitive and the modal preterit Linde ROELS Tanja MORTELMANS Johan VAN DER AUWERA Antwerp University 1 1. Dutch and German compared Even though German and Dutch are closely related, on a micro-level they show a somewhat different behaviour as far as the conjunctive and mood in general are concerned. Whereas German discriminates between an indicative and a conjunctive mood, Dutch – at least from a morphological point of view – only seems to have the indicative mood at its disposal : Das finite Verbalsystem der festlandsskandinavischen Sprachen, des Englischen und des Niederländischen weist keine dem deutschen Konjunktiv vergleichbare Flexionskategorie auf. Es gibt zwar in variiendem [sic] Ausmaß sogenannte Konjunktivformen (‘subjunctive’), aber keine ausgebaute Flexionskategorie mit vergleichbarem paradigmatischem und funktionalem Stellenwert. The finite verbal system of the mainland Scandinavian languages, English and Dutch does not have an inflected category which resembles the conjunctive in German. Even though so-called conjunctive forms do appear to varying degrees, there is no completely elaborated inflectional category with a similar paradigmatic and functional status (Fabricius-Hansen 2004 : 132, translation ours) This has not always been the case, though : in Middle Dutch, a morphological conjunctive was still part of the verbal system (see table 1, from Van der Wal & Van Bree 1994 : 144). It appears, however, that the formal distinctions in Dutch were much smaller than in German, and in fact confined to the first (for the past conjunctive) and the third person singular (for the present and past conjunctive). It is therefore not surprising that this distinctive feature gradually got lost. 1 This study was carried out within the GOA-research project on Mood and Modality, which is supported by the University of Antwerp. © Cahiers Chronos 17 (2007) : 177-196. 178 Linde Roels, Tanja Mortelmans & Johan van der Auwera Present Indicative Present Conjunctive weak strong weak strong 1sg ic kere ic neme ic kere ic neme 2sg du keers du neems du keers du neems 3sg hi keert hi neemt hi kere hi neme 1pl wi keren wi nemen wi keren wi nemen 2pl ghi keert ghi neemt ghi keert ghi neemt 3pl si keren si nemen si keren si nemen Past Indicative Past Conjunctive weak strong weak strong 1sg ic keerde ic nam ic keerde ic name 2sg du keerdes du naems du keerdes du naems 3sg hi keerde hi nam hi keerde hi name 1pl wi keerden wi namen wi keerden wi namen 2pl ghi keerdet ghi naemt ghi keerdet ghi naemt 3pl si keerden si namen si keerden si namen Table 1 : Past indicative and subjunctive in Middle Dutch (Van der Wal & Van Bree 1994 : 144) In present-day Dutch, the original conjunctive is still recognizable in formulaic remnants (e.g. als het ware ‘if it were’, ware het niet ‘were it not’, lang leve de koning ‘long live the king’), but it is no longer a productive category. Dutch basically makes use of two devices to cover the meanings expressed by the German conjunctive : the analytic zou + infinitive construction (which, according to the ANS (1997 : 132), is the default way to signal irreality) and the so-called modal preterit The conjunctive paradigm in German hinges on the traditional distinction present vs. past, reflecting the fact that the present conjunctive is built on a present verb stem, whereas the past conjunctive is built on a past verb stem. In the realm of the conjunctive, the notions ‘present’ and ‘past’ do not express the temporal distinctions traditionally associated with the indicative. The past conjunctive in particular rather functions as a marker of epistemic distance. In order to avoid the confusion that the distinction between present and past may cause, one could prefer a distinction between ‘proximal’ and ‘distal’, based on notions of proximity and distance in the epistemic sphere (Langacker 1991 : 245) 2. In this paper we will mainly 2 Thieroff (1992) and Andersson (1989) describe the conjunctive paradigm in terms of distance as well. Dutch equivalents of the German past conjunctive 179 consider the ‘past’ or distal conjunctive, which will also be referred to as Konjunktiv II (KII). The German conjunctive still has a complete morphological paradigm, although in many of its forms it overlaps with the preterit indicative. This is particularly the case for weak verbs (see table 2 below). In fact, the past conjunctive is morphologically unambiguous for strong verbs only, and more particularly for those with a vowel that can take umlaut in the stem. Strong verbs which do not allow umlaut do not distinguish between past indicative and past conjunctive in the first and third person plural. Past Indicative Past Conjunctive weak strong weak strong 1sg ich liebte ich sang / blieb ich liebte ich sänge / bliebe 2sg du liebtest du sangst / bliebst du liebtest du sängest / bliebest 3sg er liebte er sang / blieb er liebte er sänge / bliebe 1pl wir liebten wir sangen / blieben wir liebten wir sängen / blieben 2pl ihr liebtet ihr sangt / bliebt ihr liebtet ihr sänget / bliebet 3pl sie liebten sie sangen / blieben sie liebten sie sängen / blieben Table 2 : the past conjunctive in German (Duden 1998 : 114, 122) In Dutch, however, a preterit form is ambiguous, so that we may need contextual clues to interpret it in a temporal or modal way. So, in (1a), for instance, the past tense forms studeerde en kreeg can be interpreted as referring to a repeated event in the past, but they also allow – albeit rather marginally – a conditional interpretation. (1) a. Als hij goed studeerde, kreeg hij een beloning. ‘Every time he studied well, he would get / got a reward (back then).’ ‘If he studied well, he would get a reward (now, later).’ Note that in German the preterit of weak verbs is modally ambiguous as well. (1) b. Wenn er gut studierte, kriegte er eine Belohnung. Both languages have developed different strategies to solve the formal ambiguity between past indicative and conjunctive forms. They both have an alternative analytic device to replace the original synthetic forms : for German this is würde + infinitive, for Dutch zou + infinitive. (1) c. Als hij goed zou studeren, zou hij een beloning krijgen. Wenn er gut studieren würde, würde er eine Belohnung kriegen. 180 Linde Roels, Tanja Mortelmans & Johan van der Auwera Both analytic devices combine elements of the past and elements that refer to the future. Dutch zou is the past tense of the future auxiliary zullen, whereas würde + infinitive is, in the first place, the past conjunctive of werden + inf., which serves to express futurity as well. On the basis of this formal build-up it is not surprising that both devices also serve to express future in the past (see section 2.2.3), comparable to the French conditionnel, which in its formal build-up also combines past and future marking. The motivations behind the development of analytic constructions in German and Dutch are quite similar. Their rise runs parallel with an increase in the number of conjunctive modal verb constructions, i.e. analytic constructions in which a modal (marked for the past conjunctive) combines with an infinitive. Synthetic conjunctives (of lexical verbs) were thus more and more pushed into the background (see Guchmann / Semenjuk 1980 : 230 and de Vooys 1967 : 155). At the same time, analytic forms serve to avoid the formal overlap between indicative and conjunctive synthetic forms that had come about after the formal attrition of the conjunctive paradigm, both in German and Dutch. The origin of the respective auxiliaries on the other hand is different. Zou has developed out of a form that in Middle Dutch was modally ambiguous, due to a formal overlap between the conjunctive and indicative paradigm. From the start it has fulfilled both a temporal and a modal function. German würde can only be characterized as a straightforward conjunctive (at least from a morphological point of view, see Thieroff 1992 for a different position on this issue). It is generally accepted that würde came into the verbal paradigm of the indicative to replace the combination of (past indicative) wurde + infinitive, which expressed future in the past. (Leiss 1992 : 223). Past conjunctive würde + infinitive took over this function, next to its inherent function as a conjunctive of werden + infinitive. In the following section, we will contrast the use of the German conjunctive (both in its synthetic and analytical form) with the Dutch modal preterit and zou-infinitive. We will give an overview of the contexts in which the Dutch ‘modal’ constructions appear, paying attention to their respective conditions of use. As will appear, the domains in which these devices figure in Dutch do not completely overlap with those in German. Our analysis is based on a self-compiled corpus of two novels (Margit Schreiner, Haus Frauen Sex and Margriet de Moor, De Virtuoos) and their respective translations in Dutch and German. In the German original, 206 past conjunctives were found and compared with their counterparts (mainly preterit and zou + infinitive) in the Dutch translation. For Dutch, we looked at the zou-infinitives, which we compared with the German translation. As a first observation, we find that in German past conjunctives (synthetic as well as würde-forms) are indeed frequently translated by a zou-form in Dutch (57%), thus confirming the traditional grammars’ view. The modal preterit, Dutch equivalents of the German past conjunctive 181 however, isn’t marginal at all : in absolute numbers it occurs even more often than the zou-construction (zou + (present / perfect) infinitive : 104 tokens vs. modal preterit / pluperfect : 110 tokens). This is particularly due to the high frequency of preterit forms to render pluperfect conjunctives in German. Another important finding is related to the relative frequency of würde : On the whole, we find more zou-forms in Dutch than würde-forms in German. This is particularly striking if we look at the pluperfect : here, würde is marginal (only one occurrence vs. 85 ‘synthetic’ forms), whereas Dutch zou is a less frequent alternative, but is not as marginal as würde (29 zou vs. 50 ‘synthetic’ forms). The distribution of the different types is given in table 3a for German and 3b for Dutch : German HFS DV Total synthetic past simple 45 34 79 conjunctives pluperfect 75 10 85 total 120 44 164 würde + inf simple 13 28 41 pluperfect 0 1 1 total 13 29 42 Total 133 73 206 Table 3a : Corpus distribution of German past conjunctives Dutch HVS DV Total past tense simple 43 17 60 forms pluperfect 40 10 50 total 83 27 110 zou + inf simple 31 44 75 pluperfect 17 12 29 total 48 56 104 Total 131 83 214 Table 3b : Corpus distribution of Dutch equivalents for the German past conjunctives 2. Semantics of the German conjunctive and its Dutch equivalents Following Fabricius-Hansen (2000 : 91), the use of the past conjunctive in German comprises two core domains : irrealis (in a wide sense) and indirectness. Typically, then, the conjunctive signals a certain disconnection 182 Linde Roels, Tanja Mortelmans & Johan van der Auwera or distance from (the representation of) immediate reality : the state of affairs at issue belongs to a mental space different from the conceptualizer’s own reality, either because it does not exist at the moment of speaking (i.e. it is non-factual) or because it is mediated (and is thus in some sense ‘removed’ from the speaker). 2.1. Non-factual uses 2.1.1. Hypotheticality – conditionality The prototypical function of the conjunctive comes to the fore in unreal conditional constructions, as in examples (2) and (3). Bouma (1973 : 102) considers the use of the KII in the conditional construction as “the most consistent and transparent”. Conditional constructions are typical examples in which the conjunctive can be analyzed as an indication that the SoA is to be interpreted in a different mental space. The protasis sets up “matching conditions” under which the apodosis may be true (Cutrer 1994 : 186-187). Together with the typical conditional structure (if...then) it separates the proposition from speaker reality and the factual domain. In (2), the speaker creates a hypothetical space in which he wants to compare his wage to his wife’s earnings, so that they could see the considerable difference between them. The protasis sets up a ‘possible world’ in which the apodosis will hold if the condition is met. In (3) the condition is not met : the son’s friend has never spent the night at their home. As a consequence, the SoA described in the apodosis cannot map onto speaker’s reality and must be interpreted as counterfactual : the son has never been able to have breakfast with his friend outside. (2) [...] wenn ich das dann dem gegenüberstellen würde, was du jetzt seit fünf oder sechs Jahren verdienst, dann würdest du sehen, dass es ein lächerlicher Betrag ist. (HFS 24) [...] en als ik dan ook nog […] ertegenover zou zetten wat jij in de afgelopen vijf of zes jaar hebt verdiend, dan zou je zien wat een bespottelijk bedrag dat is. (HVS 23) ‘And if I opposed this to the money you have earned during the past five or six years you would see what a ridiculous amount it is.’ (3) Und wenn dann einmal ein Freund bei ihm übernachtet hätte, dann hätten die beiden am nächsten Morgen draußen frühstücken können (HFS 17) […] en als er dan een keertje een vriendje was blijven slapen, hadden ze de volgende ochtend samen buiten kunnen ontbijten (HVS 16) ‘And if a friend had ever spent the night at his place, the two of them would have been able to have breakfast outside next morning.’ Dutch equivalents of the German past conjunctive 183 In this context, both the synthetic and analytic conjunctive occur without clear meaning differences 3. Originally, the würde form in this context mainly served to disambiguate modally ambiguous forms. Up till recently, double use of würde in conditional clauses was advised against, restricting the use of würde to the apodosis 4. In present-day German, however, there is a tendency to use würde forms irrespective of the clause they appear in ; würde even appears in counterfactual conditional clauses without future time reference. (4) Du bist Christin, hast aber Gebet, Messe und Sakramente vernachlässigt und du sagst : Ich habe alles ! Du lügst dich selber an ! Wenn du alles haben würdest, dann würdest du jetzt nicht in diesem Zustand sein. http://wallfahrtsort-medjugorje.de/pfarrer_slavko.htm ‘You are a Christian, but you have ignored prayer, mass and the sacraments and you say : I’ve got everything. You lie to yourself. If you had everything, you would not be in this condition now.’ When the würde construction is used as an alternative to the synthetic KII, this results in a less ambiguous (and therefore more efficient) marker of the conjunctive, especially for weak verbs (Fabricius-Hansen 2004 : 125). For the expression of irrealis and hypotheticality (especially in conditionals), Dutch primarily adopts the zou-form. It can occur in both protasis and apodosis (e.g. 5a). Just like for German würde, though, zou initially appeared in the main clause only (Overdiep 1928 : 191), with a modal preterit in the subclause, as in (5b) : (5) a. […] als ik dan zou uitrekenen hoeveel al die uren bij elkaar kosten, dan zou je zien wat een bespottelijk bedrag dat is. (HFS 23) ‘if I were to calculate how much all those hours cost in total, you would see what a laughable amount that is.’ b. Als dit een film was, zei Maria, zou de muziek voor dat dal gespeeld worden door een cello (CGN) ‘If this was/were a film, said Maria, the music for that valley would be played by a cello.’ 3 Cf. Duden (1998 : 804) [Im Konditionalgefüge]. “Für den einfach oder umschreibenden Konjunktiv II im Haupt- oder Nebensatz tritt ohne Bedeutungsunterschied auch würde + Infinitiv bzw. würde + Partizip + haben/sein auf, und zwar vor allem [...] in einer Ausdruckweise, die der gesprochenen Sprache nahe steht”. [[in conditionals] : the simple and periphrastic conjunctives II in main or subclauses can also be replaced by würde + infinitive or würde + Partizip + haben/sein, without change of meaning, particularly in uses that resemble spoken language] 4 Cf. Frantzen (1920 : 35) : “Die Einschränkung des Gebrauchs dieser Form auf den kond. Hauptsatz erklärt sich eben durch ihre noch deutlich empfundene futurische Bedeutung”. [The confinement of the use of this form to conditional main clauses can be ascribed to its clearly perceived future meaning] 184 Linde Roels, Tanja Mortelmans & Johan van der Auwera Many grammarians of Dutch claim that the zou-construction in conditionals is interchangeable with a modal preterit without substantial difference in meaning (e.g. Niewint 1984, Haeseryn, Romijn, Geerts, de Rooij & van den Toorn 1997). Searches in the Corpus Gesproken Nederlands (CGN), a standard corpus for spoken Dutch, 5 show that the most common patterns in Dutch are indeed the sequences with zou in both clauses or the combination of zou in the apodosis with a modal preterit in the protasis (table 4, ordered according to frequency of occurrence (in absolute numbers) : Protasis – Apodosis – Total Apodosis (als…, Protasis (…, als dan … ‘if…, … ‘]…, if’) then’) zou + inf – zou 132 66 198 zou – zou + inf. zou + perf. inf.. 13 3 16 – zou + perf. inf. Total 132 69 214 mod. pret – zou 70 4 74 mod. past. - +inf. zou mod. plupf. – 33 0 33 zou + perf. inf. Total 103 4 107 zou + inf. – 15 24 39 zou – mod. mod. Pret past. zou + perf. inf. – 2 14 16 mod. plupf.. Total 17 38 55 mod. pret. – 19 2 21 mod. past – mod. Pret mod. past mod. plupf. – 27 6 33 mod. plupf. Total 46 8 54 Table 4 : Distribution of zou vs. modal past in Dutch conditionals (CGN) Note that the most common pattern – except for pluperfects – is the combination zou-zou. The predominance of a double zou-structure suggests that zou has lost most of its original features in the course of time and has acquired a high degree of grammaticalization, at least in this syntactic environment. In perfect tenses, the appearance of zou is more limited. Out of 194 clauses with protasis and/or apodosis in the perfect tense, 81 clauses with zou 5 For more information, see http://lands.let.kun.nl/cgn/home.htm Dutch equivalents of the German past conjunctive 185 occur (ex. 6), as opposed to 113 with past perfect forms (ex. 7) ; see table 4 again. (6) [...] en als iemand hem gevraagd zou hebben of hij dan niet vond dat in het leven de liefde voor alles kwam, zou hij verwonderd hebben opgekeken. (DV 93) [...] und wenn ihn jemand gefragt hätte, ob er denn nicht der Ansicht sei, die Liebe sei das Allerwichtigste im Leben, dann hätte er nur verwundert geschaut. (DV 97) ‘and if someone had asked him whether he didn’t think that in life love comes before everything, he would have looked at him in astonishment.’ (7) Je zoon was vast en zeker veel beter geworden op school omdat hij dan had gevoeld dat hij iemand is (HVS 16) Dein Sohn wäre bestimmt besser geworden in der Schule, weil er gespürt hätte, dass er etwas ist und dass seine Eltern etwas sind. (HFS 17). ‘Your son would definitely have done much better at school because he would have felt that he is someone.’ Still, the fact that zou, originally a hypothetical / future marker, appears in counterfactual perfect constructions (without future reference at all) may be seen as an indication of high grammaticalization of the form. It is also remarkable that combinations with a modal preterit in both protasis and apodosis build a rather small group compared to conditionals with zou + inf. in protasis and/or apodosis. Possibly this is due to the potential ambiguity of the preterit in conditionals, which speakers may want avoid. Under (1a) we already indicated that preterit forms may often be ambiguous as to their temporal or modal interpretation. Without contextual information (1a), repeated for convenience’ sake as (8), does not easily get a conditional interpretation : (8) Als hij goed studeerde, kreeg hij een beloning. ‘If he studied well, he would get a reward (now, later)’ There may be a diachronic explanation for this development. Als originally was a temporal conjunction. We can assume that through a process of grammaticalization in Middle Dutch the temporal meaning got reinterpreted as a conditional one in particular contexts (De Vooys 1967 : 155), such that both functions nowadays co-exist (an instance of layering). The fact that als is by default interpreted in a temporal way may affect the interpretation of a following past form : it is likely that it will primarily be interpreted as temporally past as well. If the context, however, provides one or more elements like a modal adverb (misschien ‘perhaps’ in (9a)), an adverb with future time reference or a clearly counterfactual reasoning (9b), this may induce a conditional interpretation : 186 Linde Roels, Tanja Mortelmans & Johan van der Auwera (9) a. Ik had al lang het vermoeden dat je vreemdging [...] Als jij ziek was, deed ik misschien wel hetzelfde. (www.kluun.nl/komteenvrouwbijdedokter/images/kluun_deel_2.pdf) ‘I had my suspicions all along that you were having an affair. If you were ill I would maybe do the same.’ b. Als jij er niet was, wist ik niet eens hoe een dinosaurus eruit zag. (CGN) ‘If I didn’t have you, I wouldn’t even known what a dinosaur looked like.’ Note that in the category of double pasts, most instances (33 out of 54 occurrences) occur in combination with a perfect infinitive – which easily gets a counterfactual interpretation in conditionals (Iatridou 2000 : 240). (9) c. Maar als ik alles had geweten, dan had ik het niet gedaan. (CGN) ‘But if I had known everything I would not have done it.’ Apart from its function in situations in which an unreal condition is involved, the conjunctive and its equivalents may also appear in contexts in which other kinds of irrealis are involved, often with subtle distinctions of meaning. Conjunctive forms for instance figure in independent clauses and different types of dependent clauses, or they may be connected to specific sentence parts or nouns. These contexts will be discussed below. 2.1.2. Irrealis comparisons In irrealis comparisons in German both the synthetic (10a) and analytic (10b) KII occur. (10) a. [...] während du nur wieder dein Glänzen in den Augen gehabt hast, als ob das alles ein Wunder wäre und nicht von Anfang bis Ende durchdacht (HFS 8) [...] terwijl jouw ogen alleen maar weer begonnen te glanzen alsof het allemaal een wonder was, en niet van begin tot eind doordacht. (HVS 9) ‘while you only had that glow in your eyes again, as if everything was a miracle and not thought out from A to Z.’ b. Tu doch nicht so, als ob du so viel Geld mit dem bisschen Nähen verdienen würdest. (HFS 27) Doe nou toch niet net alsof je zoveel geld verdient [simple present] met dat beetje naaiwerk van je. (HVS 25) ‘Don’t pretend that you would earn so much money with that little bit of sewing.’ For Dutch, a striking observation is the relatively low frequency of zou in irrealis comparisons. Out of 15 als (ob)-clauses in our corpus, there are 14 past conjunctives in German and 1 present conjunctive, as opposed to 8 modal preterits, 5 present indicatives and only 1 zou-infinitive, e. g. : Dutch equivalents of the German past conjunctive 187 (11) Und ich selbst war mir auch fremd. So, als stünde ich gar nicht auf dem Stadtplatz und wartete auf die Sonnenfinsternis, sondern als säße ich in Wirklichkeit vor einer Leinwand und sähe einen Film (HFS 10) Alsof ik helemaal niet op het marktplein op de zonsverduistering stond te wachten maar alsof ik in werkelijkheid naar een groot wit doek keek en een film zag. (HVS 10) ‘And I was a stranger to myself. As if I wasn’t standing on the market place at all and waited for the solar eclipse, but as if I was actually sitting in front of a white screen watching a movie.’ This may be explained by the fact that the Dutch conjunction alsof by itself unambiguously triggers an irrealis interpretation of the following SoA, whereas Dutch als can also get a temporal reading (see discussion of example 8) .As a consequence, when alsof is followed by a preterit (or even a present tense), it gets a modal interpretation 6. In irreal comparisons the zou construction seems to signal an increased degree of distance. In 12(a), the only zou-example in the contrastive corpus, the zou form indicates that the speaker distances himself from the SoA presented in the clause of comparison. The interpretation of zou as a sign of clear distance is sustained by the use of the pluperfect conjunctive in German, which according to Leirbukt (1991 : 184f) underlines the counterfactual interpretation 7. (12) a. Alsof ik een huis van twee verdiepingen met zes kamers […] helemaal alleen voor mij zou bouwen. (HFS 43) Als ob ich das Haus für mich allein gebaut hätte (HVS 51) ‘As if I was going to build a house with two floors of six rooms all for myself’ Sometimes the use of zou seems to be related to the future time reference of the irrealis subclause as in (12b). The modal preterit is not possible there, as it would invoke a simultaneity reading : 6 Compare also some frequencies based on a small search in Google (date of access 12/05/05) als hij vond : 3 modal preterits out of a selection of 100 instances (N= 745) als hij zou vinden : 3 instances alsof hij vond : 11 instances alsof hij zou vinden : 0 instances als hij ging : 17 mod. pret. out of a selection of 100 (N= 545) als hij zou gaan : 161 instances alsof hij ging : 96 instances alsof hij zou gaan : 7 instances 7 According to Leirbukt (1991 : 184ff), the additional marking of the pluperfect conjunctive more clearly (than the simple KII or würde-construction) establishes a blocking factor that prevents a realis interpretation. 188 Linde Roels, Tanja Mortelmans & Johan van der Auwera (12) b. Heel even leek het alsof de kerk zou veranderen. De verwachtingen waren hoog gespannen. (www.vrouwenmaatschappij.be/opiniedetijd-gelezen%20 door % 20 eva.doc) * alsof de kerk veranderde ‘For a very short time it looked as if the church was going to change. Expectations were running high.’ 2.1.3. Contexts of negation The past conjunctive can also occur in contexts connected with negation in German. Typically, a negative element (often in a superordinate clause) invites the past conjunctive, which indicates that the state of affairs is not grounded in the reality of the speaker. In (13a) the subject niemand ‘no one’ opens up a counterfactual mental space in which the SoA is located. Dutch translations make use of either the modal pluperfect or zou. In (13b), for instance, the simple paste tense form zei (instead of zou zeggen) would not be able to trigger a reading as irrealis, which is in line with the connection between the use of a modal pluperfect and the expression of counterfactuality under 2.1.1 Note that zou + perfect infinitive is not impossible, but seems to be a less frequent alternative here 8. (13) a. Niemand hätte gedacht, dass das alles einmal so ausgeht (HFS 13) (‘Nobody had thought that it would all end like this’) Niemand had gedacht dat het met ons zo zou aflopen. (HVS 13) ‘No one would have thought that it would end like it did’ b. Wat betreft haar looks zou niemand zeggen dat ze 28 april al weer 31 wordt. (www.filmfashion.nl/news/391.html) ‘With respect to her looks nobody would say that she’s turning 31 on April 28.’ 2.1.4. Preludic conjunctive In situations of play, würde does not seem to serve as an alternative to the synthetic conjunctive in German. (14) Das wäre jetzt wohl unser Haus und du wärst die Mutter (Knobloch 2001 : 68) ‘This is our house and you are the mother’ 8 Google : ‘niemand zou hebben gedacht’ (4 hits) vs. ‘niemand had gedacht’ (414 hits); ‘niemand hätte gedacht’ (1210 hits) v. ‘niemand hatte gedacht’ (64 hits) (date of access 14/01/2005) Dutch equivalents of the German past conjunctive 189 For Dutch, zou + infinitive doesn’t occur either, leaving the expression of preludic meanings to the preterit only. (15) “Ik was de moeder en jij was de hond”, commandeerde het kleine meisje het buurjongetje. (www.omroepzwolle.nl/index.php) *zou zijn “‘I‘ll be [lit. ‘was’] the mother and you’ll be [lit. ‘were’] the dog,’ the little girl ordered the little boy next door.” The temporal reference of (15) is not to be situated in the past, but the (near) future or the present, the idea of which is implied in the situation of the game at hand. It is not completely clear how we should account for the impossibility to use either würde or zou in this context. When such sentences are uttered in children’s play, they usually serve to impose roles upon others. Warnant (1966 : 43) has shown for French that preludic conjunctives are typically used during the part of the game that deals with the division of the roles, later on children turn towards the indicative. Possibly, the default hypothetical-conditional meaning of würde and zou is too hypothetical for this kind of use. 2.1.5. Deliberative and rhetorical questions Questions can also be subsumed under the wide heading of ‘irrealis’. For German, the past conjunctive often occurs in deliberative questions, particularly the KII of the modal verb sollen, next to the synthetic conjunctive of main verbs (würde- periphrasis is rare here – but see example 16). Deliberative questions express doubt or surprise on the part of the speaker with respect to the action (s)he has to undertake or whether a particular SoA obtains or not. Typically, they do not expect an answer from the addressee. As could be expected, the modal preterit is rare in this kind of context ; the zou-form on the other hand is very frequent. (16) Hoe zou dit zorgvuldig opgekweekte talent het er afbrengen ? (DV 100) Wie würde dieses sorgfältig ausgebildete Talent sich schlagen ? (DV 105) ‘How would this carefully raised talented person perform ?’ Rhetorical questions do not so much express the speaker’s uncertainty but rather the assertion of his own point of view. (17) a. Waarom zouden we de toespelingen van dit spookuur niet laten voor wat ze zijn ? (DV 160) ?/* Waarom lieten we […] voor wat ze zijn ? Warum sollten wir die Fingerzeige dieser Geisterstunde nicht hinnehmen ? (DV 167) 190 Linde Roels, Tanja Mortelmans & Johan van der Auwera ‘Why [would we] not leave the coincidences of this witching hour for what they are ?’ Why did we not leave […] for what they are ?’) b. Waarom zouden we ons neerleggen bij de misère ? (DV 81) Warum sollten wir uns mit dem Elend abfinden ? (DV 85) ‘Why [should we] indulge in the misery ?’ In (17b) for instance the speaker suggests that she does not want to indulge in the misery, but do something about it. By putting her opinion in the form of a question she wants to convince the addressee that there are no objections to this view. Rhetorical questions are also typically characterized by a reversed polarity : negative rhetorical questions are usually meant as a suggestion of the opposite and conversely. 2.2. Uses related to indirectness 2.2.1. Contexts of attenuation In contexts of attenuation German and Dutch are quite similar. Both languages mostly use an analytic construction with either würde or zou. For German, the synthetic conjunctive only seems to be available for modal verbs, as well as haben (‘to have’) and sein (‘to be’), see (18c). In the latter cases, Dutch can either have the modal preterit or the zou form to make the SoA less direct, more polite. (18) a. Ich würde sagen, ... * Ich sagte, ... ‘I would say,…’ b. Hmm ich würde meinen, dass die Prüfungszeit mit Französisch recht moderat begonnen hat. (andi.brenner.lu/category/dabei-gewesen/) ‘Hmm, the examination period has taken a really moderate start with French, I would think.’ c. Ich möchte / hätte gern ein Bier / würde gern ein Bier haben. Ik had graag een biertje / zou graag een biertje hebben. ‘I would like to have a beer.’ In Dutch, zou and the preterit are not always interchangeable. In (19a) the request is attenuated by presenting the volition as past. This is possible because ‘wanting to do something’ does not entail the realization of the SoA 9. Zeggen ‘to say’ in (19b), on the other hand, cannot be attenuated by turning it into the past tense, as zei would get a simple realis (past tense) reading. It needs the zou construction. 9 Compare : Ik wilde naar huis gaan, maar deed het niet (‘I wanted to go home, but didn’t) with *Ik zei iets, maar deed het niet ‘I said something, but didn’t) Dutch equivalents of the German past conjunctive 191 (19) a. Ik wilde vragen wat dat boarding house kost (CGN) = ik zou willen vragen ‘I wanted to ask what that boarding house costs’ b. Dus ik zou zeggen, mijn anderstalige collega’s zijn op de hoogte dat wij op het vlak van verdeelsleutels onze rechtmatige eisen zullen verdedigen (CGN) *ik zei (lit.: I said) ‘So I would say, my colleagues have been informed that with respect to allocations we will defend our rightful demands.’ 2.2.2. Indirect speech and quotative meanings Next to the marking of different subfields of irrealis, the expression of indirect speech is the second major function of the conjunctive in German. (20) […] aber dann, wenn es euch passt, kommt ihr daher und behauptet, wir hätten uns nicht genug um sie gekümmert (HFS 46) [...] maar dan, als het jullie toevallig zo uitkomt, beweren jullie dat we ons niet genoeg om de kinderen hebben bekommerd (HVS 41) ‘but then, whenever it suits you, you come along and claim we hadn’t taken enough care of them’ In the domain of indirect speech, German and Dutch differ considerably from each other. Neither the zou-form nor the modal preterit in Dutch cover what the conjunctive does in German contexts of indirect speech. In Dutch, indirect speech is typically unmarked, whereas according to traditional rules in German, the conjunctive is used to signal that the speaker is reporting (by default KI and KII to replace ambiguous forms) 10. In Dutch, only ‘indicative’ forms are used (present, perfect, past). The preterit cannot be claimed to have a modal function in this context, but rather expresses simultaneity with a past verb form in the main clause. (21) Später hat es dann geheißen, ich hätte wiederum einmal einen Eklat verursacht (HFS 52) Later deed het verhaal de ronde dat ik weer eens amok had gemaakt (HVS 47) ‘Later on there was a rumour saying I had caused scandal once again’ Zou + infinitive is not used as a grammatical device to signal indirect speech. It does function, however, in main clauses as an evidential marker, more specifically as a quotative one. In example (22), the use of zou signals that the speaker doesn’t (want to) commit herself to the truth value of the proposition. It can imply that the speaker doubts the propositional content, 10 Note that in German the indicative is increasingly used in such contexts as well. 192 Linde Roels, Tanja Mortelmans & Johan van der Auwera but need not necessarily do so. The stress is on the fact that the speaker signals she is not the original source of information. The modal preterit cannot be used, as it would simply shift the SoA in the past. German uses the modal verb sollen (in the indicative) in this context (see for discussion De Haan 2001). (22) Volgens schattingen zou 70 tot 80% van de vacatures hier in Vlaanderen in handen zitten van deze integratoren. (CGN) ‘According to estimates, 70 to 80% of jobs here in Flanders are [said to be] in the hands of these integrators.’ 2.2.3. Temporal and modal uses Finally würde + infinitive and zou + infinitive occur to express future in the past (23) and erlebte rede or free indirect discourse (24). Zifonun et al. (1997 : 1775) define erlebte rede as a mixture between narrated discourse in author’s perspective and rendering of thoughts from the perspective of one or more characters. In German only the analytic form with würde occurs. The synthetic KII is not possible at all. (23) […] bis sie sicher war, daß der Mann seine Handlaterne nicht anzünden würde (DV 31) […] tot ze zeker wist dat de man zijn handlantaarn niet aan zou steken (DV 31) ‘until she was sure that the man would not turn on his flashlight’ In German, another marker of future in the past is sollte. Both sollte and würde can be used to represent a SoA which is seen as lying in the future from a moment in the past. The bipartition is based on a difference in viewpoint. Würde is used when the SoA is considered from the viewpoint of one of the characters in a story or report. Sollte occurs when the SoA is viewed from the perspective of the narrator or speaker. Sollte is therefore more determined than würde : it presents the SoA in terms of a necessary development of reality. Dutch does not distinguish between different viewpoints, but uses zou throughout. (24) Ich sollte sehr schnell merken, dass die Napolitaner keine Fußgänger sind (DV 36) Ik zou algauw merken dat Napolitanen geen voetgangers zijn. (DV 35) ‘I was to realize very soon that the people of Naples are not pedestrians’ In Dutch the default form is zou + infinitive. The modal preterit does not occur here. Dutch equivalents of the German past conjunctive 193 (25) a. Dat denken, weet ik nu, daar zou ik pas later, eenmaal terug op het landgoed van mijn man, eenmaal zwanger, een begin mee maken. (DV 78) Mit dem Denken, weiß ich jetzt, würde ich erst später, nach der Rückkehr auf das Landgut meines Mannes und einmal schwanger, beginnen. (DV 82) ‘I now know that I would only start to think about that later, once I was pregnant and back on my husband’s estate.’ b. ‘Gasparo, die er een avond zingen op had zitten, voelde zich toch helemaal niet moe. Straks zou hij, verwarmd door de prettige conversatie, nog even met iemand mee naar huis gaan. (DV 112) Gasparo spürte nach einem Abend singen gar keine Müdigkeit. Gleich würde er [...] noch jemanden in dessen Haus begleiten. (DV 117) ‘Gasparo, who had sung for a whole evening, did not actually feel tired at all. He would, being warmed up by the pleasant conversation, just go home with someone later.’ In addition to being used in contexts of future in the past, zou can also reflect the primary deontic meaning zullen originally had (which becomes particularly clear through comparison with German). Uses like (26a) and (26b) correspond to German sollte, and würde is not possible. In Dutch the modal preterit cannot be used ; as the deontic component (the element of obligation imposed by a third party) would get lost. (26) a. Ik zou mij vestigen op zijn landgoed bij Altavilla, zo’n veertig mijl verderop, en niet in zijn huis in Napels. (DV 17) Ich sollte auf seinem etwa vierzig Meilen entfernten Landgut bei Altavilla wohnen und nicht in seinem Haus in Neapel. (DV 17) ‘I was supposed to stay on his estate close to Altavilla about forty miles away and not at his house in Naples.’ b. Ik zou mij, in zijn bijzijn, met amber noch muskus parfumeren. (DV 17) Und ich sollte mich in seinem Beisein weder mit Amber noch mit Moschus parfümieren. (DV 17) ‘And I was not supposed to wear ambergris or musk scents in his presence.’ 3. Some diachronic observations Following Duinhoven (1997), Dutch soude (i.e. the past tense form of sullen) originally expressed an obligation in the past, as in (27a). Typically, the realization of the SoA lies in the future with respect to the past viewpoint, i.e. the SoA is not real. (27) a. Karel sach, dat wesen soude. ('dat het zo moest zijn') ‘Charles saw it had to be / would have to be like this.’ 194 Linde Roels, Tanja Mortelmans & Johan van der Auwera In Middle Dutch, then, soude gradually extends its uses to include cases in which the element of obligation fades into the background. What remains is the construal of the SoA as not necessarily belonging to reality anymore. (27) b. Hi peinsde, hi soudet brengen voort. ‘He thought he would produce.’ Ten gevolge van het frequente gebruik in de afhankelijke bijzin, werd soude daardoor een hulpwerkwoord van modaliteit dat eerst en vooral het aspect van niet-werkelijkheid tot uitdrukking brengt. Het tijdsaspect en de woordbetekenis (‘moeten’ dan wel ‘zullen’) spelen hoegenaamd geen rol meer. As a consequence of the frequent use in the dependent clause, soude has developed into an auxiliary of modality that primarily expresses the component ‘non-reality’. The temporal component and the lexical meaning are no longer of importance (Duinhoven 1997 : 431, translation ours). To summarize, zou originates in the modal verb zullen which primarily had deontic meaning. Through semantic bleaching the temporal meaning (future) has developed out of the deontic (‘needing to be done’ => ‘still to be realised’). Würde on the other hand is derived from werden, which is not a standard modal verb. Consequently, würde does not exhibit the same modal verb characteristics as its Dutch ‘counterpart’ (e.g. some of the uses discussed in 2.1.5, 2.2.2 and 2.2.3). 4. Conclusions : Dutch and German compared The foregoing has shown that one cannot simple equate the German conjunctive (both in its synthetic and analytic form) with Dutch zou. Striking differences between German and Dutch can be discerned in at least three domains. First, in irrealis comparisons (introduced by alsof), Dutch does not seem to need an irrealis marker, as if the conjunction by itself marks the conceptual break between reality and what it is compared to. German, on the other hand, clearly prefers conjunctive mood in this context : the conjunction als ob does not suffice. Another difference concerns the marking of indirect speech. Especially in written language, German consistently marks (by means of the present or past conjunctive) the fact that the utterance’s content must be traced back to a source different from the actual speaker. Dutch does not seem to bother to give indirect speech an extra marking (by means of zou, for instance). It relies more on contextual information (e.g. an introductory phrase like Hij zei dat ‘He said that’) than grammatical marking. Third, Dutch zou still shows traces of its origin as a modal verb in some of its uses : its quotative use (for which German makes use of the modal verb sollen), its use in particular types of rhetorical questions (see examples under 17) and Dutch equivalents of the German past conjunctive 195 uses like (26) – in which zou refers to a past obligation - still reflect the original deontic meanings of the verb. In German, neither the past conjunctive nor würde can appear in these contexts. The exact relationship between German würde and the synthetic past conjunctive is not easy to establish (the issue also goes beyond the aims of this paper). Still, it seems that two ‘kinds’ of würde-uses can be detected : one in which würde ‘replaces’ (particular forms of) the synthetic conjunctive, and one in which würde has clear future-time semantics and cannot be replaced by the synthetic conjunctive. Remarkably, Dutch zou functions as an equivalent of würde in both cases : the use of zou in irrealis conditionals or as an attenuation marker would be an example for the first category, whereas zou expressing future in the past (ex. 23-24-25) represents the second one. The observation that in Dutch spoken language, the combination of a zou- protasis with a zou-apodosis is the most frequent one, points to a high degree of grammaticalization of zou in this particular context. One of the other findings of this paper concerns the status of the past tense in Dutch, the ‘modal’ meanings of which seem often to be forgotten. We have shown that it is not the form of the verb which locates the event, but rather the whole context / discourse in which it is used. It has become clear that the German conjunctive can neither be equated with the Dutch (modal) preterit nor with the zou-form. Whereas in German the conjunctive is an obligatory grammatical marker in various irrealis and indirectness contexts, Dutch – in a number of cases - seems to rely more on contextual information to do the same job. References Andersson, S.-G. (1989). Zur Interaktion von Temporalität, Modalität, Aspektualität und Aktionsart bei den nichtfuturischen Tempora im Deutschen, Englischen und Schwedischen, in : A. Werner ; T. Janssen, (eds), Tempus – Aspekt – Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen. Tübingen : Niemeyer, 27-49. Bouma, L. (1973). The Semantics of the Modal Auxiliaries in Contemporary German. The Hague & Paris : Mouton. Cutrer, M.L. (1994). Time and Tense in Narrative and Everyday Language. San Diego : University of California. De Haan, F. (2001). The Relation between Modality and Evidentiality, in : Linguistische Berichte 9 : 201-16. de Vooys, C.G.N. (1967). Nederlandse Spraakkunst. Herz. door M. Schönfeld. 7e dr. Groningen : Wolters. Doiz-Bienzobas, A. (2002). The preterit and the imperfect as grounding predications, in : F. Brisard, (ed.), Grounding. The Epistemic Footing of Deixis and Reference. Berlin : Mouton de Gruyter, 301-347. 196 Linde Roels, Tanja Mortelmans & Johan van der Auwera Eisenberg, P. et al., (eds), (1998) Duden Grammatik der deutschen Gegenwartssprache, 6, Mannheim : Dudenverlag. Duinhoven, A.M. (1988-1997). Middelnederlandse Syntaxis : Synchroon en diachroon, Groningen : Nijhoff. Fabricius-Hansen, C. (2000). Die Geheimnisse der deutschen würde- Konstruktion, in : R. Thieroff et al., (eds), Deutsche Grammatik in Theorie und Praxis, Tübingen : Niemeyer, 83-96. Fabricius-Hansen, C. (2004). Wessen Redehintergrund ? Indirektheits- kontexte aus kontrastiver Sicht (Deutsch – Norwegisch – Englisch), in : Oddleif Leirbukt (Hrsg.) Tempus/Temporalität und Modus/Modalität im Sprachenvergleich, Tübingen : Stauffenburg, 119-155. Frantzen, J. (1920). Über den Gebrauch des Konjunktivs im Deutschen, Groningen : Noordhoff. Guchmann, M. ; Semenjuk, N. (1980). Zur Ausbildung der Norm der deutschen Literatursprache im Bereich des Verbs (1470 - 1730). Tempus und Modus, Berlin : Akademie. A3I B J32 23I B $3 K 88= % & E 2 LM 0K B ?A # Knobloch, C. (2001). Wie man den Konjunktiv erwirbt, in : C. Knobloch ; H. Feilke ; K-P. Kappest, (eds), Grammatikalisierung, Spracherwerb und Schriftlichkeit, Tübingen : Max Niemeyer, 67-102. Iatridou, S. (2000). The Grammatical Ingredients of Counterfactuality, Linguistic Inquiry 31(2) : 231-270. Langacker, R.W. (1991). Foundations of Cognitive Grammar. Vol. II : Descriptive Application. Stanford : Stanford UP. Leirbukt, O. (1991). Nächstes Jahr wäre er 200 Jahre alt geworden. Über den Konjunktiv Plusquamperfekt in hypothetischen Bedingungsgefügen mit Zukunftsbezug, Zeitschrift für germanistische Linguistik 19 : 158-193. Leiss, E. (1992). Die Verbalkategorien des Deutschen. Ein Beitrag zur Theorie sprachlicher Kategorisierung. Berlin & New York : Mouton de Gruyter. Nieuwint, P.J.G.M. (1984). Counterfactuals, De nieuwe taalgids 77(6) : 542- 555. Thieroff, R. (1992). Das finite Verb im Deutschen. Tempus–Modus–Distanz. Tübingen : Max Niemeyer. Van der Wal, M. ; Van Bree, C. (1994). Geschiedenis van het Nederlands. Utrecht : Spectrum. Warnant, L. (1966). Moi, j’étais le papa..., in : Mélanges de grammaire française offerts à Maurice Grevisse pour le trentième anniversaire du “Bon Usage”, Gembloux : Editions J. Duculot, 343-366. Zifonun, G. ; Hoffmann, L. ; Strecker, B. ; Ballweg, J. (1997). Grammatik der deutschen Sprache (3 Bde.), Berlin/New York : de Gruyter. ∗ Slavic verb prefixes are resultative∗∗ Boban ARSENIJEVIĆ Universiteit Leiden 1. Introduction Many semantic and syntactic aspects of Slavic verb-prefixes (SV-Ps) and Germanic verb-particles (such as their meanings, relatedness to PPs, aspec- tual properties or properties related to arguments structure) suggest that these two types of elements form a quite uniform class. Traditionally, one of the most frequent and most numerous types of particles are those that modify the eventuality, make verbs perfective and/or telic and often introduce or specify the result state of an event. It has even been claimed, for instance in Lüdeling (1998), that there is no real theoretical difference between Germanic sepa- rable verb-particles and resultatives. These particles also typically have the potential to introduce a new argument to the argument structure of the verb. The introduced argument quite regularly has the roles Undergoer (Theme, the affected participant) or Goal. I will argue that all event-modifying SV-Ps behave in the described way, especially in being resultative. In Slavic languages, it is rather uncontroversial that the internal (Di Sciullo and Slabakova 2005), i.e. lexical (Svenonius 2004) prefixes, belong to the class described above 1. Prefixation by an internal prefix makes the verb telic, and if a new argument is introduced it is always either the Undergoer or the Goal. Another similarity between separable Germanic verb-particles and SV- Ps is that for many different reasons, e.g. selectional relations, the same type of modification or lexical similarities, they are often closely related to prepo- itions. In formal syntax there are models that generate the particles and prefi- xes as heads of phrases with a prepositional meaning and structure – in PPs ∗ I am indebted to B. Gehrke for discussions, references and data. I also thank F. Kratochvil for examples and detailed judgments and interpretations, and to N. Elouazizi and H. Reckman for making our office such a good and stimulating place for linguistic work. This paper is a part of my research within the Spreekbuis and I2RP subprojects of the TOKEN2000 project funded by NWO. 1 The classes of internal and lexical prefixes in Slavic languages, as defined in the cited papers, do not fully overlap, but still share most of the properties relevant for the current discussion. Moreover, Arsenijević (2005b) provides an account in which external, i.e. super-lexical prefixes are not a separate class, but only differ from the internal ones in taking eventualities instead of nominal expressions as their arguments. © Cahiers Chronos 17 (2007) : 197-213. 198 Boban Arsenijević or in Small Clauses (SC) which lack either a (overt) specifier or a (overt) complement (cf. among others Emonds 1985, Kayne 1985 and den Dikken 1992). In Slavic languages prefixes tend to be fully or nearly homophonous with semantically corresponding prepositions, although there are a few exceptions. In this paper, I develop an analysis in which all event-modifying prefixes in Slavic languages are resultative, i.e. related to the predicate of the result of the eventuality, which is usually specified by the resultative component of the meaning of the verb, or by a PP/SC constituent. Syntactically, I propose to analyze prefixes as markers of agreement between the value that a certain property of the undergoer of change takes in the initiating sub-event and the value that this property takes in the result sub-event of the telic VP. These values are realized through PPs, which explain the phonological similarities between the two categories. Prefixes in this model do not trigger a result interpretation – they simply reflect it through agreement. I treat telicity along the lines of the work like Krifka (1998) and Borer (2005). This means that I relate telicity to quantificational properties of the event, especially to quantization. A predicate is quantized if when it holds for an object, a) it is never the case that it simultaneously holds for some part of this object and for the complement of that part respective to the original object and b) it is never the case that it holds for a sum of this object with another object for which it holds. The way I relate quantization with telicity is however slightly different than in either of the two cited approaches. I consider an eventuality telic iff a) it involves a change and b) it entails that this change is interpreted as quantized. Change is seen as the difference in some property of one of the event participants between the initiating and the result component of the eventuality. In this way, quantization is directly related to the kind of decomposition of eventualities that divides a telic eventuality into two components, its initiating and its result sub-event. In section 2 I give a brief overview of my main claim. In 3 I discuss and partly reject Filip’s (2004) Goal-Source Asymmetry, showing that both Goal- and Source-modifying prefixes correspond to resultative interpretations. In 4 I give a semantic and a syntactic account for these facts. In the semantic part I use a binary event structure and show that prefixes all reflect some facts about the resultative sub-event. Syntactically, I propose a configuration in which prefixes appear on the verb as a consequence of agreement between the head of the VP and the PP in its specifier. Finally I conclude that Slavic Slavic verb prefixes are resultative 199 spatial prefixes are indeed all resultative, and that this is not unexpected considering their aspectual effects and effects on argument structure 2. 2. Slavic prefixes introduce telicity Verb prefixes in Slavic languages are closely related to the phenomenon of grammatical aspect, i.e. perfectivity vs. imperfectivity : when attached to an imperfective verb, the verb always becomes perfective. If we accept that perfectivity is directly related to telicity, or more precisely that all perfective VPs denote telic events, it follows straightforwardly that all Slavic prefixes co-occur with telicity. This view has one unclear point : the nature of the link between telicity and perfectivity. There are views of grammatical aspect that deny it a direct correspondence with telicity. Within the Slavic paradigm, such is the approach of Filip (2004). Filip claims that not all perfective VPs are telic, or more precisely that while all prefixes make verbs perfective, not all of them introduce telicity. Part of this argument is based on the Goal- and Source-modifying spatial prefixes, for which she claims that only the former are telic, while the latter are not. The term she uses for this distinction is Goal-Source Asymmetry (G-SA). Interestingly, one class of spatial event modifiers is left out of this classification : those related to the process-component of the eventuality. Before discussing G-SA, I will briefly address the issue of process-modifying prefixes, because conclusions I reach for them will eventually extend to event-modifying prefixes in general. The basic question is : Are there process-modifying prefixes and which ones would that be ? One good class of candidates involves verbs of movement and their path modifiers. Prefixes that realize such meanings are given in (1a-b). Interestingly, both in the Serbo-Croatian (S-C) examples and their Germanic translations, the path modifiers are interpreted as resultatives. There is not only an entailment that the movement took place along / through a certain object, but also that it reached the end of the path that this object determines. Sentences in (1c-d) show that in languages where such constructions are grammatical (which may be conditioned by different factors), path modifiers may realize end-states (results) of eventualities. (1) a. Jovan je uz-trčao (uz brdo). (Serbo-Croatian) Jovan AUX up-ran.PTC (up hill) ‘Jovan ran up the hill.’ 2 I think that all event modifying prefixes (thus more or less all Slavic prefixes except for the Czech modal ones) are resultative, but in this paper I only talk about the spatial prefixes. 200 Boban Arsenijević b. Marija je pro-letela (kroz kanjon). Marija AUX through-fly.PTC (through canyon) ‘Marija flew through (the canyon).’ c. John is up the hill. d. Mary is through the canyon. The conclusion is split. There are prefixes that semantically relate to the process part of the event, but they also introduce telicity and are interpreted as resultative. Moreover, they (possibly through perfectivity) also impose a resultative interpretation for the process-modifying PPs. It is possible of course, as in (2), to make these VPs imperfective by adding an imperfective suffix to the verb, in which case we get a real process-related non-resultative interpretation for these modifiers. This effect, however, comes from the imperfective nature of the VP, which, just like the progressive, appears to cancel all the resultative entailments of the eventuality (even Goal modifiers become directional or intensionally embedded). (2) a. Jovan je uz-trča-va-o (uz brdo). (Serbo-Croatian) Jovan AUX up-ran-IMPF.PTC (up hill) ‘Jovan was running up the hill.’ b. Marija je pro-let-a-la (kroz kanjon). Marija AUX through-fly-IMPF.PTC (through canyon) ‘Marija was flying through (the canyon).’ The interpretation of (2a) is close to the English progressive : at the time of epistemic evaluation, Jovan either ran a part of the distance to the top of the hill or he ran towards the top of the hill an unspecified number of times, each time possibly reaching its top. Throughout the paper, when I talk about resultative interpretations, if not specified otherwise, I refer only to those that are entailed by the sentence, and not to those that are intensionally embedded. The discussion above complies with Žaucer (2002), who argued that all event-modifying prefixes in Slavic languages are stative. A direct consequence of his claim is that there are no real process-modifying prefixes, i.e. that although prefixes may involve some process-related component, their meaning is always essentially stative. The weak conclusion that follows from the observed facts is that verb prefixes in Slavic languages introduce perfectivity, which then forces resultative interpretations. The strong one is that they specify or relate to the result of the eventuality, and, since having a result makes the eventuality telic, that they also reflect telicity. In this view all event-modifying SV-Ps are seen as directly related to the result. I have only observed the behaviour of process-modifying prefixes, so the next step is to see if the Source-modifying ones behave differently. There Slavic verb prefixes are resultative 201 is at least one analysis of Slavic prefixes which claims that they do : Filip’s (2004) Goal-Source Asymmetry hypothesis and the next section brings several arguments against it. 3. The Goal-Source (A)Symmetry Hana Filip (2004) introduces the notion of Goal-Source Asymmetry (G-SA) in the aspectual structure of VPs. She claims, relying on data from Czech and English, that unlike Goal-modifiers, which always derive telic eventualities, Source-modifying verb-prefixes in Czech, as well as Source-modifying PPs cross-linguistically, have the aspectual effect of making an eventuality atelic (in her terms, weakly homogeneous). In this section I will argue that G-SA, if there is such a thing, is at least not as simple as in Filip’s formulation, and that on a deeper view Source-modified eventualities are just as telic as Goal- modified ones. Filip (2004 : 21) formulates G-SA as in (3). (3) The spatial orientation of directional modifiers determines the telicity status of derived predicates. Source-modifiers form atelic (homogeneous) predicates. Goal-modifiers form telic predicates. A conceptual argument for G-SA, presented and rejected in Gehrke (2004), is that Source-modifiers introduce only the starting point for an eventuality, but not the ending point (i.e. telos, termination, culmination), and therefore they do not make eventualities telic. The conceptual argument for G-SA relies on the traditional definition of telicity, formulated in terms of the point of termination, i.e. the telos. Recent literature (among others Krifka 1998, Rothstein 2001, Borer 2005) reaches a consensus of accounting for telicity in terms of quantificational properties of eventualities, crucially their quantization vs. homogeneity. In these terms, having one or the other end of the event delimited should not make any difference. Actually, to be quantized, eventualities need to be delimited on both ends. Indeed, as I show below, units qualified as Source-modifying, although indeed related to the Source, have a resultative interpretation in all perfective VPs. In syntactic terms, if G-SA holds, not only Slavic prefixes could not all be resultative, but there should also be at least two different VP-internal syntactic positions where they are derived. Since independent arguments for this structure are neither found in the literature, nor suggested by the data, this also seems like a weakly motivated complication in the syntactic representation of the VP. 202 Boban Arsenijević 3.1. Empirical arguments Empirically based arguments for G-SA are the following (the examples are taken from the cited paper by Filip). Filip presents the English example in (4) (her 28) as a minimal pair showing that the Source-modified VP in (4a) is atelic, while the Goal-modified one in (4b) is telic. (4) a. John ran away from the car for ten minutes / *in ten minutes 3. b. John ran to the car *for ten minutes / in ten minutes. On a closer look, however, (4) is not a proper minimal pair. In the relevant reading, (4a) involves the meaning of Direction, and it is even not entailed that John was at the car at the beginning of the event. (4b) on the other hand involves a Goal : it entails that at the end of the event John is at the car. The atelic reading of (4a) makes a proper minimal pair as in (5a-b). (5) a. John ran away from the car for ten minutes / *in ten minutes. b. John ran towards the car for ten minutes / *in ten minutes. For the Direction reading, there is no asymmetry in (5a-b) : both sentences involve atelic eventualities. Examples in (6a-b), which have Source- and Goal- modification respecti- vely, do show a certain asymmetry, but this asymmetry is not aspectual. (6) a. John ran from the car *(to the house) *for ten minutes / in ten minutes. b. John ran (from the car) to the house *for ten minutes / in ten minutes. It simply appears that the Source-modified sentence is ungrammatical with- out a Goal-modification, and the Goal-modified one is not strongly sensitive to the presence of Source-modification. I will come back to this asymmetry later, but for now the reason for it seems to be that the Goal-modified sentence in (6b) involves a variable of Source which is easily bound by John (or his contextually provided location in the initiating sub-event – he runs to the car from the place at which he is before running, from his earliest position in the eventuality). It is pragmatically less likely to be the case that the Goal in (6a) and similar sentences is provided by the context : it is usually part of the new information. Crucial examples for G-SA are those in (7) (Filip’s 25- 26). 3 Asterisk comes from the original (Filip 2004 : 29) : “‘(*)’ indicates that the sentence is acceptable in the inchoative reading, i.e., ‘after ten minutes, he started to run away from the car’”. This reading is irrelevant for the current discussion so the brackets can be ignored. Slavic verb prefixes are resultative 203 (7) a. SkočilP metr od branky, aby chytilP míč jump.PAST.3SG metre from goal in.order.to catch.PAST.3SG ball ‘He jumped a metre away from the goal so that he could catch the ball.’ b. SkočilP metr k brance, aby do ní jump.PAST.3SG metre to goal in.order.to into it vstřelilP míč. kick.PAST.3SG ball ‘He jumped a metre to the goal so that he could kick the ball into it.’ c. Od-skočilP metr od okna. AWAY-jump.PAST.3SG metre from window ‘He jumped a metre away from the window.’ d. Při-skočilP ??metr k oknu. TO-jump.PAST.3SG ??metre to window ‘He jumped a metre to the window.’ On the first sight, these examples really seem to confirm G-SA : adding the Source-modifying prefix in (7c) and adding the Goal-modifying one in (7d) seems to lead to different effects on telicity (superscript letters mark grammatical aspect : perfective and imperfective). Before going into a deeper discussion, it is important to mention that a number of Czech speakers find the sentence in (7d) totally fine, but meaning that the subject jumped one metre closer to the window (thus not ending exactly at / by the window). Gehrke (2005) reports that in fact these sentences are accepted as grammatical by a majority of Czech speakers. Independently of the grammaticality issue, I will show that the metre- phrase is not a proper test for telicity because it is a modifier that determines the Goal of the event (a Goal modifier) in terms of its distance from a certain point. The from-phrase does not (only) realize Source-modification at the level of the eventuality – it is (also) a distance- or measure-phrase, which specifies the position of the Goal. 3.2. Metre-phrase as a telicity test Filip takes compatibility with the metre-phrase as a telicity test for VPs involving prefixed verbs. If a VP combines with the expression x metres to / from the window – it is atelic (homogeneous) and if not it is telic (quantized). However, this test is problematic in several ways. First, this metre-phrase resembles the classical for / in x time test. This similarity is especially strong in Slavic languages, where the counterpart of the for-phrase appears without any preposition, just like the metre-phrase 4. 4 It is interesting to see how these prefixes behave with phrases determining the length and the starting and ending points of the temporal interval of the 204 Boban Arsenijević But in the very same paper, Filip uses the latter as a test of perfectivity and not telicity. Since she distinguishes between the two, a potential clash arises, unless the parallel is given up. But this takes away the only reason why the metre-phrase should be treated as an aspectual test. Secondly, there is a problem with the test that shows up when we try to combine the metre-phrase with non-iteratively interpreted imperfective VPs, for which it is uncontroversial that they are atelic. The combination is possible only if there are both a from- and a to-PP, delimiting both ends of the distance denoted by the measure phrase, as illustrated in (8). (8) a. BĕhavalI kilometr od okna ??(k stromu). run.PAST.3SG kilometre from window to tree ‘He was running around one kilometre from the window to(wards) the tree.’ b. BĕhavalI od okna ??(k stromu). run.PAST.3SG from window to tree ‘He was running from (the direction of) the window to(wards) the tree.’ Recall that we had a similar situation with the English example in (5). I argue that this is because the PPs actually modify an abstract nominal denoting a quantized path (for a detailed discussion of this point see Arsenijević 2005a and 2005b). Imperfective in (8) produces a reading fully parallel to the English progressive. Ergo, one available interpretation is that the subject in (8a) is engaged in running, and he has to be somewhere within a certain definite path of one kilometre with the Source and the Goal as specified by the two PPs. The other reading is that the subject is involved in one of the unbounded number of iterations of running the same path. In both readings of the imperfective VP, there appears to be a telic eventuality that involves the full path, to which the meaning of the imperfective is applied. This is especially clearly suggested by the requirement of having an overt Source and Goal participant, i.e. of delimiting both the Source and the result end of the eventuality. In other words, the eventuality needs a quantized path before it can receive the imperfective interpretation. Paths are quantized due to the presence of the two PPs, which mark that they are delimited on both ends. Without one of the two PPs quantization of the path would not be achieved. Once modified by the PPs, the metre-phrase in (8a) gets a definite interpretation, and refers to a particular path that the subject of this sentence is running. Therefore, the interpretation of this phrase cannot be considered independent of the interpretation of the metre-phrase and the Goal- and Source-modifiers. It appears that all these three elements in fact modify the eventuality, such as ‘x-long time from/after period Y” or “x-long time until period Y”. I leave this for further research. Slavic verb prefixes are resultative 205 path of the eventuality. The metre-phrase modifies it for length, and the Goal- and the Source-phrase determine its starting and ending points. Crucially, it is the presence of a quantized path that makes the eventu- ality telic, and not any of the three types of modifiers. The presence of any of the three modifiers is possible only if the path is quantized, i.e. the presence of one of them is possible only if the path also has the other two elements, even if they are not specified. Therefore, it is wrong to see a Goal-modifier as a part of the eventuality, which contributes telicity, and the metre-phrase as an independent element, which serves as a test of telicity. All these three phrases should in fact be given the same status with respect to telicity. This introduces the third problem for treating the metre-phrase as a test for telicity. The metre-phrase, when combined with the Source-modifier, results in the modification of the Goal participant. If a quantized path is specified for its Source and length, then the Goal is also partly specified : it is part of the circle around the Source with the radius equal to the length of the path. If the context further suggests the direction of the path, then the Goal is even fully determined. Observe in this sense the examples in (9). (9) a. Od-skočilP metr od okna. AWAY-jump.PAST.3SG metre from window ‘He jumped (to the point) a metre away from the window.’ b. Po-stavilP se metr od okna ON-place REFL metre from window ‘He placed himself one metre from the window.’ c. SedělI metr od okna. sit.PAST.3SG metre from window ‘He was sitting one metre from the window.’ This is a further argument that the Source- and Goal-modifying PPs, together with the metre-phrase, are generated within the same phrase, which denotes a line in space. In telic eventualities, this line is interpreted as the path and in atelic eventualities as a modifier of a location. Whatever conclusion made from the behaviour of the metre-phrase, it cannot be about the aspect of the eventuality, but rather about the internal structure of the path-phrase. This means that G-SA is orthogonal to telicity. Additional support for this view comes from the standard tests of telicity and their relation with the metre-phrase. Standard tests, like for-/in-phrases, other temporal adverbials like at that moment, or Goal-modifying phrases, identify eventualities with Source-modifying prefixes as telic, giving the interpretation in which the eventuality reaches its termination 5. This is shown in (10). 5 Filip takes many of the standard tests of telicity to actually indicate grammatical aspect. I leave this question aside for the moment, but the analysis I will provide 206 Boban Arsenijević (10) a. Od-plavalP k / od břehu za hodinu / *hodinu. swim PAST.3SG to / from beach.DAT for hour.ACC / hour.ACC ‘He swam away to / from the beach in an hour / for an hour.’ b. Od-plavalP (od brehu) (*zrovna). AWAY-swim PAST.3SG from beach at that moment ‘He swam away (at that moment).’ c. Od-skočilP ke kredenci AWAY-jump.PAST.3SG to cupboard ‘He jumped to the cupboard.’ Furthermore, in Serbo-Croatian, the sentences in (11a and b) are fully equivalently interpreted. (11) a. Od-skočio je metar od prozora. (Serbo-Croatian) AWAY-jump.PTC.3SG AUX metre from window ‘He jumped (to the point) a metre away from the window.’ b. Od-skočio je na metar od prozora. AWAY-jump.PTC.3SG AUX on metre from window ‘He jumped (to the point) a metre away from the window.’ c. Od-skočio je od prozora na metar AWAY-jump.PTC.3SG AUX from window on metre (daljine). distance.GEN ‘He jumped from the window (to the point) a metre away from the window.’ d. Odskočio je od vrata na metar (daljine) AWAY-jump.PTC.3SG AUX from door on metre distance.GEN (od prozora). from window ‘He jumped from the door (to the point) a metre away from the window.’ This suggests that, underlyingly, there is always a silent preposition in front of the measure phrase in the construction of the type in (11a). Symptomati- cally, this preposition contributes Goal interpretation. The actual semantic structure of at least one reading of these two sentences, where the agent both jumps from the window and ends on a metre distance from it, seems to be the one given in (11c), with possibly two identical PPs, one of them being a real Source-modifier, and the other establishing a relation between the Goal and some other location, a PathP-internal Source. The two Sources need not be identical, as illustrated in (11d). Finally, the from-to construction in (8), for which I later show that it combines well with imperfective verbs, cannot combine with the perfective ones, no matter the type of modification. Both examples in (12a-b) are will define both lexical and grammatical aspect in terms of quantizedness and will give a more precise insight into what exactly which of these tests indicate. Slavic verb prefixes are resultative 207 grammatical, but their meanings are different from those derived with the imperfective verb. (12) a. Od-skočilP od okna ke kredenci. AWAY-jump.PAST.3SG from window to cupboard ‘He jumped from the window to the cupboard.’ b. Při-skočilP od okna ke kredenci. TO-jump. PAST.3SG from window to cupboard ‘He jumped from the window to the cupboard.’ One of the PPs is clearly realizing the Source of the event, and the other the Goal, unlike in what I originally identified as the from-to construction, where they together delimit the path and thus do not directly participate in the event. In one case there appears to be a complex from-to construction, and in the other two independent modifiers of the locations of the subject in the initiating and result sub-events. The ‘unity’ of the from-to construction can be well explained by the fact that both its parts are generated inside one single phrase, the one realizing the path. I consider the discussion above sufficient to reject G-SA at the aspectual level. I have not only shown that there is no such thing as aspectual G-SA, but also that Source-oriented event modifiers all have a resultative component (of course, just like with Goal-modifiers, it can be independently blocked by imperfectivization). This apparent paradox, that Source-oriented modifiers get a result interpretation needs an explanation. This is the topic of the next section. 4. Decomposing the paradox This section discusses the paradox observed above, namely that Source- modifying elements have resultative interpretations. Two different aspects of this question are tackled : how the resultative interpretation is derived and how verb prefixes are related to this phenomenon. 4.1. Source in the result : the event structure In this subsection I give a more concrete semantic analysis of the investigated phenomenon. It is based on event structure, more precisely on the model proposed in Arsenijević (2005b). In this model, atelic eventualities are represented as simple predicates involving temporal traces, and telic eventualities are represented as asymmetric concatenations involving one process and one state. The process is interpreted as the initiating sub-event, and the state as the result. The process and the state share one argument, which is interpreted as the Undergoer. All participants of an eventuality must be interpreted in at least one of the two sub-events. In syntax, the sub-events 208 Boban Arsenijević get structurally realized as two arguments of the predicate concat. In terms of phrase structure, the initiating sub-event is realized in the specifier position, and the result sub-event in the complement of the concat-predicate. The event structure is illustrated in (13) : (13) a. Jan od-skočil od okna. Jan away-jump.PAST from window ‘Jan jumped away from the window.’ b. concat(e1, e2, e) & add_to(Jan, S1, e1) & place(Jan, window, S1) & manner(e1, jump / bounce) & place(Jan, p2, e2). Three eventualities have the following properties. One of them (e) is built as the concatenation of the other two (e1 and e2). The sub-event e1 is built from the predicate add_to, which introduces dynamicity and thus makes this sub- event a process. The first argument of this predicate introduces dynamicity to the eventuality, and the second instantiates it, by undergoing a change. Predicate add_to takes two arguments, one of which is Jan and the other the predication S1, which specifies that the place of Jan is at the window. This eventuality is also modified for manner : its predicate of dynamicity has properties of jumping, or bouncing, which are two meanings associated with the root skok in Czech (realized in the verb). The aggregate interpretation of this sub-event is that there is a process with the properties of jumping or bouncing, which is initiated by Jan and affects the fact that he is placed at the window. The other sub-event simply specifies that Jan is at some place represented by the variable p2. The sentence in (13a) does not specify p2 : this location either stays underspecified or gets a specification from the context. The sentence does, however, entail that p2 is (part of) the complement of the location specified as the window. This is a consequence of the application of the dynamic predicate add_to to the predication S1, after which this predication cannot hold any more. The interpretation of the full eventuality is that the process of jump- ing / bouncing is initiated by Jan and it affects his location, which is specified as being at the window. This is followed by the state of John being at some other location than the window. A detailed presentation of the model is available in Arsenijević (2005b). Observe now the examples in (14), where the from-phrase appears with the metre-phrase. (14) a. Jan od-skočilP metr od okna. Jan AWAY-jump.PAST.3SG metre from window ‘Jan jumped (to the point) a metre away from the window.’ Slavic verb prefixes are resultative 209 b. ∃ e, e1, e2. concat(e1, e2, e) & add_to(Jan, S1, e1) & place(Jan, window, S1) & manner(e1, jump / bounce) & place(e2, Jan, ‘about one metre from the window’). c. Kaminek od-skočilP metr od okna. little_stone AWAY-jump.PAST.3SG metre from window ‘The little stone bounced (off the window) to the point about one metre from the window.’ d. ∃ e, e1, e2. concat(e1, e2, e) & add_to(Ø, S1, e1) (& place(stone, window, S1)) & manner(e1, jump / bounce) & place(e3, stone, 'about one metre from the window’). The presence of the metre-phrase results in having the result place of the Undergoer additionally specified. In this example, it is not just any location different from the one specified by the source, but one that also has the property of being about one metre away from the Source. Assuming the presented model, the example in (14c-d) further confirms the claim that the from-phrase does not relate only to the first sub-event and therefore is not only a Source-modifier. At the same time, this phrase bears a meaning related to the result sub-event : it specifies the Goal as being part of the complement (in the sense of negation !) of the Source, and thus this phrase is also a Goal-modifier. For a certain location L2, to be part of the complement of some other location L1, entails that it is also at a certain distance from L2. Thus every Source-modifier also introduces a distance modifier to the result sub-event. The distance is normally unspecified, but a modifier like the metre-phrase can be used to assign it a measure. Examples in (14) showed that the metre-phrase indeed narrows down the domain of the Goal, from the complement of the Source to only that part of this complement which also satisfies the measure. This confirms the Goal-related component of the Source-modifying elements. The proposed analysis is thus that Source-modifiers, when directly modifying a telic eventuality, modify both its sub-events. This relates their meaning to the result of the eventuality. 4.2. Source entails Goal It is trivial that not all eventualities, but only those entailing a change (events in the narrow sense), may have a Source participant. Source in an event is either a) the starting location L1 of a participant X, which its location to L2 in that eventuality or b) the location L1 of the initiating sub-event, which is different from the location L2 of the result sub-event. In both cases, L1 is at some distance from L2. The different place in the result sub-event for both cases gives the resultative component of the Source. The entailment of the result sub-event may be cancelled by different aspectual elements (e.g. the 210 Boban Arsenijević imperfective in Slavic languages, the progressive in English, Dutch or Portuguese), but this presents an independent problem. 4.3. The syntax It follows from the semantic representation which I proposed above, that a segment of the semantics of the Source-modifier should be present in the semantic representation of the result sub-event, and consequently in the domain of its syntactic realization. I proposed that this segment is represented as a certain distance, which can be or be not specified for measure. In this section I introduce a syntactic structure for the presented semantics and discuss how it can derive prefixes. The syntactic representation in (15c) corresponds to the proposed semantic analysis for the repeated example in (15a, b). (15) a. Jan od-skočilP metr od okna. Jan AWAY-jump.PAST.3SG metre from window ‘Jan jumped (to the point) a metre away from the window.’ b. ∃ e, e1, e2. concat(e1, e2, e) & add_to(Jan, S1, e1) & place(Jan, window, S1) & manner(e1, jump / bounce) & place(e2, Jan, ‘about one metre from the window’). c. The syntactic structure of the sentence e (VP) [concat] e2 (VP) e1 (VP) jump/bounce Jan [at] ┐p1 (PP) Jan [add_to] S1 (PP) metre [distance] the window Jan /from/ [at] p1/the window Fig. 1 : Syntactic Structure of (15) The representation corresponds to the logical form of the sentence, and considers a number of rules at the phonological interface to determine among other things the realization of the multiple copies of certain elements (e.g. Jan) and a general possibility to reduce, or fully drop, certain elements in the Slavic verb prefixes are resultative 211 structure (e.g. at in both sub-events). Here, I only discuss the semantic domain of the structure. I propose the structure in (15c) as the general structure of a telic eventuality. When only the Goal is specified overtly, it will appear in the specifier of the distance-phrase, as in (16c). In this case, the predicate distance, i.e. the preposition from, is not overtly realized, because its complement lacks overt realization. (16) a. Jan při-skočilP ke kredenci. Jan TO-jump.PAST.3SG to cupboard ‘Jan jumped to the cupboard.’ b. ∃ e, e1, e2. concat(e1, e2, e) & add_to(Jan, S1, e1) & place(Jan, p1, S1) & manner(e1, jump / bounce) & place(e2, Jan, cupboard). c. The syntactic structure of the sentence e (VP) [concat] e2 (VP) e1 (VP) jump/bounce Jan [at] ┐p1 (PP) Jan [add_to] S1 (PP) the cupboard [distance] p1 Jan /from/ [at] p1 Figure 2 : Syntactic structure of (16) Naturally, if both the Goal and the Source are overt, there will be two prepositions in the surface realization of the VP, one for each of the participants. Let me finally propose an analysis for the generation of SV-Ps. One of the main conditions for establishing a telic structure is that the result sub- event specifies the Undergoer for bearing the negative value of some property that it is assigned in the initiating sub-event. The value assigned in the result sub-event may be further specified. Therefore, at the level of concatenation, for a telic eventuality (labelled e in the structure) to be formed, it has to be checked whether the two values assigned by the two sub- events to the relevant property of the Undergoer are indeed in mutual exclusion. This matching is reflected by a morpheme that corresponds to the 212 Boban Arsenijević predicate of the relevant property in the result sub-event. In Slavic languages this morpheme is specified to be a prefix. When the Source is overtly realized or strongly contextually suggested, this morpheme can realize the relevant predicate together with the negation applied to the value specified by the Source (so-called source-modifying prefixes). The fact that Source- modifying prefixes reflect a richer structure (i.e. also include the distance, or negation) may trigger some asymmetries. These asymmetries, however, do not consider the aspectual structure : both types of prefixes appear only in telic VPs. The paradox of resultative meanings of Source-modifiers is now solved : so-called Source and Goal-modifying prefixes are derived by the same structure. This complies with the intuition that all prefixes, including the Source-oriented ones, are associated with resultative interpretations for the VP. 5. Conclusion I showed that event-modifying prefixes in Slavic languages, no matter which part of the event they relate to, necessarily co-occur with a resultative interpretation of the event. This behaviour, which would be a surprise if prefixes were indeed modifiers, becomes very normal if they are seen as reflexes of a certain relation established between the sub-events of a telic eventuality. I argued that the G-SA, which would be a problem for the presented approach, in fact does not have any effects at the level of telicity. References Arsenijević, B. (2005a). Prepositions and event participants, in : A. Vilavicencio ; V. Kordoni , (eds), Proceedings of the second ACL- SIGSEM workshop The linguistic dimensions of prepositions, 39-46. Arsenijević, B. (2005b). Inner aspect and telicity, Ms. LUCL, Leiden University. Borer, H. (2005). Structuring sense, Vol. 2 : The normal course of events, Oxford : Oxford University Press. Carlson, L. (1981). Aspect and quantification, Syntax and Semantics, Vol 14 : Tense and Aspect, P. Tedeschi ; A. Zaenene, (eds), New York : Academic Press, 31-64. Dikken, M. den (1992). Particles, Holland Institute of Generative Linguistics : Leiden. Di Sciullo, A.; Slabakova, R. (2005). Quantification and aspect, in : A. Van Hout ; H. de Swart ; H. J. Verkuyl, (eds), Perspectives on aspect, Dordrecht : Kluwer : 61-80. Slavic verb prefixes are resultative 213 Emonds, J. (1985). A Unified Theory of Syntactic Categories, Dordrecht : Foris Publications. Filip, H. (2003). Prefixes and the delimitation of events, Journal of Slavic linguistics 11(1) : 55-101. Gehrke, B. (2005). The prepositional aspect of Slavic prefixes and the goal- source asymmetry, Proceedings of the ESSLLI workshop on Formal semantics and cross-linguistic data, 47-56. Kayne, R. (1985). Principles of particle constructions, in : J. Guéron ; H.-G. Obenauer ; J.-Y. Pollock, (eds), Grammatical Representation, Dor- drecht : Foris Publications, 101-140. Krifka, M. (1998). The origins of telicity, in : S. Rothstein, (ed.), Events and Grammar, Dordrecht : Kluwer, 197-235. Koopman, H. (2003). The locality of agreement and the structure of the DP in Maasai, in : W. E. Griffin, (ed.), The Role of Agreement in Natural Language Proceedings of the Fifth Annual Texas Linguistics Society Conferenc, Texas Linguistic Forum, 207-227. Lüdeling, A. (1998). On particle verbs and similar constructions in German, Dissertation, Tübingen. Ramchand, G. (2002). First phase syntax, Ms., Oxford University. Ramchand G. ; Svenonius, P. (2002). The Lexical Syntax and Lexical Semantics of the Verb-Particle Construction, in : L. Mikkelsen ; C. Potts. Somerville, (eds), WCCFL 21 Proceedings, Cascadilla Press, 101-114. Rothstein, S. (2001). What are incremental themes?, in N% ' EH0 ( 0 <86 9= Svenonius, P. (2004). Slavic prefixes inside and outside VP, Nordlyd Vol. 32, Nr. 2 - Slavic Prefixes, 205-253. Žaucer, R. (2002). The role of verbal prefixes in Slavic : evidence from Slovenian locative denominal verbs, MA thesis, University of Ottawa. This page intentionally left blank The acquisition of aspect in child Greek : A production experiment* Sophia DELIDAKI University of Reading 1. Introduction Tense is the grammaticalisation of location in time and marks a relationship prior to (past), concurrent with (present) or subsequent to (future) speech time (Comrie 1985). Aspect, on the other hand, marks the internal temporal constituency of a situation and does not have a specific reference which is why it is considered as rather distinct from tense (Comrie 1976). According to B. Comrie (1976) there are two aspectual components : lexical and grammatical aspect 1. Lexical aspect refers to the inherent seman- tic properties expressed by the verb and its arguments. These properties are considered to be three and are expressed as contrasting sets of ‘dynamicity’ (dynamic/stative), ‘telicity’ (telic/atelic) and ‘durativity’ (durative/ instantaneous). In this study we are mainly concerned with the telic/atelic distinction. Telic predicates are the ones that have an inherent end-point, for example break, while atelic predicates lack such an inherent end-point and are only given an arbitrary one, for example ride. Z. Vendler (1967) classified the verb phrases according to their inherent lexical properties into four categories ; ‘states’ which are stable undifferentiated periods (for example love), ‘activities’ which are processes with successive stages that terminate or stop but do not finish (for example drive), ‘accomplishments’ which are processes with an outcome (for example build a house) and ‘achievements’ which are instantaneous events that result in a change of state * I would like to thank Spyridoula Varlokosta for her helpful comments and support, Laura Wagner for providing me her work, Colin Grayer from the Department of Applied Statistics for his statistical advice. I am especially thankful to the children, their parents for allowing them to participate, as well as the director and the teachers of the 29th municipal kindergaten school in Athens. I am also grateful to the adults that participated in this study. I would also like to thank the audience of the 6th Chronos Conference on syntax, semantics and pragmatics of tense, mood and aspect in Geneva, where this study was presented. 1 Smith (1991) refers to lexical and grammatical aspect as ‘situation’ and ‘viewpoint’ aspect, respectively. © Cahiers Chronos 17 (2007) : 215-227. 216 Sophia Delidaki (for example write a book) 2. It is generally accepted in the literature that atelic predicates, on the one hand, include states and activities and telic predicates, on the other hand, include accomplishments and achievements. Grammatical aspect specifies the boundaries of an event and is ex- pressed by auxiliaries and a grammatical morpheme that is usually attached to the verb. According to B. Comrie (1976) there are two types of grammati- cal aspect morphology 3 : perfective and imperfective. Perfective morphology focuses on the initial and final boundaries of an event and views a situation in its entirety (for example I ate a sandwich yesterday). Imperfective morpholo- gy focuses only on the internal temporal constituency of a situation without any reference to the initial and the final boundaries of an event and views a situation from within (for example I was eating a sandwich yesterday). 2. Acquisition studies on tense and aspect Production studies on the acquisition of tense and aspect have revealed certain restricted combinations of tense or grammatical aspect with lexical as- pect. These studies include languages like English (Bloom, Lifter & Hafitz 1980 ; Shirai & Andersen 1995), French (Bronckart & Sinclair 1976), Greek (Stephany 1981), Hebrew (Berman 1983), and Italian (Antinucci & Miller 1973) among others. The pattern observed is the following : past or perfective forms are used mainly with telic predicates, while present or imperfective forms are combined primarily with atelic predicates. Depending on the language examined and its characteristics, it is either tense or grammatical aspect that is considered to be influenced by lexical aspect. Irrespective of which of the two is actually influenced, the basic claim is that a certain type of morphology (tense or grammatical aspect) is used mistakenly by children in order to mark lexical aspect. This view has been expressed in the literature by a weak and a strong version ; what is called the ‘Defective Tense Hypothesis’ (Weist et al. 1984) is the strong version claiming that originally tense is defective in child speech and it does not express any deictic but only aspectual relations. Consequently, combinations 2 Smith (1991) identifies five situation types : states, activities, accomplishments, achievements and semelfactives, where semelfactives are the instantaneous events that do not have an outcome (atelic) and have simultaneous initial and final endpoints (e.g. cough). 3 Smith (1991) identifies one more category of grammatical aspect : neutral aspect. Neutral aspect employs neither a perfective nor an imperfective morpheme and refers to aspectually vague sentences (e.g. the Future in French). It is considered to be more flexible than the other two types of grammatical aspect; it is weaker than the perfective since it allows for an open reading, but it is stronger than the imperfective since it allows for a closed reading. The acquisition of aspect in child Greek 217 other than past-perfective-telic and present-imperfective-atelic should never be found in child speech. On the other hand, the AFH or “Aspect before Tense Hypothesis” (Bloom et al. 1980) is the weak version, which suggests that lexical aspect is acquired earlier than tense and grammatical aspect, and as a result it initially influences children’s acquisition and choice of tense and grammatical aspect. Therefore, children use in a larger proportion the above combination, but it is not claimed that no other combination can be found in child speech. It is the weak version that we will be concerned with in the current paper. J. P. Bronckart and H. Sinclair (1976) were of the first researchers to examine the production of tense and aspect with French-speaking children. They interviewed seventy-four children between the age of 3;0 and 8;6. They performed some actions in front of the children with the use of toys and then asked the children to describe what they had seen. In order not to influence the children’s choice of tense and aspect, after the end of each action the experimenter only said ‘Tell me’ (‘Raconte’ in French) to the child. The actions performed by the researchers differed in four semantic features : telicity (being perfective, imperfective and aperfective events 4), duration of the action (lasting more or less than three seconds), frequency of the event (being frequentative/non-frequentative) and whether the action had a stated aim or not (that is, if a telic event eventually reached its endpoint or not). Their results presented evidence that the choice of tense at all ages takes into consideration the telicity feature. Consequently, children’s use of tense morphology appears to some extent to reflect aspectual features rather than temporal ones. This influence is less significant after the age of 6, as J. P. Bronckart and H. Sinclair’s (1976) results suggest. According to the researchers, this is in accordance with Piaget (1969), who has suggested the age of 6 as a landmark for the child’s cognitive development and understanding of time. The role of duration and frequency appeared to be important especially for telic events between the ages of 3 and 6, while the completion or not of telic events did not influence any of the children. 4 Recall that we used the terms “perfective” and “imperfective” here for grammatical aspect, while J. P. Bronckart and H. Sinclair (1976) use them for an event’s lexical aspect. In order not to cause any confusion, we will replace throughout the paper J. P. Bronckart and H. Sinclair’s (1976) “perfective” and “imperfective” terms with “telic” and “atelic”, respectively. In addition, J. P. Bronckart and H. Sinclair (1976 : 112) call “aperfective” events the ones for which “the distinction between perfective and imperfective is not pertinent”; we use the term “auditory” here over the “aperfective”, since it succeeds in describing the nature of these events without a reference to their grammatical aspect, which could also prove confusing. 218 Sophia Delidaki 3. Greek tense and aspect Greek is a language where tense and aspect are marked with distinct morphology and both grammatical and lexical aspect are encoded. It is really difficult to segregate tense and aspect, since it is a combination of verb stem choice and inflectional ending. Due to the design of our study we are mainly concerned with the use of present and past tenses in Greek, subsequently these are the two tenses that will be examined here. As far as grammatical aspect is concerned, the present tense has only imperfective morphology, while there are two separate tenses in the past : the perfective past (‘Aoristos’) and the imperfective past (‘Paratatikos’). Verbs that denote both types of lexical aspect (telic and atelic) can be expressed in all tenses and with both types of grammatical aspect. Table 1 presents the interplay of present and past tenses with lexical and grammatical aspect in Greek. Perfective Imperfective Present n.a. pez-o (I play) [Atelic] xtiz-o (I build) [Telic] Past epeks-a (I played) [Atelic] epez-a (I was playing) [Atelic] extis-a (I built) [Telic] extiz-a (I was building) [Telic] Table 1. Interplay of present and past tenses with grammatical and lexical aspect in Greek 4. Our study : an elicited production task 4.1. Subjects This study reports an elicited production task, largely based on the methodology used by J. P. Bronckart and H. Sinclair (1976). The subjects of the study were twenty-eight Greek-speaking monolingual children and twenty-five adult native speakers of Greek. The children’s age ranged from 3;0 to 6;3 with a general mean age of 4;9. For the needs of our analysis we further divided the children in two smaller age groups : the younger group consisted of fourteen children from 3;0 to 4;10 (mean age 3;9) and the older group consisted of fourteen children from 4;11 to 6;3 (mean age 5;4). The adults were either students at the local university or recent graduates, with ages ranging from 23 to 29 years old. The children were tested at a quiet area of the kindergarten school that they attended and the adults at their homes in the area of Athens, Greece. All subjects were tested individually in a single session lasting 20-30 minutes. The acquisition of aspect in child Greek 219 4.2. Design and procedure The aim of the task is to test Greek-speaking children and adult production performance concerning their choice of tense and grammatical aspect in relation primarily to the predicate’s lexical aspect. The experimenter acted out the actions in front of the children by using as materials several toys, such as toys in the shape of animals, a car, a fence, a boat, a ball, a house, a street, a river and a garage. Children, but not adults, had to take part in a pre-test with six events ; first the experimenter asked each child to name the toys used in the task, in order to be certain that they had no difficulties with the required vocabulary. If a child did not recognize some of the toys, it was explained to him/her what they were. Then the children had to describe adequately what happened. After the children got used to this procedure the actual task began. An additional two children were excluded from the final calculations, due to their failure to complete the pre-test. One of them was not willing to communicate at all and the other child was describing the toys involved in each event and not the actual actions performed. At the beginning of the task, the experimenter explained to the subject that some actions were going to be performed with the toys and after the end of each action he/she had to describe what happened in detail. In order not to influence the subject’s choice of tense and grammatical aspect, at the end of each action the child simply heard ‘Tell me’ (‘Pes mu’ in Greek) by the experimenter. The events examined were randomly presented to the subjects, in order to achieve an alternation within the session between telic events that reached completion, ones that did not reach completion, atelic and auditory events. The events examined were sixteen in all and they differed in four semantic features. First, they were differentiated regarding their lexical aspect, being telic (with an inherent endpoint) and atelic (with an arbitrary endpoint). According to Z. Vendler’s (1967) classification of verb phrases the telic events that were acted out consisted of two achievements and two accomplishments, while the four atelic events were all activities, since states were difficult to be acted out. Secondly, the events differed in their duration, which ranged from ½ to 15 seconds. We decided to examine the semantic feature of duration by dividing the events into ones lasting less and more than three seconds. According to J. P. Bronckart and H. Sinclair (1976 : 113) : “actions lasting less than 3 seconds are apprehended in a very different manner from those lasting more than 3 seconds ; the first produce a relatively simultaneous perception of duration, the latter lead to a quantitative or qualitative estimation of duration” 5. 5 According to J. P. Bronckart and H. Sinclair (1976), the 3-second cutting point was originally suggested by Fraisse (1948). 220 Sophia Delidaki In addition, four of the events -the ones named as auditory in Table 2- differed in their frequency, two of them being repeated (the ‘barking’ events) and the other two continuous (the ‘crying’ events). These events were all activities, that is, atelic events. Finally, the telic events were performed twice, differing in their completion, that is, in whether they reached their end point or not. We added this feature in order to check if our subjects recognize and understand the semantic feature of telicity or if they are solely relying on the feature of the event’s completion for their choice of tense and grammatical aspect. Table 2 presents the events examined 6 in the task : Telic Atelic Auditory a cow crosses a street a rabbit hugs a horse a baby cries for 2 secs [2 seconds] [2 seconds] [2 seconds] a rabbit pushes a car in a man drives a car a baby cries for 10 secs a garage [3 seconds] [10 seconds] [3 seconds] a horse goes to a house a boat sails in a river a dog barks once [10 seconds] [7 seconds] [½ second] a man builds a house a dog plays with a ball a dog barks 10 times [15 seconds] [10 seconds] [10 seconds with ½ sec intervals between barks] Table 2. The events examined and their duration 4.3. Analysis As it was seen earlier, the actions performed belonged in two main lexical aspect categories (telic or atelic) and a ‘target verb’ 7 was assigned to them. Since the ‘target verb’ was not uttered by the experimenter but was only acted out, the subjects could describe the events with either the ‘target verb’ or another verb of their choice. The differences in the verbs that the subjects used were mainly due to the fact that the subjects focused their descriptions on different parts of the event presented. For example, for the event ‘a rabbit pushes a car in a garage’, two descriptions provided by two of the children can be seen in examples (1) and (2) below : 6 Table 2 does not include the four telic events that do not reach completion. These were exactly the same as the telic events in Table 2, except from the fact that they did not actually culminate. 7 “Target verbs” are considered to be those in Table 2. The acquisition of aspect in child Greek 221 (1) O laγos oδiγuse. the rabbit drive-Past/Imp/3sg ‘The rabbit was driving.’ (2) O laγos evale to aftokinito sto garaz. the rabbit put-Past/Perf/3sg the car in the garage ‘The rabbit put the car in the garage.’ The same event (‘a rabbit pushes a car in a garage’) that was categorized by the experimenter as a telic event (Table 2) is described with an atelic event in sentence (1) and with a telic event in sentence (2). In the present task, we followed the same analytical procedure that J. P. Bronckart and H. Sinclair (1976) did ; we coded the children’s answers based on the number of times that the children used a lexical aspect category (telic-atelic events 8) in their description that matched the one that the experimenter had already assigned to the predicate during the task’s design. For example, when we performed a telic action, only the children’s answers that included a telic event were counted in the results. Actually, in most of the cases, subjects chose to describe each event with the verb and the lexical aspect (telic / atelic) they were expected to, based on the design of the task, as seen in Figure 1 : CHILDREN ADULTS 100 90 80 70 60 TELIC 50 ATELIC 40 30 20 10 0 Atelic Auditory Telic Telic-No Atelic Auditory Telic Telic-No comp comp Figure 1. Lexical categories used by adults and children in the description of each set of verbs Due to the experimental design all the events that the subjects described happened in the past, since they have already finished when the subject was asked to describe them. Consequently the tenses that we would expect them to use in their description are the two past tenses. In the results section we will observe that children and adults used three tenses in order to express the 8 In order to categorize the verb phrases that the subjects provided according to their lexical aspect as telic or atelic, we used a number of diagnostic tests that D. R. Dowty (1979) suggests. 222 Sophia Delidaki events : present tense (only imperfective morphology), the perfective past tense (Aoristos) and the imperfective past tense (Paratatikos). 4.4. Results 4.4.1. Telicity In search of results that would confirm or disconfirm the AFH for the Greek language, the first feature examined was telicity and its role. In Table 3 the distribution of the three tenses used with telic and atelic events by the adults, the children altogether and each age group separately is presented. Two are the main observations to be made about the proportions in Table 3. First of all, children use a high percentage of present tense when the verb is atelic and perfective past (Aoristos) when the verb is telic, which is in accordance with the AFH. What is particularly interesting is the second observation that the same pattern holds for adults as well at a high percentage. Atelic Telic Ages Present Paratatikos Aoristos Present Paratatikos Aoristos Adults 57% 21% 22% 17% 0% 83% Children 49% 11% 40% 11% 2% 87% Older 56% 7% 37% 13% 2% 85% Younger 41% 15% 44% 8% 2% 90% Table 3. Tenses used to describe atelic and telic events The difference observed in the percentages with which the present tense and the perfective past tense are selected with telic and atelic predicates are statistically very significant for all age groups (p<0.001) 9. Adults and children used significantly more times the perfective past to describe telic events and the present tense to describe atelic events. The same happens with the difference in grammatical aspect ; in order to calculate this difference, we conflate the present tense and the imperfective past into an imperfective category and the perfective past alone represents the perfective category. It is again obvious that subjects mainly used imperfective aspect to describe atelic events (around 60% for the children altogether and 78% for adults) and perfective aspect for telic events (around 87% for the children altogether and 83% for adults). The statistical difference found between the use of perfective and imperfective 9 All the statistical tests performed were chi-squared tests. An alpha level of 0.05 was used; the differences found were considered to be statistically significant when p<0.05 and statistically not significant when p>0.05. When p<0.001 the difference obtained was considered to be statistically very significant. The acquisition of aspect in child Greek 223 morphology for telic and atelic events was statistically very significant for children altogether and adults (p<0.001). 4.4.2. Duration Telic and atelic events were divided according to their duration into ones that lasted more and less than three seconds. In Table 4 the results regarding the feature of duration for the telic events acted out can be seen. No particular difference appears to exist in the children’s performance between the events lasting more and less than three seconds, since all of them are almost exclusively expressed with the perfective past tense. Only the younger children and the adults seem to use more present tense and less perfective past for a longer lasting than a shorter lasting event ; statistically this difference is not significant for the younger children (p>0.05) and it is only weakly significant for the adults (p=0.063). Consequently, duration does not influence the choice of tense and grammatical aspect when the verb is telic. Telic less than 3 seconds Telic more than 3 seconds Ages Present Paratatikos Aoristos Present Paratatikos Aoristos Adults 8% 0% 92% 24% 0% 76% Children 11% 2% 87% 12% 2% 86% Older 14% 4% 82% 12% 0% 88% Younger 7% 0% 93% 12% 4% 84% Table 4. Tenses used with telic events lasting less and more than 3 second In order to check the role of duration in the four atelic events that were acted out, in Table 5 the percentages of the three tenses used to describe the atelic events that lasted less and more than three seconds are presented. Atelic less than 3 seconds Atelic more than 3 seconds Ages Present Paratatikos Aoristos Present Paratatikos Aoristos Adults 36% 32% 32% 73% 12% 15% Children 34% 11% 55% 64% 10% 26% Older 37% 10% 53% 77% 2% 21% Younger 30% 12% 58% 50% 18% 32% Table 5. Tenses used with atelic events lasting less and more than 3 second As can be seen in the above Table, children present a more clear-cut result in comparison to adults ; whenever atelic events have a shorter than three seconds duration the percentage of perfective past is almost double (55%) than with events with a longer than three seconds duration (26%) and 224 Sophia Delidaki the opposite happens with the present tense (from 34% when the atelic event lasts less than three seconds to 64% when the event lasts more than three seconds). A chi-squared test on the children’s answers altogether suggests that they used significantly (p<0.05) more present tense than the other two tenses with atelic events lasting more than three seconds and significantly (p<0.05) more perfective past with events lasting less than three seconds. Adults’ responses are more confusing with the shorter than three seconds duration events, but still there is a significant difference in the use of the present tense (73% with events longer and 36% with events shorter than three seconds), which in principle is their primary choice when the event is atelic (Table 3). A chi-squared test suggests that this difference is indeed significant (p<0.05). Consequently, when the verb is atelic adults and children’s tense choice is influenced by the feature of duration. 4.4.3. Frequency In order to check the role of frequency, the four auditory events were examined ; the two ‘crying’ events (which are continuous) and the two ‘barking’ events (which are repeated). Our first observation is that adults’ primary choice is the present tense for both types of events. This finding is most probably due to the fact that all the events examined here (repeated and continuous) belong in the atelic category and adults were found to mainly combine atelic events with the present tense (Table 3). Adults’ preference for the present tense does not suggest that they do not distinguish between repeated and continuous events, but rather that they just choose different means to describe the semantic feature of frequency, mainly with adverbs and adverbial phrases. In the relevant data answers like “the dog barks many times / once” or “the baby cries a lot / a little” were very common. Continuous Repeated Ages Present Paratatikos Aoristos Present Paratatikos Aoristos Adults 100% 0% 0% 82% 0% 18% Children 65% 19% 16% 51% 10% 39% Older 79% 10% 11% 64% 14% 22% Younger 50% 29% 21% 38% 0% 62% Table 6. Tenses used in continuous and repeated events In Table 6 it can also be seen that although the children altogether mainly use the present tense for both types of events, there is a small difference in the use of the perfective past tense (39% with repeated and 16% with continuous). A chi-squared test suggests that this difference is statistically significant (p<0.05). This perhaps can be looked at in combination with the The acquisition of aspect in child Greek 225 importance of the duration feature that was presented in Table 5. If this is true, then children might consider the continuous events as having a longer duration and the repeated ones as a combination of many instantaneous events, rather than a single one. If we examine the two age groups, it is observed that the younger group of children is the only one that uses mainly the present tense (50%) with continuous events and primarily the perfective past (62%) when describing repeated events. This is verified by a chi-squared test which presents a significant difference for younger children (p<0.05). This suggests that frequency plays a role in the choice of tense at a younger age and diminishes as it most probably finds other ways to be expressed, like in the adult data with adverbs. 4.4.4. Completion or not of telic events The data relating to the fourth feature examined, that of ‘completion’, are illustrated in Table 7. Adults and children mainly use the perfective past in order to describe both types of telic events and in similar proportions across the two. Reach Completion Do not reach Completion Ages Present Paratatikos Aoristos Present Paratatikos Aoristos Adults 14% 0% 86% 31% 9% 60% Children 11% 2% 87% 11% 4% 85% Older 13% 2% 85% 10% 4% 86% Younger 8% 2% 90% 13% 4% 83% Table 7. Tenses used with telic events that reach completion or not What can be observed in this table is that adults are the only ones to present a small difference in the present tense use (31% with telic events that do not reach completion and 14% with telic events that reach completion), but this difference is not statistically significant (p>0.05). Adults may be using the present tense with a future meaning when describing non-completed telic events ; since the end of the event is not reached, the present tense is preferred over the perfective past, because it leaves open the possibility for the event to be continued and the end-point to be reached in the near future. Children, on the other hand, consistently use the perfective past to express all telic verbs, irrespective of their completion. So, the actual completion of a telic event does not play a role for none of the age groups in our study. 226 Sophia Delidaki 5. Conclusions and discussion The primary aim of the present study was to investigate the role of four se- mantic features (telicity, duration, frequency and completion of a telic event) in the choice of tense and grammatical aspect in the production of Greek- speaking children and adults. Our basic findings can be summarized as fol- lows : a) A predicate’s lexical aspect (telic / atelic) influences the choice of tense and grammatical aspect. Children and adults use more frequently pre- sent tense and imperfective morphology to describe atelic events, while they primarily use past tense and perfective morphology to describe telic events. b) Duration seems to influence children and adults choice mainly when the verbs encountered are atelic and it only slightly affects adults’ choice with telic ones. When the verb is atelic, children mainly use the present tense to describe an action lasting more than three seconds and the past perfective tense to describe an action lasting less than three seconds. For adults the results are similar except for the fact that when the verb lasts less than three seconds they almost equally divide their answers between the perfective past and the present tense. c) Frequency does not appear to influence adults and older children, while it does play a role in the choice of tense for the younger group of children. d) The success or not of a telic event reaching completion does not seem to influence any of the groups involved in our study. The most intriguing finding was that adult native speakers of Greek pre- sent in their answers the same pattern that is characterized as the AFH in language acquisition. In particular, the adults in our study combine perfective morphology and past tense with telic verbs and imperfective morphology and present tense with atelic verbs, as frequently as children do. This finding creates a confusion regarding the role of the AFH, transferring it from the acquisition field to the adult grammar, at least for the Greek language. L. Wagner (2003) also presents adult data from two comprehension tasks performed to English native speakers, where she found that within the narrow ceiling performance of the adults, the rare mistakes that they made did not belong to the temporal / aspectual pattern that the AFH suggests. Maybe it is the case that data like these lead towards the fact that the combination that the LAFH suggests is an issue of performance choice rather than a competence phenomenon and not a pattern of acquisition, but part of the grammar. References Antinucci, F. ; Miller, R. (1973). How children talk about what happened, Journal of Child Language 3 : 167-189. Berman, R. (1983). Establishing schema : Children’s construals of verb-tense marking, Language Sciences 5 : 61-78. The acquisition of aspect in child Greek 227 Bloom, L. ; Lifter, K. ; Hafitz, J. (1980). Semantics of verbs and the development of verb inflection in child language, Language 56(2) : 386-412. Bronckart, J. P. ; Sinclair, H. (1976). Time, Tense and Aspect, Cognition 2(1) : 107-130. Comrie, B. (1985). Tense, Cambridge : Cambridge University Press. Comrie, B. (1976). Aspect, Cambridge : Cambridge University Press. Dowty, D. R. (1979). Word meaning and Montague grammar : the semantics of verbs and times in generative semantics and in Montague’s PTQ, Dodrecht : D. Reidel. Fraisse, P. (1948). Etude compare de la perception et de l’estimation de la durée chez les enfants et les adultes, Enfance 1 : 199-211. Piaget, J. (1969). The Child’s Conception of Time, English translation by A. J. Pomerans, London : Routledge and Kenan Paul. Shirai, Y. ; Andersen, R. (1995). The acquisition of tense-aspect morphology : A Prototype Account, Language 71(4) : 743-762. Smith, C. (1991). The parametre of aspect, Dodrecht : Kluwer Academic publishers. Stephany, U. (1981). Verbal Grammar in Early Modern Greek Child Language, in : P. S. Dale ; D. Ingram, (eds), Child Language : An International Perspective, Baltimore : University Park Press, 45-57. Vendler, Z. (1967). Linguistics in Philosophy, Ithaca NY : Cornell University Press. Wagner, L. (2003). The influence of prototypes on linguistic judgments of tense and aspect, Paper presented at the Summer LSA Workshop : It’s about time, East Lansing MI USA, July 18-20. Weist, R. ; Wysocka, H. ; Witkowska, K. ; Buczowska, E. ; Konieczna, E. (1984). The defective tense hypothesis : On the emergence of tense and aspect in child Polish, Journal of Child language 11 : 347-374. This page intentionally left blank The Thai c1a 1 : A marker of tense or modality ?* Jiranthara SRIOUTAI University of Cambridge 1. Introduction The word c1a in Thai is often translated as will in English and behaves simi- larly to will both syntactically and semantically. That is, it occurs before a verb and may combine with verbs from all aspectual classes. Like will, it can have future time reference as in (1), may co-occur with an expression of epis- temic necessity, as in (2), and may express dispositional necessity, as in (3). (1) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a:I khu’:n ph3r3ungIn3i:II Mary c1a go see opera night tomorrow ‘Mary will go to the opera tomorrow night.’ (2) m3ae:r3i:I kh3ong c1a d1u: ’1op1e:r3a:I y3u:I t1o’:nn3i:II Mary may c1a see opera PROG now ‘Mary will be in the opera now.’ (3) b1a:ngkh3r3angII m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a:I n3ai2 ch3udw3o’m sometimes Mary c1a go see opera in tracksuit ‘Mary will sometimes go to the opera in her tracksuit.’ It is frequently argued that will expresses not the placement of an eventuality on the time line in relation to some specified point, but the degree of the speaker’s commitment to something which is said. C1a, by contrast, is often classified as a future tense morpheme. However, the purpose of this paper is to study the use of c1a in naturally occurring language use including news articles, magazine articles, journals, excerpts from fiction and conversations among Thai native speakers and propose that it is a marker of modality rather than a tense marker. The paper is organised as follows. First, I mention some previous treatments of c1a, and say why they are less than adequate. Then, I present 1 The Thai transliteration system used here is that which is found in Diller (1996). * I would like to express my deepest appreciation to my supervisor Dr. Kasia Jaszczolt for her insightful guidance and comments and heartfelt thanks to my family and friends for their unfailing support and encouragement. © Cahiers Chronos 17 (2007) : 229-239. 230 Jiranthara Srioutai my analysis of c1a as a modal and not a tense marker, and discuss how it may be represented in Discourse Representation Theory. Lastly, I conclude that the facts that (i) c1a does not stand for the future tense, (ii) it optionally co-occurs with expressions of various types of modalities, and (iii) it is essential in expressing possibility in the past and giving rise to counterfactuality in the absence of modal expressions seem to suggest strongly that all its uses involve modality and that it is thus better analysed as a modal marker. 2. Previous Treatments of C1a 2.1. A future tense morpheme C1a has been analysed as both an absolute and a relative future tense morphemes. Scovel (1970) and Supanvanich (1973) agree that it conveys an absolute future tense in (4), the future relative to the utterance time. (4) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a:I khu’:n ph3r3ungIn3i:II Mary FUT go see opera night tomorrow ‘Mary will go to the opera tomorrow night.’ According to Kanchanawan (1978), c1a, as in (5), is a relative tense marker. That is, it refers to the future that has a point of time other than the utterance time as its deictic centre. In (5), the deictic centre is a past time point. (5) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a:I m3u’aIkhu’:nn3i:II Mary FUT go see opera last night t1ae:I f3o:n t1o:k h2n3ak l3oe:y m3ai1I d1ai1II p1ai1 but rain fall heavy so not can go ‘Mary would have gone to the opera last night. But it rained heavily so she did not go.’ 2.2. An aspect morpheme Muansuwan (2002) proposes that c1a, as in (6) is an aspect morpheme that means ‘be about to’ and thus encodes the preliminary stage of an eventuality. (6) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a: Mary be about to go see opera ‘Mary was / is / will be about to go to the opera.’ The Thai cla: A marker of tense or modality? 231 2.3. A modal Rangkupan (2000) is by far the most explanatorily adequate analysis presented here. C1a is identified as a modal. It may occur in various positions in a sentence below. It can occur as a single modal in simple sentences such as (7). (7) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a: khu’:n ph3r3ungIn3i:II Mary modal go see opera night tomorrow ‘Mary will go to the opera tomorrow night.’ Apart from occurring alone, c1a also co-occurs with a verb of seeming such as th3a:I ‘seem’, as in (8), which together with c1a, can be said to encode ‘possibility’. (8) d1u: th3a:I m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a: khu’:n ph3r3ungIn3i:II look seem Mary modal go see opera night tomorrow ‘It seems that Mary will go to the opera tomorrow night.’ Finally, c1a may occur after affective modals and seems to have various changing meanings depending on what affective modal it co-occurs with. That is, ’1o’:kc1a means ‘quite’, ch3akc1a means ‘start to’, k1u’apc1a means ‘almost’, kh3o:ngc1a means ‘might’, ’1a:tc1a means ‘may’, and n3a:Ic1a means ‘should’. One example is given in (9), where c1a follows n3a:I and together they mean ‘should’. (9) m3ae:r3i:I kh3uanc1a p1ai1 d1u: ’1op1e:r3a: khu’:n ph3r3ungIn3i:II Mary should go see opera night tomorrow ‘Mary should go to see the opera tomorrow night.’ 2.4. A lexical word C1a has not only been analysed as a grammatical word but also a lexical item. Savetamalaya (1988) says that it means ‘intend to’, as in (10). (10) m3ae:r3i:I c1a p1ai1 d1u: ’1op1e:r3a:I khu’:n ph3r3ungIn3i:II Mary intend to go see opera night tomorrow ‘Mary intends to go to the opera tomorrow night.’ 232 Jiranthara Srioutai To summarise, there seem to be five possible standpoints as far as the semantics of c1a is concerned. Does it express absolute or relative future tense, or lexical meaning ‘intend to’, encode aspectual meaning ‘be about to’, or modality, or is it ambiguous between these meanings ? The ambiguity position is easily rejected by Grice’s principle of Modified Occam’s Razor: Senses (linguistic meanings) are not to be multiplied beyond necessity (1978). It is more appealing to attempt to come up with a unified semantic account of c1a since it is clear that the apparent differences are not lexical, unlike, for example, the meanings of the word bank. Now I will give counter- examples to the analyses of c1a as an aspect morpheme, a tense morpheme, and a lexical word. Counter-examples of this sort are extremely easily found. All we need to do is to look back at the examples (1) – (3). First, in (1), c1a apparently does not encode the preliminary stage of an eventuality. In (2), it is clear that c1a does not express future time reference, and it does not seem right to say that c1a is always a future tense marker. Lastly, c1a does not always have the lexical meaning ‘intend to’ because it can also be used in (3), where it cannot possibly mean ‘intend to’. 3. C1a : A marker of modality In this section, more evidence will be given in favour of the analysis of c1a as a modal marker. That is, I agree with Rangkupan (2000). However, although Rangkupan (2000) also classifies c1a as a modal, she does not make it clear what semantic contribution c1a has on its own, and she does not seem to be interested in accounting for the optional nature of c1a, either. These are two points that I add to her analysis. 3.1. Will as a modal proposals Here I mention two proposals that treat will as a modal and not tense marker because the same arguments are applicable to c1a in Thai. The first one is by Enç (1996). Theoretically, will clearly has non-future uses, where it expresses epistemic necessity or propositional necessity. When it has future time reference, the modality involved is prediction. Other modal verbs such as may and must has future time reference, too. Empirically, will patterns differently from the past tense and patterns exactly like a future-shifting modal. Unlike the past tense, will does not exhibit sequence of tense effects. The present tense embedded under will and the past tense behaves differently, and the present tense embedded under modals such as must behaves like when it is embedded under will. Lastly, modals may come in pairs ; one expresses necessity and the other expresses possibility ; if will is treated as a modal expressing necessity, may, which expresses possibility, is the dual of will. The Thai cla: A marker of tense or modality? 233 Another proposal to treat will as a modal is that of Jaszczolt (2003). Will has future time reference in (11), expresses epistemic necessity in (12), and dispositional necessity in (13). The three readings are different in terms of their degrees of modality. In Default Semantics, the Primary Intention Principle (Jaszczolt, 1999) states that : The primary role of intention in communication is to secure the referent which can be either object or eventuality of the speaker’s utterance. As it corresponds to the weakest informative and referential intentions from the Primary Intention Principle, Will in (11) is most modal. Will in (12) is less modal as it corresponds to stronger intentions. Lastly, will in (13) is least modal as it corresponds to the strongest intentions. Therefore, will is modal because it displays this gradation of intentions. This gradation means that will has a unified semantics, which Jaszczolt (2003) represents by introducing a modal operator to Discourse Representation Theory. (11) Mary will go to the opera tomorrow night. (12) Mary will be in the opera now. (13) Mary will sometimes go to the opera in her tracksuit. If what Enç (1996) and Jaszczolt (2003) say about will is true, it is also the case with c1a. Therefore, it seems that, like will, c1a is not a tense but modal marker. However, more evidence can be found in support of c1a as a modal marker in Thai. The following sections deal with this evidence. 3.2. Cla does not stand for the future tense This section shows that c1a does not stand for the future tense even though it has future time reference. When it is present in an utterance with future temporal adverbs, as in (14), it looks as if it was an absolute future tense marker. However, as it is optional there, it is unlikely to be best treated as a grammatical marking of the temporal location of an eventuality. Besides, when c1a co-occurs with past temporal adverbs, that is, when the point of reference is shifted from the time of utterance to a point of time before the utterance time, as in (15a), and the meaning of counterfactuality may arise as in (15b), it is obligatory. This seems to suggest that it is essential in expressing possibility in the past and giving rise to counterfactuality. However, when c1a also co-occurs with past temporal adverbs but no counterfactuality is intended, i.e. when it seems to express what can be called a relative future tense, as in (16), it is optional. This again shows that it should not be treated as a future tense marker, whether it is taken to be an absolute or relative one. Lastly, (17) is given to confirm that the analyses of c1a as a future tense marker is not adequate as no future time reference seems to be present in the utterance. Rather, c1a co-occurs with kh3ong ‘may’, 234 Jiranthara Srioutai which, combined with the present temporal adverb t1o’:nn3i:II ‘now’, expresses epistemic necessity. (14) ph3i:Is2aow (c1a) p1ai1 ch3iangh2m3ai1I ph3r3ungIn3i:II older sister (c1a) go Chiang Mai tomorrow ‘My older sister will go to Chiang Mai tomorrow.’ (15) a. ph3i:Is2aow c1a p1ai1 ch3iangh2m3ai1I m3u’aIw3a:nn3i:II older sister c1a go Chiang Mai yesterday ‘My older sister might have gone to Chiang Mai yesterday.’ (15) b. ph3i:Is2aow c1a p1ai1 ch3iangh2m3ai1I m3u’aIw3a:nn3i:II older sister c1a go Chiang Mai yesterday t1ae:I f3o:n t1o:k h2n3ak l3oe:y m3ai1I d1ai1II p1ai1 but rain fall heavy so not can go ‘My older sister would have gone to Chiang Mai yesterday. But it rained heavily so she did not go.’ (16) ph3i:Is2aow (c1a) p1ai1 ch3iangh2m3ai1I m3u’aIw3a:nn3i:II older sister (c1a) go Chiang Mai yesterday l30e:y c1o’:ng t1uaw m3u’aI kh3u’:n k1o’:nI so book ticket when night before ‘My older sister went to Chiang Mai yesterday so she booked the tickets the night before.’ (17) ph3i:Is2aow kh3ong (c1a) p1ai1 ch3iangh2m3ai1I t1o’:nn3i:II older sister may (c1a) go Chiang Mai now ‘My older sister will be on the way to Chiang Mai now.’ 3.3. C1a is compatible with expressions of different types of modalities In fact, c1a may also optionally co-occur with expressions of other types of modalities. It can be noticed that in all these examples, time is left unspecified. This confirms that c1a does not contribute temporally to the sentences. The fact that c1a is compatible with markers for dispositional necessity in (18), epistemic necessity in (19), epistemic possibility in (20), deontic necessity in (21) and (22), and deontic possibility in (23), suggests The Thai cla: A marker of tense or modality? 235 that each of its uses here probably primarily involves a degree of modality, too. (18) th3oe: m3ak c1a b1onI w3e:l3a: r3ot t1it she often c1a complain time car stick ‘She would / will often complain when there is a traffic jam.’ (19) k1ae: c1a t1o’:ngII s2iac1ai2 n3ae:In3ae:I he c1a must sorry certainly ‘He must have been / must be / will certainly be sorry.’ (20) ’1a:c1a:n kh3ong c1a m3o’:ng w3aI teacher may c1a see that n3is2it p1en m3u’an kh3o’:mph3iwt1oe: student be like computer ‘Teachers may have thought / may think that students are like computers.’ (21) r3aw c1a t1ongII p1r3abp1r3ung r3ab1op k1a:ny3u’:mkh3u’:n we c1a must improve system circulation ‘We had to / must improve our circulation system.’ (22) r3aw kh3uan c1a m3i: w3e:l3a: m3a:k kh2u’nII we should c1a have time much up ‘We should have had / should have more time.’ (23) r3aw (c1a) d1ai1II p1ai1 d1u: l3akh3o’:ns2at d1uayIIk1an we (c1a) can go see circus together ‘We were allowed to / may go to the circus together.’ 3.4. When c1a is optional and obligatory C1a is optional when it expresses absolute futurity, as in (14), relative futurity, as in (16), and when it co-occurs with expressions of different types of modalities, as in (18) – (23). On the other hand, c1a is obligatory when, in the absence of expressions of modalities, it co-occurs with past temporal adverbs, as in (15a) and may give rise to counterfactuality, as in (15b). This pattern seems to show that it is compulsory only when it co-occurs with past temporal adverbs when there are no other modal expressions around and when it gives rise to unreal conditional, or counterfactuality. This in turn strongly suggests that its uses definitely involve modality. It is thus better to analyse c1a as a modal marker. 4. C1a in Discourse Representation Theory This section discusses how c1a may be represented in Discourse Representation Theory (DRT). To represent grammatical tenses in English, DRT assigns the value of the feature TENSE according to the verb tense of a sentence. According to Kamp and Reyle (1993 : 512-513) : 236 Jiranthara Srioutai … TENSE has three possible values, past, present, and future, signifying that the described eventuality lies before, at, or after the utterance time, respectively. The value of TENSE for a given sentence S is determined by the tense of the verb of S. When the main verb is in the simple past, TENSE = past ; when it is in the simple present, TENSE = pres ; and when the verb complex contains the auxiliary will, TENSE = fut. … To represent fut, DRT uses n < t, where n ‘now’ refers to the utterance time, t refers to the eventuality time, and < stands for temporal precedence. Now figure 1 (below) is the Discourse Representation Structure (DRS) of (17). If c1a is a future tense marker, n < t should be present in the DRS. But to put it there would be incompatible with the meaning of (17) arrived at compositionally. Rather, the temporal adverb t1o’:nn3i:II ‘now’ introduces n = t in the DRS. The other piece of temporal information in this DRS is e ⊆ t. To represent kh3ong (c1a) there, I follow Jaszczolt (2003, 2005), who places will in the framework of Grice’s Equivocality Thesis : epistemic and deontic modalities are univocal, derived from one concept of acceptability (2001). According to Jaszczolt (2003, 2005), Grice’s acceptability can be introduced n to DRT as a modal operator on eventualities in the form of ACCΔ p, which is read ‘it is acceptable of the type Δ to a degree n that p’, where Δ stands for a type of modality : epistemic or deontic. The formal way of introducing n ACCΔ to DRT is prosecuted in Jaszczolt (2005). The whole DRS can be read as, it is acceptable of the type Δ to a degree n that it is the case that the event of the speaker’s older sister going to Chiang Mai is temporally included in the eventuality time which lies at the same time as the utterance time. n e x y t n = t e ⊆ t older sister (x) Chiang Mai (y) now (t) n ACCΔ e e: x go to y Figure 1 : DRS for (17) The Thai cla: A marker of tense or modality? 237 Next, let us see how we may go about representing c1a when it is obligatory, that is, when it co-occurs with past temporal adverbs but there are n no other modal expressions. The modal operator ACCΔ may also be introduced in the the semantic representation of (15a) in (figure 2). The relationship between the utterance time n and the eventuality time t is specified by the temporal adverb m3u’aIw3a:nn3i:II ‘yesterday’ in (15a) and is represented by t < n in the DRS. The DRS can be read as, it is acceptable of the type Δ to a degree n that it was the case that the event of the speaker’s older sister going to Chiang Mai was temporally included in the eventuality time which lies before the utterance time. n e x y t t < n e ⊆ t older sister (x) Chiang Mai (y) yesterday (t) n ACCΔ e e: x go to y Figure 2 : DRS for (15a) Lastly, when c1a occurs with a modal expression but with no temporal n adverbs, the modal operator ACCΔ can be introduced to represent kh3uan (c1a). However, the construction of the DRS in figure 3, supposedly the semantic representation of (22), cannot be completed. This is because the eventuality time is not grammatically nor lexically specified in (22). The incompleteness of the DRS is shown in n ? t. We need to look at the context to see whether this should be replaced by t < n (for ‘should have had’), or n = t or n < t (for ‘should have’). This is as rightly observed by Diller (1993 : 412) : temporal and aspectual interpretation in Thai is usually ‘a matter of contextual interpretation’. To complete the DRS-construction in figure 3, one needs to establish how to incorporate utterance meaning available only from the context, and not just from lexicon and grammar, into the DRT representation. The fact that the construction of DRSs for sentences that contain c1a can be completed only when there are also temporal adverbs but cannot be completed when there are no temporal adverbs in them also suggests that c1a does not significantly make temporal contribution in the 238 Jiranthara Srioutai sentence where it appears. Once again, this is why it is not quite adequate to treat it as a tense marker. n s X y t n ? t s o t Interlocutors (X) more time (y) n ACCΔ s s: X have y Figure 3 : Incomplete DRS for (22) 5. Conclusion It has been shown that (i) c1a does not stand for the future tense, (ii) c1a optionally co-occurs with expressions of different types of modalities, and (iii) it is essential in expressing possibility in the past and giving rise to counterfactuality in the absence of other modal expressions. This suggests that perhaps all its uses involve a degree of modality, and so c1a is more adequately analysed as a modal marker than a tense marker. This is also confirmed when c1a is represented in DRT because in the absence of temporal adverbs the DRS cannot be completed. At this point, however, I must say that my DRSs here are simplified versions and are not final. Certain points for further research include how to n arrive at values for Δ and n of ACCΔ for different modal expressions and how to represent in DRT temporal utterance meaning that does not come from tenses or temporal adverbs. References Diller, A. (1993). Diglossic grammaticality in Thai, in : W. A. Foley (ed.), The Role of Theory in Language Description, New York : Mouton de Gruyter, 393-420. Diller, A. (1996). Thai and Lao Writing, in : P. T. Daniels ; W. Bright, (eds), The World’s Writing Systems, New York : Oxford University Press, 457-466. Enç, M. (1996). Tense and Modality, in : S. Lappin, (ed.), The Handbook of Contemporary Semantic Theory, Oxford : Blackwell, 345-358. The Thai cla: A marker of tense or modality? 239 Grice, H.-P. (1978). Further notes on logic and conversation, in : P. Cole, (ed), Syntax and Semantics, vol. 9, Pragmatics, New York : Academic Press, 113-128. Grice, H.-P. (2001). Aspects of Reason, R. Warner, (ed.), Oxford : Clarendon Press. Jaszczolt, K.-M. (1999). Discourse, Beliefs, and Intentions : Semantic Defaults and Propositional Attitude Ascription, Oxford : Elsevier Science. Jaszczolt, K.-M. (2003). The Modality of the Future : A Default-Semantics Account, in : P. Dekker ; R. van Rooy, (eds), Fourteenth Amsterdam Colloquium Proceedings, University of Amsterdam, 43-48. Jaszczolt, K.-M. (2005). Default Semantics : Foundations of a Compositional Theory of Acts of Communication, Oxford : Oxford University Press. Kamp, H. and U. Reyle (1993). From Discourse to Logic : Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and Discourse Representation Theory, Dordrecht : Kluwer. Kanchanawan, N. (1978). Expression for Time in Thai Verb and its Application to Thai-English Machine Translation, Ph.D. Dissertation, University of Texas at Austin. Muansuwan, N. (2002). Verb Complexes in Thai, Ph.D. Dissertation, University at Buffalo, State University of New York. Palmer, F.-R. (1986). Mood and Modality, Cambridge : Cambridge University Press. Rangkupan, S. (2000). Characteristics of Psychological Perspective in Thai Narrative Discourse, Ph.D. Dissertation, University of Buffalo, State University of New York. Savetamalya, S. (1988). A Reanalysis of Auxiliaries in Thai, University of Hawaii Working Papers in Linguistics 19 : 1-44. Scovel, T.-S. (1970). A Grammar of Time in Thai, Ph.D. Dissertation, University of Texas at Austin. Supanvanich, I. (1973). Tenses in Thai, Master’s Thesis, Chulalongkorn University. [In Thai]