Newsletter 10
No 10
June - 1997
CONTENTS
Meanings of I think: a study based on parallel corpora
Anne-Marie Simon-Vandenbergen
The expression I think has recently received a fair amount of attention from different angles. Since Urmson (1952) introduced the term 'parenthetical verb' to designate such verbs as think, believe, suppose, guess, which in the first person singular simple present tense can assume different positions in the sentence (I think that house belongs to his father; That house, I think, belongs to his father; That house belongs to his father, I think ), a number of studies have examined these parenthetical expressions, and especially I think from the points of view of syntax and semantics (e.g. Hooper 1975), grammaticalisation (especially Thompson & Mulac 1991) and pragmatics (Aijmer 1997; Simon-Vandenbergen forthcoming (a)). In a recent talk (Simon-Vandenbergen forthcoming (b)), I looked at I think as it is used in the discourse of speakers in the European Parliament and contrasted it with the Dutch equivalents as found in the translations as well as in the original Dutch speeches. In the present paper I shall restrict myself to briefly touching upon one point, viz. to showing how contrastive analysis can contribute to a better understanding of expressions in a particular language. In the same way that diachronic data may help in grasping present-day phenomena, data gathered from contrasting different languages may shed light on phenomena in the individual languages concerned.
The data which I looked at for this study have been taken from the Debates of the European Parliament (May and October 1993), which form a subcorpus of the trilingual parallel corpus compiled by Hans Paulussen (cf. Hans Paulussen 1995 for a description of the corpus). The choice of this corpus was motivated by two factors: from a 'monolingual' point of view I was interested in the use of I think by political speakers; from a contrastive point of view the corpus was ideal because the English speeches were available in their Dutch translations while at the same time the original Dutch speeches formed a parallel corpus and were available in English translation as well.
The hypothesis, formed on the basis of I think in political interviews (Simon-Vandenbergen forthcoming (a)), was that the expression would be primarily used in its 'deliberative' rather than in its 'tentative' function. The distinction between these uses can be made on the basis of syntactic, lexical, prosodic, semantic and pragmatic criteria (Aijmer 1997; Simon -Vandenbergen forthcoming (a)). The conclusion I was led to draw from the analysis of the debates was that in this genre I think is virtually exclusively used to signal authority rather than doubt or uncertainty. In terms of the nature of the proposition which is 'projected' or qualified by I think , I found it useful to distinguish between a 'factual' thesis, which is verifiable, and an evaluative one, which is non-verifiable. These possibilities are exemplified by:
a) I think he's at home now (factual; verifiable)
b) I think you're looking great (evaluative; non-verifiable)
This distinction between the uses in a) and b) above has been referred to by Aijmer as a distinction between the meanings of 'belief' and 'opinion' respectively. It is clear that, while in both cases the speaker adds his or her subjective qualification in terms of the reservation 'this is my own point of view with regard to the state of affairs', it is only in the case of 'belief' that the speaker uses the metaphor (in Halliday's 1994 sense of the term) to signal probability. While a) can be paraphrased as 'He's probably at home now' or 'He will be at home now', such a paraphrase is inappropriate for b): ? 'You're probably looking great' or ? 'You will be looking great'. It can further be argued that the meaning of 'tentativeness' may (but need not) be present in utterances which are qualified by I think expressing 'belief', while I think expressing 'opinion' does not signal uncertainty or tentativeness at all. An investigation into the types of propositions qualified by I think in the debates corpus has shown that the 'belief' meaning in the sense explained and exemplified above does not occur, except in some pragmatically special cases which for reasons of space I will not go into here. The most frequent types of proposition in the corpus examined are those expressing a subjective evaluation (type b) above) and those expressing an obligation or necessity, i.e. the speaker's reference to what needs to be done in the future. The latter type of cases cannot be classified as either 'belief' or 'opinion' but seem to be mergers of the two meanings. In one interpretation ('belief') the speaker says that it is probably necessary to act in a certain way, in the other ('opinion') that he finds it necessary to act in a certain way, but the distinction between these two meanings is not only very small but can in fact not be made in practice. In this sense my data suggest the usefulness of replacing the categorial distinction between 'belief' and 'opinion' by a distinction between 'belief', 'evaluation' and 'belief/opinion'. The merger type is illustarted in the following example from the corpus:
I think what has to be done - and it is for the Commission to do this - is to take off the velvet glove (DAEE:17:01)
In conclusion, the speakers in the debates used I think authoritatively to express either their subjective evaluation of a past or present state of affairs or their belief/opinion regarding a necessary course of action in the future.
The Dutch translations gave further support to the usefulness of the semantic distinctions which emerged from the analysis of the English data. The most frequent translation of I think was, not surprisingly, ik denk (33.50%), followed by ik geloof (14%), volgens mij (12%) and ik vind (9.50%). The verbs denken, geloven and vinden are obviously not synonymous. A questionnaire presented to native speakers of Dutch showed the following main similarities and differences:
1) Both ik denk and ik geloof are used with the meaning of 'belief' in the sense explained above, i.e. to qualify a factual thesis which is in principle verifiable:
Ik denk dat hij thuis is.
= I think he's at home.
Ik geloof dat de buren met vakantie zijn.
= I think the neighbours are on holiday.
In contrast, ik vind is not acceptable in this type of context. Such sentences as the following were unanimously rejected by informants:
? Ik vind dat dit zijn wagen is maar het kan ook die grijze zijn.
= I think this is his car but it might also be that grey one.
2) The meaning of ik vind is 'subjective evaluation', i.e. an opinion regarding a state of affairs which is based on personal experience:
Ik vind dat deze jurk je goed staat.
= I think this dress suits you.
In such contexts informants unanimously rejected ik denk and ik geloof:
?Ik denk dat dit lelijke schoenen zijn.
= I think these shoes are ugly.
?Ik geloof dat hij zo vriendelijk is.
= I think he's so nice.
3) All three verbs were accepted by informants in contexts of 'opinion based on reasoning', as in:
Ik denk /ik geloof/ ik vind dat dit de enige juiste politiek is.
= I think this is the only right policy.
This also applies to instances where reference is made to what is necessary, advisable:
Ik denk/ ik geloof/ ik vind dat we nu snel moeten handelen.
= I think we must act quickly now.
It is this type of I think which expresses a merger of 'belief' and 'opinion', as suggested above.
4) The sentences which had been constructed to find out about the difference between ik denk and ik geloof gave less clear results, i.e they never elicited unanimous responses. Nevertheless, it was possible to discern tendencies which seem to point to the plausibility of hypotheses which need to be further tested, also on the basis of corpus research. One difference is that the verbs retain features of their process meanings, which leads to a distinction between:
a) Ik geloof dat God bestaat.
= I believe that God exists.
b) Ik denk dat God bestaat.
= I think that God exists.
In other words, although both ik denk (I think) and ik geloof (I believe) have in some contexts developed into markers of modality (cf. Thompson & Mulac 1991), they retain their process meanings in other contexts. Consequently, some sentences will allow both verbs with equal probability of occurrence but with a difference in meaning, others will favour one or the other verb. Sentences which seemed to favour ik geloof were those in which the meaning of 'believe in someone/something' was more clearly present, such as:
Ik geloof vast en zeker dat hij zal slagen voor het examen.
= I firmly believe that he will pass the exam.
Collocational tendencies are an important line of research to be pursue in this respect.
Returning to the debates data, we can now explain the findings referred to above: since the context of 'I think + verifiable factual thesis' does not occur in the data (except for a few special pragmatic functions), there are few cases in which ik vind is excluded as a possible translation of I think. That ik denk and ik geloof are nevertheless more frequent translations is because first of all ik denk is the prototypical equivalent of I think, and secondly because, as is the case with English I think and I believe, Dutch ik denk and ik geloof are in many contexts perfectly interchangeable.
Data from the parallel corpus of speeches given in Dutch provide further interesting information regarding the meanings typically expressed by speakers in this sort of context. It appears that while ik denk is still the most frequent verb, ik vind is much more frequent than ik geloof and in fact comes very close to ik denk in terms of frequency. In addition, it appears that the most frequent translation of ik vind is not I find but I think.
In conclusion, both an analysis of the meanings of I think in the English original and the translations into Dutch indicate that the typical use of I think in this genre is not as a hedge or as a marker of tentativeness but as an indicator of authoritative statement of belief/opinion. Further support for this is provided by the Dutch parallel corpus.
Acknowledgement
I am grateful to Hans Paulussen for having provided me with the data from his trilingual parallel corpus. In spite of his own busy schedule he most kindly devoted some of his precious time to excerpting the texts. I also wish to thank the 25 informants, colleagues and students, who helped me by filling in the questionnaires.
References
- Aijmer, K. (1997). 'I think an English Modal Particle'. In: T. Swan and O. J. Westvik (eds). Modality in Germanic Languages. Historical and Comparative Perspectives. Berlin, New York: Mouton de Gruyter, 1-47.
- Halliday, M. A. K. (1994). An Introduction to Functional Grammar. London: Edward Arnold.
- Hooper,J.B. (1975). 'On Assertive Predicates'. In: J.B.Kimball (ed). Syntax and Semantics. Volume 4. New York: Academic Press, 91-124.
- Paulussen, H. (1995). 'Compiling a Trilingual Parallel Corpus'. Contragram 3, 10-13.
- Thompson, S. A., Mulac, A. (1991). 'A Quantitative Perspective on the Grammaticization of Epistemic Parentheticals in English'. In: E. Closs Traugott and B. Heine (eds). Approaches to Grammaticalization. Volume II. Amsterdam/Philadelphia: John Benjamins, 313-329.
- Simon-Vandenbergen, A.-M. (forthcoming (a)). 'The Functions of I think in Political Discourse'. Paper presented at the 3rd NIC symposium, Aalborg 20-23 November 1996.
- Simon-Vandenbergen, A.-M. (forthcoming (b)). 'The Modal Expression I think and its Dutch Equivalents'. Paper presented at the 4th Nordic Symposium on Text-based Contrastive Studies, Oslo 25-27 April 1997. To be published in: S. Johansson and S. Oksefjell (eds). Computer Corpora and Crosslinguistic Research. Rodopi.
- Urmson, J. O. (1952). 'Parenthetical Verbs'. Mind, Vol. LXI, No.244.
Les objets en grammaire
Ludo Melis
L'objet en grammaire peut-il être un sujet de colloque linguistique? A cette question, les communications nombreuses et variées et les discussions animées qu'elles ont suscitées apportent une réponse positive franche, mais, si cette notion peut faire l'objet d'investigations scientifiques et de débats, elle appelle également des mises en garde qui peuvent mener à sa mise à l'écart de la description rigoureuse pour cause d'imprécision ou d'ambigüité inhérente. Cette situation paradoxale, l'objet - partenaire privilégié du prédicat élémentaire - la partage avec d'autres instruments de base de l'arsenal linguistique, tels la phrase, le mot et le sujet: intuitivement clairs, ils s'imposent comme des pivots incontournables de l'analyse, mais pour qui veut s'armer de définitions strictes et de procédures fiables, ils sont insaisissables et évanescents.
Le caractère fuyant de la notion d'objet a été mis en évidence, lors du colloque, à trois niveaux au moins. L'attention a, en premier lieu, été attirée sur la lente et confuse émergence du concept au cours du XVIIIe siècle et sur les liens que le terme grammatical entretient avec l'objet en philosophie, objet de perception tangible, substantiel et donc nominal. Les divergences terminologiques, de langue en langue et d'approche linguistique en modèle théorique, témoignent également de la difficulté à s'entendre sur l'objet; si en anglais object et complement se partagent par exemple le champ des constituants centraux nucléaires, les deux termes sont associés en français, alors que la terminologie du néerlandais privilégie le terme d'objet, voorwerp, au détriment de celui de complément et l'oppose directement au terme de bepaling ou détermination accessoire. Enfin, le domaine d'application du terme varie: certes, l'existence d'objets du verbe ‘transitif' semble être généralement reconnue, mais faut-il admettre l'existence d'objets de l'adjectif ou du nom et, dans le domaine des objets verbaux, y a-t-il lieu de distinguer divers objets et selon quels paramètres? Toutes ces imprécisions, qui font qu'un certain flou artistique entoure la notion et qui offrent un fondement au jugement péremptoire, bien connu, de M. Gross selon qui le complément d'objet n'existe pas, n'ont pas empêché les linguistes d'utiliser le terme, certes avec des fortunes diverses et dans des acceptions distinctes, mais surtout avec persévérance et même avec opiniâtreté.
Face au tableau confus que j'ai essayé d'évoquer, les discussions du colloque montrent que le bon usage de la notion d'objet en linguistique doit répondre à trois exigences: efficacité, description relationnelle et démarche différentielle.
L'analyse linguistique n'impose pas le recours à la notion d'objet; celle-ci n'est pas incontournable. Elle n'est pas non plus inutile, à condition toutefois que sa mise en oeuvre permette de décrire certaines régularités. Ces régularités peuvent être formelles et syntagmatiques, tels les règles d'accord ou les phénomènes d'ordre des constituants, ou paradigmatiques, comme les alternances entre constructions. Chaque régularité mène cependant à circonscrire autrement la notion et il y a en quelque sorte autant d'objets que de régularités à décrire. De ce point de vue, il n'existe pas d'objet en linguistique, mais des objets, définis en fonction d'observations spécifiques et dont les caractéristiques varient en conséquence de langue en langue et de phénomène en phénomène.
Ces objets forment une famille d'êtres grammaticaux proches, mais distincts, liés par une ressemblance de famille dont le fil rouge est leur caractère nucléaire et l'association privilégiée avec le prédicat dont ils sont le partenaire le plus proche.
Cette caractéristique fonde l'exigence d'une description relationnelle, en étroite liaison avec le prédicat et donc avec les types de prédicat. On différenciera par conséquent les objets du verbe dit transitif, prototypes de la catégorie, et pourquoi pas du verbe copule, extension du domaine peut-être inattendue mais qui peut être argumentée, de l'adjectif et du nom. L'approche relationnelle impose en outre la mise en rapport des objets à travers le paradigme des diverses réalisations d'une même unité, paradigme des constructions d'un même verbe et l'on retrouve ici le jeu des alternances, paradigme aussi des réalisations d'une même unité sous-jacente comme verbe, comme nom ou comme adjectif.
Il ne suffit toutefois pas de poser que l'objet doit être décrit en relation avec le prédicat. Ici aussi les diverses contributions au colloque ont montré qu'il y a plutôt un faisceau de relations qui à nouveau coïncident dans une large mesure, mais ne se superposent pas et qui imposent dès lors une approche diversifiée de l'objet, des objets. Se plaçant au niveau proprement syntaxique, l'analyste retiendra au moins trois dimensions. La première est la dimension de la présence et de l'absence avec ses diverses modalités et effets de sens. La seconde concerne le caractère essentiel, actantiel, ou accessoire; l'objet dit interne apparaît dans ce contexte comme un constituant accessoire, mais qui partage d'autres propriétés avec l'objet nucléaire. La troisième dimension a trait au mode de complémentation: il convient de distinguer le mode incorporé, moyen ou normal et autonomisé, souvent souligné par une marque prépositionnelle.
Sur le plan sémantique, au moins deux couches distinctes sont à prendre en considération, celle des rôles liés aux schémas d'action et dans une certaine mesure indépendants des prédicats lexicaux et celle du couple affecté / effectué, qui entretient des rapports complexes avec les propriétés aspectuelles. Enfin, la relation du prédicat verbal et de l'objet doit être scrutée du point de vue des contraintes qu'ils s'imposent ou de la tolérance mutuelle dont ils font preuve. La complexité des phénomènes évoqués, qui doivent trouver leur place dans une description fine et précise de la valence - thème qui a été plus d'une fois abordé, e.a. du point de vue des instruments adéquats pour rendre les propriétés essentielles, tant relationnelles que catégorielles, des objets-, montre, de nouveau, qu'une approche différenciée s'impose. Tout comme il n'y a pas un objet, il n'y a pas une seule relation du prédicat et de l'objet.
L'approche grammaticale classique, opérant avec une définition unique, claire et bien circonscrite, est battue en brèche par l'ensemble des observations qui ont été évoquées plus haut; elle l'est encore plus, si l'on tient compte des plaidoyers éloquents en faveur d'une approche différentielle des relations fondamentales dans la phrase. Plusieurs intervenants ont mis en évidence qu'il existe une asymétrie nette entre le sujet et l'objet dans leur relation au verbe et à la phrase, ainsi qu'entre l'objet et d'autres constituants nucléaires comme le ‘datif'. Cette asymétrie met en jeu des phénomènes d'ancrage dans le discours, de pragmatique, de syntaxe et de sémantique. Il en ressort qu'il ne faut pas seulement tenir compte d'un écheveau de phénomènes subsumés pour la commodité sous le nom d'objet, mais qu'il est illusoire d'approcher l'objet avec les instruments qui sont adéquats pour traiter du sujet, du datif ou d'un autre constituant phrastique. Les faits à décrire sont différenciés et leurs modes d'appréhension doivent être diversifiés. La seule analyse qui convienne doit offrir libre champ à la différence.
L'éclatement des notions, des relations, des définitions, des approches qui est le fruit des travaux présentés lors du colloque sonne-t-il la mort de la notion d'objet et la question initiale recevrait-elle donc finalement une réponse négative? Je ne le crois pas. La leçon du travail de déconstruction semble se situer ailleurs. Il me paraît que c'est l'instrument classique qui a volé en éclats, celui qui prend appui sur une architecture massive faite de notions bien solides, claires et distinctes, valables à tous les niveaux, et que cet instrument unique est à remplacer par une panoplie plus fine, plus souple, plus diversifiée, mais aussi plus difficile à manier, car bien plus délicate. Le remplacement impose une conception plus granulaire des phénomènes; les unités obtenues par l'analyse y entrent dans des réseaux multiples et fonctionnent à différents niveaux tissant une toile multidimensionnelle de rapports. A l'édifice massif se substitue un ensemble composite, en apparence moins structuré, mais en réalité unifié par d'innombrables liens particuliers, que l'analyste doit dégager un à un et dont il doit estimer l'apport spécifique au fonctionnement de l'ensemble. Un tel édifice est fait de matériaux composites; les ingénieurs nous disent que de tels matériaux, qui ont l'avenir devant eux, allient la légèreté et la souplesse à la solidité et à la qualité. Qu'il puisse en être également ainsi en grammaire.
Book notice
Bart Defrancq
Schøsler, L. and S. Kirchmeier-Andersen (1997) Studies in Valency II. The Pronominal Approach Applied to Danish. (= RASK Supplement 5.) Odense: Odense University Press. viii + 208 pp.
This book is the second volume of the RASK Supplement series dedicated to valency studies. Whereas the first volume (Schøsler (1995)) offered a more comprehensive look on valency theories, the second one concentrates on the pronominal approach. This does not mean that the authors have dramatically narrowed their views; on the contrary, Studies in Valency II, is a study of the pronominal approach against the background of other approaches to valency, which are briefly sketched and commented in Chapter 1. Chapter 2 offers a detailed theoretical picture of the pronominal approach applied to Danish, while Chapter 3 illustrates its implementation in the framework of the Odense Valency Dictionary (OVD).
It is certainly a formidable and probably very unrewarding task to synthesize the many existing theories about valency: it once took Welke (1988) more than two hundred pages to present valency descriptions undertaken in the former GDR. That is probably the reason why Chapter 1, which aims at classifyinig some approaches according to their treatment of the fundamental problems of valency description, only provides a sketchy overview of some theories, criticizing their weakest aspects, especially their way of linking the syntactic to the semantic level and their definition of valency elements. The pronominal approach is said to offer at least a partial solution for both of these problems. On the one hand, it is based on semantic information provided by language-immanent (thus less ‘subjective') elements; on the other hand, the organisation of pronouns in paradigms makes it possible to achieve an economic distributional analysis in order to determine the valency status of the constituents under analysis. Note, however, that the approach is pragmatic: the authors do not hesitate to include a more classic parameter like omissibility (p. 30).
A more detailed description of the Danish pronominal ‘machinery' is offered in Chapter 2. The ultimate aim of this description is to determine which pronominal elements are most relevant in the framework of valency description, but the interest of some of its elements is such that they deserve to be highlighted. First of all, the pronominal category is expanded with a series of elements that traditionally are not considered to be pronouns, but share with pronouns an important characteristic: proportionality with lexical constituents. Secondly, the description of the semantic features displayed by the pronominal elements is exhaustive (pp. 46-47). Finally, the importance of suspensive (= interrogative) pronouns in the Danish approach is striking, especially when considering the applications of the pronominal approach to other languages. Oversimplifying, the interrogative pronouns are held responsible for the basic semantic framework of the whole system, a position that differs substantially from the one resulting from analyses of the French pronominal system (Blanche-Benveniste (1984) and Melis (1994)), but converges to a certain extent with Jackendoff's theory about the basic constituents of conceptual structure. Interrogative pronouns are also taken to be the core elements of the valency description, but the authors admit that the information they (and other pronouns) provide does not always suffice: there is a consistent group of non-full verbs that cannot be described through pronominal paradigms and even the full verbs cannot always be described in sufficient detail in order to distinguish intuitively felt meaning distinctions. That is why the instrumentation of the OVD has been enlarged with other criteria like passive, verbal aspect, existentials and extraposition.
The implementation of the theory is the purpose of Chapter 3. It involves two Danish verbs: fylde (fill) and få (get), chosen because they display features that have always been important in valency description: linked constructions (= Levin's ‘alternations') and the difference between full verbs and non-full verbs. The description of the valency patterns of fylde and få is meticulously compared, on the one hand, with descriptions provided by Danish dictionaries and, on the other hand, with corpus data from a general language and a (limited) special language corpus. This seems to be the riskiest aspect of the whole methodology: however fine-grained the theoretical description is, the confrontation with a corpus puts at least its semantic aspects fundamentally into question and it cannot be denied that some of the corpus data included here do (the corpus examples included for fylde 50.003, for instance, seem to suggest that this is a ‘waste-basket' reading of the verb). A feed-back from the corpus to the OVD could deal with this problem.
On the whole, Studies in Valency II is a the result of a thorough reflection on one of the methods of describing valency. Like other methods, it is based on a sound fund of pragmatism, but adds to a comprehensive approach (1000 verbs sofar) a fair corpus analysis, which is a new and laudable effort. It makes an important contribution to the pronominal approach and promises interesting results in the field of functional verbs (to appear in Studies in Valency III...?).
References
- Blanche-Benveniste, C., J. Deulofeu, J. Stefanini and K. van den Eynde (1984) Pronom et syntaxe. L'approche pronominale et son application au français. Paris: SELAF.
- Melis, L. (1994) Paradigmes de la valence verbale et réalisations nominales et pronominales. International Journal of Lexicography 7, 2: 142-157.
- Schøsler, L. and M. Talbot (1995) Studies in Valency I. Odense: Odense University Press.
- Welke, K. (1988) Einführung in die Valenz- und Kasustheorie. Leipzig: Bibliographisches Institut.
To the table of contents of other CONTRAGRAM issues