Narrative discourse

ven d Germanic speakers k c ative discourse 1. Introduction Journal of Pragmatics 42 (2010) 3328–3344 A R T I C L E I N F O Article history: Received 16 March 2009 Received in revised form 27 May 2010 Accepted 29 May 2010 Keywords: Information structure Discourse cohesion Anaphoric linkage Scope particles Cross-linguistic comparison Romance languages Germanic languages A B S T R A C T This paper deals with the anaphoric linking of information units in spoken discourse in French, Italian, Dutch and German.We distinguish the information units ‘time’, ‘entity’, and ‘predicate’ and specifically investigate how speakersmark the information structure of their utterances and enhance discourse cohesion in contexts where the predicate contains given information but there is a change in one or more of the other information units. Germanic languages differ from Romance languages in the availability of a set of assertion-related particles (e.g. doch/toch, wel; roughly meaning ‘indeed’) and the option of highlighting the assertion component of a finite verb independently of its lexical content (verum focus). Based on elicited production data from 20 native speakers per language, we show that speakers of Dutch and German relate utterances to one another by focussing on this assertion component, and propose an analysis of the additive scope particles ook/auch (also) along similar lines. Speakers of Romance languages tend to highlight change or maintenance in the other information units. Such differences in the repertoire have consequences for the selection of units that are used for anaphoric linking. We conclude that there is a Germanic and a Romance way of signalling the information flow and enhancing discourse cohesion. � 2010 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Journal of Pragmatics In order to understand their information structure, utterances or stretches of discourse are often analysed as if they were answering an (explicit or implicit) question (Carroll and Lambert, 2003; Erteschik-Shir, 2007; Givón, 1983; Klein and von Stutterheim, 2002; Lambrecht, 1994). Consider the fictive discourse in (1). (1) Context: There is a fire in the house of Mr. Red, Mr. Green and Mr. Blue 1 Here comes Mr. Red 2 He calls the fire brigade 3 He then jumps out of the window 4 and tries to warn his neighbours. . . Christine Dimroth a,*, Cecilia Andorno b, Sandra Benazzo c, Josje Verhagen a aMax Planck Institute for Psycholinguistics, Netherlands bUniversità degli Studi di Pavia, Italy cUniversité Lille 3, France lin Both the like ‘‘Wh * Corresp E-mail 0378-2166 doi:10.101 hanged and maintained information in narr Gi claims about new topics. How Romance an journal homepage: www.elsev ier .com/ locate /pragma discourse and the individual utterances it consists of can be understood as answering an implicit discourse question at happened then to X?’’ (Klein and von Stutterheim, 2002). This results in a prototypical narrative structure in onding author. Tel.: +31 24 3521443; fax: +31 24 3521213. address: christine.dimroth@mpi.nl (C. Dimroth). /$ – see front matter � 2010 Elsevier B.V. All rights reserved. 6/j.pragma.2010.05.009 which the time talked about1 shifts from one utterance to the next, the protagonist talked about (henceforth: the entity) is maintained and the predicate that holds for the entity at the relevant times constantly changes. Discourse cohesion in narratives is often enhanced by anaphoric means signalling reference maintenance in the domain of entities (e.g. pronouns, zero-anaphora) and the default shift of time (including connectors, adverbials, morphological tense marking). Cross-linguistic differences have been found in the formal repertoire for the encoding of information structure (e.g. availability of zero-anaphora, flexible word order) as well as the licensing conditions for their use (Ahrenholz, 2005; Fox, 1987; von Stutterheim and Carroll, 2005). In this paperwe focuson information configurations that differ from(1)because the entities are constantly changingbut the his s Germanic (Dutch, German) languages. lysin hance discourse cohesion when talking t ch osite type occur. n our rance structure proposed by Dik (1997) e fr most elementary layer, so-called first r ele nguished. At a subsequent layer these C. Dimroth et al. / Journal of Pragmatics 42 (2010) 3328–3344 3329 elementary building blocks describe a state of affairs, called ‘‘predication’’, that is ‘‘the conception of something that can be the case in some world’’ (Dik, 1997, vol. 1:51) (e.g. Mr. Green’s jumping). Predications can also be located in space and time (e.g.Mr. Green’s jumping last Sunday).3 In addition to a predication layer every utterance has a ‘‘proposition’’ layer.4 Propositions have a topic-comment structure and an operator that establishes a relation between the topic and the comment. 1 The time span about which a claim is made corresponds to the notion of ‘‘topic time’’ proposed by Klein (1994). 2 Note that the notion ofmaintenancemeans different things when applied to different components of the utterance. While with entities and time spans maintenance means a co-reference to the exact same referent, in case of predicative expressions, what is maintained are the properties characterizing a given situation, not the situation in the external world. 3 A predication is comparable to Klein’s (2006) notion of an ‘‘assertable expression’’, also called ‘‘sentence base’’. 4 Dik (1997) distinguishes a fourth layer called ‘‘clause’’ (a speech act with an illocutionary force). We do notmake a distinction between proposition and clause here, and rather assume that the hold-for relation between the comment and the topic of a proposition can be asserted (in a declarative) but also left open (in a yes/no-question). orde ments such as entities and predicates (e.g. Mr. Green, to jump) are disti in th amework of Functional Grammar (see also Hengeveld, 1989, 1990). At the abou I anging entities and time spans at which situations of a similar or an opp analysis of these linking devices, we adopt the multi-layered model of utte Ana g production data from a film retelling task, we investigate how speakers en T tudy is about anaphoric linking devices in Romance (French, Italian) and predicates are often semantically related in that they refer to similar or opposite situations. As will be shown in more detail below, a whole array of partly language specific devices are used to achieve discourse cohesion in these cases, including scope particles and adverbials, as well as intonational markings and verbal periphrasis. Consider the following example. (2) Context: There is a fire in the house of Mr. Red, Mr. Green and Mr. Blue 1 Mr. Red jumps out of the window 2 Mr. Blue does the same 3 Mr. Green on the other hand does not want to jump 4 Eventually Mr. Green jumps out of the window as well Every utterance in (2) contributes a new piece of information that is added to the listener’s discourse representation.What is special about the utterances 2–4 is that the predicates are similar to the predicate mentioned in 1 in that the same situation does or does not occur (jumping or not jumping) while the entity constantly changes. Compare a simple utterance like Mr. Green jumps out of the window to the last one from example (2), Eventually Mr. Green jumps out of the window as well. Both have the same descriptive content, but the latter evokes a similar situation holding for a different time span, thus linking the current utterance to earlier (not necessarily directly preceding) ones in the discourse. This is signalled throughout the discourse in (2) by devices such as do the same, on the other hand, eventually, as well, which relate the utterances inwhich they occur to specific units of previous information. The items used for signallingwhich parts of the information aremaintained andwhich parts have changed in comparison to what has so far been established, will most neutrally be referred to as ‘‘anaphoric linking devices’’ in this paper. Note, however, that the anaphoric linking devices in (2) are of different types. Some are explicit expressions of information maintenance (e.g. do the same)2 whereas othersmark changes between entities (on the other hand) or time spans (eventually). Change ormaintenance of information is always relative towhat has been established in prior discourse and comparisons can be made to more than one previous utterance. For example, one and the same information unit can be maintained with respect to the directly preceding utterance but it can be changed in comparison to one that was uttered a while ago. What is then maintained and what is changed information depends on which of the earlier utterances serves as a basis for comparison. Consider the last utterance from example (2) again: Eventually Mr. Green jumps as well. In this utterance as well expresses that a comparison should be made between the current utterance and an earlier one in which the same situation (someone jumping) held for an entity different fromMr. Green, so the relevant antecedent utterance is 1 and/or 2. Eventually establishes a comparison between the current utterance and an earlier one in which the opposite situation (someone not jumping) held for the given protagonist (Mr. Green), so the relevant antecedent utterance is 3. It is important to note that marking such a relation between utterances is never obligatory, but depends onwhat the speaker finds relevant to highlight. Furthermore, information can also serve as a basis for comparison if there is no overt antecedent utterance expressing it. The relevant information must, however, be part of the common ground. man nd G gs, but there are also interesting ence the a ange and maintenance. Firstly, in and rman equivalent in Italian and French. ity (s oge d in (3a) below for Dutch. We will refer se p cles a e a proposition-level comparison of the ance hic Table 1 Information structure related typological differences between Dutch, German, French, and Italian. Word order Subject anaphora Particle repertoire Intonation Dutch V2 Weak and strong personal pronouns and demonstratives Very rich Pitch accents for (verum)focus marking C. Dimroth et al. / Journal of Pragmatics 42 (2010) 3328–33443330 Second, although contrastive stress can be used to mark information structure in all of the four languages, intonational prominence clearly plays a greater role in the Germanic languages (cf. Féry, 2001). In particular, contrastive stress on the finite verb or auxiliary can be used for the expression of verum focus (Ho¨hle, 1992;Matic´ and Nikolaeva, 2009). Contrastive stress on the finite element can have a function that is very much related to the function of the above mentioned assertion-related particles. An affirmative assertion is contrasted with an earlier utterance in which no such relation held between topic and comment, e.g. because the relation was questioned or explicitly denied. This is exemplified in (3b) below for German. (3a) A: Het boek was niet op de tafel. (The book was not on the table). B: Dat klopt niet. Het boek was WEL op de tafel. (That’s not true. The book was indeed on the table.) (3b) A: Das Buch war nicht auf dem Tisch. (The book was not on the table). B: Stimmt nicht. Das Buch WAR auf dem Tisch. (That’s not true. The book WAS on the table.) In addition to these cross-linguistic differences in the repertoire of anaphoric linking devices, earlier comparative studies have revealed differences in the use of some devices that are in principle available in both Romance andGermanic languages. For example, additive scope particles (equivalents of also or as well) exist in German (auch) and French (aussi). However, in a sample of speech elicited with the same stimulusmaterials, the relevant additive particles were used about twice as often in German as compared to French (Benazzo et al., 2004). We will come back to this observation and offer an explanation in the light of the overall tendency of Germanic languages to relate utterances on the level of the proposition. utter Given the questi certain pr overview This p questions 4. The pap 5 If the as 6 A marke of the notio 7 In additi of this asser in w these cro on of wh eference of inform aper is o . The me er finis sertive rel d change o n of contra on to these tion-relate h they occur to another utterance given in the co(n)text. polar to the ee H arti weg, 2009; Karagjosova, 2006; Van Valin, 1975). This is exemplifie s assertion-related particles (see Klein, 2008a7) because they evok that tter in which they appear is in contrast6 to an earlier, otherwise com These are particles like Dutch toch/wel and German doch/wohl/schon (roughlymeaning indeed) whose stressed variantsmark the u ance parable utterance, often with opposite Dutch Ge h there is a special group of scope particles that lacks a direct translation differ s in vailable repertoire of anaphoric linking devices for the expression of ch Ro ce a ermanic languages share many of their information structure markin Following Klein (2006), we call this operator ‘‘assertion’’, as it validates the relation between the state of affairs described in the comment part of a proposition and a topic, which necessarily comprises a temporal reference point and perhaps other components (Dimroth et al., 2003). In the languages investigated in this paper, finiteness (i.e. finite verb morphology and syntax) is the typical reflex of this operation (Last Sunday Mr. Green was jumping).5 This becomes particularly clear when lexically empty finite verbs like the copula or auxiliaries carry contrastive stress. In this case, what is highlighted is either the tense component or the assertion component encoded by finiteness (Klein, 2006). All utterances comprise all of these levels, and discourse cohesion can be achieved by relating utterances at any of these layers. German V2 Weak and strong personal pronouns and demonstratives Rich Pitch accents for (verum)focus marking French SVO (+dislocations, cleft) Weak and strong personal pronouns and demonstratives Poorer No comparable marking Italian Mainly SVO (+dislocations, cleft) Zero anaphora, personal pronouns and demonstratives Poorer No comparable marking ss-linguistic differences in the repertoire and frequency of anaphoric linking devices, we will also address ether the availability of language-specific means for the expression of change and maintenance leads to s for perspective taking in discourse (see Carroll et al., 2004; Slobin, 1996). Table 1 gives a more general ation structure related properties of the investigated languages. rganized as follows. Section 2 introduces some terminological distinctions and specifies the research thodology is presented in section 3, and the results for the different information configurations in section es with conclusions (section 5) and a discussion in section 6. ation is claimed to hold (and not put into question), a falling intonation contour is also needed. f information is called a contrast when it evokes a search for a comparable antecedent utterance. See Umbach (2004) for a definition st that is based on comparability presupposing both similarity and dissimilarity. , Klein considers awhole group of particles (including again, still, already, only and their equivalents in other languages) which on top d function convey other meanings. 2. The study Our data are retellings of video clips. Speakers are invited to talk about different time spans at which similar or opposite situations occur to different entities. In order to determine maintenance versus change at given points in the unfolding discourse the information units ‘time’, ‘entity’ and ‘predicate’ are taken into account at the predication level. At the proposition layer the content of the predication is assigned a topic–comment structure; and an assertion operator establishes a relation between the topic and the comment. In narrative discourse, times and protagonist entities are typically used to define the topic situation (Klein, 2008b), while the comment consists of the predicate. Maintenance or change of information in these units are in principle independent of the topic-comment distinction. Speakers can mark contrasts in both the topic and the comment part, and the assertion operator can also be a locus of contrast. Contrastive stress on the finite verb as the carrier of assertion can have a variety of functions (e.g. expressing tense contrasts). We are mainly interested in the way speakers highlight the fact that an assertive relation holds between the comment and the topic of their utterance in contrast to context utterances in which no such relation was established, e.g. because they have a different polarity. Talmy (1985:131) distinguishes polarity incorporated in the verb root (i.e. hit vs. miss (= not hit) the target) from independent polarity elements like not. In this study we are only concerned with the latter, i.e. contrasts such as sleep vs. be awake that occurred in our data were not counted as ‘change of polarity’. Dik (1997, vol. 2:174–177) proposes that independent polarity elements can again be analysed on two different layers. With respect to positive polarity markers8 he distinguishes predicational positive polarity (as in 4a) from propositional positive polarity (4b). atio nfigu such a predication in relation to others in terms ofmaintenance and change r stu is co following three information configurations in which the comment is maintained from the C. Dimroth et al. / Journal of Pragmatics 42 (2010) 3328–3344 3331 preceding context. Table 2 Information configurations investigated. Type Antecedent (1) and subsequent (2) predication Information configuration of utterance (2) in comparison to (1) Example utterances with corresponding information structure marking Polarity Topic situation Comment Time Entity I 1: Mr. Red going to bed 2: Mr. Blue going to bed = Shift 6¼ = 1: Mr. Red goes to bed 2: Mr. Blue also goes to bed II 1: Mr. Green not jumping 2: Mr. Blue jumping 6¼ Shift 6¼ = 1: Mr. Green doesn’t jump 2: Mr. Blue on the other hand does jump III 1: Mr. Red not jumping 2: Mr. Red jumping 6¼ Shift = = 1: Mr. Red doesn’t jump 2: Mr. Red eventually jumps 8 In the case of negation the distinction between predicational negation and propositional negationwould correspond to Lyons’ (1977:768) ‘assertion of a negative claim’ (cf. it is the case that John is not rich) vs. ‘denial of a positive claim’ (it is not the case that John is rich). 9 See also the related notion of ‘‘referential movement’’ (Klein and von Stutterheim, 2002) that captures the dynamic aspects of the information flow in discourse from one utterance to the next. An utterance’s information configuration as we understand it here, can depend on the information provided in several, sometimes distant antecedent utterances. Ou dy ncerned with the lingu mea hat are used to of information. Information structure marking describes the way this relation is marked in an actual utterance, that is, the istic ns t mark the relevant information configuration in a given language (Table 2). Inform n co ration9 is used to characterize W ll pr tion what is expressed in an (4a) A: Is John rich? B: Yes, he is rich. (4b) A: John is not rich. B: (That’s not true), he IS rich. In (4a) the yes/no-question of speaker A expresses an informational gap concerning the occurrence or non-occurrence of the relevant state of affairs that is filled in speaker B’s response, whereas in (4b) speaker B rejects the truth value of speaker A’s proposition ‘‘John is not rich’’. Dik assumes that ‘‘positive propositional polarity is always formally expressed in natural languages’’ (1997, vol. 2:177). Whereas English relies on supraseg