The cognitive load of presupposition triggers: mandatory and optional repairs in presupposition failure

Abstract

If a speaker utters a sentence p containing a presupposition trigger that activates a presupposition q, and q does not belong to the common ground of presuppositions, it is a case of presupposition failure. If this occurs, speakers are required to repair the failure to make sense of the utterance. According to Glanzberg, two subcategories of being infelicitous may emerge in the case of presupposition failure: one is that strong presuppositions lead to obligatory repair, and the other is that weak presuppositions lead only to an optional repair. Following Glanzberg's suggestion, in this paper we present the results of an experiment supporting the idea that, depending on the kind of trigger, processing the information conveyed by a presupposition can be either optional or mandatory in case of presupposition failure. The conclusion of this paper is that the cognitive demands of different presupposition triggers do not primarily depend on whether they optionally or obligatorily lead to process the presuppositions activated. Rather, their cognitive demands seem to be related with the complexity of the mental representation of the presupposition required.

Keywords:

• experimental pragmatics
• presuppositions
• presupposition triggers
• cognitive load

Acknowledgements

We are grateful to Claudio Caneva and Lucia Leporatti for their valuable help. We are also indebted to Diana Mazzarella, Harris Constantinou, Sergio Morra, Fabrizio Bracco, Cristiano Broccias and Massimiliano Vignolo. We thank two anonymous referees for their highly constructive comments.

Notes

1. There are two major approaches towards the problem of triggering presuppositions in the contemporary debate. Semantic approaches claim that presuppositions are a particular type of meaning determined by the lexicon (see, for instance, Chierchia & McConnell-Ginet, 2000; Simons, 2001). However, many scholars (Abusch, 2010; Schlenker, 2010; Simons, 2001; Stalnaker, 1974) support a pragmatic view according to which presuppositions are the result of speakers' inferences as well as in cases of conversational implicatures. For instance, Abusch (2002, 2010) has proposed a pragmatic approach to presuppositions that carries out a different classification of triggers based on the distinction between soft and hard triggers). According to Abrusán (2011, p. 492) ‘Neither of the above approaches are satisfactory: the first approach is non-explanatory and posits an enormous amount of complexity in the semantic system. The second approach is theoretically attractive, however it is fair to say that to this date no satisfactory mechanism has been given that can derive based on rational rules of conversation why certain aspects of the meaning (and not others) are turned into presuppositions’. Kadmon (2001) supports a stronger claim by arguing that whether presuppositions are semantically or pragmatically triggered is irrelevant for the purpose of theoretical investigation into presupposition filtering.

2. Informally, the notation of this update instruction might be translated: check whether the context c contains the proposition p. If so, namely, if c entails p, then the result is an update of c with new information. Otherwise, if c does not entail p, the result is a failure, namely, the context cannot be updated with new information.

3. To be more precise the reduction of the attitudinal component of ‘regret’ to a ‘negative propositional attitude’ could be argued; we could better say that regret* is regret minus the implication that what is regretted is factive and then say: John regrets that p iff John regrets* that p and it is true that p (suggestion from an anonymous referee).

4. An alternative account of ‘even’ is discussed in Kay (1990, p. 84) and Karttunen and Peters (1979, p. 11). For instance, Kay (1990) claims that ‘even’ entails a scalar set of ordered propositions that are salient at the utterance time or may become salient if introduced into the context of discourse.

5. No provisions are made for the behaviour of CS. Regarding DD, Glanzberg guesses that they probably trigger strong presuppositions (personal communication) because they might pattern with complex demonstratives that are classified as triggers activating strong presuppositions in Glanzberg and Siegel (2006).

6. A possible objection to our design might be that we used as items a set of Italian presupposition triggers selected on the base of Glanzberg's taxonomy that is based on English language. We think however that it is reasonable to assume that the formal analysis of the triggers proposed by Glanzberg can be correctly applied to the corresponding Italian triggers (e.g. ‘regret’ induce the same requirement of the ‘corresponding’ Italian expression ‘dispiacersi’, as well as ‘too’ triggers a similar requirement to the Italian focal adverb ‘anche’). We acknowledge that, in order to extend our results to other languages (such as English), further researches could be devoted at validating our results by a replication of the experiment in other languages.

7. The contents were fictional so to reduce the possibility (as far as possible) that participants had already a precise mental representation of the content of the story affected by background knowledge

8. This kind of context, where the content of a presupposition is neither verified nor falsified, is called a neutral context (Tiemann et al., 2011).

9. The original Italian version of this story was: ‘“Nell”acquario di Barcellona ci sono 20 tipi diversi di squali. La guida spiega sempre ai visitatori che quelli presenti sono tutte femmine, per questo, non è possibile la riproduzione in vasca. Nell'ultimo periodo, tuttavia, si sta pensando di riprovare a inserire un maschio nella vasca principale. Gli squali toro, in particolare, sono la principale attrazione perché vengono nutriti in vasca direttamente dagli addetti. La maggior parte degli squali si nutre solo di merluzzo. Infatti, da tempo, ha smesso di cibarsi di altri pesci. In quest'acquario, gli animali vengono curati costantemente: persino gli squali toro, talvolta, vengono prelevati dalle loro vasche’.

10. A possible criticism of our experimental design might be that, perhaps, participants remembered certain propositions from the story better than others, regardless of the kind of trigger that introduces those presuppositions, just because they were more significant. In order to reduce this bias, we tried to balance the relevance of the contents of the presuppositions triggered by the different kinds of triggers contained in the stories. For instance, we did so by trying to assign to the presupposition triggered by a token of a definite description, a relevant content with respect to the story in the first trial and a less relevant content with regards to the topic of the story presented in the second trial (and we tried to do the same with all the other triggers in all the other trials). However, it's necessary to acknowledge that future researches need to control better this variable.

11. We decided to avoid the use of an option ‘cannot say’ or ‘no answer’ to force participants to select either ‘True’ or ‘False’. In this way, they were constrained to try to recover in their memory the information conveyed by the content of the presupposition and, afterwards, to decide whether they remembered that or not.

12. This setting was chosen to avoid introducing the processing of negation during the answers to questions.

13. If presupposition failure occurs, a way for repairing the context is to accommodate the presupposition required. Accommodation is the process of introducing into the common ground a presupposition that was not already presupposed before the utterance time (as in cases of presupposition failure).

14. In the introductory phase, we informed participants that in case of wrong answers to the (second) task concerning the geometrical figures, all the answers to the questions about the stories (first task) would have been evaluated as wrong too.

15. In the literature (Chemla & Bott, 2013; Schwarz, 2007; Tiemann et al., 2011), both off-line and online methods are used to evaluate presuppositions and presupposition triggers. The off-line methods are based on questionnaires or acceptability ratings about sentences in a given context. The online methods are based on reaction time measures, mostly using the self-paced reading method. In our experiment we chose an off-line method (questions) but we manipulated the online cognitive resources available for processing contents and giving correct answers.

16. Representations involving temporal relations are not likely to be visualised in a single static image. According to Schaeken, Johnson-Liard, and d'Ydewalle (1996) temporally displaced events require multiple mental representations.

17. According to Gennari and Poeppel (2003) processing the meaning of eventive verbs requires a high cognitive demand because of the complex representation of a dynamic event entailing an initial state, a change and a final state.

18. Further researches are necessary for investigating: (1) whether triggers that require a representation of temporally displaced events are more cognitively demanding because of the complex process of mental representation. (2) Moreover, it has to be analysed whether this cognitive demand characterises the whole categories of IT and CS or whether it is peculiar of certain triggers within these categories.

19. A further study is planned to compare other categories of triggers (Cleft sentences, complex demonstratives, implicative verbs, etc.).