Abstract
In this study, we provide a new empirical generalization of the meaning contribution of the Mandarin sentence-final particle de from an information maximizer perspective. Based on the cross-entropy model, we quantify the informativity associated with the prejacent that de attaches to. We further explore the plausibility of applying cross-entropy methods (as well as related Kullback-Leibler divergence-based methods) across languages, as a precise model of understanding the subtle pragmatic meanings of sentence-final particles in general.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
Notes
- 1.
De relates to a number of focus-marking construction types, with a much broader distribution than the current scope delineated by contrastive topics. The previous characterizations of de’s meaning appear to be distinct from our upcoming proposal, although we consider it likely that the contribution of de in some or all of these environments could be unified. In the future we hope to pursue if our information optimizer meaning can be reconciled with previous proposals. We briefly point out some potential connections in footnote 7 and Sect. 3.3.
- 2.
We assume here that Monday and Tuesday to Friday are contrastive topics, each indexing a subquestion that is part of the prior overall question (Büring 2003). The subquestion corresponding to Monday is resolved by a contrastive focus (teacher Cai), which is part of the comment to the contrastive topic. Similarly, the topic Tuesday to Friday pairs up with the contrastive focus teacher Wang.
- 3.
Here the judgment has to be elicited based on the particular prior context. An out-of-the-blue utterance does not yield a similar contrast of acceptability. The same applies to subsequent reported judgments.
- 4.
The acceptability of (2b) is improved when the speaker wants to emphasize Monday as particularly relevant to the question under discussion under a different belief state. We return to this issue shortly.
- 5.
An anonymous reviewer wonders if de can attach to both partial answers (of varying informativity) at the same time. To test this, we consulted five native speakers about the following de-answers in (i).
The speakers either found the examples to be less natural compared to the sentences with one de as in (4), or were not sure what role de was playing here (note again the judgments crucially depend on the particular context here). The data are obviously very subtle. As the pilot judgment survey in Fig. 1 shows, the rating difference associated with different de-placement is statistically significant, but not very far apart. This would be typical if the meaning contribution of the de particle is derived from pragmatics (game-theoretic pragmatic meaning), instead of from grammar. A more general pattern of the placement of de will have to await formal experiments in the future.
- 6.
The mild differences between the (a)-sentences and the (b)-sentences could be a factor of the linear order between partial answers. With a partition into two information chunks/units in place, holding back the more informative chunk (and uttering first the less informative chunk) could lead to a violation of the maxim of Quantity. The maxim is maintained if the more informative partial answer precedes the less informative one (afterwards, what’s left is now the most informative chunk available, so that no violation is incurred). In this way, our finding that it is more natural to utter first the more informative partial answer follows from independent Gricean grounds. Crucially, since both (a) and (b) sentences are rated better than (c), the constraint of de-placement cannot be reduced to linear precedence alone.
- 7.
An anonymous reviewer points out the following case where de cannot attach to a complete answer to a wh-question:
From this we see the above pattern in contrastive partial answers does not directly generalize to all exhaustive answers. We do not have a story of the unacceptability here. We would like to point out that in some cases where a complete answer is presented via the explicit enumeration of all members that correspond to a true answer, attaching de to the end appears to be much better. According to our consultation, in the scenario where the following three slopes in the (a) answer are all there are that allow for skiing, adding de sentence-finally is quite natural, compared to the degraded (b)-answer with no enumeration. Importantly, using de here is better than using it with an utterance that the speaker knows is not a complete answer. For example, given the same scenario, adding de to a non-exhaustive answer as in (c) is less natural (to the extent that some speakers accepted it as fine, the meaning they obtained conveyed a separate dimension of speaker attitude implied by de, unrelated to focus marking).
The pattern seems to suggest with all else controlled, de favors more informative answers. Also, we suggest whether explicit alternatives are offered, or contrasted, might influence the acceptability of de-attachment. There are other confounding factors that impact the use of the information optimizer de in complete answers: We find that adding the copula shi improves the judgment significantly in many cases, and there is also a subject-object asymmetry at play (an object focus resists de more than a subject focus).
- 8.
The RSA model predicts an interaction between (shared) knowledge and captures how detailed the speaker’s belief states could influence a listener’s interpretation. As such, it is particularly suitable for characterizing expected utility, as the latter is a quantity that depends on the speaker’s belief distribution (Frank and Goodman 2012, Goodman and Stuhlmüller 2012). This advantage also makes RSA modeling a powerful tool in capturing incremental cancellation within the classic scalar implicature, e.g. uttering some P invites the inference that not all P (Horn 2005). The ability to cancel such not-all inference is subject to the increasing knowledge given a situation. In general, the extent to which an implicature can be cancelled depends on what the listener knows about the speaker’s knowledge state. Consider a scenario with three apples in total. The speaker and the listener have the conversational goal of conveying the information about the number of red apples in the simplest way. In the case where the speaker has complete knowledge of the number of red apples, and the listener knows that the speaker has complete knowledge, then the listener is very likely to infer ‘not all apples are red’, upon hearing the speaker’s utterance ‘some apples are red’. Consider now an alternative scenario where the speaker has partial access to information about the apples, e.g. he only knows if one apple is red, and the listener knows the speaker does not have complete knowledge. Then the listener is less likely to draw a not-all inference upon hearing ‘some apples are red’, compared to the previous scenario. Thus, the implicature is ‘canceled.’.
- 9.
The freedom property means that whenever two pairs of elements are distinct, their unions are distinct (Szabolcsi and Zwarts 1993).
- 10.
We are simplifying our discussions here. In our example, the actual QuD is broken down into pair-list questions where the atomic answers take the form of weekday-teacher pairs. We want to avoid such complication.
- 11.
We acknowledge that here the example setting is unrealistic. The purpose of using oversimplified assumptions is to enable a demonstration of the way informativity is calculated in non-uniform prior cases by feeding concrete numbers into the model.
- 12.
Ne has elsewhere been shown to assume the function of a contrastive topic marker (when selecting for a noun phrase in the sentence-medial position) as well as a fragment question marker (Constant 2014). East Asian particles are known for their multifunctionality. For instance, the particle de additionally assumes the function of a relativizer, a nominalizer and a possessive marker (Chao 1968).
- 13.
No is further claimed to convey a sense of authority on the part of the speaker (i.e. a reliable evidential source) (e.g. Cook 1990). We will leave this layer out of our discussion.
- 14.
We are not strictly interested in the listener’s actual belief state, but the speaker’s belief about the listener’s belief state, since the speaker is of course the one who must decide whether she wants to use the particle.
References
Baldi, P.: A computational theory of surprise. In: Farrell, B., van Tilborg, H. (eds.) Information, Coding, and Mathematics, pp. 1–25. Norwell, Norwell (2002)
Büring, D.: On D-trees, beans, and B-accents. Linguist. Philos. 26(5), 511–545 (2003). https://doi.org/10.1023/A:1025887707652
Champollion, L.: Parts of a Whole: Distributivity as a Bridge Between Aspect and Measurement. Oxford University Press, Oxford (2017)
Champollion, L., Alsop, A., Grosu, I.: Free choice disjunction as a rational speech act. In: Proceedings of SALT, vol. 29, pp. 238–257 (2019)
Chao, Y.R.: A Grammar of Spoken Chinese. University of California Press, Berkeley (1968)
Cheng, L.L.-S.: Deconstructing the shi...de construction. Linguist. Rev. 25, 235–266 (2008)
Chu, C.: A Discourse Grammar of Mandarin Chinese. Peter Lang, Berlin (1998)
Constant, N.: Contrastive topic: Meanings and Realizations. Ph.D. thesis, University of Massachusetts at Amherst (2014)
Cook, H.M.: An indexical account of the Japanese sentence-final particle No. Discourse Process. 13, 401–439 (1990)
Dayal, V.: Locality in wh-quantification: Questions and relative clauses in Hindi. Kluwer Academic Publishers, Dordrecht (1996)
Frank, M., Goodman, N.: Predicting pragmatic reasoning in language games. Science 336(6084), 998–1008 (2012)
Franke, M., Jäger, G.: Probabilistic pragmatics, or why Bayes’ rule is probably important for pragmatics. Z. Sprachwiss. 35(1), 3–44 (2012)
Goodman, N., Stuhlmüller, A.: Knowledge and implicature: modeling language understanding as social cognition. Top. Cogn. Sci. 5(1), 173–184 (2013)
Guo, W.: The discourse function of the Chinese particle ne in statements. J. Chin. Lang. Teach. Assoc. 40(1), 47–82 (2005)
Hamblin, L.: Questions in Montague English. Found. Lang. 10, 41–53 (1973)
Herbstritt, M., Franke, M.: Complex probability expressions & higher-order uncertainty: Compositional semantics, probabilistic pragmatics & experimental data. Cognition 186, 50–71 (2019)
Hole, D.: The deconstruction of Chinese shi...de clefts revisited. Lingua 121(11), 1707–1733 (2011)
Horn, L.: The border wars: a neo-gricean perspective. In: von Heusinger, K., Turner, K. (eds.) Where Semantics Meets Pragmatics, pp. 21–48. Oxford University Press, Oxford (2005)
Jäger, G.: Game dynamics connects semantics and pragmatics. In: Pietarinen, A.-V. (ed.) Game Theory and Linguistic Meaning, pp. 89–102. Elsevier, Amsterdam/New York (2007)
Krifka, M.: Epistemic, evidential and discourse modalities in commitment space semantics. In: Not-At-Issue-Meaning and Information Structure Workshop 1, 1–13 (2017)
Krifka, M.: Layers of assertive clauses: propositions, judgements, commitments, acts. In: Hartmann, J., Wöllstein, A. (eds.) Propositional Arguments in Cross-Linguistic Research: Theoretical and Empirical Issues, pp. 1–45. Gunter Narr Verlag, Tuebingen (2019)
Kuno, S.: Nihon-Bunpoo Kenkyuu. Taisyukan, Tokyo (1973)
Lin, W.: What does the Mandarin particle NE communicate? Cahiers de Linguistique - Asie Orientale 13(2), 217–240 (1984)
Merin, A.: Information, relevance, and social decision making: some principles and results of decision-theoretic semantics. In: Ginzburg, J., Moss, L., de Rijke, M. (eds.) Logic. Language, and Information, pp. 179–221. CSLI Publications, Standford (2011)
Paul, W., Whitman, J.: Shi...de focus clefts in Mandarin Chinese. Linguist. Rev. 25(3–4), 413–451 (2008)
Scontras, G., Tessler, M.H., Franke, M.: A practical introduction to the rational speech act modeling framework. ar**v preprint ar**v:2105.09867 (2021)
Shannon, C.: A mathematical theory of communication. Bell Syst. Technical J. 27(3), 379–423 (1948)
Simpson, A.: Sentence-final particles. In: The Handbook of Chinese Linguistics, pp. 156–179, 2014
Simpson, A., Zoe, W.: From D to T-determiner incorporation and the creation of tense. J. East Asian Linguis. 11(2), 169–209 (2002)
Soh, H.L.: Mandarin Chinese sentence final de as a marker of private evidence. In: Proceedings of the Linguistic Society of America, vol. 3, no. 22, pp. 1–14 (2018)
Spector, B.: The pragmatics of plural predication: homogeneity and non-maximality within the rational speech act model. In: Proceedings of the 21st Amsterdam Colloquium, pp. 435–444 (2017)
Szabolcsi, A., Zwart, F.: Weak islands and an algebraic semantics for scope taking. Nat. Lang. Seman. 1(3), 235–284 (1993)
van Rooij, R.: Utility of mention-some questions. Res. Lang. Comput. 2(3), 401–416 (2004)
Acknowledgements
We appreciated to the anonymous reviewers for the ESSLLI Student Session 2021 for their valuable comments and suggestions. We would also like to thank the editors who put the Best Student Session Papers 2019–2021 Springer Volume together. We have also benefited from discussions with Daniel Hole, Judith Tonhauser, Fabian Bross, Sebastian Padó and Swantje Tönnis, as well as audiences at the student session and the linguistic colloquium at University of Stuttgart. All the remaining errors are our own.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
We conducted a pilot acceptability judgement task, with the aim of addressing the question of whether the placement of de-particle is subject to the type of partial answer and subject to the relative order of partial answers.
Participants. A total of 20 participants were recruited (11 male, 9 female, average age: 23.7 ± 1.13). All participants were Mandarin native speakers currently enrolled in non-linguistics undergraduate programs from a college in China.
Material. A total of 32 target sentences were distributed across 4 lists, randomised across participants, with 8 items apiece. Each list follows a Latin square design. One variable is the placement of the particle de (first vs. second prejacent). The other is the numbers of situations (i.e. informativity). All combinations of the ordered partial answers are summarized in Table 1. Each individual rates 8 target sentences and 24 fillers on a Likert scale from 1 to 7 (ascending order, 1 being the least natural/least acceptable). All items follow a prior wh-question indicating the immediate QuD.
Procedure. The task is conducted online on Qualtrics. A target sentence’s wh-context is displayed on the same page as the target, with scores (1–7) horizontally aligned at the lower half of the page. A radio button is present underneath each score. After choosing a score by mouse-clicking on a radio button, the participant clicks on the continue button to proceed to the next sentence.
Results. An ordinal mixed effect model (Christensen 2019) is adopted: a, b, c conditions are contrast-coded against the reference level of the condition d. No significant effect is observed between condition c and d by an ordinal mixed model (Christensen 2019, Tukey \(\alpha \)-adjustment), with a random intercept for participant and item and a random by-participant slope. However, a significant difference is observed between condition a and d, where the de-particle attached to a ‘more-situation’ prejacent is preferred over the ‘fewer-situation’ prejacent (see Fig. 3).
Pattern of judgment based on permutations of partial answers with placement of de. Only English translation is provided.(‘***’: \(p<\)0.001; ‘**’: \(p<\)0.01; ‘n.s.’: not significant).
Our preliminary acceptability judgment task reveals that when partial answers (i.e. two partial sentences) are juxtaposed, judgments are significantly better with de attached to the ‘more-situation’ partial answer than the ‘fewer-situation’ one. The order of the two answers is not significant, excluding the possibility that the contrast is due to the preference for uttering the more informative partial answer earlier.
Rights and permissions
Copyright information
© 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Chen, J., Papay, S. (2024). Sentence-Final Particle de in Mandarin as an Informativity Maximizer. In: Pavlova, A., Pedersen, M.Y., Bernardi, R. (eds) Selected Reflections in Language, Logic, and Information. ESSLLI 2019. Lecture Notes in Computer Science, vol 14354. Springer, Cham. https://doi.org/10.1007/978-3-031-50628-4_3
Download citation
DOI: https://doi.org/10.1007/978-3-031-50628-4_3
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-50627-7
Online ISBN: 978-3-031-50628-4
eBook Packages: Computer ScienceComputer Science (R0)
