Abstract
This squib attempts to constrain semantic theories of agree wh constructions by broadening the data set and collecting naive speakers’ intuitions. Overall, our data suggest relatively permissive truth-conditions for these constructions. They also suggest a previously undiscussed presupposition for agree wh and also indicate that agree wh is not straightforwardly reducible to agree that. Although some accounts suggest differences in truth conditions among different asymmetrical agree with constructions and symmetrical agree constructions, we do not find any indication of such truth-conditional distinctions. In the course of our exploration of the data, we offer a new approach to distinguishing between truth, falsity and presuppositional failure.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
We use blue as the most important color to be distinguished from the others in all the examples of the paper. We hope this will minimize the difficulty for most readers, including those with color vision disabilities. We will offer additional descriptions in the text whenever more needs to be known. For purposes of print presentation, we have, in all arrays of colored letters, placed descriptive notations beneath the red, green, blue, and yellow letters, and the letters of unknown color. These annotations did not appear in our experimental stimuli.
These must be distinguished from embedders like ask and wonder, which cannot embed propositional clauses and are so analyzed as fundamentally question-oriented, and embedders like believe, which cannot embed interrogative-type clauses. For these classes, the question of reducibility, at least in the sense discussed here, does not arise.
Lahiri (2002) offers a general implementation of strong exhaustivity (Lahiri’s formula (225) in section 3.6.1) that we believe would make similar or identical predictions, given Lahiri’s other assumptions about agree.
This is something of an extrapolation of Lahiri’s brief discussion of agree. Lahiri treats all question-embedding as richly ambiguous, so this is not the only semantics for agree that Lahiri derives, but we believe it is a faithful rendering of one application of Lahiri’s account. Further, among Lahiri’s proposed readings for agree, it is the only one for which we are aware that there are substantially different predictions from the alternatives discussed above. This prediction derives by combining the general semantics of agree with ... that and agree that sketched in 2.3.3 of Lahiri (2002) (especially on p. 34). With the weakly exhaustive Ans1 operator that Lahiri describes on p. 159, which differs from a Heim (1994)-style Ans1 operator in that instead of being oriented towards truth in the world, it has a contextual parameter for the source of the restriction on answers considered. Among other things, this restriction is supposed to accommodate the presuppositions of the predicate, which, for agree with, are supposed to include a belief presupposition. The reader should consult Lahiri (2002) for the full formal details of the account, but the upshot is that A agrees with B about Q will be true on this reading iff every salient mention-some-type answer to Q that B believes is one that both A and B believe.
There was also some structure to the three repetitions. (i) In one repetition the colors were assigned so that target colors were attributed to the first letters, unknown colors to the next letters, and the non-target colors to the final letters; (ii) in another repetition the colors were assigned in the opposite manner (non-target, unknown, target); (iii) and in another repetition the positions of the colors were randomly assigned (so A, C and E could be in the target color, without B or D being in the target color).
Some algorithms exist to determine a priori how many groups should be formed. However, these algorithms rarely provide a clear answer (we ran four such algorithms and obtained: 2, 3, 4 and 8 as possible optimal number of clusters). These algorithms are best suited in situations where there is no a priori reason to choose a particular number of clusters. In our case, we had a priori reasons to look for three clusters: we wanted to sort the situations into true, false and presupposition failure.
Labels ‘true’ and ‘false’ correspond to the positive versions of the sentence.
If the algorithm is not given pairs of ratings but is applied to positive and negative sentences separately, then it delivers three groups as follows:
-
conditions in Table 5, Ctrl-1 and Ctrl-2 lead to the positive sentence being classified in group 1, and the negative sentence in group 2;
-
conditions in Table 6 and Ctrl-3 lead to the reverse: positive in group 2, negative in group 1;
-
conditions in Table 7: positive in group 3, negative in group 2.
Overall, the pattern is the same. The first group recovered by the algorithm corresponds to true sentence-situation pairs, the second group to false sentence-situation pairs, and the third group corresponds to judgments obtained for positive sentence in case of presupposition failure. So, positive sentences seem to be affected by presupposition failure more visibly in this paradigm and to give rise to a new category. Negative sentences just get judged as the false sentences (by means of local accommodation, it should be judged ‘true’).
-
We did not have controls to label the third category as ‘presupposition failure.’ However, we may compare the current study to the investigation of Križ and Chemla (2015), which explores more systematically several ways to detect truth value gaps. It is applied to so-called ‘homogeneity’ of plural definite descriptions. In a pilot experiment, Križ and Chemla (2015) found a similar pattern of responses: true situations led to clearly true responses for the positive sentence and clearly not-true responses for the corresponding negative sentence, the reverse was true for false situations, and situations in which the homogeneity assumption was not met led to somewhat intermediate types of answers for both positive and negative sentences. This is not entirely sufficient for several reasons. First, this was done in a different setting with different participants. Second, homogeneity may not be presuppositional to begin with. But note that there may actually be different kinds of presupposition, leading to different reactions for presupposition failures anyway. In trying to think about presupposition controls to add, we thought about definite descriptions, factive verbs and other presupposition triggers, but quickly realized that they may not form a uniform class. What is crucial for the current purposes then is that the third group behaves differently than clearly true and clearly false cases, making it a ‘not-true and not-false’ group, and therefore plausibly a presupposition failure group.
Thanks to Michael Franke for suggesting analyses of early responses.
This is because John and Mary agree to the same possible answer propositions in all three of these, including propositions about letters being blue and not being blue, and all Boolean combinations of such propositions.
References
Beck, S., and H. Rullmann. 1999. A flexible approach to exhaustivity in questions. Natural Language Semantics 7: 249–298.
George, B.R. 2011. Question Embedding and the Semantics of Answers. Ph. D. thesis, University of California, Los Angeles.
George, B.R. 2013. Knowing-‘wh’, mention-some readings, and non-reducibility. Thought: A Journal of Philosophy 2: 166–177.
Groenendijk, J., and M. Stokhof. 1984. Studies on the Semantics of Questions and the Pragmatics of Answers. Ph. D. thesis, University of Amsterdam.
Heim, I. 1994. Interrogative semantics and Karttunen’s semantics for know. In IATL 1, eds. Buchalla R. and Mitwoch A., 128–144.
Higginbotham, J. 1996. The semantics of questions. In The Handbook of Contemporary Semantic Theory, pp. 361–383. Blackwell.
Karttunen, L. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1: 3–44.
Kratzer, A. 2006. Exhaustive questions. Linguistics Colloquium, Massachusetts Institute of Technology.
Križ, M., and E. Chemla. 2015. Two methods to find truth value gaps and their application to the projection problem of homogeneity. Journal of Semantics.
Lahiri, U. 2002. Questions and answers in embedded contexts: Oxford University Press.
Ryle, G. 1945. Knowing how and knowing that. Proceedings of the Aristotelian Society 46: 1–16.
Schaffer, J. 2007. Knowing the answer. Philosophy and Phenomenological Research 75: 383–403.
Sharvit, Y. 2002. Embedded questions and ‘de dicto’ readings. Natural Language Semantics 10: 97–123.
Spector, B. 2005. Exhaustive interpretations: what to say and what not to say. presented at the LSA workshop on Context and Content.
Spector, B., and P Égré. 2007. Embedded questions revisited: an answer, not necessarily the answer. Presentation at MIT Ling-Lunch Seminar and Journées Sémantique et Modélisation.
Spector, B., and P. Égré. 2015. A uniform semantics for embedded interrogatives: an answer, not necessarily the answer. Synthese. Forthcoming.
Stanley, J. 2011. Know How: Oxford University Press.
Acknowledgments
We wish to thank Alexandre Cremers, Paul Egré, Michael Franke, Manuel Križ, and Benjamin Spector. The research leading to these results was supported by a ‘Euryi’grant from the European Science Foundation (“Presupposition: A Formal Pragmatic Approach”), the European Research Council under the European Union’s Seventh Framework Programme (FP/2007-2013) / ERC Grant Agreement n.313610, ANR-10-IDEX-0001-02 PSL* and ANR- 10-LABX-0087 IEC.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Chemla, E., George, B.R. Can we Agree About agree?. Rev.Phil.Psych. 7, 243–264 (2016). https://doi.org/10.1007/s13164-015-0278-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13164-015-0278-8