The pragmatics of surprise-disapproval questions: An empirical study

1 Introduction

In this paper, we investigate a type of non-canonical question that is not information-seeking, but rather expresses the speaker’s surprise and/or disapproval with respect to the event denoted by the propositional part of the question. In the theoretical literature, this question type has first been investigated in more detail and from a cross-linguistic perspective by Munaro and Obenauer (1999), who illustrate the general characteristics of structures that both feature interrogative syntax and express a surprise-disapproval interpretation by examples like the following:

The main point of these examples is that in both examples, the wh-element ‘what’ is not used to refer to a syntactic argument (like in What is he reading? [He is reading a book ]), but rather expresses a meaning close to ‘why’. This is particularly clear in examples like (1b). Here, only a non-argumental reading of ‘what’ is possible because of the intransitive verb ‘to laugh’—an argumental interpretation of the wh-element (like in What is he reading? [He is reading a book ] would result in ungrammaticality (*He is laughing x).

Crucially, as soon as interrogatives feature this ‘why-like-what’ reading of the wh-element, Munaro and Obenauer (1999: 237–238) point out that this obligatory conveys “an attitude of the speaker ranging from mild surprise to strong disapproval.” In other words, these configurations express a violation of the speaker’s expectation(s), and, at the level of illocutionary force, one could thus categorize them as exclamations (i.e., expressive speech acts that convey that a particular state-of-affairs has violated the speaker’s expectations). As is well known from the literature on exclamations, they can be conveyed by a wide range of syntactic forms, including interrogative syntax. Consider the following English examples, which are taken from Rett (2011: 412):

Looking at the data in (3), we can see that exclamation is a speech act that does not correspond to a particular sentence type, and that the notion of exclamation instead might just refer to specific uses of several sentence types. According to this view, the interrogative examples in (1) could merely instantiate another option to realize an exclamation, and according to this claim we could thus say that the cases in (1) do not function as proper (read: information-seeking) questions anymore.

However, Munaro and Obenauer (1999) in their original proposal have taken a different path (and the syntactic literature adopting their proposal has as well; see Obenauer 2004, 2006; Giorgi 2016, 2018; and many others): In their account, Munaro and Obenauer’s (1999) claim that surprise-disapproval questions must be distinguished at the speech-act level from both information-seeking questions on the one hand and exclamations on the other hand. They formulate this claim by means of the so-called ‘cartographic’ framework, which represents discourse-related meaning components as left-peripheral syntactic projections (Rizzi 1997, 2014). One of the main ideas of this approach is to also encode illocutionary components of utterance meaning within the boundaries of sentence grammar (see also Krifka 2014, 2019). Adopting such a view on the relation between syntax and pragmatics, Munaro and Obenauer (1999) propose that the surprise-disapproval effect of cases like (1) above is encoded in a functional projection other than Interrogative Force. (3) represents the ordering of functional projections in Munaro and Obenauer’s (1999: 217–218) decomposed left periphery; see also Obenauer (2006: 266):

(3)

As (3) represents, there are two more functional projections above the projection encoding the interrogative force of the clause. Specifically, while SpecExcl(amation)CP is the landing site for wh-phrases in wh-exclamatives (e.g., How beautiful he is!), SpecEv(aluation)CP serves as a position for wh-phrases in non-canonical questions like (1) above. Accordingly, Munaro and Obenauer argue that special questions such as surprise questions (but also rhetorical questions etc.) are derived by adding functional structure on top of the structure encoding the question force of a clause (see also Nye 2009: chapter 2 on relevant pseudo-questions in English and a similar proposal).

There are many empirical aspects and conceptual assumptions underlying syntactic analyses like (3), which we cannot discuss in this paper; note that structural representations like (3) are controversial also within generative linguistics (Trotzke and Zwart 2014). But even if one shares the general theoretical framework behind structural claims like (3), one can still question some of its most basic assumptions (see Trotzke 2020). In particular, (3) is in line with prominent theories about exclamatives that posit a dedicated illocutionary force operator for exclamations. For instance, Rett’s (2011) ‘E-FORCE’ and Grosz’s (2012) EX operators instantiate it. Yet, unlike with questions—where a speech act operator returns a different semantic type (a question)—the alleged E-FORCE and EX return the original proposition (‘the at-issue content’) and a ‘not-at-issue’ proposition (in this case: surprise/amazement about p). One can thus wonder on conceptual grounds what exactly is achieved by postulating a distinct speech act operator for exclamations.

Note that it is a common theoretical claim in linguistic theory that so-called ‘major’ sentence types (König and Siemund 2007, 2013) like declaratives (4a), interrogatives (4b), and imperatives (4c) should be represented by distinct illocutionary operators:

Proposing an operator for exclamations (and thus assuming that ‘exclamatives’ as a distinct sentence type exist) already adds a lot of controversial assumptions to the basic inventory in (4). As we have seen above, cartographic approaches that have investigated the phenomenon of surprise-disapproval questions in more detail even go one step further and propose an operator for ‘non-canonical’ speech acts like surprise-disapproval questions (e.g., an Evaluation Phrase in [3]).

We are merely indicating our doubts here about an analysis like (3) above because our goal in this paper is not in the first place a theoretical one. Rather, although our paper starts with and is empirically based on observations that have been pointed out in the cartographic literature, our study aims at empirically assessing theoretical claims about the illocutionary status of surprise-disapproval questions by exploring their pragmatics. In particular, we investigate the illocutionary status of questions like (1) by asking how speakers can react to surprise-disapproval questions in a discourse: Do they perceive them as requests for information in the first place? Or do they mainly interpret them as exclamation speech acts expressing the surprise of the speaker? By carrying out an acceptability study on how speakers react to surprise-disapproval questions, and by comparing those reactions to the ones we observe for both information-seeking questions and wh-exclamatives, we hope to contribute to answering questions about the illocutionary status of such non-canonical questions and about which pragmatic components of these non-canonical uses are predominantly perceived by speakers. In what follows, we will first present our empirical study in Section 2, and we will then turn back to the theoretical implications of our empirical investigation.

2 An empirical study on the pragmatics of surprise questions

The more recent literature on surprise-disapproval questions has identified many more syntactic phenomena that might fall into the category of surprise (or ‘counter-expectational’) questions, and the cases featuring ‘why-like-what’ readings were thus only the starting point of a still growing literature. For instance, configurations like English How come? or What the hell is he doing? are now characterized as surprise questions too (e.g., Celle 2018; Celle et al. 2019).

However, we hypothesize that ‘why-like-what’ is still the clearest example instantiating the category of surprise questions because we agree with the cartographic literature that as soon as ‘what’ receives the non-argumental reading of ‘why’ in interrogatives, the surprise-disapproval interpretation obligatorily follows. Note that this is different in most of the other cases cited in the literature. For instance, it has been pointed out that intensifiers like the hell are not only compatible with surprise interpretations (see Pesetsky 1987: 111), but also with other types of non-canonical questions that might lack a surprise interpretation. As Obenauer (2004: 376) illustrates, the hell is also compatible with the rhetorical-question interpretation (Who the hell cares?) or with the ‘Can’t-find-the-value’ interpretation (Where the hell did I leave my keys?). Given these considerations, our study only tested surprise-disapproval questions instantiated by ‘why-like-what’ readings, as illustrated in (1) above. The language investigated by our study was German since parallel cases to the Romance ‘why-like-what’ readings have many times been pointed out in the original literature on this topic (see Munaro and Obenauer’s 1999 example above [1]).

2.2 Results

2.2.1 Critical conditions

Descriptively, the ratings for information-seeking questions were higher with answers (5.94) than affirmations (2.38). For exclamatives, ratings were lower with answers (4.38) than with affirmations (5.65), although both were still in the acceptable range. For surprise questions, ratings were slightly higher with answers (5.54) than with affirmations (5.22), although the difference was less marked than for the other question types. Mean ratings for the critical conditions are presented in Table 1 and Figure 1.

Table 1:

Mean ratings per condition across participants for critical items. Standard deviations and standard errors of the mean are given in parentheses.

Utterance type	Reaction	Mean rating
Information-seeking	Answer	5.94 (sd 1.11, sem 0.14)
	Affirmation	2.38 (sd 1.01, sem 0.13)
Exclamative	Answer	4.38 (sd 1.22, sem 0.16)
	Affirmation	5.65 (sd 1.06, sem 0.14)
Surprise–disapproval	Answer	5.54 (sd 1.00, sem 0.13)
	Affirmation	5.22 (sd 1.14, sem 0.15)

Figure 1:

Boxplots of mean ratings per condition for critical items.

Ratings were analyzed in R (R Core Team 2017) with a series of linear mixed models, using the packages lme4 (Bates et al. 2015, lme4 function), and LMERConvenienceFunctions (Tremblay and Ransijn 2015, summary function). Plots were prepared using the ggplot2 package (Wickham 2016). For the first model, we specified the main effects and interactions of factors UTTERANCE TYPE (three levels: information-seeking, exclamative, surprise-disapproval) and REACTION (two levels: answer and affirmation) as fixed effects, and PARTICIPANT and ITEM as random effects. In addition, REACTION was specified as random slope for each participant. In addition to significant main effects, there was a statistically significant interaction of UTTERANCE TYPE and REACTION (information-seeking vs. exclamative: t = −22.98, p < 0.001; information-seeking vs. surprise-disapproval: t = −15.45, p < 0.001; exclamative vs. surprise-disapproval: t = 7.53, p < 0.001). This interaction was pursued by monitoring the main effect of REACTION separately for each level of UTTERANCE TYPE. For the second model, we specified the main effect of REACTION as fixed effect, PARTICIPANT and ITEM as random effects, and REACTION as random slope for each participant. The main effect of REACTION was statistically significant for information-seeking questions (t = 17.11, p < 0.001) and exclamatives (t = −6.88, p < 0.001). For surprise-disapproval questions, the main effect of REACTION was only marginally significant (t = 1.85, p < 0.07).

2.2.2 Fillers

Descriptively, fillers were rated as expected, with high ratings for good fillers, low ratings for bad fillers, and medium ratings for medium fillers. Mean ratings for the fillers are presented in Table 2 and Figure 2.

Table 2:

Mean ratings per condition across participants for fillers. Standard deviations (sd) and standard errors of the mean (sem) are given in parentheses.

Filler type	Mean rating
Good	6.19 (sd 0.82, sem 0.11)
Medium	2.48 (sd 1.16, sem 0.15)
Bad	1.43 (sd 0.83, sem 0.11)

Figure 2:

Boxplots of mean ratings per condition for fillers.

Ratings for fillers were analyzed with a third model (parallel to the second model outlined above). We specified the main effect of FILLER TYPE as fixed effect, PARTICIPANT and ITEM as random effects, and FILLER TYPE as random slope for each participant. There was a main effect of FILLER TYPE (good vs. bad: t = −26.27, p < 0.001; good vs. medium: t = −17.94, p < 0.001; medium vs. bad: t = −7.96, p < 0.001).

3 Discussion and conclusion

In this final section, we would like to briefly discuss the results of our empirical study on the pragmatics of surprise-disapproval questions. First, we conclude that participants overall understood the rating task they were presented with. In particular, we found that ratings for all three filler conditions are different from one another, and participants’ preferences for the appropriate reaction for each condition fit our expectations. This shows that our task is sensitive enough to distinguish reactions of varying felicity, and does so in more detail than providing simple quasi-binary distinctions.

The main findings of our empirical study concern the critical items. Specifically, we found—rather unsurprisingly—that the preferred reaction to an information-seeking question is an answer; affirmations are clearly dispreferred. The preferred reaction to exclamatives is affirmation; this is in line with the theoretical literature on exclamatives showing that although an exclamative cannot be answered (6B), the descriptive content of the exclamative (which is presupposed) can be affirmed by the hearer ([6B′]; see Castroviejo Miró 2006; Chernilovskaya et al. 2012; Grimshaw 1979; Zanuttini and Portner 2003). Consider the following pattern again (Zanuttini and Portner 2003: 47), repeated from (5):

Interestingly, our data suggest that although ratings for exclamatives with affirmation were significantly higher than those for exclamatives with answers, the latter still showed higher ratings than expected. In particular, the exclamative–answer pattern suggests that this reaction is perceived as more felicitous than the information-seeking–affirmative condition. This might reflect a bias of our paradigm towards an interrogative interpretation. Another explanation is that participants perceive the descriptive content of an exclamative as still being ‘at-issue’ and not presupposed. In other words, participants might have thought that the speaker has no settled belief about the proposition conveyed by an exclamative (e.g., ‘He bakes delicious pies’ in an exclamative like What delicious pies he bakes!), and that the proposition is thus still ‘at-issue’ (see Villalba 2017 for such an account) and not so far in the conversational background than usually assumed in the theoretical literature (e.g., Zanuttini and Portner 2003 and their analysis of the relevant proposition in terms of a factivity presupposition).

Turning now to the main point of our paper, namely the results we obtained for the pragmatics of surprise-disapproval questions, we observed that both types of reactions (i.e., answer and affirmation) are rated as felicitous. While there is a small descriptive preference for answers over affirmations, the difference between both reactions is only marginally significant. We interpret this as showing that in the case of surprise-disapproval questions there is actually no preference for one of the two reaction types—neither for answer nor for affirmation.

This finding suggests that surprise-disapproval questions are indeed perceived as both question and exclamation speech acts and thus contrast with both information-seeking questions (where we found a significant preference for answer) and exclamatives (this utterance type is preferred with affirmations). The question is now how to interpret this result theoretically (and in the light of the cartographic literature sketched in Section 1) because it is rather uncontroversial that surprise-disapproval questions belong to a ‘mixed’ clause and speech act category, featuring formal and functional properties of different illocutionary forces. Our experiment thus seems to confirm this general characterization.

However, we would like to point out that our results are particularly compatible with a broad theory that assumes that pragmatic phenomena like surprise-disapproval questions are truly ‘mixed types’ in the sense that we do not have to postulate a distinct illocutionary status (or, in syntactic terms, a distinct projection) in order to explain them theoretically. By contrast, we suggest that surprise-disapproval questions might be felicitous with answers (and noticeably more felicitous than wh-exclamatives) because they share the basic word order with information-seeking questions (V2 order), and differ from our cases of wh-exclamatives in this regard (they were all verb-final). We could explain this by a broad illocutionary theory like Truckenbrodt’s (2006) account, which suggests for German that the V-to-C movement property in wh-questions results in a special pragmatics that verb-final exclamatives (like the items in our experiment) lack. Since surprise-disapproval questions in our items feature V-to-C movement (Was lacht der?!), and exclamatives lack it (Was der lacht!), surprise-disapproval question still fall into one grammatical category with their information-seeking counterparts (Was liest er?).

On the other hand, our experiment has shown that surprise-disapproval questions are also felicitously followed by affirmations, which is the reaction type that has also been preferred by our participants for exclamations. We submit that this cannot (and therefore should not) be explained by syntax because many factors at different linguistic levels add up and cumulate to yield the exclamation component of surprise-disapproval questions (certainly lexical items like the special and non-canonical reading of was, but also exclamative intonation, which we did not test in our experiment).

To be sure, factors at different linguistic levels are involved in yielding an interrogative reading in regular (read: information-seeking) wh-questions too—most notably intonation. Our claim here merely is that only the interrogative reading (of both information-seeking and surprise-disapproval questions) requires a dedicated word order and thus must involve syntax proper. This explains the pattern that surprise-disapproval questions cannot only be felicitously followed by affirmations (= signaling their exclamation component), but also by answers (= signaling their interrogative component): The V2 word order has an effect on the interpretation of surprise-disapproval questions because this is the feature that they have in common with information-seeking questions and that distinguishes them from the exclamatives we tested in our experiment.

To conclude, we submit that the only thing that is really due to word order and thus part of the syntax of mixed speech act types such as surprise-disapproval questions is the interrogative component (yielded by wh- + V-to-C movement), but the exclamation component should maybe not be part of and represented in the grammar, in contrast to previous syntactic claims along the lines of cartography. Such a ‘mixed’ theory (combining grammar/word order and non-syntactic means to yield the respective interpretations) is needed for many other mixed clause and speech act types like the more recent English How cool is that?! and its counterparts in further Germanic languages (Auer 2016; Trotzke 2020).