Surprise questions in English and French

4.1 Previous studies

There is general agreement that the hell carries expressive force. It is used in questions to ‘express surprise or bafflement’ (Huddleston and Pullum 2002) and to ‘express surprise in the answer’ (Pesetsky 1987). The hell is adjoined to the wh-word, so that the wh-phrase behaves like a constituent (Kim et al. (2019), Huang and Ochi (2004)):

The wh-the hell phrase is necessarily fronted (Pesetsky (1987), Ginzburg and Sag (2000)). It cannot appear in situ:

Expressivity conveyed by wh-the hell targets the left periphery and presupposes a negative inference. As den Dikken and Giannakidou (2002, 43) state, the hell ‘conveys a presupposition that the speaker has a negative attitude towards the value of wh-the hell and the propositional content of the wh-question’. However, there is no consensus on the precise role of the hell. The hell is variously labelled ‘emotive modifier of wh word’ (Huddleston and Pullum 2002), ‘polarity item’ (den Dikken and Giannakidou 2002), and ‘attitude intensifier’ (Hoeksema and Napoli 2008).

4.2 Results and discussion: The hell as an intensifier

In this study, the hell is analysed as an intensifier whose emotive contribution depends on the context. The nature of emotion is not fixed, the interpretation of the negative inference depending on the speaker’s knowledge state and on the nature of the scale that is involved. It is argued that two types of scale may be associated with the hell, which determines the type of expressivity. If expressivity is grounded in the speaker’s attitude, an attitudinal scale is contextually defined and may account for various emotive values. This corresponds to the emotive function of language as defined by Jakobson (1960). However, expressivity is not necessarily grounded in the speaker’s attitude (Legallois and François 2012). Expressivity may be enacted regardless of the speaker’s emotional state. In that case, the wh-the hell construction conventionally performs surprise. The scale that is involved is a scale of performativity (Celle et al. 2021).

Let us first consider emotional expressivity as the reflection of the speaker’s attitude. If the speaker does not know the answer to the question, a scale of ignorance may be postulated:

Expressivity targets the identification of an entity, intensifying the sense of ignorance. The speaker’s failure to find a value for the wh-variable may induce puzzlement, frustration, and annoyance.

If the question points to a conflict between the speaker’s expectations and contextual evidence, a judgement is formed along a scale of incongruity. In that case, the question does not aim to identify an entity:

Clearly, the question is not information-seeking, as the speaker knows what the addressee is doing. However, the addressee’s behaviour is felt by the speaker to violate a norm. The what the hell question presupposes ‘you should not be doing what you are doing’. The hell intensifies the speaker’s negative attitude along a scale of incongruity. Along with surprise, the question conveys strong disapproval. The unexpected inappropriate event induces a request for an explanation. Similarly, intensification may apply to a WXDY construction:

The secondary predicate ‘talking to him’ is in focus and indicates that the speaker knows what the addressee is doing. At the same time, this proposition contradicts the speaker’s expectations. While interrogative, the wh-word is not variable-binding. As stated by Kay and Fillmore (1999), what is deprived of a referential function in the WXDY construction. The construction conveys a judgement of incongruity, whereby the activity represented by the Y constituent is judged inappropriate. This sense of incongruity is emphasised by the hell, which intensifies the speaker’s negative attitude.

In embedded contexts with main verbs that describe the subject’s attitude or speech, expressivity marked by what the hell is displaced to the subject of the main clause:

The judgement of incongruity is temporally indexed to the shifted reported speech and the judge corresponds to the subject – Mary in (5′), Peter in (6′).

If, however, the answer to the question is known to the speaker and obvious to everyone, the rhetorical reading of the question is reinforced by the hell:

Adopting Rohde’s view (2006) that rhetorical questions convey strong bias, it may be argued that by intensifying the speaker’s negative attitude, the hell strengthens the negative bias. The rhetorical reading may be paraphrased by ‘you know, that does not solve anything at all’. This intensification may induce a sense of impatience or annoyance.

To sum up, the hell is an intensifier that targets the semantic domain associated with the wh-word. The negative inference it conveys gives rise to various negatively valenced emotions depending on the speaker’s knowledge state and the nature of an attitudinal scale that is contextually defined. It is in contexts where a scale of incongruity is at play that the sense of surprise is the clearest ((5)–(6)).

However, intensification marked by the hell is not necessarily related to an attitudinal scale. The hell may involve a scale of performativity, in which case surprise is enacted rather than felt. This type of expressivity is typically triggered by the discourse marker like. As pointed out by Fox and Robles (2010, 720), be like has various uses. In personal uses, be like has a quotative function. In impersonal uses (it’s like) however, be like does not report actual speech. It functions as an enactment marker. Utterances framed by it’s like are ‘affect-laden assessments, enacting a responsive attitude’. When used as a discourse marker, like conveys a sense of approximation that is compatible with either function. However, what the hell questions framed by like tend to be construed as response cries enacting an inner state rather than describing actual report:

In (8), what the hell is going on is not presented as a sequence of reported speech, but rather as a conventionalised utterance that stands for a judgement of incongruity. The wh-the hell question is meant to encapsulate the attitude denoted by the adjectives upset and mad. This is even clearer in (9), where like cannot be considered as a reportative. The what the hell question enacts the look referred to in the first part of the utterance. It allows construing this look as a surprised response to an unexpected event by involving a scale of performativity. The wh-question intensified by the hell functions as a token of surprise, enacting an inner state that characterises the situation denoted in the first clause. This enactment explains why a what the hell question may denote a totally virtual stereotyped situation.

The condition for construing what the hell as enacted surprise is that speaker and addressee share the knowledge of what what the hell questions stand for. The speaker relies on the addressee’s capacity to associate a specific situation with a prototypical situation able to characterise it. The enactment strategy is part of what Fox and Robles (2010) call a ‘tendency towards performance’ whereby the speaker invokes a shared representation of the world. Certain contexts, such as those framed by the discourse marker like, unpack the enactment potential of what the hell. This suggests that the use of what the hell interrogatives has become conventionalised enough to function as a token for incongruity meaning.

4.3 Stand-alone and lexicalised uses

It is generally agreed that sluicing is unacceptable in embedded what the hell clauses:

While what establishes a link to something, thus licencing a sluice, what the hell is not compatible with such a discourse link. Kim et al. (2019), in line with den Dikken and Giannakidou (2002), argue that this follows from the non-D-linked characterization of what the hell (Pesetsky 1987). This constraint is corroborated by Kim et al.’s corpus study of all the wh-the hell elliptical clauses in the COCA. According to Kim et al. (2019), sluicing is problematic with an overt correlate because it violates the non-D-linked constraint. However, they note that ellipsis is frequent in matrix clauses, more specifically in ‘pragmatically-controlled’ contexts (11) as opposed to ‘merger contexts’ (12):

Building on this, it may be argued that a correlation can be established between the use of what the hell beyond a sentential frame and the type of expressivity it carries. If an ellipsis can be postulated, as in (12), a predicate may be recovered. Intensification is then based on an attitudinal scale that is contextually defined, as in (4)–(7). If, however, what the hell occurs in a ‘pragmatically controlled’ context (11), it is questionable whether the question is truly elliptical. In line with Bondarenko and Celle (2019), such questions may be regarded as verbless, as no predicate can be recovered. In that case, the attitudinal scale that is involved has to be inferred from the context. In (13), the speaker’s negative attitude to a preceding discourse content is intensified by stand-alone what the hell:

As the wh-phrase is not associated with a content, it cannot aim at getting an informative answer. A negative inference may be drawn from the preceding context, namely that one bite is not enough to get in trouble. What the hell scopes backwards and cancels the relevance of Adam’s refusal to take a bite on the grounds that he does not want to get in trouble.

This indicates that the use of what the hell has become a conventionalised token of expressivity by intensifying the speaker’s negative attitude even in the absence of an interrogative content. This conventionalisation also accounts for the lexicalisation of what the hell. For instance, it may be lexicalised as a noun:

In (14), what the hell stands for dismay. This suggests that the morphosyntax of what the hell facilitates the conventionalisation of the incongruity meaning. Being a constituent with an incorporated intensifier, what the hell can be taken to encode incongruity beyond a sentential frame. What the hell questions crucially differ in this from surprise questions in French, which always rely on sentence structure. Before addressing surprise questions in French, a reminder on some important constraints on interrogative sentences in French is in order.

6.1 Previous studies

Dekhissi (2016, 2018) and Coveney and Dekhissi (2017) identified two structures as expressive questions – qu’est-ce que questions in their non-argumental use and c’est quoi + right dislocation. They dub the non-argumental use of qu’est-ce que a “conflictual rhetorical question”:

According to Coveney and Dekhissi (2017), the non-argumental use of qu’est-ce que has a why-like meaning. However, unlike a pourquoi (why) question, it conveys an emphatic assertion of opposite polarity concerning the reason for p. This can be paraphrased as ‘there is no reason for p’.

Dekhissi and Coveney distinguish this use from the c’est quoi + right dislocation construction, which they regard as an “exclamative question”:

Their main argument for not treating this question as rhetorical is that no assertion of opposite polarity is implied, as opposed to (18).

It is argued below that in their non-argumental use, qu’est-ce que questions are not rhetorical questions. C’est quoi questions are not considered to be exclamative questions. Both constructions are reported to be surprise questions.

6.2 Results and discussion: qu’est-ce que questions as expressive questions

6.3 C’est quoi ce N as an expressive construction

As mentioned in Section 5, the use of the in situ c’est quoi construction is on the increase and tends to be preferred over the use of the qu’est-ce que construction in information-seeking contexts. I uphold the view that c’est quoi + ce N is also a competitor to qu’est-ce que questions in expressive contexts and that both constructions have a similar illocutionary function as surprise questions. The c’est quoi construction is almost a fixed phrase – it is always in the present tense and cannot be embedded.

In the in situ structure, the interrogative word is in focus and sets up an expectation for a focal argument. The in situ structure is compatible with left or right dislocation, i.e. with a constituent that appears « outside of the boundaries of the clause containing the predication » (Lambrecht 2001, 1050). This constituent is co-indexed with the clitic pronoun c’ in subject position. While left dislocations establish the dislocated noun phrase as a topic (Lambrecht 1981), right dislocations are more controversial. Chafe (1976) and Lambrecht (2001) label the constituent that appears in right dislocations an antitopic. There is no agreement on the discursive function of the antitopic. Ashby (1988, 214) argues that right dislocations are much more frequently used to refer to situationally evoked referents than are left dislocations. However, an important difference between the topic and the antitopic is that the latter does not necessarily involve presupposition. It requires activation^[4], in the sense that it relies on an activated entity that is accessible to the addressee. The entity is made salient by the right dislocation, but does not require any shared knowledge. This explains why a left-dislocated noun phrase has to be referential (28b), while a right-dislocated noun phrase may have a referential quality in information-seeking questions (28a) or a non-referential quality in surprise questions (29a):

In (28), the question seeks information about the identity of the designated entity, as made clear by the preceding explanatory statement (j’avais pas compris le franglais ‘I did not understand the Frenglish word’). In (29) however, the noun is interpreted figuratively and not literally. The right dislocation gives rise to a purely evaluative reading with nouns that may oscillate between a referential and a figurative construal (délire ‘delusion’, ‘mess’; histoire ‘story’, bazar ‘mess’). The noun phrase in the right dislocation cannot become a topic (29b). As pointed out by Stefani and Horlacher (2017), the right dislocation has become a routinised structure that conveys the speaker’s evaluation. This is confirmed in corpora of spoken French. In the OFROM corpus, for instance, c’est quoi questions with a right-dislocated noun preceded by a demonstrative are all surprise questions. This suggests that the c’est quoi + ce N construction tends to be conventionalised as a surprise question. While the in situ question sets up an expectation regarding the properties of the focal argument, the speaker’s assessment in the right dislocation makes it impossible to provide a focal argument in the answer. The evaluation of the designated entity in the right dislocation blocks the information-seeking reading of the in situ question and quoi cannot be interpreted referentially. The question conveys the speaker’s judgement of incongruity about the situation. The answer cannot instantiate the wh-variable and may only be an explanation.

As observed by Lefeuvre (2015), c’est quoi questions may be used to seek information not only about an entity, but also about an event. However, as noted in the analysis of (16), the structure of a c’est quoi question does not convey a query about an event in an explicit way. Reference to an event has to be inferred by the addressee, as evidenced by the reformulation in (17). In the surprise reading, the judgement of incongruity may target an event in a similar way:

The postmodifier in those right-dislocated noun phrases does not have a classifying function – it is purely evaluative. In (30), the question is not about identifying the properties of a bar-mitzvah dress. The question rather points to the fact that wearing an elegant dress is incongruous in a context where there is no need to dress up. Similarly in (31), the activity judged incongruous is categorised as a practice. In both examples, the role of an agent in performing such an activity has to be inferred.

Qu’est-ce que questions and c’est quoi + right dislocation may be viewed as two available constructions to convey a judgement of incongruity. They both indicate that a proposition cannot be added to the common ground unless an explanation is supplied to allow the speaker to recompute newness. The results of a production study on string-identical questions with both an information-seeking construal and a surprise construal confirm that the two constructions have similar prosodic characteristics in the surprise condition as opposed to the information-seeking condition (Celle and Pélissier 2022). When the syntax is ambiguous between a surprise reading and an information-seeking reading, the surprise reading is characterised by an increase in lengthening, slower speech rate, and less frequent rising final contours. Surprise questions may thus be defined as a question type on the basis of a mapping between syntactic-prosodic features and functional meaning. However, Celle and Pélissier’s (2022) results also show that the prosodic differences between qu’est-ce que questions and c’est quoi questions in the surprise and information-seeking conditions are more marked in the case of c’est quoi. In particular, the c’est quoi + right dislocation questions are more sensitive to the difference in length depending on the condition. As surprise questions, they are longer and pronounced with a slower speech rate than as information-seeking questions. In contrast, qu’est-ce que questions do not exhibit a significant difference in terms of duration as surprise questions, which may be due to their greater complexity and variability. Qu’est-ce que questions turned out to be more difficult to interpret for some participants in Celle and Pélissier’s experiment, while c’est quoi questions are more easily identifiable as either surprise questions or information-seeking questions. This suggests that the in situ syntax associated with the right dislocation allows for a more clear-cut prosodic distinction between the surprise reading and the information-seeking reading. In combination with syntactic restructuring, the prosodic realisation of the focal and post-focal constituents is more easily differentiated between the two conditions, which confirms Féry’s (2001) and Boucher’s (2010) claims. Indeed, French cannot rely on prosody alone to put a constituent in focus.

From a semantic point of view, the main difference between the two constructions is that they do not conceptualise the incongruity judgement in the same way. For instance, (30) and (31) could be rephrased using qu’est-ce que without any change in the illocutionary force:

In (30′) and (31′), the focus is on an activity intentionally carried out by an agent although this activity is contrary to the speaker’s expectations. The aspectual profile of the eventuality makes it possible for the addressee to opt for another path. As explained above, there are aspectual restrictions to the formation of the surprise question.

By contrast, in the c’est quoi + right dislocation construction, a noun in a stressed post-focal position is used to denote an incongruous situation. The situation is directly designated and assessed. There is no mention of an agent involved in performing an activity. The underlying predication as well as the role of an agent have to be inferred by the addressee. The structure comes as less complex from a syntactic point of view. Nonetheless, it is more demanding from a cognitive point of view as the incongruous event is only implicitly hinted at.

To sum up, c’est quoi ce N and qu’est-ce que constructions share a non-referential use of interrogative words in surprise questions, focusing on a situation judged incongruous. Questioning is triggered by a conflict between the speaker’s expectations and contextual evidence. The incongruous situation is a dialogical issue that needs to be resolved.