Exclamation, epistemic assertion and interlocutor’s subjectivity

1 Trotzke & Giannakidou’s conclusions from emotive embedding

Paramount for T&G is the comparison between clauses tagged as exclamations and embedded under an emotive verb, on the one hand, and root exclamations, on the other hand. They claim that the latter is amenable to an analysis in terms of the former, based on a thorough exploration of their similarities. Their data are drawn from English, German and Greek. For example, consider the Greek and German examples in (1) [T&G’s (28) and (29)] and (2) [T&G’s (78a) and (74b)], respectively, and their English translations.

From this comparison, T&G draw several conclusions, which are summarized here under points (A), (B) and (C).

1. Syntactically, T&G emphasize that the (a) sentences resemble the embedded sentences in (b) (Section 3.3, p. 25–33, Section 4, p. 37–44). In particular, they observe that the Greek emotive complementizer pou features in (1) to the exclusion of the other (non-emotive) complementizers oti and na. In German, the subordinate syntax, with the verb in final position, is visible in (a) as in (b). Based on this observation, they claim that a silent emotive verb is present in the (a) sentences.

2. Interpretively, both root exclamations and complements to emotive verbs are traditionally deemed factive, i.e., to convey some kind of presupposition. In both cases, the authors take issue with the factive status of the content of the clause.^[4] As far as complements to emotive verbs are concerned, they report data points like (3) (Égré 2008: 102), in which the proposition embedded under regret is false in the context.^[5]

As for root exclamatives, like assertions, they can be denied (4) [T&G’s (13)] and are found in answers to questions (5) [T&G’s (20)].

From these observations, the authors draw the conclusion that wh-exclamatives and not just declarative exclamations^[6] involve assertion. In fact, their description even leads to the conclusion that exclamations carry two assertions, even though they do not phrase it exactly this way. Consider the conclusion of Section 2, on p. 18:

[our proposal is that] exclamations as a class (wh- and declarative ones) indeed convey descriptive content and are emotive assertions of intensity.

So, there is emotive assertion, but the conveyance of descriptive content is also an assertion, as said above on the same page:

The second major data point in the literature^[7] intended to show that [the claim that] exclamatives lack assertive content is questionable: exclamatives can indeed be used as answers to information-seeking questions provided that they occur in a relevant information-structural and discourse environment.

Thus, exclamations are taken to carry

1. An assertion of the propositional descriptive content;

2. An emotive assertion.

The second point is by far not new, as T&G acknowledge, e.g., in fn. 4. It is already found in, among others, Michaelis and Lambrecht (1996: 378): “The notion of surprise or expectation contravention comes into play because the exclamative utterance counts as an assertion that the degree of the scalar property in question is unusually high,” which they rephrase “assertion of affective stance” below on the same page. Evidence for this comes from the fact that exclamations can be prefixed with “I can’t believe” (p. 379). Likewise, Biraud et al. (2021) defend the view that one of the main ingredients of the exclamatory speech act is that “the speaker informs her/his interlocutor that s/he affectively does not accept the content of the exclamation” (p. 91, my translation). We find something similar in Rett (2011), although she is careful about describing exclamations not as assertions but as expressions (i.e., expressive acts).

Conversely, that the descriptive content of the exclamation is asserted and not presupposed is somewhat new. Only in Biraud et al. (2021), I find a comparable view: “The speaker intends to inform the interlocutor that the proposition is true (by a strong assertion)” (p. 91, my translation). Observing that wh-exclamatives pass the tests for neither conventional implicatures nor presuppositions, they adopt Martin’s (1987: 98–99) description that their content is presented as obvious. They distinguish between assertion and strong assertion. Strong assertion is a speech act that aims at enforcing the interlocutor’s acceptance, while (mere) assertion is neutral in this respect.^[8] Crucially, like T&G, the book shows that this part of the exclamatory speech act is valid whatever the clause type that harbors the exclamation.

1. Both contrary emotives and exclamations involve subjective presupposition, which divides into two parts: A counterexpectation presupposition (the speaker believed not p before the utterance t, formalized on p. 23) and a veridical presupposition (the speaker believes p at the utterance time, formalized on p. 24–25). The point is borrowed from Giannakidou (2016) and Giannakidou and Mari (2021), where it is formalized adapting the tools of modal bases to emotive modality. To my knowledge, while counterexpectation was identified a long time ago,^[9] the fashion in which a subjective-veridical-presupposition part in exclamations is identified and formalized is novel and in fact sheds new light on our understanding of exclamation. I will comment on it and try to take it a step further below.

2 Emotive embedding is limited

I would like to discuss the points that were highlighted in Section 1 (similarity with emotive embedding, assertion, subjective presupposition) in the light of the embedding data, which play a central role in T&G. As far as syntax is concerned, T&G’s proposal cannot be maintained as such, in my view. First, if there is an equivalence between (a) and (b) sentences in (1) and (2), we do not see why the structure is different when the emotive verb is overt and when it is covert. Put otherwise, we do not see why (1a) and (2a) cannot be embedded unmodified, as in (6).

Second, as the authors themselves point out (p. 20, p. 43), the (b) sentences in (1) and (2) are not unambiguously exclamatory, so no firm conclusion can be drawn from them. Moreover, one of their arguments is that Greek exclamatives can exhibit the emotive complementizer pou, which is also found in emotive declarative embedded clauses (1b). However, this data point is not unequivocal either, since pou is also the homonymous relative marker (7). Relations between exclamative and relative clauses have been observed a lot across languages, especially when degree is involved (Culioli 1974; Delfitto and Fiorin 2014; Elliott 1974; Faure 2012; Rett 2009, a.o.). The Greek structure could similarly be the relativization of the degree to which the speed of Kipchoge’s running is measured in (1a). The exclusion of oti and na follows, since they are excluded from relative clauses as well.

The unacceptability of (6) matches the robust crosslinguistic reluctance against embedding under emotive predicates. Consider the French wh-exclamative in (8a).^[11] It cannot be embedded under contrary emotive verbs (8b). Although for étonner ‘astonish’ and surprendre ‘surprise’ this restriction can be ascribed to their difficulty of embedding wh-clauses in general (9a), this is not the case for admirer ‘admire,’ which accepts wh-clauses that are not exclamatives (9b)/(9c), but does not accept exclamatives (8b). Beyond wh-exclamatives, declarative exclamations are not embeddable either in that language. Observe the mood difference of the embedded verb in (10a) and (10b).

In English, embedded exclamatives are reported with emotive verbs, like (11) (Elliott 1974: 237, ex. 57). However, most examples in the literature require that a preposition mediates between the predicate and the wh-clause (12) (Rett 2011, ex. (11a)/(11b), adapted), suggesting that the preposition is the proper selector of the clause. Worse, Rett (2011: 436–439) observes that only wh-exclamatives and nominal exclamatives are embeddable under emotives, to the exclusion of, e.g., subject-auxiliary-inversion exclamatives, and that their interpretation differs (different possible continuations; no degree restrictions).

In this section, we have seen that emotive embedding is rather limited for exclamations in the languages in which it was investigated. It seems to be almost limited to ambiguous (exclamatory or declarative) clauses. This puts in jeopardy T&G’s idea that exclamations feature a silent emotive verb. This is not the case for epistemic embedding.

3 Epistemic embedding

In contrast to emotive predicates, epistemic (aka cognitive/semi-factive) attitude predicates are at ease with embedding exclamatives.^[13] Most of Elliott’s (1974) English examples feature epistemic predicates (13) [T&G’s (59) and (69)]. The French exclamative (8a) is comfortable with savoir ‘know,’ comprendre ‘understand,’ réaliser ‘realize’ ((14) to be contrasted with (8b)). In their corpus study of Classical Greek, Biraud et al. (2021) found 75 good examples of epistemic predicates versus 6 disputable examples of emotive verbs. The hoíois-clause in (15),^[14] with the emotive predicate aiskhúnomai ‘be ashamed at,’ is best construed as a direct exclamative: It is not coordinated with the object DP (in the preceding line) tàs emautoû sumforás and exhibits the particle ára, preferably found in direct speech (16).^[15] Conversely, the hōs-kalôs clause in (17) must be construed as an indirect exclamative, parallel to the direct (18).

In (Modern) Greek as well, embedding exclamatives under epistemic verbs is way better than under emotive predicates. Compare (19) to (6a).

Summing up, embedding exclamations under epistemic attitude predicates is easier than under emotive predicates. This observation holds crosslinguistically and for clauses that unambiguously convey exclamations.

4 Discussion: interlocutor subjective orientation

Even if it seems to contradict T&G’s stance, the double embedding possibility (under epistemic and, to a lesser extent, emotive attitude predicates) actually dovetails with their account. Recall from Section 1 that, in addition to the subjective veridicality presupposition, exclamations perform two assertions: Assertion of the descriptive content, assertion of the emotive stance. Each is reflected by an embedding option: The former under epistemic predicates, the latter under emotive predicates.

This is not the only confirmation for the two-assertion stance. The presence of two assertive layers in exclamations (described in Section 1 (B)) is also manifest from the two denial options that are offered and nicely illustrated by (4): B denies either the descriptive content (B) or the emotive stance (B′). Very importantly, the emotive stance that can be rebutted is not A’s attitude, but the hypothesis that B’s attitude matches with A’s attitude.^[16] This limitation is due to A’s attitude being a presupposition, as T&G show (point (C) in Section 1 above), and presuppositions are famous for not being directly deniable. Consider the failed denial of the existence presupposition of the definite description in (20) (Beaver et al. 2024, ex. (9)).

Nevertheless, under closer scrutiny, like the two embedding possibilities, B’s double option of reaction evinces an asymmetry between the two (emotive and descriptive) assertive layers. By uttering (4), A implies that B has the same attitude towards the exclamative’s content as hers/his. This implication can be taken to be part of the descriptive assertion, and as such to prompt the same reaction as assertions: B can deny the attitude or tacitly accept it by carrying on the conversation.^[17]

However, overt acceptance returns a somewhat different picture.^[18] Whereas denials can apply to both asserted parts, the descriptive content and the attitude, as we saw about (4), overt acceptance cannot so easily target the attitude, as visible from the infelicity of B′ in (21) and (22) (too and aussi left aside for the moment). This observation is also correct when we move away from predicates of personal taste, like delicious, and consider measurable dimensions like speed in (23)B‴ versus B″ (acceptance of the description), and in B and B’ (rebuttal of both assertions).

Most interestingly, adding too/aussi is required to render the utterances (21)B′, (22)B′ and (23)B‴ felicitous. Too is an additive operator, which, in our examples, presupposes that someone else likes John’s desserts or finds Kipchoge fast. Thus, A’s belief that p is true is presupposed, to wit, this test comes as a confirmation for T&G’s view that exclamations convey a subjective veridicality presupposition. It is not surprising that this subjective presupposition must first be validated by the interlocutor (here with too) if s/he wants to express her/his acceptance of the extension of A’s emotive stance beyond A’s sole emotive state.

Beyssade (2009: 29, ex. (31)) proposes a test involving non-shared belief that goes in the same direction.^[19] Whereas the approval of the interlocutor cannot be requested on the presupposed content in (24a), this request is admissible on the content of an exclamative (24b), suggesting that it remains open to the interlocutor’s validation.

If I am on the right track in proposing that the various options of embedding are each the reflex of one of the dimensions of exclamatory utterances, we can now understand why embedding under emotive predicates is more limited than under epistemic predicates. Subordination under epistemic predicates is just the reflex of the denial or acceptance of the descriptive content. Conversely, subordination under emotive predicates requires both this deny-or-accept move and that the subjective presupposition be taken into account. Most likely, this is why, when acceptable, the examples that are found in the literature feature performative-like properties on the matrix predicate (first singular, indicative present: (11), (15)).

5 Conclusions

Although this commentary has challenged the existence of a silent emotive verb in exclamations that T&G put forth, the main dimensions of their approach are maintained: There are two layers of assertion and a subjective presupposition in exclamations. To discuss and improve on their proposal, I extended T&G’s methodologies of dwelling on the comparison with subordination cases and on replies to exclamations. In particular, I claimed that subordination cases are not the equivalents to direct exclamations. Instead, each embedding possibility explicates ONE of the dimensions of exclamations: Assertion of the descriptive content, assertion of the emotive stance. Overt acceptance was used to confirm the presence of a subjective veridicality presupposition and refine the role that it plays in exclamatory interactions. We uncovered that, along with the speaker-oriented subjective presupposition and emotive assertion, the emotive assertion contains a subjective, interlocutor-oriented part, namely a request that the interlocutor adopt the speaker’s emotive state with respect to the content of the exclamation. This result corroborates the view that exclamations too are intersubjective items,^[20] a dimension that must not be ignored in future attempts of analysis and formalization.