The landscape of exclamations: brief musings on emotive language

1 Introduction

In our target article, we proposed a new framework for the study of syntactic configurations commonly summarized under the terms ‘exclamative’ and ‘exclamation’, respectively. Our overarching conceptual goal is summarized as follows (p. 3):

We will argue that there is no illocutionary force of exclamation in the first place – and this goes against almost all the previous accounts. If we assume that there is no such thing as the illocutionary force of exclamation, then there can also be no conventional association between the (non-existing) force and a specific syntactic structure (like in accounts that propose a dedicated sentential force aka illocutionary mood for exclamatives […]).

Accordingly, we presented what we hope was convincing evidence and arguments for the claim that there is no speech act of exclamation. A corollary of that claim is that there is then also no clause type called ‘exclamative’, if such a clause type is defined by featuring a dedicated force other than assertion, question, or command (which are the major illocutionary forces according to most of the syntax-semantics literature). It is important to also note that empirically there is no specific form such as exclamative mood (unlike imperative mood or question particles). The empirical observation in our paper was that the effect of exclamation maps onto a number of distinct morphosyntactic structures, recycling existing morphology (e.g. wh-forms, degree adverbs, and expressive words).

As for the semantics of what has been called exclamation in previous work, we argued that an utterance featuring this effect is an emotive assertion that “asserts emotion, is veridically committed to the truth of the sentence, and presupposes intensity” (p. 5). Here is a summary of the semantic ingredients of exclamation:

As we noted, our approach did not come in a vacuum. Michaelis and Lambrecht’s (1996: 239) idea of “assertion of affective stance” proposed in the context of nominal extrapositions such as It’s amazing, the difference! is a precursor, and in his commentary Richard Faure highlights his earlier work arguing that there is no exclamation illocutionary operator (Faure 2012) and that exclamativity can be treated uniformly across various clause types (Biraud et al. 2021). In other words, while our program for exclamations and its specific implementation is new, building a theory of exclamations as an emotive attitude with intensity looks like a desideratum in the literature. We offered a way to implement it based on a closer look at German and Greek data. The main syntactic claim that emerged from these particular data is that the syntax of the specific structures we looked at involved ellipsis of a higher emotive predicate, and that those structures do not instantiate a distinct clause type called ‘exclamative’, but are configurations embedded under a higher assertive predicate.

In our article, we wanted to articulate a new analysis of exclamation, one that will be broad enough to be used as a framework for new sets of data, but also specific enough about what the semantic and syntactic ingredients of an exclamation are. We are very thankful to our ten commentators who took up our proposal and built, quite literally, a landscape of exclamation patterns in numerous languages, therefore substantiating our claim that exclamations are not one type of sentence but a quite rich and diverse set of structures. We expected indeed some resistance to our proposals that neither exclamation (as a distinct illocutionary force) nor exclamative (as a distinct clause type) exist, and we take it as a sportsman! Reading all these inspiring commentaries, we become only more convinced that our approach is on the right track – but of course it is a marathon, not a sprint. Some of the data pose challenges that need to be addressed in future work, and we are confident that future researchers, like the hero of our examples Eliud Kipchoge, will not lack the (analytical) stamina needed to pursue this complex task.

In this short reply, we cannot do justice to all the empirical points raised by our colleagues, but we will try our best to clarify three main issues that surfaced (in different versions and in both supportive and critical tone) in all seven commentaries. We start with the pervasive nature of ‘exclaiming’ in Section 2 where we acknowledge the significance of the diversity of linguistic means and syntactic configurations that can be used in the context of what has been traditionally called exclamation. In Section 3 we address the issue of diverging judgments and introspective data that has long formed the branch of linguistic work concerned with exclamations and exclamatives. Section 4, finally, addresses the topic from a broader perspective, concluding our discussion with a new proposal for interjections such as ouch! and wow! as super-linguistic in character, on a par with emojis (as described in recent work by Grosz et al. 2023).

2 The pervasive nature of ‘exclaiming’

In his commentary, Foolen observes that in Searle’s (1976) classification of illocutionary acts, there is no mentioning of the verb exclaim in the list of speech act verbs. Foolen hypothesizes that this is because the verb exclaim, in contrast to those verbs referring to expressive speech acts in Searle’s list, cannot be used in an explicit performative way. We share this judgment because if I would say ‘I hereby exclaim what a mess we’ve gotten into’ the performative character of the utterance could be realized by exclamation prosody, but not by just using the verb exclaim – this is different from the other verbs listed by Searle where the very use of those verbs (e.g., ‘I hereby apologize for…’) performs the respective speech act.

We start with Foolen’s remark here because we think that there is a general point in this observation that supports our approach: the meaning referred to as ‘exclamation’ does not form a speech act (more precisely: an expressive speech act) of its own. Rather, it can potentially be added to any speech act. Crucially, in support of this view, Searle and Vanderveken (1985 :179) explicitly mention the verb exclaim and highlight that it falls into one category with verbs like announce or interject (more on interjecting in Section 4 below):

The verb “announce” is a case in point: it does not name an illocutionary force since almost any illocutionary force can be announced. Announcing carries no restriction as to its illocutionary point. One can announce statements, promises, warnings, orders, apologies, etc. The same holds for the verbs “interject”, “interpose”, and “exclaim”.

In many languages, this pervasiveness of exclaiming is prominently instantiated by prosodic means, as Esipova highlights in her commentary for Russian. Villalba in his commentary calls those prosodic strategies ‘a kind of magic wand that converts the sentence it touches into an exclamation’, and he highlights that our proposal is very much compatible with this fact because we are claiming that what has been called exclamation force can actually be found across different (all?) speech acts and syntactic configurations; we provided examples for metalinguistic comparatives (3a), non-canonical questions (3b), and emphatic imperatives (3c) in our paper (p. 36); see Giannakidou and Yoon (2011); Trotzke (2023) on the first two types.

Exclamation intonation and the corresponding exclamation meaning can apply freely in all those examples, and there is no need to assume a dedicated force other than assertion (3a), question (3b), or command (3c). These remarks (and the supportive clarifications by Villalba) indicate that we fully acknowledge the important (even primary!) role prosody plays in many languages for the domain of exclamations. The emerging picture lends support to our approach which emphasizes that ‘exclamation’ can be added to potentially any linguistic utterance since it is a mode of modifying the existing type of sentence with intensity; exclamation and the act of exclaiming is therefore pervasive and operates at a different level than illocutionary forces, enhancing the illocutionary force of the respective utterance.

We do mention prosodic differences at several points in our target article, but at the same time we make it very transparent for the reader that our focus lies somewhere else. Let us clarify this focus again. The starting point of our collaboration was that we both strongly believed (for different reasons) that there is no such thing as an exclamation force – not the least because of the observations about intonation above. Coming from a syntax-semantics perspective, we quickly realized that the prime example in the syntax-semantics literature on exclamation force are degree wh-exclamatives. Those structures figure so prominently in the literature because very often they are syntactically realized by structures that seem to be confined to expressing an exclamation reading (e.g., English How fast he is!), hence the postulation of a dedicated clause type ‘the exclamative’. In order to make our case that there is no exclamation force in the first place, we focused on those structures and tried to show that they can actually be analyzed in terms of one of the three major forces (in this case: assertion), with no need to postulate an additional force called ‘exclamation’. In other words, in order to show that there is no distinct force of exclamation, we had to address syntactic configurations whose main point (prima facie and according to most of the previous literature) is to exclaim something rather than to assert something, question something, command something, etc.

One major misunderstanding we saw in some of the commentaries is that our syntactic analysis for some wh-exclamatives (and later in the paper for that-exclamatives in German) as elliptical structures of an assertive matrix predicate should also hold for any form that can be used as an exclamation. For instance, Bücking highlights for German that it ‘remains a major puzzle how to derive the emotive meaning in those cases where the verb is fronted’; see the following non-fronted (4a) and fronted (4b) examples:

We think there is no puzzle here at all. We have not claimed that our syntactic analysis for (4a), with verb-final/embedded word order, should also hold for (4b) or other cases featuring V2 syntax, see (5a). In fact, in (4b) above we are looking at a case that can be analyzed just like other cases of emphatic fronting such as (5b); see Trotzke (2017) on those marked syntactic strategies that are combined with mirative aka exclamation intonation, many times also with emphatic lengthening strategies affecting fricatives and vowels (Trotzke and Turco 2015).

In other words, the so-called wh-exclamative in (4b) above is nothing else than a declarative with an obligatorily fronted adverb phrase (i.e., [_AdvP how fast], in those cases, is equal to [_AdvP so fast]). Again, we see no puzzle here, just another linguistic means to yield/co-construct the effect of exclaiming. We do not propose our syntactic analysis as a universal syntax of exclamations; we do advance the point, after all, that there is no clause type ‘exclamative’ and arguing for a universal syntax for all exclamations would contradict this very point. We did use the embedded pu complementizer of Greek (as well as German that-exclamatives) as an argument for embedding (see also Faure’s commentary for more relevant structures in French and Modern and Classical Greek) – but in the broad landscape of exclamations we expect different syntactic strategies to emerge in a systematic way.

We also want to address some of Esipova’s comments in this context who asserts that ‘[j]ust because a segmental string is compatible with being a matrix wh-question does not mean that it can’t also be compatible with being an exclamative.’ Given our examples and remarks here, we would reply that (4b) above is indeed string-identical to a matrix wh-question, but there is a lexical ambiguity between how as a wh-pronoun and how as an adverbial denoting extreme degree (‘so’). Accordingly, (4b), if interpreted as so fast, instantiates a V2 declarative with emphatic fronting, and not a V2 wh-interrogative with fronting of a wh-pronoun for reasons of clause typing. Crucially, we do not need to postulate an extra clause type (‘the exclamative’) for those cases because we can account for them by means of our standard repertoire of clause types (e.g., declarative vs. interrogative). Again, the driving hypothesis of our enterprise is to say that ‘exclamative’ (as distinct clause type) and ‘exclamation’ (as distinct force) are extra assumptions that one can do without easily once we look at the phenomenon of exclamation as a pervasive feature that cuts across linguistic utterances of any kind. That has been (and still is) the spirit of our enterprise, and we firmly believe that the vast literature and diversity of claims about exclamatives and exclamations are in dire need of Occam’s razor like only very few other empirical domains in theoretical linguistics.

In this context, Castroviejo in her commentary summarizes ‘that languages recycle different syntactic structures for the purpose of exclaiming (wh-clauses, that-clauses, DPs, …); providing a unified compositional account seems an extremely ambitious goal.’ Our response is to recall our analysis in (2) and say that these delineate the semantic ingredients of exclamation, but there is no syntax-semantics entity ‘exclamative’ that would account for all varying forms indicated by Castroviejo in a uniform way at the syntax-semantics interface. The semantic ingredients can map onto a number of syntactic types, as it arises from our discussion: wh-forms, SO-intensity components, etc. What we do claim, in other words, is that there is a common semantics that all those different syntactic, lexical, and prosodic strategies share – and that this can be characterized at the level of speech acts as ‘emotive assertion’; in (6) applied to our key example about being surprised that Kipchoge ran so fast, cf. p. 28:

Crucially, this notion of emotive assertion dovetails nicely with observations in the commentary by Faure where he refers to the notion of ‘strong assertion’ whereby a speaker would like to yield an effect of ‘enforcing the interlocutor’s acceptance’ (Faure). Maybe this is the key difference of exclamations to utterances such as Ouch! (see Villalba’s comment and our remarks in Section 4) because they do not feature an addressee-oriented and dynamic perspective per se.

All in all, while the specific syntax of ellipsis proposed in our paper is claimed for the degree wh-exclamatives discussed in our article (and later in the paper also for verb-final that-exclamatives in German), the speech-act meaning of emotive assertion holds for all conceivable forms of exclamation. Note that we thereby avoid singling out individual syntactic structures and proposing ad-hoc notions such as ‘true exclamatives’ (see Esipova’s comment). What are ‘fake exclamatives’ then?

3 Tastes differ: a plea for running more experiments

One of the main reasons for one of us (Trotzke) to start working on exclamatives and exclamations as a fun side project was that the judgment data presented in some of the most prominent works on this topic simply seemed counter-intuitive and not accurate. Our target article mentions introspective data by Rett (2008, 2011), but also parallel judgments on German exclamations by d’Avis (2016) with which we do not agree. There are only very few studies on exclamations that tested those recurring and reproduced judgments in actual experiments; we mention Trotzke (2019), Villalba (2017), and Trotzke and Villalba (2020). From their data, we concluded in our article that the descriptive content also of degree wh-exclamatives can be denied, and that this type of exclamative can also be used as a felicitous response to an information-seeking question.

Most of the commentators have deeply engaged with those patterns, and we thank them for all the interesting data points, especially the new ones! Some of them will certainly be on our bucket list for future judgment experiments to come because this is all uncharted territory in terms of empirical studies. In what follows, we will be very selective because there are simply too many new judgment data in the seven commentaries. We decided to mention one crucial data point/comparison that has been mentioned as a blind spot in our article by both Castroviejo (in our favor) and Esipova (against our claims). At the end of this section, we will address criticisms by Wilhelmy et al. to illustrate something more general about how imperative it is to study this topic empirically in order to enter a scientific conversation at all.

Let us start with the points by Castroviejo and Esipova. Both commentaries mention that we have not provided the patterns for the non-elliptical emotive assertion in the denial contexts (Section 2.2 in our target article) and in the answering contexts (Section 2.3 in our article). Let us start with Castroviejo’s remarks about the answering contexts. She adopts our examples in (7) and completes the pattern by adding the possibility in (7d):

As predicted by our approach, (7d) can as well be used as a felicitous response to (7a), which is asking about the descriptive content (the ‘object of belief/surprise’/p ₄, according to Castroviejo). Interestingly, Castroviejo points out that all three options (which are semantically equivalent according to our account) can also be used as a response to a question which asks about the emotional state (p ₁, according to Castroviejo’s classification):

We agree with her judgment that not only the non-elliptical version in (8d), but also the elliptical version in (8c) as well as the declarative in (8b) would be felicitous responses. If that would be corroborated by more judgment data, then this would support our approach that all three utterances (including the degree wh-exclamative in (8c)) perform emotive assertions.

Esipova also highlights that one would have to contrast the elliptical versus non-elliptical versions of emotive assertions more carefully, and she provides the following judgments for the denial patterns discussed in Section 2.2 of our article:

Regarding those examples, she claims that ‘B’s responses in (10) might sound rude or overstepping, but they are by no means infelicitous in the same way as they are in (9)’. While we agree with the tendency that (9B) sounds weirder than (10B), we do think that the weak denial strategies (which are generally preferred by subjects in analogous judgments; see Trotzke 2019) sound just fine:

If Esipova would agree, then this might indicate that the conclusions she draws from the patterns in her commentary would have to be re-evaluated. But, of course, we cannot know for sure whether we find agreement with her about data patterns like that or not – and this points to a more general issue in the literature on exclamatives and exclamations: data judgments differ considerably.

For us, this was most obvious in the commentary dealing with German data by Wilhelmy et al. We pick out this particular commentary because throughout this piece, the authors criticize us for having only tested a subset of exclamation forms in German experimentally while they, at the same time, provide a huge amount of introspective data about additional forms that they have never assessed empirically whatsoever. Many times, Wilhelmy et al. do not even provide the data they are claiming to be mere facts about German exclamatives. For instance, they present the following two examples from the experiments by Trotzke (2019):^[1]

The finding in Trotzke (2019) has been that the descriptive content of (11) and (12) can perfectly be denied by a variety of strategies, preferably by something like That’s not quite right,… or Not really,…: ‘That’s not quite right, she always looks the solution up in the textbook’ (11), ‘Not really, she just uses a lot of makeup.’ (12); see Trotzke (2019: 529) for the German original examples.

Regarding those patterns, Wilhelmy et al. criticize that ‘[i]n the 2019 study, only the wh-exclamatives and dass-exclamatives containing a scale-indicating element were used […]. If these elements (e.g. schwierig and schön) are removed […], the exclamative sentence can no longer be denied’. They present this as an empirical fact, without even showing us the (allegedly) infelicitous denial pattern they are referring to. We happily do them a favor and complete one of their suggested structures with the denial pattern:

Honestly, if the denial pattern is ok for (11), which has been experimentally corroborated, we see absolutely no difference in the felicity of (13B). If Wilhelmy et al. still do, they would have to make an effort and run a study. We are perfectly aware of the theoretical reasons for why they would like (13B) to be infelicitous, but this is not the way the science of linguistics works for us – the theory must fit the data, not the other way around.

Unfortunately, Wilhelmy et al. sometimes misquote our examples (and hence our claims): In their example (17), they translate our broad-focus context Tell me, how did Eliud Kipchoge do in the race? as Tell me, how fast was Eliud Kipchoge? (German: ‘Erzähl mal, wie schnell war Eliud Kipchoge?’), thereby changing it to a narrow-focus context. Based on this misrepresentation, they go on and discuss the introductory phrase Tell me…, building an argument about broad narrative ranges instead of engaging with our argument about mismatches having to do with narrow vs. broad focus.

All in all, some of the commentaries object to some of our judgments. That was of course expected. However, in our target article we make clear that our whole theoretical argument is based on some more recent attempts to add data other than introspection to the study of exclamations. Given the heterogeneity of tastes in the commentaries of what is and what is not ok in a certain pragmatic context, we encourage all commentators to help building a scientific discourse about exclamations that is based on either experiments or also corpus data (as Foolen remarked in his commentary). We enthusiastically exclaim: There is still SO much to do!

4 Interjection and exclamation: interjections as verbal emojis

In this last section, we will address an issue raised by Xavier Villalba’s commentary who cautions that our theory ‘leaves us without a proper account for purely expressive acts like Ouch! or Wow!’ He very correctly points out that he doesn’t think that ‘anyone will be happy saying that these utterances are assertions of the speaker’s emotion, for there is no propositional content we can resume with an anaphor’, and Villalba provides the following patterns to illustrate that observation:

We agree completely that expressive interjections are not assertions. But then, Villalba asks, ‘what kind of beast are they? In classical speech act terms […], these utterances do not seem to have a direction of fit (the hallmark of expressive speech acts), unlike assertions, imperatives or questions. Hence, if we do not have an exclamation speech act, these small particles might be the elephant in the room for the proposal by Trotzke and Giannakidou.’

We will offer here some thoughts on how one can go about interjections in the framework we have developed, starting with the point that interjections, while expressives, do not qualify as exclamations in our approach. When used by themselves as in examples like the above, they seem to be more or less involuntary ‘reactions’, and in this they remind us of co-speech gestures which happen automatically as we speak, and do not by themselves have transparent propositional content. For instance, if I utter Wow!, that could mean any number of things ranging from Wow! That was an awesome movie, to Wow! That was a waste of time! or Wow! What an asshole! Likewise, Ouch! could mean Ouch! That hurts!, or Ouch! That tickles!, or Ouch! Keep away! It is therefore important to acknowledge that it is the additional proposition that gives that actual content, and the interjections are complementary to content, just like it has been discussed for gestures (see Ebert 2024 for a recent overview). Notice also that those interjections can apply to different kinds of speech acts:

If interjections are gestural, their function is rather parallel to super-linguistic objects of study, specifically similar to that of face emojis studied recently by Grosz et al. (2023). The authors in this work observe that face emojis (disc-shaped pictograms with stylized facial expressions that accompany written text) have a use in which they comment on a target proposition expressed by the accompanying text, as opposed to making an independent contribution to discourse. Notice that emojis too combine with various speech acts:

(16)

a.

How did the interview go?

b.

How are you coping?

c.

I found 100 dollars on the street!

Grosz et al. (2023) suggest that emojis are part of multi-modal discourse and interact with text. In particular, emojis contribute information about the participants’ affective attitudes towards propositions that are expressed by linguistic components of the discourse. Following their idea, we will propose that interjections are a kind of verbal emojis commenting not on text but on speech: they comment on a target proposition, in light of the way that proposition bears on a salient value, priority, or goal held by the speaker. This component is important as it shows the dependence of interjections, just like emojis’ dependence on what the speaker desires, aspires to, wishes for, or hopes for as a discourse value established in the context. Informally, we propose the following dependence analysis, analogous to what has been suggested by Grosz et al. (2023: 914) for emojis:

We are suggesting that interjections and emojis belong to the same class of commenting expressives (as opposed to independent expressives like damn etc.) and their function is super-linguistic: they are an additional part of the speech act they occur in, and their function is to comment based on an emotive value of the situation prior to it. We will not offer here a fully-fledged analysis of interjections, and we will also not address whether they are syntactically attached to the content like adverbs. With expressive speaker-oriented adverbs such as unfortunately, alas, etc. we do have syntactic attachment, but interjections seem to be exactly like emojis, i.e., separate components. In sum, we hope that this line of thought could be a first step for addressing the issue raised by Villalba above to distinguish between emotive assertions (aka exclamations, according to the previous literature) and verbal gestures like Ouch! and Wow!

To conclude, let us stress that we feel very energized with the issues being brought by the response pieces of this volume – and we thank the authors for bringing to the fore the rich and multidimensional landscape of exclamations and their relation to other forms of expressing emotion in language. Emotive discourse seems to be a central part of linguistic interactions, and we are very happy to see that our proposal for exclamations stimulated a lively and very necessary conversation. Finally, we want to thank again our editor, Hans-Martin Gärtner, for encouraging and supporting this conversation.