2.1 Humour theories

Raskin (1985) develops a theory, the Semantic-Script Theory of Humor, (SSTH), a theory of humour competence whose intended coverage is primarily short verbal jokes.

Ideally, a linguistic theory of humor should determine and formulate the necessary and sufficient linguistic conditions for the text to be funny (Raskin 1985), p.47).

The main theoretical innovation one finds in (Raskin 1985) is an attempt to explicate textual incongruity formally:⁷

Scripts were introduced by Schank & Abelson (1977) and can be explicated in various ways (Raskin 2017); the same applies to opposite. We will make a proposal of our own as regards incongruity, one that refines (2). The theoretical apparatus for explicating jokes in subsequent theories such as the General Theory of Verbal Humor (Attardo & Raskin 1991; 2017) is significantly more complex.⁸ However, (2) remains a key notion. This approach ties incongruity to the general theory of scripts, viewed as embedded in ontology construction (Raskin 2017); this might have seemed idiosyncratic when it appeared in the 1980s, but currently is closely related to frames (Fillmore 1985; Löbner 2015) and to semantics as ontology construction (cf. situation semantics (Barwise & Perry 1983; Ginzburg 2011)).

Nonetheless, work on humour theory exemplified by SSTH and the GTVH has different aims from ours: our aim is to explicate the possibilities of laughter occurrence in conversation, in particular the relation between laughter events and laughables, the triggers or objects of laughter (more on which see sections 2.2 and 3.). In contrast, the avowed quarry of humour theorists is characterizing humour competence. (3) taken from the corpus collected for (Priego-Valverde et al. 2018) exemplifies how laughter can emerge without the laughable per se having any incongruous aspect. In this case the laughable is the utterance “this morning’s semantics class”, which is an answer to the question “What would you like to speak about?”. Nothing in this question/answer sequence is incongruous relative to any script or other related notion encoding domain knowledge; the laughter arises from some implicit shared knowledge about the denoted event. Minimally, this means that a theory of laughter needs to involve contextual resolution of implicit aspects of laughables.⁹¹⁰

However, a more significant point, exemplified copiously below, is that much laughter in conversation is not related to jokes or witticisms; indeed, various corpus studies reveal a high percentage of speaker laughter which lacks the surprise element key to jokes (Provine & Yong 1991; Vettin & Todt 2004).

2.2 Conversation Analysis and laughter

Gail Jefferson initiated work within Conversation Analysis on laughter (e.g., (Jefferson et al. 1977; Jefferson 1979)). A crucial assumption that Jefferson makes is to treat laughter as a possible response in the range of responses an utterance generates. This is an assumption we adopt in the sequel. Specifically, Jefferson takes a strategic view of laughter, viewing it as a social activity that a speaker can invite others to join in, one which occurs frequently with material which has no humorous aspect as such. This is exemplified here by (4a). These are instances of antiphonal laughter, “responsive laughter” whereby the responder shares the initial laugher’s assessment of the situation. Jefferson also points out cases such as (4b), where laughter is not responded to with laughter, though the responder addresses the direct issue raised.

Capturing these contrasting options is clearly a requirement for any theory of laughter in conversation, as is the need for distinguishing laughter potential in 2-person and multi-party conversation, in light of observations by Glenn (2003). Another notion arising from (Glenn 2003) is the notion of laughable: laughter marks its referent as laughable and potentially funny. This is a key starting point for any semantic theory of laughter. Work in CA has documented the use of laughter in a variety of non-humorous settings: medical (Haakana 1999) (where doctors often avoid responding to a patient’s laughter with laughter), job interview situations (Glenn 2013) (see example (20a) below), political interviews (Romaniuk 2013), where politicians use laughter to deflect questions, exemplified here in (5), which is further discussed in section 4.

Conversation Analysis work on laughter has, as we have seen, led to many important insights. As Glenn & Holt (2013) explain CA associates laughables as “referents” for laughter but, as with CA work in other domains,¹¹ explicitly assumes no semantics beyond this (repeated from footnote 2): “… Although laughter lacks semantic or linguistic content, variations in its production contribute to its communicative value.” (Glenn & Holt 2013: 6); indeed in their survey, Glenn and Holt point out how CA work recognizes a variety of effects laughter can produce, while offering a careful description of these uses and the positions in which they occur:

But in the absence of anything more than a “referential semantics” in terms of laughables these remain an essentially arbitrary list of effects. Moreover, since CA avoids any explicit means of representing emotion, in saying that laughter can serve as a stance marker, it has no way to distinguish laughter in such cases from verbal post-completion stance markers such as yea and mmh; (see concrete discussion of such cases in section 3.6 below.). Nor can one capture the non-translatability of laughter potential (see below section 4.3.3).

3.1 Laughter and external events

The view of laughter we develop assumes that laughter is ambiguous and in many cases has propositional content that arises from an eventive source. Starting with the last point, we wish to emphasize a certain similarity in behaviour with eventive anaphors, here exemplified by demonstrative NPs. Both can take as antecedents exophoric antecedents (see (7)), the event described by the preceding utterance (see (8)), an event occuring utterance medially—e.g., a sub-utterance or co–speech gesture (see (9)), and, finally, an event described cumulatively (Asher 1993) (see (10)):

Each of the events predicated about in (7)–(10) are what we refer to subsequently as the laughable, a notion introduced in a related if not identical sense in the CA literature (see section 2). This laughable l, we suggest, is a constituent of laughter content, which informally can be either (11a) or (11b). We leave both incongruity and pleasure unexplicated for now until sections 4 (for incongruity) and 5 (for pleasantness).¹²

Indeed already the data (7)–(10) provides us with the means to distinguish laughter from unintentional bodily sound productions like sneezing and hiccuping. With respect to e.g., (7)–(8), if C enters and sees B laughing, she could pose the question in (12a). B’s response indicates that the laughter is directed at certain events she has perceived (visually, aurally, or via touch, as with tickling) with her laughter being used to express her view that this is, say, incongruous or pleasant—the difference here is immaterial; this contrasts clearly with (12b,c), where the trigger is bodily:¹³

One could be tempted also to say that there is a contrast in (12) in that sneezing and hiccuping are involuntary. However, while some cases of external–event–directed laughter seem to be as voluntary (say in the sense of “consciously planned signals”) as verbal speech, and while in some neuroscience studies of laughter under experimental conditions in a laboratory, a distinction has been drawn between genuine, “involuntary” and fake, “planned” laughter (Wild et al. 2003; McGettigan et al. 2013), this is far harder to assess for laughter in spontaneous conversation (Wierzbicka 2000; Gervais & Wilson 2005).

Nonetheless, while this dimension is significant pragmatically, we do not take it to be qualitatively more significant than it is for verbal utterances. Indeed one can find similar, “involuntary” occurrences with the latter, in other words verbal utterances that take place without the speaker’s (full) “conscious control”—apologies, as in (13a), exclamations (13b), or “errors” such as making an utterance in the “wrong” language (13c):

At the same time, the two main imports communicated by laughter—however packaged—incongruity and pleasure are both susceptible to occur “spontaneously” and “reflectively”. Thus, when someone laughs at a joke, an instance of an incongruity, we are likely to consider whether the laugh is “genuine/spontaneous” or “fake/forced”. A child laughing wildly while playing with a toy is an example that would obviously be viewed as spontaneous pleasure, as would a mutually flirtatious couple (Fuchs & Rathcke 2018), though in the latter case the issue of “genuine/spontaneous” or “fake/voluntary” is certainly one that can arise.

Given that interlocutors are aware of the possibility that laughter can be “spontaneous” or not, this is an important pragmatic issue that potentially arises in any context when non-spontaneity is associated with lack of sincerity.¹⁴ This, then, actually ties laughter (and smiling, frowning, crying etc) to verbally expressed speech acts, for which an analogous sincerity issue arises frequently, as much emphasized by Austin (1962); Searle (1969).

With some initial notion of what laughter content consists of, we now turn to data illustrating further motivation to posit such content, its interaction with verbal content, and the ambiguity of laughter.

3.2 Laughter: stand alone uses

One additional piece of evidence for the claim that laughter has propositional content is that it can be used as a stand-alone utterance whose content incorporates content from the immediate context and such that changes in the linguistic context lead to different contents for a laugh: it can convey—we will suggest in section 4 via pragmatic inference— a negative response to a polar question, as in (14a)¹⁵ or disbelief in the previously uttered assertion, as in (14b), based on an advertisement from the Paris metro, the biblical (14c), and (14d); (14e,f), taken from newspaper reports on parliamentary debates, offer essentially such a paraphrase verbatim:

(14)	a.	(Context: Bayern München goalkeeper Manuel Neuer faces the press after his team’s (Dreierkette) defense has proved highly problematic in the game just played (3-2 against Paderborn).)
		Journalist: (smile): Dreierkette auch ‘ne Option? (Is the three-in-the-back also an option?) Manuel Neuer: fuh fuh fuh (brief laugh) ↝ The three-in-the-back is not an option! (audio: https://tinyurl.com/yxhh7pvf)
	b.	A: I will take care of your savings. B: (laughs)
		↝ I don’t think you will take care of my savings!
	c.	God: You will at age 99 with your aged wife Sarah have a son. Abraham: laughs. ↝ I don’t think I will at age 99 have a son …
	d.	Theresa May: They will know we built them a better Britain. B: (laughs) ↝ that’s a ridiculous claim (I don’t believe that they will know we built them a better Britain) (https://www.youtube.com/watch?v=HxN1STgQXW8 at about 2:25.)
	e.	“While I was agreeing the future of Nato with President Trump, he was joining a protest march against him,” she said, inadvertently prompting loud laughter from MPs at the idea that her talks with Trump had secured the future of Nato (emphasis added—authors) https://www.mirror.co.uk/news/politics/mps-openly-laugh-theresa-boasts-12941471
	f.	She has listened very carefully to what has been said, she says. That triggers loud laughter. (emphasis added—authors) There is broad support for many aspects of the deal, she says. That triggers more laughter. (emphasis added—authors) (The Guardian, Dec 10, 2018)

As expected from a signal bearing propositional content, laughter can have a force that may be disputed in conversation, as illustrated in (15):¹⁶ in (15a) the audience laughter indicates that they view the speaker’s claim as non-serious; the lecturer’s retort disputes this reaction (though in a somewhat double–edged way, transferring the belief in the claim he appeared to be making to General MacArthur); in (15b) Unknown disputes Danny’s laugh and the funniness it putatively ascribes to someone being tied to a railway track:¹⁷

3.3 Meaning–affecting placement

Just as with other meaning–bearing linguistic elements, varying the placement of laughter can vary the content—this applies to other non–verbal social signals such as smiles, sighs, eyebrow raising, sniffles (see (Krahmer et al. 2002) for eyebrow raising and focus.). Here we point to the effect achievable by low arousal laughter within an utterance. This enables such laughter to serve as a scare-quoting device (Predelli 2003): in (16a(i)) the term long term friend is presented as a euphemism, whereas in (16a(ii)) it is the term friend which is at issue; (16b) exemplifies an analogous effect where in (i) the issue is what is to be pressed, whereas in (ii) it is what entity the red object is; similar remarks apply to (16c), which involves a sigh rather than a laugh:¹⁸

3.4 Laughter and conversational implicature

At least some of the implicatures that have been the focus of much work in pragmatics since (Grice 1989) and whose origin (grammar or intentional reasoning) is currently a source of much debate (e.g., (Chierchia 2006; Geurts & van Tiel 2013)) can be triggered as a result of laughter and smiling. In (17a) a quantity implicature arises by exploiting a ⟨laughter, smile⟩ scale,¹⁹ whereas in (17b) a relevance implicature arises, presumably on the basis that a laugh in such a context—unless it is seen to be a nervous laugh—expresses pleasure at the news communicated by A, which is inappropriate:

3.5 Irony

As we saw when discussing scare quotation, laughter can be used to mark irony—to indicate that an utterance is meant to deviate from its literal meaning. By the same token, irony can affect laughter, smiling, or crying: in all cases, there is a composition of a basic meaning of the non-verbal signal with a meaning roughly paraphrasable as ‘I don’t really mean this’. Mouthing “ha ha” as a reaction to a joke or humourously intended comment conveys that the speaker does not actually find the joke funny or the comment humourous, as in (18a,b); similar comments mutatis mutandis apply to smiling, crying and sighing, as in (18c,d,e):

(18)	a.	A: Tells joke. B: Ha ha.
	b.	MR. WHITE: You talked to Nice Guy Eddie? Why the fuck didn’t you say that in the first place? MR. BLONDE: You didn’t ask. MR. WHITE: Hardy-fuckin-har. (From the film Reservoir Dogs.)
	c.	Wes Brown smiles as he is mistakenly sent off. (The Independent, 28.02.2015)
	d.	A: And we were only give two macaroons each. B: (Mock cries).
	e.	Mock sighing C: [Well do] you wanna grab me a beer? P: tsk (h::) ..Guess. C: Please? P: (hx::). C: … Watch out for the lasagna. (Example (7) in (Hoey 2014))

3.6 Laughter ambiguity: a non–incongruous force

The data we have seen hitherto all involve laughter that communicates an essentially “disruptive” or “negative” force—used to communicate irony, scare quotation, question unworthiness etc. We will unify these within one meaning incongruity laughter. However, not all laughter has such a force. There is a significant class of “positive”, affiliative, empathetic laughters, which we will unify as pleasant (a notion we introduced earlier and is explicated in detail in section 5.). In his novel The Book of Laughter and Forgetting Kundera (1996) explicitly suggests the existence of two forces laughter can have, which correspond quite closely with the opposition we propose to make here.²⁰

(19) exemplifies such an empathetic laughter, where there is no incongruity to associate with the laughable, the pleasing prospect of Fred’s holiday:

(20) seems to function as an acknowledgement; the laugh is a more effective means than a simple, verbal acknowledgement; a verbal acknowledgement (e.g., Right or Yeah) would involve explicitly agreeing with a compliment and could be perceived as arrogant; Empathetic or affiliative laughter conveys pleasure of sharing the situation with the speaker and, only indirectly (by a mechanism elucidated in section 5) acts as agreement:²¹

Laughter can occur simultaneously with speech, both in cases of speech laughter, where a speaker speaks and laughs simultaneously (Kohler 2008), and where one interlocutor is speaking and the other is laughing. Is this an aspect that distinguishes speech from laughter?²² A detailed answer would need to consider and compare in a detailed way overlapping in speech and laughter, which we cannot do here. Nonetheless, at least qualitatively, one can offer a negative answer in that verbal acknowledgements/back channels, akin to but distinct in effect to empathetic laughs, can certainly occur simultaneously with speech by another interlocutor, particularly during a long turn by the latter (e.g., where the latter is a lecturer responding to questions in an academic setting.). This applies equally to other proposition–denoting gestures such as (negation–conveying) head shaking (Kendon 2002) and head nodding (Wagner et al. 2014), as well as their non-sentential analogues yes and no.

4.1 Background on KoS and TTR

4.1.1 Cognitive states in dialogue

We formulate our account within the framework of KoS (Ginzburg 1994; Ginzburg & Cooper 2004; Larsson 2002; Purver 2006; Fernández 2006; Ginzburg & Fernández 2010; Ginzburg 2012)).²³ KoS is a theory that combines an approach to semantics inspired by situation semantics and dynamic semantics with a view of interaction influenced by CA. On the approach developed in KoS, there is actually no single context — instead of a single context, analysis is formulated at a level of cognitive states, one per conversational participant. Each cognitive state consists of two ‘parts’, a private part and the dialogue gameboard that represents information that arises from publicized interactions.

The type definition—in a formal sense we will shortly elucidate— of a cognitive state is given in (21a). We focus on the public part—the dialogue gameboard, though will consider an aspect of the private part when discussing how to integrate emotion into the former. The structure of the dialogue gameboard is given in (21b) — the spkr,addr fields allow one to track turn ownership, Facts represents conversationally shared assumptions, VisSit keeps track of the visual situation including the focus of visual attention, Pending and Moves represent respectively moves that are in the process of being or have been grounded, QUD tracks the questions currently under discussion. Of these contextual parameters one, VisSit, is probably never entirely identical across participants since distinct interlocutors do not share the same pair of eyes, and moreover much of the time interlocutors have each other as their focus of attention. Nonetheless, there are various devices such as pointing or the verbal Look! to effect alignment. This point will apply later with renewed force when we extend the dialogue gameboard to incorporate emotion.

(21)	a.	TotalCognitiveState = def
	b.	DGBType (provisional definition) =def

To understand better the specification in (21), we offer a short discussion concerning the logical underpinnings of KoS. KoS is formulated within the framework of Type Theory with Records (TTR) (Cooper 2005; 2012; Cooper & Ginzburg 2015; Cooper 2020). TTR is a model–theoretic descendant of the by and large proof theoretic Martin-Löf Type Theory (Ranta 1994; Betarte & Tasistro 1998) and of situation semantics (Barwise & Perry 1983; Cooper & Poesio 1994; Seligman & Moss 1997; Ginzburg & Sag 2000). TTR enables one to develop a semantic ontology, including entities such as events, propositions, and questions, whence types characterizing questions and propositions, in (21). As we will see shortly, with the same means TTR enables the construction of a grammatical ontology consisting of utterance types and tokens and of an interactional domain in which agents utilize utterances to talk about the semantic universe. For current purposes, the key notions of TTR are the notion of a judgement and the notion of a record.

• The typing judgement: a: T classifying an object a as being of type T. Examples are given in (22). (22a,b) involve basic “atomic” types IND (ividual) and TIME. In (22c) run (arg1_IND= b, arg2_TIME = t) is a p(redicate)–type, that arises by assigning the entities b,t, respectively to the argument roles of run; arg1_IND requires its fillers to be of type IND, whereas arg2_TIME requires its fillers to be of type TIME; we will usually notate such types as (22d). Ranta (1994) proposed that elements such as s in (22c) be viewed as events or situations.

  (22)	a.	b : IND
  	b.	t : TIME
  	c.	s : run(arg1IND= b, arg2TIME = t)
  	d.	run(b,t)

• Records: A record is a set of fields assigning entities to labels of the form (23a), partially ordered by a notion of dependence between the fields—dependent fields must follow fields on which their values depend. A concrete instance is exemplified in (23b). This is a record with four fields x, e-time, e-loc, and c_{temp–at–in} to which are assigned respectively a number, a time, a location, and a situation sit1; the example is further discussed in (25). Records are used here to model events and states, including utterances, and dialogue gameboards.

  (23)	a.
  	b.

• Record Types: a record type is a record where each field represents a judgement rather than an assignment, as in (24a). The basic relationship between records and record types is that a record r is of type RT if each value in r assigned to a given label l_i satisfies the typing constraints imposed by RT on l_i. More precisely, as in (24b):

  (24)	a.
  	b.	The record
  	c.	Type inclusion:24 T1 ⊑ T2 iff for all assignments to basic types it is the case that if a:T1 then a:T2

  To exemplify this, (25a) is a possible type for (23b), assuming the conditions in (25b) hold. Record types are used to model utterance types (often referred to in formal grammar as signs) and to express rules of conversational interaction.

  (25)	a.
  	b.	-28 : Ind; 2:00AM, Feb 17, 2019 : Time; Nome : Loc; sit1 : temp_at_in (2:00AM, Feb 17, 2019, Nome, -28)

4.1.2 Conversational rules

Contextual reasoning will be important here in several ways. First, we characterize dialogue regularities (e.g., A’s assertion p gives rise to the possibility that B accepts p or alternatively that B initiates discussion of the question p?) in terms of conversational rules, mappings between two cognitive states the precond(ition)s and the effects. We will eventually suggest that laughter inter alia serves to signal that a more marked conversational option has been selected. Conversational rules can come in two flavours, rules that each interlocutor applies in the same way to their cognitive state (participant neutral) and rules that are specified only for particular interlocutors (participant sensitive). The latter kind of specification is, in principle, more general and is particularly important for an algorithmic perspective involving generation see e.g., (Larsson 2002; Cooper 2020). Most of the conversational rules we will specify will be participant neutral, as exemplfied in the rules given in (26):

(26)	a.	Ask QUD-incrementation: given a question q and ASK (A,B,q) being the LatestMove, one can update QUD with q as MaxQUD.
	b.	QSPEC: this rule characterizes the contextual background of reactive queries and assertions—if q is MaxQUD, then subsequent to this either conversational participant may make a move constrained to be q-specific (i.e., either About or Influencing q).
	c.	Assert QUD-incrementation: a straightforward analogue for assertion of (26a): given a proposition p and ASSERT (A,B,p) being the LatestMove, one can update QUD with p? as MaxQUD.
	d.	Accept move: specifies that the background for an acceptance move by B is an assertion by A and the effect is to modify LatestMove.
	e.	Fact Update/ QUD Downdate: given an acceptance of p by B, p can be unioned into FACTS, whereas QUD is modified by the function NonResolve. NonResolve is a function that maps a partially ordered set of questions poset(q) and a set of propositions P to a partially ordered set of questions poset’(q) which is identical to poset(q) modulo those questions in poset(q) resolved by members of P.

We exemplify a participant sensitive rule that relates to one of the most basic communicative interactions from infancy, namely visual attention directing, where A directs B to an object o (Lücking 2018). This is a visual situation update rule, analogous to the QUD and FACTS update rules above. The sole difference is that in this case B needs to modify her visual situation so that it includes o as the visual focus, whereas A must already have updated his visual situation to effect such an act. The notation we use for such rules is exemplified in (27a), where the rule applies to the dialogue gameboard of current addressee, with the obvious change in the case where it applies to the current speaker. (27b) provides the specification for visual situation update rule:

(27)	a.
	b.	Visual situation update:

We exemplify how certain of these rules work in (28), which involves discussion and disagreement at the illocutionary level. A poses a query, which via Ask QUD-incrementation updates Moves and via QSPEC licences B’s assertion, which in turn updates Moves via Assertion QUD-incrementation. A rejects B’s assertion, and then offers her own proposal, which B accepts. This licences acceptance, incrementation of FACTS and downdating of QUD via Accept and Fact update/QUD downdate, respectively:

(28)	a.	A(1): Who’s a good candidate?
		B(2): Petra.
		A(3): (3a) No, (3b) Pauline is.
		B(4): OK.
	b.	Utt.	DGB Update (Conditions)	Rule
		initial	MOVES = ⟨⟩ QUD = ⟨⟩ FACTS = cg1
		1	LatestMove := Ask(A,B,q0) QUD : = ⟨q0⟩	Ask QUD-incrementation
		2	LatestMove := Assert(B,A,p1) (About (p1,q0)) QUD : = ⟨p1?,q0⟩	QSPEC   Assert QUD-incrementation
		3a	LatestMove := Assert(A,B, ¬ p1)	QSPEC
			(About (¬p1,p1?))
			QUD : = ⟨¬p1?, p1?,q0⟩	Assert QUD-incrementation
		3b	LatestMove := Assert(A,B, p2)	QSPEC
			(About (¬p2,q0))
			QUD : = ⟨p2?, ¬ p1?, p1?,q0⟩	Assert QUD-incrementation
		4	LatestMove := Accept(A,B, p2)	Accept
			QUD := ⟨q0⟩	Fact update/QUD downdate
			FACTS := cg1 ∪ {p2} ∪ {¬p1}

4.1.3 Topoi and enthymemes

Conversational reasoning is important here in part because one of the relata of incongruity is in some sense an inference rule that represents “congruity” (what is expected). To capture this we use the Aristotelian notions of topos and enthymeme introduced into TTR in work by Breitholz and Cooper (Breitholtz & Cooper 2011; Breitholtz 2014). Topoi represent general inferential patterns (e.g., given two routes choose the shortest one). Enthymemes are the actual arguments conveyed in dialogue or other discourse which are drawing on topoi. In other words, they are applications of topoi in particular cases, e.g., given that the route via Walnut street is shorter than the route via Alma, choose Walnut street. We adopt the formalization of Breitholtz & Cooper (2011); Breitholtz (2014) in which topoi and enthymemes are functions from records (the context) to record types (the conclusion). (29) is a simplified illustration of the route choice topos, discussed in detail in (Breitholtz 2014).

(29)	a.
	b.

The basic relationship between enthymemes and topoi on this view is given in (30):

(30)	a.	An enthymeme E = λe : D1 . R1(e) belongs to a topos τ = λe : D.R(e) if:
	b.	D1 ⊑ D(the enthymeme’s domain is a subtype of the topos’s domain), and
	c.	for any, e : D1, E(e) ⊑ τ(e) (for any element e of the enthymeme’s domain, applying the enthymeme to e is a subtype of applying the topos to e).

4.1.4 Propositions

The final logical notion we introduce is the situation semantics notion of an Austinian proposition (Barwise & Etchemendy 1987). Deriving from Austin’s theory of truth (a true assertion involving a situation token matching a situation type), they were originally proposed to explicate assertions and relatedly beliefs. In TTR they are identified with records of the form (31a) whose truth conditions are defined in (31b):

(31)	a.
	b.	A proposition is true iff s0 : ST0

Subsequently, such propositions have been used in modelling utterance processing (Ginzburg 2012). Ginzburg (2012) proposes that dialogue interaction is, to a large extent, structured by a series of branching points where an utterance is either grounded (Clark 1996) or gives rise to clarification interaction or repair.²⁵ Ginzburg (2012) shows that the specific conditions for grounding and possibilities for repair of an utterance u can be read off the locutionary proposition defined by u and a grammatical type T_u, intuitively the sign (in the Saussaurean sense) associated with u. That the locutionary proposition involves the entire sign and not merely its semantic components is motivated, in part, by the fact that this enables the locutionary proposition to characterize the forms that are possible means to ground or request clarification about u and these exhibit significant syntactic and phonological parallelism with u (Ginzburg & Cooper 2004), as exemplified in (32b,c). (32d) exemplifies lexical entries we will posit below for laughter and its ilk. Here it is a somewhat simplified lexical entry for the particle mmh used to by B to acknowledge understanding of a prior utterance by A. It has fields for phonological and syntactic types, as well as for the contextual parameters of the utterance (DGB-PARAMS) needed to resolve the content of an utterance of mmh on a given use. In this case the contextual parameters are an utterance token and the conversational participants:

(32)	a.	A locutionary proposition is true iff u0 : Tu0, in other words iff the sign fully classifies the utterance; otherwise, repair interaction ensues.
	b.	(i) A: Do you fear him? B: Fear? (=What do you mean by ‘fear’ or Are you asking if I fear him) / #Afraid?(ii) A: Are you afraid of him? B: Afraid? (=What do you mean by “afraid”? or Are you asking if I am afraid of him) / #Fear?
	c.	A: She is an advocate. B: What do you mean an advocate/#an attorney?
	d.

4.2 A minimal theory of laughter: laughables and incongruity

We propose to view laughables as Austinian propositions, comprising the laughable event and its description, a record type. (33) offers some examples: in (33a) the laughable is simply a perceived event of a man slipping over a banana peel; in (33b) the laughable is the utterance by A ‘Bill is absurd’, whereas in (33c) the laughable is the utterance “friend”:²⁶

(33)	a.
	b.
	c.

Like Raskin we think that incongruity needs to be explicated in terms of a clash. For Raskin the clash is between two objects (scripts) at the level of types. However, given that incongruity in conversation must apply to laughables (real world events), we refine slightly a proposal by (Ginzburg et al. 2015) to yield a view of incongruity as a clash between (an enthymeme triggered by) the laughable and a topos that represents ‘congruity’, i.e., the much more probable course of action.²⁷ That is, the laughable l satisfies the domain type of an enthymeme, but there is a clash between the range of this enthymeme and that of a topos which would license an enthymeme more expected in the circumstances. In (34), p is a proposition comprised of l, the laughable event, and L a type that classifies l, E is the triggered enthymeme, and τ is the clashing topos—E’s domain is a subtype of τ, but its range (L₁) is incompatible with τ’s range:

(34)

We exemplify this definition with several distinct types of effects we have seen before:

1. Irony marking of an assertion:

   (35)	Lecturer: … And then of course you’ve got Ronald Reagan … and <laughter/> history ended with Ronald Reagan. (BNC, JSM)

   This relies on the enthymeme “If A says that history ended with Ronald Reagan, then A means that in fact it did not.” This clashes with the sincerity topos “If A says p, then A means p”. Hence it conveys the content in (36):

   (36)	Content: Incongruous (laughable:Lecturer’s utterance that history ended with Ronald Reagan, topos: If A says p, A intends p)

   How does this arise in context? We do not offer here a precise account of topos resolution, but sketch a hypothesis based on markedness.²⁸ Whenever a declarative utterance is made by A which involves a proposition p there are (inter alia) two possible understandings available: A asserts p or A intends to convey a content incompatible with p. A priori, the former is far likelier, so a laugh can be viewed as a means of signalling the choice of the much less likely possibility.

2. Query cancellation

   (37)	David Gregory; interviewing Chuck Schumer:
   	DG: (1) Is Sarah Palin the future of the Republican party?
   	CS: (2) .hh hh=W(h)well(h)heh heh heh .hhuh, I guess I shouldn’t judge and let them f(h)ight among themselves. (Romaniuk 2013)

   How does laughing enable questions to be deflected? To answer this, we need an account of how questions can be deflected. We follow the one offered in (Ginzburg 2012).²⁹ On this view given that A poses q, for B not to address q, B has to accommodate the issue ?Wish(B,q) [whether B wishes to discuss q] into QUD and provide an utterance about this issue. This can lead to q being downdated from QUD.³⁰

   In this case the deflection relies on the enthymeme if DG posed the question whether Sarah Palin is the future of the Republican party, then CS will address the issue of whether CS wishes to discuss whether Sarah Palin is the future of the Republican party. This conflicts with the topos associated with our earlier conversational rule QSPEC (if A poses q, then either A or B utter a q–specific utterance.).

   Hence CS’s laugh conveys the content in (38):

   (38)	Content: Incongruous (laughable:DG’s utterance asking if Sarah Palin is the future of the Republican party, topos: if A poses q, then A/B utter a q–specific utterance)

   As with irony marking, the laugh emerges as a signal for a marked choice: when a question is posed, addressing it is the unmarked choice.³¹

3. Scare quotation

   (39)	B: Are you interested in drama? A: well I I’m interested in it in a (pause) (laughs) ((comfortably)) re:laxed way, you know. (London Lund Corpus).32

   Here we have an interaction between laughter and self communication management.³³ Here the laughable is A’s upcoming speech event u. The pause potentially indicates that A is not entirely sure how to express the next word or phrase. A chooses a particular sub-utterance but uses the laugh to signal that this choice deviates from the standard use of this phrase. In light of semantic work such as (Ginzburg et al. 2014) and psycholinguistic work such as (Lowder & Ferreira 2018), we assume that a hesitation by A means that the immediately subsequent utterance addresses the issue of What does A mean to say after u₀?. The laugh in this case signals that A does not quite mean ‘comfortably relaxed’ as the utterance to follow “[in] a”. The default, unmarked alternative would be for this utterance to represent exactly what A meant to say after “[in] a”. This is the same mechanism as for irony marking of a statement, but applied to word/phrase choice, as a consequence of self communication management.

   Hence (39) conveys the content in (40):

   (40)	Content: Incongruous (laughable: A’s utterance ‘comfortably relaxed’, topos: if A utters u, A means μ(u), u’s conventional meaning)

   Note that different placements of a laugh within the same utterance type will result in potentially different contents being conveyed, given that which sub-utterance is the one in focus alters the issue raised and the scope of the scare quotation. Moreover, as with focus marking, the exact target of the scare quoting laughter is potentially ambiguous.

4.3 Incongruity laughter: a lexical entry

We can now offer a lexical entry for a laugh that expresses incongruity.³⁴³⁵ This lexical entry indicates about the signal whose phonological type is lphontype, on which more shortly, that given a context which supplies a laughable p and topos τ, the laugh has as content the proposition that p is incongruous relative to τ. Moreover, the laughable is incongruous to a contextually given degree δ, constrained by a relation whose other argument is the arousal encoded in the laughter’s phonetics.

(42)	incongruous laughter

FIX FIX FIX!

A number of key aspects remain underspecified in this entry for reasons we explain now:

• Form: we have mentioned above that the arousal associated with a laugh is a significant parameter. First, it distinguishes different words for laughs: a snigger represents a low arousal laugh, whereas a cackle represents a high arousal laugh; second, a high arousal laugh is inappropriate for intra-utterance scare quotation and for empathetic acknowledgement, whereas a low arousal laugh to a joke communicates a scalar implicature that the laugher does not find the joke particularly funny. Arousal is encoded by a number of parameters, including intensity, duration, and posture (Kousidis et al. 2015), but in the absence of a clear account, we will simply assume this as a scalar value readable off the phonetic representation of a laugh this value for current purposes we restrict to be H(igh arousal), M(edium arousal), and L(ow arousal). The phonological representation of a laugh is known to be subject to both large inter–subject and intra-subject variation (Urbain & Dutoit 2011), and so we leave that unanalyzed as a type lphonetype, which given what we said about arousal will have three subtypes: H, M and, L.³⁶ To the extent we need to specify laughs that have specific arousal profiles we can add restrictions on the arousal, exemplified in (43):³⁷

  (43)	a.
  	b.

• Contextual parameters Both the laughable and the topos that are the relata of the incongruity relation are specified here as contextual parameters to be resolved from information from the DGB. We have seen that laughables can originate from the content of the latest move made (examples (14), (37)), from the ongoing utterance (examples (16)), from the visual situation (example (7b)). These can be both anaphoric and cataphoric (for the latter: examples (16), (3)), so we defer attempting to formulate a precise theory of resolution to future work (see (Tian et al. 2016) for some discussion of the placement of laughter relative to speech.). A similar point applies to the resolution of the topos—as we have illustrated above, resolutions can be topoi that are conversational rules, rules potentially applicable in the current context. While this is a common case, it is not invariably so, as example (3) above shows (‘Today’s semantics class (laugh)’): in that case laughter is triggered by recollection of an event whose incongruity is not accessible in the current context, but is down to a common experience shared by the interlocutors.

4.3.1 Antiphonal laughter: sharing incongruous judgement

How to analyze the contribution of laughter in a dialogue? For now we show that merely assuming the propositional content in section 4.3, we can capture some basic patterns it manifests in dialogue pointed out by Jefferson, discussed in section 2. There is no need to postulate any special purpose conversational rules, though the framework would enable us to do so if necessary.³⁸

Armed with the lexical entry in the previous subsection, we can now consider the dialogue in (44a), for which we assume the laughable can be represented as in (44b). We view the sharing of the incongruity judgement concerning Roger’s statement, as akin to the sharing of a “normal” linguistically expressed proposition, as in (44c):

(44)	a.	Roger: you are what dey refer to in rougher circles as a chickn shit.
		Roger: hhhhehh
		Ken: heh:heh:heh
	b.
	c.	A: You are annoying. B: I am annoying / Yes!

Assume the laughter proposition has a similar force to a normal assertion. This yields a QUD update, as in step 4 below. This enables B to express the same proposition and share in incongruity classification of l:³⁹

Utt.	DGB Update (Conditions)	Rule
4	LatestMove := Assert (B,A,Incongruous (p0, e1,τ1))
5	QUD : = ⟨Incongruous(p0, τ1)?⟩ LatestMove := Accept (A,B,Incongruous (p0, e1,τ1)))	Assert QUD–incrementation Accept
	QUD := ⟨⟩	Fact update/QUD downdate
	FACTS := cg1 ∪{ Incongruous(p0, e1,τ1)}

5.1 Cognitive and computational models of emotion

We will explicate pleasure in laughter by attempting to incorporate a cognitive theory of emotion into our dialogue theory. There are a variety of theories of emotion. Although important early work on facial gestures came from researchers espousing theories postulating a small number of basic emotions see e.g., (Ekman et al. 1987), there has been little evidence to support the existence of physiological characteristics that instances of a single emotion share but that other emotions do not (Scherer & Ellgring 2007; Barrett 2017). A variety of approaches have emerged that avoid such an assumption. We mention here two influential approaches, emotion constructivism and approaches based on appraisal; both provide useful means for classifying laughables, smileables, sighables etc and the emotional episodes they give rise to.⁴² Emotion constructivism e.g., (Russell 2003; Barrett 2017) views emotional episodes as arising from a categorisation process of the triggering event in terms of previous event “exemplars” on the basis of a resemblance in terms of certain dimensions. In the account of Russell (2003) these include core affect (a two dimensional matrix of (un)pleasantness and arousal), affective quality, and the object causally involved in the event. Appraisal approaches also view emotional episodes in terms of appraising events in relation to concerns. An initial automatic appraisal takes place that does not require conscious processing. This is followed by a secondary appraisal that often includes conscious reflection and that can lead to new intentions. As an example of such an approach consider the Component Process Model (CPM) of appraisal developed by Klaus Scherer (see e.g., (Scherer 2009)). On this view, appraisal of an event and its consequences can be structured on the basis of a number of criteria or stimulus evaluation checks, each of which has certain physiological correlates:⁴³

The computational framework EMA (EMotion and Adaptation) (Gratch & Marsella 2004; Marsella et al. 2010; Traum et al. 2008) implements a view of appraisal quite close to that of the Component Process Model. There are some significant differences, nonetheless. Thus, EMA does not subscribe to the assumption that the stimulus evaluation checks are strictly ordered, and we follow EMA in this respect. EMA treats appraisal as a set of feature detectors that map features of the agent’s current view of the agent-environment relationship into appraisal variables.

5.2 Adding emotion to the DGB

In order to integrate emotion–related information into the Dialogue Gameboard, we posit an additional field we will dub Mood. Mood represents the publicly accessible emotional aspect of an agent that arises by publicly visible/audible actions (such as non-verbal social signals). This can but need not diverge from the private emotional state—as with insincere illocutionary acts, one manifestation of a “fake” laugh/smile is a laugh/smile that does not reflect genuine pleasure;⁴⁵ it can also involve the lack of a genuine belief that the laughable is incongruous.

We view Mood as a complex entity built from categorisation dimensions/appraisal values. More specifically, we assume that it is a weighted sum of these values.⁴⁶ We treat each appraisal as being a record of the type Appraisal, given in (50). We posit this type to be, following Russell (2003), the Component Process Model and EMA, a record type with three fields pleasant, responsible, power, each field corresponding to a Russellian dimension or an answer to a stimulus evaluation check. We restrict attention here to these dimensions, which seem needed for the purposes of this paper. Pleasantness is specified via a scalar predicate Pleasant which can be positively valenced (value tracked by the label pve) or negatively valenced (value tracked by the label nve) or both (when both pve and nve have non-zero values.); the latter case corresponds to the case of mixed emotions (Minsky 2007), which in practice are relatively common (Oatley & Duncan 1994). Specifically for our current interest, this can be exemplified on the one hand by the case of a joke told in a funeral and, conversely, by the case of a sigh occurring when one is generally in a rather elated mood. Power is specified in terms of a scalar predicate Powerful whose lower bound (zero) arises when the agent perceives total lack of control vis à vis the appraised event.

(50)	a
	b.

As mentioned above, this leads to the postulation of a modified type for the dialogue gameboard, with a field for Mood:

(51)

We do not offer here a general theory of how emotion categorisations/appraisals arise⁴⁷—we will restrict attention below to the effect laughter and other social signals can have on the parameters of appraisal. Nonetheless, we make a number of assumptions concerning how Mood gets updated and introduce one abbreviation for two recurring operations. We assume that the most recent value of numerical components of Mood should be given a stronger weight than its current value, which itself represents some combination of the earlier ones. Therefore we postulate a weighting between the contribution of the new appraisal and the current value of a field of Mood. If this field also has a non-zero negative value, the weighting will attenuate it, otherwise it will have no effect.

Thus, an update rule that increments the positive pleasantness recorded in Mood to an extent given by the weight ɛ is given in (52a), whereas the converse operation of incrementing the negative pleasantness is given in (52b). We exemplify the application of positive pleasantness update for δ = 12, ɛ = .25 in (52c):

(52)	a.	PositivePleasantnessIncr (δ,ɛ)=def
	b.	NegativePleasantnessIncr(δ,ɛ) =def
	c.

We will assume for simplicity that the weighting ɛ has a fixed value for any given agent and will not—as we formally should—specify it each time as a parameter of an agent’s private cognitive state.

5.3 Pleasantness laughter

We can now formulate a lexical entry for pleasantness laughter, as in (53a): the content we posit is that the laughable is pleasant for the speaker to a contextually given degree δ. The effect of such laughter on the speaker is captured in terms of an update rule that increments the (positive) pleasantness recorded in Mood to an extent given by the weight ɛ, as described earlier. Note that this conversational rule is what we dubbed above participant sensitive: it can apply only to someone laughing; this ensures that in order to boost one’s Mood.pleasantness (which represents one’s public emotional state) one is actually required to laugh (or engage in other action, as we will see below.).⁴⁸⁴⁹⁵⁰

(53)	a.
	b.

From pleasantness, we can derive three functions of laughter: affiliation, empathetic acknowledgement, and (with additional assumptions) superiority. We take these in turn.

5.4 Incongruity and pleasure

What of the pleasantness that occurs with much incongruous laughter? This can be captured by a simple refinement of (53)—adding as a disjunct that the trigger for the Mood update can also be an incongruity content, as in (42). This means that incongruous laughter will communicate (i) that the laughable p_lis incongruous relative to some topos τ and (ii) that the speaker is (momentarily) pleased.

This seems sufficient for cases like irony signalling (35) and scare quoting (39). In other cases, however, arguably the pleasure is actually incorporated in the incongruity. Thus, (54b) which denies the pleasure but not the incongruity seems a more natural reaction than (54c), which denies the latter:

One way of capturing this additional inference is by positing a topos that finding an entity incongruous involves that fact itself being pleasant:

6.1 Smiling

The smile is characterized by the upward turn of the corners of the lips, which is produced by the contraction of the zygomaticus major muscle (Ekman et al. 1987). Ever since Darwin, the relationship between laughter and smiling has been a key question. Darwin’s view was, essentially, that smiling is a low arousal version of laughter.⁵¹ One consideration in favour of this is the scalar relationship between laughter and smiling discussed earlier—when a laugh is expected, production of a smile implicates that laughter cannot be provided because the stimulus is insufficiently funny/pleasant. More generally, there is a large though not complete overlap in the effects that laughter and smiling can bring about: in addition to pleasant incongruity, these include affiliation and the expression of superiority or dominance over one’s interlocutor (Niedenthal et al. 2010)—this latter can be viewed a “low arousal” version of the effect of mockery discussed above for laughter.

As with laughter, a definitive taxonomy of uses is still a matter of controversy. An extreme case is Ekman (2009), who identifies 18 types of smiles and suggests that there might be as many as 50 in all. A more minimalist view is that of Niedenthal et al. (2010), who suggest that many smiles are simply readouts of positive internal states such as happiness; this holds also for the “play-face” in primates, such as chimpanzees (for the latter, see discussion in section 7.) Such a meaning for a smile is analogous to our pleasantness meaning for laughter. However, Niedenthal et al. (2010) distinguish this from two other types—affiliative and dominance smiles. As we pointed out above for laughter, affiliation and mockery can actually be deduced from pleasantness uses. This does not rule out positing these as basic meanings for smiles, but leaves open the possibility that these, like many other effects of smiling noted inter alia by Ekman, are actually pragmatic inferences.

These and the corresponding scalar implicatures could be captured by positing lexical entries for smiles that correspond to the ones we have posited for laughter; given our formulation, the pleasantness update rule will apply without further stipulation. There are two clear differences, however, signalled in (56) in bold. The first concerns form, which needs to be specified in terms of various parameters pertaining to facial shape/activity (the Zygomaticus major and orbicularus orbi muscles); as with laughter we abstract away from this using the label smileshape; the second difference concerns arousal—the range of arousal needs to be bounded so as to, roughly, be less than the arousal achieved by high arousal laughter; in other words, we assume that smiling is restricted to low and medium arousal.

(56)	a.
	b.

6.2 Sighing

Sighs—breaths with a volume at least twice as large as the mean volume during a surrounding representative time interval (Vlemincx et al. 2017)— are ambiguous signals in conversation. They can be produced for purely physiological reasons to allow the respiratory system to return to a homeostatic state with respect to blood gas levels, but they can also be produced with communicative intent.

Teigen (2008) suggests that the prototypical sigh conveys discrepancy (something is wrong) accompanied by acceptance (there is nothing to be done). Such a meaning is straightforward to describe in the current set up, as we will see shortly. However, the most extensive corpus study of sighs we are aware of, the study by (Cash & Ginzburg 2019) of more than 100 sighs in the BNC, points to an additional class of sighs, positively valenced sighs, that are much more frequent than commonly believed. In the data of (Cash & Ginzburg 2019) they occur 46% of the time, and moreover seem to be partly correlated with whether the sigh is same/cross-turn with respect to the sighable: 29% of sighs in answers are positive, contrasting with 58% in queries. Sighs are typically low arousal (as noted also by Teigen and Vlemincx et al)—only 11% of the sample of (Cash & Ginzburg 2019) are high arousal. However, of the high arousal sighs a very large proportion—82%—are positively valenced.

A first approximation this suggests is to distinguish non–high arousal sighs from high arousal ones, associating the former with negative pleasantness and a sense of powerlessness, the latter with positive pleasantness and possibly relief. Respective lexical entries are (57a,c), whereas (57b) is an update rule associated with (57a), incrementing the negative pleasantness and setting the power arousal level to zero. The force of a positive sigh (57c) is postulated to be simply Pleasant, which makes it trigger the same update rule as for smiling and laughter, namely (57b).

(57)	a.
	b.
	c.

6.3 Eye rolling

By comparison with other non-verbal signals mentioned here, eye rolling has not been much studied and we are not aware of a major corpus study, though see (Goodwin & Alim 2010) for some discussion.

We mention here three attested examples:

This suggest the following basic description: the eye roller views the eye-rollable as ridiculous; this effects a decrease in the eye roller’s mood for which some person or organization R is responsible. Here the notion of ridiculous is akin to, but distinct from incongruity, and thereby relative to a topos. Hence the following lexical entry:

(59)	a.
	b.

6.4 Frowning

Wierzbicka (2000) attempts to find a general meaning for drawing one’s eyebrows together and plumps for an earlier intuition of Scherer (1992), who speaks of the perception of some type of discrepancy between one’s needs or goals and one’s actual circumstances. Wierzbicka tries to explicate this as I want to do something, but I know I’m not doing it now. Kaukomaa et al. (2014) suggest with respect to turn–opening frowns that they are related to the emergence of a problem, relate to negative assessment and lack of affiliation across participants. The constructed example (60a) indicates the possibility of a frown as part of clarification interaction, whereas (60b) exemplifies a frown in self-repair:

We refine these proposals by viewing discrepancy and problem emergence as involving the frownable giving rise to a question, which can indeed be explicitly posed, as in (60); this is associated with a Mood update in which power is non-zero and pleasantness is decreased.

(61)	a.
	b.

7.1 Main points

In this paper we have refuted a view that has dominated research on formal grammar since the influential typology of Trager (1961) that views laughter as paralinguistic, hence lacking meaning akin to what words and phrases possess and not contributing to the compositional construction of meaning. We have shown that, in fact, laughter has propositional content: we show that it involves reference to external real world events, quite analogously to event anaphors and in contradistinction to bodily signals like sneezing and hiccuping, a view that can be traced back to Plessner (1970). We suggest that laughter has stand alone meanings and that it participates in semantic and pragmatic processes like repair, implicature, and irony.

We have shown how to develop a formal semantic and pragmatic account of laughter embedded in a general theory of conversational interaction. This novel view of laughter enables us to capture in a unified and rigorous manner previous insights concerning laughter, including those from Conversation Analysis and those emanating from linguists working on humour. From CA, in particular Gail Jefferson, we take the assumption that laughter can be and often is an intentionally produced signal included in the range of responses an utterance generates.

Echoing a parable of Milan Kundera’s, our account, couched within the framework KoS, posits two distinct meanings for laughter, one based on incongruity exhibited by the laughable, the other based on pleasure the laughable causes to the laugher. Building on Raskin’s work in humour theory, we explicate incongruity as a notion that relates a contextually salient entity l with a defeasible rule (a topos) in case there exists a contextually salient characterization of l that is incompatible with τ. With this view of incongruity, which is closely related to the TTR notion of negation (Cooper & Ginzburg 2015), we show how a wide range of functions of laughter can be deduced. These include signalling irony, dismissing the seriousness of an assertion, question, or suggestion, and scare quotation. This view of laughter also captures straightforwardly the potential for misunderstanding/repair associated with laughter, given the need to resolve in context both laughable and topos. It also captures the “hyper-intensional” nature of humor that does not e.g., survive translation or synonymous paraphrase, in exactly the same way as the hyper-intensionality of repair processes. This requires adopting independently motivated assumptions about the nature of the input to semantics/pragmatics, namely rich, multi-dimensional representations of utterances. This assumption is not generally assumed in most other semantic/pragmatic frameworks.

In order to explicate the requisite notion of pleasantness whose importance for laughter has been emphasized by Morreall, we build on existing work in cognitive psychology and artificial intelligence. We extend the representations of KoS so that they incorporate the outputs of appraisal reasoning/emotion construction. This allows us both to capture the incrementation of a laugher’s public level of pleasure, as well as deduce functions of laughter such as affiliation, empathetic feedback, and superiority. Using this extension of KoS, we have sketched in broad terms how our approach extends to other non-verbal social signals such as smiling, sighing, and frowning for which the arguments we offer about the content-fulness of laughter extend straightforwardly.

Given that these signals are considered intentionally produced alternatives to each other (and to verbal signals), we can also capture their potential to give rise to conversational implicatures.

We have mentioned that the current approach can deduce various potential functions that laughter has. In fact, our view of laughter meaning means that we do not assume the existence of a limited set of discrete functions that are associated with laughter; in principle, distinct laughter episodes can give rise to an unbounded number of distinct inferences, as assumed, in both constructionist and appraisal views of emotional episodes. We have not addressed attested functions here, but Mazzocconi et al. (2020) have developed a taxonomy for the functions of laughter building on the two basic meanings here. They utilize a taxonomy with four subtypes: pleasant incongruity (cases in which a clash between the laughable and certain background information is perceived as witty, rewarding and/or somehow pleasant), social incongruity (situations involving a clash between social norms and/or comfort and the laughable), pragmatic incongruity (cases when there is a clash between what is said and what is intended), and pleasantness (cases lacking incongruity where a sense of pleasantness/affiliation is either felt or displayed towards the interlocutor.). A corpus study using this taxonomy based on examination of 1072 laughs (289 instances in English (drawn from the British National Corpus (BNC)), 562 instances in French, and 221 in Chinese (the latter two drawn from the multilingual DUEL corpus (Hough et al. 2016)) achieves very high coverage (> 99%).

7.2 Implications for the grammar

At this point, a natural question to ask is: have we shown that laughter (and its close relatives smiling, frowning, sighing etc) belong to the grammar? We do not assume that this is a question that can be answered straightforwardly, not least because of potential disagreements about what constitutes the grammar. Over the last few years, several works have appeared detailing the view that grammars should be viewed as systems that classify an utterance as it occurs in conversation see e.g., (Ginzburg 2012; Ginzburg & Poesio 2016; Kempson et al. 2016; Cooper 2020). Thus, Ginzburg & Poesio (2016). argue that phenomena such as disfluencies, non sentential utterances, quotation, and co–speech gestures are as rule-governed as binding, control, and dislocation—traditional sentence-level phenomena captured in formal grammars. Given the existence of formal accounts for all these conversational phenomena within frameworks such as KoS (Ginzburg 2012), PTT (Poesio & Rieser 2010), SDRT (Asher & Lascarides 2003), Dynamic Syntax (Kempson et al. 2016), and other related frameworks, this suggests the need for a new view wherein grammar is a means for directly characterizing speech events, abolishing the performance/competence distinction (though recasting this in a way that allows maintaining a distinction between the task of describing the linguistic phenomena from the task of describing how they get processed.).

In considering whether laughter and its non-verbal cousins merit inclusion in the grammar in the above sense, one would presumably consider criteria such as the following, all of which all the above phenomena such as disfluencies, non-sentential utterances, quotation, and co-speech gestures pass:⁵²

With respect to placement, one very influential view due to Provine (1993) has been that laughter always follows the laughable and only occurs between spoken utterances; Vettin & Todt (2004) offer a more nuanced account, but assume adjacency between laughter and laughable and exclude laughs that occur in the middle of or overlap with an utterance. Mazzocconi et al. (2020) demonstrate (for French and Chinese in the DUEL corpus and English in the BNC) that only 72% of laughs immediately follow their referents. Instead, the laugh can occur before, during or after the laughable with wide time ranges. Strikingly, they demonstrate statistically significant differences between the three languages regarding the positioning of laughter in relation to its laughable. An earlier result though on a smaller scale is due to Gavioli (1995), who found that laughter used to remedy social incongruity—where a bookseller had to provide a dispreferred response to a client—occurred by and large turn initially for English (n=16, 75%) whereas turn finally in Italian (n=20,90%). To this one can add work that describes carefully the positioning of a variety of non-verbal signals such as sighs (Hoey 2014) and sniffs (Mondada 2020). At the same time, it is clear that a detailed formal account of placement and form/meaning/function correlations is still very open for laughter and other conversational facial gestures, in particular on the extent of cross-linguistic variation in this respect and its basis.⁵³

The neuroscientific study of laughter is very active: including locating multiple sites in the brain of activation for laughter and smiling (Szameitat et al. 2010), and the demonstration that there is different brain activity underlying subjects’ ability to distinguish “genuine”/spontaneous’ from “fake”/“strategic” laughter (Lavan & McGettigan 2016). However, for the moment, there has been little work on laughter processing in spontaneous conversation which would indicate how this processing relates to other elements of conversation viewed as grammar internal.

The fact that apes possess laughter–like calls (Ross et al. 2009) raises various questions about the cognitive and semantic relationship between such calls and human laughter (and smiling). These include: (i) Does chimpanzee laugh face anticipate illocutionary act seriousness cancellation laughter with humans, as already suggested by the highly prescient Bateson (1955)? More generally, (ii) do apes have incongruity laughter? This is suggested, for instance, by observation of laughter during play (tickling and rough-and-tumble) (Van Hooff & Preuschoft 2003; Davila-Ross et al. 2011), and by anecdotal data such as https://www.youtube.com/watch?v=OLrYzY3jVPY, where an orangutan apparently laughs at a magic trick. Given the semantic complexity of incongruity, as discussed here extensively, this would potentially require rethinking existing views of non-human cognitive capacities, whether in the wild or for non-humans living in close contact with humans (Griebel et al. 2016).