1This essay is aimed principally at sketching an approach to cognitive phonology. It starts from a question: what would a truly cognitive phonology look like? Nathan (2008) has written an important characterisation of cognitive phonology. His article placed emphasis on the issue of the storage and representation of phonological units. He considered Bybee's view that distinct phonetic realisations of a phonological unit are stored (e.g. Bybee 2001). This view is consistent with, for example, Langacker's position on grammatical units and the possibility of having both a schema and specific instantiations of that schema stored in memory. However, Nathan argued for an embodied view in which the distinct phonetic realisations of a phonemic category are driven by natural processes of articulation. He argued that specific phonetic realisations are, in many cases, driven by and direct consequences of the process of articulation constrained by the context of the phonetic environment in which the sound appears. He pointed to examples such as the realisation of allophonic variants in speech errors, which are invariably appropriate to the phonetic context.
2In making this argument, he made links with Natural Phonology. Indeed, phonetics and phonology are, in any case, already embedded in psychology, the psychology of hearing and of motor control. However, there are three further features that we could expect a cognitive phonology to emphasise distinctively:
Continuity of the linguistic and cognitive systems
Functional aspects of phonetic and phonological choices
Partial constructional equivalence
3The first is implicit in the traditional links between phonology and motor control or hearing. However, it can be extended by emphasising the continuity of linguistic, paralinguistic, and extralinguistic signalling. The second demands an account of phonological processes in terms of the communicative and social functions they serve. A simple example of a social function would be code switching to express a relationship of identification with a group. An example of a communicative function would be the relationship between information structure and different patterns of intonation. We can seek to apply cognitive principles, such as Gestalt principles, to understanding these patterns. The third alludes to the general principle that a given “act” can be realised in different ways. A given gesture need not be realised by precisely the same combination of movements on any two occasions. For example, if I wave to you while lying on my back, I will use a different combination of muscles and tendons than if I wave to you while sitting up. But it will be the same gesture.
4In this paper, I will use the example of speech laughter to illustrate some of these points. Speech laughter is the simultaneous expression of laughter and articulated speech. The speaker utters some words and laughs at the same time. The laughter thus becomes a suprasegmental acoustic signal, akin to prosody. Oversimplifying a little, we can say that it tends to be used when there is some disjunction between what is said by the words and what the speaker means or is prepared to commit to. For example, it may be used when the speaker is not confident that what they are saying will be accepted by the addressee.
5Only a few studies have examined speech laughter in spoken interaction, as far as I am aware. Trouvain (2001) analysed examples from the Kiel Corpus of spoken German. Kohler (2007, 2008), using examples from a small sample from the same corpus of spoken German, identified a phonetic mechanism for realising speech laughter. He observed that the vowel of the syllable being articulated was 'divided' by an energy contour. That is, the amplitude or intensity of the sound would rise, fall, and rise again. Kohler (2007) posed the question of whether the realisation of laughter, including speech laughter, would vary across languages, or might be led by individual differences between speakers. Nwokah, Hsu, Davies & Vogel (1999) studied mother-infant speech, and observed the use of similar mechanisms, together with a degree of individual variation across mothers. In the present study, we make a comparison with a sample drawn from a small number of varieties of spoken Scots. In section 3.1, a series of examples are used to illustrate the range of variation in the form of speech laughter found in Scots. In section 3.2, the communicative context of the same examples is analysed in detail. An important issue is the identification of any symbolic link between what I have termed the functional domain, which is semantic-pragmatic in nature, and the phonetic domain. A step towards this will be taken in the Discussion, using the concept of an image schema.
6The study found that speakers realised speech laughter using a range of forms, including amplitude or pitch fluctuation, and used it to express a number of functions including management of self-presentation and the expression of a subjective attitude, or modality, towards the utterance. There was some evidence that form was related to function. For example, expressions of humour using speech laughter had a form more similar to plain laughter, but concession tended to be acoustically reduced.
7The data is drawn from the Scottish Corpus (Anderson, Beavan & Kay 2007). The corpus contains a large sample of written and spoken Scots. For this study, a selection of spontaneous spoken conversations was used. In these conversations, people who knew each other very well, typically close family members, had a conversation in the presence of a researcher from the English Department at the University of Glasgow. The researcher can be heard to prompt and join in the conversation. Indeed, in several instances the researcher was also a relative of the participants. Participants were aware that the research project was gathering samples of Scots, and occasionally there are references in the conversation that make that clear. For example, participants might raise the topic of unusual dialect words from their own area. More typically, they share anecdotes about events and people they have known.
8From our point of view, it is very useful that neither the participants nor the researchers were focused on humour or laughter. For them, the research was not about laughter. This helps to provide reassurance that any laughter was natural and not influenced by demand effects. The conversations represented four varieties of Scots, with speakers from Aberdeenshire, Ayrshire, Edinburgh and Stirling. Samples of speech laughter were identified, and were extracted from the sound files in the corpus. Analysis was carried out using Praat software (Boersma & Weenink 2010). Spectrograms were constructed, together with amplitude and fundamental frequency (F0, 'pitch') contours.
9In this section, I focus first on the phonetic analysis of samples of speech laughter drawn from the Scots corpus. Then I briefly examine the relationship between form and function in speech laughter.
10Figures 1 and 2 show examples of plain laughter. In common with the other spectrograms shown, time unfolds along the x-axis, and the y-axis represents frequency (0 – 5kHz). The sample in Figure 1 lasts for approximately 3s. The other spectrograms shown in the remainder of the paper are all approximately 1s long. Amplitude at a given frequency is indicated by the darkness of the plot. These spectrograms illustrate the regular beat of intense bursts of sound that is characteristic of laughter. There is relatively little evidence of sonorance or formant structure in Figure 1. Figure 2 illustrates the very fine and close timing of laughter in conversational interaction (Kohler 2008). In Figure 2, two people are laughing, with the beats of their laughter alternating in rhythm.
Figure 1. Spectrogram of plain laughter
Figure 2. Spectrogram of two speakers laughing, showing coordinated timing (the first beat from each is marked with an arrow; they continue to alternate thereafter)
Figure 3. Spectrogram of the word 'aye' with speech laughter
11As Kohler (2007, 2008) reported for German, a common realisation of speech laughter was for amplitude to 'divide' the vowel, though sonorant would be a better term than vowel, perhaps. Figure 3 illustrates this for an utterance of the word aye (“yes”). It is easily seen from the spectrogram that the speech laughter is realised without the characteristic rhythmic phonation of laughter. The narrow alternating deep dark and light vertical bands are not present. This is quite typical. Speech laughter is not generally a plain laugh made along with speech, it is speech with an overlay that creates an impression of laughter. The red contour indicates the rise and fall of amplitude, "dividing" the vowel and glide. The green line indicates pitch, which in this example is relatively steady. Figure 4 compares two more utterances of aye, with (on the left) and without speech laughter. In this case, the speech laughter is realised with more of the rhythmic beat typical of ordinary laughter. The red curve again shows amplitude, which rises and falls rhythmically in Figure 4(a).
Figure 4. Spectrograms of the word 'aye' with speech laughter (left), and without
Figure 5. Spectrograms of the compound 'double-decker' with (left) and without speech laughter
12As well as amplitude, speech laughter was also realised using pitch and vowel-lengthening. Figure 5 compares utterances of double-decker. Figure 5(a) shows amplitude (red line) and pitch (green line) being used in combination to effect speech laughter. Figure 6 illustrates the combination of pitch dividing the vowel and vowel-lengthening to produce speech laughter: the cla:ppers. Lengthening the vowel extends the carrier of the speech laughter, allowing it to be extended in time and creating a larger acoustic canvas on which to paint the pitch division. Note that in this case, amplitude makes only a very slight contribution to “dividing” the vowel.
Figure 6. Spectrogram illustrating the use of vowel-lengthening together with pitch modulation (between the vertical blue lines) to create speech laughter
13Another, quite different, phonetic strategy observed was to reduce sonorance. Figures 7 and 8 illustrate the use of aspiration and devoicing, respectively. This kind of speech laughter is far removed acoustically from plain laughter, yet still gives the impression of laughter. It is different from plain laughter in that it lacks both the high amplitude open sound and the rhythmic beat pattern. Trouvain (2001) analysed plain laughter as a CV, (consonant vowel) sequence, with aspiration or a glottal stop occupying C, and in which the duration of the aspiration is greater than the normal duration of the consonant. Although this analogy to CV structure is controversial (Kohler, 2007), it nevertheless illuminates the reduction strategy, which can be seen as playing on an acoustic analogy to the “C” phase of plain laughter. That is, the part Trouvain labelled C can be seen as the element that remains, for example as aspiration or devoicing, when speech laughter is reduced in amplitude or force.
14In Example 7 “that’s true”, the aspiration is, however, not carried through to the vowel from the initial voiceless consonant, but begins after the vowel formant structure is first established. In Example 8, the spectrogram shows that there is no voicing during the vowel (or indeed the nasal that follows). There is slightly extended aspiration following release of the voiceless stop, about 180ms, but then the ‘shape’ of the diphthong is audible, as there would be in a whisper.
Figure 7. Spectrogram showing aspiration following the vowel of "true"
Figure 8. Spectrogram showing devoicing of the noun "time", which begins at the vertical line
15All the realisations of speech laughter described so far are restricted to small stretches of speech. Trouvain (2001) found that speech laughter was typically restricted to no more than a two syllable sequence, occasionally three, and Kohler's examples were consistent with that. However, in this corpus there were examples stretching over much longer spans. Characteristic of these is that they were uttered on a single breath (exhalation). In example 1, the second speech laugh, beginning at terrified, stretches across 21 syllables (worm is bisyllabic in this variety of Scots). In example 2, the speech laughter extends over 11 syllables. The sections involving speech laughter are italicised.
16Different explanations are possible for the occurrence of longer periods of speech laughter in this corpus than found in previous studies. An important difference is that the Kiel corpus is confined to a comparatively formal work situation which may lead speakers to contain their expressiveness. Such restraint might also occur as a function of cultural differences between the speech communities sampled, professionals from northern Germany versus older people from Scotland.
17Each of the examples of speech laughter occurred in a conversational context. In this section, we examine its function. Transcription of the conversations is taken directly from the original corpus transcripts, without correction. A phrasal glossary is provided in the Appendix for words that may be obscure. Where the extract shown begins or ends within a turn, the elision is marked with "[…]". At the beginning of each turn, the speaker is identified by an anonymous code (e.g. M636). The first character of the code indicates the sex of the speaker. Speech that overlaps with another speaker's turn is placed between slash symbols, e.g. "// this overlaps with part of another turn //".
M636: //Perk! A a wee a wee perk during the the the grim years.// But eh,
M635: We also used to cycle for half a dozen eggs.
M636: Aye. [laugh]
M635: Used to cycle to a farm about, oh, how many miles?
M636: Oh about two to three miles away.
M635: Three miles there and back. At eh Duncan's farm, for half a dozen eggs.
Excerpt 1. Source: Scottish Corpus file number 26
18Excerpt 1 contains the first speech laugh aye from Figure 3. This excerpt is from a conversation between brothers recalling a period when times were difficult, "the grim years". M635 continues by introducing a specific new anecdote. The speech laugh occurs on the following turn in which M635 accepts the anecdote. M635 acknowledges the truth of the memory, accepting also its relevance and appropriateness. The tone is warm, and the speech laugh acknowledges an incongruity between the effort they had been prepared to share, and the relatively small reward. The conversation continues with further development of details of the anecdote.
F638: //But when ye// coo- when ye made somethin special, like ye made soup, or ehm, apple tart or something, ye thought, "I'll take that home and show my mother it." And ye'd go, ye started //walkin the mile home, and ye had a wee bit tae taste it.//
M608: //[laugh]// //[laugh]//
F638: And then ye took another wee bit tae taste, so when ye got home, ye'd maybe crumbs left. Ye never let them see it, remember that? Ye used tae eat it //on the [laugh] the road home.//
F637: //Aye. [laugh]//
Excerpt 2. Source: Scottish Corpus file number 27
19Excerpt 2 concludes with another example of aye uttered with a speech laugh. F637 has listened to F638 tell a story of childhood mischief. Again, it is a scenario presented as being common to all ("when ye…", "like ye…" and so on), and F637 is acknowledging this commonality, although this time there is no reference to a specific shared event. This time the anecdote is presented in a joke form, with a set up and punchline ("ye'd maybe crumbs left") and is received with laughter. This helps to explain the phonetic realisation, illustrated by the spectrogram in Figure 4(a), which has a stronger and more direct expression of features of plain laughter.
20Excerpt 3 shows the context of the double-decker speech laugh illustrated in Figure 5. This laugh comes at the end of a humorous anecdote, which, again, was told with the broad form of a joke. However, this time it was the joke-teller who produced the speech laugh. Here the laugh functions to signal the end of the turn, to signal that the speaker wishes the anecdote, which in cold terms involves a competence failure and damage to property, to be viewed positively, and to invite the addressee to laugh.
M194: [...] the postman came out and said yer bus is fancily d- eh decorated this mornin. How? Here he was all was all these de- all these buntins and lights and everything all lying at the front of the double-decker [laugh]
Excerpt 3. Source: Scottish Corpus file number 25
21Excerpt 4 again involves a speech laugh by the raconteur. This time an addressee's laughter has already been prompted by the term fairy-cycle, used to describe the bicycle then belonging to a boy, who M608 will have only known as a mature man. The speech laugh on the clappers is not marked in the original transcription, and is a relatively gentle laugh. The phrase the clappers is an idiomatic expression in British English, and connotes a determined urgency. It creates an amusing image. The speech laugh highlights this, as well as signalling the speaker's acceptance that the anecdote can be viewed as humorous. The function is similar to Bolinger's concept of an "accent of interest" (Bolinger 1985: Ch. 7). There is possibly a small element of self-deprecation in Excerpts 1 and 4.
F638: and eh on a, I learned tae ride a bicycle on Robin Hutchison's fairy-cycle.
M608: [laugh] //[laugh]//
F638: //And my knees, and I was away doon McLean Street and up Market Street like the clappers! Wi this wee fairy-cycle bicycle. [...]
Excerpt 4. Source: Scottish Corpus file number 27
22In Excerpt 5, the speakers are deploring the quality of secondary education. F902 and M903 are quite serious ; this is not a humorous issue for them. Both left school aged 15 years. F826 is a university-educated teacher, who may not in fact completely agree that economic history is useless. Anyway, she does express agreement with the aspiration-carried speech laugh shown in Figure 7. Although this analysis involves speculating about unexpressed beliefs of F826, it appears that she found the moment slightly awkward but felt the need to signal congruence with the others.
F826: //It's typical of education.//
F902: //-cation that's the, that's the//
M903: //aye they dinnae ken whit they're daein either.// //I mean I was telt aboot eh linoleum fae Kirkcaldy, well fit good's that tae me noo?//
F826: //That's true, yeah [laugh]//
F902: //Linoleum fae Kirkcaldy?//
M903: //Aye fit good did that dae me?// [inaudible] intae ma heid aboot that //[inaudible] yes, aye right, okay,//
Excerpt 5. Source: Scottish Corpus file number 1042
23The next excerpt (Excerpt 6, see Figure 8) also shows speech laughter arising in a non-humorous social context. F889 is strongly criticising a neighbour’s parenting. The neighbour, who is not present, is expecting another child. F606, the interviewer, who does not know the neighbour, has been expressing congruence with F889, but doing so while making comments that are on the face of things objective and do not require personal judgement. For example, she accepts that it is inappropriate for a two year-old to be unsupervised, as a universal observation. The neighbour’s children are all girls, and F606 ventures the suggestion "so is she hoping it's a boy this time?" with devoiced speech laughter on time. There is a lot of meaning loaded in this, and it takes F889 a moment or two to absorb it. F606 is tentatively offering an imaginary resolution: If the next child is a boy, perhaps there will not be the same problems. This offers F889 a way out of the rhetorical path she has been following, and an opportunity to look forward with hope. F889 makes a positive but cautious acceptance, and moves on to another neighbour.
F889: //You know what I mean?// Oh she's smart, she's smart as anything, but that's, I mean her mother shouldna allow that.
F606: No, she shouldn't be wandering around on her //own at the age of two! Especially in this world, I mean,//
F890: //No, two years of age, no.//
F889: //No.// She's got two, she's got another one, she's, she's got another three girls, one fifteen, one eight an one six I think, an there's another one due in six weeks
F889: so that'll be a treat for everybody!
F606: //Yeah [laugh]//
F890: //[laugh]// //[inaudible]//
F606: //She won't have much time to// //run after a two-year-old then.//
F889: //No!// //No and she goes oot to work!//
F890: //cryin its head off.//
F889: No she's, I think she's got another three weeks to work an then she finishes up.
F606: So is she hoping it's a boy this time? [laugh]
F889: Well aye, I hope so, aye,
F889: well that's four girls she's got, //so she'll//
F889: be hopin for a boy, eh? an I was speakin to the man next door in his garden yesterday so he says, "Oh we're havin an addition in October", I thought, "Heavens, there's gonna be nothin but prams roon aboot here" [...]
Excerpt 6. Source: Scottish Corpus file number 996
24Excerpts 7 and 8 show the contexts of the extended speech laughs described above. In Excerpt 7, F637 is sharing an anecdote with her sisters, and in this anecdote there is an incidental element that could be presented to portray her in a weak light. She takes on the telling of this element, her fear of worms, herself. The speech laughter acts a defence. She accepts self-deprecation, but, by taking the initiative, controls the level. The single breath helps to maintain control of the turn. In Excerpt 8, the extended speech laugh occurs towards the end, in F902's final turn. However, the key is in her first turn, where she mentions politeness. In her conversation she describes attempts to evaluate someone in terms of their origins. In this final turn she is expressing a reservation about the person's birthplace. But she is aware of maintaining a polite face. The speech laughter provides a note of epistemic distance from the potentially discrediting allegation of Australian birth, a velvet glove indeed.
F637: [...] And I think I was the only tatty-howker that ever wore gloves, tatty-howkin, cause I was terrified I'd touch a worm, //in among the soil.//
F637: I was petrified fae worms! And I wore this old pair a gloves tae lift these tat- eh tatties.
Excerpt 7. Source: Scottish Corpus file number 27
F902: Cause I'm sayin tae her on the phone, ehm I mean, how do you be polite an say mm is your fa-, I'm sayin, "Whit aboot eh Barry, is he an ori- original Australian or was he one, was he descended fae one o the ones that had haen their passage paid by the government?" //[inaudible] a convict!//
F902: An she said, "No no, Barry's dad an parents emigrated eh many years ago", //cause I think Barry//
F902: must be seventy so he, they'd emigrated //from Aberdeen.//
M903: //[inaudible] if ye listen to the history// //if you listen to the//
M903: history that, the history o the, the Australians is aw a sort of a //criminals [inaudible] uh-huh//
F902: //Well that was what I was tryin to be, be polite an ask in a roundaboot,//
F902: I didnae like tae say //ken, are they descended fae the criminals or//
M903: //Well they spend the ten pounds on// //their passage wis still [inaudible]//
F902: //[laugh]// No so the, no no, he comes fae Aberdeen so that's
F606: He's alright //[laugh]//
F902: //he's okay, [laugh] well his parents come fae Aberdeen, maybe no, maybe they'd left the country// //you're never sure, [laugh] you're never sure.//
F826: //In a hurry as well, yes.//
Excerpt 8. Source: Scottish Corpus file number 1042
25Devoicing and aspiration were often associated with a softening of the communication, a distancing of the speaker from the full implications of the message if it were taken at face value. In Example 8, for instance, which is discussed in Excerpt 6 “[so is she hoping it’s a boy] this time?”, the speaker is commenting on the family circumstances of a third party, who she doesn’t know herself. The speech laughter allows the speaker to step back from full commitment to the utterance, allowing space for the addressee to respond and take the conversation forward as she chooses. To describe this as submissive would be going too far. However, there is an element of humility and an attenuation of assertiveness, or dominance. It is attractive to think of this as a systematic relationship because of the iconicity between the smaller, quieter, sound created by devoicing or aspiration and a submissive, non-dominant, social role. However, the sample is too small to draw a firm conclusion.
26There was evidence for a substantial effect of individual differences. Individual speakers tended to use different techniques for creating speech laughter. For example, speaker F606 used aspiration, devoicing and reduced intensity; F889 used pitch and amplitude fluctuation; and F638 used pitch fluctuations, often with amplitude fluctuation, trailing aspiration, an infixed hiss laugh, and the single breath to accommodate extended spans. Some speakers tended to stick to a small set of particular techniques, while others deployed a broad repertoire.
27The phonetic form of speech laughter in a small corpus of spoken Scots was compared to the form identified in earlier studies of conversations in the Kiel corpus of spoken German (Kohler 2007, 2008; Trouvain 2001). In the German corpus, the technique of dividing the stressed vowel with a rise-fall-rise amplitude contour had been identified. In the Scots corpus, this phonetic tactic was also observed. In addition, pitch contours, vowel lengthening, aspiration and devoicing were used to realise speech laughter. These techniques were sometimes used in combination, and to varying degrees. This result is broadly similar to the findings reported by Nwokah, Hsu, Davies & Vogel (1999) who examined speech laughter in mother's speech to their infant children. Nwokah et al. observed an increase in "breathiness", "syllabic pulsation", and "vowel elongation" on speech laughter. However, they found that "pitch changes" were less likely. Nwokah et al. used students to subjectively code these features rather than instrumental methods, which may contribute to any differences in their findings. Also found in the Scots corpus were instances where speech laughter stretched over many syllables, uttered with a single breath.
28Speech laughter is a suprasegmental layer upon speech with operates in the same category as intonation. It can be realised using a variety of phonetic techniques that give an acoustic impression of laughter, just as stress can be realised by amplitude, pitch, or both. The phonemic category of speech laughter is relatively abstract, then, and can be instantiated by a range of different particular forms. These differences are not narrowly determined by phonetic context, but are obviously constrained by the phonetic substrate. An ideal placing for speech laughter is on an open stressed vowel. Semantically, words like ja,yes, orno are good locations because they express valence in a direct way, and speech laughter is often a comment on the uttered valence.
29Speakers can alter the phonetic environment, for instance by lengthening a vowel, but seem to bring with them an individual repertoire of speech laugh types. Nwokah, Hsu, Davies & Vogel (1999) also found that different mothers used certain phonetic features in speech laughter more often than others. For example, one mother lowered the pitch of her utterance, whereas another characteristically raised it. Of course, we have to be careful to recognise that speech to infants may not be representative of adult speech, but the individuality appears to be ubiquitous.
30The characterisation of speech laughter as an abstract suprasegmental is reminiscent of Bolinger's (1985) characterisation of accent. Bolinger showed that accent could be realised by a range of phonetic contrasts, including pitch, amplitude, and vowel lengthening. He emphasised that the ordinary listener would typically find it difficult to differentiate the effects of pitch and amplitude. Just as Bolinger's casual listener was thought to perceive the foregrounding of the syllable, not the specific phonetic change, it seems likely that listeners perceive speech laughter, not pitch or amplitude fluctuation. That would be an interesting topic for future experimental work. Bolinger (1985) argued that pitch was the main technique for realising accent in American English, the prototype, one might say. Amplitude contour as a way of realising speech laughter was observed for both German and Scots, and pitch with amplitude was the most common and widely shared of the various methods observed in Scots. This evidence is far from conclusive, but it is plausible to hypothesise that the main techniques for realising speech laughter in different languages will correspond with their main techniques for realising accent. In other languages, such as French, a different set of phonetic features is used to realise stress. Contrastive studies in a still broader range of languages would, then, be interesting.
31The examples of speech laughter observed were used to perform a range of communicative functions. We can sketch two main categories. First, speech laughter can be used to manage self-presentation. As a speaker utters words, speech laughter provides an extra degree of freedom with which to express an attitude towards what is said. A large range of attitudes can be expressed, from humour (Excerpt 2 / Figure 4a) to tentativeness (Excerpt 6 / Figure 8). In this category, speech laughter serves to help convey subjectivity. Because the attitude is not articulated lexically, this layer of meaning can be added with less planning and more deniability. The second category is social signalling. Speech laughter can be used to signal things like the end of a turn, acceptance of a topic, and congruence with other participants in a conversation. These two categories are not clearly separated, and will often be combined in a single occurrence of speech laughter. For example, in Excerpt 1 (Figure 3), the speech laugh signals acceptance of the topic and a wry attitude towards the event being described.
32There is some indication of a relationship between the form and function of speech laughter. When expressing humour, it was most like plain laughter. When expressing tentativeness, it was in its most reduced form. It would be useful to have more systematic evidence on this issue, but the evidence here suggests at least an element of iconicity in the relationship between form and function. The theoretical notion of an image schema (e.g. Lakoff & Johnson 1980; Johnson 1987) would be one way to link the acoustic and functional analyses. In its simplest form, this link could use the elementary spatial schema corresponding to the vertical dimension (up versus down). In the functional domain being described in Excerpts 5 and 6, up corresponds roughly to confidence, dominance or superiority, while in the acoustic domain, the dimension is sonority.
33It is unlikely, of course, that a single image schema could effectively describe the full range of connections between various forms of speech laughter and their social function. Nevertheless, even this very simple schema can be applied also to the first example, of plain laughter described at the beginning of section 3.1 (Figure 1), where the intensity and vigour of the sound corresponds to superiority. The original textual context was that one woman had recounted an anecdote that was very much told at the expense of her sister, and concluded it with this ‘roar’ of laughter. And the use of shouting in aggressive interactions is almost too straightforward an example of the same pattern to be worth mentioning.
34Plain laughter provides an acoustic and phonetic prototype for speech laughter. Speech laughter needs to be perceived as if it was laughter, to be recognised as laughter and so must be in some extended sense similar to laughter. However, the prototype of laughter also comes with prototypical communicative meanings. Where the meaning of speech laughter is close to that of plain laughter, its form can be relatively close, as with the example illustrated in Figure 4(a), whose function was discussed in Excerpt 2. Further research is needed to investigate this systematically, but the evidence from the present study suggests that as the function begins to vary from the prototypical function of laughter, its form alters as well, while still maintaining a family resemblance structure in relation to the plain laughter prototype (cf. Rosch & Mervis 1975).
35There are other questions about speech laughter that it would be interesting to investigate with a larger dataset. For example, to what extent is it influenced by the formality of an interaction, differences in status, and demographic differences including gender. Other potential sources of variation in the form of speech laughter include speech-community differences, cross-linguistic and dialect variation, as well as individual speaker differences, and differences in function. The roles and relative weight of these influences is a topic for future research. But to conclude:
36Suprasegmentals, such as intonation, can be expected to play a more important and central role in cognitive phonology than in some traditional approaches to phonology. Cognitive phonology will also emphasise the continuity of linguistic and cognitive systems. In this, too, as Nathan (2008) emphasised, there are antecedents that cognitive phonology can build on. He pointed to Natural Phonology in particular (e.g. Donegan & Stampe 1979; Dressler 1985). Indeed, most phonetic and phonological research has been concerned with perceptual and motor aspects of the sound system of language. Moreover, in the work of linguists like Bolinger (1985), there are styles of explanation that are congenial to the cognitive linguist. For example, Bolinger talks of figure and ground, also a key concept in Dressler’s work, and of gradience. This is a rich foundation for cognitive phonology. The example of speech laughter illustrates these points, and emphasises that a cognitive phonology will be an expressive phonology.