Constructing a context with intonation

Constructing a context with intonation

Journal of Pragmatics 38 (2006) 1542–1558 www.elsevier.com/locate/pragma Constructing a context with intonation Jill House Department of Phonetics an...

159KB Sizes 1 Downloads 175 Views

Journal of Pragmatics 38 (2006) 1542–1558 www.elsevier.com/locate/pragma

Constructing a context with intonation Jill House Department of Phonetics and Linguistics, UCL, Gower Street, London WC1E 6BT, UK Received 28 February 2004; received in revised form 1 July 2005; accepted 5 July 2005

Abstract It is widely accepted that intonation makes a significant and systematic contribution to utterance interpretation. Less well understood are the linguistic or paralinguistic mechanisms by which the meaningful effects we identify and describe are communicated. Recent work by Gussenhoven (2002) illuminates how these functions may be reconciled by appealing to partly grammaticalised ‘biological codes’. An explanatory pragmatic framework for analysing intonational meaning from the hearer’s perspective may be found in Relevance Theory (Sperber and Wilson, 1986/1995), which allows linguistic expressions to be coded for either conceptual or procedural meaning (Blakemore, 2002). This paper investigates what procedural coding, typically involving constraints on inferential processes, might look like when applied to intonation. Discussion centres on the structure-building properties of intonation together with an examination of the range of functions associated with one particular tonal pattern, the high rising tone (HRT). # 2006 Elsevier B.V. All rights reserved. Keywords: Intonation; Pragmatics; Relevance Theory; Prosody; High rising tone

1. The nature of intonational meaning: form and function Whatever language we speak, we are most unlikely to do so on a monotone. The rises and falls in pitch that we produce combine with variations in length, loudness and voice quality to give prominence to some parts of the utterance above others, to establish discontinuities between chunks of speech, and to assign a particular melody to the whole. The melodies we use and their distribution will vary according to the language, or variety of the language, that we speak. There have been extensive descriptions of the recurrent pitch patterns found in many languages, particularly English, and the prosodic structures that underpin them (see Cruttenden, 1997 for an accessible introduction to the subject, especially to work on English; see Hirst and Di Cristo, 1998, for cross-language coverage). Some analyses of intonational form have been largely grounded in auditory perception (as in the British pedagogical tradition), whereas others rely E-mail address: [email protected]. 0378-2166/$ – see front matter # 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.pragma.2005.07.005

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1543

more heavily on measurable properties of the fundamental frequency contour which is the physical correlate of the pitch we perceive. A preliminary objective of any formal analysis must be to find a way of accounting for the range of patterns observed, for example by identifying a set of component units and providing rules for their combination within a specified domain. In this respect, various frameworks achieve descriptive adequacy, whether the units identified are themselves stretches of contour or tonal targets which constitute the turning points of such contours. Evaluation of the various models in terms of their theoretical adequacy and explanatory potential is another matter (see Ladd, 1996 for a comprehensive survey of phonological issues). Exactly how to identify the meaningful units of intonation is a fundamental question. What is the best way to divide up the pitch contour into its component parts? How do we know whether particular differences in pitch pattern should be regarded as phonologically distinct units or as phonetic variants of a single unit (a problem which frequently arises when patterns differ along a single dimension, such as pitch range)? We usually justify our decisions by showing how the patterns function: if we can reliably associate specific functions with the different forms, then we have initial evidence to support a view that some kind of ‘coding’ may be involved. The study of intonational meaning has progressed in parallel with that of intonational form, and again the varied approaches (for example, giving pride of place to grammatical, attitudinal or discoursal functions) have broad areas of overlap and agreement. The subject always raises a range of questions about the nature of intonation. By comparison with other aspects of spoken language, intonation seems to have special properties. We note that it occurs simultaneously with, and inseparably from, the words that make up an utterance, and as such is an integral part of the linguistic form of that utterance. But since the same string of words can be uttered with different intonation patterns, and since we may understand the utterance differently as a result, even though the lexical meanings remain unchanged, we conclude that intonation is meaningful. However, it is in interaction with the word string that we have to interpret the intonation we hear; intonation cannot convey linguistic meaning independently. Importantly, intonation has a dual nature. On the one hand it is partly iconic, comparable with gesture, with strong cross-language similarities testifying to its essentially non-arbitrary nature. Many researchers would follow Bolinger (1983) in regarding iconicity as central to understanding how intonation works. On the other hand, it has seemingly arbitrary, language-specific characteristics which are arguably morpho-syntactic in nature, like other linguistic signs. The details of how prosodic patterns are implemented and distributed vary widely across languages and across varieties of a language, suggesting that any universal iconicity is mediated by linguistic form. Notable among grammatical accounts of intonational meaning are Halliday (1967); Gussenhoven (1984); Pierrehumbert and Hirschberg (1990); Tench (1996); Steedman (2000). Linguistic theory has always had problems accommodating intonation because of its elusiveness, and the difficulty of defining morphemic components with consistent meanings. While the inventory of intonational units remains a matter of debate, the formal parameters that are varied to achieve functional distinctions are well established; for example: cues to prosodic phrase boundaries; the distribution of pitch accents (salient turning points in the pitch contour assigned to rhythmically prominent syllables) across the text; the actual pitch values and movements observed on the accents and at boundaries; the use of a wide or narrow pitch span. All these are exploited systematically in spoken communication. But in analysing their contribution to meaning we often have to ask whether it is primarily a linguistic or a paralinguistic function that is being fulfilled. Does the intonation affect the proposition expressed, making it an integral part of the linguistic content? Is it giving us non-linguistic information about the speaker’s

1544

J. House / Journal of Pragmatics 38 (2006) 1542–1558

attitude or emotional state? Does it contribute to the propositional meaning by packaging the message in a particular way? Is it best seen as a way of structuring discourse and interaction? It can be useful to think of intonation as having at least three different functional orientations, any of which may be more or less important for a given utterance. Firstly, we may observe an orientation towards the speaker himself,1 which will include indexical information, of the sort described by Abercrombie (1967), such as indicators of social and regional affiliation, gender, and idiosyncratic properties of the speaker’s voice that indicate his identity. Speaker-oriented properties also include paralinguistic information, whether intentionally or unintentionally communicated: attitude, emotion, all aspects of affect. Secondly, there is the genuinely linguistic orientation, where intonation is used as an integral part of linguistic form to contribute to the propositional content in some way, or to clarify the context which the listener should use to recover the speaker’s meaning. We may regard this as its grammatical function. Thirdly, intonation enters into the on-going process of constructing discourse, with an orientation towards interaction with the listener in dialogue and conversation generally, or towards the structure of a text, for example in scripted or semi-scripted monologue. All types of orientation may be part of the speaker’s explicit communicative intentions. 2. Linguistic interpretation Paired examples such as (1)–(5) illustrate the contribution of intonation to utterance interpretation. In presenting them in their written form we normally have to rely on differences in punctuation or conventions for indicating emphasis (bold or italic fonts, or capitalisation) to suggest the intonational differences. (1)

a b

I didn’t vote Labour because of Tony Blair. I didn’t vote Labour, because of Tony Blair.

(2)

a b c

Jenny teaches VIOLIN. Jenny TEACHES violin. JENNY teaches violin.

(3)

a b

Mary hugged Lucy, and then Jack hugged her. Mary hugged Lucy, and then Jack hugged HER.

(4)

a b

The choir was great. Their standard was really professional. The choir was great. Their standard was really professional!

(5)

a b

Coffee. Coffee?

Reading these examples aloud, respecting the punctuation and/or emphasis shown, is likely to elicit variant intonation patterns that will in turn cause us to interpret the lexically identical utterances differently. We are likely to produce (1a) as a single prosodic phrase, but (1b) as two phrases, with a boundary between them at the position of the comma; the phrasing will influence our understanding of the scope of the negative and directly affect the proposition expressed. In (2a–c), we can choose to place our main emphasis on any of the three words, to reflect their 1

To simplify the use of gendered pronouns in this paper, the speaker is described as ‘he’ and the listener as ‘she’ throughout.

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1545

relative importance in the situation in which we make the utterance. In (3b), an accent on the pronoun (indicated by the capitalisation) will suggest a different referent from the unstressed one in (3a). In (4b), the sense of surprise indicated by the exclamation mark ‘!’ may encourage us to pronounce the sentence with raised pitch on ‘professional’—suggesting that the speaker was in fact reporting on a competent bunch of amateurs whose standard was higher than expected; and in (5b) we are likely to interpret the question mark ‘?’ as denoting a rising, questioning intonation, appropriate to an invitation. The examples all illustrate well-known functions associated with manipulating the intonational parameters outlined earlier. For example, (1) illustrates the ‘chunking’ function, whereby intonational phrasing tells us which pieces of speech material to process together as a unit. The ‘focussing’ function is demonstrated by (2) and (3), in which accent placement helps us to sort out the information structure—what should be considered as background and what as foreground. In (4), we have an indication of how local choice of pitch span or ‘key’ can be used to reinforce (4a) or override (4b) certain expectations. Finally some kind of ‘speech act’ function, whereby choice of tone indicates the speaker’s intentions for that utterance, is illustrated by (5). What all the examples have in common is that the intonation we choose will constrain the hearer’s interpretation of the message, whether by clarifying syntactic structure or by organising the information so as to give access to a set of background assumptions which help to define the context of utterance. Intonation (or its poor relation, punctuation) contributes to successful communication by allowing us to eliminate a range of possibilities in favour of one interpretation in particular. 3. Paralinguistic and discoursal interpretation The examples above illustrate how intonation may enhance the linguistic content of the communication, but this may itself be only part of what a speaker is communicating at a given moment. Paralinguistic information, such as the speaker’s attitude or emotional state, may also be communicated. In (4b), an overtly communicated sense of surprise qualifies as paralinguistic information, intentionally communicated, which in this instance has implications for the hearer’s interpretation of the linguistic message. The speaker has exploited an aspect of linguistic form to communicate this attitude. On other occasions, aspects of intonation or voice quality may communicate (say) depression, or scepticism, or amusement, or anxiety, or drunkenness—either deliberately or unintentionally (see Wharton, 2003 for discussion of the showing—meaning continuum). Whereas the linguistic functions described above increase the efficiency with which the listener may gain access to the speaker’s message, paralinguistic information would seem to be a genuinely additional component. A further important function served by intonation is its role in structuring text and interaction (Wichmann, 2000; Wennerstrom, 2001). This is arguably an extension of the grammatical function, as it relies on markers of information structure and status to negotiate common ground in the communicative process. Again, there is evidence of a highly systematic exploitation of prosodic parameters to indicate continuity, finality, turn-taking cues, topic structure and so on. The process of prosodically shaping the discourse has further implications for the understanding of the linguistic communication, since participants are organising the pieces of information that make up a longer discourse to indicate dependencies, coherence and discontinuities, which may in turn have significant effects on the meaning conveyed (Hirschberg and Pierrehumbert, 1986; Silverman, 1987; Wichmann, 2000). In interaction, intonation is also used to help participants evaluate the success of the on-going communication process.

1546

J. House / Journal of Pragmatics 38 (2006) 1542–1558

All types of meaning may in principle be expressed simultaneously, though a particular orientation may predominate. Recent work by Gussenhoven (2002, 2004) makes a persuasive case for reconciling the iconic, quasi-universal aspects of intonation with the grammaticalised, language-specific sound-meaning correspondences. He proposes that form-function relations in intonation are derived from biological conditions (larynx and vocal tract size; energy; and the breathing process) which give rise to three biological ‘Codes’—of Frequency (following Ohala, 1983), Effort and Production, respectively. Universal affective and informational interpretations of pitch height (Frequency Code), excursion size (Effort Code) and phrasing (Production Code) may be exploited in language- or variety-specific ways to establish systematic, consistent, grammatical sound-meaning correspondences (summarised in Gussenhoven, 2004:95). All the Codes have a role to play in establishing information structure and status; affective meanings are restricted to the Frequency and Effort Codes; the more arbitrary grammaticalisations are associated with all the Codes. A strength of this theory is that it unifies the disparate aspects of intonation; an implication must be that a given pitch pattern may lend itself simultaneously to a number of different interpretations—affective, informational and grammatical. It is not clear whether these interpretations should be regarded as simultaneously valid, or whether a single dimension of meaning should override and cancel out the others, and how listeners resolve any potential conflicts. 4. Pragmatics and intonational meaning Pragmatics is the study of utterance meaning. In other words, we are interested not only in the words themselves, but in everything the speaker intended to communicate in the context of utterance, whether this is explicitly expressed, or implicit, derived by making inferences based on adding what is explicitly expressed to background assumptions constituting the context. The prosodic form of an utterance will make a crucial contribution to conveying the speaker’s meaning, as the examples in section 2 have demonstrated. Intonation must therefore be a matter of serious interest to pragmatic theory. In considering exactly how intonation communicates its different effects a number of theoretical questions need to be addressed. Are listeners using a decoding mechanism, inferencing procedures or some combination? If linguistic (de)coding is involved, how does it work? Is there a potential conflict between linguistic and paralinguistic communication, or can the latter be seen as an enrichment of the linguistically encoded semantic representation? As discussed earlier, intonation constrains the possible interpretations of an utterance in a number of ways. It enables the hearer to resolve ambiguities, to access and evaluate the relative importance of particular contexts (sets of assumptions) relevant to understanding the utterance, and to evaluate the status of particular assumptions. In other words, it plays a vital role in establishing the contexts within which utterances should be interpreted, and in cutting down the hearer’s processing costs by making communication more efficient. These are notions that are of central importance to Relevance Theory (RT) (Sperber and Wilson, 1995). Primarily a theory of cognition, RT incorporates a theory of communication which covers both linguistic and nonlinguistic varieties. It has been claimed to provide an appropriate theoretical framework within which to consider the role of intonation (House, 1989, 1990; Vandepitte, 1989; Fretheim, 2002; Wilson and Wharton, 2006). A few key principles and claims made by the theory, presented in a simplified form below, should help to make this clear. Underpinning RT is the Communicative Principle of Relevance: ‘‘Every act of ostensive communication communicates a presumption of its own optimal relevance’’ (Sperber and Wilson, 1995:158). The Principle has important implications. Firstly, if someone directs a

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1547

message towards us, whether verbal or non-verbal, we will try to make sense of it: our cognitive systems are geared to doing so. Secondly, we will find the interpretation that seems to be most relevant for the least amount of processing effort. Understanding communicated utterances proceeds in much the same way as our understanding of other aspects of the world which come to our attention: we will formulate an assumption about what we observe or hear; we will place this assumption in the context of some other set of assumptions (our cognitive environment), then make inferences that may or may not lead us to modify our set of assumptions about the world. The Principle of Relevance means that when a communication is overtly directed towards us, we will expect it to be relevant to us, in the sense that it will update or modify our cognitive environment, and we will attend to it with this in mind. Communication is then relevant if it has cognitive effects, as summarised by Blakemore (1992:30): 1. It may allow the derivation of a contextual implication. 2. It may provide further evidence for, and hence strengthen, an existing assumption. 3. It may contradict an existing assumption. A contextual implication may be further defined as a deductive assumption involving the synthesis of proposition P with context C (for a much fuller explanation see Sperber and Wilson, 1995:108–137). As communication progresses, so the set of assumptions in our cognitive environment – the context – is continuously updated. Linguistic communication is just one kind of ostensive communication. It has special properties, because any language is made up of arbitrary sound-meaning correspondences that have to be decoded if we are to understand the communication properly. But it is widely recognised that linguistic messages are never fully understood by a process of decoding alone. Inferencing will be needed in order to flesh out even the explicit content of the message (for example, reference assignment; co-ordinates of time and place; resolving lexical ambiguities), and contextual assumptions are needed for this to be successful. In other words, what is communicated explicitly is recovered by the hearer using a combination of linguistic decoding and pragmatic inferencing. When the hearer has formed a hypothesis about the explicit content, she can use her representation of this content to update the set of assumptions constituting her cognitive environment; then using further inferential processes she can recover the implicit content of the message. RT makes a distinction between explicatures and implicatures: explicatures are assumptions made using inference procedures to develop the linguistically encoded semantic representation of an utterance. For example, in order to understand the explicit content of an utterance such as: (6)

They lack fibre

the listener must assign a referent to the pronoun ‘they’, and determine what the speaker means by ‘fibre’. This information is not explicitly coded, but is derived by inferencing procedures, which will yield very different results in contexts (a) and (b). (a) (b)

Parents have been complaining about the quality of school dinners. They say they lack fibre. The army officers have been grumbling about the quality of their new recruits. They say they lack fibre.

1548

J. House / Journal of Pragmatics 38 (2006) 1542–1558

The assumptions made by the listener thus develop the linguistically encoded semantic representation and are explicatures. The explicatures of an utterance may then be used to make further inferences, which are the implicatures. The hearer of (6) in context (a), for example, having inferred that ‘they’ refers to the dinners, and that ‘fibre’ refers to food content (explicatures), may use her knowledge about food to infer that the meals in question are lacking in fruit and vegetables. This would be an implicature. A further distinction can be made amongst explicatures: information which is explicitly communicated but which is not part of the proposition expressed – for example, the speaker’s attitude towards the proposition – is regarded as a higher-level explicature (Wilson and Sperber, 1993). How might intonation fit into this scenario? We have already looked at claims that aspects of intonation can be grammaticalised, attaching language-specific meanings to certain pitch patterns. It has been further claimed that these are derived from universal biological codes, but insofar as they come to encode arbitrary, language-specific meanings they resemble linguistic morphemes. If we accept this, at some level our processing of the pitch patterns of an utterance must involve some decoding. At the same time, intonation is inextricably bound up with the segmental string, and one of its key functions would seem to be its role in clarifying aspects of the proposition expressed by that segmental string. What intonation appears to be encoding is a set of instructions for interpreting the morpho-syntactic content of the utterance, and to the extent that it contributes to the encoded semantic representation of the utterance, it must contribute to the explicatures of that utterance. Intentionally communicated paralinguistic information that does not directly affect the proposition will contribute to higher level explicatures. Inferencing procedures based on further aspects of the context will enable the hearer to decide whether broad, universal interpretations of the biological codes, or grammaticalised, language-specific ones, are the more important. As always, she will recover the interpretation which yields optimal relevance for minimum processing effort. Recent work in RT gives us further insights into how intonational meaning may be processed. Blakemore (2002), working primarily on discourse connectives such as ‘but’, ‘however’ and ‘nevertheless’ develops ideas outlined in Blakemore (1987), and taken up by Wilson and Sperber (1993) to account for their meaning in terms of a notion of procedural coding. The hypothesis is that some linguistic expressions do not encode conceptual meaning as such, but a procedure which involves a constraint on pragmatic inferencing. ‘‘The function of . . . ‘procedural’ expressions would be to guide the hearer towards the speaker’s meaning by narrowing the search space for inferential comprehension, increasing the salience of some hypotheses and eliminating others, thus reducing the overall effort required.’’ (Wilson and Wharton, 2006). Blakemore argues that procedural coding may directly encode the type of cognitive effect intended (as in the case of ‘but’, where the listener is led towards the elimination of an existing assumption); it may constrain the context within which cognitive effects are derived (‘however’); it may orient the hearer towards a particular set of assumptions (‘well’); it may place procedural constraints on explicit content (‘then’); or it may be an indicator of the status of non-declaratives. The exact scope of possibilities for procedural coding has still to be determined, but aspects of intonation, which as we have seen encodes instructions for interpreting the status of the text, cry out for a procedural explanation. 5. Intonation and structure Let us return to our original examples, and sketch out what may be involved in a procedural account of their interpretation.

J. House / Journal of Pragmatics 38 (2006) 1542–1558

(1)

a

I didn’t vote Labour because of Tony Blair j T% (= I did vote X but not because Y)

b

I didn’t vote Labour j because of Tony Blair j T% T% (= I did not vote X because Y)

1549

(1a) and (1b) differ in terms of phrasing. We must presume that any procedural coding is located on the phrase boundary itself, which plays a demarcative role, separating groups of linguistic expressions which belong together syntactically and semantically, allowing the hearer to infer in a cost-effective way the likely scope of the negative and thence the intended proposition. Without the contribution of phrasing we have no unambiguous access to the underlying syntactic structure and to the proposition itself. The effect of the phrase boundary in (1b) is to place a constraint on explicit content. In this instance it blocks an interpretation in which negative scope is carried over to the subordinate clause (see Fretheim, 2002 for a procedural account of negative scope in Norwegian). Note that it is the presence of a prosodic boundary (shown above as ‘T%’) that is relevant here, irrespective of the particular tone on which it is realised. A fully specified autosegmental transcription would have to assign an explicit boundary tone (e.g. H% or L%), but the inferences associated with the demarcative function are tone-neutral, which is what is indicated by the ‘T%’ notation (where T = any tone). In practice, a full interpretation of a boundary tone will involve adding in any meaning coded by the tone itself. (1a) and (1b) are also likely to differ in terms of pitch accent and boundary tone. The interpretation associated with (1a) will be reinforced by the use of a ‘fall-rise’ contour, or final H*L H% pitch accent + boundary tone pattern, but arguably the main function performed by the tonal pattern will be to draw attention to the status of the intonation phrase with respect to the discourse being constructed, rather than to the underlying syntactic structure. For example, the proposition in (1a) is only likely to be uttered in a context where it is understood that Tony Blair was a factor influencing people’s voting choice, and the speaker is thus referring to some ‘given’ information (Brazil, 1975). The interaction of tonal pattern and phrasing is not explored further here, but I would argue that the difference in interpretation dependent on syntactic structure is crucially cued by the phrasing. As is well known, it is not just in cases of negative scope where phrasing can constrain explicit content. For example, the correct interpretation of sentence adverbials, intransitive verbs and nondefining relative clauses may depend on the presence of a phrase boundary. The listener is thereby steered towards a particular interpretation of the propositional content, or in the case of relative clauses, particular inferences about their defining status. Boundaries simply direct the listener to treat the linguistic material between them as coherent, and arguably that instruction is encoded in the boundary itself. The biological origin of intonational phrasing may lie in a correlation between utterances and breath groups (Production Code), but its derived demarcative function, the chunking of linguistic material so as to create informational structure – probably a universal property of phrase boundaries – no longer necessarily coincides with the speaker’s breathing activities. The nature of the hearer’s inferencing task is procedural, and the effect is structural. Creation of the structure constrains explicit communicative content and indirectly constrains the inferencing procedures available to the hearer when processing the communication. Similarly, the presence or absence of pitch accents on linguistic expressions is a device for structuring information independently of the actual tone used. In the words of Pierrehumbert and

1550

J. House / Journal of Pragmatics 38 (2006) 1542–1558

Hirschberg (1990:288), ‘‘All pitch accents render salient the material with which they are associated’’. Conversely, linguistic material that is left unaccented is marked as relatively less salient, so that the relative importance of different pieces of structure is signalled. Pitch accents normally involve a degree of pitch excursion, and the scale of such an excursion can be related to Gussenhoven’s Effort Code. The focussing function of accents, and their role in English in differentiating foreground from background, or new from given, has been extensively discussed in the literature (see for example, Halliday, 1967; Ladd, 1980; Gussenhoven, 1984; Sperber and Wilson, 1995) and will not be dealt with in detail here. In simple examples such as (2a–c) the information-structuring function is performed in a straightforward way. By definition, accented material (here shown by the underspecified T* associated with the rhythmically prominent syllables of linguistic expressions) indicates foregrounded information, and therefore part of what is in focus. The word ‘teaches’ in (2a) is shown as optionally accented: even if it is within the domain of what the speaker presents as in focus, language-specific rules sensitive to the rhythm of English will allow the expected accent to be suppressed without affecting its information status. Additional information about the status of the salient material is potentially provided by the realisation of T* as one specific accent type or another. (2)

a

Jenny teaches violin. T* (T*) T*

b

Jenny teaches violin. T* T*

c

Jenny teaches violin. T*

In procedural terms, the accent is interpreted as an instruction to make selected linguistic material salient, or ‘in focus’. It thus contributes to the creation of another kind of structure, focal structure. The proposition itself is unaltered by the focal structure, but the implied relegation of non-salient parts of the utterance to the background guides the hearer to structure the context in a particular way, so as to infer systematically what the most relevant set of underlying assumptions making up the cognitive environment might be (see Sperber and Wilson, 1995:202ff for a detailed discussion of the focal scale). The effect is to constrain access to the context within which cognitive effects will be derived, narrowing the search space and reducing processing effort. In examples such as (3b), the salience afforded by the pitch accent on the pronoun directs the hearer to seek some referent that is different from the one she would assign in (3a). (3)

a

Mary hugged Lucy and then Jack hugged her. T* T* T*

b

Mary hugged Lucy and then Jack hugged her. T* T* T* T*

Reference assignment is an essential inferencing procedure that the hearer must perform in order to flesh out the explicit content of the speaker’s utterance (Carston, 1988). It is in the nature of pronouns to represent ‘given’ material, and their unmarked focal status is correspondingly given,

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1551

so they are unlikely to require a prosodic accent. For optimal relevance the hearer will assign the default referent, the one that is most immediately accessible—which in (2a) is most likely to be Lucy (the absence of accentual prominence on ‘hugged her’ suggests that the whole verb phrase be treated as given, and the position of ‘her’ as the direct object of ‘hugged’ parallels that of Lucy in the first clause, reinforcing the relationship between name and pronoun). A pitch accent on a personal pronoun serves a deictic function, and in the process activates a wider search space for potential referents. The effect in (2b) will be to reject the default reference candidate in favour of the next most readily accessible. This will probably be Mary. But in a context where other female candidates are physically present, the speaker could pick out someone else entirely by accompanying the accented pronoun with a pointing gesture. The accent does not itself directly encode any reference switching, but its function in assigning salience orients the hearer to update her cognitive environment in a particular way. Arguably, in interaction with a pronoun, an accent can act as a procedural constraint on explicit content, but this is achieved by the procedure of building focal structure, which in turn puts constraints on inferencing. Salience is, of course, a gradient phenomenon: we can have degrees of salience, and the relative prominence of pitch accents will lead us to infer different contexts and make different interpretations, as in examples (4a,b): (4)

a

The choir was great. Their standard was really professional.

b

The choir was great. Their standard was really professional!

For the sake of argument, let us assume that in both cases there is a H* accent associated with the second syllable of the word ‘professional’. In order to elicit the ‘unexpected’ interpretation associated with the exclamation mark in (4b), this accent needs to be ‘extra high’ in pitch, and that in (4a) merely ‘high’ (this would equate roughly to the use of ‘normal High’ or ‘emphatic High’ accents (Ladd, 1996), or to using ‘high key’ versus ‘low key’ (Brazil, 1978). A continuous gradient of possibilities is possible between High and emphatic High, but above a certain threshold the hearer will interpret the pitch height as implying a contrast between the linguistic material on which it is found and preceding material. This in turn will give access to different sets of contextual assumptions which after inferencing procedures will yield different cognitive effects: (4b) will give access to the contextual assumption that the singers concerned were a group of amateurs, which may reinforce or contradict an assumption already held by the hearer. In recent experimental work, following an earlier study by Ladd and Morton (1997), Chen (2003) investigated precisely this kind of threshold, between ‘normal High’ and ‘emphatic High’ accents in English, and found evidence for a discrete boundary between the two. In her study, the meaning distinction was simply between ‘expected’ and ‘unexpected’—there was no preceding material in the test sentence, so any contrastiveness implicit in the ‘unexpected’ interpretation must be with an unspecified set of background assumptions. In Gussenhoven’s (2002, 2004) account, emphasis, realised as a relatively wide pitch excursion, is seen as an informational interpretation of the Effort Code; this fits with the local expansion of pitch span in examples like (4b). The Effort Code thus appears simultaneously to be operating linguistically (through accent assignment and the building of focal structure), informationally (through emphasis) and affectively (by denoting surprise). In interpreting the utterance the three aspects do not seem to be mutually exclusive. The effect must again be to reduce processing effort for the hearer, who will be guided by the relative saliencies of the linguistic expressions she hears to make inferences about the contexts in

1552

J. House / Journal of Pragmatics 38 (2006) 1542–1558

which the speaker’s utterance will be optimally relevant. It is not clear how extra-high accents should be represented phonologically,2 or exactly what procedure they encode other than to draw extra attention to a particular piece of structure. Pitch height must always be perceived relative to something, whether to some normal expectation of pitch span in an isolated utterance, or to an expectation of declination already set up by earlier accented material. In an utterance like (4b), the extra-high accent disrupts normal patterns of declination. Where this occurs, it would seem to instruct the hearer to adjust the context she is constructing by re-ranking the salient material; the inferences that she will make in consequence will depend on that context. It would be tempting to see the extra-high accent as directly encoding a cognitive effect – the contradiction of an existing assumption – in a way similar to that proposed by Blakemore (2002) for ‘but’ – but this would be misleading. In other contexts an emphatic accent could be used simply to confirm an existing assumption in an enthusiastic manner (extra-high H* on ‘BRILLiant’): (7)

I gather you enjoyed the play?

– Yes, it was BRILLiant. H*

Rather, any coding relates to a structure-building procedure, and the cognitive effects are derived through inferences made based on that structure. It may be argued that the way in which accents in general draw attention to particular linguistic expressions, by making them salient, should be regarded as a type of natural highlighting rather than involving any linguistic coding. If this were the case, we might expect essentially the same patterns of intonational prominence to occur across languages, and language varieties. Instead, we find a range of language-specific conventions for indicating focus and emphasis, conventions that have to be learned by non-native speakers seeking native-like competence. The conventions of one language do not necessarily translate into another, suggesting some arbitrariness about linguistic form, which in turn supports the possibility of patterns being coded. Useful evidence for cross-linguistic variation may be found in Ladd (1996:168–197). Accent assignment, like boundary assignment, is designed to make life easier for the hearer, whose interpretation3 of the relative saliencies will be used to construct a relevant context with minimal effort. For the speaker, exploiting the Effort Code in this way is a deictic activity. The structure he builds directs the listener to the relevant contexts independently of the particular pitch accents used. An understanding of this is essentially separate from the interpretation of the individual pitch accents themselves, or the pitch contour as a whole, which will tell us about the status of the salient material. 6. Pitch patterns and the high rising tone Our examples (5a,b) depend for their interpretation not on salience as such, but on the direction of the pitch movement, falling in (5a) and rising in (5b): 2 The most obvious candidate in the autosegmental-metrical framework is L + H*, which has been associated with a contrastive function (Pierrehumbert and Hirschberg, 1990; Wennerstrom, 2001), but this is characterised by a low leading tone, giving the accent a ‘spikier’ shape, rather than by extra pitch height. 3 This interpretation will depend on language-specific conventions associated with the grammar of accent placement (see inter alia Halliday, 1969; Ladd, 1980; Gussenhoven, 1984; Cruttenden, 1995; Pierrehumbert and Hirschberg, 1990).

J. House / Journal of Pragmatics 38 (2006) 1542–1558

(5)

1553

\

a

Coffee H* L L%

b

/

Coffee H* H H%

This pitch movement is represented iconically by tone marks (‘\’ and ‘/’) placed before the word, consistent with a ‘high fall’ or ‘high rise’ nuclear tone in the British contour analysis tradition (see for example, O’Connor and Arnold, 1973). The autosegmental representations denote sequences of tones that describe essentially similar contours. In (5a), there is a relatively high starting point on the accented syllable, falling to low pitch on the second syllable. In (5b), the starting point is again relatively high, but the pitch then rises further through the rest of the word up to the intonation phrase boundary. The phrase accent components (H) are redundant to the phonetic realisation, since they overlap with the high boundary tone (H%) and will therefore not affect the interpolation between the pitch accent and boundary tone; but they are required in the representation because of the potential contrast with e.g. H* L H% (fall-rise). Up till now this paper has had little to say about the use of high or low pitch, or about the changes of direction in pitch essential in describing any speech melody, which are central to any characterization of intonation. In a given (non-tonal) language, researchers investigating the contribution to utterance interpretation of the pitch patterns themselves look for some kind of systematic relationship between pitch and meaning, whether they consider the meaningful units to be contours or sequences of tonal targets. The assumption that there is a paradigmatic choice to be made amongst a set of forms with corresponding meanings (however abstract) supports the view that there may be some kind of code involved. The ‘meanings’ identified may involve exploitation of the Frequency Code for grammatical or paralinguistic purposes. Intonation languages like English or Dutch have no frequency constraints imposed by the lexicon, leaving them relatively free to exploit a choice between H and L tones for intonation purposes. The story will clearly be more complicated in languages with lexical tone, or lexical pitch accent, where frequency patterns enter into arbitrary sound—meaning correspondences, and therefore restrict the availability of pitch patterns at the intonational level. Fretheim (2002) warns against making strong claims about universal meanings based on a choice between H and L when there are language-specific constraints on intonational form. A fully worked out account of intonational meaning in the RT framework will need to evaluate the merits of competing theories of intonational phonology. For example, autosegmental models consider the pitch contour to be derived from interpolation between fixed tonal targets (pitch accent, phrase accent, and boundary tone in (5a,b)); contour models, on the other hand, treat the pitch movements themselves as the primitive units (‘high fall’ in (5a) and ‘high rise’ in (5b)). The analyses make very different predictions about what can be coded, since in the autosegmental framework we have the option of attaching meaning to the tonal targets themselves, and regarding any meanings attached to tonal sequences (contours) as a composite of the meanings associated with the individual tonal targets; such a compositional approach is found in Pierrehumbert and Hirschberg (1990). Additionally, or alternatively, tonal sequences may also be assigned meanings comparable to those proposed for contours in contour models. Within either theory of intonation, there is no consensus about the precise inventory of phonologically distinct patterns, and there are important variations across regional and social varieties. These issues are unresolved. To explore questions about the procedures involved in the listener’s interpretation,

1554

J. House / Journal of Pragmatics 38 (2006) 1542–1558

this section concentrates on the type of English phrase-final high rising pattern expected in (5b): the high rising tone (HRT).4 An important characteristic of this rising pattern is that its end-point comes high in the speaker’s pitch range. The level at which the rise begins in the accented syllable may be more variable: typically it is also relatively high (well above the speaker’s baseline), as we assumed when characterizing it as H* H H%. Alternatively, a rising pattern covering a wider pitch span, but also having a high termination, may be achieved with a low starting point: L* H H%. There is some evidence that the two patterns are phonologically distinct, though accounts of HRT do not always recognize the distinction. In the initial representation of (5b) in section 2, the punctuation mark ‘?’ instructs the reader to interpret the utterance as a kind of question. It was predicted that in reading aloud, a rising tone would be used. But if we start instead with the spoken utterance, realised with the high rising pattern, a range of other interpretations is possible, depending on what is accessible in the context. The speaker may be in the process of making suggestions for a shopping list, for example, where the high boundary merely signals non-finality. The interrogativity interpretation represents a possible grammaticalisation of the pattern, but cannot be regarded as the only possible meaning. This is not to say that the rises used in lists and questions are necessarily phonetically identical, but that there would seem to be a genuine overlap, at least in Southern British English, in the forms used for both functions (a matter for empirical investigation). It may be revealing at this point to reconsider the different orientations outlined in section 1, and explore the implications of these for interpreting the high pitch at the boundary. An orientation to the speaker himself, or the communication of indexical information, suggests a range of Frequency Code-inspired affective possibilities: politeness, deference, timidity, vulnerability, or friendliness, for example. If the orientation is to the linguistic message, the speaker may be using high pitch to express uncertainty about the status of the current intonational phrase. Gussenhoven (1984) assigns a ‘testing’ function to simple rises: the speaker does not commit himself to the relevance of adding the information expressed by the linguistic material to the discourse model, but leaves it up to the hearer to decide. (Note that Pierrehumbert and Hirschberg (1990) assign a similar function to the L* accent – which in all the examples they give is followed by H%, so effectively surfaces as a rise.) Finally, if the orientation is towards building the on-going discourse, or towards interaction with the hearer, the high boundary tone (H%) may simply signal open-endedness, continuity, non-finality, or a need to check with participants in the interaction that they are successfully negotiating common ground. This fits in with the ‘‘forwardlooking’’ generalisations about phrase accents and boundary tones proposed by Pierrehumbert and Hirschberg (1990), and brings us back to the notion that a basic function of intonation is to build structure. Non-finality implies that the utterance in question is part of some larger piece of structure, which is in the process of being constructed. A procedural hypothesis concerning the high boundary tone (H%) itself is that it encodes an instruction to interpret the preceding phrase as part of a larger piece of structure, thereby giving it an open-ended status, and indicating a wider context. The hearer must use her cognitive environment to make appropriate inferences about the speaker’s intentions, to work out what the wider context might be. As always, the hearer will select the first interpretation that is relevant enough to yield cognitive effects. This hypothesis has the advantage of being consistent with the other structure-building functions identified for intonation – the demarcation into coherent chunks, the distribution of salient material within those chunks to indicate focal structure, and the 4

In the Antipodean literature (e.g. Guy and Vonwiller, 1984; Fletcher and Harrington, 2001), ‘HRT’ stands for ‘High Rising Terminal’; I follow Cruttenden (1995) in applying the same abbreviation to ‘High Rising Tone’.

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1555

ranking of that salient material – all of which conspire to organise the linguistic content so as to cut down processing effort for the hearer. The boundary tone itself is only part of the story, of course, when it comes to analysing the different tonal patterns that precede the boundary. Open-endedness is just one component of the interrogative interpretation strongly associated across a wide range of languages with a high rising pattern, phonologised as H* H H%. The pattern is so well established as an interrogative marker that it may be considered part of the grammar of utterance modality. If there is any coding involved with this interpretation, should it be attached to one of the component tones or to the sequence? Similarly, should the connotation of uncertainty, claimed by Gussenhoven to be an informational interpretation of the Frequency Code, be attached to the rising shape or simply to the high pitch? It is outside the scope of this paper to analyse the range of pitch configurations (pitch accents; nuclear tones) that may occur in a non-tonal language. Plausibly their varied contribution to utterance interpretation may also depend on procedural coding of some kind which characterises the non-declarative status of the intonational phrase, or of the focussed components of that phrase. How should the hearer decide whether affective functions of high pitch (deference, vulnerability, etc.) should be given more weight than linguistic interpretations like interrogativity? Affective and informational/linguistic interpretations cannot be mutually exclusive, but the balance between them shifts, and insofar as different interpretations are available, there is potential ambiguity. RT claims that the hearer will opt for the first interpretation relevant enough to bring about cognitive effects; but as with any potentially ambiguous utterance, mistakes in interpretation may sometimes be made. A complicating factor in the case of HRT is its widespread use, primarily by the younger generation, in situations where older generations might not think to use it: for example, on declarative utterances presenting new information, where an interrogative interpretation makes no sense. The example below is taken from narrative data collected by Shobbrook (2002:28), where the speaker describes a scuba diving trip: ‘‘and um because you’ve got like loads of diving equipment that you don’t know how to use, they – the snorkelers like dive /down and they like adjust your /things all the /time .... You just suddenly realise that you’re rising and it’s because somebody’s come down and changed the pressure in your /thing ‘n’ you’re like oh!. . .’’ Described variously as ‘talking up’, ‘uptalk’ or ‘upspeak’, the usage of this pattern is a relatively recent innovation in British English, attributed variously to the influence of Antipodean or North American (‘Pacific Rim’) speech (Cruttenden, 1995). It is described as fulfilling a sociophonetic ‘bonding’ function amongst users: ‘‘It signals a wish to include and be included and is reinforced by the acceptance it brings’’ (Bradford, 1996:23); it signals a co-operative orientation within the peer group, and may be used as a kind of ‘checking’ device to elicit feedback from the listener (Cruttenden, 1995). Increasingly accepted as time goes by as a completely normal pattern in interaction, uptalk is a sociophonetic pattern that has aroused disapproval, irritation and sometimes hostile reactions among non-users, who may regard it as a legitimate target for fun. As the pattern becomes established, the reaction described below becomes less common: ‘‘Tonight we have a special?’’ she said. ‘‘I don’t know, I’m sure,’’ said Bernard, examining the menu. But apparently the girl’s rising intonation did not signify a question, for she proceeded to tell him what the special was: spinach lasagna. (Lodge, 1991:104)

1556

J. House / Journal of Pragmatics 38 (2006) 1542–1558

Misunderstandings potentially occur when hearers mistakenly attribute a questioning function to uptalk, or perceive uncertainty where none was intended (Guy and Vonwiller, 1984, reported in Cruttenden, 1995). This must be because the cues that they hear are consistent with the cues for interrogativity or uncertainty with which they are familiar. Whether these are procedurally coded cues, indicating something about the status of the phrase, or a matter of conventional association, is open to argument. A further implication would be that there was a kind of lexical homophony involved, giving rise potentially to ambiguity. A more conservative proposal might be that only the most general, abstract properties are coded by the HRT pattern, and that a more precise specification of function is derived by inferencing, based on any contextual cues to the appropriate orientation. Membership of a particular regional or social language group may be a relevant part of that context. Any misinterpretation will inevitably involve increased processing costs for the hearer, who must backtrack and start again if she is to infer the correct orientation in the end. We expect intonation to provide us with clear guidance to interpretation by constraining our inferencing processes appropriately, and to find that we have mis-interpreted intonational cues is disconcerting. It may be particularly disconcerting when a linguistic innovation commandeers a familiar pattern and puts it to some new use. Once we have ‘tuned in’, our interpretation quickly adjusts, but misgivings may remain, together with a persistent feeling that the speaker has got it wrong, as if he were mis-using a well-established vocabulary item (compare the use of ‘wicked’ to mean ‘good’ among many younger speakers). For the hearer to choose the interpretation which best matches the speaker’s intentions implies access to a common coding system, as well as to a context constructed and negotiated co-operatively. We may predict that if an expanding range of functions becomes associated with a particular pattern, such as HRT, it will become necessary in time to distinguish between those functions by systematically using different phonetic variants of the pattern to express them. There is some evidence that for some speakers a phonologisation of two pattern variants is indeed taking place. For example, Fletcher and Harrington (2001) found that in Australian English map task data the starting point of the HRT contour – the pitch accent itself – was systematically higher in genuine questions (i.e. H* H H%) than in declaratives (L* H H%). In a study using comparable map task data Shobbrook and House (2003) found no such consistency yet amongst speakers of Southern British English, but at some point in the future some kind of phonological split may emerge. 7. Future directions Intonation’s role in constraining the hearer’s access to particular contexts and in imposing procedural constraints on inferencing needs to be integrated into a wider theory of communication. This paper has argued in favour of a relevance-theoretic account. The mechanisms of procedural coding and the phonological categories to which codes are attached are still imperfectly understood. Furthermore, pitch melody and its organisation is only part of the prosodic aspects of speech; the contributions of rhythm, timing and voice quality all need to be considered by a communication theory. It is the author’s belief that a pragmatic account of intonational meaning must ultimately be founded in and supported by natural speech data. Important insights have come from the study of intonation in spoken corpora (Aijmer, 1996; Couper-Kuhlen and Selting, 1996; Wichmann, 2000; Wennerstrom, 2001), where different discourse situations can be seen to have their own prosodic characteristics. Good descriptive accounts of intonation in use are an essential resource.

J. House / Journal of Pragmatics 38 (2006) 1542–1558

1557

More empirical investigation using natural speech will illuminate the interactive function of intonation, and a better understanding of the process of communication itself. Acknowledgments The author would like to thank two anonymous reviewers for their very helpful comments. References Abercrombie, David, 1967. Elements of General Phonetics. Edinburgh University Press, Edinburgh. Aijmer, Karin, 1996. Conversational Routines in English. Longman, London. Blakemore, Diane, 1987. Semantic Constraints on Relevance. Blackwell, Oxford. Blakemore, Diane, 1992. Understanding Utterances. Blackwell, Oxford. Blakemore, Diane, 2002. Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Markers. Cambridge University Press, Cambridge. Bolinger, Dwight, 1983. The inherent iconism of intonation. In: Haiman, J. (Ed.), Iconicity in Syntax. Benjamins, Amsterdam, pp. 97–109. Bradford, Barbara, 1996. Upspeak. In: Changes in Pronunciation: special issue of Speak Out! IATEFL Newsletter. Brazil, David, 1975. Discourse Intonation. Discourse Analysis Monograph 1. University of Birmingham English Language Research, Birmingham. Brazil, David, 1978. Discourse Intonation II. Discourse Analysis Monograph 2. University of Birmingham English Language Research, Birmingham. Carston, Robyn, 1988. Implicature, explicature and truth-theoretic semantics. In: Kempson, Ruth (Ed.), Mental Representations: The Interface Between Language and Reality. Cambridge University Press, Cambridge, pp. 155–181. Chen, Aoju, 2003. Reaction time as an indicator of discrete intonational contrasts in English. In: Proceedings of the Eurospeech 2003 Geneva. pp. 97–100. Couper-Kuhlen, Elizabeth, Selting, Margret (Eds.), 1996. Prosody in Conversation. Cambridge University Press, Cambridge. Cruttenden, Alan, 1995. Rises in English. In: Windsor Lewis, J. (Ed.), Studies in General and English Phonetics. Routledge, London, pp. 155–173. Cruttenden, Alan, 1997. Intonation, second ed. Cambridge University Press, Cambridge. Fletcher, Janet, Harrington, Jonathan, 2001. High-rising terminals and fall-rise tunes in Australian English. Phonetica 58, 215–229. Fretheim, Thorstein, 2002. Intonation as a constraint on inferential processing. In: Proceedings of the Speech Prosody 2002 Aix-en-Provence. pp. 307–310. Gussenhoven, Carlos, 1984. On the Grammar and Semantics of Sentence Accents. Foris, Dordrecht. Gussenhoven, Carlos, 2002. Intonation and interpretation: phonetics and phonology. In: Proceedings of the Speech Prosody 2002 Aix-en-Provence. pp. 47–57. Gussenhoven, Carlos, 2004. The Phonology of Tone and Intonation. Cambridge University Press, Cambridge. Guy, Gregory, Vonwiller, Julie, 1984. The meaning of intonation in Australian English. Australian Journal of Linguistics 4, 1–17. Halliday, M.A.K., 1967. Intonation and Grammar in British English. Mouton, The Hague. Hirschberg, Julia, Pierrehumbert, Janet, 1986. The intonational structuring of discourse. In: Proceedings of the 24th Annual Meeting of the Association for Computational Linguistics, Columbia, New York, pp. 136–144. Hirst, Daniel, Di Cristo, Albert (Eds.), 1998. Intonation Systems: A Survey of Twenty Languages. Cambridge University Press, Cambridge. House, Jill, 1989. The relevance of intonation? UCL Working Papers in Linguistics 1, 3–17. House, Jill, 1990. Intonation structures and pragmatic interpretation. In: Ramsaran, S. (Ed.), Studies in the Pronunciation of English. Routledge, London, pp. 38–57. Ladd, D. Robert, 1980. The Structure of Intonational Meaning. Indiana University Press, Bloomington, London. Ladd, D. Robert, 1996. Intonational Phonology. Cambridge University Press, Cambridge. Ladd, D. Robert, Morton, Rachel, 1997. The perception of intonational emphasis: continuous or categorical? Journal of Phonetics 25, 313–342.

1558

J. House / Journal of Pragmatics 38 (2006) 1542–1558

Lodge, David, 1991. Paradise News. Secker and Warburg, London. O’Connor, J.D., Arnold, Gordon, 1973. Intonation of Colloquial English, second ed. Longman, London. Ohala, John J., 1983. Cross-language use of pitch: an ethological view. Phonetica 40, 1–18. Pierrehumbert, Janet, Hirschberg, Julia, 1990. The meaning of intonational contours in discourse. In: Cohen, P., Morgan, J., Pollack, M. (Eds.), Intentions in Communication. MIT Press, Cambridge, Mass, pp. 271–311. Shobbrook, Katherine, 2002. Rising intonation in British English: are the high rising tones used for statements and questions phonetically identical? Unpublished MSc. Dissertation, UCL. Shobbrook, Katherine, House, Jill, 2003. High rising tones in Southern British English. In: Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, pp. 1273–1276. Silverman, Kim, 1987. The structure and processing of fundamental frequency contours. Ph.D. thesis, University of Cambridge. Sperber, Dan, Wilson, Deirdre, 1995. Relevance: Communication and Cognition, second ed. Blackwell, Oxford. Steedman, Mark, 2000. Information structure and the syntax-phonology interface. Linguistic Inquiry 31 (4), 649–689. Tench, Paul, 1996. The Intonation Systems of English. Cassell, London. Vandepitte, Sonia, 1989. A pragmatic function of intonation. Lingua 79, 265–297. Wharton, Tim, 2003. Natural pragmatics and natural codes. Mind and Language 18, 447–477. Wennerstrom, Ann, 2001. The Music of Everyday Speech. Oxford University Press, Oxford. Wichmann, Anne, 2000. Intonation in Text and Discourse: Beginnings, Middles and Ends. Longman, London. Wilson, Deirdre, Sperber, Dan, 1993. Linguistic form and relevance. Lingua 90, 1–25. Wilson, Deirdre, Wharton, Tim, 2006. Relevance and prosody 38, 1559–1579. Jill House is a senior lecturer in phonetics at UCL, where she teaches on linguistics, phonetics, English language, and speech and language therapy programmes. Her research interests have centred on prosody, exploring both the formal properties of f0 contours, including their alignment to text, and their function in discourse. She has worked on designing and implementing intonation models for text-to-speech synthesis in dialogue applications (JSRU; Infovox/SUNDIAL), and on the use of synthesis to test the naturalness of a prosodically based model (ProSynth). Current work includes the investigation of the intonation of ‘social rituals’ in naturally occurring speech.