This is a very useful contribution to the literature on intonation and prosody, though possibly not in quite the way it was intended. As a scholarly book, it has all the attributes one would hope for: careful editing and proofreading, a decent index, and extras like a CD with sound files of all the examples from each chapter. As a resource for anyone wanting clear and authoritative summaries of the way the autosegmental-metrical (AM) theory of prosody has been applied to more than a dozen languages, it is without equal. In this role as a reference work, the book should have a long shelf life, and despite my considerable delay in writing this review, I can still strongly recommend to readers that they (or their institution's library) should add this title to their collection. However, I think the book is less successful in its announced aim of making a contribution to prosodic typology, because a secondary aim – promoting the development of ToBI transcription systems – gets in the way. I will devote the second part of the review to supporting this latter judgement.

But first a summary of the book, and of the good things about it. The book had its origins in a satellite workshop (‘Intonation: Models and Transcription’) held in conjunction with the 14th International Congress of Phonetic Sciences in San Francisco in 1999. It consists of sixteen chapters: a brief introduction (Ch. 1) and a somewhat longer concluding synthesis (Ch. 16) by Jun as editor; a historical and theoretical chapter (Ch. 2) on the development of the ToBI framework by Mary E. Beckman, Julia Hirschberg & Stefanie Shattuck-Hufnagel; and in between, thirteen chapters on as many different languages. The authors of the language chapters are in almost all cases the obvious experts on the AM description of the language in question – the people you would have sought out in the corridors of the Cathedral Hill Hotel in San Francisco that summer if you'd had a question about prosody in one of these languages. They are: Martine Grice, Stefan Baumann & Ralf Benzmüller on German (Ch. 3), Amalia Arvaniti & Mary Baltazani on Greek (Ch. 4), Carlos Gussenhoven on Dutch (Ch. 5), Svetlana Godjevac on Serbo-Croatian (Ch. 6), Jennifer J. Venditti on Japanese (Ch. 7), Jun herself on Korean (Ch. 8), Shu-hui Peng, Marjorie K. M. Chan, Chiu-yu Tseng, Tsan Huang, Ok Joo Lee & Mary E. Beckman on Mandarin (Ch. 9), Wai Yi P. Wong, Marjorie K. M. Chan & Mary E. Beckman on Cantonese (Ch. 10), Matthew K. Gordon on Chickasaw (Ch. 11), Judith Bishop & Janet Fletcher on Bininj Gun-Wok (Ch. 12), Martine Grice, Mariapaola D'Imperio, Michelina Savino & Cinzia Avesani on Italian (Ch. 13), Janet Fletcher, Esther Grabe & Paul Warren on high rising tunes in English (Ch. 14) and Gösta Bruce on Swedish (Ch. 15). For anyone not implacably opposed to the AM approach, the list of contributors alone gives the book prima facie authority.

Moreover, the integration and coherence of the book is unusual for an edited volume. A substantial majority of the chapters (the ones on German, Greek, Dutch, Serbo-Croatian, Japanese, Korean, Chickasaw and Bininj Gun-Wok) present details of a well worked-out AM transcription system for the language they discuss. Of the others, three (Bruce on Swedish and Fletcher et al. on English high rises, and to some extent Grice et al. on Italian) are more concerned with specific issues that arise in transcribing different varieties of the language they cover, while the two chapters on Mandarin and Cantonese deal with broader issues that arise in talking about the sentence-level prosodic features of languages that have lexical tonal specifications on almost every syllable. But even these other chapters are clearly integrated into the book's demonstration of what happens when the AM approach to intonation is applied to different languages. It seems clear that the authors have really been listening to one another and engaging in what they see as a common enterprise; among other things, the quantity of useful cross-references from chapter to chapter attests to this mutual attention.

This attitude on the part of the authors must have made the editor's task easier, but there is plenty of evidence of her work beyond simply getting a group of enthusiastic participants together. For example, she encouraged the contributors to provide summary lists of the prosodic labels used in each of the systems described, which greatly ease the reader's task in consulting the book for comparative detail. More generally, the careful editing and proofreading that is apparent throughout is no mean feat in a book that deals with a lot of different languages and that depends on accurate reproduction of transcriptions of prosodic detail. (One curious lapse in this respect is the form of the abbreviation for ‘fundamental frequency’: in most of the chapters this is Fo, but the editor herself in her chapters uses fo, and Godjevac's chapter uses F_o.) While I have no ready way of checking up on the transcriptions of, say, Chickasaw or Serbo-Croatian on my own, the overall absence of errors in the material I am in a position to evaluate suggests that what you see is what you were meant to get. So I repeat my recommendation that this is a book that any serious university library and any serious prosody researcher should own.

Now let me try to explain why I think the book, despite its conspicuous virtues, doesn't really tell us very much about prosodic typology beyond providing some pretty solid empirical documentation. The problem is that in order to do typology, you have to have a set of agreed descriptions cast in comparable terms. That kind of consensus is still lacking in the description of prosody. The broad AM approach is certainly leading us toward such a consensus, but we're not there yet – it's only the practical and collegial cohesiveness of the ToBI movement that makes the progress seem greater than it is. Eventually we will have to confront the disagreements more explicitly.

This was the overarching theoretical point of my book Intonational phonology (1996). I argued that it is important to distinguish the fundamental theoretical tenets of the overall AM approach to intonation from the specific descriptive claims of Pierrehumbert's analysis of English intonation (e.g. Pierrehumbert Reference Pierrehumbert1980, Beckman & Pierrehumbert Reference Beckman and Pierrehumbert1986). I discussed a variety of proposals and analyses of English that disagree with Pierrehumbert but are still, in my view, based solidly on AM principles and ideas, and I argued that we should be considering these alternatives to Pierrehumbert's specific descriptive proposals as a way of refining our theoretical understanding. But this has not really happened. Instead, Pierrehumbert's analysis, metamorphosed into the system now known as MAE_ToBI (‘Mainstream American English ToBI’), has become both the established analysis of (American) English and the model for analyses of other languages, and ToBI has thus become a kind of theory-within-a-theory. Jun seems to acknowledge this when she describes the ToDI system for Dutch intonation as ‘based on Gussenhoven's model, another AM model of intonational phonology’ (p. 4; emphasis added).

This is an unfortunate development. If Gussenhoven's work can be treated as part of a different ‘model’, it is easier to ignore his arguments about how to analyse nuclear accent patterns than if we acknowledge that Gussenhoven and Pierrehumbert share the same general theoretical outlook and force ourselves to decide whose analysis is superior. By the same token, if we regard self-identified ToBI descriptions as all being based on the same ‘model’, we play down the significance of disagreements over Pierrehumbert's analysis of English, and we exaggerate the cross-linguistic comparability of analyses bearing the ToBI stamp. That is, when the ToBI description of Language L differs from Gussenhoven's description of Dutch in some respect, we do not automatically assume that Dutch has typological property x and Language L has typological property y; rather, we will be alert to the possibility that the languages are not being described in the same terms. But the idea that the descriptions in all the other chapters are ‘ToBI’ descriptions lulls us into thinking that they are based on approximate agreement on terminology and analysis. Ultimately, this may create even more problems for insightful typology.

To see why, consider the concrete example of what I'll refer to as ‘sustained level phrase-final pitch’ (SLPFP), which is found in a variety of contexts in a variety of languages. In English, one common context for SLPFP (and the easiest one to illustrate on paper) is the ‘calling contour’ (cf. Ladd Reference Ladd1978), stereotypically used to call children from a distance, in which the stressed syllable is produced on a sustained (almost chanted) high level pitch and the remaining syllables are produced on another sustained level that is often three semitones lower.

1. (1) Ma–   ry–

Pierrehumbert's original analysis of SLPFP, which is still the official MAE_ToBI analysis, treats it as the phonetic realisation of a downstepped high phrase accent on -ry followed by an ‘upstepped’ low boundary tone. High phrase accent is also said to upstep a following high boundary tone, yielding the continuous sentence-final rise at the end of many questions. This analysis works, in the sense that it can be made to generate the contrast between rising and sustained level phrase-final pitch. However (as extensively discussed in Ladd Reference Ladd1996: ch. 4, 2008: ch. 3), it leaves no obvious way of describing numerous British English intonation patterns in which the pitch gradually falls or ‘slumps’ after a nuclear accentual rise. Moreover, it requires a fairly abstract notion of the relation between phonological categories and phonetic realisation that not everyone feels is justified. Consequently, other analyses of SLPFP have been proposed. For example, Gussenhoven (p. 124) analyses SLPFP in Dutch as involving the absence of any boundary tone, noting matter-of-factly that ‘the advantages of making T% optional were convincingly argued for by Grabe (Reference Grabe1998) for German and English’. Grice et al.'s analysis of SLPFP in German is more like Pierrehumbert's, in that the contrast between rising and sustained level phrase-final pitch is treated as a difference between different boundary tones, but their distinction is between upstepped and non-upstepped high, rather than between upstepped high and upstepped low. And so on.

Yet none of these alternative proposals have made us go back and reconsider the MAE_ToBI analysis. To some extent this is because some of them can be pushed off to one side as part of ‘another AM model’, but increasingly the central motivation for not considering the mutual implications of competing analyses is the principle that each language has to be analysed in terms of its own contrasts. ToBI's theoreticians clearly favour the view that different analyses reflect genuine differences between languages.

This case is set forth at some length in the introductory historical and theoretical overview chapter by Beckman et al. They emphasise that ToBI is not a phonetic transcription system like the IPA, but rather a framework for the phonological analysis of each language. If SLPFP is transcribed H-L% in English and H-% in German and 0% in Dutch or Chickasaw, we should assume that that is because phonologically it isn't the same thing in those languages. As Beckman et al. put it (p. 42):

Now, this is clearly an intellectually respectable position – a radically Saussurean interpretation of language structure, so to speak, whereby what counts is not the physical manifestation of signs but the network of oppositions between signs in a specific language. But if taken seriously, it casts doubt on any sort of generalisation across languages or even varieties of the same language. With respect to typology generally, this point of view is the basis of Haspelmath's recent argument (Reference Haspelmath2007) that rigorous typological comparison cannot be based on the same categories used for the description of individual languages. With respect to ToBI, it's the basis of the progressive narrowing of the claimed applicability of the original ToBI system: in the early 1990s it was said to cover General American, Southern British and Australian, but today it only claims to be about ‘Mainstream American English’. With respect to the specific extended example I have been discussing, the each-language-in-its-own-terms principle provides a clear basis for treating Gussenhoven's analysis of SLPFP in Dutch or the existence of ‘slumping’ contours in British English as irrelevant to Pierrehumbert's analysis of SLPFP in Mainstream American English. The heart of the issue is whether there is any basis for identifying SLPFP as a cross-linguistically comparable phenomenon to be transcribed in comparable ways.

If we are really going to undertake linguistic typology, I believe we need to look for inspiration in the arguments outlined in Newmeyer's reply (Reference Newmeyer2007) to Haspelmath, or in Huddleston & Pullum's assumption (Reference Huddleston and Pullum2002: 31) that it is both appropriate and possible to distinguish between ‘general’ and ‘language-particular’ definitions of grammatical terms. With specific reference to SLPFP, I assume with Gussenhoven that SLPFP is a single identifiable phenomenon that is broadly comparable at least in all the West Germanic languages, and that it is legitimate and meaningful to argue about how best to analyse it in the framework of AM assumptions. I recognise the profundity of the theoretical issues raised by Beckman et al. with regard to any kind of generalisation across languages, and I'm perfectly willing to accept, as suggested by Arvaniti & Baltazani in their analysis of SLPFP in Greek, that there may actually be differences between languages in this respect as in any other. But my point here is not to try to resolve this specific issue in a short book review. It's just to say that ToBI can't have it both ways. If the analyses on which the transcription systems are based are truly language-specific (or indeed, variety-specific), then they are strictly speaking incommensurate, and typological generalisations are at best difficult and at worst useless. If on the other hand typological comparison is appropriate, then we can't avoid talking about things like SLPFP and deciding whether we want our typology to be able to treat them as some sort of cross-linguistically identifiable phenomenon (like adjective, subject or sibilant in other areas of linguistic typology). And if we decide that they are cross-linguistically identifiable, then we need what I named above as the fundamental prerequisite for typological comparison, namely ‘a set of agreed descriptions cast in comparable terms’. It won't do in the long run if you call something an upstepped low boundary tone and I say there's no boundary tone there at all.

That's why I think ToBI-identified researchers should do some serious theoretical thinking about what it actually means to have a common ‘general framework for the development of prosodic annotation systems’ (Beckman et al., p. 9).They should acknowledge that there are some serious disagreements about principles of transcription that are not just disagreements over the analysis of a specific contrast in a specific language variety. Until they do, a book like the one under review, however carefully conceived and edited and produced, makes only a limited contribution to the study of prosodic typology.