This substantial volume presents 28 papers arising from the 10th Conference of Laboratory Phonology (LabPhon 10), held in Paris in 2006. Though the LabPhon conferences, proceedings and commentaries (e.g. Pierrehumbert et al. Reference Pierrehumbert, Beckman, Ladd, Burton-Roberts, Carr and Docherty2000) render definitions hardly necessary, two papers in the 10th anniversaryFootnote ¹ session offer reflections on and summarise the successes of the movement. Abigail C. Cohn argues that ‘bridging the gap’ between phonological theory and experimental phonetics and understanding the relationship between the cognitive and physical aspects of human speech require movement away from over-simple idealisations (e.g. the ideal speaker-hearer in a homogenous speech community) towards more accurate but complex models taking in grammar, knowledge, acquisition, usage and change, as partially attempted in this and other LabPhon volumes.

Cohn emphasises that laboratory phonology fosters debate that allows for the recognition of approximately correct ideas about speech, rather than polarisation of Model A vs. B. While I think Cohn is approximately right in this, it is important in an experimentally grounded discipline to identify and purge wrong ideas, rather than politely agreeing to disagree. The first section of Janet B. Pierrehumbert & Cynthia G. Clopper's chapter, ‘What is LabPhon? And where is it going?’, asks ‘Who is LabPhon?’, providing a citation network analysis of the publications of 23 ‘heavily cited laboratory phonology authors’ as evidence of the notice taken of laboratory phonology in neighbouring disciplines. It is noteworthy that just over half the papers in this volume are by authors from previous LabPhon conferences; nevertheless, every LabPhon conference has reached out to bring new people into that federation of scholars, and to introduce new themes. The present volume, for example, has a section on disordered speech.

Pierrehumbert & Clopper propose that ‘like all free-trade zones, [LabPhon] has important commodities … The original commodity that made laboratory phonology possible is the attunement of mind to matter’ (p. 116). The value of this commodity is demonstrated by the fact that LabPhon attracts the participation of experimental psychologists, not just linguists. Pierrehumbert & Clopper claim that autosegmental-metrical phonology is a second commodity tying the community together: this could be broadened to prosodic phonology more generally, including articulatory phonology (Browman & Goldstein Reference Browman, Goldstein, Kingston and Beckman1990, etc.) and, for some LabPhon practitioners, Firthian phonology. In all of these flavours of prosodic phonology, LabPhon researchers have developed models of the mapping between abstract phonology and acoustic or articulatory phonetics, such as the speech synthesis proposals of Pierrehumbert & Beckman (Reference Pierrehumbert and Beckman1988: ch. 7), or the metrical analysis of articulatory differences of Beckman & Edwards (Reference Beckman, Edwards and Keating1994), to pick out some instances. I think LabPhon offers two other such ‘commodities’: first, its social network – the meetings and the publications are successful, intellectually rewarding, selective and esteemed, so many people want ‘in’; and second, the importance of fine phonetic detail to improved understandings of phonology, examined in the longest section of this volume but also central to many papers earlier in the series. Pierrehumbert & Clopper propose a new commodity for LabPhon researchers to begin working on: the link between prosody and social structure. In a later chapter, Benjamin Munson, Adriane L. Baylis, Miriam O. Krause & Dongsun Yim remind us that fine phonetic detail is used to convey social-indexical information. As for the three other papers from the anniversary session, Dani Byrd & Susie Choi's chapter advances their model of prosodic structure within articulatory phonology, while Rachid Ridouane's chapter is a classic ‘crossover’ study of various phonetic exponents of gemination configurations in Tashlhiyt Berber, in lexical geminates (even word-initially) and junctural (word-join) geminates, including geminate junctures arising from voicing assimilation. Finally, the chapter by Anne Cutler, Frank Eisner, James M. McQueen & Dennis Norris shows that listeners can rapidly retune their [s]/[f] perceptual boundary when biased with the relevant stimuli; they argue that this demonstrates prelexical abstraction of phonological categories beyond the mere cloud of exemplars supposed in ‘an extreme episodic model’ (p. 99) – a version of Hintzman's (Reference Hintzman1986) ‘multiple-trace’ memory model MINERVA-2, so not just a straw man, but almost: exemplar modellers indeed recognise the need to account for evidence of phonological abstraction, and have made proposals to that effect (e.g. Johnson Reference Johnson, Solé, Beddor and Ohala2007: 38). This chapter exemplifies Cohn's plea for the recognition of approximately correct ideas, rather than disciplinary polarisation.

The four other sections examine ‘variation, phonetic detail and phonological representation’ (p. ix). The papers in Section II are concerned with ‘Variation and language universals’, i.e. cross-linguistic generalisations. Ian Maddieson's paper relates co-variation in phonetic parameters to the ‘reduced’ – ‘clear’ speech continuum (Lindblom Reference Lindblom, Hardcastle and Marchal1990). Alan C. L. Yu's paper examines the relationship between pitch movements and perceived duration. Rina Kreitman proves by example that a putative universal constraint against [+voice][−voice] initial consonant clusters proposed on theoretical grounds by Lindblom (Reference Lindblom and MacNeilage1983) and Lombardi (Reference Lombardi1995), for example, is certainly not universal, as such clusters are attested with phonetic evidence from Khasi and Tsou, to which Kreitman here adds Modern Hebrew.Footnote ² (In Section V, Adamantios I. Gafos, Philip Hoole, Kevin Roon & Chakir Zeroual mention [#bk] and [#bt] in Moroccan Arabic, but it might be retorted that [b] is not contrastively [+voice] in Arabic, as there is no /p/.) In addition to the original papers, full discussion papers (commentaries) are a hallmark of LabPhon proceedings and are vital for stimulating debate and dialogue. Citing further data that supports Kreitman, i.e. [#gk], [#bp] and [#dt] in Begak, Juliette Blevins' commentary attempts to explain the findings of the other papers in this section in terms of her ‘evolutionary phonology’ perspective.

Section III deals with ‘Variation and the emergence of phonology’, opening with a proposal by Paula Fikkert (in the vein of Waterson Reference Waterson1987) that 14-month-old infants have extremely underspecified mental representations of Dutch monosyllables. Fikkert and her co-workers tested predictions such as ‘infants will not perceive the difference between bin and din but will hear the difference between bon and don’ using looking-time data, explaining it in terms of the FUL model of word recognition. As usual in model testing, some of the data is nicely consistent with the presumed model; other data is more challenging. There are puzzles in the account: heard [bon] matches lexical /don/, whereas [don] mismatches lexical /don/, as it is a ‘labial word’ (from its rounded vowel). The child form [dœyt], however, despite also having a rounded vowel, is not a ‘labial word’ (p. 236). In their commentary chapter, Franck Ramus, Sharon Peperkamp, Anne Christophe, Charlotte Jacquemot, Sid Kouider & Emmanuel Dupoux note that since /bin/ is held to have no place features at all, it could be matched by extremely different heard words, such as [gom] (p. 321). Recalling Cohn's advice to recognise what is almost correct in a model, it would be valuable to see how other views of the same data fare; for example, the bon – don F2 difference is usually greater than that of bin – din: could acoustic distances explain the children's behavioural data? Turning to the fact that symbolic models often address the problem of variation by ‘abstracting away from it’, Matthew Goldrick & Meredith Larson's chapter demonstrates an alternative response, deploying a stochastic approach to model phonetic variation during acquisition. Marilyn May Vihman's chapter examines the idea of phonological ‘templates’ in children's lexicons in the single word period, with a meta-analysis based on studies of British and American English, Finnish, French, Italian and Welsh. More cross-linguistic studies of this sort would be welcome, in order to avoid the pitfall of framing overly general observations on the basis of one-language studies. Ramus et al. comment that a recurrent difficulty of developmental and other psycholinguistic studies is a tendency to focus on just one or two levels of representation; they remind us that many interacting levels of representation are implicated in linguistic processing, and they advertise the merit of a rather more complex information-processing model of speech perception and production. Though explicit models are to be welcomed, Ramus et al.'s boxes-and-arrows/levels-of-processing model seems implausible in some respects. They attempt a reconciliation of the evidence for episodic memory with an abstract lexicon by putting each in its own box and drawing arrows between them; the more common idea coming out of exemplar-based modelling sees the lexicon as grounded in episodic memories, with abstraction derived via clustering and categorisation/labelling of clusters, for example.

Though the papers in this section deal with the problem that phonetic variation presents to children, it is striking how little attention is given to specific phonetic details of the children's productions, or of their adult models, or the stimuli used in perception experiments. For example, in Fikkert's presentation of the puzzle-puzzle (p. 229), it would be informative to examine the acoustic differences between [dl] and [gl] (which are in fact very subtle, as Kawasaki-Fukumori Reference Kawasaki-Fukumori1992 noticed). Equally, examination of the articulation of the [d] in the adult model ‘puddle’ (which Amahl pronounces as velar) may show it to be laminal or palatal, not a canonical alveolar stop (Schmidt Reference Schmidt1888: 64); a more posterior, laminal or possibly palatal stop might reasonably be perceived as dorsal, and thus inferred to be velar. Is the ‘puzzle’ an illusion arising from excessively broad transcriptions, i.e. the ‘measurement’ weaknesses of impressionistic transcriptions?

Section IV, ‘Variation at the crossroad between normal and ‘disordered’ speech', opens with Grzegorz Dogil's ‘Hard-wired phonology: limits and latitude of phonological variation in pathological speech’, which sweeps from neurological data based on impairment and neuroimaging to recently proposed integrated models of speech production (partly meeting Ramus et al.'s plea). Dogil cites Guenther et al.'s (Reference Guenther, Gosh and Tourville2006) DIVA model of speech production, possibly the most important new phonetic model of the decade, the influence of which should surely be felt in future LabPhon papers. Munson et al.'s chapter shows how impaired perception can lead to impairments in phonological development and thence to inaccurate speech production. A question of observational accuracy and detail recurs here: is the [t] of [dʌt] ‘duck’ apical or laminal? Munson et al. note earlier findings that some cases of such mispronunciations involved a merged lingual gesture intermediate between alveolar and velar places of articulation. We might ask again whether impressionistic assessment of articulatory accuracy misses errors in phonetic variation by too gross, phonemic recording. Timothy Arbisi-Kelm's chapter investigates prosodic factors in stuttering, finding that stutterers seem to be aware of prosodic ‘breakdowns’ well before articulation of the problematic stretch is reached, supporting a model of prosody generation in which large prosodic structures (e.g. Intonational Phrases) are available before the phonological content is fully retrieved. Striking agreement with this is found in Karen Croot, Claudia Au & Amy Harper's chapter on prosodic factors in tongue twisters. Ray D. Kent's commentary paper looks at these papers from the context of a selection of models of serial order in speech production, again approving of the DIVA model as well as competitive-queueing models of movement planning, which may account for errors of serial ordering in speech production by proposing that ‘before a serial act is initiated, there is an active parallel representation of all components’ (p. 466); if the components are not properly inhibited or deleted in their due turn, an error occurs.

The final section V, ‘Phonetic detail, processes and representation’, runs to ten chapters and 310 pages. If LabPhon offers prosodic structure as a ‘commodity’ for understanding phonetic variation, close observational studies of fine phonetic differences are the currency that pays for it. Sarah Hawkins opens with a wide-ranging position paper emphasising the coupling of fine phonetic detail to prosodic structure, its relevance to speech perception, and grounding in neurophysiology. Mark Pluymaekers, Mirjam Ernestus, R. Harald Baayen and Geert Booij present a corpus-based study of reduction patterns in the Dutch suffix -igheid, relating variation in reduction to morphological factors: structure and informativeness. Interestingly, a prosodic structure hypothesis is unexplanatory here. To segment the numerous (432) occurrences of -igheid in a 100-hour corpus, the authors used a Hidden Markov Model speech recogniser, an increasingly common tool in phonetics data processing nowadays. Tamara Rathke & Jonathan Harrington's chapter is a standard lab study of H+!H* and H+L* intonation accents in Standard German, arguing that the difference between these two types is phonetic, not phonological, and that H+!H* is not an appropriate accent category for German. With some similarities to Maddieson's and Pluymaekers et al.'s papers, Rebecca Scarborough's chapter relates details of vowel quality to their predictability from the sentence context. She found that vowels in ‘hard’ words (high neighbourhood density) and in non-predictable contexts were longer and more dispersed (somewhat more peripheral) than ‘easy’ words in predictable contexts, though the effect of neighbourhood density greatly outweighs that of contextual predictability. Edward Flemming's commentary on Pluymaekers et al.'s and Scarborough's papers advances Bayesian interpretations of their findings. Whereas Yu's chapter examined the effect of tone on vowel duration, Maria-Josep Solé & John J. Ohala examine whether the durational differences found between different vowel heights are under the speaker's control. Finding different results in different languages, they argue that in Catalan and American English, vowel duration is actively controlled by speakers as a cue to vowel height, whereas in Japanese the differences between high and low vowel duration can be explained by biomechanical factors alone. This is reminiscent of Chen's (Reference Chen1970) finding that vowel durations before voiced vs. voiceless consonants differ by at least 10% cross-linguistically, but that some languages exaggerate this difference.

Gafos et al.'s chapter examines articulatory overlap in consonant clusters in Moroccan Arabic, comparing variation in overlap timing that is grammatically conditioned, i.e. by OCP (phonology) or morphological affixation, with non-grammatical conditioning, e.g. the position of the cluster in the word. In the former type, they see evidence that aspects of phonetic timing have the same cognitive status as grammatical principles. Jeff Mielke, Adam Baker & Diana Archangeli look at variation in the articulation of American English /ɹ/ and its effect on preceding /s/, using ultrasound imaging and acoustic measurements. They find a number of distinct, speaker-specific patterns, with very complex conditioning environments. Claudia Kuzla, Mirjam Ernestus & Holger Mitterer's chapter examines the perception of devoiced fricatives in various prosodic environments in German. Jean-Luc Schwartz responds to Solé & Ohala, Kuzla et al., Mielke et al. and Gafos et al., in each case asking ‘what does the listener know about a speaker's gesture?’. Rather than attempting to summarise several complex and detailed arguments here, I conclude by noting how valuable it is to have such incisive and thoughtful critical commentaries to accompany this rich collection of scientific papers.

Has LabPhon been a success? Only partly, I think. ‘Phonology’ – the discipline and the journal of the same name – continues to be theory-heavy, abstract and sometimes fairly light on empirical evidence. Publications devoted to experimental phonetics, such as the Journal of Phonetics, carry very little ‘straight’ phonological theory. But there have been undeniable sociological shifts: as Cohn's chapter notes, some linguistics job advertisements now list laboratory phonology as a subfield, and some graduate students and a few established academics state laboratory phonology as their specialisation. Though a number of colleagues who once might have been described as theoreticians have become regular participants in LabPhon, one rarely if ever sees experimentalists ‘defecting’ to abstract theory. To some extent, the LabPhon programme is emulated in other areas of linguistics, e.g. the ‘spoken syntax’ subfield. It seems entirely possible now (and would not have seemed so in 1987) that LabPhon's matchmaking of mind and matter will come to appear quaint. As a psychiatrist I met at a party recently commented, ‘well, we don't believe in that kind of dualism any more: mind is matter’.