EDITORS: van de Weijer, Jeroen; Nanjo, Kensuke; Nishihara, Tetsuo TITLE: Voicing in Japanese SERIES: Studies in Generative Grammar 84 PUBLISHER: Mouton de Gruyter YEAR: 2005
Mark Irwin, Faculty of Literature & Social Sciences, Yamagata University
'This book presents a number of studies which focus on the [voice] grammar of Japanese, paying particular attention to historical background, dialectal diversity, phonetic experiment and phonological analysis' (p. 1), with the initial version of some of the papers having being presented at the Linguistics and Phonetics 2002 workshop at Meikai University, Urayasu, Japan. Of the 14 papers in this volume, ten deal with consonant (Part I) and four with vowel voicing (Part II), with seven of the contributions to Part I dealing wholly or largely with the well-documented allomorphic phenomenon of sequential voicing, or 'rendaku' in the Japanese tradition (by which term the phenomenon will be referred to henceforth). Perspectives within a variety of theoretical frameworks are offered.
Part I of the volume opens with Kubozono's review (pp. 5-24) of past work on rendaku. Within the space constraints available, a comprehensive summary is necessarily impossible and the author confines himself to a survey of the conditions which define the phenomenon's synchronic domain. He deals first with 'Lyman's Law', treating it as an example of the Obligatory Contour Principle and arguing that its standard domain requires widening in some instances. A second condition is the branching constraint, whose status Kubozono questions for rendaku while offering supporting evidence for this constraint in general from accentual phrasing. The final condition examined is that of the mora, the most recent and perhaps the weakest constraint on rendaku, where the author seeks to integrate the behaviour of one specific morpheme, hon 'book', into a wider framework in which it is claimed the prosodic word in Japanese is optimally up to four moras in length.
Couched within the representational framework, Rice (pp. 25-45) argues that the voicing features active in rendaku and post-nasal voicing should not be considered as one single type. Instead, she posits a 'dual mechanism' hypothesis, under which a laryngeal voicing feature is identified with rendaku and a sonorant voicing feature with post-nasal voicing. The author also considers problems highlighted by her hypothesis which are inherent in recently proposed lexical stratification models for Japanese.
Ohno (pp. 47-69) traces the historical orthography of what is known in the Japanese tradition as the sei-daku distinction and which, in the contemporary language, marks the voiced-voiceless feature in obstruents. While 'provid[ing] neither new data nor new findings' (p. 47), he outlines the three historical stages of sei-daku orthography and reviews the two main arguments which seek to explain them. The author closes with brief summaries of the sei-daku distinction in the modern language and the important issue of whether it may have been one of prenasalization rather than voicing in the past.
Within the framework of Element Theory, Nasukawa (pp. 71-87) presents an analysis of laryngeal source contrasts. Of the two autonomous melodic categories available for cross-linguistic source contrasts under this theory, the author claims that Japanese exploits the category contributing prevoicing in representation of phonation- type contrasts. Support for this claim is offered from assimilatory and concatenating processes, early language acquisition and aphasic deficit.
Vance (pp. 89-103) provides a statistical analysis of rendaku incidence in inflected words, i.e. verb+verb compounds and compounds containing adjectives. Contrary to the established opinion that verb+verb compound verbs tend away from and verb+verb compound nouns tend towards rendaku, he reports that, in fact, in the vast majority of cases rendaku does not occur in either. He also finds that compounds containing adjectives tend overwhelmingly towards exhibiting rendaku.
Fukazawa & Kitahara (pp. 105-121) re-examine work on Japanese core-periphery vocabulary models (specifically Ito & Mester (1995, 2001)) developed to counter the problem of single invariant ranking in Optimality Theory (cf. Rice in this volume, who re-evaluates essentially the same body of work). They present three ranking paradoxes in consonant voicing which they claim cannot be accounted for in these models, arguing that etymologically motivated sub-lexica cannot exist in Japanese phonological grammar and that these must be replaced with sub-lexica based on standard morphophonological categories.
Yamane-Tanaka (pp. 123-156) draws parallels between synchronic cross-dialectal variation in intervocalic prenasalized stops and their probable diachronic development since the Old Japanese period. After presenting an Optimality Theory analysis of these two continua, she tackles the issue of prenasalization versus voicing in Old Japanese and beyond (i.e. the historical nature of the sei-daku distinction: see Ohno in this volume) and suggests that a gradual loss of prenasalization has gradually led to an increased role for voicing contrast.
Zamma (pp. 157-176) undertakes a detailed examination of the complicated rendaku patterning seen in Japanese surnames. He first discusses the generalization put forward by Sugito (1965), that accented surnames tend to undergo rendaku while accentless surnames do not, as well as the suggestion made by Kubozono (this volume) as to the influence of the onset of the final mora of the preceding morpheme. He then moves on to examine these general claims within the context of a much wider database of surname head morphemes and concludes that their widely varying behaviour presents considerable theoretical problems.
A survey of rendaku incidence in loanwords is offered by Takayama (pp. 177-190), who argues that those few loanwords from the foreign word group (the author eschews 'stratum') that exhibit rendaku do so only because they have merged into non-foreign word groups due to their phonotactics or their semantics (they represent 'an unsophisticated object'). Takayama further claims that the larger number of Sino-Japanese lexemes that exhibit rendaku can be explained by setting up a 'vulgarized SJ' word group, which can be differentiated stylistically from a 'formal SJ' word group and whose members may be rendaku targets.
Part I of the volume closes with Suzuki's (pp. 191-204) presentation of the results of his study into the problems of the automated speech recognition of Japanese numeral-classifier compounds. Many classifiers beginning in a voiceless obstruent (e.g. hon, cylinder- shaped object) exhibit context-dependent (morphophonemic) voicing after certain numerals (e.g. san.bon, '3...'), as well as context- independent (free) variation (e.g. ichi ~ hito, 'one'), and both these present performance problems for the author's Large-Vocabulary Continuous Speech Recognition engine. The solution and the test results that Suzuki puts forward show that the engine's performance can be improved by making probability adjustments in order to cover unseen data in the corpus.
Part II is opened by Maekawa & Kikuchi (pp. 205-228), who present an interim report on their corpus-based analysis of vowel devoicing in spontaneous speech, which utilizes some 23 hours of the Corpus of Spontaneous Spoken Japanese containing approximately 427,000 vowel segments. Although, as the authors point out, it is often stated in introductory texts that high vowels are devoiced between two voiceless consonants, previous research has indicated that this is not the case. Maekawa & Kikuchi's results are in broad agreement with this research and show that such devoicing is not exceptionless in this environment and that its rate varies considerably according to the manner of the surrounding consonants. The authors also present their statistical findings for the phenomenon of consecutive devoicing, as well as for single devoicing in two atypical environments.
Kondo (pp. 229-245) investigates the duration and intensity of vowels in the standard devoicing environment between two voiceless consonants. Her experiment on duration shows that those vowels which become devoiced are also shorter than their voiced counterparts, the entire duration of the mora of which the vowel is part also being reduced. Furthermore, her experiment on intensity indicates that a vowel in a single devoicing environment which remains voiced has a lower intensity than a vowel in a non-devoicing environment: the same was not the case, however, for vowels in a consecutive devoicing environment. From this she concludes that devoiceable vowels are first reduced in length and intensity before being devoiced. Kondo also examines the syllable constraints on devoicing, in particular how desyllabification, demoraification and resyllabification may control which vowel is devoiced in cases of potential (but banned) triple consecutive devoicing.
Sugito (pp. 247-260) presents the results of acoustic and physiological experiments on the effects of speech rate on devoiced accented vowels in Osaka Japanese, whose accentual system is radically different from that of standard Tokyo Japanese and where vowels may be both devoiced and accented. In her acoustic experiments with bimoraic HL-accented lexemes, she shows that speech rate affects both vowel voicing and accentedness, with devoiceable accented vowels occurring more frequently devoiced in faster speech and, additionally, a tendency for the accentual pattern to be perceived as shifting from HL to HH in the fastest speech. The results of her physiological experiments suggest that this change in accentual pattern may be due to the short duration of the second vowel in the lexemes in question, whereby there was insufficient time for sternohyoid muscle activity and thus no F0 fall, the perceptual trigger for accentedness.
The final paper in the volume is that of Tanaka (pp. 261-278), who examines the interaction of voicing and accent, specifically the outcome of the 'in principle' incompatible situation of an accent- bearing vowel being in a devoiceable position. This interaction is investigated through reconsidering it in the wider perspective of a general theory of prominence involving tone, length, sonority, accent and voicing. From this may be derived the phonological Harmonic Scale of Prominence which posits the order of prominence accent > tone > sonority > voicing and where, furthermore, an element always presupposes the existence of any element to its right (thus accent presupposes tone, sonority and voicing). After examining the interaction of accent with tone and of accent with sonority, the author proceeds to that of accent with voicing, where he views accent shift as a 'repair strategy for upholding harmonic completeness and prominence'. However, Tanaka claims to have solved the problems posed by the numerous examples of the accent not shifting and of the vowel being both devoiced and accented by resorting to the notions of sympathy and reranking within the Optimality Theory framework.
This collection of papers deserves a slot on the bookshelf of any scholar of the Japanese language, particularly those whose research focuses more specifically on phonology or phonetics, or to a lesser extent the lexicon, orthographical history or automated speech recognition. Those with no background in Japanese but with a strong cross-linguistic interest in consonant voicing or vowel devoicing should also find much of interest in this volume. What is particularly beneficial about this collection is its undoubted usefulness as a portal for the reader who is unable to access or understand the vast mines of previous research material in Japanese on many of the subjects tackled.
The editors deserve much credit for bringing together a stimulating and valuable selection of papers covering the many aspects of voicing in Japanese (rendaku; post-nasal voicing; the diachronic relationship between voicing and prenasalization in consonants; morphophonemic voicing; the orthographical representation of voicing; vowel devoicing and its relationship to accent, environment and intensity) from a wide range of theoretical (representational, Element Theory, Optimality Theory) and non-theoretical perspectives. Rendaku itself, with which half the papers in this volume concern themselves, is examined with reference to post-nasal voicing, inflected words, core-periphery vocabulary models and vocabulary strata, anthroponyms, accent, and laryngeal features.
In a volume that is part of a generative grammar series it is perhaps unsurprising that the vast majority of papers deal with voicing from a synchronic perspective, with only the contributions from Ohno, Yamane-Tanaka and, to a certain extent, Takayama, offering the reader a diachronic take on voicing and its historical development. Since, as Blevins (2004: 3) has recently remarked, an 'essentially ahistorical perspective leads to considerable redundancy' because 'many patterns with a well-understood historical basis or origin must be re-encoded in synchronic accounts', the relevant dearth of diachronic analyses in this collection and its rather overly ahistorical perspective is perhaps somewhat regrettable. This is especially so with Part II of the volume, where, to the reviewer's knowledge, relatively little diachronic research has been published on the phenomenon of vowel devoicing and where the inclusion of a least one paper examining this from a historical perspective would have been useful.
Inevitably, in a volume with so many individual contributors, there are areas where more judicial editing would have been preferred. Although just under half of the papers are wonderfully free from any typos whatsoever, and most of the remainder have only a small number of minor errors, statistical discrepancies in tables and potentially serious typos in examples and maps lurk in the papers by Yamane-Tanaka and Maekawa & Kikuchi. A further two papers, those by Fukuzawa & Kitahara and Yamane-Tanaka, are clearly in need of native speaker editing. In the latter paper especially, the author's argument is occasionally difficult to follow due to some weak English and insufficient proofing. Yamane-Tanaka's paper also unfortunately suffers from at least six endnotes being incorrectly referenced, as well as the errors in tables and maps already pointed out above. This is a shame, since, for the reviewer, this paper was one of the most stimulating and original in the collection. On the more mundane level, the 58 endnotes in Ohno's 15-page (omitting appendices, acknowledgements and the endnotes themselves) paper is, by any standard, rather excessive: do we really need to be told in an endnote that 'lit.' is an abbreviation for 'literal translation'?
On balance, however, the rich and varied content this volume offers, as well as the stimulation and leads for further research it provides, more than outweigh the relatively minor gripes I have outlined in the previous two paragraphs.
Ito, Junko & Mester, Ralf-Armin (1995). Japanese Phonology. In John Goldsmith (ed.), Handbook of Phonological Theory, pp. 817-838, Blackwell, Cambridge.
Ito, Junko & Mester, Ralf-Armin (2001). Covert generalizations in Optimality Theory: the role of stratal faithfulness constraints. In Proceedings of 2001 International Conference on Phonology and Morphology, 3-33, Yongin, Korea.
Sugito, Miyoko (1965). Shibata-san to Imada-san: tango no chookakuteki benbetsu ni tsuite no ichi koosatsu, Gengo Seikatsu 165: 64-72.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Mark Irwin is an associate professor at Yamagata University, Japan. His research interests include the historical phonology of Japanese, especially rendaku, gemination, Sino-Japanese and Japanese sinography.