LlNCOM Studies in Phonetics

full text research abstracts of all titles monthly updates LlNCOM webshop

www.lincom·europa.com

Quantitative approaches to problems in linguistics

Studies in honour of Phil Rose

Cathryn Donohue, Shunichi Ishihara,

William Steed (editors)

2012 LlNCOM EUROPA

Published by LlNCOM GmbH 2012.

LlNCOMGmbH Gmunder Str. 35 D-81379 Muenchen

LlNCOM.EUROPA@t-online.de www.lincom-europa.eu

webshop:www. lincom-shop.eu

All rights reserved , including the rights of translation into any foreign language. No part of this book may be reproduced in any way without the permission of the publisher.

Printed in E.C. Printed on chlorine-free paper

Die Deutsche Bibliothek - CIP Cataloguing-in-Publication-Data

A catalogue record for this publication is available from Die Deutsche Bibliothek (http://www.ddb.de)

LlNCOM EUROPAacademic publications are a productofLiNCOM GmbH.

Contents

List of Contributors ........................................ .. _ ............................................... ........ ... ............ ···· 111

Phil Rose: A short biography ...... . ...... --------- .................................... ···········································v A NN K UMAR, CATHRYN D ONOHUE, SIIUNICII I ISHII lARA

MOllY vo ices, many tones ... -............. , , ............ .......................... ........ . . ........ ··········· 1

CATIIRYN D ON01IUE, S I-I UNICI-II iSI-II II ARA, WILLI AM STEED

Part I. Tones and Acoustic Phonetics

The shape Hncl sp read of lonc .......... . M ARK D ONOHUE

Qinglian Wu lexica l tOile sandhi : Voice less depressors and allatones .... ································· 21 WI LLI AM Sn : ED

The rea lisa ti on of the stopped tone in North-Centra l Victnamcsc ................ . K OIC H1 H ONDA

Tone altemat ions in Ugong (Tha iland) .......... . . ................................... ·· 55 D AVID BRADLEY

The role of contour and phonation in Fuzhou tonal iden tifi ca tion ......................... .. 63 CATHRYN D ONOHUE

OsakH nnd Kagoshima Japanese citation tOile acoust ics: A linguist ic tonctic cOlllpara li ve study ... .................... ........... .. ... ................ ..................................... ...... ------ ----. --------- -77

SI-IUNICl-l l l s HIIIARA

Off-thc-Chart vowc l changes in Chinese -................................... ............................. ..... .... 103

XIAONONG Z IIU

The critical period hypot hes is and phonological acqui sition of Japancsc TAKAKa T ODA

··· 123

On the phonetics o f long, thin phonologies ..................................................................... ··· ...... 133

ANDREW B UTCHER

Part II. Forensic voice comparison

Lincar-sca li ng cffccts of phonet ic contex t 0 11 vowc l form fUlts: A tutoria l --· - -- -- ............ · .. · 155 FRANTZ CLERMONT

Japanese formant frequencies in mobi le phone transmission: Implica ti ons for Forensic Voiee Comparison ............. ..... .. ......................................... .... . . ....... .. ·~· 171

MICHAEL CARNE

"

Fine-grained automatic speaker recognition lIsing ccpslrai-trajcclori cs in phone units ........ 185 JAVIER FRANCO-PEDROSO, JOAQUIN GONZALEZ-RODRIGUEZ, JAVIER GONZALEZ-D OMINGUEZ,

DANIEL RAMOS

Automatic speaker identification using the magnitude and phase spectra of inverse-fi ltered voiced speech ........................................................................................... ································· 197

MICHAEL WAGNER

Forensic voice comparison using Chinese Ijau/ ····································· ······ ········· ···················207 CUlLING ZHANG, G EOFFREY MORRISON, THARMARAJAH THIRUVARAN

Bayes and beyond: The complex challenges of LA DO and the ir relevance to forensic speaker com parison ................. ................... ........................................................ ······· 215

HELEN FRASER

Part III. Bayes and beyond

ua and variation in expression ................. .......... ......... ... ...................................... ···················223

A VERY D . ANDREWS 3RD

' Humble aux iliarics' in Old Japanese: Javanese derivations, context, and sign ificance ······· 241 ANN KUMAR

Eating and drinking in Mandarin and Shanghainese: A Lexical-Conceptua l analys is···········265 ZHENGDAO YE

Index ................................ .................................................................................... . ······ 281

Contributors

AVERY D. ANDREWS 3RD: School of Language Studies, Baldcssill Prec inct Building (1 10), Austra lian Na tional Uni versity, Canberra ACT 0200, Australia. Email : avery . andrews@a nu. edu. a u

DAVID BRADLEY: Linguisti cs, La Trobc University VIC 3086, Austra li a. Email : d . bradley@latrobe . edu. a u

ANDREW BUTCHER: Speech Pathology & Audiology, School of Medic ine, Flinders University, GPO Box 2 100, Adelaide SA 500 1, Australia. Emai l : a ndy. b u tcher@fli nders . edu. au

MrCIlAEL CARNE: Department of Linguistics, School of History, Culture and Language, College of As ia and the Pacific, Austra lian Na tiona l Un iversity, Cnl1bcrra ACT 0200, Australia. Email : michael . carne@anu . edu. au

FRANTZ CLERMONT: King Fahd Universi ty of Petroleum & Minera ls, College of Applied & Supporting Stud ies, PEP· Box 5026, Dhahran 3 126 1. Kingdom of Saudi Arabia. Ema i l: akusti kfo neti ks @ya hoo . com . a u

CATHRYN DONOHUE: Department of Lingui stics, School of History, Culture und Language, College of Asia and the Paci fic , Austra lian Na tional Univers ity, Canberra ACT 0200, Australia. Emai l: cat h ryn . donohue@a nu . edu. a u

MARK DONOHUE: Department of Linguist ics, School of History, Culture and Language, College of Asia and the Paci fi c, Australian Na tiona l University, Canberra ACT 0200, Austral ia. Ema il : mar k. do no hu e@a nu . e du. a u

JAVt ER FRANCO·PEDROSO: Escucla Pol itccnica Superior, Unive rsidad Autonoma de Madrid, Calle Francisco Tomas y Va liente, 11 ,28049 Madrid , Spain. Email : javier . franco@ ua m. es

HELEN FRASER: Faculty of Law, University of New South Wales NSW 2052, Australia. Email : he!en @helenf ras e r . com . a u

JAV IER GONZALEZ·DoMINGUEZ: Escucla Politcc nica Superior, Univcrsidad Aut6nom<l de Madrid, Ca lle Francisco Tamils y Valiente, 11 ,28049 Madrid, Spain. Email : javier . gonzalez@ ua m. e s

JOAQUIN GONZALEZ- RODRIGUEZ: Escuda Pol itccnica Superior, Uni vers idad Autonoma de Madrid, Ca llc Francisco Tomas y Va li ente, 11 , 28049 Madrid, Spain. Email : joaquin . gon zalez@ uam . es

III

IV

KOICHI HONDA: No.8, e u xa Nguyen Canh Chan, P. Nguyen e ll Trinh, District I , Ho Chi Minh City, Vietnam. Email : hondako@gmail . com

SIiUNICHl ISI-IlHARA: Department of Linguistics, School of History, Culture and Language, College of Asia and the Pacific, Australian Nationa l University, Canberra ACT 0200, Australia. Email : s hun ichi . ishi hara@a nu.edu . a u

ANN K UMAR: School of Culture, History and Language, College of Asia and the Pacific, Australian National Un iversity, Canberra ACf 0200, Australia. Emai l: a nn . kumar@a nu. edu. au

Gr::OI'FERY STEWART MORRISON: Forensic Voice Comparison Laboratory, School of Electrical Engineeri ng & Telecommunications, Un ivers ity of New South Wales, Sydney. Australia Emai l : geoff- morrison@forensic- voice - comparison . net

DANI EL RAMOS: Escuela Politecni ca Superior, Un iversidad Autonoma dc Madrid, Ca lle Francisco Tomas y Va liente, 11 , 28049 Madrid , Spa in. Email : da nie l . ramo s @u arn . e s

WILLIAM STEED: Disc ipline of Speech Pathology, School of Public Health, Tropical Mcdicine and Rehabilitation Science, James Cook Uni versity, Douglas QLD 4810, Austra li a. Email : william . ste ed@jcu. edu . a u

THARMARAJA I-I T I·IIRUVARAN: Forensic Voice Com parison Laboratory, Schoo l of Electrical Engineering & Telecommunications, University of New South Wa les, Sydncy, Austra lia Email : t. t hiruvaran @un s w. edu . a u

TAKAKO TODA: Wascda Un iversi ty, 1-7-14 Nishiwaseda, Shinjuku-ku, Tokyo, 169-8050 Japan . Email : toda@waseda . j p

MICHAEL WAGN ER: Human-Centred Computing Laboratory, Faculty of Information Sciences and Enginecri ng, Un iversity of Canberra ACT 260 I Email : michael . wa gner@canbe rra . e du. au

ZIIENGDAO YE: School of Language Studies, Ba ldess in Prec inct Building (1 10), Ell ery Crescent, Australian National University, Canberra ACT 0200, Austra lia . Email : zhengdao .ye@ a nu. edu . a u

CUlLING ZHANG: Forensic Voice Comparison Laboratory, School of Electrical Engineering & Telecommunications, Universi ty of Ncw South Wales, Sydney, Austra lia ; Departmcnt of Forensic Sc icnee & Technology, China Criminal Police University, Shcnyang, Chi na . Ema il : c ui l i ng- zha ng@ forensic -voice-compa r ison.net

XIAONONG ZHU: Division of Humanities, The Hong Kong Un iversity of Sciencc and Technology, Clearwater Bay, Kowloon , Hong Kong. Email : hrnxzhu@h kust .hk

Phil Rose: A short biography Ann Kumar, Cathryn Donohue, Shunichi Ishihara

Phil was born in London, though his fami ly later moved to the countrys ide, which was more congenial than the town for an athle tic boy. Phil 's father came from a we ll-ta-do cou llIry famil y and was duly sell t ofT to a public schoo l as a boarder - hi s school days endi ng prematu rely when, [l mere 14 yea rs old , he enli sted by fals ify ing hi s age. )-Ie was a decorated war hero pa rt icul arly for hi s work with the French Resistance. Phil's mother was al so heroic and ac ti ve ly in vo lved in the wa r effort, driving ambulances for the Wo men's Auxiliary Air Force in the war to pick up inju red pilots. These ambulances we re huge, cumbersome vehicles requiri ng spec ial tra ining and a specia l licence to drive, and del.l ling with the seriously wounded wus no picnic either.

Phil 's father real ized ea rly in the piece tilat he had an except ional SOil , who was precociously gifted both as a cricketer and as a chess player. Besides coaching him in criekct and chess he also prov ided Phil with II wel l- rounded educati on dcveloping a ll his abilities. Pcrhaps becausc o f his own country ori gins, he was an enthus iast ic naturali st and took young Phil on field trips identifying the fauna frolll TIle /Joys Book oIBeasls. (One of Phil 's hobb ies is assembling animal ske letons !) In secondary school Phil was an ou tstand ing studell t and was Head Boy in hi s final yea r, but he was also passion<lle about play ing sport - he was Captain of Cricket and Cuptain of Rugby and was even recru ited to play in ad ul t cricket s ides frOIll the age of about th irteen or fourteen. Though he was always proud to contribute to his tea m's success, his mother wus far less inclined to discount the cost in inj uries, whi ch on one occasion included a bloody broken nose.

Phil a ttended the Uni versity of Manches ter for his undergraduat e studi es. His major spec iali zat ion was German, in which he had already acquired considerab le competence at school , though during the in-coulltry peri od of hi s German training he started studyi ng Chinese and Japanese thu nks to a scholarship bc rece ived to study at Erl angen. He also stud ied Russian, and graduated with First Class Honours. He subsequently ga ined the London Dip. IPA (First Clnss).

Phil conti nued to study Chinese, which was to be at the centre of his academic career, teaching hi mself from an o ld Li nguaphone course with 78-rptn records and working at it whil e commuting be tween work and his home in East Grinstead . He undertook hi s postgraduate studi es at Cambridge Unive rsity, focusing on Chi nese and solidifyi ng his career path in tone lind Ilcoustics. He completed both his M.A. and Ph.D. under the direction of Paul Kratochvil , author o f The Chinese Language Today and an expert on tones and tone sandh i, who also, Phil says, laught him to write Engli sh wel l.

Phil was recruited to the A.N.U. by Bob Dixon in the year Ma rgaret Thatc her became Prime Mini ster of Ihe U.K. ( 1979). He spent ten years as a tutor though he was ac tua lly the lecturer in charge of major courses. At the same lime he did an enormO lIS amount of reading across the who le di sc ipl ine of linguistics. Although he always elaims to know very little of other branches of the di sc ipline than phonet ics, he has in fael been inte llectuall y engaged well beyond hi s own spec ia lty, and hi s opinions arc always we ll-cons idered .

v

VI KIII1fOl; C Donolllle. Ishihara

Teaching has always been a vocat ion and not just a job for Phil. In a review of the Department of Linguistics it was clear that he was a quite exceptional teacher: among many appreciative assessments one student commented "Even in a department of good teachers, Phil Rose stands out", His dedicated and inspiring teaching was nowhere pu t to better effect than in the first-year class, whi ch was the bedrock of the linguistics program. He taught thi s over a long period, and in sett ing the numerous ass ignments for this course he succeeded in making them not only challenging, bUI also fun. In his phonetics courses, he would always spend a lot of time with every student selling up and monitoring their course project, ensuring that they produced the best possible work despite the loss o r the phonetics lab technic ian. Compromise was anathema to him and he cont inued to mainta in thc high standards he set himselr ror both his undergraduate and his graduate students, even as the linguistics program was progressively cut down despite. glowing extemal reviews recommendi ng inereascd sta ffing.

Despite his very heavy tcaching load Phi l produced a strenm or publications or the highest quality. His publications on tone covered not on ly Chinese but also Tha i, and other languages; he even made a study or Tibetan spell ing chant. In contrast to his notab le patience with students or a ll ca libres, he had no patience at all with the ever- increas ing number or unstructured and rambling meeti ngs on university procedures and 'restructuring' that impinged on time ava ilab le ror activities such as teaching and research that he considered much more important.

The second main rocus of Phil's academic career bcgan when hi s rormer student Mark Durie recommended him to someonc with a rorensic problcm as far and away the bcst qualifi ed and most careful and conscient ious pcrson. How right this judgment was to prove! After work ing in th is field for some yenrs Phil produced the pioneering book Forensic Speaker Identification (Rose 2002) introducing a ncw, Bayesian, mcthodology to thc fie ld. The book was dedicatcd to hi s mum, who died the day before she could see the finished book, and his dad 'who taught him how to observe'. He established this Bayesian approach in the discip line of forens ic voice comparison in Austra lia. His pionecring work led to him being chosen as keynote speaker at Odyssey 2004-The Speaker and Language Recognition Workshop in Toledo, Spain, which was hosted by the Biometric Recognition Group of Un ivcrs idad Autonoma de Madrid. This in turn led to an ongoing co ll aboration with Spani sh Bayes ian speech scientists at the Univers idad AUlonoma de Madrid. Onc of thc significant outcomes from this collaboration is the publication o r Gonza lez-Rodriguez, Rose, Ramos, To ledano & Ortega-Garcia (2007) in which it is demonstrated how traditional and automatic forensic voice comparison methodologies meet the court admissib ility rcquirements. The award of a British Academy Professorship enabled him to further improve his statistical competcnce in collaboration with colleagues (such as Colin Aitken and David Lucy) of the Joseph Bcll Centre for Forensic Statist ics and Legal Reason ing at the University of Edi nburgh. His research coll aborat ions are not limited to Western countries. In some of Phil 's papers on forensic voice comparison (e.g. Rose, Lucy & Osanai 2004; Rose 20 11 a; 20 11 b), Japanese is used as the target language, duc to the success ful co llaboration with the Nat ional Research Institutc of Police Science, Japan.

As his reputation grew, Phil undcrtook an increasingly heavy load of forensic cascwork. Among his many cases the Bain case in New Zealand is perhaps the most strik ing example of his contribution to the di spensation of justice. David Bain had scrvcd more than 13 years in prison for the murders of his parents and siblings, before a retrial was ordered in 2009. Ph il was thc court's expert witness and rather than focusing on what he heard , Phi l ri ghtl y pointed

Phil Rose: If .~hO"f hiogmphy VII

out that what anyone hears is irre levan t, whatillatters is what Baill said , something that can bc establi shed with acoustic analysis. Phil's meticu lous ana lysis of tile disputcd utterance thatthc policc claimed was clcarly an admission of gu ilt by David Bain was the most tell ing exposure of the danger o f thi s claim by any expert. Bain was subsequently acquitted after his long ordca l and a second trial.

In 2007 Phil was awarded a large ARC grant on a forensic voice comparison project ent it led Catching Criminals by Their Voice - Combilling All/omatic and Traditional Methods for Optimum Peljol'llumce ;11 Forensic Speaker identification, and appoin ted Geoffrcy Morrison as hi s research associate. He also brough t Cui li ng Zhang from the China Crimina l Police University to the ANU thanks to a small grant from the Intcmutionu l Centre for Excel lence in As ian Studies, leading to the introduction of Baycsiun fo rensic voice comparison to Chi na . They produced many important publications advancing the professional standards of forens ic voice comparison including Rose and Morrison (2009), Morrison , Zhang and Rose (20 II ), Morrison, Rose and Zhang (20 12).

Phil spcnt the first semester of 2012 at the Hong Kong Uni ve rs ity of Science and Technology, tcaching an undergraduate course on the Chinese language and a postgraduate course aimed at cSl<Iblishing whcthcr likelihood rnlio-based forcnsic vo ice comparison was possible for Chinese - and whethcr it could be taught in one semester to studcnts wi th no previous training. The outcome of thi s iallcr coursc was 110 less thAIl six pnpers by the studcnts, with Ph il liS co-author, submillcd to and accepted for the 14th Aust ral as ia n Inlcmational Confercnce on Speech Science and Tcchnology (where thi s vo lullle will bc presentcd). Subjccts covcrcd includc 'Like li hood ratio-bascd forcnsic vo ice comparison with segmental cepstrn from Cantonese and Jilpancse sy ll ab ic/mora nasals'; and ' Likelihood ratio­based forens ic voice comparison wi th Cantonese short-term fundamcn tal frcquency distribu ti on parameters' . Phi l himself had only one single-author paper accepted - he submilled two, but was to ld one was the conferencc li mit! On return ing to Canberra, he immcdiately co ll aborated with Hugh Sclby, the leg.t l expert on evaluation of evidence, to write a cri tique of thc draft standards proposed for fo rens ic voice comparison in Austra lia .

Phil's current major project is a book on southern Wu tOllC sandhi . With in that framcwork he has already complctcd II comprehensive study of WenzhOll (iJ.'.i~i'I), on whi ch he has alrcady publ ished some papers, and is now writing a monograph focuscd on WCl1zhou morphotoncmics. The plan of the book is to ana lyze both righ t- and lefl-branching tone sandhi varieties, as we ll as intcmlcdiate varieties, illus trating fca turcs of all sys tems.

Phil always has Illany irons in the fire and th is book, along with othcr projec ts , includi ng a tcx tbook, will keep him busy for many years to come in his retiremcnt from the ANU (but thankfu ll y not from the field!).

References Gonzalez-Rodriguez, JOl.1q uin, Phil Rose , Danie l Ramos, Dorotea T. Toledallo and Jav ier

Ortcga-Garcia . 2007. Emulating DNA : Rigorous quantification of cvidential we ight in transparent and testab lc forensic speaker recogn it ion. IEEE Trallsactions 011 Audio Speech and Language Processing 15 (7). 2104-2 11 5.

vii i KUlI/aJ; CDonohue. Ishihara

Morrison , Geoffery S., Cui ling Zhang and Phil Rose. 20 11. An empirical estimate of the precision of likelihood ratios from a forcnsic-voice-comparison systelll. Forensic Science II1ternational208 (1- 3).59- 65.

Morrison, Geoffrey S., Phil Rose and Cui ling Zhang. 2012. Protocol fo r the co tiection of databases of record ings for forens ic-voice-comparison research and practice. Australian Journal oj Forensic Sciences 44 (2). l 55~167 .

Rose, Phil. 2002. Forensic speaker identification. London: Taylor & Francis.

Rose, Phil. 20 II a. Forensic voice comparison with secular shibboleths - A hybrid fused GMM-Multivariatc likelihood ratio-based approach using alveolo-palata l fricative cepstral spectra . In Proceedings oj IEEE International ConJerence on Acollstics. Speech. and Signal Processing (ICASSP), 5900- 5903. Prague, Czech Republic.

Rose, Phil. 2011 b. Forensic voice comparison with Japanese vowel acoustics - A li kelihood rat io-bascd approach using segmental cepstra. In Proceedings oj the 21" International Congress oj Phonetic Sciences, Hong Kong. 1718- 172 1.

Rose, Phil , David Lucy and Takashi Osanai. 2004. Linguistic-acoustic forensic speaker identification with likelihood ra tios from a multivariate hierarchical random effects model: A "non-idiot's Bayes" approach. In Proceedings oj the 101h Australian !nfemational ConJerence on Speech Science and Technology, Sydney, Austra lia . 492-497.

Rose, Phil and GeofTcry S. Morrison. 2009. A response to the UK posi tion statement on forensic speaker comparison. International JOl//"I1al oJSpeech. Language and the Law 16 (1). 139- 163.

Phil Rose: China, 2007 (Courtesy of Zhu Xiaonong)

Ningbo

Westlake, Haogzhou

IX

Many voices, many tones Cathryn Donohue", Shunichi Ishihara", William Steedb

:'Australion National University. bJames Cook University

This volume is a collection of papers by friends and co lleagues of Phil Rose to honour him for hi s significant contribution to the field of linguistics on the occasion of his recent retirement from the ANU. Phil has influenced us all in Illany important and lasting ways. His work in both (especially Chinese) tone and forens ic voice comparison has always nmllagcd to improve upon the accepted standard through the insistence all stat isti ca l significance and the introduction of stati stical techniques not previous ly employed in the field . He is a true sc ientist and always makes sure that his work was both ambit ious and of the highest calibre.

Phil has a long history of working on tona l phenomena, since hi s Master 's thes is on the phonology of Ningbo (Ch inese). Despite having somc of the best cars around, Phil quickly identified the shortcomings of auditory impress ions for seriolls tona l descripti ons, turning to the aeollstie signal for data that can bc propcrly quantified and measured (e.g. Rose 1982). He used many speakers for hi s early studies to ensure taking the mcan eliminatcd performance qu irks. However, Phi l had also secn efforts to normalize acoustic data for vowels and had the great idea of extend ing the genera l practise to tona l descripti ons as we ll. Needless to say, Phil's 1987 paper, "Considerat ions on the normali sation of the fundamental frequency of linguistic tone", had a major impact on the field . Phil used a s imple s tati stic - the z-scorc transfoml - to normalize hi s acousti c data of the tones to abstract awuy from the variation found between speakers, making il poss ible 10 identify the lanai contou rs representative of the variety as (I whole. He has sincc made studies of many varieties of Chinese (and other languages), clarifying isslies for tonologica l rcscareh as well as providing quality quantificd descripti ons for the field of linguistics . Phi l 's approach is now the go ld standard for tona l descriptions, and 110 doubt p!lvcd the way for hi s work in forens ic vo ice comparison.

Phi l's work on fo rensic voice compmison (FYC) has turned the field Oil its head, quite literally. In a lypic,)1 FVC easc, you havc a speech recording from an unknown voice (e.g. an offender), and a speech sample from a known subject (c.g. a suspect). Beforc the introduction of the likelihood ratio-based approach, the problem of FYC was cons idered a classification/identification task; dctcrm ining, for example, whether the specch from the offender and suspect we re s imi lar enough to be identificd as coming from the same person (or more "precisely", the probabi lity of this being the case). In tcchnica l terms, this is equiva lent to trying to eva luate the probabi li ty of a hypothes is (c.g. the sllspect is gu ilty), given Ihe ev idencc (c.g. the observed similarities and differenccs between speech pallems of the suspect and the offender). Indeed, FYC was long considered a subfield of automatic speaker recognit ion and even referred to as forensic speaker recognition/ identificmion/verificat ion.

However, Phil has shown that there arc serious problems (both technical and Icga l) with the o ld approach. Not only is it imposs iblc for the forcnsic specialis t to direct ly calculate the (conditiona l) probabilities of such hypotheses (e.g. the suspec t is guilty) given the evidence available to him/her, but also, statemen ts on the probabi lity of such hypotheses being true potentially change the scientist's ro le from that of cxpert witness, to onc of " trier-of-fac t", norma lly reserved for judges and juries.

2 CDol/olllle. Ishihara. S/eed

Phil is one of the first people who introduced the likelihood rat io-based evidence evaluation to forensic speech science, and demonstrated thaI it works wi th speech evidence. Using likelihood ratios one can eva luate the odds of obtain ing the evidence under the two competing hypotheses (e.g. the suspect and the offender are the same person, or not). Thus, one obtains a measure of the strength of the evidence in support of either hypothesis, as is done in cases of DNA profiling. Ana lysing and presenting the evidence in Ihis way docs not violate the role of the trier-of-fact, since , following Bayes' theorem, the judge (or jury) must not only consider the likelihood rat ios presented by the forensic spec ialist, but also factor in the prior odds (i.e. the other evidence in the case) to reach a fina l conclusion regarding the overall odds in favor of the hypothesis. Indeed, Phil has been a fervent advocate for the eva luation and presentation o f FVC evidence in tenns of Bayesian likelihood ratios, emulating the long~standing pract ice. in the field of DNA forens ics. Thanks to Phil, the likelihood ratio~based FVC is more than pure ly an area of academic research, but has been used in practice in several trial s and is becoming widely accepted as the standard procedure in the Austral ian Courl.

The title of thi s vo lume could well have been the same as that of this introductory piece­"Many voices, many tones" - instead, it is the broader Quantitative approaches to problems in linguistics. This was necessalY to encompass all the papers ineluded here; however, it also reflects Phil's engagement with the field. While his publications, professiona l ach ievements and accolades highlight his work in the phonetic sciences, a conversat ion with him on any topic will reveal a much broader interest and knowledge. He is genuinely interested in other areas and is singula rly capable of not on ly immediately comprehending issues in that sub field he may be introduced 10, but also of identifying ways in which that area could benefit from a particular approach or ana lysis or the further scrutiny of some related data. His own work now includes not just synchronic desc l'i ptions and explorations of lona l phenomena as well as forensic voice comparison in a variety of languages, but also a novel approach to thc historical reconstruction of tones, also inspired by his background in stati stics and norma lization.

This volume is divided into three scctions. The first two represent the two main areas of Phil 's research: tonal studies and acoustics, and forcnsie voice comparison, and the third shows Phil's influence in wider fie lds of research.

The first section slarts with Mark Donohue's paper "The shape and spread of tone" which examines the integration of lone with the rest of the phonological system ofa language. Investigating the co~occurrencc of contrastive tones, contrastive vowels, and syll able shapes in a large sample of languages, he shows Ihat there is a statistically significant skewing of the distribution of languages such that large numbcrs of contrastive toncs are associated with large numbers of contrast ive vowels, and not correlated wit h complex codas. However, when the correlations are examined on a family~by~family and area~by~area basis, it appears that the associations are not uniform, suggesting that different processes of tonogenesis have <lpplied in different regions (or alleast that the circumstantial evidence fo r tonogenes is is different).

William Steed's "Qingtian Wu lexical tonc sandhi: Voice less depressors and allotones" builds on Ph il's work describing depressor phenomcna in Wenzhou (Rose 2002a). The evidence from Qingtian data affirms his assertion that the lowered onset of depressed tones is not synchronically connected to the syllab ic onset's voice feature. The loss of phono logical vo icing in the data, whi le retaining the depressed tones that merge with the ir undepressed allotones. simplifies the underlying phonology of Qingtian. The dcscription continues the

1v/(lIIY voic:es. lIIony lones 3

good example Phil has set of providing quantified aeousl ic data when descri bing compl ex. tone phenomena.

Koichi Honda 's detai led desc ription of Lam River Vietnamese shows more subtl e interac ti ons between tones and segments in "The reulisation of stopped tOile in North-Central Vietnumese". He shows thai in addition to phonation ty pe variation, vowe l length separates two Vietnamese tone ca tegories further in Lam Ri ver region, partly condit ioned by the final consonant. The secondary conclusion shows that the stopped and unstopped tones foml allotonic pairs. rather than being tonemicall y separate. These conclusions re in force that tonal descript ion should not rely entirely on pilch.

In hi s puper "Tone Altemations in Ugong (Thailand)". David Bradley separates the complex. real isations into two processes, and describes the variation , including the loss of the altcmations, thtH occurs in the speech of semi-speakers. These scmi-spcukers instead overlay a tone system from the more dominant Thai language on their Ugong speech. The description of not only the altemat ion, but ulso the variation found contri butes to both tone study und the study of language contuct.

In her paper "The ro le of con tour and phonation in FlIzhou tona l ide nt ifi cation", Cathryn Donohue presents resu lts frol11 a discriminution task Ihat shows that both a s li ght contour in FO in (phonologica lly) Icvel tones and the usc of non-modal phonation ure statistieully signi ficant fac tors in tonu l identification in FlIzhou. The results pose particu lar challenges to stundard tonologica l unul yses of Chinese by suggesting that phoneti call y grounded approaches should incorporate both these perceptually s ignificant fea tures.

Shunichi Ishihara clai ms that littl e lingu istic phonetic descriptive research has been undertaken on the accept types of Japancsc di,llccts in his paper "Osaka and Kagoshima Japanese c itation tone acoust ics: A linguistic !onetic comparati ve study". In this study, focusing 0 11 Osaka (OJ) and Kagoshima Japanese (KJ) which share some same pitch pallcms (LH, LHL, LLH und LLLH), 1) he derived linguist ic tonetic representations of OJ and KJ tonali ties frolll the normulised acoustic representations of LH, LHL, LLH and LLLH pitch patterns, and then 2) he exp lored some impl ications of the identified differences in the acoustic reali sation ofl-llL units between OJ and KJ for the surface tona l representation ofKJ within the autosegmental metrical theory. The usefulness of the z-score normalisation technique in lonul studies was theoretica lly argued for and empi rica lly tested in Rose (1987, 1991 ). The tec hnique has been widely lIsed as the standa rd technique in Illany tonetic studies now.

Like Phil , Xiaonong Zhu has a lo ng-standing inte rest in Wu acoustics, both tones and segments. In the puper included here, "OfT-the-chart vowel chunges in Chinese", Zhu describes and una lyses the modern reflex.es that arise from rais ing high vowels. He describes fi ve processes tha t Chinese varieties have undergone as u rcsult of rais ing vowels, based on principles of art icul at ion and perception . He uses Phil's description and 'curl ed j ' transcription of Zhenhai 's 'h igh vowel with friction ' for Shanghai's equiva lent vowe l (Rose 1982).

Takako Tod:1 invest igated the role of uge in acquiring the phonology of a second language in her puper "The critical pcriod hypothesis and phonologica l acquisition of Japuncse". In her study, the common notion of ' the earlier the belter' for language lcaming ­of whi ch the theoret ieu l ex.planation is the well -known crit ical period hypot hes is - was tested in terms of Japanese pronunciati on. Speech samples and background information were collected from 86 L2 Japanese speakers vary ing in thei r LI languages and 16 native speakers of Japanese (control group) . Her results demonstrated that there was a llegut ive correlation between age o f onset ulld level of phonological lIequi sition, supporting the notion of ' the

4 CVonohlie. Ishihara. Sleed

earlier the better' for language leaming. However, she also found Oul that there were some learners who success fully attained the nati ve-level pronunciation even though they started studying Japanese after the cri tica l period.

Andrew Butcher 's paper, "Long, thin phonologies" uses acoustic ev idence to explain the cooccurrcncc of many places of articulation and few manners of articul ation in Austra lian languages. He uses acoustic and articulatory evidence to demonstrate that speakers of these languages do not spread cues of manner of art icu lation to vowels, a llowing the vowels to give mOfC salient cues to place of articu lation.

The second sect ion contains papers linking to Phil's research on forensic vo ice compari son (FVC). As the author of one of the major textbooks (Rose 2002b) written for non-lingu ists 10

understand how FVC works in the I~gal system, Phi l has necessarily had a large impact on people working in this area.

Frantz Clermont's paper " Linear-scaling cffects of phonetic context on vowel formants: A tutorial" is a fundamental research on a source of wi thin-speaker variabil ity, namely phonetic-context effect on fonnant transitions, which is highly relevant to Ph il's work on FVC. Clcnnont and hi s colleaguc hypothesise that fo r a given spcaker and a given formant, the relativc spacing betwcen vowels is invariant, and unaffected by consonantal con text or by location within a syllable. They call this linear-scaling hypothesis. In hi s tutorial paper, first of all, Frantz Clennont explains the key concept of the lincar-sea ling approach, namely vowe l-formant cnsemble and then illustrates the various stagcs of thi s approach using F2 data obtained from 5 Arabic vowels in 13 fricative eontcx ts.

Telephone transmission of any kind introduccs an additiona l acoustic filter that is not related 10 a speaker's individual is ing infonnation. Phi l (Rose 2002b, 2003) provides detailed descriptions as to how fonnant va luc cstimates arc affected by landlinc telephone transmiss ions in comparison to direct recording. Largely motivatcd by these works of Phil's, Michael Carne 's paper, "Japanese fonnant frequencies in mob ile phonc transmission: Implications for FVC", describes an ex periment that in vestigates the impact of mobi le phone transmissions on vowe l fonnant frequencies (FI -FJ) in Japanesc in tenns of the direction of freq uency shifts associatcd with spectral distortion. Although formant frequencies have becn cons idered fairly robust to different channel effects, the resu lts of his experiment show significant di fferences between the mobile phone and microphone fonnant measurements. He emphasises the need for the FVC analyst to exercise caution when lIsing speech samples from different record ing conditions.

Phil tested the effectiveness of a fricative as a FVC feature in Rose (20 II ), to potentially add to the standardly used vowel spectra l feature for phonetically motivated FYC stud ics . In their paper, "Fine-grained automatic speakcr recognition using cepstral trajectories in phone units", Javier Franco-Pedroso, Joaquin Conzalez-Rodriguez, Javier Conzalcz­Dominguez and Daniel Ramos conduct a series of experiments in order to investigate to what ex.tent different phone units contribute to speaker's identity. Their experiments foclls on the temporal trajectorics of the mel-frequency cepstral coefficient (MFCC) extracted from different phone units. Thcir experimental results arc useful , in particular, whcn forensic speech scientists need to deal with uncontrolled scenarios where only some short scgments are avail able to be compared . In that case, the speech scientist can infer a conclusion about the speakcr identi ty in the speech samplc from the experimental results of the current study.

Phil 's work on FVC demonstrates that exploring and testing new acoustic features are very important for FVC. In his paper "Automatic speaker identification using the magnitude

A1l111Y lIoice.f, IIWIIY tull(!.\· 5

and phase spcct ra of in verse-filtered voiced speech", Michael Wagner reports the results of speaker identificat ion experiments based on speakers' glotta l source characteri sti cs, whieh arc not standard features of contemporary spcflkcr rccognition systcms. From these prom ising results, he argues that the inclusion of glotUl I source characteri stics will resul t in a better pcrformance in conventional speaker recognition, in which the standard features are rel ated to the vocal-tract transfer function, such as MFCC. In hi s papcr, he also proposes an algorithm to obtain an approximation of the glottal exc itation functi on from each vo iced frame of an utterance.

Chinese tones and FVC arc Phil's two main areas of expertise, so this study. by close co llaborators of hi s, is a golden combination for Phil. Focusing on thc fOnllant traj ectories of thc Standard Chinese triphthong liaul on tOile I, Cuiling Zhang, Ceoffrey Morrison and Tharmarajah Thiruvaran explore the effec ti vcncss of the formant trajectory in FVC in their paper "Forensic voice comparison using Chincse l ill u/". They use di screte cosine tranSfOnllS (DCTs) in order to model fo rmant trajectories, and then usc the deri ved DCT coefficient values for li kelihood ratio calcu lation. They rcport a substantial improvement in system validity but a dcel inc in system reliabi lity when th is information is added to a generic automatic FVC system.

Helen Fraser states in her paper, " Bayes and Beyond: The complex chall cnges of LADO and their relevancc to forensic speaker comparison", tha t one of the Illany significant achievements of Phil 's work is that he has demonstrated the importance of presen ting FYC evidence in terms ofa Bayes ian Likelihood Ratio (Rose 2002b). Relating to the use of speech samples as lega l ev idcnee, she writes about some issues surrounding LA DO (Language Analys is for Determ inat ion of the Origin of asylum seekers). She draws attention to the current status of the area, and outlincs some of the issues that are most pertincnt to the field . She calls for the tes ting of va lidity and rigollr in current methodo logies and statistical cducation in lega l profess ions as two primary resea rch nceds with regards to LADO. The need for public education on the lega l use of sta tistics is part icularly relevant not just to LA DO, but also for criminal forcnsics.

The final section includes three papers. While the topics of these p.lpcrs fall outside Phil 's main research areas, they touch upon his resenrch program through thcir relcvancc to Chincsc culture, language change, and Bayesian statisti cs, and are testimony to Phil's broadcr engagement with thc academic community and the sort of pos it ive impact he has on his colleagues.

Avery Andrews 's paper, "UG and vari ation in expression", proposes a principle call ed ' Mi rroring the Vari ation in Ex pression ' which allows some forms of indi rect negat ive evidence to be taken into account when try ing to understand the architecture of Universal Grammar. This Bayes ian approach 10 learnability, inspired by Phil 's interest in Bayes' Theorem, promiscs to be very fruitful , al ready overcollling sevcral of the problems of other approaches.

Ann Kumar 's chapter on "Contex tua li zing thc Old Javanese influence on Old Japanese" is a sequel to an earl ier paper published in co llaboration wi th Ph il (Kumar & Rose 2000). 11 builds on thei r prev iously establi shed phonological link between between Old Japanese and Old Javanese. This chapter adds lexica l evidence to the phonologica l evidence, through a detai led examination of the ' humble auxi li aries' and other morphosyntacti c fcaturcs shared by both languages. Kumar also adduces 'Irchaeologica l and historica l evidcnee, further confirming the continued contact between the two civilisat ions.

6 C.Donohue. Ishihara, Steed

Wu Chinese being one of Phil's foci of research (e.g. Rose 2001). Zhcngdao Ve ts paper "Eating and drinking in Mundarin and Shanghaincsc" is a good paper to round oul the work on Wu presented here. Ye contrasts Mandarin Chinese, with separate words for eating and drinking, with Shanghai Chinese, wh ich only uses a single word to describe both activities. Usi ng Natural Semant ic Metalanguage, she describes prototypes and extended meanings of eating and drinking words in both languages in order to demonstrate how the Shanghai word lIZ chyq does not match the conceptual space of its Mandarin cognate chi.

We arc grateful to Hsin-tien Liao from the ANU for the beautiful ca lligraphy to be presented to Phil at the same time as thi s volume, and reproduecd on the front covcr. It is a pocm by thc poct He Zhlzhang (f8J"l~; 659~754 CE) who came from Yuezhou~Yongxing, in what today is Xiaoshan County, in Zhejiang Province wherc Phil has conducted much of his own fieldwork, He was one of a group of four poets from the Lower Yangtze Basin that were known as the "Four Scholars from Wuzhong". A politiCian and a poet, he retired from politics at age eighty~five to become a Daoist hermit near Lake Jinghu in Zhejiang Province. He was a great friend of Li Bni, and in fact gave him his appellation of the "Banished Immortal", He was himself called one of the "Eight Immortals of the Winecup" by Du Fu and as he was known for his id iosyncrasy, also earned the appellation "Crazy Zhang". He was known for his openhearted love of the lower classes, and for his freethinking, and in later Daoist tales is presentcd as a man who achieved immortality. Only 19 of his poems remain. I This poem is about local accents, something that this particular set of contributors - like Phil - is especially gratcful for!

§Ili$ 111111 & 'H!ltli<;t:k §I. li$1l-1!! "UI 'B:tI: . .R.rMIlYl.'f'ffllIII. ~ p.~ ~r.ti"J!13 *. HUI xillng ~u shn SMo xi~o Ii jia l~oda hui, xiiingyin Wli SAi bin mao shuai. Ert6ng xiang jian bil xiangshi, xiao wen ke cong he ehu lai.

Retllming lrome I left home young and returned old. accent unchanged, but my hair now thin and grey. Little kids do IIot know me at all- with a big smile they ask "Where are you from, stranger? ..

We wou ld like to thank LlNCOM for accepting our proposal to publish this collection so enthusiast ically, We arc deeply grateful to all our revicwers for their time and energy in providing extensive comments on the papers, increasing their quality immensely. Our reviewers include Kanae Am ino, Steven Bird , Heather Buchan, Cathryn Donohue, Mark Donohue, Helen Fraser, Margaret Hughes, Larry Hyman, Shuniehi Ish ihara, Yuki ltan i­Adams, Mark Johnson, Yuko Kinoshita, Tom Mylne, Daniel Ramos, Wi lli am Steed, Kimiko Tsukada, Tom Wasow, Kristine Yu, Jie Zhang and five additional anonymous rev iewers.

We would like to thank OUf contributors, not only for their contributions but also for working within our time constraints. Several others wanted to contribute to the vo lume but could not due to other pressing commitments. We wish to express our gratitude to the

I Thanks (0 the A/Ichor book o/Chinese poefly web companion, hostcd by Whittier College. for most of Ihis information [hup:llweb. whittier.edu/academic/english/chinescl].

MallY voices, 111(111), IOIlC.f 7

Department of Linguistics, ANU for their fimlllcia l support of thi s proj ect. We would fmiher like to acknowledge Geoffrey Morrison 's invo lvemen t in the init ial stagcs of thi s project.

Finally, we would like to end by wishi ng Phil all the very best in th is new chapter of his life and as he focuses hi s energy more exelusively on his research . Whi le your absence in the teaching ranks at the AN U wi ll be sorely felt , we look for,vard to enjoying - and no doubt benefiting from - the fruits ofyollr la bour. Certainl y your textbook wi ll be a wonderful legacy for the studcnts who will miss the opportunity to work with yOll directl y [and perhaps, when that book ofWu is complete you' ll finally head soulh to Min! -CD]

References Kumar, Ann and Phil Rose. 2000. Lex ica l evidence for early contact between Indonesian

languages and Japanese. Oceanic Lingllistics 39 (2) . 219-255.

Rose, Philip. 1982. An acoustically based phonet ic desc ription of the syllable in the Zhcnhai dialect. PhD di ssertation: Cambridge Universi ty.

Rose, Philip. 1987. Considerations on the normali sation of the fundamental frequency of li nguistic tone. Speech Communicatioll 6 (4). 343- 352.

Rose, Philip. 1991 . How eflect ive arc long term mean and standard dev icllion as normalisation parameters for tonal fundamental frequency? Speech Commllilicatioll 10.229- 247.

Rose, Philip. 2001. Chinese languages: Wu. In Ja ne Garry and Carl Rubino (cds) Facts abollt tlte worlds languages, pp. 158- 16 1. New York: H. W. Wilson Press.

Rose, Philip. 2002a. Independent depressor and register effects in Wu dialect tonology. JOII/'l1al o/Chille!"e Lingllistics 30 ( 1). 39- 8 1.

Rose, Philip. 2002b. Forensic speaker identijication. New York: Taylor & Francis.

Rose, Philip. 2003 . The technica l compari son of forensic vo ice samples. In Inn Freckclton and Hugh Selby (cds) Expert Evidence. chapter 99. Sydney: Thomson Lawbook Company.

Rose, Philip. 201 1. Forensic voice comparison wi th secular shibboleths - A hybrid fused GMM-Multi variatc likelihood mlio-based approach using alvcolo-palatal fricati ve eepstral speetnl . In Proceedings o/the lOll IEE£ Il1tel'l1(1tional Conference on Acollstics, Speech and Signal ProceSSing. 5900-5903.