On Some Determinants of Error in Child Phonology

Kobe Shoin Women’s University Repository

Title

On Some Determinants of Error in Child Phonology

Author(s)

Yukimasa Hattori

Citation

Shoin Literary Review，No.10：75-90

Issue Date

1976

Resource Type

Bulletin Paper / 紀要論文

Resource Version

URL

Right

On 

Some 

Determinants 

of

Errorin 

Child 

Phonology

Yukimasa  HATTORX Inthispaperwewouldliketoexploresomeofthephonological featuresoflanguagedevelopmentofanative-Japanese-speakingchild fromtheagesofthirty-onethroughthirty-sixmonths.Oneofthe authorshaspreviouslypublishedapaperregardingthedevelopment ofthesamechild,NojiSumihare,fromtheagesoftwenty-fourthrough thirtymonths.Themajorreasonforusingthespeechofthisparticular childasadatabase,istheexistenceofthevoDuminousandmeticulously-lceptrecordofhislanguagedevelopmentoveralongperiodoftime fromitsearlieststagesuntiltheageofsixyears.Twovolumesof thisrecord,representingyearsthreeandfour,havealreadybeenpub-lisped(BunkaHyoronShuppan).Whilethispaperisbasedonthe speechrecordofonlythisonechild,weplaninthenearfutureto expandtheanalysistoincludedatafromotherchildren,aswellas datafromtheearliermonthsofSumihare'sspeechrecord. Onewaytoapproachastudyofchildphonologyistolookatwhat Thispaperwasco-authoredbyMarySanches,DepartmentofAnthropology, UniversityofTexas-AustinandYukimasaHattori,however,accordingto theregulationsoftheShoinScientificResearchSociety,onlythenamesof membersofthesocietycanappearonpaperspublishedintheShoinLiterary Review.Thus,onlyMr.Hattori'snameappearsabove.

thechildproducesascorrectandincorrectfromthestandpointofan adultidealmodeltowardwhichthechildcanoethoughtofasmoving _, andintermsofthedevelopmentofarticulatoryprocessesdetermined byphysiologicalmaturation .Herehowever,wewouldliketoconsider alsosomeotherpossibledeterminantsofthe"errors"whichwefind inSumihare'sphonologyatthisage, Onemightconsiderthatbytheageofthirty-onemonths ,achildhas masteredthephonemicdistinctionsofthelanguageheorsheis learningand,evenifnotabletoproduceallofthembecauseofsorne articulatoryproblems,thatthechildshouldstillbeabletodistinguish amongthemconceptually.Ifachildhasmasteredtheconceptual phonemicdistinctionswhichliebehindthephonologyoftheparticular languagebeinglearned,thenithastraditionallybeenassumedthat those"errors"whichdooccurmustbetheresultofsomearticulatory difficulties.Inthispaperwewillnotruleoutthispossibilityof articulatorydifficultiesasdeterminingtheoccurrenceofphoneticforms whichwouldnotbeconsidered"correct"foranadult _,butwewill suggestthattherearealsootherkindsofdeterminantsproducingthese deviations,namelythosehavingtodowith:(1)ot-erallphonemicand morphophonemicpatternsinthelanguage-inthiscase _,inthe patternsofautomaticalternationandcertaincontractionpatternsof colloquialstyleintheadultlanguage,and(2)themeaningwhich certainphoneticalternationscarry,alsotruefortheadultlanguage . Wewillfurthersuggestthatthemostfrequenttypesof"errors"are patterned,andraisethequestionofhowthesepatternedtypesof phonologicaldeviationsmayberelatedtoonekindo￡languagechange inJapanese.

1・Typesandamoun彦 〔ゾ ♪ 加 πoZO9'Cα'``errors"ZnSumiha .re'5spee`h. Therearernanyoccurrencesofwhatmight,byanative-speaking

adult,bethoughtofaserrors,especiallysoifthespeechrecordis lookedatfromanexclusively"phonemic"pointofview.Wewillfirst presentallofthevarioustypesof"errors",theircontextsandfre-quencies,andthendiscussthevariousfactorswhichmustbeconsidered ifwearetoaccountforthem.Intermsofprocessesofproduction, errorsareoftwotypes:vowelandconsonantlength,andphonetic substitutionanddeletion-alsoofbothvowelsandconsonants. A.Vowelandconsonantlength.OneofthefeaturesofSumihare's phonologynoticeablydifferentfromanidealmodelisinthesometimes deviantlengthofconsonantsandvowels.Someofthesedeviations areconstant,someunpredictable,andsomelooktobedeterminedby semanticfactors.Undueconsonantlengthconstantlyoccurs,inallof theformsofdeicticspeci丘cation:thus,acci(idealαcの,kocci(idea1 々o`の.Thephenomenonofinappropriatevowellengthalsooccurs throughouttherecordofthesesixmonths.Syinappropriatewemean thatlengthisaddedwherenonewouldbeexpected,phonemically. Someexamplesfromthethirty-fourthmontharerepresentati、 ・e: a→a/simクsu,osakana,・..,fut4tu u→{i/homu,taberu,iru,aru,ik露,surd e→2/kot∂,tore,de o一 う6/unkδ,fuitagyo,misetagyδ,no,1σzu Interestingly,thereisnooccurrenceinthisrecordofinappropriately lengthened,!i/. B.Phoneticsubstitutionanddeletion:Theconsonanterrorsmade bySumihareareasfollows ChartI. InitiallyMedially 1.s→c:canrincaikina`ai

Gurukimacu ceklcalcucumimacen coacoko 2・v3→cφsanrinoa citeocikko `ukociφ

 

の  

COICCO

ソ

3・s→vsφminsai

ヘ

ノ

ソ

surunbasu

ソ

ロ

 

のmisete

v

sorottoφ 4.ts→ccuretefutacu 5・j→ φ φoican 6・w→ φ φmaatta 7.h→ φ φTakθ αra 8・t→cφkicanai 9・t→ 善 φki"sanai

 

ソ

10・z→j(o)ブ6riφ 11.g→kφkomorikasa

ソ

12・h→v3Slt・tuφ 13.s→kkakuramboφ 14・t→nφha"abl 15.r→gφagigato 16.b→mφa脚naku

・7・ト

・ ・a{盟aψ

18.t→dmuide

19・ky→kk6(〈kyO)senk6(<senky6)

20。ky→

乙CO(<ky6)φ

21、hy→vsvsomenφ

一 一78一

22.by→bibiyoinφ

ね

23.y→vSsobocusaψ 24.y→ φ φkureon 25,si→ φ φdotte 26.su→ φ φde」kka 27.o→-w→ φ φkonda(<kondowa) 28.ki→ φ φTerucan(くTerukican) 29.ra→ φ φokoretan(〈olioraretan) makumoto(makuramoto) Thevowelsubstitutionsanddeletionsinhisspeechareasfollows 1,i→uhifuku 2,i→eenpetu 3.iaclocca 4,e‐3ciagyo,misetageru J,e→ ユoma: 6.e→atoramasen 7.a→ φsinsai 8,a→esakurembo 9.a→ubutabuta 10,a→okoun 11.o-→ias∫kov}7a 12.o-一>aperapera 13.o→ φraji(rajio) 14.u→ibiranko 15.0→uiku 16.yu→iokaican 17.u→eperopero amongthesevariousdeviationsfromwhatcanbeseenasthe correctadultphoneticrepresentationofforms,wethinkthatwecan

distinguishanumberofdifferentkindsoferrors,thatis,different basesforproducingtheerroneousphoneticrepresentations.These areasfollows,andnotethattheyarenotallbasedonarticulatoryor evenphonemiccriteria 1.Articulatoryandphonemicsimilarity:Thisdeterminantof erroneousarticulationmaybebecausethetwocategoriesofsounds areclosecognitivelyforthechildandthustendtobeconfused,orit maybebecausetheyarearticulatorilycloseandgetsubstituteddue tothechild'sasyetnotfullydevelopedarticulatorycontrol.Ehamples ofthesetypesoferrorsare: a.voicing(eitheracorrectlyvoicedsoundisunvoiced,orvice-versa)一#17」 →c;#12g→k;#18t→d b,1abio-dentalfricativebecomesgrooved。#12h→s c,fricativebecomesaffricated-#10z→l d,stopbecomescontinuant-(bilabia1)#16b→m;(dental)#14 t→n e.problemswithpalatals-Thereisageneralproblemwithpalatals forSumihare,basedontheevidenceofsixkindsoferrorshe haswiththem.In#s23&24y→sinitiailyandφmedially; thepalatalizedconsonantskyandbyhealsohastroublewith, producingforthefirsteitherkorcandaCVsequence,bi, forthesecond;forthepalatalizedvelarfricative,hy,healso producesvS(see#s19-22)・ f.semivoweldeletionintervocalically・#6,7w,h→ φ g.assimilation-#15r→gprecedingg. 2,Whiletheseerrorsmaybeaccountedforbyinvokingaprinciple ofphonemiccloseness,whichmayorinaynotbedistinguishablefrom articulatory"slips",thisdoesnotaccountforalltheerrors.The assumptionwhichweareemployinghereisthatthisprincipleworks :1

bestwhenthesounddifferencesareonlyonefeatureawayfromeach otherinthecorrectanderroneousforms.However,thereareanumber oferrorsinwhichthesoundssubstitutedaremorethanonefeature-whetherarticulatoryordistinctive-awayfromthecorrectsound; therearealsodeletionsofindividualsoundsandsyllableswhichcannot beaccountedforintermsofphonemiccloseness. The丘stinstanceofthisis#14・Wefeelthatthisexample ,along withthevowelerrorrepresentedin#80fthevowellist _,issimply lexicalconfusion:thechildconfusesthesetwolexicalrepresentations inbothcontextswhereoneortheotherwouldbeappropriate _.That is,sometimeshesayskakurambowhenhemeanssakuramboandsome-timeshesayssaku-rembowhenhemeanskakurembo.Lexical ,rather thanphonemic,confusionwouldappeartobethecauseofanumberof theothervowelerrors,particu]ary詐s5,12&17wherehehasconfused peroperowithperaperaandperoperowithpuropuro.Further,thereason heproducedrajiforrajio-whichmightsuperficiallylooktobeinvolving thedeletionofo,-seemsrathertobebecausethatishisrepresentation forthatlexeme;thereisnoevidencethatheeverusesanyotherform _. Thefactorofconfusinglexemesmayalsoaccountfortheerrorin#9 , wherebuta,avalidforminhislexicon,isconfusedlexicallywith Bataandreduplicatedtoformbatabata. 3.Athirdkindofcauseofphoneticerrorseemstobemorpho-phonemicandconnectedwiththeprocessofcontraction.Whilewe cannotgointothisprocessfullyhere,itiscommonincolloquialstylein Japanesetoproduceformsliketabecaufromwhatcanbethoughtof asamoreslowandcarefulstyle(ifnotunderlying)formtabete 十simau.Thisprocessseemstobeevidentinthe"voweldeletion  of#7inthelistofvowelerrors:sinasai>sinsai(andsometjmes sincai,whichisadifferentproblemtobediscussedbelow),and#4

ageyo>agyo,etc.Theseseemtobebu[anextensionofadult, colloquial-stylecontractionprocesses.Alsosuspiciouslyresembling theseare#s26and27inthelistofconsonanterrorswhereasyllable involvings十Visdroppedandtheresultingformevidencesgenination ofthefollowingconsonant,i.e.,desuka-dekka,#26,0rdosite-dotte, #27.#30fthevowelerrorscouldlikewiseheacontractionfroman attempteddoci-ryaordocira→docca.Andthisalsoseemstobethe caseinoneoftheexamplesof#300ftheconsonanterrorfistwhere rawas"dropped":probablyintheprocessofinternalizingthe derivationalruleforcombiningmalcura+moto,thechildmaybe workingonaruleanalogoustothatforderivingnominalcombinations fromverbalforms. Anotherdeterminantofwhatsuperficiallyappearstobephoneticor articulatoryerrorinSumihare'sspeechappearsoncloserinspection tobeaccountableforbyhisnot-as-yet-perfectunderstandingofmor-phophonemicrulesinvolvedininflectingverbs.Webelievethatthis accountsfortherepresentationoftorαmasen-#60fthevowelerrors-wheretoriinasenwouldbemorphophonemicallyproperforthe-masen negative,buttoranai,withwhichheseemstohaveconfusedtheroot form,wouldbeapplicableforthe-nainegative.Likewiseitappears thattheerrorinvolvedin#300ftheconsonanterrors,wherehedrops -ra- _{,inonecasetheendofthestemwhichisfollowedby-re-,the} passive,hesimplydoesnotunderstandtheuseofthepassiveinflection yet.Wealsosuggestthattheerroneousoin#100fthevowelerrors istheresultofconfusionbetweentwomorphophonemicalternatives inhisrepertoire:whenheformstheperfectofsomeforms-kau amongthem・hesometimesusesthedialectk6ta,andsometimesthe " standard"katta,Wesuggestthatheissimplyinthisinstance extendiDgthelogicalphoneticrepresentationforimperfectfromthe -82一

dialectperfect:thuskota--kou,whichisnotaphonetically,or articulatorilymotivatederroratall,thoughitmaysoundlikeit. Thisanalysisleavesuswithanumberofvowelerrorsforwhichwe cannotaccount:#1.i/u;#2.i/e;#14,u/i,andafewconsonant errors-#s1-4,whichwewouldliketodiscussinmoredetail. Inadditiontothe``errors"inSumihare'sphonologydiscussedabove, thefourlistedatthebeginningofthelistofconsonanterrors,areso preponderant-theiroverallfrequencyofoccurrenceissomuchgreater thanthatofalloftheothersputtogetherthatwehavefeltitnecessary togivethemspecialconsideration.Theirrelativeoccurrenceover thesixmonthsperiodofthisspeechrecordispresentedinChartI1. ChartII.Frequenciesoffour・mostcommon"errors",bymonth. AgeinMonths 313233343536 Errortype s→c66%32%9%3%0.6%φ s→VS3.1%2.9%9.9%9.9%2.5%1,7% ts→c33°028%9%9%10%3% § →c49%27%φ φr;iφ Whatthischartshowsisthat(1)thedeviationsofs→c,s→ 百, V S→vC,andts→coccurfarlnorethandotheothers.somuchso thatinthefirstmonthofthissegmentofthespeechrecordtwoof them:s→candS→caccountforatleasthalfof亡hetotaloccur-rencesoftheformsofsands,andcisone-thirdoftheoccurrence ofis;(2)whiletheothererrorscanbeconsideredtobeforthemost partoccasionalarticulatorlly-orotherwise-determindphonological "errors"ofaperformancenature ,thesefirstfivemustbeconsidered, duetotheiroverwhelmingoccurrence,asdetermindbysomethingelse than"performance"mistakes;(3)furthermore,thischartdemonstrates

thattheerrorsalmostentirelydisappearbythethirty一 丘fthmonth _, indicatingthatwhatevertheircauseis,ithasbeeneliminatedby then. Onefactorwhichwethinkisofimportancetoconsiderinany discussionofthedeterminantsofthesefourpredominanterrorsis thewayinwhichtheyarerelatedtosomeautomaticphonological alternationsinadultJapaneseconsonant-initialsyllables _.Aseveryone knows,theseare:(1)t/u→ts;(2)t/i→cand(3)S/i→vS _.In termsofarticulatoryfeatureswecanchartthisalternationasfollows ChartIII.Articulationsq1㌧4"θ プπα庭oπ dentalalveolaralveopalatal

Stops〈full

affricated二

し

聾/量

皇1⊥

-V・ice5ノ

ε

彗!i

Fricatives1一

一

(grooved)o

＼+V。iceZ!忌Y

1/i

'u o InSumihare'scase,atthebeginningoftherecordusedhereat anyrate,hedoesnotseemtohaveahardandfastruletoprovidefor thesealternationsinallcases.Hismodelmightberepresentedas thefollowing -84一

ChartIV.ArticulationsofAlternation:Sumihare'smodel dentalalveolaralveopalatal fu1Lし/0

-5tvps

・ff・i・a・ ・d"-66拓 一 ・・uu}i a,。o) -VOlce且5∂ 拓 Frica仁iviess!e・36a (grooved)01已 Inotherwords,whilein66%ofthecases,Sumihare'st→is/u,in 33°,60fthecasesofthisform,heshiftsthearticulationfurtherback toanalveopalatalc,appropriateinthecontextof_/i,almostasthough hehasacompetingrulewhichcouldberepresentedast/i,u→c. Furthermore,inthecaseofs,insteadofarulewhichsays5/i→ 善, hisactualspeechin50%ofthecasesproduced∫/i→candin66% ofthecaseswherebyadultrulessisappropriate,i.e.,!/(a,u,o,e), heproducedC,andin3%ofthesecases,vS.Thus,thereisafar greater"deviation"ofsthanoft.Wemightconceptualizethese deviationsoftandsas"drifting";inthiscasedriftingtoward alveopalatals. Whyshouldthisbeso?Whyshouldt/u→cinanalogytot/i, insteadoftheotherwayaround?Whyshouldsinallenvironments tendtobecomec?Whiletheremaybesomebasisinthearticulatory featuresoftheseunitswhichinfluencesthedirectionoftheirdevia-tion,wethinkthatratherthanascribingthephenomenontotallyto thiskindofafactor,wemustconsider{1)thatthedetermination isinpartmadebytheoverallpatternofalternationintheadult

f language,and(2)themeaningswhichatleastoneofthese"errors" (s・ →vC)hasinadultJapanese. Sythefirstpointwemeanthatevefeela"naturalness"factorcannot bewellinvokedtoaccountforthe"drifting'°phenomenawefindin Sumihare'sphoneticproductions.Ifnaturalnessweretobeinvoked, wemightaskforevidencethatcisanyeasiertoproduce,inany environmentinwhichitoccurs,thant,s,ors.Ifthechild'serrors occurredonthebasisofwhatmightbeexpectedasmorenatural,then wemightexpectthephoneticgaps-si,tu,ti-tobefilled;thatis forthechildtoproducesiwheretheadultlanguagecallsforsi,to wheretheadultlanguagespecifiestsuandtiwheretheadultlanguage specifiesci.Thus,inonesense,wethinkthatthedirectionofthese particularerrors,whatwehavecalled"drifting",isduetothedirection ofautomaticalternationintheadultphonology,thatis,ofstopsto affricatesandofbothdentalsandalveolarstoalveopalatals. $ythesecondpointwemeanthatinadditiontothisphonological patternbaseforthedirectionofthesedeviations,thereisalsoa "meaning"basewhichactsasanadditionaldeterminerfortheoc-currenceofthese"errors."Thisisespeciallyevidentinthecaseofthe s→caltemation-the``error,'withthehighestpercentageofoc. currence,Theexamplewewouldlilcetociteinsupportofthisview isinthealternationsoftheforms-sanand-can.Thefigureswehave representedaboveinChartIIdonottakeintoaccountthealternation ofs→cintheseforms;iftheydid,thepercentageoftheoccurrence ofcasan"erroneous"productionofswouldbehigherthanthe66% thatitis. Intheadultlanguageofcourse,-canissometimesproducedby adults-notwiththeimplicationthattheyhavemadeaphonetic error,butasaformwithameaningofdiminutive,intimateaddress -86一

andreferencetochildren,etc.IfwelookatSumihare'suseofthese twoalternatives,wefindthefollowing (1)Thereareformswithwhich-canoccursexclusivelyandothers withwhich-sanoccursexclusively.Thus, A.-can-suffixedforms immediatefamily (o)tocan,"father" (o)kacan,"mother" terukican,"yr.bro" bokucan,"self,1" (0)bacan,``gmo." jican,"gfa." neighborsandothers Kyokocanlo如kican,``moon"

 

Takeharacanyuyakecan,"sunset" Maricanakacan,"baby" Makocan"namesofosanbacan,"midwife" Miccangirls"yaoyacan,"grocer" Kecanhitoyubican,"hisfinger" TomocanIkyupican,"doll" Qbacan,"aunt,neighbor" Totoobacan"Hamamotoneighbor" O(j)ican,"uncle,neighbor" otofu-no-ojican,"tofuseller" taitai-no-ojican,``丘sh・storeprop." Akionican Kaauboccan S。ican"・ ・m…fb・ys" Torucan onican,"el.bro.,visitingstudent (o)pecan;"visitor,el.sister" animals hikikaerucan,"frog(s)" kokkocan,"hen,chicken" nyankocan,"kitty" kamecan,"turtle" kumacan,"bear"

wanican,"aligator" memecan,"goat" ノOods amecan,"candy" (o)mamecan,"beans" oimocan,"potatoes,variousroots" B.-san-suffixedforms people oisasan,"doctor" okomeyasan,"rice-seller" jorijorisan,"barb.er" sanpatuyasan,"barb°r" ringoyasan,"apple-seller" sentakuyasan,"laundryman" akusuriyasan,"pharmacy/cist" cindonyasan,"advertisers" omawarisan,"policeman" untenyasan,"busdriver" gyunyu-no-ojisan,``milkman" omocayasan,"toystore/seller" rajioyasan,"radiostore/seller" ameuriyasan,"candystore/seller" sa"sosan,"trainconductor" Fujiisan,"landlord'sname" uta-no-obasan,"womanwhosingsonradioprogram,`songauntie' obasan,characterinastory momotarosan,characterinastory ohosisan,"star(s)" onisan,"demon" minasan,"everybody" hotokesan,"ancestorspirit" ouma-no-kasan,"motherhorse",characterinastory Therearetwopaintswhicharetobemadeaboutthedifference betweentheentitieswhicharereferredtowiththetwodifferentforms, -sanand-can:generally ,those"beings"andthingswhichrate-can seemtobethosewhicharecloseandimportanttoSumihare、towards ..

で

whichhefeelsaffection,whichheisfamiliarwithandnotafraidof. Goodexamplesare:hismostcommonandbestlikedfoods,most immediaterelatives,playmatesandtoys.Itseemsintuitivelyobvious that-can,havingthisconnotation,shouldbeusedinreferringto theseentities.Whatisnatsoclear,fromourconceptsofthem,is whyentitieslikewani,"crocodile",orbeya,"bear"shouldbeincluded inthisset.PerhapsitistheforminwhichtheyoccurinSumihare's world,i.e.aseither"cute"toysorpictorialrepresentationsinabook. Likewise,itseemsthatthosethings,orentitieswhichratea-san suffix,arethosewhichseemtobemoredistant,unfamiliar,orfearful toSumihare.Thisseemsborneoutbythefactthatmanyofthe -san-suffixedformsincludethee工ement一 ッα。,signallingthatthe referrentcanbethoughtofaseithertheinstitution-astroe,etc.-or theperson{s)representingtheinstitution.Ontheotherhand,only oneoftheformssuffixedwith-cayaincludesthisreference.Thisis yaoya,whichoccursasbothyaoyacanandyaoyasan"grocer".When welookedatthecontextofthesetwoformswefoundthattheformer occurredwhenSumiharewasplayingatbeingagrocer,andthelatter alwaysoccurredwhenhewasmakingreferencetothe"real"grocer. Anothercorroboratingbitofevidenceforthispointisthedistinction hemakesbetweentaitai-no-ojiCan,"fish-seller,(fish-uncle)",whose referencemustbeunambiguouslytoapersonandsakanaya8atx,"fish store/seller"whosereferencecouldbeeithertothestoreitselforto theperson(s)representingthestore. Whilethisformulationaccountsformostofthealternationbetween sandcinthismorphologicalcontext,itdoesnotaccountforallofit. Thereareafewotherreferentswhicharesometimessuffixedwith -canandsometimeswith-san.Theyare darumacan/darumasan,"figureofDaruma"

kaminarican/kaminarisan,"thunder" okaican/okayusan,"ricegruel" Oneofthefactorswhichmayberesponsiblefortheformofthesuffix intheseinstancesisthemodelwhichtheparentspresenttothechild onoccasiontheparentsvarythesuffixwhichtheyuseinreferringto anentity.Wewouldhypothesizethatforsomeentitiesthemeaning differenceiscleartothechild,but#orothersitisstillnotproductive, ortheparentalmodelisoverridingofthisconnotativedistinction. Inconclusion,wehypothesizethatthedevelopmentofthemeaning distinctionbetweenthesuffixalternatives.sanand.canmayberelated tothepatternofphonetic"drift"ofsoundstowardaffricationand palatalizationasfollows:intheprocessofinternalizingarulewhich relatesthealternationofthesetwoforms,-sanand-canwithamean-ingfuldifferenceof``intimacy,closeness,cuteness",versus,``distance, formality,respect,andfearfulness",thechildgeneralizesthemeaning toothersoundsnearby-especially-t/u,andsinallcontexts, Thisclearlyraisesaquestionaboutatypeofphonologicalchange inJapanese.Didthes→cinthe-san/-cartsuffixalternation,with itsmeaningof"diminutive,etc."originallygettherebecauseparents perceivedthetendencyofchildrentoproducecsinawiderangeof phoneticcontexts?Thisiswhatthenaieveobserverwouldexpect. Zilehavetakenanalternative,andmorecomplexviewhere-inthe lightofoiztogeneticdata-andhypothesizedthat(1)thechildlearns toassociateadifferenceinmeaningwiththeformaldifferencewhichis overgeneralizedintheprocessoflearningandthat(2)forboth childrenandadultsthisparticularvariant-whichcanbeexpressedas thetendencytopalatalizationaridaffrication-takestheformitdoes becauseoftheoverallpatternofthephonemicinventoryandautomatic alternations. -90一