Phonetically driven phonology The role of Optimali

PhoneticallyDrivenPhonology:TheRoleofOptimalityTheoryandInductiveGrounding1BruceP.HayesUCLAFinalversion:August1997Writtenfortheproceedingsvolumeofthe1996MilwaukeeConferenceonFormalismandFunctionalisminLinguisticsAbstractFunctionalistphoneticliteraturehasshownhowthephonologiesofhumanlanguagesarearrangedtofacilitateeaseofarticulationandperception.Theexplanatoryforceofphonologicaltheoryisgreatlyincreasedifitcandirectlyaccesstheseresearchresults.Therearetwoformalmechanismsthattogethercanfacilitatethelink-upofformaltofunctionalwork.Asothershavenoted,OptimalityTheory,withitsemphasisondirectlyincorporatingprinciplesofmarkedness,canserveaspartofthebridge.Anothermechanismisproposedhere:analgorithmforinductivegroundingpermitsthelanguagelearnertoaccesstheknowledgegainedfromexperienceinarticulationandperception,andformfromittheappropriatesetofformalphonologicalconstraints.1IwouldliketothankMarcoBaroni,PatriciaKeating,DoncaSteriade,andmembersoftalkaudiencesatMIT,Arizona,Tilburg,Utrecht,Cornell,UCSantaCruz,UCLA,andtheFormalism/Functionalismconferenceforhelpfulinputinthepreparationofthispaper.Phonetically-DrivenPhonologyp.21.PhonologicalFunctionalismThedifferencebetweenformalistandfunctionalistapproachesinlinguisticshastakendifferentformsindifferentareas.Forphonology,andparticularlyforthestudyoffully-productivesoundpatterns,thefunctionalistapproachhastraditionallybeenphoneticincharacter.Forsometime,workinthephoneticliterature,suchasOhala(1974,1978,1981,1983),OhalaandOHALA(1993),LiljencrantsandLindblom(1972),Lindblom(1983,1990),andWestburyandKeating(1986),hasarguedthatthesoundpatternsoflanguagesareeffectivelyarrangedtofacilitateeaseofarticulationanddistinctnessofcontrastingformsinperception.Inthisview,muchofthepatterningofphonologyreflectsprinciplesofgooddesign.2Incontemporaryphonologicaltheorizing,suchaviewhasnotbeenwidelyadopted.Phonologyhasbeenmodeledasaformalsystem,setuptomirrorthecharacteristicphonologicalbehavioroflanguages.Occasionally,scholarshavemadeanodtowardsthephoneticsensiblenessofaparticularproposal.Butonthewhole,thedividebetweenformalandfunctionalistapproachesinphonologyhasbeenasdeepasanywhereelseinthestudyoflanguage.Itwouldbepointless(albeitfun)todiscussreasonsforthisbasedonthesociologyofthefieldsofphoneticsandphonology.Morepertinently,Iwillclaimthatpartoftheproblemhasbeenthatphonologicaltheoryhasnotuntilrecentlyadvancedtothepointwhereaseriouscomingtogripswithphoneticfunctionalismwouldbeworkable.2.OptimalityTheoryThenovelapproachtolinguistictheorizingknownasOptimalityTheory(PrinceandSmolensky1993)appearstooffertheprospectofamajorchangeinthissituation.HerearesomeofthebasicpremisesofthetheoryasIconstrueit.First,phonologicalgrammarisnotarrangedinthemannerofChomskyandHalle(1968),inessenceasanassemblylineconvertingunderlyingtosurfacerepresentationsinaseriesof2Thelastsentencedefineswhatismeanthereby“functionalism”.Unfortunately,thistermhascometodenoteaparticularviewpointonalargenumberofotherissues,forinstance,thestabilityandintegrityofthegrammarasacognitivesystem,thedesirabilityofexplicitformalanalysis,thevalidityofintuiteddata,andsoforth.Itisquitepossibletoholdmainstreamgenerativistviewsonalloftheseotherissues(yes,thegrammarisastable,cohesivecognitivesystem;yes,formalanalysisiscrucialtoresearchsuccess;yes,intuitivejudgmentswhencarefullygatheredareveryuseful;etc.)butbefunctionalistinthesenseofbelievingthattheformalsystemofgrammarcharacteristicallyreflectsprinciplesofgooddesign.Thisismoreorlessmyownposition.Phonetically-DrivenPhonologyp.3steps.Instead,thephonologyselectsanoutputformfromthesetoflogicalpossibilities.Itmakesitsselectionusingalargesetofconstraints,whichspecifywhatis“good”aboutanoutput,inthefollowingtwoways:(1)a.Phonotactics:“TheoutputshouldhavephonologicalpropertyX.”b.Faithfulness:“TheoutputshouldresembletheinputinpossessingpropertyY.”Phonotacticconstraintsexpresspropertiesofphonologicalmarkedness,whicharetypicallyuncontroversial.Forexample,theyrequirethatsyllablesbeopen,orthatfrontvowelsbeunrounded,andsoon.TheFaithfulnessconstraintsembodyadetailedfactorizationofwhatitmeansfortheoutputtoresembletheinput;theyarefullysatisfiedwhentheoutputisidenticaltotheinput.Constraintscanconflictwitheachother.Often,itisimpossiblefortheoutputtohavethedesiredphonotacticpropertiesandalsobefaithfultotheinput;orfortwodifferentphonotacticconstraintstobesatisfiedsimultaneously.Therefore,allconstraintsareprioritized;thatis,ranked.Prioritizationdrivesaspecificwinnowingprocess(notdescribedhere)thatultimatelyselectstheoutputofthegrammarfromthesetoflogicalpossibilitiesbyrulingoutallbutasinglewinner.3IwilltakethegenerallinethatOptimalityTheoryisagoodthing.First,itsharesthevirtuesofotherformaltheories:whenwellimplemented,suchtheoriesprovidefalsifiability,sothattheerrorsinananalysiscanleadtoimprovementorreplacement.Further,formaltheoriescharacteristicallyincreasethepatternrecognitioncapacityoftheanalyst.Forexample,itwasonlywhentheformaltheoryofmoraswasintroduced(Hyman1985)thatitbecameclearthatcompensatoryphonologicalprocessesalways