Download - An articulatory theory of sound change
Anarticulatorytheoryofsoundchange
Hypothesis:Mostcommoninitialmotivationforsoundchangeistheautomationofproduction.Tokensreducedonline,areperceivedasreducedandrepresentedintheexemplarclusterasreduced.
Thereforeweexpectsoundchangestoreflectadecreaseingesturalmagnitudeandanincreaseingesturaloverlap.
Whataresomewaystotestthearticulatorymodel?Thetheorymakespredictionsaboutwhatisapossiblesoundchange.Thesepredictionscouldbetestedonacross-linguisticdatabase.
Soundchangesthattakeplaceinthelanguagesoftheworldareverysimilar(Blevins2004,Bateman2000,Hajek 1997,Greenbergetal.1978).
Weshouldconsiderbothcommonandrarechangesandtrytoexplainboth.Commonandrarechangesmighthavedifferentcharacteristics.
Amongthepropertieswecouldlookforaretypesofphoneticmotivation,typesoflexicaldiffusion,gradualness,conditioningenvironmentandresultingsegments.
Commonvs.raresoundchange?
Weneedadatabasethatallowsustotesthypothesesconcerningwhattypesofchangesarecommonandwhattypesarenot.
Adatabaseofsoundchanges?
Mostsoundchangeshaveoccurredinundocumentedperiodssothatwehavenorecordofthem.
Evenincaseswithwrittenrecords,thephoneticinterpretationmaybeunclear.
Onlyasmallnumberoflanguageshavehistoricrecords.
Soanysampleofknownsoundchangeswouldbebiasedtowardsthoselanguages.
Adatabaseofsoundchanges?
Soundchangesareknownonlyforsomelanguagesoftheworld:
Languageswithwrittenhistories.
Soundchangescanbereconstructedbycomparingrelatedlanguages.
TraditionalreconstructionofPIEstops(Lehmann1955)
*p *t *k *kw*d *g *gw
*bh *dh *gh*ghw
The‘glottalic’ reconstruction (Hopper1973)
*p *t *k *kw*t’ *k’ *kw’
*b *d *g *gw
Adatabaseofsoundchanges?Reconstructedchangescannotbethebasisofatheoryofsoundchange.
Manylanguageshavenorecordedhistory.
Whatcanwelearnfromsynchronicdescriptions?
Mostphonologicalrulesthatdescribesoundalternationscomeaboutthroughsoundchange.
Allophonic‘rules’ordistributionscanbethoughtofasdescribingsoundchange.
Phoneticconditioningonly:Soundchangeshavehistories,too.
Asoundchangestartsasasmallphoneticchange,butthenmaycontinueuntiltheaffectedsegmentsaremoredistinctphonetically.
Palatalizationof/k/ inLatin>Spanishk>c>tʃ >ts >s/θ beforeafrontVorglide
Spanish:eléctri[k]ovs.electri[s]idadRule:k>s/___frontvowel(telescoping)
Rule:k>s/___frontvowelTelescoping(Hyman1975)
Notethattheruleisnotproductive.Onecannowhave/k/beforeafrontvowelwithoutachangeoccurring:Spanish: quince [kinse]‘fifteen’
queso [keso]‘cheese’
Soundchangeshavehistories,too.
Inversion:(Vennemann1972)IfwethinkoftheEnglishpluralmarkeras/z/,thenitisnecessarytoaddavowelinwordslikesclasses,wishes,churches.
Historically,however,thevowelwasthereandadeletionoccurredinmostcontexts,butnotwheretheprecedingCwasasibilant.
Soundchangeshavehistories,too.
Today,thevowel-insertionismorphologically-conditioned,asitisrestrictedtoapplyingwhere/z/isaplural,possessiveor3rd sg.Verbsuffix.
Thekà s‘rule’inSpanishislexicallyrestricted.
Phoneticconditioning
Inordertobesurethesynchronicprocesseswearelookingatarephoneticallyconditioned,wehavetoexcludealternationsthataremorphologicallyorlexicallyconditionedandnotphoneticallyproductive.
ALLOPHONDatabase
SheleceEasterdayandIconstructedadatabaseofphoneticallyconditioned(allophonic)processesin82languages.
Thelanguageswerechosentobemaximallyunrelatedgenealogically.ThesampleisbasedontheGRAMCATS sampledescribedinTheEvolutionofGrammar (1994).
ALLOPHONDatabase
Thedatawascollectedfromreferencegrammarsandcodedasfollows:
1. thesegmentundergoingchange,2. theresultingsegment,and3. theconditioningenvironment
wereallcodedforphonologicalfeaturesaccordingtothetraditionalclassificationsofconsonantvs.vowel,voicing,placeofarticulationandmannerofarticulation.
ALLOPHONDatabase
2.Multiplesegmentsundergoingachangeweredescribedinasingleprocessif
a.theinputsegmentsconstitutedaphoneticallydescribableclassb.theoutputsegmentsconstituteda
phonetically describableclassc.thefeatureschangedarethesamed.theconditioningenvironmentisthesame.
ALLOPHONDatabase
3.Whethertheprocessoccursacrossword-boundariesornot.4.Whethertheprocessoccursinrapidorcasualspeechorotherspecialsociolinguisticcontexts.
26%ofthe800+processescodedwerenotedtooccurintheconditionsmentionedin(4),suggestingthatthedatabaserepresentsashallowtime-depthofsoundchange.
Assumptions
1. Allophonicprocessesandphonetically-conditionedprocessesarethefirststageinthephonologizationofsoundchange.
2. Suchprocesses,likesoundchange,havephoneticexplanations.
3. Whilephoneticprocessesareverysimilaracrosslanguages,werecognizethattheremaybelanguage-specificdifferencesamongthem.
4. However,wealsoassumeinlookingforphoneticexplanations,thatsimilarprocessesacrosslanguageswillhavesimilarexplanations.
Somegeneralresults
49%ofprocessesareassimilatoryandmaybegesturalretiming.
35%arelenitionsorthereductionofgesturalmagnitudeorduration.
3.5%arestrengtheningGlidestrengthening 19Fricativestrengthening 9
Resultsonassimilation
Of391assimilationprocessesAnticipatory 56% (SeealsoJavkin 1978)Carryover 27%Both 17%
Placeassimilation:C-to-Vassimilation
Anticipatory 59%C-to-Cassimilation
Anticipatory 67% (SeealsoBlevins2004)
Strengtheningorfortition
Aphonologicalprocess(orsoundchange)thatincreasesthemagnitudeordurationofagesture.
NB:definedarticulatorily,notacoustically.t>ts orp>pf isnotastrengthening,itisaweakening.
Closurepointiseroding.Onapathtowardsfurtherweakening:pf >f;ts >s.
Apparentstrengtheningorfortition
ExcrescentCs:OEþunrian ‘tothunder’;bræmle ‘bramble’OldSpanish:salir +á ‘leave+FUT’>salrá >saldrávenir +á ‘come+FUT’>venrá >vendráLatin:homine ‘man’>Spanishhombre
ExcrescentCdevelopsoutofarticulationsthatarealreadypresent,asshownbythepointsofarticulation.Itisnotanincreaseinmagnitudeofagesture.
Blust 1990,inasurveyofknownsoundchangesinAstronesianlanguagesreports thatfortitionsarerestrictedtothesemivowels*yand*w.Thesearedetailedonpage50.
Theyinclude*wbecoming [gw]and*ybecomingacoronalaffricateHeevencitescaseswherethetransitionalglidebetweentwovowelsstrengthens into[b]or[g].
GlidestrengtheningAlloPhon database
MostfrequenttypeofCstrengtheninginAlloPhon database.12casesofapalatalglidestrengthening
ex:Pech /j/>[dj](word- ormorpheme-initial) (Holt1999:16)
ex:Slavey (Diné)variation:[j]~[ʒ][ʔajilá] ~[ʔaʒilá]‘3sdidto3s’(Rice1999)
Apinajé (Macro-Jê)(Oliveira2005:58-59)• a./ajet/[aˈʑet]‘tobesuspended onasurface’• b./atkaje/[atkaˈʑe]‘tocrack;tofissure’• c./jar/[ˈʑari]‘that(one)!’
Glidestrengthening
7casesofalabiovelarglidestrengthening
ex:Pech /w/>[gw](word- ormorpheme-initial)(Holt1999:16)
ex:Apinajé (Oliveira2005:60a)• a./kuwe/ [guˈvej]‘bird’
b./aw/ [aˈvəɾ]‘towardsyou’
Glidestrengtheningoccursinsyllable-initialposition.
Fricative>stoporaffricate
3casesofpalato-alveolartoaffricateorstopPech (Chibchan)/ta-ʃùna/>[ta-tʃùna]‘mynail’(Holt1999:16youngerspeakersuse[tʃ];possibleinfluenceofSpanish.)
Sheko (Omotic)(Hellenthal 2010:86) Free variationbāʒà [bāʒà ]~[bādʒà ]~[bāɟà ]‘work’
Garífuna (Arawakan)(Taylor1955:235)/ʃ/inunstressedsyllablesvarieswith[tʃ]instressedsyllables.
Fricative>stoporaffricate
3casesof/f/,/ɸ/,/β/to/p/.
Ningil (Sepik)[p]and[ɸ]fluctuatefreely(Manning andSaggers 1977:57)
Koiari (Papuan)(Dutton1996:6)/f/mayhave[p]asafreevariantwith[ɸ]word-initiallyprecedingbackvowels.[ɸuˈɸuri]~[puˈɸuri]‘Fufuri (nameofarock)
Oksapmin (Papuan)(Loughnane 2009:33)/ɸ/has[p]asanallophonesyllable-finally(no/p/phonemeashistorically,*pand*ɸmerged.
Fricativetostoporaffricate
Arethesestrengthening?
2casesofaglottalfricative>palatalfricativebeforehighfrontvowels
Maidu:j>ç /___highfrontV(Shipley1964)Tinrin:j>c/___highfrontV(Osumi 1995)
Fricativetostoporaffricate
Mostofthecasesdescribevariation.Directionalityunclearinmostcases.
InAlloPhon
Weakeningof/p/isdocumentedin8casesWeakeningofapalato-alveolaraffricateisfoundin3cases
Weakeningofstoptofricativeisfoundateverypointofarticulation,butreportedstrengtheningonlyatthelabialandpalato-alveolarpointsofarticulation.
Strengthening
Evidenceconcerningfricativetostop/affricateisdoubtful.
Overwhelmingly,bestdocumentedtypeofstrengtheningisglidestrengthening.
Strengthening
Evidencethatapalatal(high,front)articulationplaysasubstantialroleinstrengthening.
Arole,butalesserone,isplayedbylabialarticulations.
Notongue-tiparticulationsinvolvedinstrengthening
Inthefollowingwecomparethepointsofarticulationfoundinstrengtheningwiththosefoundinassimilation.
Strengthening
Givenhowspecific‘strengthening’processesseemtobe1. Notjusttheconverseof‘weakening’2. PerhapsrelatedtoPalatalizationand
labializationassimilations1. Bothtendtooccurinsyllable-initial position2. Boththeytendtooccurinthesamelanguages
Palatalization
AnassimilationofaCconditionedbyahighand/orfrontvowelorglide.
Atypicalpalatalizationhasasitsoutcomeanarticulationinthepalatalorpalato-alveolarregion,butourstudyencompassesotheroutcomesaswell.
SecondarypalatalizationcanoccuronanyCbutchangesinplaceusuallyaffectcoronalsandvelars.
Palatalization
Fullpalatalizationaffectscoronalsanddorsals.
(Allegedfullpalatalizationoflabialsappearstobeattributabletoglidestrengthening,Bateman2010)
SecondarypalatalizationcanaffectanyC(Bateman2007;2010).
Palatalization
ThoughanticipatorypalatalizationofaCtoaVisthemostfrequentplacefeatureassimilation,itonlyoccursinabouthalfthelanguagesoftheworld.
Bateman’s2007sampleof117lgs yieldsabout50%.
Allophonesampleof82languagesyields55%.
PalatalizationThemostcommonpalatalizationprocessesareanticipatory,meaningitisasyllable-initialCthatpalatalizesinmostcases.
Whydoesahigh,fronttonguepositioncausemoreassimilationthananyotherarticulation?
Recasens’2014theoryofarticulatoryconstraintorresistance:owingtothemassofthearticulatorinvolved,somearticulatorygesturesresisttheinfluenceofneighboringgesturesmore,andalsoinfluenceothergesturesmore.
Palatalizationandstrengthening
Evidencethatplaceassimilationandstrengtheningarerelated.
1. Thesamepointsofarticulationarefavoredforbothprocesses.
2. Someprocessesinvolvebothassimilationandstrengthening.
3. Aswewillseelater,glidestrengtheningtendstooccurinlanguagesthathavedevelopedpalatalization.
Palatalizationandstrengthening
Someprocessesinvolvestrengtheninginapalatalizingenvironment:AlloPhon:thefollowinghaveglidestrengtheningbeforeafrontvowel:Tinrin (Austronesian)Carib (Ge-Pano-Carib)Apinayé (Ge-Pano-Carib)Selepet (Indo-Pacific)Lahu (Sino-Tibetan)
Palatalizationandstrengthening
Intheconveniencesample,thefollowinghaveglidestrengtheningbeforeafrontvowel:
Atayal (Austronesian,NorthernFormosan)Awa(Kainantu,Barbacoan)Fanti (Niger-Congo,Akan)German(Indo-European)Nupe (Niger-Congo)
Strengtheningbeforeahighvowel:LowerGrandValleyDani (Trans-NewGuinea)
Dopalatalizationandstrengtheningoccurinthesamelanguages?
Somelanguagesshowalonghistoryofpalatalizationprocesses,withglidestrengtheningdevelopingafterotherprocesses(Romancelanguages).
Therefore,wewanttocheckthepresenceofsynchronicpalatalizationprocessesand
Thepresenceofpalatalandpalato-alveolarCsinthephonemeinventory,becausepalatalizationprocessesproducenewphonemes.
Languageswithglidestrengthening, AlloPhon sample.Allofthesehavecurrentpalatalizationprocesses,orCsinthepalato-alveolarregion intheirCinventories, indicatingeithercurrentorpastpalatalizationprocesses.
Languageswithsynchronicordiachronicglidestrengthening andthepresenceofpalatalizationprocessesorpalatal/palato-alveolarCsinphoneme inventory.
ChangesinCphonemeinventories
Frenchadds[ʃ],[ʒ]and[ɲ]
Spanishadds[tʃ],[ɲ]and[lj]
Portugueseadds[ʃ],[ʒ],[lj]and[ɲ]
PalatalglidestrengtheninginRomance
Latin:didnothaveaconsonantalglideaccordingtoKent1945.
PortugueseLatin iustum > Portjusto [ʒustu]‘fair’
iam > Portjá [ʒa] ‘now,already’
ArgentineSpanishModSpanishya ‘now’>[ʒa]
yo ‘I’ >[ʒo]llamar >[ʒamar]
PalatalizationandstrengtheninginCclusters
Thissequencestartsaspalatalizationandendsupasstrengthening(inPortuguese).
Latin Spanish Portugueseplicare ‘toarrive’llegar [ʎ]>[j] [tʃegar]>[ʃegar]chegarpluvia ‘rain’ lluvia [ʎ] [tʃuva]>[ʃuva]chuvaclamare ‘tocall’ llamar [ʎ] [tʃamar]>[ʃamar]chamarclave ‘key’ llave [ʎ] [tʃave]>[ʃave]chaveflamma ‘flame’ llama[ʎ] [tʃama]>[ʃama]chama
Menéndez-Pidal:the[l]palatalizesandtheinitialCislost.
Isthereapalatalizingtypeoflanguage?
Aswesawabove,allthelanguageswehavecheckedthathaveglidestrengtheningalsohaveassimilationprocessesconditionedbyahighand/orfrontVorglide,or
Csintheirphonemeinventoryinthepalatalorpalato-alveolarregion,evidenceofsuchprocesseshavingoccurredinthepast.
RecallthatonlyabouthalfofthelanguagessampledbyusandbyBatemanhavepalatalizationprocesses.
Isthereapalatalizingtypeoflanguage?
Possibleunderlyingcauses:1. Greaterco-articulationofsyllable-initialCandV.Inthatcase
otherplaceassimilationswouldalsooccur.
2. Articulatorysetting:perhapstheselanguageshaveahighfronttonguepositionthatishigherandmoretothefrontthanotherlanguages.(Gick,Bryan,IanWilson,KarstenKoch&ClareCook.(2004).Language-specificarticulatorysettings:evidencefrom inter-utterancerestposition.Phonetica 61.220-233.)
3. Articulatoryconstraint:perhapsthehighfronttonguepositionhasgreaterarticulatoryconstraintinsomelanguagesthaninothers.(Recasens,Daniel&Aina Espinosa.(2009).Anarticulatoryinvestigationoflingual coarticulatoryresistanceandaggressivenessforconsonantsandvowelsinCatalan.JASA 125.2288–98.)
Directionality
Lenition isverycommonFortitions areveryconstrained,occurringonlywithspecificarticulations,whichhappentobearticulationscommonlyinvolvedinassimilations.Assimilation isverycommonDissimilation israreandoftennotlexicallyregularMetathesisisrare.
CrosslinguisticpatternsStronglysuggestphoneticexplanationsforsoundchange.
But,aretheseexplanationsgroundedinarticulationorperception?
Myviewisthatsoundchangetakesplacebytheautomationofproduction.
1.highfrequencywordsundergosoundchangeearlierthanlowfrequencywords2.assimilationandlenitionstronglyresemblechangesthatoccurinotherhighly-practicedneuromotoractivities.
Aresoundchangesmotivatedbyarticulationorperception?
Ohala2003
Variation:‘essentiallyinfinite’(p.672)
Present-dayvariationparallelssoundchangeExamplesin(2)arepalatalizationsofcoronalbeforeahighfrontvowelorglide.
Ohala:misperceptionCompensationinperception:normalization
Experimentshaveshownlistenerscompensateforthephoneticenvironment.
Beddor etal:/ɛ/- /æ/continuumunderthreeconditions:/ɛd/- /æd//ɛd/- /æd/(subjects heard more/æ/vowels)/ɛn/- /æn/
ListenersparsethenasalizationwiththenasalC.
Misperception:failuretonormalize
Appliestoassimilativesoundchanges.
Presupposespriorarticulatoryvariation.The‘misperception’phonologizes thevariation.
1. Whywouldhighfrequencywordschangeearlierthanlowfrequencywords?
2. Howdomisperceptionsspreadacrossthelexicon?
3. Givenhowfinely-tunedtheperceptualsystemis,whywouldlistenersmakemistakes?
4. Canmisperceptionaccountforglidefortition?
Whywouldlistenersmakemistakes?
Maybetheyarenotmistakes.
Maybeoncethearticulatorychangehasadvancedtoacertainpoint,thereisachangeinwhatthelistenerandspeakertaketobethemainfeatureorlocusofthefeature.
Beddor 2009showsthisforṼNsequencesinEnglish.
Arethereanyregularsoundchangesthathavetobeduetomisperception?
Thetatof:throw>frow
Other‘longdistance’changes:
Czechdialects:p’>t
dissimilation
Latin Spanish glossarbor árbol treerobre roble oakmarmore mármol marblecarcere cárcel prisonlocale lugar place
dissimilation
Latinsuffix–ālis >-ārisWhenthestemendsin/l/
national popularmarginal regularoptimal modular
DissimilationOhala2003:Dissimilationisperceptualhypercorrection
Listenersdoalotofnormalizationorcorrectionwhendecodingspeech.Whenafeaturefromonesegmenttendstospreadoveralargestretchofspeech,listenerslearntodisregardit.
Iftheydisregardafeaturethatisnotredundant,thismightleadtodissimilation.
Iftherhoticqualityofthefinalvowelofarbor isattributedtotheearlier/r/,then/l/mightbeproduced.
Dissimilation1. Featuresinvolvedwouldbethosewhoseacoustic-
perceptualcuestendtospreadoverrelativelylongtimeintervals.Aspiration,glottalization,retroflexion,palatalization,pharyngealization,labialization,etc.
2. Conditioningenvironmentispreserved.
3. Doesnotproducenovelsegments.
4. Occursonlywithinwords.
1. Notusuallylexicallyregular
dissimilation
Lexicallyregularexamplesinvolvelaryngealfeaturessuchasglottatalization oraspiration.AncientGreek:onlyoneaspiratedstopper
wordHausa(Chaddic)andYucatecMayan:cannot
havetwoejectivesperwordunlesstheyareotherwiseidentical
Quecha:onlyoneglottalizedCperword.
dissimilation
Notethattheresultisalwaysa‘plain’C.
CasesdonotoccurinwhichoneoftwoplainCsbecomesaspiratedorglottalized.
Couldthisalsobeareduction?
Metathesis
Metathesis:twosegmentsappeartochangeposition(Hume2004)
Latinriparia >[ribeira]>Sp [riβera] ‘riverbank’
Latin casium >caiso >Sp queso ‘cheese’[keso]
sporadicEnglish:axian ‘toask’>ask[aksian]
Metathesis:ModernHebrewMorpheme surfaceform gloss
hit-nakem hitnakem hetookrevengehit-raxec hitraxec hewashedhimselfhit-balet hidbalet hebecameprominenthit-darder hiddarderhedeclined,rolledoverhit-sader histader hegotorganizedhit-zaken hidzaden hegrewoldhit-calem hictalem hetookpicturesofselfhit-ʃamer hiʃtamer hepreservedhimself
Metathesis:Hume’saccount
1. Thesequenceinvolvesacousticfeaturesthatcanspreadovermorethanonesegment,makingtheirassignmenttoasegmentdifficult.
2. Listeners(andspeakers)tendtointerpretambiguoussequencesintheordertheyaremorefamiliarwith—thesequencethatismorecommoninthelanguage.
Otherchangesmotivatedbyphonotactics
Figure1:PepsiaddusedinArgentinafrom2010.Theaddsays‘drinkPecsi,save.’Downbelowitsays‘drinkPepsialso’.
Spanishsyllable-finalobstruents
FinallabialC:concepto,concepción ‘concept’‘conception’obtener ‘obtain’
FinalvelarC:acción [ks] ‘action’ exacto [ks] ‘exact’técnico [kn] ‘technical’ fragmento [gm]proyecto [kt] ‘project’ signatura [gn]
Brown’spointisthatwhenphonotacticsequencesarerare,speakers/listeners/learnersmayreinterpretthemassequencesthataremuchmorecommon.
Othernon-lenitionchanges
Vowelepenthesishappensinconsonantclustersthat(usually)involvesonorantconsonants.
Vowellengthening
Consonantgeminationhastwosources:CCassimilationrhythm