Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Chapter16:Mid-centuryAmericanphonology:thepost-Bloomfieldians
D.RobertLadd
RevisedApril2018[CorrectedMarch2019]
PreparedforinclusionintheOxfordHistoryofPhonology,editedbyElanDresherandHarryvanderHulst.
1
Abstract
Fromtheearly1940stotheearly1960s,Americanphonology,ledbyBloch,
Trager,andHockett,wasremarkablymonolithic.Itspractitionersnominally
followedBloomfield,butinpracticerejectedhisclearlyexpressedviewsby
treatingtheelementsofphonetictranscription(‘phones’)asanintermediate
levelofabstractionbetweenthephonemeandthephysicalsignal.Theyalso
acceptedastrongversionofwhatHockettcalled‘dualityofpatterning’,
accordingtowhichindividualutterancesaresimultaneouslyarrangementsof
phonemesandarrangementsofmeaningfulunits;thisledtotheirinsistencethat
phonemicanalysishadtobecarriedoutwithout‘mixinglevels’,i.e.withoutany
referencetogrammaticalandlexicalcategories.Amongotherthings,this
principlemadeitdifficulttodealinsightfullywithmorphophonological
regularities.Thiswasperhapsthemajorsourceofdisagreementbetweenthe
post-Bloomfieldiansandthegenerativephonologistswhofollowed;bycontrast,
relianceonphonetictranscriptionrepresentsamajorsourceofcontinuity
betweenthetwo.
2
16.1Introduction
Forabout25yearsinthemiddleofthe20thcentury–roughly1940-1965–
Americanlinguisticswasdominatedbyaremarkablyuniformtheoryof
phonology,whichIwillrefertoas‘post-Bloomfieldianphonemics’.Thetheory
tookonitscanonicalshapeinthe1940samongagroupofscholarswho
identifiedthemselvesasfollowersofLeonardBloomfield,andwasgiven
textbookpresentationinthetwomajorintroductorytextsoftheperiod(Gleason
1955a/1961andHockett1958).ThelinktoBloomfield’sownideasabout
phonologywasactuallyquitetenuous,butthesenseofcommonendeavour
amongthegroupwasveryreal.
Themostinfluentialmembersofthisgroupduringtheperiodunder
considerationwereundoubtedlyBernardBloch(1907-1965),GeorgeTrager
(1906-1992),andCharlesHockett(1916-2000).Othercontributorstothis
phonologicalworldviewincludedMartinJoos(1907-1978),HenryAllanGleason
Jr.(1917-2007),HenryLeeSmith,Jr.(1913-1972),RulonS.Wells(1918-2008),
andWilliamG.Moulton(1914-2000).1Thesemen(thecastofcharactersis
virtuallyallmale)weremostlyscatteredinlanguagedepartmentsatuniversities
allovertheeasternhalfoftheUnitedStates,buttheywereunitedbytheir
almostmissionaryenthusiasmforthenewscientificlinguisticspropoundedin
Bloomfield’sLanguage(1933),andbeginninginthemid-1930s2theyalsohada
regularopportunitytoworktogetherinformallyattheLinguisticSocietyof
America’sannualsummerinstitutes.Manyofthemalsosharedtheexperienceof
workingonstrategicallyrelevantlanguageteachingmaterialsduringtheSecond
WorldWar(seeHall1991;Murray1994:144-151).
1Joosseemstohaveplayedavarietyofbehind-the-scenesrolesinmid-centuryAmericanlinguistics.Hewasinfluentialinbringingwartimeresearchonthesoundspectrographtotheattentionoflinguists(Joos1948)andbringinglinguisticstotheattentionofengineers(Joos1950);healsoeditedtheinfluentialcollectionReadingsinLinguistics(Joos1957).TheinfluenceofGleason’sworkbookofanalysisproblems(1955b),whichinitiatedagenerationoflinguisticsstudentsintophonemicandmorphophonemicanalysis,shouldnotbeunderestimated.SmithcollaboratedwithTragerforseveralyearsandtheirOutlineofEnglishStructure(1951/1957)wasthestandarddescriptionofAmericanEnglishforatleastadecade.WellsandMoultonarelesscentralbutspecificcontributionsoftheirsarediscussedlaterinthechapter.2Itistruethattheinstitutesstartedin1928,butafter1931therewasahiatusuntil1936.Atthetimeofthe1928-1931institutes,manyofthescholarswearediscussingwerestillinhighschoolorcollege.
3
Thereweretwootherscholarsworkingonphonologicalquestionsin
NorthAmericaatthetimewhorequireseparatemention.KennethPike(1912-
2000)raisedimportanttheoreticalobjectionstothemainstreamBloch-Trager-
Hockettorthodoxythroughoutthe1940sandearly1950s,buthewastreatedas
anoutsider,oratbestasadevil’sadvocate(Hockett1949);Ireturntodiscuss
hiscontributionsinsection16.3.4.ZelligHarris(1909-1992)wasaregular
participantinthetheoreticaldiscussionsofthe1940sand1950sand,as
Chomsky’sPhDsupervisor,providesalinkfromthepost-Bloomfieldianperiodto
whatfollowed.Severalfriendlycriticshavesuggestedtomethatmyaccount
hereunjustlyslightsHarris’swork,buttherearebothpracticalandsubstantive
reasonsformyfocusonBloch,TragerandHockett.Thepracticalreasonisthat
Harris’scontributionsareclearlysketchedinDresherandHall’schapteron
“Developmentsleadingtowardgenerativephonology”(thisvolume).The
substantivereasonismyviewthat,duringmostoftheperiodunderdiscussion,
Harriswasmorerespectedthangenuinelyinfluential.Hockett’sandGleason’s
textbookpresentationsofphonologyowemuchtoBlochandTragerandlittleto
Harris;Harris’sinfluencewasprimarilyonsubsequentdevelopments,whichis
whyheismoreappropriatelydiscussedinconnectionwiththeoriginsof
generativephonology.However,Iconcedethatmyunderstandingofthe
intellectualhistorymaybedistortedbymyownexperience.3
TherapidgrowthofAmericanlinguisticsimmediatelyfollowingthewar
involveddevelopmentsacrossthewholebreadthofthefield.Americanlinguists
3Althoughitisneithernecessarynorusualtoprovideanaccountofone’squalificationsforwritingonaparticulartopic,IofferabriefautobiographicalfootnotetoanchormyownphonologicalcareerintheeraIdiscussinthischapter.IstudiedlinguisticsasanundergraduateatBrownUniversityfrom1964to1968,wheremyteachersincludedFreemanTwaddell(1906-1982)andNelsonFrancis(1910-2002)andwheremytextbooksincludedbothHockett(1958)andtoalesserextentGleason(1955a/1961).Amongotherthings,thisundoubtedlymakesmeoneoftheyoungestlinguistsstillworking(b.1947)tohavebeentaughttheTrager-SmithanalysisofEnglishphonology(Trager&Smith1951/1957)ascurrenttruthratherthanhistoricalcuriosity.From1970-1972and1975-1978IwasaPhDstudentatCornell,andthereIspentalotoftimeinclassesandindiscussionwithHockett,thoughhewasneverofficiallyinvolvedinsupervisingmygraduatework.BothTwaddellandHockettwerefairlyfreewithanecdotesandopinionsabouttheircolleaguesandaboutthefield,andmyunderstandingofwhatwentonintheheydayofpost-BloomfieldianphonemicsisunquestionablycolouredbythingsIheardfromthemduringmytimeastheirstudent.Someoftheflavouroftheinteractionsamongthemid-centurypost-BloomfieldianscanbegleanedfromJoos’sobituaryofBloch(Joos1967),fromHockett’sobituaryofTrager(Hockett1993),andfromtheextensivecommentsinJoos1957;seealsothesummaryinMakkai1972:3-6.
4
continuedtoengageingeneralistfieldworkbothasacontinuationofthe
Boas/Sapirtraditionofanthropologicallinguistics(onwhichseeSilverstein,this
volume)andthroughChristianmissionarywork,andanumberofbasicideasin
morphologyandsyntax,suchasimmediateconstituents(Wells1947a)and
‘morphemealternants’(Harris1942a)or‘allomorphs’(Nida1948),werefirst
clearlyairedatthistime.Nevertheless,muchofthetheoreticalandempirical
focusofmid-centuryAmericanlinguisticslayinphonology,andtherewasaclear
sensethatgenuineprogresshadrecentlybeenmadeinunderstandingsound
systems.ThissenseisconveyedbyBloch’sarticle‘ASetofPostulatesfor
PhonemicAnalysis’(1948),whichattemptedtocodifytheconsensusthathad
emergedintheprevious15or20years.InhisintroductionBlochemphasizes
thatheisnotpresentinganythingnew,butsaysthatastatementofpostulatesis
‘aformofstocktaking:apausing,atsomecrucialpointinthedevelopmentofa
science,tolookmorecloselyintothesubstructureofitsmethodsandtorepair
whateverlogicalflawsmayappearinit’.Inhisconclusion,havingfinally
presentedthedefinitiveversionofhisphonemictranscriptionofoneshort
utterance,hesays:
Toaskwhetherthisisa‘correct’ora‘true’transcriptionofthegiven
utteranceismeaningless.Thetranscriptioneitheris,orisnot,anaccurate
recordofthephonemesthatwebelievetooccurinthisutterance,andof
theirorder.Andtheanalysisonwhichthetranscriptionisbasedeitheris,
orisnot,inaccordwithoursetofpostulates.
Anyobjectionstosuchatranscription…mustthereforebestatedand
answeredwhollyintermsofthesepostulates.Whoeverprefersa
differenttranscription…mustshoweitherthatouranalysisviolatesone
ormoreofthepostulatesthatwehavestated,orelsethatthese
postulatesareuntenable.Ifhetakesthelatterposition,wemay
reasonablyaskhimtostatehisownassumptionsinequaldetail,andput
offallargumentuntilhehasdoneso.
Hisconfidenceintheimportanceofthenewsynthesisisclearlyconveyedbythe
uncompromisingtoneofthesefinalparagraphs.
5
InthischapterIwillsketchthemaintenetsofpost-Bloomfieldian
phonemics,drawinginparticularontheworkofBloch,Trager,andHockett,on
Pike’scritiquesoftheirwork,andonJoos’scommentariesonthearticles
anthologizedinReadingsinLinguisticsI(Joos1957),aswellasonmyowndirect
experience.Iusetheterm‘phonemics’todrawattentiontothefactthatthe
theorycoveredarathernarrowsubsetofthephenomenathatmightreasonably
beregardedasfallingwithinthescopeof‘phonology’.Amongotherthings,Iwill
attempttoexplainwhatwasbehindthesescholars’willingnesstopromotesuch
anarrowperspectiveonsoundpatterns.
16.2Phonemes,allophones,andthephone
Itisrathergenerallytakenforgrantedthatgenerativephonologysuperseded
post-Bloomfieldianphonemicsandthatwhatwentbeforeisnowofprimarily
historicalinterest.However,thereisagooddealofcontinuitydespitethestormy
theoreticalupheavalsofthe1960s.Eventhoughtheirtheoreticallegitimacyhas
neverreallyrecoveredfromthegenerativecritiqueof‘taxonomicphonemics’
(especiallyChomsky1964,butalsoHalle1959),thebasicdescriptiveconstructs
ofpost-Bloomfieldianphonemictheoryarestillthoroughlyfamiliar.Thisis
because,ignoringforamomentthespecificaspectsofthemid-century
orthodoxythatattractedChomsky’sandHalle’scriticalfire,someversionofthe
phonemicprincipleremainsattheheartofphonology(seeSchane1971foran
earlygenerativestatementofthisassessment).
Thecorephonemicideamaybesummarizedasfollows.Thesound
systemofalanguageinvolvesaninventoryofabstractunitsofsoundthatstand
incontrastwithoneanotherandthatcanberealizedphoneticallyindifferent
waysdependingonthephoneticand/orstructuralcontext.Intheterminologyof
thepost-Bloomfieldians,theabstractunitsarephonemesandthedifferent
realizationsaretheirallophones;theirversionofcontextuallyconditioned
variationwasconceptualizedascomplementarydistribution.Standard
examplesfromEnglisharestillusedinbeginninglinguisticscoursesthroughout
theAnglophoneworldtoillustratethesefundamentalconcepts.Theyincludethe
differencebetweenclearanddarkallophonesof/l/foundinmanyvarietiesof
6
English(clear[l]inonsetpositioninasyllable,dark[ɫ]incodaposition)andthe
differencebetweenaspiratedandunaspiratedallophonesofthevoicelessstops
(aspiratedstopsinabsoluteinitialposition,unaspiratedstopsfollowingsyllable-
initial/s/).Somesuchnotionofcontextuallyconditionedvariationinthe
phoneticrealizationofabstractphonologicalelementsremainscentraltothe
phenomenathatoccupyphonologists’attention.
Alessobviousbutequallyimportantelementofthepost-Bloomfieldian
synthesis,whichalsoremainscentraltomuchphonologicalthinking,isanother
abstraction:thephoneticsegmentorphone.Thetheorytookforgrantedthe
scientificvalidityofasegment-basedidealisedphoneticrepresentationofspeech.
Ihavediscussedtheemergenceofthephoneconceptanditsrolein20thcentury
phonologyatgreaterlengthelsewhere(Ladd2011),andIonlybriefly
summarizethatdiscussionhere.WiththeexceptionofPike’streatiseon
phonetics(1943),nooneinvolvedindevelopingphonemictheory–oneither
sideoftheAtlantic–seemstohaveworriedmuchaboutthebasisofthephone
idealization.Pikestatedtheproblemsuccinctly:
Speech,asphoneticianswellagree,consistsofcontinuousstreamsof
soundwithinbreathgroups;neithersoundsnorwordsareseparated
consistentlyfromoneanotherbypauses,buthavetobeabstractedfrom
thecontinuum.Phonemicistsconcurinthebeliefthatsomeunitofspeech,
thephoneme,canbediscoveredasthebasicconstituentofalinguistic
system.…Isthereasignificanthalfwaypointbetweenthecontinuumand
thephoneme?Isthereareal,nonfictitioussegmentofsoundwhichisnot
aphonemicone?(p.42)
Hedevotedseveralpagesofhismonographtodeveloping‘aworkablemethod
forthedelineationofnaturalphoneticsegmentation’inwhichthe‘segmental
unitistobedeterminedentirelyapartfromphonemicfunction’,andhe
understoodthathavingsuchamethodwouldmeanthat‘animpressionistic
phoneticrecordofanewlanguageprovestheoreticallylegitimateaswellas
practicallyvaluable…forthephonemicist…’(p.53,emphasisadded).However,
forthemostpartthe‘phonemicists’continuedtouseastringofsegments–
7
phones,thatis–astheirrepresentationoftheprimaryspeechdatawithout
worryingverymuchaboutitstheoreticallegitimacy.Blochbuiltthetheoretical
legitimacyofphonetictranscriptionintohisPostulate11:
Theseriesofperceptiblearticulationsofanygivenvocalorganduringan
utterancecanbedividedwithoutaresidueintosuccessiveparts….In
[this]postulatewedonot[disregardtheinstrumentalevidencethatthe
articulatorsareincontinuousmovement];rather,weimplythata
phoneticallytrainedobservercaninterprettheauditoryfractionsofan
utteranceintermsofarticulationsthatseem(tohisperception)tobe
staticorunidirectional.(Bloch1948:§11.1)
Sincetheemergenceofthephoneconcept,onecomponentofvirtuallyall
phonologicaltheorieshasbeenasegment-basedsymbolictranscriptionof
speech–whatChomsky(1964)called‘systematicphonetics’.Thatis,the
centerpieceofmanyphonologicaltheoriesisamappingbetweenasymbolic
representationexpressedintermsofabstractelementssuchasphonemesor
featuresandanothersymbolicrepresentationexpressingthephoneticdata.This
characterisationappliesveryclearlytoPragueSchoolphonology,tothepost-
Bloomfieldians,toclassicalgenerativephonology,andtomorerecent
developmentsfromclassicalgenerativephonologysuchaslexicalphonologyand
variousversionsofOptimalityTheory.Thesedifferintheirconceptionofthe
mapping(derivationalordeclarative?),inthenatureoftheabstractelements
(phonemesorfeatures?‘taxonomic’or‘systematic’phonemes?),andinvarious
otherways,butallassumethattheprimaryphoneticdatacanbeexpressedin
termsofasegmentedsymbolicrepresentation.
Giventhisconceptionofphonology,thestudyofhowsystematicphonetic
representationsaremanifestedinquantitativeacousticandarticulatorydatais
relegatedtothephoneticslaboratoryandadifferentsetoftheoreticaland
methodologicalprecepts.(Ineffect,thephoneconcepthasrelievedgenerations
ofphonologistsoftheneedtocometogripswithfinephoneticdetail;see
Pierrehumbert&Beckman1988:1-5forsomediscussionofthispoint).Themost
prominentcurrentexceptiontothisgeneralizationisArticulatoryPhonology
8
(Browman&Goldstein1986,1989,andnumerouscolleaguesandfollowers
sincethen),whichassumesthatthesurfacephoneticdata–the‘output’,asit
were–canbestbeexpressedintermsofquantitativemodelparameters.Butthe
centralroleplayedbythephoneinpost-Bloomfieldiantheoryisotherwise
widelyreflectedmorethanhalfacenturylater.
Itisworthpointingoutthattheideaofanchoringphonologytothephone
seemsnottohavebeensharedbyBloomfieldhimself.Infact,Bloomfielddrew
attentiontothemanifoldinconsistenciesofclosephonetictranscriptionand
insistedthat‘onlytwokindsoflinguisticrecordsarescientificallyrelevant.One
isamechanicalrecordofthegrossacousticfeatures,suchasisproducedinthe
phoneticslaboratory.Theotherisarecordintermsofphonemes,ignoringall
featuresthatarenotdistinctiveinthelanguage…’(Bloomfield1933:85;andcf.
Hockett1965,footnote23).Amonghisfollowers,however,thephoneoccupieda
centraltheoreticalrole.Speech–theprimarydata–wasassumedtoconsistofa
stringofphones,andinsomesensethegoalofphonologicalanalysiswasto
assignphonestophonemes.Theterm‘complementarydistribution’isespecially
revealing,becauseitreferstothedistributionofphones.Phoneswerereal;
phonemeswereanalyticconstructs4.
Thebroadacceptanceofsystematicphoneticsbyboththepost-
Bloomfieldiansandtheearlygenerativistsisamajorpointofcontinuity.Itseems
tomethatitoutweighsanissuethatbecameoneofthemajorpointsofdispute
betweenthem,namelythequestionofwhetherphonemes(andallophones)
wereatomsorwerethemselvescomposedofsmallercomponentsthatarenow
universallyknownasdistinctivefeatures.Thereareatleasttworeasonsfor
seeingthelatterissueassecondary.First,thepost-Bloomfieldianswerecertainly4Thisisnottosuggestthatthephonemicistswereunawarethattherelationbetweenphonesandthespeechsignalmightberathercomplex.Joos,asnotedinfootnote2,wroteamonograph(1948)aimedatinforminglinguistsofthenewdevelopmentsalmostassoonaswartimesecrecysurroundingthesoundspectrographwaslifted.Hockett,inhisManualofPhonology(1955:180-211),attemptedtocomegripswiththefactthat,despitethe‘naturalaprioriexpectation…[that]eachallophonewillappearonaspectrogramasarecognizablepattern…theresultsofactualexperimentationarestaggeringlyatoddswithanysuchexpectations.’Nevertheless,theydidnotabandonthephoneidealization,onlysoughttorefineitorjustifyit.Blochinparticular(postulate9)insistedthat‘anobservercanbetrainedtomakeaphoneticdescriptionoftheutterancesofanydialect,orofasufficientsamplethereof,withouttheaidoflaboratorydevices,thatwillbeadequateandvalidforthepurposesofphonologicalanalysis’(1948:§9.1).
9
awareofworkbyTrubetzkoy,Jakobson,andothers,andconsideredwhatplace
featuresmighthaveintheirdevelopingunderstandingofthephoneme;Bloch
devotesanentiresectionofhis1948articletosixpostulates(§§47-52)on
‘features’.Second,despitetheostensiblegroundingofthedistinctivefeaturesin
thephysicaldetailsofspeech,theactualpracticeswithrespecttophonetics
differedlittlebetweenmostgenerativephonologistsandmostoftheir
predecessors.Instead,themorefundamentalriftbetweenpost-Bloomfieldian
andearlygenerativeviewsinvolvesthedegreeofabstractioninvolvedin
definingthephoneme,andtheconsequentrelationbetweenphonemes(whether
post-Bloomfieldian‘taxonomic’phonemesorgenerative‘systematic’phonemes)
andthephonesthatmakeupthesurfacephoneticrepresentation.Thisisthe
threadrunningthroughthetopicsdiscussedinthenextsection.
16.3Theheartofthematter
16.3.1Biuniqueness
Withinpost-Bloomfieldianphonemics,thetheoreticalimportanceofthephone
wasenhancedorexaggeratedbytheprincipleofbiuniqueness.(Thisterm,
whichisalsousedinmathematics,seemstohavebeenfirstusedinconnection
withphonologybyHarris,e.g.1942b,1944.)Accordingtothisprinciple,every
phonemeofalanguagecanbemanifestedbyavarietyofphones,following
regulardistributional(‘allophonic’)rules;everyphoneinanutterancecanbe
identifiedasthephoneticmanifestationofaspecificphonemeofthelanguage;
andcrucially,inagivencontextnophonecanmanifestmorethanonephoneme.
Together,theseprinciplesmeantthatthereisa‘biunique’orone-to-one
correspondencebetweenanystringofphones(i.e.any‘phoneticrepresentation’,
inpresent-dayterminology)andanystringofphonemes.Forexample,giventhe
predictableallophonicdistributionofaspiratedandunaspiratedmanifestations
ofvoicelessstopsinEnglish,thephonemicstring/kɪt/necessarilybeginswith
thephoneticsegment[kh]andthestring/skɪt/necessarilyincludesthephonetic
segment[k].Intheotherdirection,thephoneticsegment[k]followingthe
phoneme/s/in/skɪt/canonlycorrespondtothephoneme/k/.
10
ThefirstclearstatementofthisviewcameinBloch’sshortpaper
‘Phonemicoverlapping’(1941).Hebeganbydiscussingseveralcasesinwhich
therealizationsofphonemesmightappeartointersect,butwhichcanbe
analysedinotherways,orwheretherealizationsaredistinguishedbybeingin
complementarydistribution.(Hisexampleofcomplementarydistribution
contraststhealveolarflapasamanifestationofthephoneme/r/after/θ/asin
threeandasamanifestationof/t/betweenvowelsasinbutter.)Hethen
developshis‘mostseductiveexampleofapparentintersection’,involving
ostensiblyallophonicvariationinvowellengthinsomevarietiesofAmerican
English.Itisnecessarytoquotehimatsomelengthtoconveytheessenceofthe
argument.
Thepairsofwordsbitbid,betbed,batbad,butbud,bitebide,beatbead,etc.
haverespectivelythesamevowelphoneme,butexhibitaregularandfairly
constantdifferenceinthelengthofthevowelallophones.…Thealternation
betweenlongerandshorterallophonesrunsthroughthewholephonemic
system.Thevowelofpotisaffectedbythesameautomaticalternation:…
thereisnothing,sofar,toshowthatapairlikepotpodisnotineveryway
comparabletobitbid.
Inmyspeechbombisdifferentfrombalm,botherdoesnotrimewithfather,
andsorrydoesnotrimewithstarry:thevowelqualityisthesameinallthese
words,butinthefirstwordofeachpairthevowelisshort(justasitisin
pot),andinthesecondnoticeablylonger.Sincethedifferenceinlength
cannotbeexplainedasanautomaticalternation(likethedifferenceinbit
bid),weconcludethatbombandbalm,botherandfather,sorryandstarry
havedifferentvowelphonemes;andwenaturallyidentifythevowelofbomb,
bother,sorrywiththephonemeofpot.Thevowelofbalm,father,starry
appearsalsoinalms,palm,pa,star,card.Again,thereisnothing,sofar,to
showthatthephonemicorganizationisinanywayabnormal.Butnow
comesahitch.
InthesentencePa’dgo(ifhecould),theutterancefractionpa’dmustbe
analyzed,accordingtowhatwehavejustsaid,ascontainingthephonemeof
11
balm.InthesentenceThepodgrows,theutterancefractionpodmustbe
analyzed,againaccordingtowhatwehavesaid,ascontainingthephoneme
ofpot.Butpod…isphoneticallyidenticalwithpa’d!Twooccurrencesofx
underthesameconditionshavebeenassignedtodifferentphonemes.
Blochthengoesontoconcludethattheonlysolutiontotheproblemisto
concludethatthevowelsofpotandpodarenotallophonicvariantsofthesame
phoneme,butmustbephonemicallydifferent.Bydoingso,heacknowledges,‘we
destroytheneatparallelismofthepairsbitbid,bet,bed,bitebide,potpod…[b]ut
bysacrificingthissymmetryweareabletoaccountforallthefactsof
pronunciation,whichissurelythemoreimportantrequirement.’
Inpractice,atleastoneclearclassofpartialexceptionstobiuniqueness
wasacknowledgedbythepost-Bloomfieldians,namely‘freevariation’.Theusual
exampleinAmericanEnglishwasthevariationbetweenreleasedandunreleased
finalstops.Itwasacknowledgedthatboth[bith]and[bit̚]reflectthesamestring
ofphonemes(/biyt/intheTrager-Smithanalysis)andthatthevariation
betweenthetwotypesofutterance-finalstopwasofnophonological
significance.BlochdiscussedfreevariationinhisPostulates(§§27-29).
Otherphenomenathatmighthaveledtoaweakeningofthepreceptof
biuniqueness,however,didnotdoso.Themostobviousexampleis
neutralization.IfweassumethatAmericanEnglish‘flapping’neutralizesthe
/t~d/distinctioninpairslikebettingandbeddingorlatterandladder,wecannot,
giventheprincipleofbiuniqueness,treatthisastheconsequenceofallophonic
ruleswherebyboth/t/and/d/arerealizedintervocalicallyasaflap(whichIwill
representhereas[ɾ]).Thetheorywouldrequirethephone[ɾ]tobeassigned
unambiguouslyasthephoneticmanifestationofonephonemeonly.Suchan
analysiswasrigorouslyappliedinthepronunciationtranscriptionsgiveninthe
ThirdInternationaleditionoftheunabridgedMerriam-Websterdictionary
(1961),wherewordsnormallypronouncedwithaflaparetranscribedwith/d/;
thischoiceisjustifiedatsomelengthinthedictionary’s‘Guidetopronunciation’
(p.41a).Thisanalysismeansthatmorphemeslikebethavetwophonemically
12
distinctallomorphs,/bɛt/and/bɛd/,andthealternationbetweenthetwowould
havetobetreatedasamatterofmorphophonemicsormorphology5.
16.3.2Separationoflevels
Thearchitectsofthetheorydidnotregardtheseconsequencesasflaws.Onthe
contrary:toputitinmodernterms,theysawthetheoreticalconsequencesof
biuniquenessnotasabugbutasafeature.Counterintuitiveconclusionslike
Bloch’sanalysisofpotandpod,whichmightreasonablyberegardedasa
problemorevenareductioadabsurdum,weresometimesheldupasevidenceof
theinsightaffordedbyrigorousadherencetothelogicofthetheory.Inthe
specificcaseofphonemicoverlapping,theinsightwastheprinciplethat
generallywentbythenameof‘separationoflevels’.AsJoosputitinassessing
theimpactofBloch’spaper(Joos1957:96),Bloch‘madeclear,asitneverhad
beenbefore,thatphonemicsmustbekeptunmixedfromallthatliesonthe
oppositesideofitfromphonetics’.Accordingtothisprinciple,thesoundsystem
ofalanguagewastobeanalysedentirelyinitsownterms,withoutanyreference
tothegrammarorthelexicon.Thetwolevelsweretobekeptrigidlyseparate.
Failingtoseparatethemwascommittingthesinof‘mixinglevels’,atemptation
thatproveddifficulttostampout.ToquotefurtherfromJoos’scommentary,‘the
ghostoftheslaindragoncontinuedtoplaguethecommunityoflinguistsunder
suchnamesas“grammaticalprerequisitestophonemicanalysis”[adirectswipe
atPike;seesection3.4below]andhasnotbeencompletelyexorcizedtothis5Mydiscussionofthispointishedgedwithconditionalsbecause,remarkably,thisissuewasseldomdiscussedinpost-BloomfieldianworkeveninanalysesofthephonologyofAmericanEnglish.Neutralization(andtheconstructofthearchiphoneme)loomedlargeintheoreticaldiscussionsinEasternEurope,wherefinaldevoicinginmanylanguagesmeansthatneutralizationiswidespread,buttheloneexampleofflappinginAmericanEnglishwassimplyignoredorassumednottoinvolveneutralizationatall.Bloch’sdiscussionofphonemicoverlapping(1941)clearlyimpliesthatthe‘alveolarflap’inbettingorkittycontrastswith[d]inbeddingandkiddy.Similarly,Trager&Smith’sanalysis(1957)claimsonp.32that‘Inbutterwefindavoicedfortis[t̬]inmostAmericanspeech’,andonp.34mentiononly[d]asamedialallophoneof/d/.Doggedempiricalinvestigationbye.g.Haugen(1938)andOswald(1943),showingclearlythatintervocalic/t/and/d/wereindistinguishablebylisteners,seemstohavehadlittleimpactonpost-Bloomfieldiantheoreticalthinking.BythetimeofGleason’stextbook,however,wefindaclearstatementthat‘insomedialectsthecontrastbetween/t/and/d/hasbeenlostbetweenvowels’.Gleasonlaysoutthepossibleanalyses,notingthat‘Americanlinguistsgenerallyhavepreferredtomatch[theflapwitheither/t/or/d/butnotboth]whereverpossible;someEuropeans,tomaintainunitslike/T/inwhichthevoiced:voicelesscontrastissaidtobeneutralized.’(Gleason1961:295).ThesolutioninWebster’sThirdiscertainlyofGleason’sfirsttype.
13
day’6.
Seenfromthepointofviewofanindividualhearingasentence,theidea
ofseparatingphonologyandgrammarmakesacertainamountofsense.The
phonemicists’assumptionwasthathearersmustfirstidentifythephonemes
theyarehearingandthenparsethemintowords.Thiswasclearlystatedby
Wells:
Phonemicstakesthepointofviewofthehearer.Nowthehearer,inorderto
interpretcorrectlyanutterancethathehears,mustrelyontwoseparate
sourcesofinformation:(a)theheardsounds…;(b)theextra-linguistic
context….[P]honemicsmakesapointofrecordingnothingbutwhatis
conveyedby(a).Allelsebelongstogrammar(andlexicography.)(Wells
1947b:271.)
Ifthisisindeedthehearer’stask,thenthereisaconundrumunlesssomething
likebiuniquenessisatwork.Considertheinputstring[bɛɾɪŋ].Confrontedwith
[ɾ]inaphonologywhereitcouldbeanallophoneofeither/t/or/d/,thehearer
hasnobasisfordecidingwhichphonemeisintendedwithoutknowingwhether
theintendedmessageisbettingorbedding.But–giventheassumptionthat
sentenceunderstandingproceedsstrictlyfromthebottomup–theonlywayto
knowtheintendedmessageistohaveidentifiedthephonemesfirst.The
circularityisapparent.
Theimportanceofthehearer’spointofviewwasathemeofHockett’s
thinking7.Forexample,partofhisresponsetothesuccessofChomsky’searly
6ThemoralistictoneofJoos’scommentsherecanbefoundrepeatedlyinmid-centurydiscussionsofphonemictheory.Laterinthesamecommentary,forexample,Joosgoesontopreachthevirtuesofpost-Bloomfieldiandescriptivism:‘Trubetzkoyphonologytriedtoexplaineverythingfromarticulatoryacousticsandaminimumsetofphonologicallaws…,andoffer[ed]toomuchofaphonologicalexplanationwhereasobertaxonomywouldserveaswell.Childrenwantexplanations,andthereisachildineachofus;descriptivismmakesavirtueofnotpamperingthatchild.’Trager’sreaction(1950:158)toPike’smonographPhonemics:Atechniqueforreducinglanguagestowriting(Pike1947a)echoessimilarthemes:he‘condemn[s]thebookasatheoreticalwork,andevenmoreasatext-book–sinceasthelatteritwillleadastraymanywhomightotherwisebevaluableworkersinlinguisticscience.’NotealsoBloch’sowncharacterizationofhispodpa’dexampleas‘seductive’.
7WhileIwasatCornell,Hockettsuggestedmorethanoncethatanimportantdifferenceinoutlookbetweengenerativeandpost-Bloomfieldianlinguisticswasthattheformertookthepointofviewofthespeakerandthelatterthepointofviewofthehearer.However,exceptfor
14
workwasanarticleentitled‘GrammarfortheHearer’(Hockett1961a),inwhich
heargued,instrikinglycontemporary-soundingterms,thatforthehearer,
parsingcanindeedbeseenasastochasticprocessandmodelledbyafinitestate
grammar.
‘Chomskyhasshownthat,ifweacceptcertainveryreasonableempirical
assumptionsaboutEnglish,thenEnglishisnotafinitestatelanguage.He
hasalsoclaimedthatnofinitestateapproximationtoEnglishcanmatch
theknownfactsofthelanguagecloselyenoughtobeofanyinterest.This
secondpointis,Ibelieve,false.Itwillbeshownlaterinthispaperthatit
isintheorypossibletomatchthefactsofEnglishascloselyaswewish
withafiniteMarkovchain.’(1961a:220)
Yetfromtheoutsetofthearticleheeffectivelyexemptedthesoundsystemfrom
thisoutlook,explicitlyadoptingthefollowingassumption:‘Theheareralways
hearscorrectly:thatis,hehearstwowordsorsequencesofwordsasdifferentif
andonlyiftheyarephonemicallydistinct.’(1961a:221).Hedescribedthis
assumptionas‘customary’,thoughhealsonotedthatitis‘rarelystated
explicitly’.ThisassumptionencapsulatestheviewexpressedbyWellsasquoted
justabove:hearersmustfirstidentifythephonemestheyarehearingandthen
parsethemintowords.Giventhisassumption,theinputtothestochasticprocess
isnotanacousticsignalbutaphonemicrepresentation.
Withhindsightbasedondecadesofpsycholinguisticresearch
(comprehensivelysummarizedbyCutler2012),itisclearthatthisassumption
cannotbemaintained.Speechprocessing–allspeechprocessing,notjust
grammaticalparsing–isinmanyrespectsastochasticprocess.Weknownow
thathearerscontinuouslygeneratehypothesesabouttheincomingspeech
stream,andtrytoidentifywordsastheycomein;theydonot–ornot
necessarily–trytoidentifyphonemesfirst.Indeed,asanyonewhohasever
listenedtothestimuliina‘gating’experimentknows,itisoftendifficultor
impossibletoidentifyphonemesuntilafteronehasidentifiedthewords.the1961apaper,hedidnottomyknowledgedevelopthisideauntilmanyyearslater(Hockett1987).
15
Furthermore,inalmostanycontext,itwillbeclearwhether[bɛɾɪŋ]isintendedto
representbettingorbedding;thephonemicidentityofthe[ɾ](totheextentthat
thisevenmatterstothehearer)followsfromthat.Buttheassumptionthat
hearerssomehowhearphonemesdirectlyleftthepost-Bloomfieldiansunableto
acknowledgetheinterdependenceofthephonologicalandthegrammatical
levelsoflinguisticstructure.8
16.3.3Morphophonemics
Nowhereistheinterdependenceofphonologyandgrammarmoreobviousthan
inthecaseofphonologicallyconditionedalternationsintheformsofwordsor
morphemes.Theprincipleofbiuniquenessmadeitimpossibleforpost-
Bloomfieldianphonemicstotreatmostautomaticalternationsaspartofa
unifiedaccountofphonology.Neutralization,aswesawearlier,couldnotbe
analysedintermsoftwodifferentphonemesconvergingonthesameallophonic
realization;itwastheoreticallynecessarytotreattheneutralizedallophoneas
belongingonlytoonephonemeortheother.Inthesameway,phonologically
exceptionlessalternationslikethe[-s]/[-z]/[-ɨz]variationintheformofthe
Englishpluraland3rdpersonsingularmorphemescanonlybetreatedaspartof
morphologyoraspecificmorphophonemiccomponentofthegrammar,notas
partofthephonology.Oncethetheoreticaldefinitionofthephonemewas
established,thedescriptiveconsequencesforneutralizationandotherautomatic
alternationsdidnotleadtoanyreconsideration.
Thisisnotaprioriunreasonable.Forexample,asdiscussedbyIosad(this
volume),theMoscowandLeningradschoolsofSovietlinguisticsdiffered
preciselyonthematterofwhattodoaboutneutralization;Leningradprivileged
thesurfacesoundofaneutralizedsegment,andpositedasinglephonemic
representation,whileMoscowfocusedonlexicalidentity,andacceptedthat
distinctphonemescouldberealizedinthesameway.Iwouldstatethe
underlyingquestionhereasfollows:isthesoundsystemofalanguageprimarily
8Aresidualsensethatphonologicalandgrammaticaldescriptionshouldbekeptseparatemayberesponsibleforthefactthat,severaldecadeslater,Kaisse’sstudyofconnectedspeech(1985)wascriticizedformakingdirectreferencetosyntacticstructure,andthatmostsubsequentworkontheseproblems(e.g.Selkirk1984,Nespor&Vogel1986,andmuchworksincethen)haspositedasimilarbutseparate‘prosodicstructure’inthephonology.
16
asetofphoneticgeneralizationsoverthelanguage’slexicon,orisitatemplate
thatplaysaroleinguidinglisteners’perceptsofspeechsoundsindependentlyof
thelexicon?Thereisplentyofevidenceforbothpositions;thepost-
Bloomfieldianscamedownfirmlyonthesideofphoneticidentityratherthan
lexicalidentity.
However,therewasplentyofdiscussionbeforegeneralagreementonthis
issuewasreached.Bloomfieldhimself,notablyinhis1939paperonMenominee,
distinguishesclearlybetweenalternationsthatarephonologicallyconditioned
andthosethatarelexicallyidiosyncratic,andseemstosuggestthatthe
phonologicallyconditionedonesarepartofphonetics;heexplicitlystates–only
twoyearsbeforethepublicationofBloch’spaperbanningphonemicoverlapping
–thatthereis‘someoverlappingbetweenphonemes’(1939:§38).Hockett
(1993:787)saysthat‘earlyon[Trager]believedthatmorphophonemicsbelongs,
withphonemics,inphonology;partlyundermyinfluence,helaterchangedthat
opinion.’Joos(1957:92)describesthedebatesaboutthisissueasfollows:
WhenwelookbackatBloomfield’swork,wearedisturbedatthisandthat,
butmorethananythingelseBloomfield’sconfusionbetweenphonemesand
morphophonemesdisturbsus.Bloomfieldkepthimselfoutoftroublehere,
usually,bydescribingjustonelanguageatatime,oroneareawithineachat
atime,adjustingfortheeffectsoftheconfusion.…
Theescapefromthisconfusionwas,naturally,itselfaconfusedaswellasan
arduousjourney,likethatoftheIsraelitesfromEgypt.Mostofitremains
undocumented,consistingofendlesshoursofdiscussionandofprivate
pondering.…
Hethengiveshisowndefinitionofthedistinction,whichagainmakesreference
toassumptionsaboutwhathearersdo:
Thenativelistenermaybesaidtoperceive–tosomehowexploitfor
message-understandingends–itemsinwhathehears.Insofarasthis
processdoesnotdependonunderstanding,theitemsarephonemic;insofar
asitemscannotbeperceivedwithoutunderstanding,morphophonemicsat
least(perhapsmore)isinvolved.IborrowanexamplefromC.F.Hockett:
17
Onceheheardsomeonesay‘Shehaspoise’and,momentarilyinsufficiently
attentive,innocentlysaid‘What’sapoy?’Thephonemicitemshadbeen
apprehendedperfectly,but,throughalapseinunderstanding,the
morphophonemicitemshadnot.
Nevertheless,althoughtheissuewastemporarilysettled,theplaceof
morphophonemicsinpost-Bloomfieldianworkremainedawkward.Gleason’s
textbookmentionsthetermmorphophonemics(ormorphophonemicrules)at
severalscatteredplacesinitscoverageofmorphology,butitnevermentionsthe
issueofwhythelinebetweenphonologyandmorphologyisdrawnwhereitis,
simplytakingitforgrantedthatthemorphemealternantsunderdiscussionare
phonemicallydifferent.Hockett’stextbookprovidesslightlymoreunified
coverageoftheactualphenomenainagroupofchaptersentitled
‘Morphophonemicsystems’,andfurthermoremakesaseriousattempttospecify
theplaceofmorphophonemicsintheoveralldesignoflanguage(1958:137-
144);cf.alsothemoretechnicaldiscussioninHockett(1961b).Butforboth
writersthedefinitionofthephonemeinvolvingbiuniquenesswastakenas
settled,andthetreatmentofmorphophonemicsfollowedasaconsequence.
ThiswasthebasisofHalle’sfamouscritiqueofthepost-Bloomfieldian
phonemeinTheSoundPatternofRussian(1959)9.Underthephonemicists’
definitions,phonologicallyconditionedautomaticvoicingalternationsinRussian
maybetreatedinafewcasesasinvolvingthecomplementarydistributionof
allophones,butmustbetreatedinmostcasesasmorphophonemiceffectsonthe
phonemicrepresentationofmorphemes.MostRussianobstruentsarepartof
phonemicallydistinctvoiced/voicelesspairs(/pb,td/etc.),andinthesecases
automaticvoicingalternationwillchangethephonemicrepresentation.Three
obstruentphonemes,however,namely/ts,tʃ,x/,havenocontrastivevoiced
counterpart,soheretheoccurrenceincomplementarydistributionof
correspondingphoneticallyvoicedandvoicelessobstruentscanbeanalysedas
involvingvoicedandvoicelessallophonesofasinglephoneme.Somethingthat
istransparentlythesamefairlylow-levelprocessofphoneticassimilationhasto
betreatedintwodifferentways.Thecaseisalmostperfectlyanalogousto9ForamoredetailedexaminationofHalle’scritiqueseeDresherandHall(thisvolume),sec.3.2.
18
Bloch’spodPa’dexample,exceptthatHalleandBlochdrewoppositeconclusions.
Halleprivilegedtheunityofthephoneticassimilationprocess,andarguedfor
ignoringthepresenceorabsenceofsurfacecontrastsofvoicinginphonemic
analysis.Blochprivilegedsurfacecontrastsandarguedfor‘sacrificingthe
symmetry’ofallophonicdurationeffectsinaphonemicanalysis,insistingthat
thedifferencebetweenpotandpodmustinvolvethesamephonemiccontrastin
durationfoundinbombandbalm.
Whatisnoteworthyaboutthedifferencebetweenthepost-Bloomfieldian
andthegenerativeapproachesisthatbothapproachmorphophonemicanalysis
inverysimilarways–based,asithappens,ongeneralprinciplesthatwere
alreadyclearlyspelledoutbyBloomfieldhimself(1933,chapter13).Both
GleasonandHockettgivesummariesofhowtodescribemorphophonemic
alternationsintermsof‘baseforms’(i.e.underlyingforms)andsketchthe
principlesinvolvedinselectinganappropriatebaseform;Hockett(pp.277-278)
discussesthedifferencebetweeninternalandexternalsandhiandthesensein
whichoneprocess‘precedes’theother;Gleason’sworkbook(1955b)includesa
longsectionofdataproblemsthatgivestudentspracticeinidentifyingbase
formsandusingorderedrules.Themaindifferencebetweenpost-Bloomfieldian
morphophonemicsandearlygenerativephonologyconcernedthetheoretical
statusoftheelementsmanipulatedintheseanalyses,notactualdescriptive
practices.ForsomediscussionofthecontinuityinvolvedhereseeKilbury1976.
16.3.4‘Grammaticalprerequisites’
Oneconsequenceoftheradicalseparationoflevelsespousedbythepost-
Bloomfieldiansisthattherearenowordsinthephonology.Phonemiccontrasts
servetodistinguishoneutterance(or‘utterancefraction’)fromanother;words
arepartofgrammar.BlochbuiltthisassumptionintohisPostulates(especially
§2)andthisiswhy,forexample,hethoughtitrelevantandimportantto
comparepodwithPa’dinhisessayonphonemicoverlapping.Sowhatdowedo
whenwefinddirectphoneticevidenceforgrammaticalboundaries?
Themostobviouscasesofthissortinvolveallophonicvariation
19
apparentlyconditionedbypositioninmorphologicalstructure.Thestandard
exampleatthetimewasthedifferenceinpronunciationbetweennitrateand
nightrate,wheretherearetwodifferentallophonesofthephoneme/t/.Sinceit
wastheoreticallyoutofboundstodefinetheconditioningfactorsforthis
allophonicdifferenceintermsofgrammaticalfactslikewordboundaries,the
solutionwastoposittheexistenceofa‘juncture’phonemebetweenthe/t/and
the/r/innightrate.Thatis,aphenomenonthatmightreasonablybeanalysedas
thephoneticeffectofagrammaticalboundarywasrecastasordinaryallophonic
variationconditionedbythepresenceorabsenceofaphonologicalentity,a
‘phoneme’usuallysymbolisedby/+/andreferredtoas‘plus-juncture’.Thisidea
wasfirsttentativelyproposedbyTrager&Bloch(1941:§4)andwaseventually
widelyadopted,thoughreluctantlybysome(e.g.byWells,whonotedthat‘the
validityofjuncturephonemesisopentogravedoubtsonphoneticgrounds’
1947a:§64).
Forexample,plus-junctureplayedacrucialroleinthestandardpost-
Bloomfieldiananalysis(Moulton1947)oftheGermandiminutivesuffix–chen
([çən]).Ingeneral,[x](or[χ])and[ç]areincomplementarydistribution,the
formeroccurringafterbackvowelsandthelatterelsewhere,including(rarely)
word-initially.Thediminutivesuffixposedatheoreticalproblembyusingthe
palatalallophone[ç]evenafterbackvowels,inawordlikeFrauchen‘mistress’
[inthesenseof‘femaledogowner’].Plainly,theexistenceofthemorphological
boundarybetweenstemandsuffixisrelevantinsomeway,butthiscouldnotbe
acknowledgedindirectmorphologicalterms;instead,aplus-juncturephoneme
wassaidtooccurbetweenthe/au/diphthongandthedorsalfricativephoneme,
conditioningthepalatalallophone[ç].Moulton’sanalysisdrewimmediate
criticalcomment(Leopold1948),andthetwopapersarereprintedtogetherin
Joos(1957),butJoos’sowneditorialcomments(1957:216)makeclearthathe
regardsthejuncturalanalysisascorrect.
Analysesofthissort–andwiththem,thewholedoctrineofthe
separationoflevels–werecalledintoquestionbyPikeintwoarticlesonwhathe
called‘grammaticalprerequisites’tophonologicalanalysis(Pike1947band
1952).Pike’sargumentwasthat‘whenphonologicalandgrammaticalfactsare
20
mutuallydependent,thetreatmentofphonologywithoutreferencetogrammar
isaconcealmentofpartofamostimportantsetofstructuralfactspertinentto
phonology’(1947b:§0).10Heemphasizedthatinpracticetheclassicalminimal-
pairtestpresupposedlexicalorgrammaticalunits;hepointedoutthatby
focusingonutterance-initialcontrasts,asfieldworkersoftendo,‘theanalystcan
becertainthatheisatthebeginningofaphoneme,ofasyllable,ofastressor
rhythmgroup,ofanintonationcontour,ofaphonologicalsequenceofsometype’
andthat‘soundsatthebeginningofutterances…aresimultaneouslyatthe
beginningofaword,andatthebeginningofaconstruction.’(§3)Specifically
withregardtojuncturephonemes,heposedseveralquestionsthatnever
receivedsatisfactoryanswers:‘Ifajunctureisaphoneme,canonedescribeits
variantformsorindicationsasallophones?Andhowwillonetreatallophonesof
ajuncturephonemeiftheyhavenothingphysicallyincommonwitheach
other…?’(§4)Hecitedanumberofpapersinwhichanalystspositedjuncture
phonemes,andhedrewattentiontocontradictionsbetweentheirsupposed
theoreticalbasisandthewaytheywereactuallydeployedintheanalysis.
ManyofPike’sspecificcriticismsofjuncturephonemeswereraisedby
others,andtherelianceonmeaningasa‘shortcut’todeterminephonemic
contrastwasgenerallyrecognizedassomethingofaproblem.Buthisexplicit
rejectionoftheseparationoflevelsmadehimahereticratherthanjustacritic.
Bloch,aseditorofLanguage,rejectedthefirstgrammaticalprerequisitespaper
(seeLadd2015:133).Pike’sworkwascompletelyexcludedfromtheselectionof
papersinReadingsinLinguistics(Joos1957)and,aswesawabove,Joos
lamentedthefactthatthenotionofgrammaticalprerequisitescontinuedto
‘plaguethecommunityoflinguists’11.Infact,whilePikecalledformoreexplicit
recognitionoftheroleofgrammaticalfactorsinphonemicanalysis,therewere
othervoicescallingforanevenmoreradicallimitationofphonemicanalysisto
whatwasconveyedbythesoundsalone.10EssentiallythisviewlayatthefoundationofChomsky,HalleandLukoff’spaper(1956)onthephonologyofEnglishstress.ForfullerdiscussionseeDresherandHall(thisvolume).11ItseemspossiblethatPikewasmarginalizedinpartbecauseofhiscommitmenttoChristianmissionarywork(Murray1994:174,189f),butcommentsIheardasastudentsuggestthathewasalsoregardedasratherunsophisticated;amongotherthings,hisprosewasinelegantandpronetooccasionalmalapropisms.
21
Specifically,somescholarsdrewadistinctionbetween‘wordphonemics’
and‘utterancephonemics’,implyingthatbyfocusingoncontrastsbetween
wordsphonologistswerenotfulfillingthepromiseofBloch’spostulates.There
wereneverclearlylaid-outcompetingtheoriesofwordphonemicsandutterance
phonemics,andpublishedreferencestothedistinctionarerare12:forthemost
part,everyonetheoreticallybelievedinutterancephonemicsbutinpractice
workedmostlywithcontrastsbetweenwords.Thistheoreticaldisconnectwas
whatledPiketoargueforacceptingthetheoreticallegitimacyofword-level
phonology,butitalsoencouragedafundamentalistcontraryviewthatsawword
phonemicsas,atbest,apreliminarysteponthewaytotrueunderstanding,and
atworstabetrayalofbasicprinciples.Trager,forexample,grumbledthat‘many
linguistshaveremainedcontenttodowordphonemics’(1962:13).
16.3.5DualityofPatterning
Somewherebehindthepost-Bloomfieldians’insistenceontheseparationof
levelslaythenotionofdualityofpatterning.Thiswasoneoftheseveral‘design
features’thatHockettlater(Hockett1958,ch.64;Hockett1960;Hockett&
Ascher1964)identifiedascharacteristicofhumanlanguage,andasaproperty
thatdistinguisheslanguagefrommostorallothercommunicationsystemsinthe
naturalworld.Thekeyideaisthatallutteranceshavebothagrammatical
structureandaphonologicalstructure,andthatthesestructuresareinherently
distinctandevenincommensurate.Inthisconception,anutterancecanbe
studiedeitherasastringofwords(ormorphemes,orgrammaticalformativesof
somesort)orasastringofphonemes.AlthoughHockett’sworkondesign
featuresappearedsometimeafterthedebatesovergrammaticalprerequisites
andtheseparationoflevels,thereisplentyofreasontobelievethathewas
thinkingaboutsuchissuesearlier(Radick2016),andplentyofreasontobelieve
thathehelpedleadthepost-Bloomfieldianstotheconvictionthatphonological
andgrammaticaldescriptionmustbeentirelyindependentofoneanother(cf.
12Probablythemostthoroughdiscussionofthisissueinprintappearedmanyyearslater(Hockett1987,esp.section5.3).
22
theclearearlystatementofthisprincipleinHockett1942,andHockett’sown
opinion,citedearlier,thathehadbroughtTragerroundtothispointofview).
Treatingtheseparationofgrammarandphonologyasafundamental
designfeatureortheoreticalpreceptwasalsounderdiscussioninEurope,inthe
formofHjelmslev’sdistinctionbetweenthe‘contentplane’andthe‘expression
plane’(Hjelmslev1935,1975)andMartinet’snotionof‘doublearticulation’
(Martinet1949,1960).AlthoughHjelmslev’sworkwasinfluentialprimarilyin
hisnativeDenmark,itdefinitelyinfluencedHockett’sthinkingaboutdualityof
patterning(Ladd2014:108f).Martinet’s‘doublearticulation’andHockett’s
‘dualityofpatterning’arebothnowwidelyunderstoodtorefertothesame
propertyoflanguage,andarestillwidelytakenforgranted,thoughseldom
consideredinanydepth.Itisobviousthatifdualityofpatterningisa
fundamentaldesignfeatureoflanguage–thatis,ifgrammaticalstructureand
phonologicalstructurearecompletelydistinctandincommensurateaspectsof
anutterance–thentherequirementofseparatinglevelsinalinguistic
descriptionfollowslogically.
However,closerconsiderationoftheconceptofdualityofpatterning
(Ladd2014,ch.5)suggeststhatthetwotypesofstructureareactuallynot
completelydistinctandincommensurate.Rather,Ihavearguedthatthereisan
intrinsicallyhierarchicalrelationbetweenthem,whichisimplicitinMartinet’s
term‘doublearticulation’butnotinHockett’sorHjelmslev’sversionsofthe
samegeneralidea:inthe‘primaryarticulation’,utterancescanbesegmented
into(orbuiltupfrom)words,andinthe‘secondaryarticulation’wordsarethen
segmentedinto(orbuiltupfrom)phonemes.Thatis,itmaybepossibleto
describewholeutterancesassequencesofphonemes,butourdescriptionis
simplerandmoreinsightfulifwetreatwordsassequencesofphonemesand
utterancesassequencesofwords.Ifwedothat,weimplicitlyacceptPike’s
argumentsfor‘grammaticalprerequisites’:forexample,wehavenodifficulty
talkingaboutword-initialorword-finalallophones;wehavenoneedfor
‘juncture’phonemes;andwehavenoproblemdescribinginternalsandhiand
externalsandhiasinherently‘ordered’.
23
Solongasdualityofpatterningisunderstoodasinvolvingcompletely
parallelandindependentstructures,thenTrager’slackofregardforthosewho
‘haveremainedcontenttodowordphonemics’isjustified;butifwe
acknowledgetheinherentlyhierarchicalrelationshipbetweengrammatical
structureandphonologicalstructure,then‘wordphonemics’maybenotonly
legitimatebutalsoappropriate.Attheveryleast,itseemsclearthattheissuesof
separationoflevels,grammaticalprerequisites,andwordphonemicsare
ultimatelyrelatedtothequestionofwhether‘phonology’isprimarilyabout
perceptiblydifferentsoundsinutterancesoraboutnetworksofcontrastand
generalizationsovertheformsofwords.Thepost-Bloomfieldianstookthefirst
view,andmostofthedetailsoftheirtheoryfollowedfromthat.
16.4Postscript
Ifpost-Bloomfieldianphonemicsemergedfairlyrapidlyaround1940,therewas
neverthelessnodecisivebreakwiththedevelopmentsthatprecededit.Wecould
citeTrager&Bloch1941orBloch1941asthestartingpoint,butthatwould
ignoretheextenttowhichlinguistsinthe1920sand1930s,onbothsidesofthe
Atlantic,wereallactivelyengagedincomingtogripswiththeconceptofthe
phoneme.Bloch(1948)listedanumberofworksthathis‘postulates’attempted
tosynthesize;theseincludednotonlyseveralfromwithinthepost-
Bloomfieldianfold,butalsoseveralnotablepredecessors(e.g.Sapir1933,Chao
1934,Swadesh1934,Twaddell1935)andtwoworksbyPike(1945anda
preliminaryversionof1947a).
Theendofthepost-Bloomfieldianperiodwasmoreabrupt.Blochdied(at
only58)in1965,andTragereffectivelywithdrewfromtheoreticallinguisticsin
themid-1960s(Hockett1993:786).AsforHockett,hispresidentialaddressto
theLinguisticSocietyofAmericainDecember1964(publishedas‘Soundchange’
inLanguagein1965)madeonelastattempttoovercomethebitternessofthe
yearsthatfollowedthepublicationofChomsky’sSyntacticStructures(1957)13.
13Inhispresidentialaddress,Hockettdrewacomparisonbetweentheatmosphereoftheearly1960sandthatofthe1870s,whentheneogrammarianswereexpoundingtheirideas.Hemadeit
24
Heidentifiedfourkeybreakthroughsthatdefinemodernlinguistics,thethird
andfourthbeingroughlythephonemicprincipleandthemathematical
formalizationoflinguistictheory;thetalkseemstohavebeenintendedasan
olivebranchtothegenerativists,acknowledgingthattheirworkhadledto‘the
firstnontrivialmathematizingoflinguistics’,butalsopleadingforrecognitionof
theimportanceofwhathadbeenachievedinphonologyinthe1940sand1950s.
Ifthiswasindeedtheintention,itfailed,andonlyafewyearslaterHockett
(1968:3)burnedhisbridges,dismissing‘Chomskyan-Hallean“phonology”’
[scarequotesintheoriginal]as‘completelybankrupt’.Afewyearslaterhe
movedhisCornellofficefromthelinguisticsdepartmenttotheanthropology
department(hehadalwayshadajointappointment),andturnedhisattentionto
otherprojectssuchasanintroductoryanthropologytext(1973)andothertopics
suchastheevolutionoflanguage(e.g.Hockett1978).
Nevertheless,hispresidentialaddresssuggeststhathewasreadyto
abandonsomeofthedogmasofthepost-Bloomfieldianyears.Heconcedesthat
weneedtworepresentations,onerespectinglexicalidentityandtheother
respectingperceptualidentity,andsaysthatitisofnoconsequencewhetherwe
callthemappingbetweenthetwo‘phonology’or‘morphophonemics’.Healso
suggeststhattherepresentationrespectingperceptualidentityneedstobe
expressedintermsoflocalfrequencymaximainamultidimensionalphonetic
space,implyingasignificantshiftawayfromthephone-basedstringsofthepost-
Bloomfieldianyearsandtowardarepresentationthatcanmorereadily
accommodatetherelationshipbetweenthesurfacecategoriesandthe
continuoussignal.Intheend,hecouldnotovercomethebitternessthatgrewout
ofthetransitiontothegenerativeascendancy,ortheindifferenceofgenerative
phonologiststothequestionofhowphonologicalcategoriesrelatetothephysics
ofspeech.Buthisfaiththatprogresswaspossiblemayyetbevindicated.
clearthatheobjectedtoboththetoneandthecontentofsomeoftheearlygenerativework,butatthesametimeheacknowledgedthat‘theattackonthenewdispensationfromsomequartersisviciousindeed’.Hedeclinedtorepeatthesubstanceoftheseattacksonthegroundsthatthey‘werelargelymadeininformalconversation’.Adoptingthesamepolicyhere,Inoteonlythatthetemperofthetimeswassufficientlypolarizedthateminentprofessorsfeltitappropriatetosharescurrilousanecdoteswithundergraduates.
25
References
Bloch,Bernard.1941.Phonemicoverlapping.AmericanSpeech16:278-284.
ReprintedinJoos1957,pp.93-96andinMakkai1972,pp.66-70.
Bloch,Bernard.1948.Asetofpostulatesforphonemicanalysis.Language24:3-
46.ReprintedinMakkai1972,pp.167-199.
Bloomfield,Leonard.1933.Language.NewYork:Holt,RinehartandWinston.
Bloomfield,Leonard.1939.Menominimorphophonemics.Travauxducercle
linguistiquedePrague8:105-115.ReprintedinMakkai1972,pp.58-64.
Browman,CatherineP.&LouisM.Goldstein(1986).Towardsanarticulatory
phonology.PhonologyYearbook3:219-252.
Browman,CatherineP.andLouisM.Goldstein(1989).Articulatorygesturesas
phonologicalunits.Phonology6:201-251.
Chao,Yuen-ren.1934.Thenon-uniquenessofphonemicsolutionsofphonetic
systems.BulletinoftheInstituteofHistoryandPhilology,AcademiaSinica4:363-
397.ReprintedinJoos1957,pp.38-54.
Chomsky,Noam.1957.Syntacticstructures.TheHague:Mouton.
Chomsky,Noam.1964.Thenatureofstructuraldescriptions.CurrentIssuesin
LinguisticTheory,Chapter4.TheHague:Mouton.ReprintedinMakkai1972,pp.
401-423.
Chomsky,Noam,MorrisHalle,andFredLukoff.1956.Onaccentandjuncturein
English.InMorrisHalle,HoraceG.Lunt,HughMcLean,andCornelisH.van
Schooneveld(eds.),ForRomanJakobson:Essaysontheoccasionofhissixtieth
birthday,11October1956,65-80.TheHague:Mouton.
Cutler,Anne.2012.Nativelistening:languageexperienceandtherecognitionof
spokenwords.CambridgeMA:MITPress.
26
Gleason,HenryAllan,Jr.1955a.(Revisededition1961).Anintroductionto
descriptivelinguistics.NewYork:Holt,Rinehart,Winston.
Gleason,HenryAllan,Jr.1955b.Workbookindescriptivelinguistics.NewYork:
Holt,Rinehart,Winston.
Hall,RobertA.,Jr.1991.156Broadway:AcrucialnodeinAmericanstructural
linguistics.Historiographialinguistica18:153-166.
Halle,Morris.1959.ThesoundpatternofRussian.TheHague:Mouton.
Harris,ZelligS.1942a.Morphemealternantsinlinguisticanalysis.Language18:
169-180.ReprintedinJoos1957,pp109-115.
Harris,ZelligS.1942b.ReviewofL.Spieretal.(eds.),Language,culture,and
personality:EssaysinmemoryofEdwardSapir.Language18:238-245.
Harris,ZelligS.1944.Simultaneouscomponentsinphonology.Language20:
181-205.ReprintedinJoos1966,pp.124-138andinMakkai1972,pp.115-133.
Haugen,Einar.1938.Noteson‘voicedT’inAmericanEnglish.DialectNotes6:
630-631.
Hjemlslev,Louis.1935.Ontheprinciplesofphonematics.InProceedingsofthe
SecondInternationalCongressofPhoneticSciences,London,pp.49-54.
Hjemlslev,Louis.1975.Resuméofatheoryoflanguage(translatedandeditedby
FrancisJ.Whitfield).Madison:UniversityofWisconsinPress.
Hockett,CharlesF.1942.Asystemofdescriptivephonology.Language18:3-21.
ReprintedinJoos1957,pp.97-108andinMakkai1972,pp.99-112.
Hockett,CharlesF.1949.Twofundamentalproblemsinphonemics.Studiesin
Linguistics7:29-51.ReprintedinMakkai1972,pp.200-210.
Hockett,CharlesF.1955.Amanualofphonology.Baltimore:WaverleyPress.
27
Hockett,CharlesF.1958.Acourseinmodernlinguistics.NewYork:Macmillan.
Hockett,CharlesF.1960.Theoriginofspeech.ScientificAmerican203:88-111.
Hockett,CharlesF.1961a.Grammarforthehearer.InR.Jakobson(ed.)Structure
oflanguageanditsmathematicalaspects(Proceedingsofsymposiainapplied
mathematics,vol.XII).Providence,R.I.:AmericanMathematicalSociety.
Hockett,CharlesF.1961b.Linguisticelementsandtheirrelations.Language37:
29-53.
Hockett,CharlesF.1965.Soundchange.Language41:185-204.
Hockett,CharlesF.1968.Thestateoftheart.TheHague:Mouton.
Hockett,CharlesF.1973.Man’splaceinnature.NewYork:McGraw-Hill.
Hockett,CharlesF.1978.InsearchofJove’sbrow.AmericanSpeech53:243-313.
Hockett,CharlesF.1987.Refurbishingourfoundations:Elementarylinguistics
fromanadvancedpointofview.Amsterdam:JohnsBenjamins.
Hockett,CharlesF.1993.GeorgeLeonardTrager(obituary).Language69:778-
788.
Hockett,CharlesF.&RobertAscher.1964.Thehumanrevolution.Current
Anthropology5:135-168.
Joos,Martin.1948.Acousticphonetics(LanguageMonograph23).Baltimore:
WaverleyPress.
Joos,Martin.1950.Descriptionoflanguagedesign.JournaloftheAcoustical
SocietyofAmerica22:701-708.
Joos,Martin(ed.).1957.ReadingsinLinguisticsI.Chicago:UniversityofChicago
Press.Republishedwithcorrections,1966.
Joos,Martin.1967.BernardBloch(obituary).Language43:3-19.
28
Kaisse,Ellen.1985.ConnectedSpeech:Theinteractionofsyntaxandphonology.
NewYork:AcademicPress.
Kilbury,James.1976.Thedevelopmentofmorphophonemictheory.Amsterdam:
JohnBenjamins.
Ladd,D.Robert.2011.Phoneticsinphonology.InJ.Goldsmith,J.Riggle,A.Yu
(eds.)Thehandbookofphonologicaltheory,secondedition.Wiley-Blackwell.
Reprintedaschapter2ofLadd2014.
Ladd,D.Robert.2014.Simultaneousstructureinphonology.Oxford:Oxford
UniversityPress.
Ladd,D.Robert.2015.TheAmericanfour-levelanalysisofintonationcontours:
Historicalpostscript.Historiographialinguistica42:119-137.
Makkai,ValerieBecker(ed.).1972.Phonologicaltheory:Evolutionandcurrent
practice.NewYork:Holt,Rinehart&Winston.
Martinet,André.1949.Ladoublearticulationlinguistique.TravauxduCercle
LinguistiquedeCopenhague5:30-37.
Martinet,André.1960.Élémentsdelinguistiquegénérale.Paris:ArmandColin
Merriam-Webster.1961.Webster’sThirdNewInternationalDictionary.
SpringfieldMA:G.&C.Merriam.
Moulton,WilliamG.1947.JunctureinmodernstandardGerman.Language23:
212-226.ReprintedinJoos1957,pp.208-215.
Murray,StephenO.1994.TheorygroupsandthestudyoflanguageinNorth
America.Amsterdam:JohnBenjamins.
Nespor,Marina&IreneVogel.1986.Prosodicphonology.Dordrecht:Foris
Publications.
29
Nida,Eugene.1948.Theidentificationofmorphemes.Language24:414-441.
ReprintedinJoos1957,pp.255-271.
Oswald,VictorA.,Jr.1943.‘VoicedT’:Amisnomer.AmericanSpeech18:18-25.
Pierrehumbert,Janet&MaryE.Beckman.1988.Japanesetonestructure.
Cambridge,MA:MITPress.
Pike,KennethL.1943.Phonetics:Acriticalanalysisofphonetictheoryanda
technicforthepracticaldescriptionofsounds.AnnArbor:UniversityofMichigan
Press.
Pike,KennethL.1945.TheintonationofAmericanEnglish.AnnArbor:University
ofMichiganPress.
Pike,KennethL.1947a.Phonemics:Atechniqueforreducinglanguagestowriting.
AnnArbor:UniversityofMichiganPress.
Pike,KennethL.1947b.Grammaticalprerequisitestophonemicanalysis.Word
3:155-172.ReprintedinMakkai1972,pp.153-165.
Pike,KennethL.1952.Moreongrammaticalprerequisites.Word8:106-121.
ReprintedinMakkai1972,pp.211-223.
Pike,KennethL.1958.Ontagmemes,néegramemes.InternationalJournalof
AmericanLinguistics24:273-278.
Radick,Gregory.2016.Theunmakingofamodernsynthesis:NoamChomsky,
CharlesHockett,andthepoliticsofbehaviorism,1955–1965.Isis107:49-73.
Sapir,Edward.1933.Laréalitépsychologiquedesphonèmes.Journalde
psychologienormaleetpathologique30:247-265.Reprintedintranslationas
‘Thepsychologicalrealityofphonemes’inD.G.Mandelbaum(ed.),EdwardSapir:
Selectedwritingsinlanguage,culture,andpersonality,Berkeley:Universityof
CaliforniaPress,1949,pp.46-60.TranslationreprintedinMakkai1972,pp.22-
31.
30
Schane,SanfordA.1971.Thephonemerevisited.Language47:503-521.
Selkirk,ElisabethO.1984.Phonologyandsyntax:Therelationbetweensoundand
structure.CambridgeMA:MITPress.
Swadesh,Morris.1934.Thephonemicprinciple.Language10:117-129.
ReprintedinJoos1957,pp.32-37andinMakkai1972,pp.32-39.
Trager,GeorgeL.1950.ReviewofPike1947a.Language26:152-158.
Trager,GeorgeL.1962.Somethoughtson‘juncture’.StudiesinLinguistics16:11-
22.
Trager,GeorgeL.&BernardBloch.1941.ThesyllabicphonemesofEnglish.
Language17:223-246.ReprintedinMakkai1972,pp.72-89.
Trager,GeorgeL.&HenryLeeSmith,Jr.1951.AnoutlineofEnglishstructure.
Republishedwithcorrections1957.Washington:AmericanCouncilofLearned
Societies.
Twaddell,W.Freeman.1935.Ondefiningthephoneme.Languagemonographno.
16.ReprintedinJoos1957,pp.55-79.
Wells,RulonS.1947a.Immediateconstituents.Language23:81-117.Reprinted
inJoos1957,pp.186-207.
Wells,RulonS.1947b.ReviewofPike1945.Language23:255-273.