Postwww.gegeshe.com

UniversityofPennsylvaniaScholarlyCommonsPubliclyAccessiblePennDissertationsFall12-22-2010OpacityandTransparencyinPhonologicalChangeJonathanGress-WrightUniversityofPennsylvania,gressmeister@gmail.
comFollowthisandadditionalworksat:http://repository.
upenn.
edu/edissertationsPartoftheComparativeandHistoricalLinguisticsCommons,andthePhoneticsandPhonologyCommonsOpenAccessabstractdiffersslightlyfromtextindissertation.
ThispaperispostedatScholarlyCommons.
http://repository.
upenn.
edu/edissertations/266Formoreinformation,pleasecontactlibraryrepository@pobox.
upenn.
edu.
RecommendedCitationGress-Wright,Jonathan,"OpacityandTransparencyinPhonologicalChange"(2010).
PubliclyAccessiblePennDissertations.
266.
http://repository.
upenn.
edu/edissertations/266OpacityandTransparencyinPhonologicalChangeAbstractModernHighGermanfinalobstruentdevoicingisusuallythoughttodescendfromMiddleGermandevoicingwithoutanychronologicalbreak,despitethefactthatthegraphicrepresentationoffinaldevoicingceasedintheEarlyModernperiod.
However,analternativeaccountholdsthatthespellingchangereflectstheactuallossofthedevoicingrule,andthatthereforethemodernrulehasanindependentorigin.
Inparticular,apocopeoffinalschwahasbeensuggestedasthecauseofthelossofdevoicinginEarlyModernGerman.
Accordingtothistheory,lossofdevoicingoccurredbecauseschwaapocoperenderedthedevoicingruleopaque,andhencehardtolearn.
Iftrue,weexpecttoseesomeevidenceforopaquedevoicingduringtheperiodthatapocopewasinprogress.
Inaccordancewiththisprediction,wefoundastatisticallysignificantcorrelationbetweenapocopeandabsenceoffinaldevoicinginanumberofGermantextsofthe14thand15thcenturies.
Afterthe15thcentury,devoicingislostacrosstheboard,whichcorrelateswiththecompletionofschwaapocopeandthelossoftheopaquedevoicingrule.
Thisconfirmsourtheoreticalpredictions.
Ifapocopehadnotrendereddevoicingopaque,wewouldhavetoconcludethatEarlyModernGermanschwaapocopewasaninstanceofruleinsertion.
However,thestructuraldescriptionofneitherapocopenordevoicingleadsustoexpectinsertion.
Instead,ModernGermanfinaldevoicingappearstobeaninstanceofrulere-affirmation,whichentailsthatthedevoicingrule,thoughopaque,remainedproductiveinsomedialects.
DegreeTypeDissertationDegreeNameDoctorofPhilosophy(PhD)GraduateGroupLinguisticsFirstAdvisorDonaldRingeKeywordsphonology,opacity,languagechange,ruleinsertion,German,finaldevoicingSubjectCategoriesComparativeandHistoricalLinguistics|PhoneticsandPhonologyCommentsOpenAccessabstractdiffersslightlyfromtextindissertation.
ThisdissertationisavailableatScholarlyCommons:http://repository.
upenn.
edu/edissertations/266OPACITYANDTRANSPARENCYINPHONOLOGICALCHANGEJonathanGress-WrightADISSERTATIONinLinguisticsPresentedtotheFacultiesoftheUniversityofPennsylvaniainPartialFulfillmentoftheRequirementsfortheDegreeofDoctorofPhilosophy2010SupervisorofDissertationDonaldRinge,Professor,LinguisticsGraduateGroupChairEugeneBuckley,AssociateProfessor,LinguisticsDissertationCommitteeDonaldRinge,Professor,LinguisticsEugeneBuckley,AssociateProfessor,LinguisticsRolfNoyer,AssociateProfessor,LinguisticsiiABSTRACTOPACITYANDTRANSPARENCYINPHONOLOGICALCHANGEJonathanGress-WrightDonaldRingeFinalobstruentdevoicingisattestedinbothMiddleandModernHighGerman,andthemodernruleisusuallyassumedtohavebeendirectlyinheritedfromthemedievalrulewithoutanychronologicalbreak(Reichmann&Wegera1993),despitethefactthatthegraphicrepresentationoffinaldevoicingceasedintheEarlyModernperiod.
However,analternativeaccountholdsthatthespellingchangereflectstheactuallossofthedevoicingrule,andthatthereforethemodernrulehasanindependentorigin(Mihm2004).
Inparticular,apocopeoffinalschwahasbeensuggestedasthecauseofthelossofdevoicinginEarlyModernGerman(Kranzmayer1956),which,iftrue,alignsdevelopmentsinGermanwithcontemporarydevelopmentsinYiddish(King1980).
LossofdevoicinginYiddishsupposedlyoccurredbecauseschwaapocoperenderedthedevoicingruleopaque,andhencehardtolearn(Kiparsky1972).
IfschwaapocopeisthecauseofthelossoffinaldevoicinginEarlyModernGermanaswell,thenweexpecttoseesomeevidenceforopaquedevoicingduringtheperiodthatapocopewasinprogress,whichispreciselywhatwefind.
AstatisticallysignificantcorrelationbetweenapocopeandabsenceoffinaldevoicingcanbeshownforanumberofGermantextsofthe14thand15thcenturies,i.
e.
wordsthatneverhadfinalschwastilltendtoshowdevoicing,whilewordsthatformerlyhadafinalvoweltendnottoshowdevoicing.
Afterthe15thiiicentury,devoicingislostacrosstheboard,whichcorrelateswiththecompletionofschwaapocopeandthelossoftheopaquedevoicingrule.
Thisconfirmsourtheoreticalpredictions.
Ifapocopehadnotrendereddevoicingopaque,wewouldhavetoconcludethatEarlyModernGermanschwaapocopewasaninstanceofruleinsertion(King1969).
However,thestructuraldescriptionofneitherapocopenordevoicingleadsustoexpectinsertion(King1973).
Instead,ModernGermanfinaldevoicingappearstobeaninstanceofrulere-affirmation(Hock1991),whichentailsthatthedevoicingrule,thoughopaque,remainedproductiveinsomedialects.
ivTableofcontents1LossofdevoicinginEarlyModernGerman.
11.
1DescriptionandoriginofGermanfinaldevoicing.
11.
1.
1SynchronicdescriptionofGermanfinaldevoicing.
11.
1.
2Contemporaryregionalvariationinfinaldevoicing41.
1.
3OriginsoffinaldevoicinginMiddleGerman.
51.
2LossofdevoicinginEarlyModernGerman71.
2.
1Lossofdevoicingduetomorphemicspelling.
71.
2.
2ProblemswithPaul'stheory.
101.
2.
3Contemporaryattestationsoflossofdevoicing111.
3Lossofdevoicingthroughlenition.
131.
4Lossofdevoicingthroughapocope.
182LossofdevoicinginYiddish202.
1YiddishandGerman.
202.
2ApocopeandlossofdevoicinginYiddish.
222.
3Lossofdevoicingandparadigmleveling242.
4Opacityandlossofdevoicing.
272.
4.
1Opacityandmarkedness.
292.
4.
2Lossofdevoicingandparadigmuniformity312.
4.
3Opacityandlossofruleproductivity412.
5Productivityofdevoicingandevidenceforsynchronicapocope.
432.
5.
1Apocopeandrestructuringofunderlyingrepresentations432.
5.
2Apocopeasavariablerule473ApocopeandlossofdevoicinginEarlyModernGerman493.
1MiddleGermanapocope503.
2UpperandWestCentralGermanapocope553.
3InteractionoffinaldevoicingandschwaapocopeinEarlyModernGerman583.
3.
1TheBonncorpus583.
3.
2TheAugsburgerStadtbuch.
913.
4ApocopeandlossofdevoicinginGermandialects963.
5ApocopeandlossofdevoicinginEarlyModernDutch1003.
6EastCentralGerman.
1023.
6.
1EastCentralGermanapocope.
1043.
6.
2MorphologicalfactorsinEastCentralGermanapocope1133.
6.
3FinaldevoicinginEarlyModernEastCentralGerman.
1143.
6.
4OpacityandlossofdevoicinginEastCentralGerman.
1164RetentionoftransparentdevoicinginEastCentralGerman.
1184.
1Ruleaddition,ruleinsertionandrelativechronology.
1184.
1.
1Ruleaddition1224.
1.
2Ruleinsertion1244.
2Apocopeasamorphophonologicalrule.
1274.
3Apocopeasalexicalphonologicalrule.
1304.
4Finaldevoicingasaphoneticrule134v4.
4.
1Finaldevoicingasincompleteneutralization.
1374.
5Finaldevoicingasapostlexicalrule1384.
6Intrinsicruleordering1394.
7Phonologizationandopacity1414.
8Conclusions.
1445OpacityandrestorationoftransparentdevoicinginEastCentralGerman.
.
.
.
.
1455.
1Analogicalrestorationoffinaldevoicing1465.
2Rule-basedrestorationoffinaldevoicing1475.
2.
1Rulere-ordering1485.
2.
2Rulere-affirmation.
1515.
2.
3EastCentralGermanapocopeandre-affirmationofdevoicing.
1585.
3Conclusions.
1626SecondfinaldevoicingasanindependentchangeinModernGerman.
1626.
1Secondfinaldevoicingassoundchange.
1636.
2LowGermanoriginsofModernGermandevoicing.
1646.
3Generalconclusions16611LossofdevoicinginEarlyModernGermanFinaldevoicinghasbeenheavilystudiedfromasynchronicperspective,usingmanydifferenttheoreticalframeworks,bothinGermanitself(e.
g.
Wiese1996,Brockhaus1995,Jessen1997,Féry1999,Iverson&Salmons2007)andcross-linguistically(Mascaró&Wetzels2001).
Diachronically,thecross-linguisticoriginsoffinaldevoicinghavebeenextensivelystudiedinBlevins2004asprovidinggoodevidenceforan"evolutionary"approachtophonology.
However,lesstheoreticalworkhasbeendoneonthedevelopmentofthefinaldevoicingruleinEarlyModernandModernGerman,whenthelanguagewasundergoingsignificantchangesthatwouldpotentiallyhaveaffectedtheapplicationofdevoicing,inparticular,whateffecttheapocopeoffinalunstressedvowelsshouldhavehad.
Thepresentdiscussionaimstoelucidatetheproblem,providesometentativesolutionsandpointthewaytofurtherresearch.
1.
1DescriptionandoriginofGermanfinaldevoicing1.
1.
1SynchronicdescriptionofGermanfinaldevoicingSynchronically,Germanfinaldevoicingcanbecharacterizedasfollows(Wiese1996:201):(1)Finaldevoicing2[+obstruent]→[–voice]/––]σExamplesinclude(Wiese1996:200):(2)'praise'Lo[p]~Lo[b]es'wheel'Ra[t]~Ra[d]es'coffin'Sar[k]~Sr[g]e'active'akti[f]~akti[v]e'grass'Gra[s]~Gr[z]er'orange(adj)'oran[]~'orange(noun)'Oran[]eAsyoucansee,theruleappliesonlytoobstruents,i.
e.
stopsandfricatives(cf.
Icelandic,wheredevoicingalsoappliestosonorants).
Theenvironment,moreover,isthesyllableedge,ratherthanthewordedgeperse1(Vennemann1972);nevertheless,thetraditional,thoughambiguousterm"finaldevoicing"willberetainedforthesakeoffamiliarity.
SyllabificationappliescyclicallyinGermanmorphology,upuntiltheword-levelcycle,afterwhichthereisnore-syllabification(Rubach1990,Wiese1996:49-56).
Hence,word-finalobstruentsarenecessarilyalsosyllable-final,evenifthefollowingwordisvowel-initial.
Atthesametime,syllableboundarieswithinthewordtriggerdevoicing,e.
g.
A[p]teilung,withinunderlyingprefix/ab+/.
1Wiesegoesontorevisetheformulationoftheruleintermsofautosegmentalrepresentations,andthechoicebetweensyllablecodaandsyllableedgeasenvironment,butthisisnotrelevanttoourdiscussion.
3Butdevoicingcanalsooccurwithinwordsevenwheretheobstruentcouldbesyllabifiedasanonset,e.
g.
/ab+/inA[p]laut,'apophony',but[b]lau'blue'.
Insuchcases,weseethatresyllabificationdoesnotoccuracrosscertainmorphologicalboundarieswithintheword.
Fordetails,seeWiese1996,whoarguesthatsuchcompoundelementsandderivationalaffixesareinfactphonologicalwordsforthepurposesofsyllabification.
Bycontrast,therearecaseslikehab'ich'Ihave',wherethefinalunderlying//ofthefirstwordhasbeendeletedbyapostlexicalrule.
Thefinalunderlying/b/maysurfaceasvoiced[b]inthiscontext,althoughitvarieswith[p].
Thisdoesnotconstituteatrueexceptiontothedevoicingrule,however,sincethisonlyshowsthatthefollowingencliticpronounichmaybeoptionallyincludedinthesamephonologicalwordastheverbhab(e),whichisexpectedbehaviorforclitics:i.
e.
thesyllabificationis(ha)(b),fromunderlying/hab/+/x/.
Ifthetwomakeuponeprosodicword,thenweexpectnormalword-internalsyllabificationtoapply,i.
e.
the/b/willbetreatedasanonset,anddevoicingwillregularlyfailtoapply.
Wiese1996pointsoutthatsomeconsidertheGermanlaryngealoppositiontoinvolvenotvoicing,i.
e.
the[voice]feature,buttensing(Kloeke1982,andrecentlyIverson&Salmons2007).
Indeed,traditionalGermanlinguistshavelabeledthevoicingalternationasoneof'tense'versus'lax',or'fortis'versus'lenis'.
Thedebatebetweenthesetwoanalysesliesbeyondthescopeofthisdiscussion,however,becauseallwewillbeconcernedwithisthefactthatthereissomekindofrule-governedlaryngealfeaturealternationinGerman.
Whichfeatureweusetodescribethatoppositionisnotimportant4forourpurposes,sowewillcontinuetousethepopularlabelamongEnglish-languagephonologists,i.
e.
devoicing.
1.
1.
2ContemporaryregionalvariationinfinaldevoicingMihm2004notesthatfinaldevoicinginModernGermanisonlycharacteristicofnorthernpronunciation,whichhappenstobethebasisforthestandardpronunciation(Siebs1957).
Theoreticaldiscussionsaboutfinaldevoicingreflectthisnorthern-basedstandardpronunciation,ratherthanthecentralandsouthernvarieties,whichtypicallydonotshowdevoicinginfinalposition.
ThisisduetoasoundchangethattookplaceinmostoftheHighGermanareasintheEarlyModernperiod,knownas"innerHighGermanlenition",orbinnenhochdeutscheKonsonantenschwchung(cf.
Keller1978:407).
Inthelenitingregions,thereisnovoicingcontrastineithermedialorfinalposition,butbothserieshavebeenmergedinasingle,voicelesslaxseries:thus,inRhineFranconian,standardModernGermanlei[t]en'tolead'rhymeswithlei[d]en'tosuffer',witha[d]inbothcases(Keller1961,Mihm2004).
Acontrastisusuallyretainedininitialposition,butthisisnotrelevanttothediscussionbecauseinitialconsonantsneveralternatewithfinalconsonants.
AccordingtoMihm,theCentralGermanlenitingpronunciationwasthestandarduntilthe19thcentury,whenthespeechofthenortherncitiesofBerlin,HannoverandHamburg,whereboththevoicingcontrastandfinaldevoicinghadbeenpreserved,becamethebasisforthenewreceivedpronunciation.
ThisfactwillplayanimportantroleinthediscussionofthedevelopmentoffinaldevoicinginEarlyModernGerman.
51.
1.
3OriginsoffinaldevoicinginMiddleGermanThewrittenevidenceshowsthatthedevoicingrulebegantoaffectsomeobstruentsearlierthanothers(Paul2007:131-3).
Thechangebeginswithcoronalstops,withthechangemoreorlesscompletebythebeginningoftheMiddleGermanperiod(ca.
1100):OldGermankleid>kleit'cloth'.
Bythattimealsothevelarstopshavebecomedevoicedinfinalposition,atleastinBavariantexts,andlaterinotherregions:tag>tac/tach'day'2.
Thelabialstopswereapparentlythelasttoundergodevoicing:kalb>kalp'calf'.
However,althoughtheoriginofthedevoicingrulepossiblyliesinseveralsoundchangesoperatinginchronologicalseries,thefinalresultwasasingledevoicingruleaffectingallword-finalvoicedstops3.
Fricativescametoparticipateinthedevoicingrulebyadifferentroute.
InOldHighGerman,allfricativeswereapparentlyvoiceless,whetherinfinalornon-finalposition4.
IntheMiddleGermanperiod,however,word-internalfricatives,whenprecededandfollowedbya[–obstruent]segment(i.
e.
asonorantconsonant,aglideoravowel),becamevoiced:hofe>hove'court',hse>h[z]e'house'5,bothdativesingular.
AlthoughthechangethatgaverisetothefricativevoicingalternationinMiddleGerman2Thefinalrepresents[k],andthefinaltheaffricate[kx],sinceinsomesoutherndialects,especiallyMiddleBavarian,final[k]fromanysourceappearstohaveundergoneaffrication.
3Thiswouldbeacaseofrule"simplification"(King1969):severalphonologicalrulesarisingfromseveralindependentchangesmaybere-analyzedasasinglerule,ifsuchare-analysissimplifiesthegrammarwhileremainingdescriptivelyadequate.
4InsomeOldCentralFranconiantexts,wemayhaveevidenceforavoiced[](fromunderlying/θ/),writtenor,butbytheendoftheOldHighGermanperiodthisfricativehadmergedwiththestop[d],andsubsequentlyparticipatedinthedevoicingrulealongwiththeotherstops.
5Philologistsdebateoverwhether[z]wasorthographicallydistinguishedfrom[s]atallinMiddleGerman;manyhandbooksuseaspecialsymboltoindicatethevoicedsound,butithasdoubtfultextualsupport(seediscussioninPaul2007).
6wasproperlyspeakingavoicingrule,weassumethatthealternationbetweenmedialvoicedandfinalvoicelessfricativeswassubsumedunderthedevoicingrulealreadygoverningstopalternations,givingtheobstruentdevoicingruleasdescribedinstandardMiddleGermangrammars(Wright1907,Paul2007)6.
ThisMiddleGermandevoicingruleappearstohavehadthesamestructuraldescriptionastheModernGermanrule.
Thisincludestheconditioningbythesyllableboundary,asseenincompoundformslikejuncherre'gentleman'(ModernGermanJunker),from/jung/'young'+/herr/'lord',cf.
junc,butinflectedjunger(cf.
fn.
2).
AsinModernGerman,thedevoicingrulewasprobablytransparent7.
Whiletherehasalwaysbeensomevariabilityinthespelling,themedievalwritingsystemonthewholerepresentsdevoicingconsistentlyenoughthathistoricallinguistshavefeltjustifiedinassuminganexceptionlessdevoicingruleforMiddleGerman8.
UnlikeModernGerman,MiddleGermanclearlydistinguishedbetweenalternatingfinalvoicelessandmedialvoicedstops:e.
g.
nominativesingulartacagainstnominativepluraltag+e.
Thisisdespitethefactthatboththefinal[k]ofthesingularandthestem-final[g]ofthepluralcorrespondtoanunderlying/g/.
ThisindicatesthatMiddleGermanspellingusedaclassicalphonemicsystem,i.
e.
mostorallsurfacecontrastswereorthographicallyrepresented.
Bycontrast,ModernGermanspellingusesa6InEarlyModernGerman,all/v/revertedto/f/byalaterdevoicingrule,andsoonafter/w/became/v/.
Thisiswhytherearewordsusingtheoldspellingfor/f/,e.
g.
Volk,Vater.
Voiced/z/,asinhs/huz/'house',ModernGerman/haz/Haus,didnotrevertto/s/,however,butremaineddistinct.
7Paul2007notesthattheremaybeexceptionstofinaldevoicinginMiddleGermanwheretheword-finalunderlyingvoicedobstruentisfollowedbyavowel-initialclitic,e.
g.
neiger'heleaned'.
SeeaboveforasimilarcasefromModernGerman.
8Thespellingrepresentsthealternationsinstopsandthefricative/v/quiteconsistently.
Thefricative[z],ontheotherhand,asinhs~h[z]e,wasnotsoclearlydistinguishedfromthevoicelesscounterpart[s].
7"morphophonemic"or"morphemic"system,inwhichtheunderlying,notthesurfaceformisrepresented:thus,singularTag,pluralTage(seebelow).
1.
2LossofdevoicinginEarlyModernGerman1.
2.
1LossofdevoicingduetomorphemicspellingModernGerman,unlikeMiddleGerman,doesnotrepresentfinaldevoicinginthespelling:(3)MiddleGermanModernGermantac~tageTag~Tagewp~wberWeib~Weibereit~eitEid~Eide(withinnovative–eplural)ThestandardinterpretationofthelossoffinaldevoicinginEarlyModernGermancanbefoundintheclassicreferencegrammarslikePaul19899orReichmann&Wegera1993.
ThistheoryholdsthatdevoicingwasneveractuallylostinGerman,butremainedautomaticandtransparentcontinuouslyfromMiddleGermanthroughtoModernGerman.
ThisaccountcontinuestobeacceptedbymanyhistoricalphonologistsworkingonthedevelopmentoffinaldevoicinginGerman(e.
g.
Iverson&Salmons2007).
9ThemostrecenteditionofPaul'sgrammarofMiddleGerman(Paul2007)doesmentiontherecentdebateovertheexactsignificanceoftheEarlyModernGermanorthographicchanges,withoutcomingtoanydefiniteconclusions.
8ThechangesinspellingfromOldGermantoMiddleGerman(seediscussiononriseofMiddleGermandevoicing)revealthat,ratherthanmemorizethespellingsofeachwordbyrote,aswedomuchofthetimeinModernEnglishspelling,medievalGermanscribesappliedgeneralorthographicprinciplestotherepresentationoftheirphonologicalsystem.
Thus,whenthephonologicalsystemchangedsothat[tag]waspronounced[tak],thespellingchangedwithit:tag>tac.
Bythislogic,whentacbegantobewrittenTaginEarlyModernGerman,itisreasonabletosupposethatthisreflectedachangeinpronunciationfrom[tak]backto[tag].
However,the"standard"theoryofthelossofEarlyModernGermandevoicing(epitomizedbytheorthographictheoriesofHermannPaul;seeMihm2004:134-140)holdsthatthisparticularorthographicchangehadnothingwhatsoevertodowiththelossorretentionofthedevoicingrule.
Rather,itrepresentedaneworthographicprinciplethatgainedgroundintheEarlyModernperiod,whichwecallthe"morphemic"principle(dasmorphematischeormorphologischeSchreibprinzip).
Accordingtothestandardtheory,MiddleGermanspellingwasgovernedbytheso-called"phonemic"principle,whichwastheprinciplethatthespellingshoulddistinguishbetweensurfacecontrastiveelements,i.
e.
phonemes,butshouldnotrepresentmoreabstractdistinctions,i.
e.
"morphophonemes",or,ingenerativeterms,underlyingrepresentations.
Thus,since/k/and/g/contrastedinMiddleGerman(e.
g.
eithersoundcanoccurbetweenvowels),thespellingshouldhavedistinguishedbetweenthetwosounds,wherevertheyoccurredonthesurface.
Inthealternationbetweensingular[tak]andplural[tag],thestem-finalobstruent,whichis/g/inmedialposition,becomes/k/infinalposition.
Althoughinfinalpositionthecontrastisalwaysneutralized,nevertheless9the/k/phonememustbewritteninthispositionsincethecontrastobtainselsewhereinthelexicon10.
Thestandardtheoryconsidersthelossofwrittendevoicingtobeduetotheincreasinginfluenceofthe"morphemic"principle.
Instructuralistterms,whereasthephonemicstructureofthesingularof'day'was/tak/,themorphophonemicstructurewas|tag|,because/tak/and/tag/werealternatingmembersofasingleparadigm,andtherewasthereforeamorphophonemicrelationshipbetweenthephonemes/k/and/g/.
Morphemicwritingrepresentsthismorphophonemiclevelofstructure,ratherthanthephonemicleveloftheearlierstageofwriting.
Thereforethespellingrepresentsthesinglemorphophoneme|g|inbothsingularandplural,despitethephonemicneutralizationinthesingular.
Translatedintogenerativeterminology,wewouldsaythattheoriginalsystemrepresentedthosesoundsthataredistinctiveonthesurface,i.
e.
thatparticipateinbinaryoppositions,whichareinturndeterminedbydistinctive,binaryfeatures.
Thus,[+voice][g]iswritteng,but[-voice][k]iswrittenk(orc)11.
Thelater,morphemicsystemrepresentsonlysoundsthataredistinctiveintheunderlyingrepresentation.
Wepredict,inthiscase,thatallautomaticphonologicalrulesthatrewritebinaryfeaturespecifications10Paul'sorthographictheorieswerebasedinturnonconceptsfrompre-structuralisthistoricallinguistics,inwhichonerecognizedthatalphabeticsystemsareoriginallydesignedtorepresentactualpronunciation,butthattheregularcorrespondencebetweenthealphabetandthephonologybreaksdownwiththeactionofsoundchangeontheonehand,andtheinherentconservatismofwritingontheother.
11Forthepurposesoffinaldevoicing,oneexpectssurfacecontrastanddistinctiveoppositionstoberepresentedinthesameway,sincefinaldevoicingcanbetreatedeitherasphonemicneutralizationorneutralizationofafeatureopposition.
Therearesomecaseswherethetwotheoriesmakedifferentpredictions,however,e.
g.
wheretwoallophones,i.
e.
soundsthatareincomplementarydistributionandthereforenotcontrastiveinstructuralistterms,mayneverthelessbedistinguishedbyabinaryfeatureopposition,e.
g.
inOldEnglishmedial/h/andfinal/x/arefeaturallydistinctivebutincomplementarydistribution.
Phonemicwritingshouldnotbeabletorepresentsuchadistinction,butdistinctivefeature-basedwritingshould.
Thisauthorbelievesthattheevidenceinfactsupportstheideathatpre-modernalphabeticwritingisbasedonfeatureoppositions,whichisfortunatebecausetheauthoralsobelievesindistinctivefeaturetheory,notstructuralistphonemictheory.
Forsomecases,seeElmentaler2003,intheexcursusonallophonicwriting.
10willbe"undone"forthepurposesoforthographicrepresentation.
Hence,representationofsurface[tak]astacwillbereplacedbyrepresentationofunderlying/tag/asTag.
1.
2.
2ProblemswithPaul'stheoryWecancertainlyexplainhowsuchanorthographicshift,oncebegun,couldhaverapidlyspreadthroughouttheGerman-speakingregion.
AsGermangainedprestigeinthelateMiddleAges,thenotionof"correct"orthographyandastandardliterarydialectgraduallydeveloped.
Therefore,webegintosee,especiallyfromthemid-16thcenturyon(i.
e.
theReformationandCounter-reformation)arapidstandardizationofthewrittenlanguage(Keller1978,Reichmann&Wegera1993).
Thesedevelopmentspredictatendencytoincreasinguniformityofspellinginthefaceofcontinuingphonologicalvariationandchange.
OncethemorphemicprinciplebecomesestablishedintheregionwherethemostprestigiousGermanvarietyisused,itiseasytoexplainhowtherestoftheGerman-speakinglandswouldadoptthiswritingsystem,regardlessoftheirnativedialect.
ButwhatmotivatestheintroductionofthisneworthographicprincipleinthefirstplacePaul'stheoryofthedevelopmentofwritingwasbasedonquitetrueobservations,whichwerethatalphabeticsystemsdotendovertimetoshiftfromrepresentingsurfacerepresentationstorepresentingunderlyingrepresentations(andinturnfromunderlyingrepresentationstostylized,phonologicallyarbitraryrepresentationsofthewholewordormorpheme).
However,thequestionis:dothesesystemschangespontaneously,orarethereanyoutsideinfluencesthattriggersuchchanges11Wenotedabovethatwritingsystemsareinherentlyconservative.
Anuntriggeredchangefromaphonemictoamorphemicsystemwouldseemtoviolatethisprincipleoforthographicconservatism,andtheproponentsofthe"morphemicspelling"theoryneverexplainwhyitisthatthespellingsystemshouldhavespontaneouslyshiftedinthisway.
Therefore,weshouldlooktoseewhatotherchangesoccurredinthelanguagethatmighthavecausedtheshiftinspelling.
1.
2.
3ContemporaryattestationsoflossofdevoicingWealsohaveEarlyModernGermangrammarianswhocommentedonthephonologicalchangestakingplaceinthelanguageoftheirday.
OneparticularlystrikingexampleisanEastFranconian,ValentinIckelsamer(Mihm2004:180),whoinhis1534grammar(EineTeutscheGrammatica)condemnedthecontemporarytendencytopronouncefinal/d/andfinal/t/bothas,bywhichitappearshemeantavoicelesslenisstop,suchaswefindinmoderndialectsthathaveundergonetheinnerHighGermanlenition(seebelowontheAugsburgerStadtbuch).
AswewillseetheuseofthedigraphinlateEarlyModernGermantextsprobablydoesnotrepresentare-assertionoffinaldevoicing,sincetheoldergraphshouldhavesufficedtorepresentfinalfortis[t].
Rather,thedigraphrepresentsanewvoicelesslenissoundthatarosefromlenition,whichneutralizedthecontrastbetween/d/and/t/inallnon-initialpositions.
Butwhetherornotrepresentsaseconddevoicingorlenition,Ickelsamer'sprescriptivepronunciationindicatesthatthereceivedpronunciationofhisdaymaintained12avoicingcontrastinbothmedialandfinalposition.
WhileMihminterpretsthistomeanthatfinaldevoicinginMiddleGermanwasnotatrueneutralization,ourinterpretationisthat,afterapocopeoccurred,devoicingwasrenderedopaque,andsubsequentlywaslost,leadingtolevelingofthevoicedalternantandanalmostcompleterestorationofthevoicingcontrastasithadexistedinpre-devoicingOldGerman.
Aswesawintheprevioussection,thisprocesswasmainlycompletedbythe16thcentury,thetimeatwhichIckelsamerwrotehisgrammar.
Othersimilarevidencethatsupportstheexistenceofaprestige,non-neutralizing,non-lenitingdialectincludestheprescriptionsofJohannChristophGottsched(Mihm2004:177),asexpressedinhisAusführlicheRedekunst(1748).
Mihmconsidershistestimonytobeevidencethatthenormativedialect,theso-calledmeinischeAussprachenorm("Meissnischpronunciationnorm",aftertheUpperSaxoncapitalregionofMeissen),wasalenitingdialect.
Certainly,theothersourcesMihmadducessupportthis,suchastherhymingschemesofearlymodernpoets,whodidnotdistinguishbetween/d/and/t/,ortheprescriptionsofGoetheorJakobGrimm(Mihm2004:176-79).
Gottsched'spronunciationrulesstandout,however,becausehewascarefultodistinguishlenissoundslikethefinalginTagfromfortissoundslikechinTach,'roof'(inEastCentralGerman,bothwerepronouncedasfricatives).
Thus,thereappearstobesomeevidencethatnormativeHighGermanmaintainedafinalcontrastevenaslateasthe18thcentury12.
12InresponsetoboththetestimonyofIckelsamerandthetestimonyofGottsched,itmaybearguedthattheseprescriptivegrammariansweresimplyenjoiningaspellingpronunciation.
Whilethispossibilitycannotbelogicallyexcluded,itrequiresthatwereadsuchaninterpretationintotheircriticismsofcontemporarypronunciation.
IckelsamerandGottscheddidnotsayexplicitlythatafinalcontrasthadtobemaintainedinpronunciationbecauseitwasmaintainedinthespelling;rather,thefinalcontrasthadtobepronouncedbecausethatwassimplythecorrectpronunciationastheysawit.
ThetestimonyofIckelsamer131.
3LossofdevoicingthroughlenitionMihmalsodemonstratedthattheassumptionthatlossofdevoicinginEarlyModernGermanwasmerelyachangeinspellinghasbynomeansbeenuniversallyheldamongGermanphilologistsandhistoricallinguists.
Onthecontrary,heidentifiestwoschoolsofthought:a"northern"school,epitomizedbyHermannPaul,whichheldthatthelossofdevoicingwasmerelyorthographic,anda"southern"school,whichheldthatlossofdevoicingrepresentedaphonologicalchange.
SincePaulandthenorthernschoolendedupasthesourceforthestandardgrammarsandhandbooksofMiddleGermanandEarlyModernGermanthatmostnon-specialistsrefertowhendiscussingthehistoricalphonologyofGerman,thenorthernschoolwonoutbyahistoricalaccident.
Asithappened,thenorthernschoolmemberscamefrompartsofGermanywherethespokenHighGermandidinfacthaveafinaldevoicingrule,whichpronunciationalsobecamethebasisforthe"stagepronunciation"(Bühnenaussprache),i.
e.
receivedpronunciation,ofthemodern(19thcenturytopresent)standard,ascodifiedinSiebs1957.
ThesouthernschoolmemberscamefromregionsofGermanywherethespokenstandardHighGermandidnothaveadevoicingrule.
ThisnodoubtplayedasignificantroleinthedifferentinterpretationsoftheEarlyModernGermanchanges13.
Thesouthernspokenvariantstypicallyshowtheso-called"innerHighGermanlenition"mentionedearlier,i.
e.
theunconditionedneutralizationofvoicedandvoicelessobstruentsinbothmedialandfinalposition.
Sincespeakersfromthesouthwouldnotisprobablymorereliable,however,giventheearlierdate,andthefactthatthepronunciationofGottsched'sprestigiouscontemporary,Goethe,hadneitherfinaldevoicing,noranobstruentvoicingcontrastatall.
13Wearespeakinghereofregionalvariationsofthestandard(Umgangssprachen),nottruedialects.
14havefinaldevoicingforthisreason,itmakessensethatthesouthernschoolwouldbelessinclinedtobelieveintheunbrokencontinuityofdevoicingfromMiddleGermantoModernGerman.
MihminfactattributesthelossofdevoicinginEarlyModernGermantothislenition,whichthesouthernschooloriginallydidnotdo(Mihm2004:135-7).
Rather,manyattributedthelossofdevoicingtotwofactors:apocopeoffinalschwa,e.
g.
gabe>gab,andlevelingofalternations,e.
g.
tac>tagontheanalogyofpluraltage.
Thecausalconnectionbetweenthetwowasnotdiscussed,exceptinsomerarecasestobediscussedbelow.
Nevertheless,Mihmoffers,asanoriginalthesis,theideathatthelossofdevoicinginEarlyModernGermanwasduetolenition.
Hismainevidenceforthisisthedifferentratesatwhichdevoicingwaslost,accordingtotheplaceofarticulationofthefinalobstruents(Mihm2004:153-165).
ThetablebelowrepresentsthecompletestatisticsgatheredbyEwald.
ThesymbolsinthecurlybracketsrepresenttheProto-West-Germaniccorrespondence,andthesymbolsintheangularbracketsrepresenttheuseofvoicelessgraphs,i.
e.
devoicing.
Sometimesmorethanonevoicelessgraphcanbeusedtorepresentavoicelesssegment,e.
g.
final[k]canbewritten,,orevenor.
Ofcourse,theproportionofspokendevoicingmaybemuchhigherthanthatindicatedinthespelling.
(4)Proportionofdevoicingaround1500(Mihm2004&Ewald1997)1414{b}signifiesWestGermanic/b/,whichwas[b]mediallyand[p]finallyinmostHighGerman,butinCentralFranconianwas[v]mediallyand[f]finally.
{g}signifiesWestGermanic/g/,whichwas[g]mediallyand[k]finally,exceptinCFwhereitwas[]mediallyand[x]finally.
{,nd}signifiesWestGermanic/θ/and/d/afternasals,whichwas[d]mediallyand[t]finallyinalldialects,includingCF.
{d}signifiesWestGermanic/d/otherthanafternasals,whichwas[t]inallpositionsinmostHighGerman,exceptinCFwhereitwas[d]mediallyand[t]finally.
15{b}={g}={,nd}={d}=Bonn1547100%22%58%80%{b}={g}={,nd}=Leipzig14988%6%88%Bamberg15072%24%58%Straburg15074%2%9%Basel149733%7%25%Wessobrunn15050%2%42%(Bonn=CentralFranconian,Leipzig=EastCentralGerman,Bamberg=EastFranconian,Straburg=LowAlemannic,Basel=HighAlemannic,Wessobrunn=CentralBavarian)Thus,whenwelookattherelativenumbersofvoicelessgraphslikep,k,tagainstvoicedgraphslikeb,g,dintextsfromabout1500(Ewald1997),wefindthatthereisamuchhigherproportionofoverinallthetextsthanofover.
InthetextfromBasel,Switzerland,wehappentofindafairlylargeproportionofover,althoughelsewheretheproportionismuchsmaller,wherewetypicallyfindmorecasesofoverthanofover,e.
g.
Bamberg.
InthetextfromBonn(RipuarianorMiddleFranconian),wherefinal/b/wasphonologically[f],andwritten,wefindevenmorecasesofoverthanofover.
16Mihmconsidersthisplace-specificretractionofdevoicingtoindicatetheactionofsoundchange,i.
e.
innerHighGermanlenition.
ThissoundchangeaffectedmostofwhatiscalledUpperGerman(Alemannic,EastFranconian,Bavarian),withtheexceptionofHighAlemannic(Swiss)andSouthBavarian(southernAustria,Tirol).
Certainly,thecontext-specificnatureofthisretractionsitsoddlywiththeideathatitonlyreflectedanewmorphemicspellingprinciple.
Whywoulddevoicingbelostatdifferentratesaccordingtophonologicalcontext,ifthelosshadnothingtodowithphonologyHowever,Mihm'slenitiontheoryfallsintodifficulties.
Firstly,lenitiontargetedbothmedialandfinalvoicelessstopsandfricatives,buthisdiscussionfocusesonlyontheretractionofdevoicinginfinalstops.
ThereissomeevidenceoflenitioninmedialpositioninEarlyModernGermantexts,butoverallthecontrastwasmaintainedintheliterarylanguage.
Thelossofdevoicing,iftrulyphonological,doesnotrepresentunconditionedmergerofthevoicingcontrastinfavorofthevoicedseries,butratheranundoingofthecontextualmergerinfinalposition,i.
e.
thecontrastbetweenvoicedandvoicelessstopswasrestoredinfinalposition,whileinmedialpositionitwasneverlost15.
Forexample,MiddleGermanratcouldindicateeither/rat/'council',or/rad/'wheel',buttheirModernGermanreflexesarewrittendifferently,asRatandRad,respectively,althoughinthestandardnorthernpronunciationtheybothhavefinal[t],whileinthecentral,lenitingregionstheyhaveavoicelesslaxstopwhichcanberepresentedas[d].
YetMiddleGermandistinguishedthestopinleiten'tolead'fromthat15ThiswouldappeartocontradictGarde'sPrinciple(Labov1994),thatmergerscannotbeundonebylinguisticmeans.
However,laterwewillshowingreaterdetailhowcontextualmergerscanbeundonebylinguisticmeans.
17inlden/lidn/'tosuffer',andstandardModernGermancontinuestodoso:leitenversusleiden,butinthelenitingregionstheyarepronouncedthesame,i.
e.
voiceless[d].
Mihmacknowledgesthatthemodernlenitingvarietieshavemergedthecontrastinmedialposition,andaddressesthisprobleminacoupleoffootnotes(nn.
122,123)wherehearguesthatMiddleGermandevoicinginfactdidnotrepresentphonemicneutralization!
Instead,thewritingofe.
g.
/tag/astac,or/to:d/'death'astt,ModernGermanTod,onlyindicatedakindofphoneticapproximation.
Inotherwords,infinalpositionthesevoicedphonemesunderwentaphoneticdevoicingthatdidnotinvolveamergerwith"true"voicelessphonemes,e.
g.
/t/intt'dead',ModernGermantot.
ThisinterpretationoftheMiddleGermanspellingisoriginalbutimplausible.
Mihmisclaimingthatfinaldevoicingineffectdidnotneutralizeabinary[voice]distinction,andyetscribeswereconsciousenoughofthesub-featuralphoneticdistinctiontorepresentitgraphically.
Thisgoesagainsttheoverwhelmingevidenceofotherphonemicwritingsystems,inwhichonlydistinctivephonologicalinformationisencodedinthespelling(Rogers1995).
Mihm'sevidenceisbasedoncertaininterpretationsofsomespellingsintheBonnCorpustexts,#111and#133(seenextsection),whichareunfortunatelyinsomecaseswronganalyses16.
ThesecondproblemwithMihm'shypothesisisthatheignorestheotherpossibleexplanationforlossofdevoicing:apocopeoffinalschwaandlevelingofalternations.
Whilehementionsthisfactoroccasionally(e.
g.
p.
173-4),interestinglyhedoesnotseem16Thefaultyanalysesincludeanidentificationofsankchwiththestrongverbpreteritsank,wheninfactitisthenoun(vor)sang,i.
e.
weareseeingtheoutputofdevoicingandnotunderlyingfinal/k/.
Hedoesnotnoticethatanderwaidistheoutputofapocope(MiddleGermananderwaide),andheseemstothinkthatthenounglidcannothavebeenremodeledontheanalogyofarelatedform,eventhoughthepluralofglidisglider.
Theideaisthatifthefinalvoicedobstruentsoftheseformscouldnotbeaccountedforbylevelingorapocope,theonlyexplanationwouldbethattheetymologicalvoicinghadbeenpreservedthroughouttheperiodofMiddleGerman"devoicing"(cf.
fn.
60).
18tonoticetheexplanatorypowerofapocopewithregardtolossofphonologicaldevoicing17.
Theproblemremainsconcerningthedifferentratiosofvoicedandvoicelessgraphs,accordingtotheplaceofarticulation.
Since,aswehaveargued,MiddleGermandevoicingaffectedallobstruents,thestructuraldescriptionoftheruleshouldnothavereferredtoplaceofarticulation,i.
e.
theMiddleGermandevoicingruleisassumedtohavehadthesamedescriptionastheModernGermandevoicingrulegivenonpage2.
Therefore,whenthisrulewaslost,forwhateverreason,weexpecttheruletodisappearatthesamerateinallstops.
HowdoesthissquarewiththedatafromEwald1997Ewald'sdatacomesfromonlyonepointintime.
Ifwemeasurelossofdevoicingoveralongerperiod,however,wefindthatitwaslostatthesamerateinallcontexts(Fruehwald,Gress-Wright&Wallenbergforthcoming).
ThissuggeststhatthedifferentratiosfoundinEwaldarenotrelatedtothelossofdevoicingitself.
Whiletheymayreflectsomephoneticorphonologicalcontextualeffectsthatremaintobeworkedout,theydonotalterthefactthatfinaldevoicingasawholewaslostatthesamerateinallcontexts,showingthatasinglephonologicalrulewasbeinglost.
IfMihmisrightthatthesedatarepresentvariouslenitingsoundchanges,theyshouldhaveproceededatdifferentrates18.
1.
4Lossofdevoicingthroughapocope17Mihm'sinterpretationofMiddleGermandevoicinglookssimilartorecentattemptstodescribemodernGermandevoicingas"incompleteneutralization",whichwewillcometolateron.
18Thepointhereisnotthatplaceofarticulationdoesnothavesomeeffectonthefrequencyofapplicationoftheruleinquestion,butthatthedifferentfrequenciesofapplicationaccordingtoplaceofarticulationarenotevidencethattherewasnotasingledevoicingrulebeinglost.
19EberhardKranzmayerinhissurveyofSouthBavariandialects(Kranzmayer1956),whichhadnotundergonetheinnerHighGermanlenition,noticedthat,amongthosedialectsthathadretainedMiddleGermanfinalschwa,finaldevoicingwasalsoretained.
Thus,youwouldfindanalternationbetweennominativesingular[tkx]anddativesingular[tg].
Ontheotherhand,dialectsthathadlostfinalschwathroughapocope,sothatthedativesingularwasnow[tg],hadalsolostdevoicing(Kranzmayer1956:79),sothatthenominativesingularwaslikewise[tg].
Kranzmayer'sowntheory,whichcorrespondscloselytothegenerativeconceptsofopacityandruleloss,wasthatinpre-apocopeBavarian(whichincludesafewcontemporarydialectslikepartsofSouthTirolian)thefinalschwaintheparadigmaticalternationbetweene.
g.
nom/accsingular[tkx]19'day'anddativesingular[tg](cf.
MiddleGermantac~tage)insomewaymilitatedagainstanalogicalleveling(wewouldsaythatthealternationcouldbeeasilyaccountedforbyaphonologicalruleoffinaldevoicing).
Afterapocopeoccurred,withtheresultthatnom/acc[tkx]nowalternatedwithdativeandnom/acc/genpl[tg],analogicallevelingwasmorelikely,sincespeakerscouldnolongeraccountforthevoicingalternationbythepresenceoftheschwainthedative,i.
e.
thenom/accsingularformbecame[tg]ontheanalogyofthedative.
FromthisheinferredthatthelossoffinalledtothedisappearanceoffinaldevoicinginlatemedievalBavariantexts,e.
g.
Perg,'mountain'(cf.
ModernGermanBerg),from19InSouthBavarian,devoicedfinal/g/mergedwithformer/kk/togivetheaffricate[kx].
InHighAlemannic,ontheotherhand,devoiced/g/mergedwithformer/gg/togive[kk],whileformer/kk/became[kx]asinBavarian(Keller1961).
20olderperch.
UnlikePaulandthe"northern"school,Kranzmayersawthelossofwrittendevoicingasevidenceforlossofphonologicaldevoicing.
Yetaswesawearlierinthissection,historicallinguistsveryrarelyhitonapocopeasthecauseofthelossofdevoicinginlateMiddleGermanandEarlyModernGerman(totheextenttheyevenbelieveddevoicinghadbeenlost).
Themainreason,ofcourse,isthatthemodernstandardpronunciationstillhasfinaldevoicing,so,evenifweacceptthatdevoicingwaslost,westillhavetoaccountforhowdevoicing"re-applied"toallthenewinputscreatedbyschwaapocope.
Inotherwords,howdiddativesingulartage,whichnowhasnofinalvowelininformalspokenGerman,endupwithfinal[k],andnotfinal[g]AsMihmnoted,ofcourse,mostoftheHighGermanregionunderwentlenition,sowedonotknowforafactwhetherdevoicingactuallyre-appliedeverywhere.
Butwithrespectatleasttothestandardlanguage,whichmaintainsanunderlyingvoicingcontrast,weneedtoestablishtherelationshipbetweenapocopeandthis"seconddevoicing.
"2LossofdevoicinginYiddish2.
1YiddishandGermanKranzmayer'shypothesisaboutthelossofdevoicinginBavarian,i.
e.
thatitwascausedbyapocopeoffinalschwa,recallstheworkofPaulKiparskyandRobertKingontherolethatopaqueruleorderinghadintriggeringlossofdevoicinginYiddish(Kiparsky1968,1971,King1980).
However,whereasithasbeenuntilrecentlyuncontroversialthat21schwaapocopecausedthelossofdevoicinginYiddish,ithasyettobeestablishedforGerman.
BothYiddishandModernGermanaregenerallyagreedtodescendfromacommonancestor,MiddleGerman(Weinreich1980,Jacobs2005,butcf.
Wexler2002).
LikestandardModernGerman,Yiddishhasavoicingcontrastamongobstruents;unlikeGerman,Yiddishdoesnothavefinaldevoicing:(5)GermanTa[k],Yiddishto[g],'day'GermanLan[t],Yiddishlan[d],'country'GermanWei[p],Yiddishvay[b],'woman'GermanBrie[f],Yiddishbri[v],'letter'GermanHau[s],Yiddishhoy[z],'house'but,(6)GermanSa[k],Yiddishza[k],'bag'Germanmi[t],Yiddishmi[t],'with'GermanSie[p],Yiddishzi[p],'sieve'(butseebelow)Germantie[f],Yiddishti[f],'deep'Germangro[s],Yiddishgroy[s]'big'ThesecondsetofexamplesshowsthattheexistenceoffinalvoicedobstruentsinYiddishisnotduetosomefinalvoicingrule(whichwouldinanycasebeanunnaturalsound22change).
WealsoknowthatthefinalvoicedobstruentsinYiddisharenotduetothefactthatdevoicingsimplyneveroccurredinYiddish,becauseofexampleslikezip,whichwassibinOldGerman20,oravek'away',whichwasinweginOldGerman.
BothwordshadundergonedevoicinginMiddleGerman,i.
e.
beforeaseparateJudeo-Germandialectbegantocomeintoevidence(startinginthelate14thandearly15thcenturies).
KingalsoshowsthattheearliestJudeo-Germantextsshowdevoicing.
Forexample,a1382manuscriptwritesthefinalconsonantof'woman',/wi:b/,withtheHebrewcharacterfora/p/,nota/b/.
Therefore,thelossofdevoicingmustpost-datetheearliestattestationsofYiddish(King1980:374,401).
2.
2ApocopeandlossofdevoicinginYiddishStartinginthe15thcentury,weseeevidenceofapocopeoffinalvowelsinYiddish.
Apocopewasveryregular,affectingallfinalschwa,withtheapparentexceptionoftheinflectionalschwafoundinattributiveadjectives:gut'good',aguteman'agoodman'.
Leavingthatasideforthemoment,weseethatYiddishapocopewasaregularsoundchange.
Thisisimportantwhenwecometothediscussionofruleaddition.
Theresultofapocopewasthatallinflectionalcategoriesandwordclassesthatwerepartlyorwhollydesignatedbyfinalschwahadtobedrasticallyre-analyzed.
We20Final/p/isexceedinglyrareinMiddleGerman,occurringonlyinrecentborrowingsfromLatinorotherlanguageswithunshifted/p/.
Proto-Germanic/p/or/pp/wereshiftedtofricativesoraffricatesinallnon-initialpositionsinUpperHighGerman.
TherearemoreinstancesintheGermancomponentofYiddish,e.
g.
kop'head',butetymologicallythisfinal/p/isfromMiddleGermangeminate/pp/,whichbecame/pf/inUpperGerman,andfromtherepassedintoModernGerman,henceKopf.
Theretentionofunshifted/p/inYiddishisaninterestingexampleoftheCentralGermancomponentofYiddish,whichisotherwiselargelyUpperGermaninorigin(Jacobs2005).
ModernStandardGerman,bycontrast,hasinheritedmuchmorefromCentralGerman.
23canillustratethisusinghypotheticalpost-apocopeMiddleGermanparadigms,sinceMiddleGermanismoreorthelesstheimmediateancestorofYiddish,asitisofModernGerman.
Forexample,theMiddleGermanparadigmsof/tag/beforeandafterapocopewouldbeasfollows(borrowedfromKing1980):(7)Pre-apocopeSgPlNomtaktagAcctaktagGentagstagDattagtagnPost-apocopeNomtaktagAcctaktagGentagstagDattagtagnOrconsiderthepresentandpastindicativeofthestrongverbgeben'togive':24(8)Pre-apocopePresentPastSgPlSgPl1gebegebengapgaben2gipstgebetgbegabet3giptgebengapgabenPost-apocope1gebgebengapgaben2gipstgebetgbgabet3giptgebengapgabenAccordingtoKing,thissituationbroughtabouttheeventuallossofthedevoicingrule(owingtotheopacityofthedevoicingrule;cf.
below).
Formslikenominativesingular[tak],derivedfromunderlying/tag/,revertedto[tag].
Inotherwords,theunderlying[voice]specificationoftheobstruentwaspermittedtosurfaceinanenvironmentwherepreviouslyitcouldnot.
2.
3Lossofdevoicingandparadigmleveling25TherelativechronologyaccordingtoKiparsky,KingandKranzmayer,whichdatesparadigmlevelingafterapocope,overturnedanolderrelativechronology(Sapir1915,Sadock1973,andrecentlyagainAlbright2008),accordingtowhichlevelingprecededapocope:(9)Sapir/SadockchronologyMiddleGermantagetakLevelingtagetagApocopetagtagYiddishtegtog(10)Kiparsky/KingchronologyMiddleGermantagetakApocopetagtakLevelingtagtagYiddishtegtogNotonlythechronologicalorder,butalsothecausalorderwasreversed.
Intheolderview,thelevelingofthevoicedalternanttothesingularparadigmwasheldtobethecauseoflossofautomaticdevoicing,andhencethereasonwhythenewfinalobstruentsresultingfromapocopedidnotbecomedevoiced.
Preciselywhyparadigmlevelingwould26haveoccurredatall,despiteautomaticdevoicing,isunclearinSapir'sformulation,butneverthelesswehavetoconsiderlevelingasapossiblealternativeexplanationforthelossofdevoicing.
Onereasonforpreferringtotreatlossofdevoicingastrueruleloss,ratherthananalogicalleveling,isthefollowing:KingnotesthatYiddishfinalvoicedobstruentsonlyturnupinalternatingforms,e.
g.
MiddleGermanwc~wge>–>Yiddishveg~vegn.
Non-alternatingforms,likeavek'away',ultimatelyfromOldGermaninwegthroughMiddleGermanenwc,retaindevoicing.
Inotherwords,restorationoffinalvoicedobstruentsdoesnotoccureveryplacewheredevoicinghadappliedhistorically,butonlywherethefinalvoicelessobstruentwasstillderivedsynchronicallyfromanunderlyingvoicedobstruent.
I.
e.
enwcwouldnowhaveunderlying/k/intheUR,/nwk/,butwcwouldstillhave/g/,i.
e.
/wg/,becausethereisevidenceforunderlyingstem-final/g/inthepluralwg+e.
Ifthelossofdevoicingweretheproductofanalogicalleveling,wewouldnotexpectsuchastrongcorrelationbetweenlevelingoffinalvoicingandsynchronicderivationofdevoicing.
Forexample,sinceavekissimilarinformtoveg'way',thereisnoreasonforlexicalanalogy,whichoperatesonthesimilaritiesbetweensurfaceforms,totreatthetwodifferently.
Thepatternmakesperfectsense,however,oncewestartreferringtounderlyingrepresentations,butonlyatheoryofsynchronicderivationsallowsustodothat21.
21ThisargumentfollowsHock1991,whodemonstratesthatthiskindof"regular"analogyisimpossibletopredictunderpre-generativeassumptions,notwithstandingattemptstoformalizeprinciplesofanalogicalchangewithintheoldertradition(e.
g.
Kurylowicz1949).
27Theotherreasontoconsiderthisacaseofrulelossisthatwhilenon-alternatingformsresistlossofdevoicing,bycontrastalmostallalternatingformsfailtoresistit.
Thatis,almostallunderlyingvoicedobstruentsnowsurfacedasvoiced(withahandfulofexceptions,includingzipmentionedabove).
Iflossofdevoicingwereentirelytheproductoflexicalanalogy,wewouldexpectalternatingformsliketac~tag,wc~wegtobeaboutevenlysplitbetweenparadigmsthatlevelthevoicelessalternant,andthosethatlevelthevoicedalternant.
Itistruethat,asnotedabove,therearesomeparadigmswherethevoicelessalternantisunexpectedlyleveled:asidefromzip,thereise.
g.
hant~hent,formerlyhant~hende,cf.
land~lender,fromearlierlant~lender.
Kingnotes,however,thatsuchexceptionsarealmostentirelyconfinedtostemsendingin/nd/inMiddleGerman,andthat/nd/stemswheredevoicingislost,likeland,allhaveMiddleGermanpluralsin–er,not–e,i.
e.
theywereunaffectedbyapocope.
IfwerecallthatinOldGerman,manyoftheseformshad/t/,e.
g.
hant~henti,itisconceivablethatthevoicingof/t/afternasalsdidnotoccurinYiddishuntilafterapocope.
Inanycase,thisgroupofexceptionsisdefinedbyacommoncharacteristic,andotherwisethelossofdevoicingisoverwhelminglyregular,i.
e.
itdidnotfollowalexicallyarbitrarypath,aswewouldexpectfromsimpleanalogy.
2.
4OpacityandlossofdevoicingTheMiddleGermandevoicingrule,aswenoted,wastransparent.
Thatis,allmembersofaparadigmwhereanunderlyingvoicedobstruentsurfacedfinally,surfaceasvoiceless,28withtheresultthatdevoicingwascompletelypredictableaccordingtothephonologicalenvironment:ifanunderlyingvoicedobstruentsurfacessyllable-finally,itloses[voice].
Apocope,however,rendereddevoicingopaque,i.
e.
itwasnolongerpredictableaccordingtophonologicalenvironment.
Phonologicalopacityisdefinedasfollows(Kiparsky1971,cf.
Bakovic2007):-AruleA→B/C__Disopaquetotheextentthattherearesurfacerepresentationso(i)oftheformAinenvironmentC__D,oro(ii)oftheformBinenvironmentotherthanC__DAccordingtoKiparskyandKing,thenon-predictabilityofarule,i.
e.
itsopacity,appearstohavesomecausalrelationshipwiththelaterlossofthatrule.
Yetelucidatingthatrelationshipcanbedifficult.
Forinstance,wecannotsimplysaythatopacitypreventedthedevoicingrulefrombeingacquiredatall.
Althoughtheruleisnolongerpredictableonthesurfaceinphonologicalterms,itmightstillbepredictableinmorphologicalterms,especiallywhenweconsiderhowthedevoicingrulecontinuestoapplyproductivelyincertaininflectionalorderivationalcategories,orincertainmorphologicalclasses.
Thisiswhatwecallrulemorphologization.
Forexample,devoicingwouldstillapplyinthenom/accsingofcertainsnouns,theuninflectedpredicativeformsofadjectiveslikeblint,blinder'tired',andthe1sgand3sgpasttenseofstrongverbs:29(11)'day'SgPlNomtaktagAcctaktagGentagestagDattagtagen(12)'togive',presentandpastindicative1gebgebengapgaben2gipstgebetgbgabet3giptgebengapgabenInthetwoparadigmsabove,wecouldargueforacontinueddevoicingrule,butonethatappliesonlyinthenominativeandaccusativesingular,orinthefirstandthirdpersonpastindicative.
Becausewecandemonstratethatadevoicingruleofsomedescriptioncontinuedinthelanguageafterapocopehadoccurred,westillhavetospecifywhatthecausalconnectionbetweenapocopeandlossofdevoicingwas,i.
e.
whynominativesingular[tak]waseventuallyreplacedby[tag].
2.
4.
1OpacityandmarkednessOneanswertothisisthattheapocopeledtoamarkedgrammar.
Whenapocopeoccurred,itwasaddedtothegrammar,afterfinaldevoicinginthederivationalseries:30(13)UR/tag//tag/FD--takApocopetag--SR[tag][tak]Notethatinthisgrammar,finaldevoicingremainsphonologicallypredictable.
Theonlyreasonitisopaqueisbecauseanotherrulehasbeenorderedafteritinthederivation.
Assumingthatlearnerswouldhavenoproblemacquiringbothphonologicalrules,whatreasonisthereforthemtoalterthegrammarbyeliminatingoneoftherulesAccordingtoKiparsky1971,agrammarismarkediftherulesareincounter-feedingorbleedingorder.
Andtheorderingofapocopeafterdevoicingisindeedacounter-feedingordering:thatis,ifapocopewereinsteadorderedbeforedevoicing,asin(13),thenapocopewouldfeeddevoicing.
Kiparskylaterrevisedhisdefinitionofgrammaticalmarkednesstosaythatagrammarismarkedifitisopaque,i.
e.
iftherulesareorderedinsuchawaythatoneofthemisnolongerpredictableonthesurface,asindicatedinthedefinitionofopacitygivenabove.
Ineithercase,the(idealized)languagelearnerrepairsthismarkedsituationbyeliminatingtheopaquerule.
Ofcourse,rulelossisnottheonlywaytorepairopacity.
Anotherwayistore-ordertherulessothattheyareinatransparent,feedingorder:(14)UR/tag//tag/31Apocopetag--FDtaktakSR[tak][tak]Whyshouldrulelosshavebeenpreferredtorulere-orderinginthecaseofYiddishThereisnoprincipleofsynchronicgrammarthatwouldhelpushere.
Rulelossiscertainlyinevitablewhereverthesimplicitycriterionrequiresit.
Forexample,iftheoutputsgeneratedbyagrammaraftertheadditionofarulecanalsobegeneratedbyarestructuredgrammarwithoutthatrule,thenweexpectrestructuringandrulelosstooccur(Halle1962).
But,aswesawabove,apocopedidnotcompletelydestroytheevidencefordevoicing,butonlyrenderedtheevidencelesstransparent.
Itispossible,though,thatthechoicesofrulelossandrulere-orderingarenottheonlyonesavailabletous.
Oneoptionistosaythatopacityledtoirregularparadigms,whichinturntriggeredparadigmleveling,meaningthattherewasnoactualrulelosstakingplace,butratheraradicallydifferentprocess,whoseresultsjusthappentoappeartoidenticaltotheresultsofruleloss.
Anotheroptionistosaythatopacityledtoalossintheproductivityofthedevoicingrule,andthattherule'sproductivityisthefactorpredictingeitherlossorre-ordering.
Weshalldealwitheachinturn.
2.
4.
2LossofdevoicingandparadigmuniformityAlbright2008hasarguedthatthelevelingofvoicingalternationsinYiddishwasduetoparadigmpressure,whichinOTcanbehandledbyparadigmuniformityconstraints.
The32constraintinquestionisBase-Identity,whichisviolatedifanoutputdiffersinsomerespectfromthe"base"memberoftheparadigm,whichinthecaseofMiddleGermanandYiddish,wasthestem-formoftheplural/tag+e/.
ThechangeinvolvespromotionofBase-IdentityabovethemarkednessconstraintFinDevoi:(15)MiddleGermangrammar/tag/FinDevoiBase-Ident(/tag+PL/)Ident-IO[vce]tag*!
tak**Yiddishgrammar/tag/Base-Ident(/tag+PL/)Ident-IO[vce]FinDevoitag*tak*!
*Thepluralstem/tag+/=[tag],withthevoicedobstruent,wasselectedastheparadigmaticbasebythelearner,accordingtoanindependentlearningprinciple(expoundedinAlbright2002)wherebythemostphonologically"informative"memberof33theparadigmisalwaysselectedasthebase.
Themostphonologicallyinformativememberofaparadigmistheoneinwhichthemostunderlyingcontrastssurface.
Thus,inthecaseoftheparadigmof/tag/,themostinformativememberistheoneinwhichtheunderlyingvaluefor[voice]surfaces,i.
e.
theplural.
Bycontrast,inthesingularform/tag/=[tak],theobstruentisinfinalpositionandtheunderlying[voice]cannotsurface,sothatitisless"informative"abouttheunderlyingrepresentation.
Butwhatmotivatesthepromotionofbase-identityBaseIdentityisakindoffaithfulnessconstraint,soweexpectittoberankedbelowmarkednessinUniversalGrammar(Kager1999,Tesar&Smolensky2000).
SincepromotionofBase-Identityresultsinamoremarkedgrammar,wedonotexpectittooccurspontaneously22.
Theonlythingthatwouldactuallypromptlearnerstopromotethisfaithfulnessconstraintwouldbe,asweexpect,someotherchangethatcompelslearnertodeviatefromtheunmarkedgrammar.
Thatotherchangeis,ofcourse,apocope,whichresultedinvoicingalternationssurfacinginfinalpositionthatcannotbeaccountedforbyatransparentfinaldevoicinggrammar.
Werepresentapocopebythepromotionof*FinalSchwaoverMAX,andwepreventdevoicingfromapplyingbypromotingIdent-IO[vce]overFinDevoi:(16)Post-apocopeMiddleGerman/Proto-Yiddish22Thisisassumingthatspontaneousre-rankingsaretheprimarycauseofchange.
McMahon2000a,however,showedthatOTaccountsofchangefrequentlydonotspecifywhetherconstraintre-rankingsarethecauseofchange,orratherchange,i.
e.
phoneticshiftoutsidethephonologicalgrammar,isthecauseofthere-ranking.
Ifthelatteristrue,thentheexplanationforthechangewillnotlieintheframeworkofourphonologicaltheoryatall.
34/tag/*FinSchwaIdent-IO[vce]FinDevoiMAXtag**tag*!
tak*!
*tak*!
*ByrankingFinDevoibelowIdent-IO[vce],learnerspresumablynolongerhadreasontotreatFinDevoiasaninviolableconstraint,asitwasinMiddleGerman,andthereforethepromotionofBase-Identwasmadeeasier.
Butevenifweattempttobypassthecausalrelationshipbetweenapocopeandlevelingofvoicingalternationsbyappealingtoparadigmpressure,westillneedapocopetoexplainwhydevoicingceasedtobetransparentandautomatic,orwhyFinDevoiwasdemotedbelowIdent-IO[vce]23.
Inparticular,whyshouldthepromotionof*FinalSchwaoverMAX,whichresultsinapocope,leadtopromotionofIdent-IO[vce]overFinDevoi,whichresultsinlossoftransparentdevoicing2.
4.
2.
1OpacityandlossofdevoicinginOTTheproblemofcapturingthecausalrelationshipbetweenapocopeandopacityislinkedtothewiderproblemofrepresentingopacityinOTingeneral.
Varioussolutionshavebeenproposedforthis,allofwhichrelyinsomewayonstipulatingformalmechanisms,23Albrighthimselfadmittedtheinabilityofhisanalysistoexplainthecausalrelationshipbetweenapocopeanddevoicing,whichheseesisstillnecessarytoexplainthedemotionofFinDevoi.
Paradigmoptimality,inotherwords,canonlyaccountforthelevelingofvoicingalternations,notthelossofobligatory,transparentfinaldevoicingitself.
35overandaboverankedconstraints,thatallowtheexpressionofsynchronicopacityinapurelyparallelsystem(themorewell-knownmechanismsincludeOutput-Outputcorrespondence,Sympathy,andinterleavedorStratalOT).
ThemostpromisingapproachisStratalorinterleavedOT.
InStratalOT,eachmorphologicallevelhasitsownconstraintranking,i.
e.
grammar.
Theoutputofthegrammaratonelevelservesastheinputtothegrammarofthenextlevel.
Bermúdez-Otero&Hogg2003arguedthatStratalOTcouldcapturetheopaquerelationshipbetweendevoicingandapocopeinYiddish,providedthatdevoicingandapocopewereascribedtodifferentmorphologicallevels,i.
e.
devoicingtotheword-level,andapocopetothephrase-level.
Butinorderforapocopenottofeeddevoicing,wemustalsosupposethatIdent-IO[vce]outrankedFinDevoiatthephraselevel,asfollows:(17)MiddleGermanpre-apocopeWord-level/tag/FinDevoiIdent-IO[vce]tag*!
tak*Phrase-level36/tak/Ident-IO[vce]FinDevoitaktag*!
*(18)MiddleGerman/Proto-Yiddishpost-apocopeWord-level/tag/FinDevoiMAXIdent-IO[vce]*FinSchwatage*tag*!
*take*!
*tak*!
*Phrase-level/tag/*FinSchwaIdent-IO[vce]FinDevoiMAXtage*!
tag**take*!
*tak*!
*37However,itisnotclearwhyIdent-IO[vce]shouldhaveoutrankedFinDevoiatthephrase-levelinMiddleGerman.
Onepossiblereasonisthatwearelookingattheeffectofinputoptimization,i.
e.
deviationsfrominput-outputidentitywerejustifiedattheword-levelbecauseofalternationslike/tag/~/tak/,butnotatthephrase-level.
Theeffectofinputoptimizationcanbeillustratedbycomparingstem-levelandword-levelgrammarsinModernGerman:(19)ModernGermanStem-level/tag+Sg/Ident-IO[vce]FinDevoitag*tak*!
/tag+Pl/tag*tak*!
Word-level/tag+/FinDevoiIdent-IO[vce]tag*!
38tak*/tag+e/FinDevoiMAXIdent-IO[vce]*FinSchwatage*tag*!
*take*!
*tak*!
*Accordingtotheprincipleofinputoptimization,ateachlevelweassumelearnersstartwiththeassumptionofinput-outputidentity,unlessanalternationforcesadeviation.
Soatthestemlevel,final/g/in/tag/isthecorrectoutputinallpartsoftheparadigm,whichiswhythepluralis/tag+e/,not*/tak+e/.
Butatthewordlevel,where[tag]alternateswith[tak],thisalternationforcesadeviationfrominput-outputidentity,resultinginthehigherrankingofFinDevoi.
Theoutputofthewordlevel,[tak],thenbecometheinputtothephraselevel,i.
e.
/tak/.
Now,though,therearenoalternationsthatforceadeviationfrominput-outputidentity,sothelearnercontinuestoassumesurface[tak]=/tak/.
Thus,thehigherrankingofIdent-IO[vce]atthephrase-levelisindependentlymotivatedbythelearningprincipleofinputoptimization.
Whentheapocopeoccurs,then,wedonotneedtosupposethatIdent-IO[vce]wasre-rankedatthesametimeas*FinSchwa.
Wedonotneedtoappealtoanadhocmechanismlikeconjoinedconstraintre-ranking.
Ontheotherhand,wecannotfailtonoticethattheabilityofStratalOTtorepresentthiskindofopacityinanelegantfashionderivesfromitsmimicryofanorderedruleapproach,i.
e.
withtheuseofintermediaterepresentations.
39Moreover,thereisanothercriticismtakesbroaderaimattherepresentationofchangeinOT,namelythatthecausalorderofchangeandre-rankingisunclear(McMahon2000a:97;cf.
fn.
22).
Doeschangecausere-ranking,ordoesre-rankingcausechangeIfthere-rankingofconstraintscausesthechange,wehavetoexplainwhytheconstraintsshouldeverbere-ranked.
If,ontheotherhand,thechangesrepresentedhereoriginateoutsidethephonology,thenOToffersnoexplanatoryadvantageoverrule-basedphonology,andthereisnoparticularreasonforustopreferOToverorderedrules.
Onepossiblemotivationforre-rankingisthetendencytoreplacemarkedgrammarsbyunmarkedones.
Forexample,ifweassumethatacompletelyunmarkedgrammarranksmarkednessconstraintsoverfaithfulnessconstraints,wecanmotivatethechangefromOldGermantoMiddleGermanquiteeasily,sinceitwasinMiddleGermanthatthemarkednessconstraintFinDevoiwaspromotedoverthefaithfulnessconstraintIdent-IO[vce].
Theproblem,ofcourse,isthatthedemotionofFinDevoiinYiddishgoesinpreciselytheoppositedirection,sinceinthiscasewehavefaithfulnessbeingpromotedovermarkedness.
WehavealreadyexplainedhowOTmighthandlethelowerrankingofFinDevoi,namelythroughinputoptimizationatthephrase-level.
However,ifthelearnerstartsbyassumingMarkednessoutranksFaithfulness,thereisinfactnoevidenceatthephraselevelforthemtodeviatefromthisranking,either,becausefinaldevoicinginMiddleGermanwastrueatboththeword-andphrase-level.
Theresultisthatbothinputoptimizationandanunmarkedgrammarobtainatthephraselevel:(20)40Word-level/tag/FinDevoiIdent-IO[vce]tag*!
tak*Phrase-level/tak/FinDevoiIdent-IO[vce]tag*!
*takThisphrase-levelranking,ofcourse,givesthewrongpredictionsfortheoutputofapocope:/tag/*FinSchwaFinDevoiIdent-IO[vce]MAXtage*!
tag*!
*take*!
*tak**41WecanrankFinDevoiaboveIdent-IO[vce]onlyifweabandontheideathatlearnersapproachacquisitionwithcompletelyunmarkedgrammars.
Thisispossibleifweassumeconstraintsarelearned,ratherthaninnate(Hayes1999).
Inthatcase,wedonotassumelearnersstartbyassumingMarkedness>>Faithfulness,butratherthattheystartbyassumingFaithfulness>>Markedness.
However,assoonasweadoptthatapproach,wecannolongerexplainchangeastheproductofconstraintre-ranking:ifastablesynchronicconstraintrankingmustbelearned,thenachangingrankingmustalsobelearned.
Wecancertainlychoosetodescribethechangesintermsofconstraintre-ranking,buttheframeworkitselfdoesnotofferanymoreexplanatoryvaluethanarule-baseddescription.
2.
4.
3OpacityandlossofruleproductivityWesawabovethatitwasnotenoughtoappealtogrammaticalmarkednesstoexplainruleloss,sincerulere-orderingisanequallysufficientsolutiontotheproblemofmarkedgrammars.
Theremustbesomeotheraspectoftheopaqueorderingoffinaldevoicingandschwaapocopethatbroughtaboutthelossofthedevoicingrule.
Thataspectislearnability.
AsKiparskyexplains(Kiparsky1971),opaquerulesarehardtolearn.
Rulesthatarehardtolearnare,obviously,morepronetoloss(seealsoStampe1979).
Ifarulehasmanyexceptionstoitssurfaceapplication,i.
e.
ifitappliesinenvironmentswhereitisnotexpected,orifitfailstoapplyintheexpectedenvironment,thereiscorrespondinglyless42evidenceavailabletothelearnerthatsucharuleexistsatall,andthereforethelikelihoodthatthelearnerwillfailtoacquiretheruleincreases.
Ifweturnbacktothepost-apocopeparadigmsof/tag/and/geben/onpp.
21-22,weseethattheoutputsofapocopeprovidelessevidenceforfinaldevoicingthentheMiddleGermanparadigmsonp.
18.
Wealreadymentionedthefactthattheirregularparadigmsresultingfromapocopewouldhavetriggeredmorphologizationoftherule,i.
e.
thedevoicingrulewouldnowbeinterpretedasapplyingonlyincertainmorphologicalcategories.
Weassumethatsuchmorphologizedrulesarehardertolearn(Kiparsky1971).
AsKingnotes,however,atsomepointthenon-periphrasticpasttenseinYiddishhadbeenreplacedbytheperiphrasticpast(usingthecopulaandpastparticiple),sothatmuchoftheevidenceforadevoicingrulewouldhavebeenlost24:MiddleGerman'hegave',ergap,wasreplacedbyerhotgegebn,'hehasgiven'.
Thismeansthatthedevoicingrule,whichwehaveseenwasalreadymorphologizedasaresultofapocope,hasceasedtoapplyinthepasttenseofstrongverbsaswell,withtheresultthatthereiscorrespondinglyevenlessevidenceavailabletothelearnerforafinaldevoicingrule.
Moreover,thereisanotherlikelytriggerforthelossofdevoicing,namelytheeffectapocopehadinparadigmsinwhichnomembershadeverundergonedevoicing.
E.
g.
thesingularformsofweaknounslikegabe'gift'>Yiddishgob,pluralgaben>gobn,wouldnotshowdevoicingineithersingularorplural.
Therefore,devoicingwouldhaverelativelymanyexceptionsevenwithinaparticularinflectionalcategorylike24Thechronologicalrelationshipbetweenapocope,lossofdevoicingandlossofthenon-periphrasticpasttenseisunclear,however.
Conceivablydevoicingremainedasanopaqueruleforsometime,andonlythelaterlossofthenon-periphrasticpasttriggeredthecompletelossofdevoicingandconsequentparadigmleveling.
SeediscussioninKing1980:402.
43"nominativesingular",orinaparticularwordclass,like"nouns".
Itwouldnotbepossible,thatis,toaccountfornominativesingular[tak]byarule"devoiceinthenominativesingular",becausethenyouwouldalsoget*[gap]fromunderlying/gab/.
Instead,youwouldhavetospecifythatdevoicingapplyonlytosomenominativesingularforms,butnottoothers.
However,itisworthnotingthatthelackofevidenceforsynchronicdevoicing,asaresultofalltheexceptionsthataccruetodevoicingfromopacity,onlyobtainsifthelearnerisunabletoacquiresynchronicapocopeatthesametime.
Aslongasthelearnerhasbothrulesinhisorhergrammar,thenheorshewilltreattheexceptionstofinaldevoicingasmerelyartifactsoftheopaqueruleordering.
Wehavetocomeupwithareasonforlearnerstofailtoacquireasynchronicapocoperule.
2.
5Productivityofdevoicingandevidenceforsynchronicapocope2.
5.
1ApocopeandrestructuringofunderlyingrepresentationsThereisonegoodreasontosupposethatlearnersfailedtoacquiresynchronicapocope,andthatisthefactthat,afterapocopeoccurred,therewerenoalternationsbetweenfinalschwaandzero,whichwouldotherwisemotivateasynchronicapocoperule(seee.
g.
Bermúdez-Otero2006).
Inotherwords,afterapocopeoccurs,learnersareconfrontedwithaparadigmlikethefollowing:(21)SgPl44NomtaktagAcctaktagGentagstagDattagtagnThereisnoreasonforthemtoknowthatthefailureofdevoicingtoapplyinthedativesingularorthenominative/accusative/genitivepluralhasanythingtodowithafinalschwa.
WecanillustratetheproblemwithanotherexamplefromProto-Germanic,namelyGrimm'sLaw(Hock1991:251f.
).
Thisconsistedofthreechanges:(22)+stop>+fric/–obstr__–voicee.
g.
bhrātēr'brother'>brōēr,butstā-'stand'>–>stand-(23)+stop>–voice+voicee.
g.
gews-'taste'>keus-'choose'(24)+stop>+voice+asp–asp45(±fric)e.
g.
see'brother'above,initialsegmentInthisway,Proto-Indo-Europeanvoicelessstopsbecamevoicelessfricatives,unlessprecededbyanobstruent,voicedstopsbecamevoicelessstops,andvoicedaspiratedstopsbecamevoicedunaspiratedstopsorfricatives(alternationbetweenvoicedstopsandfricativeswasallophonicinProto-Germanic;fordetails,seediscussioninRinge2006).
Hockpointsoutthatonlythefirstchangeresultsinasplit,whichcanbedemonstratedbytheexampleofthepastparticiplesof'think'and'save':(25)Pre-Proto-Germanictonk-tó-nose-tó-Grimm'sLawθanx-tá-nasi-θá-Byotherchangesweeventuallygete.
g.
Gothic[θaxta],[nasia],andOldEnglish[θoxte],[nerede].
AttheperiodofdevelopmentafterGrimm'sLawoccurred,butbeforethelaterchangesoccurred,thealternationbetween[ta]and[θa]formsoftheparticipialsuffixwouldhavemotivatedasynchronicruleanalogoustoGrimm'sLaw,i.
e.
underlying/t/shouldbespirantized(becomeafricative),unlessprecededbyanobstruent(suchas[x]).
46Theothertwochangesdidnotresultinsplits:thus,Proto-Indo-European[g]became[k]inallenvironments25,e.
g.
Proto-Indo-European/gews/'taste,choose'>Proto-Germanic/keus/.
Ifnosplitresultsfromachange,thenthereisnopossibilityofalternationbetweentwoconditionedoutputs;ratherwehaveasingle,unconditionedoutput.
Underthesecircumstances,wecanspeakofrestructuringoftheunderlyingrepresentationsofthesounds,asintheexampleofunderlyingProto-Indo-European/gews/,whichbecameunderlyingProto-Germanic/keus/directly.
Ofcourse,wemustnowfigureouthowtoreconcilethepossibilityofsuch"restructuring"changes,wherenonewsynchronicrulehasresultedfromthechange,andtheequationofsoundchangewithruleaddition.
Oneconceivablesolutionistosaythatthechangeitselfconsistsintheadditionofarule,whetherunconditionedorconditioned,butthattherestructuringofrepresentationsoccurssubsequently,whenthesurfaceoutputsarere-analyzedbythenextgenerationofspeakers.
WithrespecttoYiddishapocope,thismeansthatthechangeitself,theinnovation,consistsinadditionofsynchronicapocope:(26)UR/tag//tag/FD--takApocopetag--SR[tag][tak]25Thelastchange,theonethataffectedPIEvoicedaspirates,didresultinanallophonicsplitbetweenvoicedfricativesandstops;however,aspirationwaslostacrosstheboard,sothisfeatureofthecorrespondingProto-Indo-EuropeanstopseriesnolongerparticipatedinalternationsinProto-Germanic.
47However,becauseapocopedoesnotresultinsynchronicalternations,thisgrammarisessentiallyunlearnable,sincethereisnoevidenceforasynchronicapocope.
Therefore,restructuringofthegrammaroccurs,withsimultaneousmorphologizationofdevoicing:(27)UR/tag+PL//tag+SG/Devoicing--takWeassumethatthedevoicingruleissensitivetomorphologicalfeatureslike[+plural].
Nevertheless,themorphologicalconditioningrendersithardertolearn,andthusmoreliabletoloss.
2.
5.
2ApocopeasavariableruleTheproblemwenowhavetoconsideristhis:ifthegrammarin(25)isunlearnable,howdiditariseinthefirstplaceToanswerthis,wemustturntothefindingsofvariationistlinguistics.
Asweknowfromstudyingsoundchangeinprogress,thereisalwaysaperiodofvariation,whenolderformsthathavenotyetundergonethechangealternatewithformsthathaveundergoneit(Labov1994).
Thisvariationcanbeexpressedastheoperationofavariablerule(Labov,Weinreich&Herzog1968).
Variablerules,justlikecategoricalrules,canbeordered,andsotheyreadilyfitintoanorderedruleframeworkthatallowsforsynchronicopacity(seealsothediscussioninHock1991,ch.
20)26.
26HockalsonotesthatLabovianvariablerulesareanalogousto"optional"rulesingenerativetheory.
Thedifferenceisthatclassicgenerativephonologycannotpredictwhenarulewillbeoptionalorobligatory,withtheresultthatnewrulesaddedtothegrammarcantheoreticallybeeither.
Laboviansocio-historicallinguistics,ontheotherhand,considersvariablerulestobeanecessarypreliminarystagetocategoriality;48IfweapplytheuniformitarianprincipletotheYiddishproblem,i.
e.
ifweassumethatapocopepassedthroughavariablestagebeforebecomingcategorical,thenwecanjustifytheadditionofavariableapocoperulethathadbeenaddedtotheendoftheseriesofrules,namelyafterthedevoicingrule.
Thevariablerulewouldbesynchronicallyjustifiedbecausevariantswithandwithoutfinalschwawouldbothbeavailabletothelearner:e.
g.
foratime,thepluralof/tag/wouldhavetwosurfacevariantsincompetition,[tag]and[tag].
Onlyuponcompletionofthechangewastheapocoperulelost,simultaneouslyreplacedbytherestructuringofunderlyingrepresentationsandmorphologizationofthedevoicingrule.
2.
5.
2.
1AbstractfinalschwaWeshouldbrieflyaddressthepossibilitythatanabstractfinalvowelstillexisted,evenafterapocopehadeliminateditonthesurface,sinceabstractsegmentshavebeenproposedbeforetoexplaincertaincomplexsurfacepatterns(e.
g.
Brame1972).
Thiswouldsolvetheproblemofhowtojustifyasynchronicapocoperulewhenfinalschwadidnotparticipateinanysurfacealternations.
Acasecanbemadeforthisanalysis,ifweallowthat,atleasttemporarily,word-finalvoicedobstruentspersistedinformslikeplural[tag].
Sincethesewouldberightlyperceivedasexceptionstofinaldevoicing,itisquiteconceivablethatlearnerspositedanabstractfinalvowelinordertoaccountforthefailureofdevoicingtoapply,i.
e.
theunderlyingrepresentationwouldstillbe/tag/.
noruleentersthegrammarasobligatoryfromtheoutset.
Moreover,variablerulesdonotalwaysbecomecategorical;undersomesocialconditions,theymayremainvariableindefinitely(Labov2001).
49Onemightobjectthat,sincethesegmentinquestionisabstract,thelearnerhasnowayofknowingwhetheritisaschwa,ratherthansomeothersegment.
However,ifweconsiderfinaldevoicingtobeconditionedbythesyllable-edge,andsinceonlyafinalvowelwouldtriggerre-syllabificationofthefinalobstruentintoonsetposition,bleedingdevoicing,itstandstoreasonthatthelearnerwouldpositanabstractvowel,notaconsonant.
Thisvowel,moreover,wouldmostlikelybeunspecifiedastoplaceofarticulation,makingitaschwaratherthananyfullvowel.
Themainobjectiontotheanalysiswithanabstractvowelisthatthesubsequentlossofthedevoicingrulecannotbeexplained.
Iflearnersarestillabletorepresentthepluralof/tag/,/tag/,withafinalschwa,thentheabsenceoffinalschwaonthesurfacecontinuestobemotivatedbyasynchronicapocoperule.
Andiflearnersarestillacquiringsynchronicapocope,thenthedevoicingrulewillnothaveacquiredtrueexceptions,norwillithaveundergonemorphologization.
Andifthedevoicingruleisstillphonologicallypredictable,thenitshouldnotbeliabletoloss.
Therefore,theeventuallossofdevoicingarguesagainstananalysiswithabstractfinalschwafollowingapocope.
3ApocopeandlossofdevoicinginEarlyModernGermanApocopealsoaffectedEarlyModernGerman,sowepredictthatthesamethingwouldhavehappenedhereasinYiddish:apocopewouldhaverenderedfinaldevoicingopaque,leadingtotheeventuallossofthedevoicingrule.
ThisisindeedwhatKranzmayerconcludedfromthelossofdevoicinginEarlyModernGermantexts.
However,wealsohavetoaccountforthefactthatModernGermanhasfinaldevoicing,andwewill50considerthepossibilitythatthepresenceofdevoicinginModernGermanisduetothedifferentwayapocopedevelopedinGermanasopposedtoYiddish.
3.
1MiddleGermanapocopeTheprocessbeganintheMiddleGermanperiod(1100-1400),whenfinalunstressedvowels,i.
e.
theschwa[],writteneinthetexts,begantodisappearinthewrittenrecord.
Theconditioningwasoriginallyphonological,indicatingitwasasoundchange,orperhapsseveralsoundchanges(Wright1907:77f.
):-Afterlandrindisyllableswithshortstems:are>ar,wole>wol-Afterlandrintrisyllabicandpolysyllabicformswithlongstems:blindere>blinder-Afternasalsinthefinalsyllableoftrisyllabicformswithlongstems:blindeme>blindem-Intheunstressedformsofdisyllables:ane>an,obe>ob(bothprepositions)Theterm'stem'intraditionalMiddleGermangrammarindicatestheshapeofthemorphologicalrootsyllable,whichwereinmostcasesalsothesyllablesofprimarystress.
Shortorlightsyllablescontainedshortvowelsandnocodaconsonant,i.
e.
CV;longorheavysyllablescontainedlongvowels,diphthongs,orcodaconsonants,i.
e.
CVVorCVC.
51Themetricalconditioningappearstobetheeliminationoftheunstressedvowelwhereitrepresentsthesecondmoraofatrochaicfoot,probablywiththemorare-assignedtotheprecedingsonorantconsonant:((ar)(e))Ft>(ar).
WhetherornotitispossibletosubsumealloftheMiddleGermanchangesintoasinglephonologicalstatementisnotrelevantatthemoment,althoughitappearsthatinthreeoutofthefouritemssomekindofsonorantschwaassimilationisresponsible.
Also,thesyllableprecedingthedeletedvowelwasalwaysunstresseditselfatthisearlystage.
Infact,Wrightdoesnottreatapocope(deletionoffinalvowels),andsyncope(deletionofword-medialvowels)separately,butattemptstoformulaterulesgoverningallcasesofschwadeletion,incontrasttoothertreatments(e.
g.
Reichmann&Wegera1993),asfollows:(28)MiddleGermanapocopeandsyncopeofschwaAfterliquids,/l,r/,indisyllableswithshortstems:are>ar'eagle',butname>name'name'wole>wol'well'(adverb)Afterliquidsinpolysyllableswithlongstems:blindere>blinder'blind'(genitiveplural)engele>engel'angels'(nominativeplural)52wandelete>wandelte'hechanged'Afternasals,/m,n/,inpolysyllableswithlongstems:blindeme>blindem'blind'(dativesingular)gevangene>gevangen'prisoner'Afternasalsinshortstemformsbeforefollowingt:nmet'you(pl.
)take'>nmtwonet>wont'hedwells'wonete>wonte'hedwelt'Inunstresseddisyllables:ane>an'on'abe>ab'off'unde>und'and'Wrightdoesnotsayhowthesefivechangesarechronologicallyorderedrelativetoeachother,butitappearsthattheinitialconditionwasprecedingliquidifthesyllableprecedingtheunstressedsyllablewasheavy.
Themoraicschemasuggestedabovesupportsthis:53(29)en.
ge.
le=((én))Ft((gé)(lè))Ftna.
ge.
le=((ná)(gè))Ft((lé))FtInengele,thefinaleisdeletedsinceitfallsontheweakbranchofthesecondfoot.
Innagele,however,thefinaleisnotonaweakbranch,soitisnotoriginallysubjecttoapocope.
Wrightmentionsthatapocopewaslateranalogicallyextendedtothiseaswell,butitisnotclearwhetherthisistodowithamorphologicalorphonologicalre-analysisoftheearlierapocope.
Apparentlytheenvironmentwasextendedtoprecedingnasals,butonlyiftheprecedingsyllablewasunstressed.
Sowehaveblíndeme>blíndem,likeblíndere>blinder,butnameremainsname,whileáre>ar.
Withrespecttoprecedingliquids,theschwaonlyneedstobeintheweakbranchofthefoot;withrespecttoprecedingnasals,however,theschwaneedstobeintheweakbranchofthefoot,andthefootneedstobeintheweakbranchoftheprosodicword.
Adoptingahierarchicalmetricalstructureasfollowscancapturethisconditioning.
Germanmetricalstructureisleftdominant,sowefindthattheschwadeletedisintherightmostbranch,i.
e.
theweakestmetricalslot.
Themoraattachedtothefinalsyllable,–me,isre-attachedtotheprecedingsyllable,whichisnow–dem:54(30)PrWdFootFootMoraMoraMoraMorablinddemeAftertheprimarystressedsyllable,aswesaw,deletionoftheschwaoccursifprecededbyaliquid,butnotanasal.
Inunstressedwords,ontheotherhand,apocopeoccurredeveniftheprecedingsegmentwasanobstruent:abe>ab,unde>und.
Therefore,weshouldbejustifiedinsayingthat,asschwadeletionwasextendedtotheenvironmentafterthestressedsyllable,atfirsttheprecedingsegmenthadtobealiquid,i.
e.
whereithadoriginallyoccurredafterunstressedsyllables;meanwhile,afterunstressedsyllables,theenvironmentwasextendedtoanysegment.
Sotheenvironmentofthedeletionrulewasbeingextendedintwodirections,asitwere.
Therearetwoproblemswiththisdescription.
Exampleslikewonete>wonteshowthatsometimesdeletioncouldoccurafterstressedsyllablesfollowedbynasals,ifthefollowingsegmentwas/t/.
Thisleadstothesecondproblem,namelythatthefinalvowelintheweakpreteritsuffix–tedoesnotundergoapocope.
Thus,wehavewandelete>wandelte,notwandelet.
55However,Keller1978:406showsthatwehaveformslikewandlenalongsidewandeln,andthatapocopeofschwain–tedidinfactoccur.
ThealmostuniversalexistenceoffinalschwainModernGerman–teisduetoanalogicalrestoration,orpossiblydialectborrowing.
Astowhywefindexamplessuchaswonet>wont'hedwells',butname>name,thisagainmayhavesomethingtodowithmoraicstructure.
I.
e.
thefinalschwainnameisthesolemora-bearerinthatsyllable:(na)(me),butinwonet,boththeschwaandthefollowingtcanbearmoras,meaningthatthefinal–etcouldbeconsideredheavy:(wo)(net).
Soifthechangehadbeenoneofreducingsyllableweightinunstressedsyllables,thenitwouldmakesensetohave(wo)(net)become(wont).
Thesetentativeinterpretationsofthisextremelycomplicatedpictureareonlymeanttodevelopthepointthatapocopewascertainlyinoriginaphonologicallydrivenprocess,evenifatalaterstageanalogyandborrowingobscuredthepicture.
ThisisanimportantthingtokeepinmindespeciallywhenwecometothequestionofwhetherfinaldevoicingappliedtotheoutputsofapocopeinEarlyModernGerman.
3.
2UpperandWestCentralGermanapocopeThedatareviewedbyWright1907focusesalmostexclusivelyonapocopeasitoccurredinliteraryEarlyModernGermanandModernGerman.
Fromthesixteenthcenturyon,themostprestigiousvarietyofGerman,thevarietyuponwhichWright'sgrammarisbased,wasthatofEastCentralGerman,i.
e.
UpperSaxony,Thuringiaandlateron56variousformerLowGerman-speakingnortherncitiesthatadoptedEastCentralHighGermanintheearlymodernperiod:Berlin,Hamburgetc(seeKeller1978,Knig1978).
Asithappens,itwasmainlyintheEastCentralGermanregionthatwefindsignificantirregularitiesintheapplicationofapocope,whichweshallturntolater.
Inmostofthesouthern,UpperGermanregion,andalsoeventuallyinWestCentralGerman(RhineFranconianandCentralFranconian),apocopeappliedmuchmoreregularly.
Notonlythat,butearliertextsfromtheEastCentralregionshowmoreapocopethanlatertexts,indicatingthattheoriginalEarlyModernGermanapocopewasmuchmoreobviouslyaunified,regularsoundchange,andthatthephonologicalandmorphologicalpeculiaritiesfoundinlaterEastCentralGerman,asdescribedinWright1907andotherstandardsources,aretheresultsofchronologicallysubsequentchangesthatobscuredtheregularpatternsoftheoriginalchange(seebelow).
Thegeneralpatternscanbeshowninthefollowingtable(fromLindgren1953:178;Bav=Bavarian,EFranc=EastFranconian,Swab=Swabian,HAlem=HighAlemannic,LAlem=LowAlemannic,Boh=Bohemian,RhFr=RhineFranconian,EastCentralGerman=EastCentralGerman):(31)EarlyModernGermanapocope27%-eBavEFrancSwabHAlemLAlemBohRhFrEastCentral27TheblanksintheEastCentralGerman(EastCentralGerman)columnindicatethefactthatapocopedidnotbeginuntilaround1500,whichisLindgren'scut-offpoint.
Similarly,theblankcellsunderBohemianandRhineFranconianindicatethatfrequencyoffinalereached10%onlyafter1500intheseregions.
57German901200130013001325132513501400--501275137513751400142514001425--1013751425142514251450------Theleftmostcolumnrepresentsthepercentageofretentionoffinalschwainformswhereitshouldhaveoccurredin(standardized)MiddleGerman.
Theyearsindicatethemid-pointsofthecorrespondingperiod.
Aswecansee,apocopeisfirstattestedinBavarianintheearly13thcentury,andhadproceededsignificantlybythelate13thcentury(Lindgren1953:177f.
).
Fromthenon,apocopeappearstospreadtoneighboringregions.
Theblankcellsindicatewhereretentionofschwahadnotyetdeclinedtotherespectivepercentageby1500.
Lindgrenchose1500asthecut-offpointbecauseconvergenceofthevariousliterarytraditionsbegantotakeplaceinthe16thcentury,undertheinfluenceofthenowprestigiousEastCentralstandard,whichhappenednottohaveundergonecompleteapocopebythattime.
Forthisreason,afterabout1500finalschwabegantoberestoredintheliterarydialectofmostoftheseregions(althoughnotinthespokendialects).
OneofthefindingsofLindgren1953wasthatapocopeinalltheseregions(exceptEastCentralGerman)proceededinanunambiguouslyregularfashion,whentheshiftinpercentagesthatareoutlinedinthetablearerepresentedgraphically(Lindgren1953:182-185).
Lindgrenmeasuredthedifferentpercentagesinretentionoffinalschwaandgroupedthembymorphologicalcategory,inordertoascertainwhetherthereweregrammaticaleffectsontheprogressofthechange.
Infact,therewerenosignificant58morphologicaleffectsuntilneartheendofthechange.
Apocopeproceededatthesamerateinallcategories,followingaveryclearS-shapedcurve,whichisthetelltalesignofanykindofregulargrammaticalchange,includingphonologicalchange(seeKroch1989,Labov1994).
Lindgrentreatedapocopeaftertheliquids/r/and/l/separately,sincehefound,asWrightdid,thatapocopeoccurredearlierintheseenvironments.
Unfortunately,hedidnotalsomeasuretheratesofapocopeaccordingtopositioninthemetricalstructure.
Tosomeextent,thatisnotimportantforourpurposes,becauseheprovedbeyondadoubtthatEarlyModernGermanapocopewasaregularsoundchange.
Thisisimportanttorememberwhenwecometodiscusstherelationshipbetweensoundchangeandadditionofphonologicalrules28.
3.
3InteractionoffinaldevoicingandschwaapocopeinEarlyModernGerman3.
3.
1TheBonncorpus3.
3.
1.
1MethodologyanddataTheBonnerFrühneuhochdeutschkorpus(BFK)isacollectionofEarlyModern(1350-1700)HighGermantexts,producedbytheUniversityofDuisburg-Essenunderthe28Somegrammaticalconditioningappearstowardstheend,whenthepercentageofretentionoffinalschwabeginstoriseagainininflectedadjectives,butnotinothergrammaticalclasses.
Wecertainlyneedtoexplainhowthiscanhavehappened,despitetheNeogrammarianregularityhypothesis,butatthesametime,therestorationofschwaininflectedadjectivesisclearlytheanomalous,ratherthanthepredominantelementinthepicture.
Therefore,wearejustifiedintryingtoexplaintheirregularitiesthatwedoseebypositingsubsequentchangesthatmayhaverestoredschwainthoseformswhereitappearsinthemodernlanguage.
59directionofMichaelElmentaler,inwhichwordsareindividuallytaggedwiththeirlemmataandrelevantmorphologicalinformation29.
Thisallowsustocompareattestedwordformswithexpectedforms,thatis,whattheMiddleGermanandEarlyModernGermangrammarstellustoexpectforthatwordinthatmorphologicalcategory.
Thus,thenominativepluralof/tag/shouldbetage;wethensearchforthenominativepluraloftaginthecorpus,andturnuptheactualattestedforms,whichcanvaryfromtagetotag,tach,dagandsoon.
Thetextsaredividedintotendialectregions.
Itshouldbeborneinmindthat,inthecontextofMiddleGerman,'dialect'meansregionalvariantsofapartlystandardizedwrittenlanguage.
Notethatregionalvariantsarenotdevelopmentsfrom,letalone"corruptions"ofaformerlypristineanduniformliterarystandard,asscholarssincetheGrimmbrothershavebeenwonttoclaim(Keller1978:252-55,370-80);rather,thereneverwasanoriginalliterarystandard.
AstandarddidgraduallyemergeduringtheMiddleAges,becomingmoreorlesscodifiedduringthesixteenthcentury.
Thestandardizationprocesswasdrivenfrombelow,followingthegradualsocial,culturalandeconomicunificationthatgatheredpaceinthecenturyimmediatelyprecedingtheReformation.
Continueddialectaldifferentiationwasduetotheinterferenceoftheauthor'snativedialectinthegrammarofthisprestigevariety.
Thisprestigevarietyordialect,atleastwithrespecttoitsobstruentsystem,wasbasedmainlyonmedievalEastFranconian,adialectthatpreservedtheoverallcontrastbetweenvoicedandvoicelesscategories,butwhichneutralizedthiscontrastinword-finalposition.
Thereislittleevidencefortheeffectoflenitioninthisvariety(althoughsee29Thecorpusisavailablefordownloadathttp://www.
korpora.
org/Fnhd.
Notallthetextsoftheoriginalcorpus,compiledinthe1970sand1980s,havebeentagged,hencethepeculiarnumberingsystemyouseebelow(111,113,115,117,121,123,etc.
).
60discussioninMihm2004onsomeoftheevidenceinthelateMiddleGermanperiodforlenition).
Nevertheless,despitethisincipientstandardization,thereisalsoevidenceforvariationandchangeeveninthewrittenvariety,atleastduringthispre-modernstagewhenspellinghadnotbeenrigidlycodified.
Thegreatestvaluetousistheinsightofferedintotheextentoffinalneutralizationontheonehand,andthelossoffinalschwaontheother,bothofwhicharereflectedinthespelling.
Theinterpretationofthelossofdevoicingisnaturallymorecontroversial,butinthefollowingsectionsIwilldemonstratethatvariationinthespellingoffinalneutralizationdoesreflectphonologicalvariationandchange.
Thereare40textsinall,dividedintotensetsoffoureach.
Eachsetrepresentsaparticulardialectregion,withtheresultthattendialectregionsofEarlyModernGermanarerepresentedintotal.
Ineachset,onetextisselectedfromeachoffourperiods:1350-1400,1450-1500,1550-1600,and1650-1700.
Theresultisthatthetextsrepresentbroadlyequalintervalsof100yearsforalldialectregionsunderconsideration.
Inthissubsection,Iwillpresenttheresultingstatisticsbydialectarea,anddiscusstheevidencetheyshowforlossofdevoicinginthenextsubsection.
Thesearchconsistedofcomparingattestedformswiththelemmataandrelevantmorphologicaltags30.
Ifthelemmashowedastem-finalvoicedobstruentandtheattestedformshowedthestem-finalobstruentsurfacingword-finally,thatformwasincludedintheanalysis.
Allothers,wherethestem-finalobstruentdidnotsurfacefinally,wereexcluded.
Theselectedformswerethendividedaccordingtowhetherthefinalobstruent30Lemmataconsistedeitheroftheunderlyingformofthestemfornounsandadjectives,sowrittentacwouldbetaggedtag,orelsetheunderlyingformoftheinfinitiveforverbs,sowrittengap(apasttenseformof'togive')wouldbetaggedgeben.
61waswrittenwithavoicelessgraph('p,ph,f,t,th,k,kh,c,ch')oravoicedgraph('b,d,g,gh');theresultswerelabeled'Devoiced'and'Voiced',respectively31.
Thesewerethensubdividedaccordingtotwocategories:'Apoc'(i.
e.
apocopated)and'NoApoc'(i.
e.
non-apocopated),thatis,whereweexpectthefinalzerotobeduetoapocopeornot.
Theformercategoryconsistsofthefollowing:-nounswhoselemmaendsinzero:odativesingularandnominative/accusative/genitivepluralofmasculinesopluralsoffemininesodativesingularandgenitivepluralofneuters-nounswhoselemmaendsinschwa(ja-stems,ō-stems,andweaknouns)-inflectedadjectives-verbs:ofirstpersonsingularpresentindicativeofstrongandweakverbsofirst/thirdpersonsingularpresentsubjunctiveofstrong,weakandpreterit-presentverbsosingularimperativeofweakverbsosecondpersonsingularpreteritindicativeofstrongverbs(excludingformswithinnovative–stending)ofirst/thirdpersonsingularpreteritsubjunctiveofstrongverbs31Ihaverestrictedthesearchtostops,sincefricativeswerenotasconsistentlymarkedforvoicinginMiddleGermanorEarlyModernGerman,especiallyforthecontrastbetween[z]and[s].
Also,therewasaparalleldevoicingofvoiced[v]inEarlyModernGermanthatwouldhaveinterferedwiththealternationbetweenmedial[v]andfinal[f](Keller1978,Paul2007).
Theterms'Fort'and'Voiced'standliterallyfor'Fortition'and'NoFortition',sincefortitionisanothertraditionaltermusedforMiddleGermanandEarlyModernGermandevoicing.
62opresentparticiplesThelattercategoryconsistsofthefollowing:-nounswhoselemmaendsinzero:onominative/accusativesingularofmasculinesosingularsoffemininesonominative/accusativesingularandpluralofneuters-uninflectedadjectives-verbs:ofirst/thirdpersonsingularpresentindicativeofpreterit-presentverbsosingularimperativeofstrongverbsofirst/thirdpersonsingularpreteritindicativeofstrongverbsSomeunwantedformsmighthavebeencollectedevenwithsuchfine-grainedcategories:forinstance,ja-stemadjectiveslikemüdeshouldhavehadfinalschwaeveninuninflectedforms.
However,thecorpustagsthemwithschwa-lesslemmata,sothatIhadtotreatthemthesameasa-stemslikeblind.
Likewise,uninflectedadjectivesinMiddleGermanwerefrequentlyusedinsteadofthestronginflectedformsformasculineandfemininenominativesingular,orneuternominativeandaccusativesingular:so,blintinsteadofblinder,blindiu,blindez.
Thecorpustagsthemasinflected,however.
Nevertheless,Ididnottreattheseadjectivalcategoriesasnon-apocopatedbecausethenominativesingularoftheweakadjectivedeclensionhadfinalschwainallgenders63(blinde),andthecorpusdoesnottagadjectivesforstrongorweakdeclension.
Therefore,Iincludedallinflectedadjectivesinthe'Apoc'category,andalluninflectedinthe'NoApoc'category,withthehopethaterrorsineachcategorywouldcanceleachotherout.
Similarly,nounswerenottaggedforinflectionalclass,butforgender.
InMiddleGerman,paradigmaticdifferenceswithingenderswerebeingleveledout,butitispossibleonewouldstillfind,forexample,genitiveanddativesingularfemininei-stemswithfinalschwaandumlautofthestem-vowel,e.
g.
hende,ratherthantheinnovativeModernGermanpattern,inwhichallsingularcategoriesareidenticaltothenominativesingular(noschwa,noumlaut),e.
g.
hant.
Nevertheless,Ichosetotreatallsingularfemininenounswithschwa-lesslemmatathesameway,asnon-apocopated,inordertofacilitatethesearch.
Thesearchtookinonlynouns,adjectivesandverbs.
Thereareasizeablenumberofotherwordsthatoriginallyhadfinalschwa,namelytheunstresseddisyllableslikeunde>und,abe>ab,whichincludeonlyprepositions,conjunctionsoradverbs.
SincethesewordshadalreadylostfinalschwainMiddleGerman,theirlemmataaregivenwithoutschwainthecorpus.
Interestingly,theynevershowfinaldevoicing,whichmaybeduetoopacity,orelsemayhavesomethingtodowiththeirlackofstress,whichmightcorrelatewithlenitionandhencevoicedgraphs.
Whatevertheexplanation,theywerenotabletoskewtheresults.
InModernGerman,finalschwawasinfactintroducedintosomecategoriesinwhichitwasnotoriginallyfoundeveninMiddleGerman:thenominativeandaccusativepluralofneutera-stemnounsisagoodexample.
Itispossiblethatthisinnovativefinal64schwacouldthenhavebeensubjecttophonologicalapocope.
However,Ichosetoincludeallsuchcategoriesinthenon-apocopatedgroup,sinceIwasprincipallyinterestedinthetransitionfromMiddleGermantoModernGerman.
Insum,inordertofacilitatethegatheringofdata,IassumedafairlystandardizedMiddleGermaninflectionalsystem,basedonstandardhandbookslikeWright1907andPaul2007.
Aftertherawnumbersofeachcategorywerecollected,thepercentagesofdevoicingintheapocopatedandnon-apocopatedcategorieswerecalculated,aswellasthepercentageoftotaldevoicing.
Thentherawnumbersweresubjectedtoachi-squaredtesttoseeifthedifferentproportionsofdevoicinginthetwocategoriesweresignificant.
Ap-valueoflessthan0.
05isgenerallyconsideredsignificant.
Whatwehopetoseeisthatdevoicingshouldbesignificantlymorefrequentinwordsthatdidnothaveafinalvoweloriginally,sincethatwouldcorrespondwithourpredictions:opaqueorderingofapocopeafterdevoicingwouldresultinexceptionstodevoicingonlyinthosewordsthataretheoutputsofapocope,i.
e.
thosewordsthatformerlyhadfinalschwa.
Ofcourse,wealsopredictthatsuchopacitywouldtriggerrulelossandlevelingofvoicedobstruentsthroughouttheparadigm,leadingtocompletedisappearanceofdevoicingfromtherecord.
Nevertheless,thereshouldbeanintermediateperiodoflexicalizeddevoicingapparentinthetextsinordertosecureourcase.
MiddleBavariantextsTextno.
11165ApocNoApocTotalDevoiced194261Voiced16420136510%17%14%X-squared=3.
5095,df=1,p-value=0.
06102Textno.
113ApocNoApocTotalDevoiced124456Voiced2522635155%14%10%X-squared=14.
2831,df=1,p-value=0.
000157366Textno.
115ApocNoApocTotalDevoiced82533Voiced1062433497%9%9%X-squared=0.
2879,df=1,p-value=0.
5915Textno.
117ApocNoApocTotalDevoiced000Voiced521041560%0%0%X-squared=Inf,df=1,p-valuedigraph.
Aswewillseelater,thisdigraphprobablyindicatestheeffectofinnerGermanlenition,notdevoicing.
92Glaser'smethodwasverysimilartotheoneusedtoanalyzetheBonncorpus.
TheactualspellingswerecomparedwithMiddleGermancitationforms,suchastac'day'.
Glasercoinedtheterm"meta-phonemic"todescribethephonologicalstructureofthecitationforms.
Tosomeextentthiscorrelateswithunderlyingrepresentation,where,forexample,thecitationformrepresentsahistoricallyolderformthatmaystillberecoverablefromalternations,suchasthepluraltage.
Thefinal"meta-phonemic"vowelcouldstillberecoveredwheretheformvariedwithapocopatedtag.
Inothercases,themeta-phonemedoesnotreflectunderlyingrepresentation,butahistoricallyoldersurfacerepresentation,e.
g.
thefinalcintac,wherewesupposethatevenattheearlieststagetheunderlyingobstruentwas/g/.
AlthoughGlaserdidnotmakeuseofgenerativephonology,andhenceoftheopportunitytodescribethechangesintermsofshiftingsurfaceandunderlyingrepresentations,shediddistinguishbetweenalternatingandnon-alternatingphonemes.
Thiswasparticularlyusefulindistinguishingword-finalsurface[t]derivingfromunderlying/t/,andsurface[t]derivingfromunderlying/d/36.
Theresultwasasetofsixalternatingmeta-phonemes:/b/,/d/,/g/,and/p/,/t/,/k/,whichGlasercomparedwiththeirgraphicrepresentations,focusingonlyonword-finalposition.
Theformerthreerepresentedthosevoicedstopsthatoriginallyonlysurfacedmedially,butafterapocopecouldalsosurfacefinally;herrestrictionofthecomparisontoword-finalpositionmeansthatonlyapocopatedformswereconsidered.
Thelatterthreerepresentedthosevoiceless36Glaserconsideredmostcasesword-finalunderlying/k/toinvolvegeminate/kk/,i.
e.
phonemicallydistinctfrom[k]derivingfromunderlying/g/.
Surface[p],ontheotherhand,couldintheoryreflecteitherunderlying/b/or/p/,butinrealityexamplesofunderlying/p/wereveryrare.
Ageminate/tt/didexist,butwasrestrictedtomedialposition,whereitcontrastedwith/t/and/d/.
93stops(derivingfrometymologicalvoicedstops)thatwerefoundoriginallyinfinalposition.
Inpost-vocalicfinalposition(afterapocope),/b/waswrittenb90%ofthetimein1276,80%ofthetimein1373,andin1483and1523100%ofthetime.
Final/d/isonlyattestedfrom1373,whereitiswrittend100%ofthetime;in1483itisalso100%d,butin1523itisonlydabout45%ofthetime.
Final/g/isnotattestedin1276,butisfoundasg100%in1373,1483,and1523.
Inpost-consonantalposition37,wefind/b/asb100%inallfourtexts;/d/isd0%in1276,85%in1373,andabout90-95%in1483and1523;finally,/g/isg0%in1276,but100%afterwards.
Asfortheoriginalvoicelessphonemes,alternating/p/(fromunderlying/b/)inpost-vocalicpositionisb0%in1276,35-40%in1373,and100%in1483and1523.
Alternating/t/isd0%in1276,90%in1373,80-85%in1483,and10%in1523.
Alternating/k/isgisabout25%in1276,80%in1373,and100%in1483and1523.
Inpost-consonantalposition,/p/isb0%in1276,and100%afterwards;/t/isd0%in1276,10-15%in1373,70-75%in1483,and15%in1523;while/k/is0%in1276,20%in1373,80%in1483,and60%in1523.
Interpretedingenerativeterms,herstatisticsrevealthat,atleastintheearliestversionofthetext(1276),finalunderlyingvoicedobstruentsinapocopatedformsweremuchmorelikelytobewrittenwithvoicedgraphsthanthoseinnon-apocopatedforms.
Inotherwords,exceptionstofinaldevoicingwereonlyfound,forthemostpart,incaseswherewepredictsuchexceptionstohavearisenfromtheopacifyingeffectofschwa37Glaserdoesnotdistinguishbetweenobstruentsandsonorantsinherdefinitionofpost-consonantal,sothatpost-consonantalinstancesofdevoicingmayindicateeithervoicingassimilation(aftervoicelessobstruents)orfinaldevoicing.
However,sincesheonlyincludedvoicelessstopsthatalternatedwithvoicedones,inpracticepost-consonantalcouldonlyincludepost-sonorant,sinceonlyafterasonorantwouldvoicedandvoicelessobstruentscontrast.
94apocope.
Inlatertexts,thesignificantcorrelationofexceptionalfinalvoicingandapocopeislost,asthevoicedgraphsareincreasinglywrittenfortheunderlyingvoicedstopsinbothapocopatedandnon-apocopatedforms,i.
e.
weareseeingtheeffectofleveling,orlossofthedevoicingrule.
ThisstronglycorroboratesourfindingsfromtheBonnCorpus:apocopeleadstoopacityorexceptionstothedevoicingrule,whichinturnleadstolossoftheruleandlevelingofalternations.
Themajoranomalyhere,ofcourse,isthebehaviorofthemeta-phonemes/d/and/t/,i.
e.
underlying/d/(andtoamuchlesserextent,/g/and/k/).
Althoughtheuseofthevoicedgraphclimbsdramaticallyinthesecondtext(1373),wefindacuriousrestorationofthevoicelesstinthelasttwotexts.
ForthosewhobelievethatMiddleGermandevoicingsurvivedintoEarlyModernGermanthroughapocope,thismightbestrongempiricalconfirmationofthishypothesis.
Perhaps,afterall,apocopedidnotcausethelossofdevoicing,butonlyatemporaryperiodofopacity,latercorrectedbyre-assertionofthedevoicingrule.
Theproblemwiththishypothesisisthatitonlyaccountsforthere-assertionoffinaltfor/d/;finalkfor/g/isinfactonlysparselyattested,whilefinalpfor/b/isnotattestedatall.
There-assertionofthedevoicingrulewouldthusberestrictedtoonlyoneunderlyingvoicedstop,namely/d/.
Sincethestructuraldescriptionoftheoriginaldevoicingruledidnotdistinguishbyplaceofarticulation,butaffectedallthreeunderlyingstopsmoreorlessevenly,itisunlikelythatwearelookingatthere-assertionoftheoriginaldevoicingrule.
Weneedtofindanotherwaytoexplainthisphenomenon.
ItisherethatperhapsMihm'shypothesisaboutlenitionhasaparttoplay.
Lenitionwasatruesoundchange,andthereforewewouldnotbesurprisedifit95proceededatdifferenttimesandratesaccordingtoplaceofarticulation.
Althoughinthosedialectswherelenitiontookplace,allthreestopswerelenited(Keller1961),whilethechangewasinprogressitisplausiblethatwewouldseeitaffectoneplaceofarticulationbeforeanother.
Couldtherestorationoffinaltinfactrepresenttheonsetoflenition,restrictedatthattimetocoronalsAtfirst,thisseemsincorrect,sincelenitionshouldhavereplacedtbyd,notviceversa.
Nevertheless,acloserexaminationoftheselateinstancesoftrevealthatinfacttheyreflectthenewdigraphdt,largelyunknowninMiddleGerman,butverycommoninEarlyModernGerman.
Theuseofadigraphstronglysuggeststhatscribesandprinterswereattemptingtorepresentasoundsomewhereinbetween[d]and[t],whichhappenstomatchinnerHighGermanlenited[d]verywell38.
Theuseofoccasionalgktowritefinal/g/alsosuggeststheoperationoflenition.
Theabsenceof*bpmightindicatethatlenitionhadnotyetspreadtothelabialseries,andbythetimeitdid,standardizationofthespellingpreventedanyfurtherinnovationsinorthographythatmightreflectongoingphonologicalchange39.
Someadditionalstrongevidencethatdtrepresentedtheeffectoflenitioniswhereitwasusedtorepresentoriginalunderlying/t/,e.
g.
thegenitivesingularmasculinetten38HistoricallinguistsofGermangenerallyholdthatthemedievalcontrastinvolvedasignificantphoneticdistinctionbetweenfortis(tensevoiceless)andlenis(laxandprobablyvoiceless).
Theamountoftruevoicingisdebated(seediscussioninPaul2007),butwasprobablyrelativelyinsignificant,judgingbysoutherndialectsthatpreservethedistinction(e.
g.
Swiss).
TheinnerHighGermanlenitionsimplyeliminatedmarkedfortesinfavoroftheunmarkedlenes,aprocessthatforconveniencewehavebeendescribingasmergerinfavorofthevoicedseries.
SeeIverson&Salmons2007andreferencesforwhylenitionshouldnotbedescribedasvoicing,althoughtheyidentifyfortisortensearticulationwithaspiration,eventhoughaspirationwasprobablynotasalientfeatureofMiddleGermanfortes(Paul2007:121).
39TheuseofdtremainedcommonthroughtheEarlyModernGermanperiod,beinggraduallyeliminatedinthe19thcentury,asdbecamethepreferredgraphforunderlying/d/.
Atthispointwecanprobablyspeakofanewmorphemicprincipleofspelling,since,accordingtothisauthor,itwasinthe19thcenturythatfinaldevoicingwasfinallyreintroducedintothestandardliterarylanguage,owingtonorthern,LowGermaninfluenceintheHighGermanspeechofcitieslikeBerlinandHamburg.
96'dead',writtenastoten,totten,andtodten.
Theaccusativesingularneutertteinturniswrittentotez,tottes,todes,andtods.
Soweevenfinddformedial/t/!
Theonlyplausibleexplanationfortheuseofthevoicedgraphhereislenitionof/t/to[d].
Therefore,wearejustifiedinattributingtheinnovativeuseofdt,nottore-assertionofdevoicing,buttoinnerHighGermanlenition40.
3.
4ApocopeandlossofdevoicinginGermandialectsWehavefoundrobustevidenceinthetextualdataforthehypothesisthatthelossoffinaldevoicinginEarlyModernGermanwasarealphonologicalchange,broughtaboutbyapocopeoffinalschwa.
Apocoperesultedinmanysurfaceexceptionstodevoicing,whichinturntriggeredthelossoftheruleandeventuallevelingofvoicingalternations.
However,itwouldbeevenmorehelpfulifwecouldfindevidenceofthissequenceofeventsindevoicingdialectsthatunderwentapocopeinrecenttimes.
Thebestplacetolook,ofcourse,isdialectsinareaswhereapocopeisstillspreading.
ThemapsgivenintheDTVatlasoftheGermanlanguage(Knig1978:158f.
)showtheapproximatebordersofapocopearound194041.
Thereareinfacttworegionsof40NotethatthisevidenceoflenitioninourlatesttextsdoesnotsupportMihm'stheorythatthelossofdevoicinginEarlyModernGermanisduetolenition.
Theevidenceoflenitiononlyappearsafterapocopehaslargelygonetocompletionanddevoicinghasalreadybeenlost,andthecorrelationbetweenapocopeandexceptionstofinaldevoicingis,aswesaw,toorobustlyattestedforourearliertextstodismisstheapocopetheory.
Intheearliesttext,thevoicingcontraststillobtains,andthedevoicingruleisstillactive,thoughithasgainedsomeexceptionsthroughapocope;inthesecondtext,therearemanymoreexceptions,andlevelingofthevoicedalternantshasensued;inthethird,levelingofthealternationsismostlycomplete,butlenitionisstillincipient;inthefinaltext,lenitionhasproceededconsiderablyincoronals,andhasstartedtoshowindorsals.
41PopulationmovementsafterWorldWarTwoconsiderablydisruptedthedialectsituation.
Theeasternthirdofpre-warGermanywasalmostentirelystrippedofitsGerman-speakingpopulation,anditisfrompreciselytheseareas(Silesia,Pomerania,EastPrussia)thatthedialectalevidencepresentedinthissubsectionisdrawn.
97apocope,oneinthenorth,whereLowGermanisspoken,andonecomprisingmostofthesouthernHighGermanarea.
TheareawhereapocopedidnotoccurcomprisesSilesia,BrandenburgandpartsofsoutheasternPomerania,northernThuringia,mostofUpperSaxony,Westphalia,Eastphalia,southernLowerSaxonyandtheregionsborderingthenortheasternNetherlands.
Asthemapsindicate,theareaofnoapocopeisshrinking,meaningthatinthosedialectsstraddlingtheborderoftheapocopatingregion,apocopeislikelytobearecentlyintroducedrule,andthereforewillbelikelytobeopaquelyorderedwithrespecttodevoicing42.
Ofcourse,therangeofdialectareaswemightobserveopaquedevoicingwillbemorerestrictedthanthat,becauseofinnerHighGermanlenition,whichneutralizedthevoicingcontrastinallnon-initialenvironments.
Mostofthedialectsinthesouthernregionwhereapocopeoccurred,i.
e.
UpperGermanandWestCentralGerman,haveundergonethislenition.
Therefore,evenifatanearlierstageofthelanguageapocopehadcausedthelossofdevoicing,wecannotnowrecovertheevidenceforthis,sincethefinalvoicedobstruentsthatwouldhavesurfacedfinallyowingtoapocopearenowindistinguishablefromoriginalvoicelessobstruents.
Hence,wearecompelledtolookinonlyafewrestrictedareaswhereapocopehasrecentlyoccurred,butwheretheoriginalvoicingcontrasthasalsobeenmaintained.
Onepromisingregionisintheeast,namelythenorthwesternborderofSilesia,whereonKnig'smapwecanseethatapocopeisadvancingsouthfromtheLowGerman-speaking42Infact,themapalsoindicatesdiscrepanciesbetweenapocopeasitappliesinuninflectedwords,likethepredicativeformoftheadjectivemüd(e)'tired',andinflectedwordslikethedativesingularHaus(e)'house'.
Apocopeinsomeareashasaffectedthelatter,butnottheformer.
Thismaybeevidenceofmorphologizationoftheapocoperule,butforthepresentargumentthediscrepancyisunimportant:thedialectswewillbeexamininghaveallundergoneregular,phonologicalapocope.
98region43.
Andindeed,asurveyofalocaldialect(Messow1965:15-20)revealsjustwhatweexpect:apocopehascausedthelexicalizationofthedevoicingrule.
Sinceapocopehassorecentlyoccurredinthisdialect(fromthetownofZüllichauandthesurroundingregion),thedevoicingrulehasnotbeenlostcompletely.
Thus,whilethesingularof'day',ModernGermanTa[k],whichnevercontainedafinalschwa,is[tak],thewordfor'eye',whichdidhaveschwa(cf.
ModernGermanAu[g]e),is[og].
Thecorrelationbetweenapocopeandexceptionstodevoicingisverystrong,althoughtherearesomeanomalies,e.
g.
thewordfor'mountains'is[bark],cf.
ModernGermanBer[g]e.
Thelattermightbecausedbyhaphazardextensionofthelexicaldevoicingruleonaword-by-wordbasis(seebelowforthediscussionofasimilarcaseinFinnish),orelseplainlexicalanalogy.
Thatsomekindofdevoicingruleisstillactiveinthephonologyisshownbyexampleslikethefollowing:singular'goose',ModernGermanGan[s],is[gns],buttheplural,ModernGermanGn[z]ewithfinalvowelandumlaut,is[gnz]44.
Thiscannotbeanalyzedasapluralvoicingrule,becausewealsohave[tk]'stick',plural[tk],cf.
ModernGermanSto[k],St[k]e.
Opacityhascausedthere-analysisofthedevoicingrule43AlthoughLowGermanisusuallyconsideredaseparatelanguage,thereisinfactmutualintelligibilitybetweenthesoutherndialectsofLowGermanandthenortherndialectsofcentralHighGerman,sothatinnovationsconstantlyspreadbackandforthovertheborderbetweenthetwo.
44Historically,itwasthe[s]thatwastheoriginalProto-Germanicsound,with[z]derivingfrommediallenitionoffricativesintheMiddleGermanperiod,i.
e.
thesoundchangethatgaverisetothealternationinfricativeswasindependentofMiddleGermandevoicing,whichatfirstappliedonlytostops(OldGermanfricativeswereallvoiceless).
Afterbothchangesoccurred,however,thestopdevoicingruleandthefricativevoicingrulewerere-analyzedasaunifiedfinalobstruentdevoicingrule,sincethesurfacealternationpatternsforstopsandfricativeswereidentical:medialvoicedalternatingwithfinalvoiceless.
99asalexicalphonologicalrule,applyingonlyincertainmorphologicalcontexts(suchascertainnounsingulars).
Itisarguablethatwearemerelylookingatarestructuredlexicon,withtheolderdevoicingrulenowfossilizedincertainwords.
However,otherpiecesofevidence,suchastheexampleof'mountains'above,wherethedevoicingrulewasexceptionallyappliedtooneoftheoutputsofapocope,suggestthattherulemusthavebeenproductivetoatleastasmalldegree.
Nevertheless,wepredictthatsuchasituationisinherentlyunstable,andthatsuchalexicaldevoicingrulewouldsoonerorlaterbelost,leadingtothelevelingofalternations,asinYiddish.
Inconclusion,thisSilesianexampleindirectlysupportsouranalysisofEarlyModernGerman.
45Otherdialectsurveyscorroboratethisanalysis.
Inthesamevolume(no.
56)ofDeutscheDialektgeographie(Tita1965,Kuck/Wiesinger1965),therearesurveysofdialectsfromPomeranianLowGerman(Bublitz)andEastPrussianHighGerman(Ermlandish)thatshowpreciselythesamekindofpattern:recentlyoccurredapocopehasresultedinthelexicalization,butnotcompleteloss,ofthefinaldevoicingrule.
TheLowGermanstudyisespeciallyinstructive,sincetheauthorofthesurveyprovidesmorphologicalparadigmsshowingwherethedevoicingruleisstillproductive,e.
g.
instrongverbs,like[rv],'todig'cf.
ModernGermangra[b]en,wherethefirstperson45ItalsounderminesthehypothesisofSapirandSadockconcerningthelossofdevoicinginYiddish,accordingtowhichlevelingprecededapocope.
Thetrendisclearlyforapocopetorenderdevoicingopaque,withthepossibilityofthecompletelossofdevoicingatalaterstage.
100singularpresentindicative,ModernGerman(ich)gra[b]e,is[rv],butthepast,ModernGerman(ich)gru[p],is[rɑf]46.
Notallsuchsurveyssupportourtheory,however.
AnolderstudyofSilesianfoundthatdevoicingappliedtotheoutputsofapocopeinanotherareathatunderwentit(vonUnwerth1908:44).
Similarpatternscanbefoundine.
g.
Luxemburgish,a(semi-standardized)MiddleFranconiandialectthathasretainedthevoicingcontrast,undergoneregularphonologicalapocope,andyetwhichshowsfinaldevoicingalsoinformswhereitisnotexpectedaccordingtoourhypothesis(Keller1961)47.
Giventhatthesevariousdialectshavebroadlysimilarlexiconsandgrammars,itishardtoexplainwhyapocopewouldhavesuchoppositeeffectsondevoicing.
3.
5ApocopeandlossofdevoicinginEarlyModernDutchOldDutch(pre-1200)hadaruleoffinaldevoicing,asthespellingsattest(Goossens1974:65-66),justasinMiddleGerman48.
ThefactthatdevoicingappearedinDutchataroundthesametimeasinGermansuggeststhetwophenomenaarerelated,althoughit46Thefinalschwaoftheinfinitivederivesfromoriginal-en,asintheModernGermaninfinitive.
Notethatthenon-periphrasticpreteritwasretainedinLowGerman,whileitwaslostinUpperGermanandYiddish.
Conceivably,thiswouldhavehelpedalexicaldevoicingruletosurvivelonger.
47AsinYiddish,inLuxemburgishthefinalschwaofinflectedattributiveadjectiveswaspreservedfromapocope.
Otherwise,apocopewasquiteregular,andoughttohaveintroducedenoughopacityindevoicingtohinderproductivity.
TheremayevenbesomeinternalevidencethatLuxemburgishlostfinaldevoicingcompletelyafterapocopeoccurred.
Thus,historicalfinal//wasdevoiced,e.
g.
ewe[]'away',butinalternatingWe[],We[]en'path(s)'thefinalfricativehasdisappearedcompletely(Bruch1953:172).
Althoughthelossofthedorsalfricativeisusuallyfoundintervocalically,thelossinfinalpositioninalternatingformssuggeststhatperhapsitwasapostvocalicchangeonly,butonethatoccurredafterapocopehadrenderedfinaldevoicingunproductive.
However,thiscouldalsobeinterpretedasarestructuringoftheURof'path'as/ve/,followingthelossof//intheplural/ven/>/ven/.
48SinceDutchisnotattestedasearlyasGerman,the"Old"periodofDutchcorrespondsintimetoearlyMiddleGerman,while"MiddleDutch"correspondstolateMiddleGermanandearlyEarlyModernGerman.
101ispossiblethatDutchdevoicingaroseindependently.
Forourpurposes,itisenoughtoknowthatDutchinitsearliestperiodhadarobustlyattesteddevoicingrule.
Thus,fan[t]'found',butfin[d]an'tofind';utgan[k]'exit',butgan[g]a'passages';ga[f]'gave',butge[v]an'togive'49.
ThisrulewasreinforcedintheMiddleDutchperiod(post-1200)bythevoicingoffricativesinmedialposition,e.
g.
hui[s]'house',buthui[z]en'houses'50.
InearlymodernDutchfinalschwawasapocopated,asinUpperGerman(Goossens1974:55).
ThisprocessmayhaveoriginatedseparatelyfromtheUpperGermanapocope,sincesomeapocopeisattestedalreadyintheMiddleDutchperiod(vanLoey1966),atthesametimethatsomerestrictedapocopewasalreadyunderwayinUpperGerman(Keller1978:273-4).
Dutchapocope,likethatofUpperandWestCentralGerman,wasaregularsoundchange,althoughtheeffectsofthechangewereobscuredbylatermorphologicallevelingsandre-analyses51.
AswithEarlyModernGerman,schwaapocopecanbeshowntohaveinteractedopaquelywithfinaldevoicinginEarlyModernDutch.
Moreover,weevenhavesomeevidencethatvoicingwasleveledincertainparadigms.
Thus,aroundthetimeapocopeoccursintheearlymodernperiodoneseesspellingsreflecting,forexample,acontrastbetweenfinal[]andfinal[χ],e.
g.
icklagh'Ilay'(fromliggen'tolie'),buticklach'I49DutchandLowGermanhadvoicedfricativereflexesofProto-Germanic/b/innon-initialposition,whereasHighGermanhadvoicedstops.
50GoossensnotesthattheOldandMiddleDutchspellingdidnotalwaysreflectthedevoicingrule.
Heassumesthatthiswasmerelyamatteroforthographicirregularity,thoughitcouldalsomeanthatthedevoicingrulewassubjecttovariation.
51ApocopeinDutchshowsmoreexceptionsthanYiddishandLuxemburgish,thoughfewerthaninEastCentralGerman.
Deadjectivalslikebreedte'breadth',theweakpreteritin–te,andinflectedattributiveadjectivesareallexceptions(vanHaeringen1937).
Russ1982arguedthatthisevidenceshowsapocopewasagrammaticallyconditionedsoundchange.
RecentlyMondon2009hasarguedforadifferentanalysisthatbreaksdownunconditionedapocopeintodifferentphonologicallyconditionedchanges,whoseresultswereaffectedbyanalogicalleveling.
Bethatasitmay,thereisenoughevidencethatapocopewasaphonologicallyconditionedprocessthatwecansafelycallitasoundchange,andthereforeassumethatitwasaddedtotheendofthegrammar,afterdevoicing.
102laugh',cf.
MiddleDutchicklach,icklache(vanBree2003:263ff.
andreferences).
Again,vanBreereportsthatcertainearlymoderngrammariansreportedfinalvoicedpronunciationinwordslikewe[]'path'(vanBree2003:265),whichwashistoricallyschwa-less.
Voicinginfinalpositioninthiscasecanonlyhavecomeaboutthroughleveling,notthroughapocope.
Earlymodernorthoepistsconfirmthisinterpretationofthespellingpatterns.
Itistruethat,likeGerman,Dutchspellingbecamenormalizedinthe16thand17thcenturies,sothatonecouldconceivablyattributethelevelingofvoicingalternationstotheadoptionof"morphemic"spelling,alongthelinesofHermannPaul'sargumentsforthedevelopmentofGermanspelling.
However,Dutchspellingisnotentirelymorphemiceventoday,andtherepresentationoffinaldevoicinghasbeenmaintainedforsomealternations,namelythoseinvolvingunderlying/v/and/z/:huis~huizen,ikgeef~geven.
ThisshowsatleastthatPaul'stheoryofthedevelopmentofwriting,thatphonemicsystemsinvariablyturnintomorphemicovertime,isnotwatertight.
IfwewishtoadoptPaul'stheoryofthedevelopmentofwriting,wewouldhavetoaccountforwhyGermanspellingbecame"purely"morphemic,whileDutchspellingremainspartlyphonemictothisday.
3.
6EastCentralGermanEastCentralGermanapocopedidoccur,butwithmorerestrictivephonologicalandmorphologicalconditioningthanasfoundinUpperandWestCentralGerman.
AccordingtoKeller1978,apocopeactuallybeganinthesamewayitbeganintherestofthe103German-speakingregion:asanunconditioneddeletionruletargetingallfinalschwa.
However,whereasinotherregionsapocopewenttocompletion,withtheexceptionofinflectedadjectives,inEastCentralGermanapocopestoppedprogressingatamuchearlierstage,andthenwasreversedinseveralcategories.
Eventually,finalschwawasevenaddedtocategorieswhereithadnotexistedinMiddleGerman.
FromtheexampleofYiddish,wecouldpredictthatwhereverunconditionalapocopeoccurs,itshouldinteractopaquelywithdevoicing,andthatthisopacityshouldleadtotheeventuallossofthedevoicingrule,becausethenumberofexceptionsintroducedbyopacitywouldhindertherule'sproductivity.
WhatdowepredictshouldhappenwiththemorerestrictedformofapocopethatoccurredinEastCentralGermanApocoperenderedfinaldevoicingopaquebecauseitwasaddedtothederivationalseriesafterthedevoicingrule.
Thishappenedbecauseapocopewasaregularsoundchange,whichwerepresentbytheadditionofaphonologicalrule.
However,lossoffinalschwamighttakeplacewithoutthehelpofaphonologicalruledeletingfinalschwa:-becausearuleofinflectionalorderivationalmorphologyaddingfinalschwafailstoapply:e.
g.
thedativesingularisformedbyaddingschwatothestem,Tag–>Tage;ifthisruleislost,TagebecomesTagwithoutthehelpofaphonologicaldeletionrule-alternatively,wheretheschwaispartofthestem,restructuringofthestemthroughlexicalanalogycancausetheschwatobelostwithoutphonologicaldeletion:e.
g.
ja-stemclassadjectiveschrgebecomea-stemclassschrg104IfitturnsoutthatEastCentralGermanapocopeinvolvedmorphologicalandlexicalchangesliketheabove,thenwewouldpredictthatsuchchangeswouldnotrenderdevoicingopaqueandliabletoloss.
YetasKellernotes,apocope,atleastinitsearlystages,wasclearlyphonological,eveninEastCentralGerman,andaswesaw,therewasonetextfromUpperSaxonythatshowedevidenceofopaquedevoicingasaresultofapocope.
Inaddition,thefinaloutcomeofapocopeinEastCentralGerman,althoughlesssweepingthaninUpperGerman,neverthelessshowedunambiguousphonologicalconditioning.
3.
6.
1EastCentralGermanapocopeAfterthebackandforthbetweenapocopeandrestorationofschwahadsettleddown,Germanendedupwithpermanentapocopeinthefollowingcategories(Wright1907:79ff.
):-Inthesyllableprecedingorfollowingthesecondaryaccent(Behaghel'sLaw):boúmgàrte>boumgart,állerhànde>allerhand,átemùnge>atmung-Inadjectivesandnouns,finalewaslostindisyllablesiftheprecedingconsonantwasavoicelessobstruentorasonorantorvowel:spte>spt,küele>kühl,niuwe>neu,buttrge,weise5252Therestrictiontoadjectivesandnounsappearstobeduetothefactthatfinalschwawasanalogicallyrestoredinallverbformswherethisapocopeoccurred,e.
g.
intheimperatives.
105Wrightfurtherreportsthatfinalschwainthedativesingularofnounsandadjectivesfluctuateswithzeroindisyllabicforms,dependingon"sentencerhythm",buthenotesthatinthemodernspokenlanguageitisusuallylost(andbythelate20thcenturyinthewrittenlanguage,too).
Thus,eitherTageorTagispermissibleasthedativesingularofTag,thoughonlythelatteristypicallyfoundtoday53.
3.
6.
1.
1Behaghel'sLawWithrespecttoBehaghel'sLaw,theconditioningispurelyphonological,andhencewemaylabelthisaregularsoundchangewithoutdifficulty.
However,therequirementfortheeliminatedvoweltoeitherprecedeorfollowthesyllableofsecondarystressseemsalittlehardtocaptureinanon-disjunctivestatement.
Also,accordingtoWright'sdescriptionthelawtargetedbothword-final(apocope)andword-internalschwa(syncope),e.
g.
atemunge>atmung.
Othergrammars,e.
g.
Reichmann&Wegera1993,donotgroupsyncopeandapocopetogether,suggestingthatperhapsWright'sdescriptioncouldbedisputedonthegroundsofaccuracy.
Keller1978andReichmann&Wegera1993alsoarguethatapocopetargetedfinalschwaafteraderivationalsuffix,e.
g.
nounsin–unge,-nisse,oradjectivesin–liche,becoming–ung,-nis,-lich,respectively.
Tosaythatapocopewasblockedincertainmorphologicalenvironmentscontradictsthetraditionalassumptionofexceptionlesssoundchange,sothisdoesnotseemasignificantimprovementonWright'sdescription.
53"Sentencerhythm"suggestssomekindofphonologicalconditioning.
Ontheotherhand,therestrictiontothedativesingular,andnotthephonologicallyidenticalnominative/accusativepluralTage,suggeststhisisananalogical,morphologicallyconditionedchange.
Asensiblesolutionistosaywearedealingwithmorphologicalvariants,whiletheratioofvariantsisaffectedbyphonologicalorphoneticfactors(cf.
Kiparsky1989onconstraintsinphonologicalvariation).
106Also,thisanalysisignorescaseswherethesecondcompoundelementwasprobablysemanticallyopaque,e.
g.
herzoge>Herzog'duke',inwhichthesecondelementzog(e)nolongeroccursonitsown(theoriginalmeaningwas'leader',herzogebeing'armyleader',OldHighGermanheri-zogo).
Thetwodescriptionscanbereconciledifwerevisethedescriptioninthefollowingway:ratherthansayschwaisdeletedbeforeorafterthesecondarystress,wesayinsteadthatschwawasdeletedafteranystressedsyllable,whetherprimaryorsecondary.
Ofcourse,thereareexampleslikeZunge'tongue',whereapocopeofschwaappearstohavebeenblocked.
However,sincetherootsyllablewasoriginally/tsng-/,withavoicedobstruent/g/,thepresenceofschwainthiswordcanbeaccountedforbythegeneralrulethatapocopewasblocked(orfinalschwarestored)aftervoicedobstruents54.
Onecouldarguethatsince/ng/becomesthevelarnasal//inModernGerman,thatapocopeshouldnotinfactbeblocked,butthenonewouldhavetoarguethatassimilationof/ng/hadalreadyoccurredbeforethetimeofapocope.
Giventhatwestillfindclearevidenceofsurface[k]fromunderlying/ng/inEarlyModernGerman,e.
g.
junckfrau'virgin',from/jung/,thiswouldbeahardcasetoargue.
Butevengrantingthatassimilationhadalreadytakenplace,wecouldarguethatschwawasrestoredbecausezungewasafeminineweaknoun,andWrightobservesthatfeminineweaknounsingeneralpreserveschwa.
Sozunge>zungcouldhaveoccurred,butlaterfinalschwawasrestoredbymorphologicalanalogy.
54Thereisanotherpossibility,namelythat/ng/hadalreadyassimilatedmediallytothevelarnasal,asinModernGerman.
Ifso,'tongue'shouldhaveendedinasonorant,whichdoesnotblockapocope.
107Theredonotappeartobeanycounter-examples,e.
g.
awordlike*fór-e-de-rùng-e>for-e-drung-e,wherethedeletedvowelwasnottheonefollowingtheprimarystress,butpreciselytheoneprecedingthesecondarystress.
Rather,theexamplesWrightoffersforBehaghel'sLawcangenerallybeaccountedforbypositingdeletionofschwaafteranystressedsyllable.
CasescitedbyWrightwherethisdoesnotoccurcanbeattributedtolatervowelepenthesis,analogyorborrowing:(32)vorderunge>forderung'claim'küneginne>knigin'queen'krbeze>krebs'cancer'fremide>fremd'foreign'hemidi(OldHighGerman)>hemd'dress'kebisa(OldGerman)>kebse'concubine'muniza(OldGerman)>münze'coin'lbete>lebte'helived',andotherformsoftheweakpastin-teThesecondvowelinforderungisprobablyduetovowelepenthesis,fromfordrung,orperhapsbecauseoftheinfinitivefordern.
The–iginknigandderiveknigin'queen'hadsecondarystressandthusdidnotundergosyncope.
Whilefremide>fremde,orhemidi>(hemede)>hemdeiseasytoexplain(withasecondarystressonthefinalsyllable),thesubsequentdeletionofthefinalvowelishardtoexplainusingWright'sdescription;thefinalvoiced/d/shouldhavepreventedapocope.
Butthiscouldbeduetolexicalanalogy,alongthelinesofschrge–>schrg(seebelow).
108Likewise,whydowehavekrebsbutkebseOneofthetwomustbeadialectborrowingoranalogicalremodeling,sinceyoucannotgetbothformsoutofasingleregularsoundchange.
Ifkebse,asislikely,istheborrowing(seeKluge1975),thenwestillneedtosayhowwegetkrebsfromkrebeze,sincedeletionshouldonlyhavetargetedthemiddlevowel,krebse.
However,Kluge1975showsthatthereweretwovariantsinMiddleGerman,krebezandkrbeze(withdifferentrootvowels),somodernkrebsisprobablyderivedfromtheformervariant.
Inthatcase,thelossofthevowelcouldbeduetothesamesyncopethatcreatedvariantsliketagsfortages(seeKeller1978:406).
Aconsequenceofthisapproachisthatthereisnospeciallawtargetingschwaintheenvironmentofsecondary,asopposedtoprimarystress,butratherasingle,unifiedschwadeletionruleconditionedbyprecedingstress.
ItseemsthemainreasontotreatBehaghel'sLawasaseparateprocessisasfollows:wheretheschwaoccurredafterthesecondarilystressedsyllable,the[voice]specificationoftheprecedingsegmentwasirrelevant,soallerhande>allerhand,herzoge>herzog.
Whereapocopeoccurredaftertheprimarilystressedsyllable,ontheotherhand,apocopewasblockediftheprecedingsegmentwas[+voice,+obstruent],sospte>spt,buttrge>trge.
Thereisalsothefactthatschwaislesspronetorestorationafterderivationalsuffixesthanafterstems.
Therefore,thereissomeevidencethatapocopeaftersecondarystressistreatedsomewhatdifferentlyfromapocopeafterprimarystress55.
3.
6.
1.
2Blockingofapocopebyprecedingvoicedobstruent55Cf.
aboveonMiddleGermanschwadeletion,whichoccurredafterprimarystressonlyinlimitedconditions.
109AfterBehaghel'sLaw,apocopeseemstohavebeengeneralizedtoallword-finalschwa,exceptwheretheprecedingconsonantwasavoicedobstruent(providedweautomaticallyattributethelexicalexceptionstoanalogy).
Althoughthesoundchangecanbestatedintermsofphonologicalconditioningonly,thestatementisdisjunctive:schwawaslostaftervoicelessobstruentsandsonorantsandvowels,butnotvoicedobstruents,i.
e.
theprecedingsegmentmusteitherhavethefeaturespecifications[–obstruent]or[+obstruent,–voice].
Thisconditioningismoststrikinglyattestedintheclassofformerneutercollectivessignifiedbytheprefixge-,allofwhichendedin–einMiddleGerman(beingja-stemsoriginally):(33)Gebude,Gebinde,Gebirge,Gedrnge,Gefilde,Gefolge,Gelnde,Geleise,Gehnge,Gemüse,Geprge,Gesinde,Getreide,Gewebe,Gewerbe,Gewlbebut,(34)Gebiet,Gebüsch,Gedicht,Gefhrt,Gefecht,Gehirn,Gehft,Gehlz,Geleit,Gelüst,Gepck,Gert,Gerusch,Gerüst,Geschft,Geschenk,Geschirr,Geschlecht,Geschrei,Gesetz,Gespann,Gespenst,Gestirn,Gewchs,Gewicht,GlückOnecouldarguethatsonorantconsonantsandvowelsareredundantlyvoicedinGerman,sothatwearelookingatsomekindofunderspecificationeffect.
Ifweuseprivative110featuretheory,wecansaytheapocoperuleappliesaftersegmentsnotspecifiedfor[voice];the[voice]featureisthenaddedto[–obstruent]segmentsredundantlyatalaterstage,e.
g.
kühl.
Underthisapproach,theblockingeffectisillusory,sincereallywearedealingwithaphonologicalconditioningenvironmentattheunderlyinglevel.
However,suchanapproachbegsthequestionofwhethersoundchangeshouldbeabletooccuranywhereelsethanatthesurfacelevelofrepresentation(seediscussionbelowonruleinsertion)56.
Thefactthatapocopeseemstobeavoidedwhereitmightprovideinputtofinaldevoicingsuggeststhatthetwoareperhapsinvolvedinaphonologicalconspiracy.
Therearetwowayswecantrytoexplainthisapparentconspiracy.
Onewayistostipulatetheblockingeffectinthedescriptionoftherule,andsaythatapocopewasblockediftheschwawasprecededbyavoicedobstruent.
Theproblemwiththissolutionisthattheblockingeffectremainsunexplained.
Anotherwayistoadoptteleologyandarguethatapocopedidnotoccuraftervoicedobstruentsinorderthattheobstruentsshouldnotbesubjecttodevoicing(thisisWright'sapproach).
Yetteleologyasamotivatingfactorforsoundchangeishighlycontroversialinhistoricallinguistics,andanon-teleologicalapproach,ifavailable,shouldbepreferred.
Theteleologicalproblemcanbecircumvented,however,byadoptinganoutput-orientedtheoreticalframeworklikeOptimalityTheory.
WesimplyrankthemarkednessconstraintFINALDEVOICING(whichtargetsobstruentsonly)abovethefaithfulnessconstraintIDENT-IO[voice],sothatwhenapocopeoccurs,bypromoting*FinalSchwa56Notethatunderthisanalysis,finaldevoicingwouldhavetobleedredundantvoicing:the/g/in/tag/,withspecified[voice],wouldlosethatfeaturefirstinthederivation.
Thenredundantfeature-fillingwouldnotapplybecausethe/g/isspecified[obstruent](ortheappropriatenodeinthefeaturehierarchy),andredundantvoicingdoesnotapplytosegmentsspecifiedas[obstruent].
111aboveMAX,itisautomaticallyblockedwherethedevoicingconstraintwouldbeviolated:(35)/gebiete/FINDEVOI[obst]IDENT-IO[vce]*FINALSCHWAMAXGebiete*!
Gebiet*Gebiede*!
*Gebied*!
**/gebinde/FINDEVOIIDENT-IO[vce]*FINALSCHWAMAXGebinde*Gebind*!
*Gebinte*!
*Gebint*!
*Thefactthatvoicedobstruentsdidnotblockapocopeinsomecases,likeo[b]e>o[p],orallerhan[d]e>allerhan[t],canbedealtwithbyrefiningtheconstraintsusedforenforcingapocope.
Forexample,insteadofasingle*FinalSchwaconstraint,wehaveoneconstraintthatonlyaffectsschwaafterunstressedsyllables,whichisrankedhigherthan112Ident-IO[voice],whileanotherconstrainteliminatesschwaafterstressedsyllables,whichisrankedlowerasintheabovetables.
(36)/obe/FinDevoi*FinSchwa(NoStress)Ident-IO[vce]MAXo[p]**o[p]e*!
*ob*!
*obe*!
Dealingwithclearanalogicalremodelingslikeschrgistrickier,buttheabilityofOTtohandletheregularphonologicalpatternsisclearenough.
Simplyfine-tuningthedescriptionoftheconstraintscancapturethemorefine-grainedconditionsonMiddleGermanapocope(seeabove).
However,withrespecttodialectswhereapocopeledtotheeventualloss,ratherthanrestorationofdevoicing,wesawearlierthatOTstrugglestoaccountfortheopacityofdevoicing.
AlthoughinthiscaseOTappearstoprovideanelegantdescriptionofthechange,inlightofthedifficultieswithYiddishopacity,wewillattempttofindanotherexplanation.
Anon-teleologicalexplanationfortheblockingeffectthatalsoavoidsthepitfallsofOTisasfollows:apocopeaffectedallfinalschwas,andtheresultwasthatinsomecases,voicedobstruentsnowsurfaceinword-finalposition,e.
g.
trge>trg.
Later,113schwawasinsertedaftertheresultingword-finalvoicedobstruents,thustrg>moderntrge.
Insomeisolatedcases,devoicingappliedtothenewfinalvoicedobstruents,e.
g.
schrg>schr[k],withtheresultthatschwawasnotinserted,thusmodernschrg,withfinal[k].
Thisrequiresustosupposethatthedevoicingruledidbecomeopaqueasaresultofapocope.
However,italsorequiresustosupposethatdevoicingre-appliedtoatleastsomeoftheoutputsofapocope.
3.
6.
2MorphologicalfactorsinEastCentralGermanapocopeTheroleofanalogyinthesubsequentdistributionoffinalschwaisalsoclearfromWright'sdescription.
Inseveralcategories,finalschwawasactuallyaddedbyvariousanalogicalprocesses,whereitwasnotfoundinMiddleGermane.
g.
thenominative/accusativepluralofneutera-stemnouns,wort'words'→worte.
Moreover,inthecategorieswhereschwawaslostregularly,accordingtotheaboveconditions,analogyappearstohaverestoredfinalschwainsomecases:e.
g.
amongtheimperativeformsofverbs,WrightnotesthatformswhichlostschwaaccordingtothelateMiddleGermanapocope,e.
g.
vare>var(imperative),laterregainedschwaontheanalogyofformsthatneverlostit,thusmodernfahre.
Insomecases,apocopeoccurswhereitshouldnothaveoccurredbyregularsoundchange,e.
g.
schrge–>schrg(cf.
numberTwoabove),ordativesingulartage–>tag:thesecanonlybeaccountedforbysomekindofanalogy,orperhapsdialectborrowing.
1143.
6.
3FinaldevoicinginEarlyModernEastCentralGermanAlthoughtheevidenceofEarlyModernUpperGermanstronglysupportsthehypothesisthatapocoperendereddevoicingopaque,causingitseventualloss,aswehaveseenthedevelopmentsinEastCentralGerman,whichhasbeenthemaininfluenceonliteraryGermanfromthe16thtothe19thcenturies,areratherdifferent.
Apocopedidnottargetallfinalschwa,asitdidinUpperandWestCentralGerman,butwasevenblockedwheretheprecedingsegmentwasavoicedobstruent,i.
e.
apotentialinputfordevoicing.
IstheconditioningofEastCentralGermanapocopedifferentinsuchawaythatwewouldnotexpectdevoicingtobecomeopaqueNotquite,sincewehaveseenevidenceofunambiguouslyphonologicalconditioningine.
g.
Behaghel'sLaw.
Moreover,Behaghel'sLawshouldhaveresultedinnewfinalvoicedobstruentsinatleastsomecases,sowehavetodealwiththepossibilitythatdevoicingbecameopaque.
Finally,ourtentativeexplanationfortheretentionoffinalschwaindisyllableaftervoicedobstruents,e.
g.
intrge,requiresthatapocopehadresultedinfinalvoicedobstruents,i.
e.
opaquedevoicing.
Ontheotherhand,thereisevidencethatfinaldevoicingwasrestoredfairlyearlyinEastCentralGerman.
SourcesincludethegrammarofJohannesClajus(GrammaticaGermanicaeLinguae,1578),whoexplainedthespellingconventionsofcontemporaryGerman,inwhichfinaldevoicingwasnolongerrepresented,byaprinciplethatdoescorrespondcloselytothe"morphemic"principleencounteredearlier,theso-calledExplikationsregel("explanationrule").
Thisruleexplainedthate.
gKalb'calf',althoughpronounced[kalp],mustbewrittenwitha,becausethepluralKelberispronounced115with[b].
ThisindicatesthatClajus'audience,andpossiblyClajushimself,stillemployedfinaldevoicing,althoughthenormativepronunciationquitepossiblydidnot.
ThissuggeststhatforthenewEastCentralGermanstandard,thespellingreallydidreflectashifttoamorphemicsystemfromtheMiddleGermanphonemicsystem.
However,forUpperGerman,thedialectoftheEastFranconianIckelsamer,webelievethespellingreflectedtheiractualpronunciation,i.
e.
adialectthathadlostdevoicing.
TheuseofthesamespellinginEastCentralGermanisdue,inthatcase,tothepurelyliteraryinfluenceoftheoldUpperGermannormonthenewEastCentralGermannorm.
Inotherwords,adialectthathadneverlost,orhadrecentlyrestoredfinaldevoicing,didindeedadoptaspellingsystemthatreflectedadifferentphonologicalsystem.
Thiswasnotduetoaprincipledchangeinorthography,asPaulheld,buttotheconvergenceofliterarystandards.
Sowehavetoconsiderhowdevoicingmayhavere-enteredthelanguagequitesoonafterapocopeended.
Wealsohavetoconsiderthereasonsthatapocopedidnotgotocompletion,andendedupwiththerestrictedconditioningthatithasinModernGerman,andwhatrelationshipthismighthavewithdevoicing.
Finally,wehavetoconsidertheeffectsthattheshiftingcentersofprestigiousHighGermanhadonthepronunciationoffinaldevoicing.
3.
6.
3.
1LenitioninEastCentralGermanWhatabouttheevidencethatEastCentralGermanwasinfactalenitingdialectTheoldMiddleGermannorm,totheextentitexisted,maintainedavoicingcontrast.
Butthereis116evidencethatlenitionbegantoaffectEastCentralGermanalreadyinthe14thcentury(Mihm2004:175),andwecanpresumethatlenitionalsoaffectedotherUpperGermanareasatthistime;thecurrentareaoflenitionincludesthevastmajorityoftheUpperGermanregion.
However,sincelenitionwasstillaninnovationinthelateMiddleGerman/earlyEarlyModernGermanperiod,itisnotsurprisingthatthenormativepronunciationwouldholdtoamoreconservativestandardthatmaintainedthecontrast.
TheevidenceofJohannesClajus'showsthatnormativeEastCentralGermaninthe16thcenturyhadfinaldevoicing,i.
e.
ithadnotyetundergonelenition.
ButbythetimeofGoethe,lenitionhadaffectedevennormativepronunciation.
3.
6.
4OpacityandlossofdevoicinginEastCentralGermanTheevidencefromtheBonnCorpusshowedthatapocopemostlikelyhadthesameeffectonEarlyModernGermandevoicing,oratleastUpperGermandevoicing,asitdidinYiddish:theadditionoftheapocoperulecauseddevoicingtobecomeopaque,resultinginthepatternswesee,wheredevoicingappliesatasignificantlyhigherfrequencyinwordsthathistoricallyhadnofinalschwa.
Thisrepresentsthemorphologizationofthedevoicingrule.
Subsequently,thelevelingoffinalvoicingintothoseformsthatoriginallyhaddevoicingtookplace,representingtheactuallossofthedevoicingrule,whichresultedinthesurfacingofunderlying[voice]infinalposition.
ButwestillhavetoexplainhowandwhyModernGerman(andModernDutch)hasdevoicingtoday,despitetheevidencethatdevoicinghadalreadybeenlostintheEarlyModernperiod.
Thereareseveralpossibleanswers,allofwhichhingeonthefact117thatModernGermandoesnotdescenddirectlyfromtheliteraryMiddleGermandialect,whichwaslargelyUpperGerman,butfromEastCentralGermanandtheHighGermanofthenorthern,formerlyLowGermancities(Keller1978).
ThefirstpossibleanswerdependsonthedoubtfulnessofthewrittenevidenceforEastCentralGerman,whereapocopedidnotgotocompletionandwherewehaveevidencethatdevoicingstillexistedinthelanguageatleastbytheendofthe16thcentury.
ThissuggeststhepossibilitythatdevoicingremainedtransparentandproductivethroughouttheperiodofapocopeinEastCentralEarlyModernGerman,thoughnotinYiddishorUpperGerman.
Ifthatistrue,weneedtodeterminewhetherornotthephonologicalcharacteristicsofeitherfinaldevoicingorapocopeweresuchastoleadustoexpectthiskindofoutcome.
ThesecondistoacceptthatdevoicingbecameopaquetemporarilyinEastCentralGerman,butthatitwasnotlostcompletelyandsomehowremainedproductiveinthegrammar,andthatsomekindofgrammarrestructuringisresponsibleforthe"restoration"oftransparency.
ThatwouldmeanthatthewrittenevidenceisaccuratewithrespecttotheopacityofEarlyModerndevoicinginthe14thand15thcenturies,butinaccuratewithrespecttotheapparentlossofdevoicingandlevelingofalternationsinthe16thand17thcenturies.
Again,weneedtoexaminethephonologicaldescriptionofthedevoicingruleandtheapocoperuleinordertodeterminewhethersuchadevelopmentisplausibleornot.
ThethirdpossibilityisthatdevoicingwasindeedlostcompletelyeveninEastCentralGerman,andthatthetransparentdevoicingruleofModernGerman(andperhapsDutch)hasindependentorigins.
Thereareinfacttwowayswecanaccountforsuch118independentorigins.
OneisthatfinaldevoicingaroseagainasasecondsoundchangewithinGermanandDutch.
TheotheristhatfinaldevoicingenteredGermanandDutchthroughcontactwithlanguagesordialectsthatstillhadfinaldevoicing.
4RetentionoftransparentdevoicinginEastCentralGermanInthissection,wewillassumethatourinterpretationofthewrittenevidenceisnotnecessarilycorrectforEastCentralGerman,andthatthelossofdevoicingintheEarlyModernperiodonlyatteststoachangeinspelling.
Undersuchanassumption,wewouldhavetodiscoveraframeworkinwhichfinaldevoicingwouldbepermittedtoapplytotheoutputsofschwaapocope,despitethefactthatapocopewasaddedtothegrammarlaterintimethandevoicing.
Ifthetheoreticalargumentsarestrongenough,wemaybejustifiedindoubtingtheconclusionswedrewfromtheEarlyModernevidenceinthepreviouschapter.
4.
1Ruleaddition,ruleinsertionandrelativechronologyOnetheoreticalsolutiontoourconundrumistheconceptof"ruleinsertion",wherearuleisaddedsomewherebeforetheendoftheseriesofrules,not,asinYiddish,attheveryend.
Iftheapocoperulehadbeeninsertedbeforethedevoicingruleinthederivationalseries,thenapocopewouldautomaticallyfeeddevoicing,anddevoicingwouldhaveremainedtransparent.
Itonlyremainstobedeterminedwhethereitherthedevoicingruleortheapocoperulewassuchastomakeruleinsertionalikelyscenario.
119Itisworthemphasizingthatthepossibilityofruleinsertionisunavailabletotraditionalmethodsofinternalreconstruction.
Ruleinsertionmeansthatgrammarscanchangeinsuchawaythattheactualchronologyofeventsisobscured,asthefollowingchronologiesshow:(37)MiddleGermangrammarUR/tag//tag/Devoicingtak--(38)EarlyModernGermangrammarUR/tag//tag/Apocope--tagDevoicingtaktakAlthoughwewouldstillholdthatapocopeenteredthegrammarafterdevoicinghadenteredthegrammar,wewouldnotbeabletodeducethisfromthesynchronicorderingofrulesafterinsertionhadoccurred.
Accordingtothemethodsoftraditionalinternalreconstruction,ontheotherhand,wewouldbeforcedtopositthatapocopefirstcreatedexceptionstofinaldevoicing,i.
e.
tage>tag,andthatonlyafterwardswasdevoicing"extended"totheseexceptionsbysomeanalogicalprocess.
Alternatively,wecansupposethatasecondfinaldevoicingoccurred,i.
e.
asoundchange.
Butwecannot120supposethattheapplicationofdevoicingtotheoutputsofapocopewasimmediateandautomatic,whichiswhatruleinsertionwouldpredict.
Wecanillustratethisbyanotherexample,namelytheProto-Indo-EuropeanruleknownasSievers'Law(Ringe2006:16,118-122).
Thisrulevocalizedahighfrontglide,orpossiblyinsertedahighfrontvowelbeforetheglide,iftheglidewasprecededbyaheavysyllable:(39)Sievers'Law:VV.
CjV>VV.
Ci.
jVVC.
CjV>VC.
Ci.
jVThus,examplesusingthesuffix/-jos/:(40)*/pedjos/57'offeet;onfoot',Greek/pesdos/=[pezdos]'onfoot',where/dj/>[dz]>[zd]=/sd/,but(41)*/neptijos/'ofgrandsons',Greek/anepsios/'cousin',where/t/>/s/before/i/InthetransitionfromProto-Indo-EuropeantoProto-Germanic,Sievers'Lawappearstohavere-appliedtotheoutputsofavowelepenthesisrule,whichhadcreatednewheavysyllables:(42)Proto-Indo-European*wr.
gje.
ti'heworks'57Theasteriskisbeingusedtosymbolizereconstructedforms,notungrammaticalforms.
121Sievers'Law*wr.
gje.
ti(nochange),Avestan/vrzjeti/pre-Proto-Germanic*wur.
gje.
ti(epenthesize/u/beforesonorant)Sievers'Law*wur.
gi.
je.
tiProto-Germanic*wur.
ki.
θi(/ije/>/i/),Gothic/workiθ/Aswesee,inProto-Indo-European,syllabicsonorantsaretreatedasshortvowels,sothat/wrgje-/hadthesyllabicstructureCV.
CjV,i.
e.
itwasnotaninputtoSievers'Law.
However,inpre-Proto-Germanic,a/u/wasepenthesizedinfrontofallsyllabicsonorants,andthesonorantbecamenon-syllabic58:(43)u-EpenthesisR>uRAsweseeinourexample,u-EpenthesiscreatednewheavysyllablesofVC.
CjVstructure(cf.
1).
Ratherthanu-EpenthesissimplycreatingexceptionstoSievers'Law,theProto-Indo-Europeanruleappliedtothesenewinputsthatarose,notfromborrowing,butfrominternalsoundchange.
Ifweapplytraditionalmethodsofhistoricalreconstructiontothiscase,wemightconcludethattheouputsofEpenthesiswereanalogicallyremodeledonotherforms58ThechangelooksquitesimilartowhathappenedinthetransitionfromOldIcelandictoModernIcelandic:vocalic/r/became/r/,written–ur,e.
g.
mar'man'>maur.
Phonologically,theremusthavebeenatleasttwopartstothischange:first,somekindofexcrescenthighvowelmusthaveappearedinthearticulationofthesyllabicsonorant,whichwasthenre-analyzedasanepenthetichighvowel/u/.
Oncetheepentheticvowelfounditswayintothephonologicalrepresentation,thesonorantbecamere-analyzedasnon-syllabic,owingtothesonorityhierarchy(/u/ispreferableto/r/asthesyllablenucleus).
Theselectionof/u/torepresenttheexcrescentvowelwouldhavebeensupportedbythefactthatanunderlying/u/alreadyexistedinProto-Indo-European,inrootsthatparticipatedinablaut,e.
g.
*bhewg-'flee',*bhug-,cf.
Greek/pheug/'Iflee',but/ephugon/'Ifled'.
122wheretheeffectofSievers'Lawwasstillvisible.
E.
g.
hypotheticalpre-Proto-Germanic*[wurgjeti],wasremodeledonthe"analogy"of(non-hypothetical)formsthatstillobeyit,like[sagijeti]'heseeks'(Proto-Germanic[sokiθi]).
Theregularityofthisanalogicalremodeling,however,doesnotfitverywellwiththetraditionalconceptionofanalogyasalexicallygradualprocess.
Itseemsmoreprobablethatwearelookingattheeffectsofsynchronicruleapplication.
4.
1.
1RuleadditionTheusualwayforrulestobeaddedtogrammarsisattheendofthederivationalseries(Halle1962,King1969).
BelowwehavethetworuleadditioneventsthatintroducedSievers'Lawandu-Epenthesis.
Firstinpre-Proto-Indo-European:(44)AddtheSievers'Lawrule[]→[i]/V:C__jVVCC(45)Pre-Proto-Germanic:AddsonorantvowelepenthesisR→uR123Atthispointwehavetoconsiderwheretheruleisadded.
Ifwesupposetheruleisaddedattheendoftheseriesofrules,wewillhavetoassumethefollowingorderingofrulesinpre-Proto-Germanic:(46)Sievers'LawEpenthesisThisruleorderingwouldderivethefollowingsurfacerepresentation(SR)fromtheunderlyingrepresentation(UR)/wrgjeti/:(47)UR/wrgjeti/Sievers'LawwrgjetiEpenthesiswurgjetiAsyoucansee,theresultoftheadditionofEpenthesisisanexceptiontotheSievers'Lawrule.
AlthoughSievers'Lawisstillactiveinthegrammar,itnolongerobtainstransparentlyonthesurface,i.
e.
ithasbecomeopaque.
However,wealsoknowthatsuchexceptionstoSievers'Lawarenotattested.
ThesameproblemconfrontsuswithrespecttoEarlyModernGermanapocopeandthefinaldevoicingrule.
Aswesawabove,apocopeconsistedinoneormoresoundchanges,inadditiontovarioussubsequentanalogicallevelingsandextensions.
Thesoundchangesattheveryleastmustbetreatedinthesamewayaspre-Proto-Germanic124Epenthesis,i.
e.
asadditionofnewrules.
Addingtheapocoperuleattheendofthegrammar,moreover,oughttohaveresultedinthesamekindofopacity:(48)UR/ob//tag/Devoicing–takApocopeob–Inorderfor/ob/toendupasModernGermandevoiced[p]59,wehavetoconsiderthepossibilitythattheapocoperulehadbeeninsertedbeforedevoicing.
4.
1.
2RuleinsertionAsmentionedabove,Halle1962(supportedbyKiparsky1965)believedthatrulesshouldbeaddedattheendoftheseriesofrules.
However,Hallequalifiesthisbysayingthatrulesmaybeaddedattheendoftheserieswithintherelevantcomponentofthegrammar.
AccordingtoHalle,therearethreecomponentsofphonology,startingwiththemorphophonologicalcomponent,inwhichre-adjustmentrulesoperated.
AnexamplerelatedtoProto-Indo-EuropeanandProto-Germanicisablaut.
Thiswasarulewherebythevowelofthelexicalrootwaschangedaccordingtoderivationalorinflectionalclass,e.
g.
Proto-Indo-European/es/'be'→/s+enti/'theyare',but/es+ti/'heis',withthe'+'indicatingmorphologicalboundarybetweenrootandsuffix.
Theinputandoutputofthe59AndalsoYiddishop,wherethefinal[p]showsthatdevoicingmusthaveexceptionallyappliedtothiswordafterapocope,whiletheotheroutputsofapocopeinYiddishdidnotundergodevoicing.
125ablautrulecouldbedescribedinpurelyphonologicalterms,buttheconditioningenvironmentwasnotphonological,butmorphosyntactic.
Thenthereisthephonologicalcomponent.
Unlikethemorphophonologicalcomponent,input,outputandenvironmentcanallbestatedinphonologicalterms,meaningthatthestatementmustbeintermsofbinarydistinctivefeaturesorotherwisediscretecategories.
Thus,bothSievers'LawandGermanfinaldevoicingcanbestatedintermsofbinaryfeaturesonly:intheformer,theconditioningenvironmentincludesthemetricalquantityoftheprecedingsyllableoftheglide;inthelatter,theenvironmentistheedgeoftheprosodicword(seethebeginningofthefirstchapterforaformalstatementofthedevoicingrulewithbinaryfeatures)60.
Finally,thereisthephoneticcomponent.
Phoneticrules,unlikephonologicalrules,donotdealinbinaryfeaturesordiscretecategories,buttypicallyconvertbinarydistinctivefeaturesinto"n-ary"values,i.
e.
theyconvertonevalueofaphonologicaloppositionintothephoneticallygradientvaluesthatweactuallyseeinreal-lifespeech.
Thesegradientrulesmayberule-governed,i.
e.
conditionedbyparticularphonological,phonetic,orevensemanticorpragmaticcontexts,buttheyarenotconsideredphonologicalbecausetheyarenotbinary,andalsobecausetheyarenotcruciallyordered.
Everyphoneticruleshouldapplytoeverypotentialinputwithoutexception(King1973).
Becauseoftheseparationofphonologyintothesethreecomponents,theadditionofnewrulesmaynotnecessarilyreflecttheactualchronologyofchanges.
Thatis,arulethathadenteredthegrammaratanearlierdatemayendupsituatedafteralatelyaddedrule.
Thisisknownasruleinsertion.
Halle'sguidelinessuggestthatruleinsertionshould60Asnotedabove,theprosodicworddoesnotnecessarilycorrespondtothelexicalword,e.
g.
wherewearedealingwithclitics.
However,usuallythetwosharethesameboundaries.
126beconstrainedbythesedistinctcomponents,althoughstrictlyspeakingtheydonothavetobesoconstrained.
Forexample,Kiparsky'scitationofLachmann'sLawinLatininvolvestheinsertionofonephonologicalrulebeforeanother.
AccordingtoKiparsky,Halle'sconstraintsonruleinsertionrepresenttendenciesratherthanabsoluteuniversals61.
CoulditbethatEpenthesiswasaddedbeforeSievers'Law,andthatthisiswhywedonotseeanyexceptionstoSievers'LawInthatcase,wewouldgetthecorrect,transparentsynchronicderivation:(49)UR/wrgjeti/EpenthesiswurgjetiSievers'LawwurgijetiThiswouldbeaplausibleanalysisifSiever'sLawwerephonological,whileEpenthesisweremorphophonological,orelseifSievers'Lawwerephonetic,andEpenthesiswerephonological.
ThatSievers'Lawisphonologicalseemsprettyclear:theinput,outputandenvironmentoftherulecanbeexpressedinpurelyphonologicalterms,sotherulecannot61Inbrief,Lachmann'sLawwasthelengtheningofrootvowelsbeforeunderlyingvoicedobstruentsinformsderivedfromthe"thirdstem"(Aronoff1994).
Thesestemsconsistedoftherootsyllablefollowedby/t/or/s/,e.
g.
āctus'havingbeendone',āctio'actofdoing,action',bothfromagō'Ido'.
Intheseforms,voicingassimilationalwaysdevoicedtheroot-finalobstruentonthesurface.
VoicingassimilationalsooccurredinProto-Indo-European,however,whichiswhyKiparskyattributedLachmann'sLawtoruleinsertionintheprehistoryofLatin.
Althoughthephonologicalelementoftheruleisclear,theclearformulationofaLachmann'sLawruleishinderedbythefactthatsuchlengtheningoccursonlyinoneclassofstems,andnotinotherformsofidenticalphonologicaldescription,andeveninthiscategorythereareseveralstemsmarkednottoundergoLachmann'sLaw,e.
g.
fidō,fssus.
Moreover,rootvowelsthatundergolengtheninginthethirdstemtypicallyundergosomekindoflengtheninginthepastindicativeaswell,e.
g.
agō,ēgī'Idid'.
SeeStrunk1976forwhyLachmann'sLawshouldnotbetreatedasasinglephonologicalphenomenon,butcf.
Jasanoff2004forwhyitshould.
127belabeledmorphophonological.
Moreover,inputandoutputmaybeexpressedindiscretephonologicalcategories,inthiscasethesegmentalslotsCandV.
Differentframeworksmightexpressthecategoriesas[±syllabic],usingmoras,X-slots,orbracketsandedges,butthefactremainsthatthecategoriesbeingmanipulatedbythisrulearediscrete,notgradient.
Epenthesisforitspartalsoseemstobephonological,ratherthanmorphophonologicalorphonetic.
Inputsandoutputsmaybeexpressedintermsofdiscrete,phonologicalcategories:inputis[+sonorant,+consonantal,+syllabic],whileoutputisasequenceof/u/and[+sonorant,+consonantal,–syllabic].
Thereisnoconditioningenvironment,butwecansaythecriterionforphonologicalenvironmentisfulfilledbecausethereisnoexplicitlymorphosyntacticconditioning.
Inconclusion,ifpre-Proto-Germanicu-Epenthesiswereacaseofruleinsertion,thiswouldhavetobedespitethefactthatbothEpenthesisandSievers'Lawweretobefoundinthephonologicalcomponentofthegrammar.
4.
2ApocopeasamorphophonologicalruleForthesamereason,EarlyModernGermanapocopecouldbetreatedasaninstanceofruleinsertion,butwewouldneedtoshowthateitherapocopeordevoicingwereassociatedwithdifferentmodules,asHalle'smodelpredicts.
Butthedescriptionsofbothrulessuggesttheyarephonological.
Thus,weknowthatapocopeisaphonologicalruleinthesensethatasegment,i.
e.
abundleofdistinctivefeatures,isdeletedinagivenenvironment,i.
e.
attheedgeoftheprosodicwordinagivensetofmetricalpositions.
128Finaldevoicingisalsophonological,asshownatthebeginningofchapter1:inputandoutputareexpressibleindiscretecategories.
The"irregular"aspectsofapocope,ontheotherhand,asseeninthevariousrestrictionsandextensionsaccordingtoinflectionalorgrammaticalcategory(seeabove),suggestthepossibilitythatatleastsomepartsofapocopemayhaveinvolvedmorphophonologicalorre-adjustmentrulesofthekindthatHallewouldpredictwouldbeaddedbeforethephonologicalseriesofrules.
Instructiveexamplesincludeschr[k],fromschrge,orgescheit,fromgeschde,i.
e.
apocopeafteravoicedobstruent,whereitshouldhavebeenblocked,assumingwecandescribethesecasesofirregularapocopeastheoutputofamorphophonologicalrule.
Ourtheorycertainlypredictsthatifsuchamorphophonologicalruleisaddedtothegrammar,itshouldfeedpurelyphonologicalruleslikefinaldevoicing.
Recallthatthetraditionalhistoricallinguiststrictlyseparatesregularsoundchangeandanalogy,andmoreparticularly,alwaystreatsanalogicalchangeaschronologicallysubsequenttosoundchange,wherevertheoutcomeofsoundchangeappearstohavebeenobscuredbyanalogy.
Therefore,whenwespeakofadditionofmorphophonologicalapocope,weareimaginingalmosttheinversionofthetraditionalchronologicalsequence62.
Therearetwoproblemswiththenotionofmorphophonologicalapocope.
Thefirstproblemissimplythat,aswesaw,mostofapocopeactuallyappearstobephonologicallyconditioned,asweexpectfromregularsoundchange,e.
g.
apocopeinunstressed62Additionofamorphophonologicalruleisanotherwayofsaying"grammaticallyconditionedsoundchange",aphenomenonexcludedbytheneogrammariantheoryofsoundchange.
However,whenworkingwithingenerativeframeworkswemustadoptgenerativistassumptions,whichdonotincludetheexceptionlessnessofsoundchange.
SeeSihler1977andRuss1982forsomeallegedcasesofgrammaticallyconditionedsoundchange.
129disyllables(obe>ob),andafterthesecondaryaccent(allerhande>allerhand).
Soevenifwecancharacterizepartsofapocopeasmorphophonological,andhenceaccountforsomeinstancesofseconddevoicingbyruleinsertion,wecannotaccountforallthecasesofapocopeinthisway.
Thesecondproblemisthattheanalogicalextensionsofapocopethatwedoseearehardtocharacterizeasmorphophonologicalorre-adjustmentrules.
Takeschrgandmild,forexample.
ThesetwoadjectiveshadfinalschwainMiddleGerman.
Aswesaw,phonologicalapocopeshouldnothaveappliedtothesewords,becausetherootsyllablesendinvoicedobstruents63.
Buttheexceptionalextensionofapocopetothissmallsubsetofja-stemadjectivesisnotcharacterizablebycommonmorphologicalorlexicalfeatures.
Rather,theseappeartobecasesofgenuineword-by-wordlexicalanalogy64.
Anotherexampleisthelossofthefinalschwainthedativesingular,againstitsretentionine.
g.
thenominative/accusativeplural:dativesingularTa[k]vsolderTage.
Ratherthantheoutputofare-adjustmentrule,thelossoftheschwasuffixinthedativesingularisbettercharacterizedasthelossofthere-adjustmentruleaddingschwainthedativesingular.
Inotherwords,thedativesingularhascometobemorphologicallyderivedinthesamewayasthenominative/accusativesingular,Tag:withazerosuffix,/tag+/.
63Thefinalschwasthemselvesgobackoriginallytothesuffix/-ja/.
64Itispossiblethattheseparticularadjectivesweremorefrequentlyfoundintheiruninflected,predicativeform,wheretheywouldbesituatedattheright-edgeoftheprosodicphrase.
If,asisplausible,apocopeoriginatedasaphrase-edgephenomenon,onlylaterbeingextendedtotheword-edge,thentheseadjectiveswouldhaveundergonephonologicalapocopeearlier,andhencewouldbeliabletore-analysisasschwa-lessforms,i.
e.
asa-stemadjectives.
Butthissupposesthatapocopeoriginallydidnotrespectthe[voice]valueoftheprecedingobstruent.
WewillattempttoworkoutwhatreallyhappenedwithEarlyModernGermanapocopelateron.
130InthecasesofdativesingularTag,ortheadjectivesmild,schrg,theoutputsofanalogicalapocopemaybelegitimatelycharacterizedasinputstofinaldevoicing,butnotbecauseoftheinsertionofamorphophonologicalre-adjustmentrule.
Rather,thelexicalre-analysisofadjectivesmeansthattheunderlyingrepresentationoftheseadjectiveswillnolongerhavefinalschwa.
Therefore,theyarenotinputsforapocope,buttheyareinputsforfinaldevoicing:(50)OldUR/mld/NewUR/mld/Devoicing–mltApocopemld–Similarly,whentheschwa-addingre-adjustmentrulefortheMiddleGermandativesingularislost,wearenolongerdealingwithunderlying/tag/,i.
e.
withaninputtoapocope,butratherunderlying/tag/,i.
e.
aninputtofinaldevoicing.
Inconclusion,theevidencethatapocopewasamorphophonologicalrulefromtheoutsetisnotcompelling,andthereforewecannotplausiblyexpectittohavebeeninsertedbeforedevoicingonthatground.
4.
3ApocopeasalexicalphonologicalruleAlegitimateobjectionisthatHalle'smodelofgrammar,withstrictlyseparatedmorphophonologicalandphonologicalcomponents,hasbeensupersededinrule-based131phonologybyothermodels.
ThemorphophonologicalrulesofHalle'smodelnowcorrespondforthemostparttothelexicalphonologicalrulesofLexicalPhonologyandMorphologyorLPM(Kiparsky1982,1988;Mohanan1986).
Asignificantdifferencebetweenmorphophonologicalandlexicalrulesisthatlexicalrulesareexpressibleinpurelyphonologicalterms,includingintermsofconditioningenvironment.
Morphologicalconditioningisaby-productoftheinterleavingofthephonologicalruleswiththerulesofmorphologicalderivationandinflection.
Forexample,thephonologicalenvironmentofk-palatalizationinelectri[s]ity,cf.
electri[k],isthefollowingfronthighvowel;howevertheruleonlyappliesatthemorphologicallevelatwhichthe–itysuffixisadded.
TheenvironmentofProto-Indo-Europeanablaut,ontheotherhand,cannotbeexpressedinphonologicaltermsatanylevel;theruleisentirelyconditionedbymorphosyntax.
InLPM,lexicalrulesarestrictlyseparatedfrompostlexicalrules,accordingtowhethertheruleappliesduringthebuildingofthewordinthemorphology,orafterthewordsarelinearlyorderedaftersyntax.
Forexample,Englishk-palatalizationdoesnotapplyaftersyntax,henceki[k]it.
Bycontrast,t-flappingdoesapplyaftersyntax,hencehi[]it.
UnlikeinHalle'smodelofphonology,theseparationoflexicalandpostlexicalcomponentsisrelatedtohowthephonologyinteractswithmorphologyandsyntaxasaparallelsystem,nottowhatkindsoffeaturesareassociatedwiththeruledescriptionsthemselves.
Bothlexicalandpostlexicalrulesarestrictlyphonologicalintheirstructuraldescription.
Letussupposewerepresentapocopeasalexicalphonologicalrule.
Wecanimaginethattheapplicationofapocopetothedativesingular,butnottothenominative132pluralofmasculinenounsisduetothefactthatitwasaddedtothelexicalphonologywithoutgoingthroughastageasapostlexicalrule.
Althoughassigningtwoinflectionalaffixestodifferentmorphologicallevelsmayseemadhoc,andofitselfanargumentagainstthelikelihoodthatapocopewasaddedasalexicalrule,wewillallowitforthesakeofargument.
Thederivationwouldbeasfollows:(51)UR/tag+DS//tag+NAPl//tag+NAS/Level1tag+----Apocopetag----Level2--tag+--FDtak--takSR[tak][tag][tak]Weassumethattheschwa-addinginflectionalruleswereassignedtodifferentmorphologicallevels,dependingonwhetherwearespeakingofthedativesingular(level1)orthenominative/accusativeplural(level2).
Wethenalsoassumethattheapocoperuleislevel1,withtheresultthatonlythedativesingularschwaislostthroughapocope.
Finaldevoicingisunaffected,sinceitisapost-cyclicrule,andappliestotheoutputofthislexicalapocope,whichhasbeeninsertedbeforeitinthederivationalseries.
However,similarissuesarisewith"lexical"apocopeaswith"morphophonological"apocope:wearehardputtodeterminetheprecisemorphologicalconditioningoftheapocoperule.
Exceptionalapplicationorfailureofapocopeappearsto133bedueonthewholetoword-by-wordlexicalre-analysis(i.
e.
old-fashionedanalogy),orelsethelossofsomeinflectionalruleaddingschwa(e.
g.
thelossofthedativesingularinflection).
Thereisnogoodevidencethatapocopeappliedonlyataspecificmorphologicallevel.
Thereisoneaspectofschwaapocope,though,whichpointstoalexicalrule,namelythefactthatitappliesattheword-edge.
Sincetheruleislimitedtothedomainoftheword,forthisreasonitcanbetreatedasalexicalrule.
Wedonothaveunambiguousevidenceintherecordforanythingpointingtopostlexicalapocope,suchasapredominanceofapocopeatthesentenceorphraseedge.
However,evenifapocopewasaddedasalexicalrule,wewillseebelowthatfinaldevoicingwasalreadyalexicalrule,andthereforethatapocopeshouldnothavebeeninsertedbeforeitintheseries,ifwerespectHalle'sconstraintsonruleaddition65.
Inanycase,apocopemostlikelybeganasapost-lexicalrule,acquiringlexicalconditioningonlylateron.
Ifthisisthecase,thereshouldhavebeenanintermediatestageinwhichFDwasorderedbeforeapocope,givingoutputslike*[tag].
Thiswouldconformtotheconsensusamonghistoricallinguistsconcerningtheriseoflexicalphonologicalrulesinthegrammar,whichisthatlexicalrulesariseoutofthere-analysisofolderpostlexicalrules,i.
e.
rulesthatwerepreviouslytransparentandpurelyphonologicalintheirconditioning.
Thisisknownasthe"life-cycle"ofphonologicalrules(Kiparsky1989,McMahon2000b)66.
65In(15),werepresentedfinaldevoicingasapostcycliclexicalrule,whichisnotthesameaspostlexical.
66Kiparsky1989alsoshowedthatthelexical//-tensingofNewYorkandPhiladelphiacanbeexplainedasthelexicalizationofanolder,postlexicaltensingrulethatmorecloselyresembledtheNorthernCitiespattern.
Hedoesnotexplicitlyruleoutthepossibilitythatalexicalrulemightbeaddeddirectlytothegrammar,buttheimplicationofhisargumentisthatlexicalrulesonlycomefromthelexicalizationofpostlexicalrules.
1344.
4FinaldevoicingasaphoneticruleHavingconcededthatEarlyModernGermanapocopewasprobablyneitheramorphophonologicalnoralexicalphonologicalrule,wemightstillsupposeruleinsertiontohaveoccurred,followingthemodelofHalle1962,ifwecanshowthatfinaldevoicingwasphonetic,ratherthanphonological.
Forthis,weneedtoexaminewhetherfinaldevoicingfulfilsanyofthecriteriaforaphoneticrule.
AccordingtoKing1973,phoneticrulescanbedistinguishedfromphonologicalrulesbythefollowingcriteria:theyareunordered,i.
e.
theyallapplyatonce;theyadmitofnoexceptions;theydonotnecessarilyrefertobinaryfeatures,i.
e.
theycanassignintegervaluestofeatures;andtheyare"diachronicallypersistent",i.
e.
resistanttoloss.
Thelastcriterion,ofcourse,meansthatanynewphonologicalruleswillautomaticallybeaddedbeforesuchphoneticrules.
LetusreturnforthemomenttoSievers'Law.
Certainly,inProto-Indo-EuropeanthereisnoevidencethatSievers'Lawiscruciallyorderedwithrespecttoanyotherphonologicalrule;Proto-Indo-EuropeansyllabificationofsonorantsequencesfedSievers'Law(Ringe2006:16),butSievers'Lawdidnotfeedanyotherrules.
Moreover,Sievers'Lawhadnoexceptions,asfaraswecantell,thoughthereisthequestionofwhetheritappliedtoword-initialsequencesaswellasword-internalones(seediscussioninRinge2006:16-17).
Onotherhand,thereisnoevidencethatSievers'Lawreferredtonon-binaryfeatures.
Whetherwethinkofitasinsertionofavowelsegment,orasrewriting[-135syllabic][j]as[+syllabic][i],thesecategoriesremainbinaryanddiscrete.
Finally,thediachronicpersistenceoftheruleispreciselywhatisunderquestionhere:weknowtheruleexistedbothinProto-Indo-EuropeanandinProto-Germanic,butthequestioniswhethertherulecontinuedtobeexceptionlessinthetransitionbetweenthetwostages,orelseiftherewasanintermediateperiodofopacityasin(10).
SoSievers'Lawfulfillsatleasttwoofthecriteriaforaphoneticrule.
Theproblemremainsthatthedistinctionbetweenphonologicalandphoneticrules,aspresentedbyKing,isunconvincing.
Firstly,thereisnoreasontothinkphonologicalrulesmusthaveexceptionsorbecruciallyordered,asopposedtophoneticruleswhichmustnot.
Rather,theymayormaynothaveexceptionsorbecruciallyordered;inotherwords,absenceofexceptionsorcrucialorderingarenotdiagnosticofphonetic,asopposedtophonologicalrules,paceKing.
Itwouldbemoreaccuratetosaythatphoneticrules,asdefinedbythelattertwocriteria,alsohappentofulfilltheformertwo,butthatthesetwoconditions,transparencyandlackofordering,donotapplyonlytophoneticrules.
MiddleGermanfinaldevoicingwasalsoclearlyaphonological,notaphoneticrule.
Thefactthattheneutralizationwasrepresentedinthewriting,e.
g.
wc,wp,etc.
,stronglyarguesforatruephonemicneutralization,i.
e.
aneutralizationofadistinctivefeatureopposition(seealsonextsubsectiononincompleteneutralization).
Thestructuraldescriptionoftherule,inotherwords,itselfforbidsusfromattributingdevoicingtothephoneticcomponentonsynchronicgrounds.
Bywayofcomparison,agoodexampleofatruephoneticrule(oftenknownas"phoneticimplementationrule")isthetensingof//inNorthernCitiesAmericanEnglish(Labov,Yaeger&Steiner1972,Labov1994).
Thereisnoclear-cut,binary136distinctionbetweentenseandlax//;instead,wefindquitefine-grainedphoneticconditioning,e.
g.
thereisastatisticallysignificantdifferenceindegreeofphonetictensingdependingonwhetherthevowelisprecededbyacoronaloradorsalconsonant,asingleconsonantoracluster,orwhetherthevowelisinanopenoraclosedsyllable.
Theseconditionsinvolvemorethansimplybinaryoppositionsandcannotbeeasilycapturedinadistinctivefeatureframework67.
Theonlycriterionlefttous,inthatcase,isthediachronicone.
Ifweknowofacasewhereanolderruleappliedatonceandwithoutexceptiontotheoutputsofanewerrule,thenthatconstitutesevidencethattheolderruleisphonetic,notphonological.
However,inthecaseofreconstructedlanguageslikeProto-Indo-EuropeanandProto-Germanic,wedonothavetheevidencetomakethisargument,sincenotalltheintermediatediachronicstagesareavailabletous,butonlythosethatcanbedeterminedbycomparativeorinternalreconstruction.
SoSievers'Law,andEarlyModernGermanfinaldevoicing,mightbecalledphoneticrulesonthegroundsthattheyseemtoapplyautomaticallytotheoutputsofphonologicalapocope,butthisisonlybecausewearecomparingthesituationsbeforeapocopebeginsandafterithasconcludedandanyanalogicallevelingmayhaveoccurred.
Ifitwerenotforthesediachronicfacts,therewouldbenogoodreasontoconsidereitherruleaphoneticrule.
67Note,however,thatKiparskytreatsthisgradientphoneticimplementationruleasapostlexicalphonologicalrule(Kiparsky1989).
Forhim,itappears,thereisnomeaningfuldistinctionbetweenphoneticandphonologicalrules,butratherbetweenpostlexicalandlexicalrules.
Theessentialcriterionisthereforenotoneofdiscretenessversusgradience,butofwhethertheruleappliesbeforeoraftersyntaxandlinearizationofthesyntacticinput.
1374.
4.
1FinaldevoicingasincompleteneutralizationAwordmightbesaidhereabout"incompleteneutralization"(e.
g.
Port&O'Dell1985).
SomeexperimentalstudiesofspokenModernGerman,andotherlanguageswithfinaldevoicing,purporttoshowthattheneutralizationruledoesnotcompletelyeliminatetheunderlyinglaryngealcontrast.
Thus,thefinalcoronal[rat]fromunderlying/rad/'wheel'hasslightlybutsignificantlymorevoicingthenthecoronalin[rat]from/rat/'council'.
Ifthisistrue,thenourcharacterizationofMiddleandEarlyModernGermandevoicingaspurelyphonologicalmaybeincorrect,sincetherulewouldnotbeneutralizingabinaryfeaturedistinction,butwouldinsteadbeapproximatingoneofthefeatures,i.
e.
[+voice],totheother,i.
e.
[–voice],insuchamannerthatthereispartialoverlapinperception.
However,Fourakis&Iverson1984showedthatthepartialrealizationoftheunderlyingcontrastfoundintheseexperimentsisonlymanifestedwherespeakersareconsciousoftheunderlyingcontrastthroughthereadingoftherelevantminimalpairs,sinceModernGermanorthography,asnotedabove,representstheunderlyingcontrast,anddoesnotshowfinalneutralization.
Wherespeakersarenotmadeconsciousoftheunderlyingcontrast,i.
e.
wheretheydonothaveaccesstotheorthographicrepresentation,theneutralizationappearstobecomplete,i.
e.
wearelookingatneutralizationofdiscretephonologicalcategories,notthephoneticapproximationofonecategorytoanother.
GiventhatthedevoicingruleofModernGermanisclearlyphonological,wehavenogoodreasontosupposetheMiddleGermanrulewasanydifferent(cf.
below,ontheorthographicevidence).
1384.
5FinaldevoicingasapostlexicalruleAlthoughwehavenogoodreasontosupposeMiddleGermanfinaldevoicingwasphonetic,ratherthanphonological,itispossiblethatitwasapostlexicalphonologicalrule.
Ifthisisthecase,thentheadditionofanapocoperulemayhaveresultedininsertionofapocopebeforedevoicinginthederivationalseries.
Wecannotaccountforthisbythefactthatapocopewasaddedasalexicalphonologicalrule,sinceweshowedabovethatthemorphologizationofapocopeisnotconsistentwithasinglemorphologicallevel,andinanycase,themorphologicalfactorsinthedistributionoffinalschwacanbeattributedtoanalogicalchangestakingplaceafterapocopeoccursquasoundchange.
Butwemaybeabletoclaimthatbothfinaldevoicingandschwaapocopewerepostlexicalrules,inwhichcase,undertheassumptionsofLPM,theywouldhavetobearrangedintransparent,feedingorder,becauseopaqueruleorderingisnotallowedinthepostlexicalcomponentofphonology.
Wecouldconceiveofpostlexicaldevoicingasfollows:althoughcyclicsyllabificationappliesonlywithintheworddomain(includingthesyllabificationofcliticswithintheprosodicword),devoicingmayapplypostlexicallyontheassumptionthatthesyllableboundariesestablishedbylexicalsyllabificationarestillvisibleaftersyntax,sincedevoicingproperlyappliesatthesyllableedge.
However,ifthisisthecase,apocopeoughttohaveappliedatthepostlexicallevelaswell,sinceitalsoappliedtotheoutputofthecyclicsyllabificationandfoot-buildingrules,andtheevidencethatdevoicingwaspostlexicalisnotgood.
Paul2007claimsthattherearecasesofdevoicingbeingblockedwherethewordinquestionisfollowedbyaclitic,e.
g.
neiger,not*neicer,'heleaned'.
Yetcliticizationisusuallytreatedasadistinctprocessfromsyntacticlinearization,i.
e.
itisaspecialphonologicaloperationthat139takesplaceafterwordbuildingbutbeforesyntacticphrasing.
Soneigercanbethoughtofasasingleprosodicwordforthepurposesofword-levelsyllabificationandfinaldevoicing68.
Inthatcase,wefindthatdevoicingalwaysandonlyapplieswithinthedomainoftheword,makingitstrictlyspeakingalexicalrule,albeitapostcyclicone,becauseitappliesafterallthesyllabificationcyclesthattakeplaceaftereachlevelofmorphology.
InGerman(unlikesomeotherlanguages),thereisnosyllabificationcycleaftersyntax.
Therefore,wedonotexpecttheapocoperuletohavebeeninsertedbeforefinaldevoicingintheseriesofrules.
If,ontheotherhand,wetreatfinaldevoicingaspostlexical,thenwemustalsotreatapocopeaspostlexical,sincetheconditioningenvironmentforthetwoisthesame,i.
e.
therightedgeoftheword.
4.
6IntrinsicruleorderingTheprecedingdiscussionhasassumedthatrulesareorderedextrinsically,i.
e.
thattheorderofrulesinanygivenlanguageisessentiallyanaccidentofhistory.
Undertheassumptionsofextrinsicruleordering,thesynchronicorderofrulesshouldreflectthechronologyofchanges,unlessotherfactorstriggerrestructuringorinsertion.
Restructuringoccursifthegrammarresultingfromruleadditionismorecomplexthannecessarytoaccountforthefacts,whileinsertionoccursiftherulebeingaddedbelongstoadifferentmoduleofgrammarfromtheolderruleinquestion.
Wehavealreadyshown68Asithappens,thisbleedingeffectoffollowingvowel-initialcliticswasnotobservedinmysearchoftheBonncorpus,i.
e.
gapetc.
werefoundbeforecliticslikeerasoftenasbeforenon-clitics,whethervowel-initialorconsonant-initial.
140thattheconditionsforruleinsertion,asestablishedbyHalle1962,arenotmetwithrespecttothefactsofapocopeanddevoicingasweknowthem.
Analternativeapproach,however,istoassumeintrinsicruleordering(Koutsoudas,Sanders&Noll1974).
Accordingtothistheory,rulesshouldnormallybeorderedinsuchawaythattheyallapplymaximally.
Inotherwords,universalgrammarpredictsthatthereshouldbenoopacity.
Thistheoryhelpstoexplainwhyopaquegrammarstendovertimetoberestructuredintotransparentorders.
Thistheoryshouldalsopredictthat,whenrulesareaddedtothegrammarthroughsoundchange,theyareaddedatthepointinthederivationalserieswheretheywillapplytothemaximalextent.
So,accordingtointrinsicorderingtheory,weexpectapocopetohavebeenaddedtothephonologybeforedevoicingbecausethatwaydevoicingwouldapplymaximally,i.
e.
transparently,totheoutputsofapocope.
Theempiricalfailureofintrinsicruleorderingwasnotedalreadyinthe1970s(King1976).
InthecontextoftheEarlyModernGerman,wealreadyhavetheexamplesofYiddishandUpperGermantoshowthatintrinsicorderingisnotjustifiedbythedata.
Inresponsetheproponentsofintrinsicruleorderingmightstipulatemechanismsthatwouldpermitatleasttemporaryopaqueordering,withtheresultthattheconceptunderlyingintrinsicorderingtheory,MinimizationofOpacity(King1976:79)wouldberelegatedtoanoverallprinciple,similartomarkedness,thatwoulddrivelearnerseventuallytorestructureopaqueorderingsintotransparent,feedingorderingswherepossible.
Inshort,intrinsicruleorderingisnotstrongenoughasatheorytoleadustobelievethatapocopewasinsertedbeforedevoicinginthegrammar.
1414.
7PhonologizationandopacityThereisonelastreasontobeskepticalaboutruleinsertion,independentofthepredictionsofdifferenttheoreticalframeworks:whatdoesourunderstandingofphonologizationpredictconcerningruleinsertionThediscussionsofarhasassumedthatphonologicalrulesareaddedinstantaneouslytothegrammar,butthefactisthattheadditionofphonologicalrulesisitselfaprocess,andwithinthatprocesswemightfindaclueastowhywepredicttheadditionofapocopetoresultinopacity.
Phonologizationistheprocesswherebyphonologicalrulesenterthegrammar,throughthereanalysisofperformance-relatedeffects,likeco-articulation,asgrammaticalvariables(Hyman1976,cf.
Bermúdez-Otero2007).
Forinstance,inalllanguages,dorsalstopsareslightlypalatalizedbeforefronthighvowels.
Thisisbecauseofthepredictableacousticeffectthatoccursduringthetransitionfromthestoparticulation,wherethebodyofthetongueisraisedtowardsthebackoftheoralcavity,tothevowelarticulation,wherethebodyoftongueismovedtowardsthefrontofthecavity.
Evenifthelanguagehasnopalatalizationrule,thiseffectissignificantuponacousticanalysis.
Whyisitthatsomelanguagesneverthelesshaveapalatalizationruleoverandabovethiscoarticulationeffect,e.
g.
Italian,SwedishThisoccurswhenthelearnerinterpretstheacousticeffectasphonologicallysignificant,i.
e.
theslightphoneticpalatalizationisinterpretedastheoutputofaphonologicalrule,whichrewritesunderlyingstopassurfaceaffricateorfricative.
Inthisway,theadditionofphonological142rulesisnolongerspontaneous,butisitselftheproductofre-analysisofextra-linguisticphoneticeffects69.
WhenweturntoEarlyModernGerman,wecanimaginethephonologizationofapocopeinthefollowingway.
Sincethefinalschwaisunstressedandcentralized,wecanimaginethatitwouldoftenbeonlyfaintlyarticulated,ornotatall.
Thiseffectwouldoriginallyhavenothingtodowithadeletionrule;thevowelwouldstillbepresentinthespeaker'sphonologicalsurfacerepresentation.
Butthelearnerdoesnotknowwhattherepresentationissupposedtobe,soheorshewillmisperceivethisirregularacousticdeletionasphonologicaldeletion.
Hence,arulewillbeaddedtothegrammartoaccountforthedeletion.
Ofcourse,thisrulewillhavetobevariable.
Sinceonlysomeoftheschwaswillhavedisappearedintheacousticsignal,therulethatthelearnerpositstoaccountforthedisappearancealsohastotakeintoaccountthefactthatelsewherethereisavowel.
Theinputoftheruleisthevariantwithfinalschwa;theoutputisthevariantwithoutfinalschwa.
Theratiooftheformersetofvariantsoverthelatterwillbeexpressedasaprobabilityvalueattachedtothevariablerule.
However,westillhavenotaccountedforwhythevariableschwadeletionruleisaddedtotheendoftheseriesofrules,resultinginacounter-feedingorder,andnotearlierinthederivation,sothatitfeedsdevoicing.
Thedeletionofschwaisnottargetingthe69Infact,thetransitionfromphoneticeffecttophonologicalrulemaybemorecomplicatedthanthis.
Aswementionedbefore(p.
17),thereisalsosuchathingasagradientphoneticrule,i.
e.
arulemanipulatingphoneticfeaturesthataremorefine-grainedthanbinarydistinctivefeatures,yettheoutputofsuchagradientruleisacousticallymoresignificantthanamerecoarticulationeffect(seediscussionofNorthernCities//-tensinginLabov,Yaeger&Steiner1972).
Sothecourseofeventsisasfollows:coarticulationeffect,thengradientphoneticrule,andthencategoricalphonologicalrule.
Whenconsideringchangelikeschwaapocope,ontheotherhand,anditsrelationshipwithdevoicing,itonlymatterswhetherthevowelispresentornotpresent,soweonlyneedtodealwiththetransitionfromphoneticeffecttophonologicalrule.
143inputsoroutputsofthedevoicingrule,sowhydothelearnersfailtoapplydevoicingtotheoutputsofschwadeletionIstherereallysomeuniversalformalconstraintongrammarchangethatforbidsrulesfrombeingaddedanywhereotherthantheendoftheseriesThisquestionassumesthatlearnerswillalreadyhaveacquiredatransparentdevoicingrule,evenastheypositadeletionruletoaccountforthevariableabsenceofexpectedfinalschwa.
Yettheabsenceoffinalschwamustnecessarilyaccompanyanapparentexceptiontofinaldevoicingincaseslike[tag],forexpected[tag].
Giventhatacousticdeletionofschwanecessarilyresultsinsurfaceexceptionstofinaldevoicing,wenowseehowopacityisanecessaryoutcomeofthiskindofphonologicalinnovation.
Wecannotpredictforcertainthatlearnerswillbeeasilyabletoacquirerulesthatareorderedopaquely,ifbothrulesarephonologicallypredictable.
Kiparsky1978notedthattheeffectofrulere-orderingmightarisewherelearnersacquirerulesinthe"wrong"order.
Takefinaldevoicingandapocope.
Inorderforopacitytoarise,learnersmustacquirethedevoicingrulefirst.
If,onthecontrary,theyacquiretheapocoperulefirst,thentheoutputsofapocopeshouldprovidetheinputstodevoicingwhenitisacquiredlater.
However,ifweacknowledgethatevenanearlieracquiredapocoperulecreatesexceptionstothesurfacepatternofdevoicing,thenitwillbeimpossibletoacquiretransparentdevoicingeitherbeforeorafterapocopeisacquired.
Thelearnerwillthenfaceachoicebetweenananalysisinwhichthedevoicingruleexceptionallyfailstoapplytotheoutputsofapocope,oroneinwhichdevoicingissimplyorderedbeforeapocopein144thederivation.
Thelatterissimplerandthereforewepredictlearnerswillchooseit,allelsebeingequal.
4.
8ConclusionsTheproblemisthatthereare,aswehaveseen,unambiguousexamplesofregularsoundchangeintroducingopaquegrammars.
WehavetheexamplesofdevoicingandapocopeinYiddishandEarlyModernGerman.
Otherwell-knownexamplesincludeVerner'sLaw,apre-Proto-Germanicchangewherebyfricativesbecamevoicedwhenprecededbyanunaccentedsyllable:lateProto-Indo-European/patér/'father'>/faθér/>/faér/,cf.
/bhráter/'brother'>/bróθer/>Gothic/broθar/.
Subsequently,theaccentwasshiftedtotheinitialsyllableinallwords:/fáer/,fromwhichweeventuallygetGothic/faar/,andEnglishfather.
Afteraccentretraction,Verner'sLawappliedopaquely;wedonotfind/faer/changingto*/faθer/,whichiswhatshouldhaveoccurredifruleswereorderedintrinsically.
Thegeneralconclusiontobedrawnfromthisreviewcomesdowntothefollowing:themeretheoreticalpossibilityofruleinsertiondoesnotoutweighthefirmempiricalevidenceagainstruleinsertioninthecaseofEarlyModernGermanfinaldevoicingandschwaapocope.
Tothisisrelatedthebroaderobservationthatrulesdevelopinverypredictableways.
AsnotedinKiparsky1989andtheninMcMahon2000b,rulesalwaysseemtobeginasgradientphoneticrules,laterundergoingphonologizationaspostlexicalrules,thenmorphologizationaslexicalphonologicalrules,145beforefinallybeingcompletelylexicalizedandremovedfromthegrammar.
TheevidencefromEarlyModernEastCentralGermansupportsthishistoricalmodelanddoesnotprovidegoodevidencetosupporttheruleinsertionhypothesis.
5OpacityandrestorationoftransparentdevoicinginEastCentralGermanInthissection,wewillacceptthatthewrittenevidenceconcerningthe14thand15thcenturyEastCentralGermantextsisaccurate,andthatdevoicingreallydidbecomeopaque.
Thisincludesboththeevidenceforopaquedevoicing,i.
e.
significantlyhigherratesofdevoicinginthenon-apocopatedgrammaticalcategoriesofwords,e.
g.
tac>tac,buttage>tag,andtheevidenceforactuallossofdevoicing,i.
e.
thelevelingoffinalvoicingtothenon-apocopatedcategories,e.
g.
tac>tag.
Thatthewrittenevidenceof14thand15thcenturyEastCentralGermanshouldbetreatedasreliableisindependentlyconfirmedbythefactthatfinaldevoicingcontinuedtoberobustlyattestedinthespellingupuntilthe16thcentury(Ewald1997,Mihm2004),whereasinUpperGerman,aswehaveseen,itwasmostlylostbytheendofthe15thcentury.
WhatwefindinEastCentralGerman,inotherwords,isarapidtransitionfromasituationwherefinaldevoicingisrobustlyattestedandapocopehasonlybeguntomakeinroads,toasituationwheredevoicingislostcompletely,andyetapocopestillhasnotgonetocompletionandhasevenbeenreversedinsomecases.
ThisstronglysuggeststhatthelossofdevoicinginEastCentralGerman,unlikeUpperGerman,wasnotdirectlycausedbyapocope,butwasduetosomefundamentalchangeinspellingconventions.
146Thereforewewillconsidertheevidenceoflossofdevoicinginthe16thand17thcenturytextstobeinsufficienttoprovethelossofdevoicinginEastCentralGerman,becauseapocopeneverwenttocompletioninthoseareas.
WecansupposethattheEastCentralGermanprintersadoptedtheconventionofnotrepresentingdevoicingfromtheUpperGermanprintersinEastFranconia,wheredevoicinghadbeenlost.
Inthatcase,wemustexplainhowopaquedevoicingmanagedtobecometransparentagaininthephonologyofEastCentralGerman.
5.
1AnalogicalrestorationoffinaldevoicingAsinDutch,thelossoffinalschwainEarlyModernGerman,whetherbyregular,conditionedsoundchange,orelsebyanalogicalre-modeling,willinmanycaseshaveintroducednewfinalvoicedobstruents:obe>ob,unde>und,allerhand>allerhand,herzoge>herzog70.
Althoughthisisnotreflectedinthemodernspelling,finaldevoicinghasappliedtoallsuchwordsinspokenModernGerman.
Onepossibilityisthatweattributethisseconddevoicingtoanalogy.
Inotherwords,thefinal/g/ofDSgTagisremodeledontheanalogyoftheotherformsthathavefinaldevoicing,suchasNSgTa[k].
However,onlyformsthatparticipatedinalternationswithinparadigmswouldbeliabletosuchanalogicalremodeling.
Thus,dativesingular[tag]couldbecome[tak]ontheanalogyoftheparadigmaticallyrelatednominativesingular[tak].
However,wordsthatdidnotparticipateinparadigmaticalternations70Rememberthatthefailureofapocopetooccurwheretheprecedingsegmentisavoicedobstruentonlyappliestodisyllables,i.
e.
wheretheschwacomesbetweentheprimarystressedsyllableandthewordboundary.
Afterthesecondarystressedsyllable,Behaghel'sLawtakesprecedence,asinallerhand;inthecaseofobe>ob,wearelookingatthelateMiddleGermanapocopeaffectingunstresseddisyllables.
147cannotbeexpectedtoundergoanalogicalremodeling,andyetdevoicingisseentoapplytothemaswell,e.
g.
obe>o[p].
Forsuchcases,anexplanationotherthananalogicalremodelingisrequired.
Afurtherproblemisthatanalogicalchangeismeanttoproceedwordbyword,andalexicallygradualchangeisnotexpectedtoyieldtheoverwhelminglyregularresultsthatsecondfinaldevoicingshow.
Ifdevoicingwereanalogicallyextendedtothenewfinalvoicedobstruents,wewouldexpectsomeformsnottoundergoanalogicalremodeling,suchastheoccasional*Ta[g],*o[b]andsoon.
Thefactthatfinaldevoicingissopervasiveinthemodernlanguagecallsoutforanotherexplanation.
5.
2Rule-basedrestorationoffinaldevoicingRatherthantheproductoflexicallygradualanalogy,wewouldarguethatsecondfinaldevoicingwastheresultofasynchronicdevoicingrule,ofthekinddescribedatthebeginningofthischapter71.
Asnotedabove,MiddleGerman,asfaraswecantellfromthemanuscripts,hadarobustlyattesteddevoicingrule,anditmakessensetosaythatthisruleisresponsibleforthelateranalogicalextensionofdevoicingtonewfinalvoicedobstruentsthatarosefromlateMiddleGermanandEarlyModernGermanapocope.
However,theevidencethatwehaveseenarguesstronglyforopacityinEarlyModernGermanfinaldevoicing,asaresultoftheapocopeoffinalschwa.
Inother71Hock1991hasausefulchapter(pp.
238-280)onthecorrespondencebetween'regular'analogicalchange,andthegenerativeconceptofthesynchronicrule.
Whileinfactcasesofanalogicalchangecouldbeeitherregularorirregular,therewasnothingintheanalogicalmodelthatwouldpredictregularoutcomes,sinceitsupposedlyalwaysproceededwordbyword.
Hockshowshowtheconceptofsynchronicruleshashelpedtoexplainmanyformerlyinexplicablecasesof'regular'analogicalchange,butatthesametimehasthrownupproblemssuchastheonedescribedhere,whichwouldnothaveoccurredtolinguistsworkingundertraditionalassumptions.
148words,EarlyModernGermancorroborateswhatwereconstructfortheancestorofYiddish.
Assumingthatthisistrue,weneedtoexplainhowdevoicing,althoughnowopaqueasaresultofapocope,neverthelessre-appliedtoalltheoutputsofschwaapocope,ratherthanbeingeliminatedfromthegrammarasYiddishdevoicingwas.
5.
2.
1Rulere-orderingWehaveseenthat,althoughruleinsertioncannotbecompletelyruledout,neitherthestructuraldescriptionoffinaldevoicingnorofschwaapocopeleadsustoexpectinsertion,evenwerewetoleaveasidethestrongempiricalevidenceagainstinsertionofapocopethatwediscussedinthepreviouschapter.
Anotheroption,then,isrestructuringofthegrammarafterruleadditionoccurs,whichwouldconsistinthere-orderingoftherulesfromanopaque,counterfeedingordertoatransparent,feedingorder.
Thusthere-applicationofbothSievers'LawandEarlyModernGermanfinaldevoicingcanbeexplainedastheoutcomeofrulere-ordering.
ConsidertheexampleofSievers'Law.
FirstwepredictthatEpenthesiswouldbeaddedattheend,andthentheorderofruleswouldbereversed,resultingintheapplicationofSievers'LawtotheoutputsofEpenthesis.
Epenthesismustbeaddedattheendbecauseitisacategorical,phonologicalrule.
AftertheadditionofEpenthesisattheendofthederivation,however,re-orderingoccurs,triggeredbysomeconsiderationofmarkednessorelselearner-drivenreanalysis.
Onemotivationforrestructuringofgrammar,includingre-ordering,isthedesiretoreduceexcesscomplexity.
However,thereisstrictlyspeakingnodifferencein149complexitybetweenanopaquegrammarandatransparentgrammar:complexityisonlyincreasediftherearemorerulesthanareneededtoaccountforthesurfacealternations.
Therefore,othermotivatingfactorsforgrammarrestructuringmustbesoughtout.
AccordingtoKiparsky1968,rulesshouldbere-orderediftheyareincounter-feedingorbleedingorder,tofeedingandcounter-bleedingorders,respectively,owingtotheinherentmarkednessofcounter-feedingandbleedingorders.
AndtheorderingofEpenthesisafterSievers'Lawisindeedacounter-feedingordering:thatis,ifEpenthesiswereinsteadorderedbeforeSievers'Law,thenEpenthesiswouldfeedSievers'Law.
So,evenaccordingtoKiparsky'soriginalconditions,Proto-GermanicSievers'Lawisagoodcandidate,sinceanunmarkedorderingreplacesamarkedone.
Accordingtotherevisedconditionsonre-orderingthatKingfollowsinhisarticle(seealsoKiparsky1971)),wefindthatorderingu-EpenthesisafterSievers'Lawisstillmarked.
Therevisedconditionsholdthatre-orderingisprompted:-bythedesiretoeliminateopacity-bythedesiretominimizeallomorphy(paradigmpressure)AswesawbeforeinthediscussionofYiddishapocopeanddevoicing,opacityisdefinedasfollows(Kiparsky1971):-AruleA→B/C__Disopaquetotheextentthattherearesurfacerepresentationso(i)oftheformAinenvironmentC__D,or150o(ii)oftheformBinenvironmentotherthanC__DTheformertypegenerallycorrespondstocounter-feedingorder(seealsoBakovic2007).
LetAbethesequenceCj,andBthesequenceCij,i.
e.
theinputandoutputofSievers'Law.
Asfortheenvironment,CistheprecedingsequenceVC,andDisthefollowingsegmentV.
Exampleslike*/wurgjeti/wouldbeexamplesofopacitytype(i),becausethesequenceA=Cj,appearsintheenvironmentVC__V,ratherthanthesequenceB=Cij.
Thus,wesaythatthecounter-feedingorderingof(11),withsurface[wurgjeti],isopaque,i.
e.
theSievers'Lawruleiscontradictedonthesurface,sinceEpenthesiscreatessegmentsequencesthatfitthestructuraldescriptionofSievers'Law,andyetSievers'Lawdoesnotapplytothem;inotherwords,Sievers'Lawisnotapplyingtoitsmaximalpossibleextent.
Undertheassumptionsthatopaqueruleorderingsaremarked,weexpectare-orderingintothetransparentorderingof(13),withsurface[wurgijeti],inwhichtheSievers'Lawruleappliesmaximally,i.
e.
theorderinginwhichtherearenosurfaceexceptionstoeitherrule72.
Inthesamemanner,theopaqueorderoffinaldevoicingandschwaapocopein(12)shouldhavetriggeredre-orderingintothefollowingunmarkedorder:(52)UR/ob//tag/Apocopeob–72Theotherconditionofrulere-ordering,theminimizationofallomorphy,maybecontradictedhere,sinceSievers'Lawcreatesallomorphybetweentwosurfaceformsoftheunderlyingsuffix/je/:[je]and[ije].
Therefore,wemightexpecttheparadigmconditiontoeliminateSievers'Lawcompletely,ratherthanencourageitsmaximalapplication.
However,sinceSievers'Lawwasonlyaffectingsuffixes,ratherthanthestems,onemightarguethattheparadigmconditiondoesnotapply,becauseonlystemallomorphsshouldbesubjecttoparadigmleveling(DonRinge,p.
c.
).
151DevoicingoptakThissolution,asithappens,requiresustoreconstructahypotheticalperiodwhereanopaquegrammardidexist,butforwhichtheevidencehasbeenlostowingtore-ordering.
5.
2.
2Rulere-affirmationHock1991:269speaksof"rulere-affirmation"toexplaincaseswherearuletemporarilyacquiresopacityasaresultofsomechange,butinsteadofbeinglost,somehowretainsproductivityandre-appliestotheexceptionsthathaveresulted(seealsoKrishnamurti1978).
Inhisformulation,thesynchronicmotivationforre-application,orre-affirmation,isunclear,becauseallthathappensisthatacopyoftheopaqueruleisaddedtotheendofthederivationalseries:(53)UR/ob//tag/Devoicing(1)--takApocopeob--Devoicing(2)op--Thereisnoexplanationofwhytheruleiscopiedinthatparticularexample,andnotinanotherexample,suchasYiddishfinaldevoicing.
Ifwehypothesize,however,that,forsomereason,theopacityofdevoicinginYiddishhinderedtheproductivityoftherule,whiletheopacityofdevoicinginEarlyModernGermandidnot,thenwemaybeableto152findanexplanationforthedifferentbehaviorofthedevoicingruleinthesubsequenthistoriesofthetwolanguages.
5.
2.
2.
1Finnisht-spirantizationKiparskynotedthefollowing(Kiparsky1971):"Themoreopaquearule,themorelikelyitistodevelopexceptions.
"Thismeansthat,whilesurfaceexceptionstoarulemaybeunderstoodbythelearnertobetheartifactsofsynchronicopacity,thereisalwaysthepossibilitythatthelearnerwillinsteadinterpretthesurfaceexceptionsastrueexceptionstotherule,i.
e.
whatwereferredtoearlierasthemorphologizationoftherule.
However,thereisnowayforusinprincipletodeterminewhenthelearnerwillmakethiskindoferrorornot,aslongasbothexceptionmarkingandopacityarepossibleanalysesofthedata73.
Kiparsky1973providesdetailsonaparticularhistoricalcasethatpointtoasolution.
TherulesinquestionarefromFinnish,namelyt-spirantizationandvowelcontraction:(54)t-Spirantizationt→s/__i(55)VowelContractionViAnexampleoftheformerishalusi'(he)wanted',from/halut+i/,whileanexampleofthelatterispiti'(he)held'from/pit+i/.
73SeeKiparsky1971and1978.
SeealsotherecentdiscussioninVaux2008andreferencesonopacityandlearnabilityinrule-basedphonology;forthesameissueinOT,seeBermúdez-Otero2003.
153Aswecansee,bothrulesapplybeforeunderlying/i/,andhencetheymaypotentiallyinteractinanopaquemanner.
Thus:(56)UR/haluti//piti/Spirantizationhalusi--Contraction--pitiTheopaqueorderingiscounter-feeding,thusitis"type(i)"opacityaccordingtoKiparsky'sformula.
Thischronologicalorderingmeantthatspirantizationbecameopaque.
Asaresultofthisopacity,spirantizationacquiredexceptions.
Atthesametime,however,spirantizationremainedaproductiverule,andsomeoftheexceptionsresultingfromopacityweregraduallyeliminatedatlaterstagesofthelanguage.
Thus,theverb'(he)sped'canbeeitherkiitiorkiisi,bothfromunderlying/kiit+i/.
Moreover,examinationofearlierrecordedstagesofFinnish,aswellasofregionaldialectspreservingmorearchaicpatterns,showsthatthecounterfeedingorderingismorecharacteristicofolderstagesofthelanguage,whereasinmorerecentFinnishwefindthefeedingorderbelow:(57)UR/kiit+i/ContractionkiitiSpirantizationkiisi154Thiswouldbeeasytoexplainastherepairofamarkedgrammarbyrulere-ordering,butforonething:productiveapplicationofspirantizationtonewinputsismorphologicallyrestricted.
Inotherwords,weonlyfinditinthepasttenseofverbswhosestemsendin/t/,andeveninthatcategorytherearemanyexceptions(e.
g.
whereas'sped'canbeeitherkiitiorkiisi,'held'isalwayspiti).
Othercategorieswherespirantizationneveroccursincludepastimpersonalverbs,conditionals,derivedverbs,derivedadjectives,andderivednouns:(58)Pastimpersonal:/men+t+i+hen/–>mentiin'onewent'Conditional:/tunte+isi+n/–>tuntisin'Iwouldknow'Derivedverbs:/sota+i+ta/–>sotia'towagewar'Derivedadjectives:/vete+i+nen/–>vetinen'watery',cf.
vesiDerivednouns:/sonta+iai+nen/–>sontiainen'dungbeetle'Nounplurals:/sota+i+na/–>sotina'wars'(essive)Eventhoughinallthesecategorieswefindoutputsofcontractionmeetingthestructuraldescriptionofspirantization,itisonlyinthepasttenseofpersonalverbsthatwefindthere-affirmationofspirantization,andevenhereitdoesnotapplyatthesamerateinallverbs.
Kiparsky'ssolutionistorejectre-orderingcompletely,andinsteadtoclaimthat,aftercontractionoccurredandrenderedspirantizationopaque,spirantizationacquiredmorphologicalandlexicalexceptions(the"exception-marking"solution),andwhatweseeisthegradualeliminationoftheseexceptionsovertime.
So,wherevercontraction155resultedinabsenceofexpectedspirantizationthroughoutanentirecategory,suchasderivedverbs,wefindnore-applicationofspirantization.
Thisiseasytoexplainifwesupposederivedverbstobecategoricallyexcludedfromundergoingspirantizationbythe"marking"theory,butifwesupposere-ordering,wehavetoexplainwhyre-orderingproducedexpectedtransparentspirantizationonlyinoneparticularcategory,namelythepasttenseofpersonal,non-derivedverbs.
Inotherwords,wehavetointroducemarkingofexceptions,evenifweinterpretthisasacaseofrulere-ordering.
Wecanchoosetosaytherulewasmarkedwithexceptionsfromthebeginning,orelsewasmarkedwithexceptionsimmediatelyafterre-ordering.
Theformersolutionissimplerandhencepreferable74.
Fromtheseobservations,wecansuggestthefollowinghypothesisconcerningEarlyModernGermanseconddevoicing:whentheadditionofschwaapocopecausedfinaldevoicingtobecomeopaque,theopaqueruleacquiredexceptions.
Sinceexceptionsincreasearule'scomplexity,therulebecamehardertolearn.
Learnersthenattemptedtoreducethecomplexitybyeliminatingtheexceptions75.
5.
2.
2.
2SecondfinaldevoicinginDutch74Notethatbothcontractionandspirantizationareproductiverules,butdespitethefactthattheycouldbeorderedwithrespecttoeachother,Kiparskyisclaimingthattheyapplyindependentlyandinanunorderedfashion.
Theexceptionstospirantizationhavebeenencodedinthedescriptionoftheruleitself;theyarenotartifactsofopaqueordering.
75Thisfitswithrecentacquisition-relatedworkbyCharlesYang(Yang2008)onexceptionsandtheproductivityofrules.
Thisanalysisofre-orderingasrestructuringoftherulefollowedbygradual"re-application"alsofitsthecaseofProto-GermanicSievers'Law.
Thus,afterpre-sonorantvowelepenthesisoccurred,thenumberofexpectedexceptionstoSievers'Lawwouldhavebeenverysmall;thisauthorcountedabouttenpossiblecasesthatshouldhavesurvivedfromProto-Indo-EuropeantoProto-Germanic,basedontheProto-Indo-Europeanverballexicon(Rixetal.
2001),outofwelloverahundred.
156ThishypothesisaboutsecondfinaldevoicinginEarlyModernGermanisconfirmedbyEarlyModernDutch.
AsinEarlyModernUpperGerman,apocopeshouldhaverenderedthedevoicingruleopaque,sothatdevoicingwouldhavebecomeunproductiveinatleastcertaincontexts,e.
g.
innounsandpartsoftheverbalparadigm.
YetinmodernDutch,devoicingisatransparentrule76.
Forthisreason,Goossensspeaksofa"secondfinaldevoicing"(Goossens1974:82-3),toaccountforexampleslikewe[b]e'web'>we[p].
Inhisformulation,seconddevoicingwasanactualsoundchange,independentoftheODsoundchange.
Inthis,Goossensfollowedstricthistoricalmethodology,anddidnotappealtotheapplicationofasynchronicdevoicingrule.
Otherphonologistsdidraisethispossibility,however(Goossens1977:3-4andreferences,allinDutch).
Itisnoteworthythattheyassumedanautomaticdevoicingrulewouldapplyatoncetoallnewinputsderivingfromapocope,i.
e.
theydidnotnoticetheopaque,counterfeedingorderingthatwouldhavearisenfromaddingapocopeafterdevoicing,asgenerativehistoricalphonologypredicts.
Goossensunfortunatelydoesnotraisethiscriticisminhisresponse(Goossens1977:5f.
),butratherdisplaysskepticismtowardsgenerativephonologyandsynchronicrulesystemsingeneral.
Yethisdismissalofthesynchronicrulehypothesiscannotbeitselfdismissedasmeretraditionalistreaction.
Rather,GoossensbroughtforthevidencefromdialectsurveysintheNetherlands,showingthatwhereapocoperecentlyapplied,devoicingdidnotapplyatoncetotheoutputsofapocope,butinsteadapocopegaverisetoexceptionstodevoicing,whichonlylaterwereeliminatedasdevoicingre-asserteditself.
76Ontheotherhand,asinModernGerman,Dutchspellingdoesnotrepresentthevoicingalternationinstopsandthefricative//,thoughitdoesrepresentsthealternationinthefricatives/v/and/z/,unlikeinGerman.
157AswiththeGermandialectalevidence,theDutchevidence(neatlysketchedonthemapsinGoossens1977:11,15)drawsfromdialectsstraddlingthebordersdividingapocopatingfromnon-apocopatingdialects.
ThelatterincludethedialectsofthenortheastNetherlands,commonlycalledDutchLowSaxon(sincetheyaremorecloselyrelatedtotheLowGermandialectsacrosstheborderinGermanythantostandardDutch),andtheWestFlemishdialectsoftheextremesouthwestoftheNetherlandsandthewesternendofDutch-speakingBelgium.
Intheseareas,owingtotherecentnessofapocope,theresultingexceptionstodevoicingarestillinevidence,e.
g.
be[d],cf.
MiddleDutchbe[d]e'bed',orru[g]'back',MiddleDutchru[g]e(ModernStandardDutchru[χ],withMiddleDutch/g/entirelyreplacedby//).
So,Goossenscertainlyshowsthatdevoicingdoesnotautomaticallyapplytotheoutputsofapocope;rather,sometimemustpassafterapocopehasoccurredbeforedevoicingre-appliestotheapocopatedforms77.
Whetherdevoicingre-asserteditselfwithinthedialect,orwhetheritspreadbackintotherecentlyapocopateddialectfromneighboringdialectswhereithadneverbeenlostisunclearfromGoossens'discussion.
Thelatterisadistinctpossibility,becauseapocopeshouldhavegivenrisetomanyexceptionsthatwouldhavemadeitdifficultforlearnerstocontinueanalyzingdevoicingasatransparentphonologicalrule.
Thenon-apocopatingdialects,wheredevoicinghadnotacquiredexceptionsandremainedtransparent,wouldstillbeincontactwiththeapocopatingdialects,wheredevoicingwasbecomingopaque,andsotherewouldhavebeenmuchopportunityfordevoicingtospreadback.
Thiskindofexplanationmaywellalsoliebehindtherestorationof77InGoossens1977:14ff,weseehowdevoicingwasobservedtore-applygraduallytotheexceptions.
158devoicinginapocopatingdialectsofGerman(seeaboveonSilesianandLuxemburgish),aswellasliteraryModernGerman(seebelowonLowGermaninfluence).
Thatbeingsaid,anotherexplanationfortherestorationoffinaldevoicinginthesedialectsisthatdevoicingremainedproductive,despitetheexceptions,inthesamewayFinnishspirantizationremainedproductivedespitetheexceptionsintroducedbyvowelextraction.
Suchanexplanation,ofcourse,dependsontherebeingtoofewexceptionstopreventthecontinuedproductivityofthedevoicingrule.
SinceweknowthatapocopeappliedabitmorehaphazardlyinDutchthaninUpperGermanorYiddish,thismayexplaintheretainedproductivityoffinaldevoicing.
Similarly,sinceEarlyModernEastCentralGermanapocopewasquitehaphazard,there-affirmationofdevoicingiseasiertoexplainthereby.
5.
2.
3EastCentralGermanapocopeandre-affirmationofdevoicingWhenapocopeoccurredinYiddish,weconcludedthatfinaldevoicingbecameunproductive,becauseofalltheexceptionsthatwouldhaveaccruedinthesurfaceapplicationofdevoicing.
Suchanunproductiverulewouldhavebeenliabletoloss,aslearnersgraduallyeliminatedcomplexityfromthegrammarbyceasingtoapplythebynowfullymorphologizedandlexicalizeddevoicingrule.
YetaswesawwithFinnishandprobablywithDutch,merelyhavingexceptionsisnotenoughtohinderproductivity.
Asufficientnumberofexceptionsarenecessary,suchthatthelearnerwillfailtoperceiveanyphonologicalregularityintheruleinquestion.
SinceapocopeinYiddishappliedtoaverylargenumberofforms,learnerswouldhave159beenforcedtoconcludethatdevoicingwasamorphophonologicalruleatbest.
Ontheotherhand,becausevowelcontractiononlyaffectedafewformsofthepasttenseinFinnish,learnerswerestillabletoperceivethephonologicalregularityofspirantizationinthatcategory.
Forthem,thecomplexitylayinthesmallnumberoflistedexceptionstospirantization,whichtheynaturallyeliminatedinlatergenerations.
EastCentralGermanapocopeappearstobemoreakintoFinnishspirantization.
TheMiddleGermanapocopetargetingunstresseddisyllableslikeundeandabemustindeedhaveresultedinword-finalvoicedobstruents.
Yet,sincefinaldevoicingwouldstillhavebeenoverwhelminglyregularandpredictable,itisnothardtounderstandwhydevoicingthenappliedtotheseoutputs.
Yiddishopalsoatteststothis;apocopehadappliedtothisformsignificantlyearlierthantootherforms,sothatdevoicingremainedproductiveafterabe>ab,butnotaftergabe>gab(seee.
g.
King1980,Weinreich1980).
ThatdevoicingbecametemporarilyopaqueinEastCentralGermanisprovenbytwothings:the15thcenturyUpperSaxontextweexamined,whereapocopehadprogressedfarenoughtorenderdevoicingsignificantlyopaque(inotherEastCentralGermangroupoftexts,theThuringiantexts,apocopeneverprogressedfarenough).
TheotherpieceofevidenceforopacityinEarlyModernEastCentralGermandevoicingisthepeculiarrestorationofschwaaftervoicedobstruentsindisyllablesthatweobservedearlier,sothatwehavespatfromspte,buttrgefromtrge.
Aswesaw,unlesswewishtoappealtoteleology,ormakeuseofOT(whichwesawdidnotsucceedverywellindescribingtheYiddishdevelopments),aplausible160explanationisthatapocopedidapply,butthatalaterchangerestoredfinalschwainsuchforms.
Thiscouldbeanindependentsoundchange:(59)→/[+obst,+vce]]Wd___Analternativeexplanationisthatthepatternweseearosefromacontact-inducedchange,notaregularsoundchange.
Consideringthattheapocopatingdialectwouldbeincontactwithnon-apocopatingdialects(e.
g.
Silesian),wemightbelookingatsomekindofL2phenomenon,wherethepresenceoffinalvoicedobstruentsintheapocopatingdialectwasinterpretedasapoorlyarticulatedfinalschwa,i.
e.
aperceptionerror.
Alternatively,speakersofanon-apocopating,finaldevoicingdialectwouldattempttoacquireanapocopatingdialect,togetherwithfinalvoicedobstruents,butwouldbeunabletoproducefinalvoicedobstruentsintheiroutput,owingtoL1interference,andsotheywouldautomaticallyaddafinalschwa,whichwouldhavebeenfreelyavailableintheirnon-apocopatingdialectanyway.
Thismakeshistoricalsense,becausetheoldnormwasEastFranconian,i.
e.
anapocopatingdialectthathadlostdevoicing.
ThiswouldbethedialectthatspeakersinEastCentralGermanywouldattempttoacquireupuntiltheReformation.
TheseEastCentralGermanspeakerswouldhavespoken,astheirnormativeregister,aconservativevarietywithfinaldevoicingandnoapocope78.
AsapocopetookplaceinUpperGerman,78WediscussedearlierthattherewassomeevidencethatinnerGermanlenitionhadbeguntoaffectEastCentralGermanalreadyintheMiddleGermanperiod.
Ontheotherhand,itisnotdoubtedthatfinaldevoicingwasattestedlongerinEastCentralGermanthaninUpperGerman,rightupuntiltheperiodofstandardizationinthe16thcentury.
WeshouldprobablyconcludefromthisthatthereweretworegistersinEarlyModernEastCentralGerman:anupperregisterwithavoicingcontrastandfinaldevoicing;andalowerregisterwithnocontrastandlenition.
Neitherregisterwouldhaveundergonefullapocope.
The161renderingdevoicingopaque,theEastCentralGermanspeakerswhoacquiredsuchadialectwouldmostlikelyrenderthenewfinalvoicedobstruentswithfinalschwa.
Theywouldhavenotroubleacquiringoutputsofapocopethatendedinvoicelessobstruentsorsonorants,however.
ThatleavesustoaccountfortheoutputsofBehaghel'sLaw,i.
e.
theregularapocopeaftersecondarilystressedsyllables,wherethe[voice]specificationoftheprecedingsegmentdidnotmatter(e.
g.
allerhande>allerhand),andthoselexicalexceptionstotherulethatprecedingvoicedobstruentsblockedapocopeafterprimarystress,e.
g.
milde>mil[t],schrge>schr[k].
Wecanmakesenseofthispatternifweallowthatapocopeaftersecondarilystressedsyllablesoccurredbeforetheapocopeafterprimarystressedsyllables.
Theresultwouldbeonlyafewinstancesofnewfinalvoicedobstruents,e.
g.
adjectivesin–ig,whichwouldreadilyundergostill-productivedevoicing.
ThisanalysisiscorroboratedbytheobservationthatinYiddish,the–igsuffixdidinfactundergodevoicingas–ik.
Asforthehandfulofotherexceptionstoblockingbyprecedingvoicedobstruents,thesecanbeexplainedasthehaphazardapplicationofdevoicingbyEastCentralGermanspeakerstoapocopatedformsinUpperGerman.
I.
e.
whenanEastCentralGermanspeakerwasconfrontedwithformslikeschr[g]andtr[g],thereweretwooptions:eitherattempttoarticulatethefinalvoicedobstruent,triggeringautomaticadditionoffinalschwa,orelsetreatthefinalvoicedobstruentasan"error"andapplyregularfinaldevoicing,whichwastheoptionchosenforschrg.
evidenceforlenitioninEarlyModernEastCentralGermanwouldbeduetointerferenceoflocaldialectpronunciationinthatcase.
1625.
3ConclusionsWehavefoundthatthedatasupportingthere-affirmationoffinaldevoicinginEastCentralGermanisquitestrong,unliketheevidenceforinsertionofapocopebeforedevoicing,whichwasweak.
Sinceapocopewasnotanunconditionedchange,asitwasinUpperGerman,therewouldhavebeensignificantlyfewersurfaceexceptionstodevoicingasaresultofapocope,meaningthattherewouldhavebeenconsequentlythatmuchmoreevidenceforproductive,phonologicaldevoicingavailabletothelearner,evenafterapocopeoccurred.
However,wefoundthattherewasalsogoodevidenceforcontacteffectsinEastCentralGermanapocopeanddevoicing,especiallyintheblockingofapocopeaftervoicedobstruentsinwordsliketrge.
RatherthantheUpperGerman-typescenario,whereapocoperendereddevoicingopaque,inEastCentralGermanitappearsthatonlyapocopeaftersecondarilystressedorunstressedsyllableswasaregularsoundchange,whileapocopeafterprimarilystressedsyllablesismoreclearlyacontact-inducedchange.
ThisraisesthequestionofwhethersecondfinaldevoicinginEastCentralornorthernHighGermanwasitselftheresultofcontact,ratherthaninternaldevelopments.
6SecondfinaldevoicingasanindependentchangeinModernGermanThediscussionsofarhasassumedthatfinaldevoicingwaseitherneveropaqueinEastCentralGerman,orelsehadbecomeopaquebutnotunproductive,laterre-affirmingitselfinthelocaldialect.
Wehavedeterminedthatthelatteristhemostprobableaccountofevents.
However,wewillbrieflyconsidersomealternativehypothesesoftheoriginsof163ModernGermanfinaldevoicing,whichassumethatfinaldevoicingwasindeedcompletelylostfromEastCentralGerman,andthatthelossofdevoicingmostlikelyreflectsthatchange.
Inthatcase,finaldevoicinginModernGermanmusthavecompletelyindependentorigins,whichcouldincludeeitheranindependentsoundchange,orelseinterferencefromLowGermaninthenewnortherncentersofnormativeHighGerman.
6.
1SecondfinaldevoicingassoundchangeAnotherwaytoaccountforsecondfinaldevoicingisbypositingasoundchange,alongthelinesofMiddleGermanfinaldevoicing,andalongthelinesofGoossens'reconstructionofEarlyModernDutchseconddevoicing.
Thatis,afterapocopehadoccurredinEarlyModernGerman,anewsoundchangeorchangesofsimilaroridenticalstructuraldescriptiontotheearlyMiddleGermandevoicing(s)tookplace,withtheresultthatModernGermanonceagainhasfinaldevoicing,justasitsMiddleGermanancestorhad.
Thetroublewiththisisthatapocopeappearstohavebeenalongdrawnoutprocess,withdifferentcategoriesundergoingthechangeatdifferentstages.
Wemayneedtopositmorethanonedevoicingoccurringaftereachstageofapocopetakesplace:forinstance,onceafterthechangedeletingschwainunstresseddisyllables,andthenonceagainafterBehaghel'sLawapplied.
Sincefinaldevoicingisaphoneticallynaturalchange,however,itmaynotbesoimplausibletopositthiskindofscenario.
164Theproblembecomesmoreacute,however,whenwearedealingwithanalogicalextensionofapocope.
WhenthefinalschwaofdativesingularTageislost,forexample,itishardtoarguethisispartofanyregularsoundchange,sinceapocopeisaffectingonemorphologicalcategoryonly.
Moreover,asWrightnotes,apocopeisvariable,andyetdativesingularTagisapparentlyalwayspronouncedwithafinalvoiceless[k].
Ifsecondfinaldevoicingisduetosoundchangeonly,itappearswehavetopositanewsoundchangeeverytimeoneinstanceofTageisreplacedbyTaginthelanguage.
6.
2LowGermanoriginsofModernGermandevoicingAnotherpossibilityisthatthedevoicingruleofModernGermancouldhavearisenoutofcontactwithLowGermanintheHighGerman-speakingnortherncitiesofBerlin,Hannover,Hamburgandsoon.
Upuntilthelate19thcentury,theMeissnischstandard,whichbynowwasleniting,wasthenormthroughoutGermany,owingtotheprestigeofGoetheandothers.
Butthen,aspoliticaldevelopments(i.
e.
thegrowingdominanceofPrussiaandtheunificationofGermany)movedtheculturalcenternorth,toBerlin,thenormativespeechofthelatterbecamethenewstandardofthecountry.
Andthisnormhadfinaldevoicing.
TheuseoffinaldevoicingintheBerlindialectmayhavetwosources.
OneisthatdevoicingmayhavebeenretainedinthenorthforthesamereasonitwasretainedinSaxony,andwhywefindevidenceforitasearlyasClajus'16thcenturygrammar:apocopeneverwenttocompletion.
Evenifdevoicingwasmomentarilyrenderedopaque,thefactthatapocopeneverwentbeyondthevariablestagemeansthattherewouldhave165endedupbeingmanyfewerexceptionstoapocopethaninUpperGerman,withtheconsequencethatre-assertionofdevoicingwouldhavebeenthatmuchmorelikely.
Atthispoint,weneedtobrieflydescribetheoriginofHighGermanspeechinnortherncities.
Asweknow,thenorthofGermanywasformerlyLowGermanspeaking.
Startinginthelate14thcentury,however,LowGermanwasreplacedbyHighGermaninallthemajornortherncities.
TheparticulardialectsourceofnorthernHighGerman,moreover,wastheEastCentralGermanofSaxonyandThuringia,i.
e.
non-apocopatingorpartiallyapocopatingdialects.
Forthisreason,wecanimaginethattheHighGermanofthenorthfollowedthesamepatternsasthoseofEastCentralGerman.
Thedifferenceisthat,whilelenitionaffectedEastCentralGermaneventually,itdidnotspreadtonorthernHighGerman,withtheresultthatwhenthenormativephonologyshiftedtoBerlinandHamburg,thelenitingphonologywasreplacedbyadevoicingphonology.
Oneoftwothingsmayhavehappened:nativeLowGermanspeakers,afteracquiringHighGermanasasecondlanguageintheEarlyModernGermanperiod,appliedtheirnativedevoicingruletothenon-devoicingEarlyModernGermannorm,throughtheprocessknownas"interference"(influenceofL1onimperfectlyacquiredL2).
InasituationwhereenoughL2speakersdisplaysuchinterference,theinterferingphonologicalpatterncanbeacquirednativelybythenextgeneration.
TheotherthingthatmayhavecausednorthernHighGermantogaindevoicingisspreadofthedevoicingrulefromneighboringLowGermanspeakingareastotheHighGermancities.
ThisrequiresacertaindegreeofcontactandmutualintelligibilitybetweenLowGermanandHighGermanspeakers,whichisnotimpossible,butperhapsless166likely,giventhatthesociolinguisticpressureseemstohavebeenprettyone-way,withHighGermanspreadingattheexpenseofLowGerman.
6.
3GeneralconclusionsWithrespecttoUpperGerman,wehavefoundfairlyrobustevidencethatapocoperendereddevoicingopaque,resultingintheeventuallossoftherule,ashappenedinYiddish(whichappearstobemostlydescendedfromEastUpperGermanitself;seeJacobs2005).
WithrespecttoEastCentralGerman,theregionfromwhichstandardModernGermanarose,wehavealsofoundevidencethatapocoperendereddevoicingopaque,althoughinthedatawehaverevieweditislargelyconfinedtooneUpperSaxontextofthelate15thcentury.
Wehavenotfoundgoodevidencethatapocopeautomaticallyfeddevoicing,however,sothatthedevoicingofthemodernlanguagemusthavesomeotherexplanationthansimplyretentionoftherulefromMiddleGerman.
DevoicingappearstobeattestedagaininEastCentralGermanalreadybythelate16thcentury,ontheevidenceofthegrammarofJohannesClajus.
Therearetwopossiblecausesforthisswiftre-establishmentoffinaldevoicing:there-affirmationoftheruleafteraperiodofopacity,alongthelinesofthere-affirmationofFinnisht-spirantizationaftervowelcontractionoccurred.
TheothercauseistheclosecontactthatEastCentralGermanwouldhavemaintainedwithneighboringdialects,bothHighGermanandLowGerman,wherefinaldevoicingwasneverlostandfromwherethedevoicingrulecouldhaveeasilyre-enteredthelanguage.
167WefoundthatfinaldevoicingappliedtotheoutputsofBehaghel'sLaw,i.
e.
apocopeafterasyllableofsecondarystress,ortheoutputsofMiddleGermanapocope,whichoccurredafterunstressedsyllable.
However,wealsofoundthatdevoicingappliedtotheoutputsofapocopeaftersyllablesofprimarystress,asinUpperGermanandYiddish,butonlywheretheprecedingsegmentwasavoicelessobstruentorasonorant.
ThisstrangepatternsuggeststhatapocopeafterprimarystressinEastCentralGermanwasadifferentprocessfromapocopeafterothersyllables,unlikeUpperGermanandYiddish.
Weconcludedthatthemostlikelyscenariowasthatapocopefirstappliedaftersyllablesotherthansecondarystressthroughregularsoundchange.
Thisresultedinsomeexceptionstofinaldevoicing,butthenumberofexceptionswassofewthatfinaldevoicingremainedproductive,andeventuallyre-appliedtotheoutputsofapocope.
ThischangewenttocompletioninallofUpper,WestCentralandEastCentralGerman.
Apocopeaftersyllablesofprimarystress,ontheotherhand,onlywenttocompletioninUpperGerman.
Theprecisesocialconditionsthatallowedthissituationtooccurarehardtorecover,butthefactthatapocopeafterprimarystressneverwentbeyondthevariablestageinEastCentralGermancannotbedoubtedafterconsideringthetextualevidence.
Undersuchconditions,ascenariocanbepostulatedwherebymanyspeakersofanon-apocopating,devoicingvarietywouldhaveimperfectlyacquiredanapocopatingvariety.
OwingtoL1interference,theywouldhavebeenunabletoarticulatethefinalvoicedobstruentsoftheformervariety,resultingintheretentionoffinalschwainthisenvironment.
168AlthoughtheClajusgrammarwitnessestothepresenceofdevoicinginhighregisterEastCentralGermaninthelate16thcentury,weknowthatbythe18thcenturyEastCentralGermanhadundergonelenitioninallregisters,withevenGoetheprescribingthispronunciationinhispoetry.
FinaldevoicingonlybecomesthemarkofhighregisterGermaninthe19thcentury,whenthecenteroflinguisticprestigemovedfromtheEastCentralregionofMeientothenortherncitiesofBerlinandHamburg.
Inthesecities,lenitionneveroccurred.
However,wewerenotcertainintheendwhetherthepresenceofdevoicinginNorthernHighGermanwasaholdoverfromthepre-lenitionEastCentralstandard,orwhetheritaroseindependently,perhapsduetoLowGermaninfluence.
WecautiouslycomedownonthesideofretentionfromolderEastCentralGerman,sinceindependentevidenceconfirmsthattheHighGermanvarietyofthenortherncitiesisclearlymoreofanEastCentralthanaWestCentralorUpperGermanvariety.
Moreover,thereisotherwisenotmuchevidenceforsignificantLowGermaninterferenceinNorthernHighGerman;theEarlyModernperiodapparentlysawfull-scalerapiddialectshiftinthenortherncities,ratherthanthesituationofextendedcontactthatwesawinEarlyModernEastCentralGerman,i.
e.
betweenUpperandCentralGermanvarieties.
SomeoftheclassicmarkersofNorthernHighGerman,suchasthefricativereflexofMiddleGerman/g/,e.
g.
/taχ/for'day',arealsocharacteristicofCentralGerman.
UnambiguouslyLowGermanfeatures,likeunshiftedinitial/p/or/t/,areentirelyabsent,exceptinanegligiblenumberofloanwords.
Tosumup,wehavebeenabletoadvancetheknowledgeofthehistoryofGermantoasmallextentbydrawingattentiontotherobustcorrelationbetweenapocopeand169absenceofdevoicinginEarlyModernGermantexts,whichraisesverystrongdoubtsaboutthe"standard"historyofthelanguage,whichholdsthelossofdevoicingtobemerelyachangeinspellingconvention.
Atthesametime,thisdiscoveryunderminesoneprimeexampleofruleinsertion,andhencethenotionthatphonologicalinnovationscanenterthelanguageatanyotherlevelthanthesurface.
Whilewedonotdenythatsubsequentrestructuringcanobscurethechronologyofchanges,thereappearstobegoodreasontotreatphonologicalchangeasawholeasakindof"life-cycle"ofrules.
Evidenceforruleinsertionawaitsmuchstrongerempiricalconfirmation.
170BibliographyAlbright,Adam.
2002.
Theidentificationofbasesinmorphologicalparadigms.
PhDdissertation,UCLA.
-.
2008.
"Inflectionalparadigmshavebases,too:evidencefromYiddish".
InA.
Bachrach&A.
Nevins(eds.
),Thebasesofinflectionalidentity.
Oxford:UniversityPress.
Aronoff,Mark.
1994.
Morphologybyitself.
Cambridge,MA:MITPress.
Bakovic,Eric.
2007.
"Arevisedtypologyofopaquegeneralizations".
Phonology24:217-259.
Bermúdez-Otero,Ricardo.
2006.
"Phonologicalchangeinoptimalitytheory".
InK.
Brown(ed.
),Encyclopediaoflanguageandlinguistics:volume9.
Oxford:Elsevier.
-.
2007.
"Diachronicphonology".
InP.
deLacy(ed.
),TheCambridgehandbookofphonology.
Cambridge:UniversityPress.
Bermúdez-Otero,Ricardo&Hogg,RichardM.
2003.
"Theactuationprobleminoptimalitytheory".
InD.
E.
Holt(ed.
),Optimalitytheoryandlanguagechange.
Dordrecht:Kluwer.
Blevins,Juliette.
2004.
Evolutionaryphonology.
Cambridge:UniversityPress.
Brame,MichaelK.
1972.
"Ontheabstractnessofphonology:Maltese".
InM.
Brame(ed.
),Contributionstogenerativephonology.
Austin:UniversityofTexasPress.
Brockhaus,Wiebke.
1995.
FinaldevoicinginthephonologyofGerman.
Tübingen:MaxNiemeyer.
171Bruch,Robert.
1954.
GrundlegungeinerGeschichtedesLuxemburgischen:BandI.
Luxembourg:PublicationsscientifiquesetlittérairesduMinistèredel'Educationnationale.
Elmentaler,Michael.
2003.
StrukturundWandelvormodernerSchreibsprachen.
Berlin:WalterdeGruyter.
Ewald,Petra.
1997.
"ZurAusprgungdesmorphemidentifizierendenPrinzipsinfrühneuhochdeutschenDrucken".
InG.
Schmirber(ed.
),SpracheimGesprch.
Munich:Hanns-Seidel.
Féry,Caroline.
1999.
"FinaldevoicingandthestratificationofthelexiconinGerman".
InvandeWeijeretal.
(eds.
),Thephonologicalspectrum:segmentalstructure.
Philadelphia:JohnBenjamins.
Fourakis,M.
&Iverson,G.
K.
1984.
"Onthe'incompleteneutralization'ofGermanfinalobstruents".
Phonetica41:140-149.
Fruehwald,Josef,Gress-Wright,Jonathan&Wallenberg,Joel.
Forthcoming.
"Phonologicalrulechange:theconstantrateeffect".
InProceedingsofNELS40.
Cambridge,MA:MITPress.
Goossens,Jan.
1974.
HistorischePhonologiedesNiederlndischen.
Tübingen:MaxNiemeyer.
-.
1977.
"Detweedenederlandseauslautverscherping".
Tijdschriftvoornederlandsetaal-enletterkunde93:3-23.
Halle,Morris.
1962.
"Phonologyingenerativegrammar".
Word18:54-72.
172Hayes,Bruce.
1999.
"Phonetically-drivenphonology:theroleofoptimalitytheoryandinductivegrounding".
InFunctionalismandformalisminlinguistics:volumeI.
Amsterdam:JohnBenjamins.
Hock,HansHenrich.
1991.
Principlesofhistoricallinguistics.
Berlin:MoutondeGruyter.
Hyman,Larry.
1976.
"Phonologization".
InA.
Juillandetal.
(eds.
),LinguisticstudiesofferedtoJosephGreenbergontheoccasionofhis60thbirthday.
Saratoga,NY:AnmaLibri.
Iverson,Gregory&Salmons,Joseph.
2007.
"DomainsanddirectionalityintheevolutionofGermanfinalfortition".
Phonology24:1-25.
Jacobs,Neil.
2005.
Yiddish:alinguisticintroduction.
Cambridge:UniversityPress.
Jasanoff,Jay.
2004.
"Plusachange…Lachmann'sLawinLatin".
InJ.
W.
Penney(ed.
),Indo-Europeanperspectives:studiesinhonorofAnnaMorpurgoDavies.
Oxford:UniversityPress.
Jessen,Michael.
1998.
PhoneticsandphonologyoftenseandlaxobstruentsinGerman.
Philadelphia:JohnBenjamins.
Kager,René.
1999.
Optimalitytheory.
Cambridge:UniversityPress.
Keller,RudolfErnst.
1961.
Germandialects.
Manchester:UniversityPress.
-.
1978.
TheGermanlanguage.
AtlanticHighlands,NJ:HumanitiesPress.
King,Robert.
1969.
Historicallinguisticsandgenerativegrammar.
EnglewoodCliffs,NJ:Prentice-Hall.
-.
1973.
"Ruleinsertion".
Language49:551-578.
173-.
1976.
"Indefenseofextrinsicordering".
InA.
Koutsoudas(ed.
),Theapplicationandorderingofgrammaticalrules.
TheHague:Mouton.
-.
1980.
"ThehistoryoffinaldevoicinginYiddish".
InM.
Herzogetal.
(eds.
),ThefieldofYiddish:fourthcollection.
Philadelphia:InstitutefortheStudyofHumanIssues.
Kiparsky,Paul.
1965.
Phonologicalchange.
PhDdissertation,MIT.
-.
1968.
"Linguisticuniversalsandlinguisticchange".
InE.
Bach&R.
Harms(eds.
),Universalsinlinguistictheory(Bach,E.
&Harms,R.
edd.
).
NewYork:Holt.
-.
1971.
"Historicallinguistics".
InW.
O.
Dingwall(ed.
),Asurveyoflinguisticscience.
CollegePark,MD:UniversityofMarylandPress.
-.
1973.
"Productivityinphonology".
InM.
Kenstowicz&C.
Kisseberth(eds.
),Issuesinphonologicaltheory.
TheHague:Mouton.
-.
1978.
"Analogicalchangeasaproblemforlinguistictheory".
InB.
Kachru(ed.
),Linguisticsintheseventies.
Urbana,IL:UniversityofIllinoisPress.
-.
1982.
"Lexicalphonologyandmorphology".
InI.
-S.
Yang(ed.
),Linguisticsinthemorningcalm.
Seoul.
-.
1989.
"Phonologicalchange".
InF.
Newmeyer(ed.
),Linguistics:theCambridgesurvey,Volume1.
Cambridge:UniversityPress.
Kloeke,WusvanLessen.
1982.
DeutschePhonologieundMorphologie.
Tübingen:MaxNiemeyer.
Kluge,Friedrich.
1975.
EtymologischesWrterbuchderdeutschenSprache.
Berlin:WalterdeGruyter.
174Knig,Werner.
1978.
DTV-AtlaszurdeutschenSprache.
Munich:DeutscherTaschenbuchVerlag.
Koutsoudas,Andreas,Sanders,G.
&Noll,C.
1974.
"Theapplicationofphonologicalrules".
Language50:1-28.
Kranzmayer,Eberhard.
1956.
HistorischeLautgeographiedesgesamtbairischenDialektraumes.
Vienna:HermannBhlaus.
Krishnamurti,Bhadriraju.
1978.
"Arealandlexicaldiffusionofsoundchange".
Language54:1-20.
Kroch,Anthony.
1989.
"Reflexesofgrammarinpatternsoflanguagechange".
Languagevariationandchange1:199-244.
Kurylowicz,Jerzy.
1949.
"Lanaturedesprocèsdits'analogiques'".
Actalinguistica5:121-138.
Labov,William.
1994.
Principlesoflinguisticchange:internalfactors.
Oxford:Blackwell.
-.
2001.
Principlesoflinguisticchange:socialfactors.
Oxford:Blackwell.
Labov,William,Yaeger,Malcah&Steiner,Richard.
1972.
Aquantitativestudyofsoundchangeinprogress.
Philadelphia:U.
S.
RegionalSurvey.
Lindgren,Kaj.
1953.
DieApokopedesmittelhochdeutschen–einseinenverschiedenenFunktionen.
Helsinki:SuomalainenTiedeakatemia.
Mascaró,Joan&Wetzels,Leo.
2001.
"Thetypologyofvoicinganddevoicing".
Language77:207-244.
McMahon,April.
2000a.
Change,chanceandoptimality.
Oxford:UniversityPress.
175-.
2000b.
LexicalphonologyandthehistoryofEnglish.
Cambridge:UniversityPress.
Messow,Theodor.
1965.
"Zurdialektgeographiedesschlesisch-brandenburgischenGrenzgebietesbeiZüllichau".
InDeutscheDialektgeographie56.
Marburg:N.
G.
Elwert.
Mihm,Arend.
2004.
"ZurGeschichtederAuslautverhrtungundihrerErforschung".
Sprachwissenschaft29:133-206.
Mohanan,K.
P.
1986.
Thetheoryoflexicalphonology.
Dordrecht:Reidel.
Mondon,Jean-Franois.
2009.
Thenatureofhomophonyanditseffectsondiachronyandsynchrony.
PhDdissertation,UPenn.
Paul,Hermann.
2007.
MittelhochdeutscheGrammatik.
Tübingen:MaxNiemeyer.
Port,Robert&O'Dell,M.
1985.
"Neutralizationofsyllable-finalvoicinginGerman".
Journalofphonetics13:455-471.
Reichmann,Oskar&Wegera,Klaus-Peter.
1993.
FrühneuhochdeutscheGrammatik.
Tübingen:MaxNiemeyer.
Ringe,Donald.
2006.
FromProto-Indo-EuropeantoProto-Germanic:alinguistichistoryofEnglish,volumeI.
Oxford:UniversityPress.
Rix,Helmutetal.
2001.
LexikonderindogermanischenVerben.
Wiesbaden:LudwigReichert.
Rogers,Henry.
2005.
Writingsystems:alinguisticapproach.
Oxford:Blackwell.
Rubach,Jerzy.
1990.
"FinaldevoicingandcyclicsyllabificationinGerman".
LinguisticInquiry21:79-94.
Russ,CharlesV.
1982.
"Thegrammaticalconditioningofsoundchanges".
Leuvense176bijdragen71:141-150.
Sadock,Jerrold.
1973.
"Word-finaldevoicinginthedevelopmentofYiddish".
InB.
Kachruetal.
(eds.
),Issuesinlinguistics:papersinhonorofHenryandRenéeKahane.
Urbana,IL:UniversityofIllinoisPress.
Sapir,Edward.
1915.
"NotesonJudeo-Germanphonology".
TheJewishquarterlyreview6:231-266.
Siebs,Theodor.
1957.
DeutscheHochsprache.
Berlin:WalterdeGruyter.
Sihler,Andrew.
1977.
"MorphologicallyconditionedsoundchangeandOEparticiplesin–en".
Generallinguistics17:76-97.
Stampe,David.
1979.
Adissertationonnaturalphonology.
NewYork:Garland.
Strunk,Klaus.
1976.
LachmannsRegelfürdasLateinische.
Gttingen.
Tesar,Bruce&Smolensky,Paul.
2000.
Learnabilityinoptimalitytheory.
Cambridge,MA:MITPress.
vanBree,Cor.
2003.
"AuslautverscherpingbijdeoudegrammaticienindedialectenenhetFries".
InE.
Ruijsendaaletal.
(eds.
),BonjoursNeef,ghoedendaghCozyn.
Münster:Nodus.
vanHaeringen,CoenraadB.
1937.
"Opmerkingenbijdeapocopevan–e".
Denieuwetaalgids31:241-250,322-333.
vanLoey,Adolph.
1966.
MiddelnederlandsspraakkunstI:Vormleer.
Groningen:Wolters-Noordhoff.
Vaux,Bert.
2008.
"Whythephonologicalcomponentmustbeserialandrule-based".
InB.
Vaux&A.
Nevins(eds.
),Rules,constraintsandphonologicalphenomena.
Oxford:UniversityPress.
177Vennemann,Theo.
1972.
"Onthetheoryofsyllabicphonology".
LinguistischeBerichte18:1-18.
vonUnwerth,Wolf.
1908.
DieschlesischeMundart.
Breslau:M.
&H.
Marcus.
Weinreich,Max.
1980.
HistoryoftheYiddishlanguage:VolumeII.
NewHaven:YaleUniversityPress.
Weinreich,Uriel,Labov,William&Herzog,Marvin.
1968.
"Empiricalfoundationsforatheoryoflanguagechange".
InW.
Lehmann&Y.
Malkiel(eds.
),Directionsforhistoricallinguistics.
Austin:UniversityofTexasPress.
Wexler,Paul.
2002.
Two-tieredrelexificationinYiddish.
Berlin:MoutondeGruyter.
Wiese,Richard.
1996.
PhonologyofGerman.
Oxford:UniversityPress.
Wright,Joseph.
1907.
HistoricalGermangrammar.
Oxford:UniversityPress.
Yang,Charles.
2008.
"Threefactorsinlanguagevariation".
Ms.
,toappearinLingua.