WPC $.=~C9ycÙuLXw,#+v.BM-#6S08%HD Ѕuv({9)*,ް:F{hUxnRAƛ*Q<ɑվ oGjZKA@r"޳cRZzeGјٓ|[Xo76 *j,8ܵL{ \1]#m2B#`~N1 -]F #Ȱ)LY%Wp?~[aFvz2x+3/N0 6?j#jy%y=.>Q҆bNAY7ّFS\_J{`gӆ0t'>R@A,U A0fF(Hp-UN!UN"pUNaf  BUN+yyUN 0D@!!U$a&&UN)B*UNX* B2* 0g**+ A-UNZ. A.UNQ//UN2UN2fA3 BC3UN`33UN6 B9*6UN7UNf7 0'x7UN,8zz8 0%D98:UNN:UN::a;UN;UN"<UNp< 1m< 0c+==f=UN=f=P= B5FAf{Aa}AAAaBCaVCjC 0PI D+I hJKUNL;M h&S 0KSUNTUNdTTw@W4W W RY ChYhY]UN] hcG^UN` hS` Kb Qblo;w h:y` #{UNeхUN҅UN fnapfa hfpar h ha hK`D0Ft\АUNUN43IUNUNTUNUNS&4ZUN$UNrrrUNfUNtUN°UNUN^^^UNUNUNHUNUNUN2UNUNγUNUNjUNUNUNTUNUNUN>UNUNڶUN(UNvvUNķUNUN`UNUNUNJJJJJUN 0PUN66UNUNUN5UNUNѻUNUNmUNUN UNWUNUNUNAUNUNݾUN+++++++++UNyyyyyyyyyyyUNǿUNUNcccccUNUNUNMUNUNUN7UNUNUN!UNoUNUN UNYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY&Lexmark Optra PS2LEXPS ,, 06R Z 6Times New Roman RegularX( U$  ӁLevel 1Level 2Level 3Level 4Level 5&d9 Z 6Times New Roman Regular*'Z&IPAKiel RegularQXR I&mage <=8C HKKKK  H   6$      24    ݀uXXuuThemotivationforMorphologicalOpacity(MO)isthataftertheapplicationoftheContinuativeAssociation  RulethereareonlythreeavailableCpositionsinthetemplateCgCVCbutfiveunassociatedconsonantsunderlinedinthemelodicsequencesl<gsl<g.MOthusexcludespossibleoutputslikesgs<g,where/l/remainsunassociated. ` NoticethatMOisaconstraintontheoutputofthederivation,andessentiallyapredecessorofMaxIO,oneofthe $t undominatedconstraintsoftheOTanalysisproposedearlier.#uup#    ##Xd#(9 Z 6Times New Roman Regular  6$      3    ݀uXXForthemotivationofrevisingtheoriginalOTParse/Fillconceptionoffaithfulness,seeMcCarthy&Prince  (1994b)andMcCarthy(1995).(9 Z 6Times New Roman Regular  H      6$      4    ݀uXXItuuԀistobekeptinmindthatanindependentsetofthesameconstraintsholdsfortheInput/Output  correspondencerelation,namely,#uuq#uuMaxIO#uuJ#uuԀandDepIO#uu#uu.#uu#ԀForextensionsofcorrespondencetheorytofaithfulnessrelations  betweenoutputformsseeBenua(1995),Flemming&Kenstowicz(1995),andMcCarthy(1995)(cf.Burzio1994).6R Z 6Times New Roman RegularTABLE D d6R Z 6Times New Roman Regular o 6$      18    ݀uXXIassume,ofcourse,asuperordinateconstraintthatrequiresemptyRootnodesintheinputtobefilledwith  segmentalmaterial,technicallyFillSegmentorFillPlaceofPrince&Smolensky(1993),section9.1.2.6R Z 6Times New Roman Regular($$  0    6$      4    ݀#bXXIt*bbԀistobekeptinmindthatanindependentsetofthesameconstraintsholdsfortheInput/Output  correspondencerelation,namely,##bbq#*bbMaxIO##bbJ#*bbԀandDepIO##bb#*bb.##bb#ԀForextensionsofcorrespondencetheorytofaithfulnessrelations  betweenoutputformsseeBenua(1995),Flemming&Kenstowicz(1995)andMcCarthy(1995)(cf.Burzio1994). K 6$      6    ݀uXXSeeuuDiffloth(1976a,b)forbriefdescriptionsofJahHutandSemairespectively.#uut#ԀItisclearfromthese  descriptionsthatthemorphologiesofJahHutandSemaiareverysimilartothatofTemiar.Finally,NicoleKruspeattheUniversityofMelbourne,currentlyinvolvedinfieldworkontheSouthAslianlanguageSemelai,informsmeabouttheclosesimilaritiesofthislanguagetoTemiar.TABLE D  6$      10    uXXuuԀTheallomorphyoftheaffix/tr/intheperfectivesisprosodicallydriven.#uuo#IassumeinGafos(1996)thatthe  perfectiveoutputsaresubjecttoaconstraintthatrequiresthemtobenomorethantwosyllables.Tosatisfythisconstraint,partofthecausativeaffix/tr/maybeleftunparsed.SimilarinstancesofallomorphyarefoundoutsidetheAslianbranchofMonKhmer,asforexample,inKammucausativeandnominalaffixation(Svantesson1983).Theoverarchinggeneralizationinallthesecasesisthataffixallomorphyispreferredtobaseallomorphy.6R Z 6Times New Roman Regular   6R Z 6Times New Roman Regular d((3c$ U!     0  .  6$      21    ݀XUXXXuXXUComparethiswithMakassarese(McCarthy&Prince1994),whichhasthesameconstraintonplace  assimilationofnasals,but#uuQ#uuԀthereduplicatedform#uu@#uuԀbulambuladmonths(wheremcorrespondstod)showsthatthe  inverserankingisinvolved,i.e.NC>>Ident(PL).#XUXup##XUXXXU#QR&OLE Box<<=8C6R Z 6Times New Roman RegularQXR&OLE 2.0 Box <=8C HKKKK&d9 Z 6Times New Roman Regular  6$      4    ݀,bXXIt5bbԀistobekeptinmindthatanindependentsetofthesameconstraintsholdsfortheInput/Output  correspondencerelation,namely,#,bbq#5bbMaxIO#,bbJ#5bbԀandDepIO#,bb#5bb.#,bb#ԀForextensionsofcorrespondencetheorytofaithfulnessrelations  betweenoutputformsseeBenua(1995),Flemming&Kenstowicz(1995),andMcCarthy(1995)(cf.Burzio1994).6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular d6R Z 6Times New Roman Regular  6$      14    ݀XUXXXuXXUComparethistoaparametrictheoryofinviolableconditionsandrepairstrategies,whereconstraintsareeither   onor offforthewholelanguage(Paradis1988).Insuchatheorythissituationcannotbecoherentlycharacterized.Ons,forexample,is onforeverysyllableinTemiar,i.e.itisneverviolated.Otherconstraints,however,like1v#uuQ#uuԀmay ` beviolatedundercertainconditions.#uuE##XUXup#6R Z 6Times New Roman Regular&0 d d6R Z 6Times New Roman Regular&d9 Z 6Times New Roman Regular(((3$ U!       0  (9 Z 6Times New Roman Regular p 6$      5    ݀XUXXXuXXUSeeJenner(1969)onKhmer,#uuQ#uuԀSvantesson(1983)onKammu,andLombardi(1991)onKambodian.#XUXup##XUXXXU#(#$  0     6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular 6$      9    ݀uXXIntheprestressedposition,thevowelinventoryofItalianreducesfurthertothreevowels.TABLE D6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular6R Z 6Times New Roman RegularC#ce37=CIQYag1.a.i.(1)(a)(i)1)a)(;3$2ce  0  .3  0    H 6R Z 6Times New Roman Regular F 6$      13    ݀uXXuuThesamerankingaccountsforthefactthatdespitetherequirementof1v,inthelexiconofTemiarandother  Aslianlanguages,onefindsbisyllabicwordswithphonologicallyspecifiedpenultimatevowels(JahHut:talonpython;Semai:tak<<yalizard;Temiar:halabtogodownriver,Diffloth1976a).Theseprefinalvowelsareunderlyingly ` presentbecausetheirqualitiesareunpredictable,andhencetherankingMaxIO>>1vensuresthattheywillappearon $t thesurface.SuchwordsarenotfoundintherestofMonKhmer.#uup# 8 1Large Dashed dTABLE E  H   6$      32    ݀uXXComparethistothecaseofTemiarwhereSroleisrankedhigherthanuuAnchor#uu#uuing#uu#uu.#uuN#TABLE ITABLE I  6$      35    ݀XUXXXuXXUTheanalysispresentedhereissimilartothatofBatEl(1994)inthisrespect.Itdiffers,however,fromitin  onesignificantrespect.HerruleofMelodyOverwritting,apeculiaraspectSemiticmorphology,isshownheretobetheeffectofasimpleconstraintrankingMaxAFFIXIO#uuQ#uuԀ>>MaxBASEIO#uu#uu,requiringthatthesegmentalcontentofanaffixal ` formativebeexpressedfullyintheoutputtotheexpenseofthefullexpressionofbasesegmentalmaterial,duetothebisyllabicconstrainton#uuP#uutheoutput.ThissameideaextendsstraightforwardlytoothercasesofArabicmorphologyaswell. 8 Forexample,thepassiveofkatabwrote,kutibisderivedbyaffixationof/ui/under#uul#uuԀtherankingMaxAFFIXIO#uu#uuԀ>>  MaxBASEIO#uu#uu,forcingreplacementofthevocalismofthebasebythevocalismoftheaffix.#uuX##XUXup#((3$ U!   'dxd ?   ?   ?  ? TABLE D   6$      2    ݀uXXThispapermakesnoclaimsabouttherepresentationoftruegeminateconsonants,generallyassumedto  involvedoublelinkingbetweentwoskeletonadjacentpositions.ItisonlylongdistancegeminatesthatIargueshouldnotberepresentedasdoublylinkedstructures.SeeIt=&Mester(1993)onthestatusoftruegeminatesinOT.6R Z 6Times New Roman Regular  6$      25    ݀uXXAnalternativemodeofassociationintheTAmodeldoesnothelpeither.Yip(1988)hasproposedanedgein  linkingmode,inwhichthesegmentsattheedgesofacopiedmelodyarefirstlinkedtothepositionsattheedgesofatemplate.Oneofthe#XUXuQ#uXXUcontinuativepatterns, uu#XUXu#uXXUc1 #uu#F9u#uF9h#uu c2 .c1vc2#uu#,suggests#XUXu#uXXUԀedgeinlinking,assumingthatthetemplateissomesort ` ofreduplicativeheavysyllable.However,edgeinlinkingdoesnothelpin#XUXu9#uXXUuuc1#uu4#F9u#uF9~#uu c3 .c2vc3#uu#,w#XUXu#uXXUhereonlyoneconsonantgets &v copied.#XUXuH#uXXU?@?TABLE D( $ Figure  1  6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular  6$      28    ݀uXXPlacefeaturesofconsonants,ithasbeenargued,spreadovervowelsinthecaseofconsonantharmony  systems,wheretheharmonizingfeaturesareusuallythoseclassifiedunderthecoronalplaceofarticulation(Shaw1991;butseefootnote30).Thisdoesnotaffecttheargumentinthetext.ConsonantharmonydoesnotrequireV/Cplanarsegregationbutinsteademploystiersegregation,wheretheRootsofvowelsandconsonantsareonthesameplanebuttheirfeaturesmaylieondifferenttiers.Incontrast,LDCspreadingrequiresV/Cplanarpegregation,because spreadingtargetsthewholesegment. E  bb|BLE AB  6$      20    ݀uXXExpressivesinTemiarandotherAslianlanguagesemploysimilartypesofaffixationandcopyingasfound  intheverbalmorphologyofthelanguage.compare !dDi 'dd ^ 6$      33    ݀uXXuuAnaccountinsimilartermscanbegivenforthealternativepattern,c1vc2. c1 v c2 ,whichaccordingtoBatEl  (1994)isselectedarbitrarilybysomebiconsonantalverbsofModernHebrew.ThispatternisalsofoundinthePilpelandHitpalpelofClassicalHebrew,andforalimitednumberofverbsinArabic(e.g.zalzaltoshake),asdiscussedinsection5.Iassumethatverbswhichfollowthispatternconformtoabisyllabicprosodictargetwhichisfurtherspecified $t  tobeasequenceoftwoheavysyllables(seeBatEl1994foranalternativeanalysis).Allotherconstraintsareasintheanalysisgivenintext.Thegreatercapacityoftheprosodictargetallows,inthiscase,for#uup#uuԀbothconsonantsofthebase t togetcopied.#uu##XUXuQ#3|x'W Z 6Times New Roman Regular..?.....?....?....?..?..?...?.@..@.TABLE D.(9 Z 6Times New Roman RegularEE?EEEE?EE?EE?EEE?ETABLE IEHL 6$      34    ݀uXXԛuuAsiswellknown,Arabichasanabsoluteprohibitionagainstrootsbeginningwithtwoidenticalinitial  consonants,asin*sasam,althoughitappearstoallowrootsendingwithtwoidenticalconsonants,asinsamam  (Greenberg1950,McCarthy1979,1986,1988).ThisdistributionisexplainedinMcCarthy(1979)onthebasisoftwoassumptions.#uuq#uuFirst,underlyingformsaresubjecttotheOCP,prohibitingrootswithtwoadjacentidenticalconsonants $t (*ssm,*smm).Hence,theunderlyingformofsamammustbesm.Second,whenmappingthisbiconsonantalroottoa 8 triconsonantalCVCVCtemplate,aruleofrightwardLDCspreadingspreadsthefinalconsonanttogivesamam.Clearly,  thisanalysisisnotavailableinOTbecauseconstraintsapplyingstrictlyonunderlyingformsdonotexist.Theanalysispresentedhere#uu>#Ԁoffersanalternativeexplanationfortheskeweddistributionofthesocalledgeminateroots,whileatthe 4  sametimeavoidingtheuseofthespecialphonologicalmechanismsofLDCspreadingandV/Csegregation.#XUXuQ#uXXU  AnotherimplicationfortheOCPisrelatedtothenotionoflocality.Pierrehumbert(1992)hasshownthatthe    cooccurrencerestrictionsbetweenhomorganicconsonantsinSemiticcannotbeadequatelydescribedbyanOCPthatstrictlyreferstogeometricallyadjacentidenticalelements.Instead,OCPeffectsdocrossinterveningspecificationsofthesamefeature.Hence,theOCPrestrictionsofSemiticareofthesametypeasthosefoundinRussian(Padgett1991)andEnglish(Berkley1994),whereinterveningmaterialaffectsthestrengthoftheOCP.TheimplicationisthatOCPeffectsareingeneralnonlocal,preciselyaspredictedbythenotionofarticulatorylocalitydiscussedinsection5.2:articulatorylocalitytreatsthetwoconsonantsinaC1VC2#XUX,u#uhXUconfigurationasnonlocal,becausetheirarticulationsarenot T  contiguous.#XUXu #  6$      31    ݀uXXOtherlessknownlanguagesthoughttoemployV/CsegregationincludeuuAinu(It=1983),SierraMiwok  (Smith1985),and#uu#GtaF9u#uF9B#uuԀ(McCarthy1982)#uu#uu.ThesehavealsobeencontroversialasdiscussedinSteriade(1986)#uu#uuԀforAinu  andSierraMiwok,andOdden(1987)for#uu`#GtaF9u#uF9#uu.#uu1##XUXuR# v 6$      19    ݀uXXItisimportanttorememberthatthevowel[D]isjustthephoneticrelizationofaminorsyllable.uuAlso,two  otherrelevantcandidatesneedtobeconsideredhere.Thefirstrealizestheaffixbyspreadingofthesegmentallyadjacentconsonant#uu#Ԁ...uuc1.c1vc2#uu#uu,wherethetwoinstancesofc1#uu%#uuԀsharethesameRoot.Temiar,however,disallowsgeminate ` consonantsandthiscandidatecansafelybeignored(dependingontherepresentationofgeminatesinOT,thiscandidatemayalsoincuraviolationofundominatedSrole#uu#uu;seeIt=&Mester1993forrelevantdiscussion).Thesecondis 8 candidate#uu#uu c1 .c1vc2#uul#uu,#uu#Ԁwhere uuc1 #uu#Ԁisintheonsetpositionoftheprefinalsyllable.Thiscandidateviolatestheundominateduu  head#uu#,because uuc1 #uu7#Ԁbeingtheonsetofitsminorsyllableisseparatedbytheconsonantattheleftedgeofthemajorsyllable r  bytheemptynucleusnodeoftheminorsyllable(seetherepresentationofminorsyllablesin10).   6$      22    ݀uXXuuTheminimalpairsd.l<<ktohuntsuccessfully(basesl<<kwithanasalizedcopyofbasefinalconsonant  /k/infixed),andsn.l<<ksuccessinhunting(affix/n/withthebasesl<<k)clearlydemonstratesthedifferencebetween  thetwonasals.#uup#GG?GGGG@GG@G?GGG@GG@G?GG?G?G?TABLE IG V 6$      11    ݀uXXThereisonecase,outofthetwentyfourtotalpatterns,wherethispropertyisnotobservedinonedialectal  variantofthecontinuativeintheKelantanregion.Thisisthepatternuun#uu<# uuc3 .c1#uu#uu.c2vc3#uu#,wherethecopyofthebasefinal  consonantisnotalignedwiththeleftedgeofthemajorsyllableuuc2vc3#uu#.TheregularpatternisuuԀc1#uu#uu.n#uuc# uuc3 #uu#uu.c2vc3#uu#,foundinthe b Perakregion.Thisisastrikingcaseofdialectalvariationbecausetherestofthepatternsarethesameinthetwodialects.Inforthcomingwork,Ishowhowthisvariationfollowsfromsupposingthatthetwodialectshavethesameconstraintrankingbutdifferentparadigmstructures,formalizedbycorrespondencerelationsbetweentwolexicalforms.Specifically,thePerakpatternuuc1#uu#uu.n#uu=# uuc3 #uu#uu.c2vc3#uu#Ԁisinacorrespondencerelationwiththenominalperfectivebaseuuc1#uug#uun#uu#uu.c2vc3#uu#Ԁwhile t  theKelantanpatternuun#uux# uuc3 .c1#uu#uu.c2vc3#uu #Ԁisinacorrespondencerelationwiththeverbalperfectivebaseuuc1#uu #uu.c2vc3#uu #Ԁ(Gafos1995b).6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular6R Z 6Times New Roman RegularTABLE DTABLE EJJ?JJ?J?JJ?JTABLE EJTABLE Dr rr?r?rr?r?rr?r?TABLE Drt tt?t?tt?t?tt?t?tt@t?TABLE Etw w?w?www?www?www?w@TABLE Dw V 6$      27    ݀uXXFortheHebrewpatterns,itwillbeshowninsection6thatthesizeofthebisyllabictemplateimposedonthe  output,ratherthanthesegmentalmarkednessconstraints,determinesthenumberofcopiedsegments.  6$      36    ݀XUXXXuXXUTheanalysisabovethenessentiallyshowsthatthesinglecaseinArabicwhereLDCspreadingwasthought  toapplycanbeanalyzedwithoutthismechanism.#uuQ##XUXup#  H   H  j 6$      1    ݛuXXIwishtothankStephenAnderson,LindaLombardi,JohnMcCarthy,JayePadgett,PaulSmolensky,and  CherylZoll,fordiscussionsand/orconstructivecriticismsonpreviousversionsofthispaper.SpecialthankstoLuigiBurzioforprovidingdetailedcommentsonpreviousdrafts.Theusualdisclaimersapply.Thispaperisarevisedversionofachapterfrommyforthcomingdoctoraldissertation, TheArticulatoryBasisofLocalityinPhonology.#XUXuQ#   6R Z 6Times New Roman Regular6R Z 6Times New Roman Regular 6$      15    ݀uXX*PL//is,inotherwords,theencapsulationoftheMarkednessHierarchy.SeePrince&Smolensky1993,  section8.4.  6$      7    uXXԀuuSoutheastAsianistJamesMatisoffwhointroducedthetermsesquisyllabic(meaning,oneandahalf  syllables,Matisoff1978)forwordslike#XUXuO#ԀuXXUs.l<gtoliedown#XUXuY#ԀuXXUalsoagreeswiththisinterpretationofminorsyllablevowels t (p.c.,May20,1995)#XUXu#uXXU.#uup#     6$      12    ݀XUXXXuXXUThealignmentconstraintontheplacementoftheaffixesinTemiaressentiallyrequiresthattheRootnode  oftheaffixbeintherimepositionoftheprefinalsyllable.Hence,inoperationalterms,t#uuR#uuheaspectualmorphologyof  Temiarcanbeseenastheadditionofamoratothebase.Lombardi&McCarthy(1991)havearguedonthebasisofcrosslinguisticevidencethatthe theorymustrecognizeanoperationofmoraprefixation(1991:61).ThisoperationisfoundintwoMuskogeanlanguages,ChoctawandAlabama,andalsointwoAustronesianlanguagesBalangaoandKeleyi.#uu##XUXuq#ԀXUXXXUuXXUIntheselanguages,asinTemiar,theaddedmoraisalsorealizedwithsegmentalmaterialofthebase#XUXu#uXXU#uu#.#XUXuO#  6$      8    ݀uXXuuThereisevidencesuggestingthatBenjaminscategorical[B]inopen/[D]inclosedtranscriptionofminor  syllablerealizationsmaybeanoversimplificationoftherangeofminorsyllablevowelqualities.InthecloselyrelatedSenoiclanguages,SemaiandJahHut,Difflothnotesthatminorsyllablevowelshavevarioustransitionalqualities,dependingontheircontext.Forexample,inSemai,/k.DDp/redcentipedeisrealizedphoneticallyas[kDDDp].As $t Diffloth(1976a:233)characteristicallynotes, themainvowelD:startswhere/k/endsandendswhere/p/begins;the 8 glottalstopissuperposedatsometimeduringtheutteranceofthevowel.ThisanticipationofthevowelofthemajorsyllableDifflothfindstobeacharacteristicofallAslianlanguages.OthertransitionalminorvowelqualitiesnotedbyDifflothincludearound[u]inthecontextoflabialconsonantsandahighfront[i]beforepalatalconsonants.#XUXuP##XUXXXUo#(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular6R Z 6Times New Roman Regular  6$      23    ݀uXXAssumingthatthenasalinfixisspecifiedforsomeplaceofarticulationintheinput,thenthefactthatit  surfacesassimilatedtothefollowingstoprequirestherankinguuNC>>#uu2#uuIdentIO(PL)#uuz#.Therulebasedapproachwouldalso  requiresomeruleofdelinkingandreassociationoftheplaceofarticulationofthenasalinfix.Ontheotherhand,ifthenasalinfixisnotspecifiedforplaceunderlyingly(aswouldbenecessaryintheOTapproach,accordingtoInkelas1994,butcouldalsobethecaseintherulebasedapproach)thenbothframeworkswouldagainfareequallyinthemechanismsneededtoaccountforthisfact.(9 Z 6Times New Roman Regular I 6$      29    ݀uXXSeeBrowman&Goldstein(1989)uuԀandSproat&Fujimura(1993)#uu#Ԁfortwoexplicitproposalsonhow  articulatorygesturescanbedefined.  6$      30    ݀uXXThepictureismorecomplexthandepictedherebutneverthelessthebasicresultholds.First,VtoV  contiguityinaVCVsequencecanactuallybeinterruptedbycertaininterveningconsonants,e.g.palatalswhichemploythetonguebody(Recasens1984),thesamearticulatorusedforvowels.Second,aswasnotedinfootnote28,consonantharmonysystemsarelimitedtospreadingoffeaturesusuallyclassifiedunderthecoronalplaceofarticulation,suchas[anterior]and[distributed].SpreadingofmajorplacefeaturesuchasLabial,forexample,isnotattestedoverthevowelinaCVCsequence.Thegeometricnotionoflocalitydoesnotcapturethesefacts:theLabialnodeoftheconsonantliesonadifferenttierfromtheDorsalnodeofthevowelandshouldthusbeabletospreaduninterruptedbythevowel.Ontheotherhand,thearticulatorynotionoflocalityinGafos(1996)explainsthesefacts.TheLabialnodeoftheconsonantcannotspreadoveravoweltothesecondconsonantinaCVCsequencebecausethevowelinterruptsthearticulatorycontiguitybetweenthetwoconsonants.SpreadingoftheLabialnodecantakeplaceonlyifitgoesthroughthevowel.Thearticulatorylocalityalsopredictscorrectlythefactthatconsonantharmonyisfoundintermsofthefeatures[anterior]and[distributed].Thesefeaturesdescribethestateofthetonguetip,anindependentarticulatorfromthetonguebodywithwhichvowelsarearticulated.Hence,thetonguetipconfigurationofaconsonantcanspreadthroughthevowel,maintainingitsstatewhiletheinterveningvowelisarticulatedwiththetonguebody.SeeGafos(1996)fordetaileddiscussionoftheseissues. 6$      17    ݀uXXThis,infact,hasbeensuggestedintheanalysisofBroselow&McCarthy(1983).(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular \ 6$      26    ݀uXXItshouldbementionedthattheargumentsinthissection,especiallythoseagainstthedistinctionbetween  reduplicationandLDCspreadingdrawnfromMcCarthy(1979,1981),arenotmeanttounderscoreweaknessesinthespecificanalysesinquestionbutratherthelimitationsofthegeneralframeworkthoseanalysespresupposed,whichobscuredthesimilaritiesbetweenLDCspreadingandreduplication.#XUXuQ#uXXU(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(O Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(O Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular((3$ U!   (O Z 6Times New Roman Regular   (9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(O Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(O Z 6Times New Roman Regular(O Z 6Times New Roman Regular(O Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(O Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular  6$      16    ݀uXXAvariantofthiscandidate, uuc1v . c2#uu#uu a#uu#uu.c1vc2#uu0#,avoidstheonsetlesssyllablebutcausesafatalviolationofanother  undominatedconstraint,Srole,tobeintroducedbelow. U!   ݛ  =%7UXXdd7  XUXXX?TOa:Oa   OaOa ќ?T ?TOa"Oa ?TћD?T;L?T9񛀀  Ѐ#XUXXXU#XUXXXUOntheProperCharacterizationofNonconcatenativeLanguages#XUXXXU#XUXXXU  #XUXXXU#XUXXXU AdamantiosGafos 8  DepartmentofCognitiveScienceTheJohnsHopkinsUniversity,BaltimoreMD21218,USA 9#  1      #XUXXXU#XUXXXU  `   #XUXXXU#XUXXXU uXXU[ o December10,1995!  ]#XUXu#  $  `  _Contents H   1 Introduction  p  2 CorrespondenceinOptimalityTheory   3 TemiarVerbalMorphology:AUnifiedAccountofSegmentalCopying 8"  @  3.1 BasicProsodicandMorphologicalProperties X  3.2 SegmentalCopyingDerivedbyCorrespondence   4 TemiarinPreviousAnalyses (x  5 OntheNecessityofEliminatingLDCspreading   5.1 TheApparentNeedforReduplicationandLDCspreading H  5.2 TheExceptionalStatusofLDCspreading `   6 TypologicalConsequences "  6.1 FurtherAnalyses 0!$  6.2 AtemplaticAffixation "&  7 Conclusion  $P( ̜ ,(|#.     ћ Q  #XUXXXU#XUXXXUOntheProperCharacterizationofNonconcatenativeLanguages#XUXXXU #  Abstract  t i   uXXU  Nonconcatenativelanguageshavebeenclaimedtoemployaspecialtypeofphonologicalspreadingofa ` consonantoveravowel,whichassumesarepresentationthatsegregatesconsonantsandvowelsondifferentplanes.Iarguethatthistypeofspreadingcanandmustbeeliminatedfromthetheory,byreducingittosegmentalcopyingasinreduplication.CrucialtothisreductionisthenotionofgradientviolationofconstraintsinOptimalityTheory(Prince&Smolensky1993),andthenotionofCorrespondencewithitsparticularapplicationtoreduplicativemorphology(McCarthy&Prince1995a).ThereductionisdemonstratedindetailforTemiar,oneofthemainindigenouslanguagesofMalaysia,notoriousforthecomplexityofitscopyingpatterns.ExtensionsoftheproposaltoSemiticlanguagesarealsodiscussed.Twomaintheoreticalimplicationsofthisreductionarethendeveloped.First,thedistinctionbetweenconcatenativeandnonconcatenativelanguagesneednotandshouldnotbeencodedintermsofthespecialphonologicalmechanismsofconsonantalspreadingoveravowel,applyingunderplanarsegregation.Second,thelocusofthedistinctionisfoundtobe,instead,inthemodeofaffixationemployedinnonconcatenativelanguages,namely,atemplaticreduplicativeaffixation.Thistypeofaffixationispredicted,thoughheretoforeundocumentedinthetypologyofwordformation.#XUXu #  T  М̛     v   hXUXXXU h  1.Introduction  ,  3Currentphonologicaltheoryclaimsthatnonconcatenativelanguagesarephonologicallyspecial.Intheselanguages,aconfigurationsuchasCiVCi,wherethetwoconsonantsareidentical,mayresult  fromanautosegmentaloperationthatspreadstherootofasingleunderlyingconsonantovertwoCpositions(see1).Thistypeofspreadingisknownaslongdistanceconsonantalspreading,henceforthLDCspreading.#XUXXXUe#XUXXXULDCspreadingisthoughttoproceedunobstructedbytheintervening d  vowelbecausevowelsandconsonantsarerepresentedondifferentplanes.ThisrepresentationalhypothesisisknownasV/Cplanarsegregation.BothLDCspreadingandV/Cplanarsegregationareconsideredtobeuniquetononconcatenativelanguages. h1.V/Cplanarsegregation 3 3   `    VRoot  22@(#Vowelplane '%#+  3   `    |   `  X  X h X  (output:CiVCi#XUXXXU#XUXXXU)     x  Skeleton f)$-    `  |    `  CRoot h   <(#Consonantplane @  >+&/  c 3 +'0  hTheeffectofLDCspreadingisthustocreateacopyofasegmentoverinterveningsegmentalmaterial,aneffectsimilartothatfoundinthephenomenonofreduplication. 3XӀSimilaritynotwithstanding,LDCspreadingandreduplicationhavebeenattributedtounrelatedmechanismsofthetheory.InLDCspreading,copyingistheapparenteffectofdoublelinkingofasingleconsonanttotwoskeletalpositions.Inreduplication,copyingliterallycreatesasecondinstanceofaconsonant.Thetwomechanismsexistwithindifferentcomponentsofthegrammar,LDCspreadinginthephonologicalandreduplicationinthemorphologicalcomponent. hMygoalinthispaperistoshowthatthereisaredundancyinthetheorywhichadmitstwodistinctoperationsofsegmentalcopying,andadvocateitseliminationbyreducingLDCspreadingtothesameformalmechanismasinreduplication.CrucialtothisreductionisthenotionofgradientviolationofconstraintsintheframeworkofOptimalityTheoryofPrince&Smolensky(1993).Ithenexaminetheimplicationsofthisreductionforthetypologicaldistinctionpositedbetweenconcatenativeandnonconcatenativelanguages.IarguethatthisdistinctionisnotformallyexpressedintermsofthespecialphonologicalmechanismsofLDCspreadingandV/Cplanarsegregation,butratherintermsofthespecialmodeofreduplicativeaffixationemployedinnonconcatenativelanguages.Reduplicativeaffixesintheselanguagesareatemplatic,inthesensethatthereisnoprosodicrequirementimposedontheshapeoftheaffixes.TheidentificationofthistypeofaffixationfillsapredictedbutheretoforeundocumentedtypologicalgapinthetheoryofProsodicMorphology.Thispaperisorganizedasfollows.Section2introducestheOptimalitytheoreticnotionofcorrespondence,whichhassofarbeensuccessfullyemployedtocharacterizethecrosslinguisticfactsoftemplaticreduplication,andwhich#XUXXXU#XUXXXUԀwillplayacentralroleintheanalysesthroughoutthis )$( paper.Section3beginsthemainpartofthispaperbyadetailedexaminationoftheredundancy h+&* betweenthetwocopyingmechanismsinTemiar,oneofthemainindigenouslanguagesofMalaysia(Benjamin1976)#XUXXXUw #XUXXXU.IchosetoconsiderTemiarbecauseofthreereasons.#XUXXXU!#XUXXXUԀFirst,thelanguageis  notoriousforthecomplexityofitscopyingpatterns,anddespiteimportantattempts(McCarthy1982,Broselow&McCarthy1983),ithassofarresistedasatisfactoryaccount.#XUXXXUc"#XUXXXUԀSecond,Temiarhasbeen 8  arguedtorequirethefulldeploymentofbothcopyingmechanisms,LDCspreadingandreduplication#XUXXXU##XUXXXU.#XUXXXUO$#XUXXXUԀFinally,thefactsofthelanguageillustrateinastrikingwaytheinadequacyofthe  8  derivationalapproachtocopying,whileatthesametimedemonstratingthespecialtypeofaffixationthatIwillidentifyasthegeneralcharacteristicofreduplicativeaffixationinnonconcatenativelanguages.̜Themainpartofsection3presentsaunifiedanalysisofsegmentalcopyingintheverbalmorphologyofTemiar.#XUXXXU$#XUXXXUԀ#XUXXXU&#XUXXXUTheanalysisbuildsonanunderstandingofthebasicprosodicand  p morphologicalpropertiesofthelanguagedevelopedhereforthefirsttime.#XUXXXU&#XUXXXUԀAllinstancesof H segmentalcopyingareanalyzedintermsofasinglenotionofcorrespondence.Thefullrangeofpatternsemergesfromtheinteractionofcorrespondenceconstraintswithotherconstraintsexpressingindependentlyestablishedregularitiesofthelanguage.Section4discussespreviousanalysesofthelanguage,demonstratinginparticularthesuperiorityofthecorrespondenceapproachtocopyingoverderivationalalternatives.Section5arguesthattheeliminationofLDCspreadinganditsgeometricpremiseofV/Csegregationisnotonlypossiblebutinfactnecessary.#XUXXXU'#XUXXXUԀReconstructingtheargumentforhavingboth %0!$ LDCspreadingandreduplicationinthetheory,Iexposetheweaknessesofitspremisesandtheconceptualproblemstheyraise#XUXXXUP*#XUXXXU.#XUXXXUP+#XUXXXUIargue,inaddition,thatthetheoryadmittingLDCspreadingand )$( V/CplanarsegregationfailstoexplainthefactthatwheneverLDCspreadinghasbeenclaimedto h+&* apply,itspreadsthewholeconsonantandneveroneofitsindividualfeatures.#XUXXXU+#XUXXXUTheseproblemsare  resolvedundertheproposedunificationsinceitisclearthatsegmentalcopying,asinreduplication,targetsonlywholesegments,notindividualfeatures.#XUXXXU,#XUXXXU ` Section6developsthetypologicalconsequencesoftheeliminationofLDCspreadingandV/Csegregation.ItbeginsbyexaminingsomebasicpatternsofsegmentalcopyinginSemitic.TheselanguageswerethoughttoprovideanothertypeofmotivationforV/Csegregation,duetothetraditionallyassumeddistinctmorphologicalstatusattributedtoconsonantsandvowels.#XUXXXU-#XUXXXUԀIrrespective   ofthestatusofthisassumption,thesamebasicreductionofLDCspreadingtocopyingviacorrespondenceextendstotheselanguagesaswell.IconcludethatthedistinctionbetweenconcatenativeandnonconcatenativelanguagesneednotandshouldnotbeencodedintermsofthespecialphonologicalmechanismsofV/CsegregationandLDCspreading.#XUXXXU/#XUXXXURather,thedistinctionis  p identifiedwithaspecialtypeofreduplicativeaffixationemployedinnonconcatenativelanguages,wherereduplicativeaffixesarenotspecifiedforanyprosodictarget,theirexactrealizationbeingdeterminedbytheconstraintsoftheparticularlanguage.Finally,section7concludeswithasummaryofthemainargumentsandresultsofthepaper.#XUXXXU~1#XUXXXUT 9#  2      #XUXXXUL3# XUXXXU   h  2.#XUXXXU3#XUXXXUCorrespondenceinOptimalityTheory#XUXXXUY4#XUXXXU 0"   &4 3FaithfulnessinOTexpressesthefactthatrelatedgrammaticalforms,suchasInput/Outputand $X"  Base/Reduplicant,tendtobeidentical.CorrespondencetheoryofMcCarthy&Prince(1995a)gives %0!$ formalcontenttothenotionoffaithfulness 9#  3      .Correspondencedefinesarelationbetweentwoforms,   asstatedin(2)below.#XUXXXU4#XUXXXU   h42.Correspondence#XUXXXU7#XUXXXU: 35 3GiventwosegmentalstringsS1and#XUXXXU7#XUXXXUԀS2#XUXXXU/8#XUXXXU,correspondenceisarelation4from  l thesegmentsof#XUXXXUz8#XUXXXUS1#XUXXXU9#XUXXXUԀto#XUXXXU_9#XUXXXUthoseofS2#XUXXXU9#XUXXXU.SegmentsofS1#XUXXXU9#XUXXXUԀandof#XUXXXUU:#XUXXXUԀS2arereferredtoascorrespondents  D ofoneanotherwhen4#XUXXXU:#XUXXXU#XUXXXUM;#XUXXXU.#XUXXXU;#XUXXXU    #XUXXXU;# 37XUXXXU 3Acorrespondencerelationimposesanumberofconstraintsrequiringidentitybetweenthetwo `  segmentalstrings.Twobasiccorrespondenceconstraintsaregivenin(3)and(4)fortheBase(B)/Reduplicant(R)correspondencerelation@+ 9#  4      . ` #XUXXXUD<#XUXXXU3.MaxBR  e> 3c<  EverysegmentofBhasacorrespondentinR.>4.DepBR   3  EverysegmentofRhasacorrespondentinB.#XUXXXUE>#XUXXXU  -? 3h?Perfectcorrespondenceistotalreduplication,asinAxinincaCampa #XUXXXU?#XUXXXUnata nata#XUXXXU@#XUXXXUcarry(copied  segmentsareboldfaced),whichfullystatisfies#XUXXXU@#XUXXXUMaxBR#XUXXXUA#XUXXXUԀandDepBR#XUXXXUA#XUXXXU.Deviationsfromperfectionare f foundwhenthereduplicantcopieslessthanthewholebase,violatingMaxBR#XUXXXU=B#XUXXXU,orwhenthe >! reduplicantcontainssegmentswhicharenotpartofthebase,violatingDepBR#XUXXXU C#XUXXXU.Bothcasesofviolation  f# correspondtowellattestedphenomena,partialreduplicationandprespecifiedreduplicationrespectively.InTemiar,thesimulfactiveaspectform c1 #XUXhXUC#XUXhXUa.c1vc2#XUXXXUD#XUXXXUԀ#XUXXXU.E#XUXXXUderivedfromtheverbalbase#XUXXXUqE#XUXXXUc1vc2#XUXXXUE#XUXXXU, #' copiesonlyasingleconsonantofthebase#XUXXXU#F#XUXXXU,causingtwoviolationsof#XUXXXUF#XUXXXUMaxBR#XUXXXUG#XUXXXUԀbecause/v,c2#XUXXXUVG#XUXXXU/arenot % ) copied.#XUXXXUG#XUXXXUԀMoreover,theoutputcontains/a/whichisnotpartofthebase,aviolationofDepBR#XUXXXUH#XUXXXU.   3Otherconstraintsevaluatethequalityoftheidentitybetweencorrespondentsegmentsoverfeaturalandprosodicdimensions,asin(5)and(6)respectively.#XUXXXUH#XUXXXU5.IdentBR#XUXhXUI#XUXhXU(F)#XUXXXU)J#XUXXXU  D I 3H  AsegmentinRanditscorrespondentinBmusthaveidenticalvaluesforthefeature[F].#XUXXXUqJ#XUXXXU    J6.Srole  K    3AsegmentinRanditscorrespondentinBmusthaveidenticalsyllabicroles. 3K#XUXXXUkK#K 3XUXXXU 3LFeaturalidentitymaybeviolatedbecauseofhigherrankedconstraintsimposingspecificdemandsonthefeaturalmakeupofacorrespondentsegment.#XUXXXUL#XUXXXUԀInTemiarvoicelessstopsare   nasalizedtobecomemoresonorousinthecodapositionduetoaconstraintspecifictocodas,Coda  Cond.Whenacopyofthebasefinalconsonantisaffixed,#XUXXXUM#XUXXXUasin#XUXXXUN#Xl XXXUԀ#XUX XXlO#XUXXXUyaap#XUXXXU[O#XUXXXUԀtocry#XUXXXUO#XUXXXUyD m .yaap#XUXXXUO#XUXXXU,the ^ consonantisthusnasalized:#XUXXXURP#XUXXXUIdent#XUXXXUP#XUXXXUBR(nasal)isviolatedbecauseofthehigherrankedCodaCond#XUXXXUQ#XUXXXU.#XUXXXUQ#XUXXXU J #XUXXXUQ#XUXXXUSrole#XUXXXUp 3}q 3Biconsonantalbasesconsistentirelyofonesyllable.Triconsonantalbasesarebisyllabic,exhibitingthespecialtypeofunderlyinglyvowelesssyllablefoundinAslian(andMonKhmer)languages.In#XUXXXUr#XUXXXUs.l<g#XUXXXUt#ԛtheconsonantXUXXXU/s/#XUXXXUrt#XUXXXUistheonsetofasyllable,calledaminorsyllable,asopposedto   thefinalmajorsyllableoftheword,whichcontainsaphonologicallyspecifiedvowel.#XUXXXUt#XUXXXUExamplesof   wordswithoneconsonantminorsyllablesareshownin(8ac)(ignoreforthemomentthephoneticϜforms,whichIdiscussbelow).#XUXXXUu#XUXXXUIn(8a), t .l#XUXXXUv#Xl XXXUD#XUX XXlv#XUXXXUk, t istheonsetoftheminorsyllable,followedbythe 2  majorsyllablel#XUXXXU6w#Xl XXXUD#XUX XXlw#XUXXXUk.Morphologicaloperations,involvinginfixationsofconsonants,cancreateclosed n  minorsyllables,consistingoftwoconsonantsasshownin(8df).#XUXXXU$x#XUXXXUԀIn(8d),theminorsyllableis br #XUXXXUy#XUXXXU  Z with/b/asitsonsetand/r/asitscoda. 3s#XUXXXUzy#XUXXXU8.a. t. lDk toteach h [tB.lDk]  d. br. caa  tofeed x [bDr.caa] ~   b. b. huj guilty h [bB.huj]  e. cb. niib  going x [cDb.niib] 6   3Ӏc. s. l<g toliedown h [sB.l<g]  f. t. taa  oldmen x [tD.taa]#XUXXXU&z#XUXXXUԛ N yuXXU#XUXu|#AccordingtoBenjamin,phoneticallyminorsyllablessurfacewithtwopredictablevowel [ qualities.Openminorsyllablesaretranscribedwiththevowel[B]asin(8ac),andclosedminor 3  syllableswiththevowel#XUXXXU|#XUXXXU[D]#XUXXXUY~#XUXXXUԛasin(8df)#XUXXXU~#XUXXXU.Itisclearthatminorsyllablevowelshavenophonological  [" status.Thiscanbeseeninthewaythesevowelssurfaceinmorphologicalvariantsofaword.#XUXXXU~#XUXXXUԀIn(9), !3$ showingpartofthevoice/aspectparadigmof#XUXXXU#XUXXXUs.l<g#XUXXXUo#XUXXXUtoliedown,#XUXXXU#XUXXXUԀtheminorsyllablevowels#XUXXXU#XUXXXU[B/D]#XUXXXUo#XUXXXUԛare # & freelysubstitutedbyothervowelsprovidedbythemorphologyorbyeachother,conditionedapparentlyonlybysyllablestructureinthelattercase.uXXU#uu#uuԀ#XUXu# k'"* #XUXXXÛ#XUXXXU#XUXXXUG#XUXXXU9.a.[sB.l<g] toliedown h     c.[sDr.l<g]  toliedownCAUS.#XUXXXU#XUXXXUԛ )O%.  3{ #XUXXXUb#XUXXXU  b.[sa.l<g] toliedownSIM.    d.[sB.rDg.l<g]toliedownCAUS#XUXXXU#XUXXXU.CONT.#XUXXXU#XUXXXU I+&0 #XUXXXU#XUXXXU  h#XUXXXUP#XUXXXUIn(9b),whenthevowelaffix/a/isaddedtotheverbalbaseof(9a),the[B]intheminorsyllableof  М[sB.l<g]changesto[a].In(9c),whentheconsonantaffix/r/isadded,thevowel[B]changesto[D].  Finally,(9d)showsthesamealternationintheinversedirection,wherethevowel[D]of[sDr.l<g] {  turnsto[B].This[B]/[D]oscillationhasnophonologicalbasis.#XUXXXU#XUXXXU S   h#XUXXXUՈ#XUXXXU 3FollowingthegenerallyacceptedviewinMonKhmerlanguages,Iwillthusassumethatthese + {  vowelsarenotspecifiedunderlyingly,butarethephoneticrealizationsofasyllablewithnophonologicallyspecifiedvowel#XUXXXU>#XUXXXU 9#  7      .#XUXXXU#XUXXXUԀThisisalsotheassumptioninDifflothsdescriptionsoftwoother +  Senoiclanguages,SemaiandJahHut,#XUXXXU#XUXXXUwhicharecloselyrelatedtoTemiar 9#  8      .#XUXXXU؋#XUXXXUԀSpecifically,Iwill  assumethatthephonologicalrepresentationofminorsyllablesisasshownin(10),wherethenucleushasnophonologicalspecificationandisinterpretedbythephoneticcomponentofthegrammarbasedonthesyllabiccontext.#XUXXXU#XUXXXU ; 10. `     h   %#XUXXXUߍ#XUXXXU       `     h   n#XUXXU#XUXXU   i #XUXXXU<#XUXXXU 3]    `     h C  (C) p ܏#XUXXXU#XUXXXU c  3SyllabicstructurecanalwaysbeunambiguouslyassignedbecauseeverysyllableinTemiarmusthaveanonsetandcomplexsyllabicmarginsarenotallowed.Thuswhenthereisonlyoneconsonant, !c" asin #XUXXXU#XUXXXUt. lDk#XUXXXU#Xl XXXUԛ#XUX XXl7#XUXXXUtoteach#XUXXXU{#XUXXXU,thatconsonantistheonsetofthesyllableandwhentherearetwoconsonants,  asin #XUXXXU͒#XUXXXUbr. caa#XUXXXU~#Xl XXXUԛ#XUX XXl͓#XUXXXUtofeed,#XUXXXU#XUXXXUthefirstistheonsetandthesecondisthecodaoftheminorsyllable.#XUXXXUd#XUXXXUThis  simplesyllabicstructurecanbecapturedbythetwostandardconstraints#XUXXXU#XUXXXUԀofthebasicsyllabictheory  ofOTshownin(11)and(12).Thesetwoconstraintsare#XUXXXU#XUXXXUԀneverviolatedinthislanguage. `  Л11.Ons  D   3Ԑ  Everysyllablemusthaveanonset.%12.*Complex 7    3Nomorethanonesegmentmayassociatetoanysyllabicmargin.#XUXXXU>#XUXXXUԛ #  R#XUXXXU # 3XUXXXU 3AnotherprosodicpropertyofTemiaristhatmajorsyllablesappearalwaysinthefinalposition,  bearingthewordstress,andprecededbyanoptionalsequenceofminorsyllables.ThisisageneralpropertyoftheMonKhmerfamily.SincestressinMonKhmeristypicallyfinal,thispropertycanbeseenasaspecialcaseofawidelyattestedtendencyoflanguagestoreducetheirvowelinventoriesinunstressedpositions.Typicalexamplesofinventoryreductionsfromstressedtounstressedpositionscrosslinguisticallyinclude,theseventofivevowelreductionofItalian= 9#  9      ,#XUXXXU#XUXXXUԀthesixtofour J reductionofRumanian,thefivetothreereductioninSicilian,andtheeighttosixreductioninTurkish.InTemiarunstressedpositions,correspondingtominorsyllables,thevowelinventoryisreducedtojusttwopredictablequalities[B/#XUXXXU#D]XUXXXU.Theissueofhowtoexpressthisgeneralizationusing !" currentresourcesisaproblembeyondthescopeofthispaper.HereIwillsimplyusetheconstraint Z#$ 1v#XUXXXU#XUXXXUԀin(13),asacovernameforthesetofconstraintsthatmayliebehindtheTemiar(andMon 2% & Khmer)generalization. )$, ?T H  13.#XUXXXU#XUXXXU1v#XUXXXU#XUXXXU   3 3Thereisonlyonespecifiedvowel(henceonemajorsyllable)perword.#XUXXXUZ#XUXXXU J Inowturntothemorphologicalpropertiesoftheverbalparadigms.#XUXXXU#XUXXXUTherearetwovoices,active  andcausative.Foreachvoice,therearethreeaspects,perfective,simulfactive,andcontinuative#XUXXXU#XUXXXU.The f  ?T H  ?T H  aspectualparadigmoftheactivevoiceisshownin(14). 3 3˺14.#XUXXXU#XUXXXU `     h Biconsonantal p   Triconsonantal "   3~*qr$s ddd Xdd Xdd X(#(#q,G dd ,( dd ,( dd +   w  a.Perfective %+  %#XUXXXU͢#XUXXXUc1vc2󀀀 +  Мk<<wtocall %  %c1.c2vc3 +  Ӝs.l<gtoliedown#XUXXXU#XUXXXU %  %b.Simulfactive %G  %#XUXXXU#XUXXXUc1a.c1vc2 G  Мka.k<<w %3 %c1a.c2vc3 G  Мsa.l<g#XUXXXU#XUXXXUԛ %3 %<c.Continuative %  %#XUXXXU#XUXXXUc1c2.c1vc2  Мkw.k<<w %  %c1c3.c2vc3  Мsg.l<g#XUXXXUL#XUXXXUԛ&   & 3 h 3Theunmarkedperfectiveaspectconsistsoftheverbalbasealone,(14a).Thisperfectiveisthenthebasefortheformationofthetwootheraspects,thesimulfactiveandthecontinuative.Thesimulfactiveaspectin(14b)ismarkedbythe#XUXXXU:#XUXXXUԀvowel/a/andinthebiconsonantal#XUXXXUު#XUXXXUԀcaseacopiedbase $ consonant.#XUXXXUC#XUXXXUThecontinuativeaspectin(14c)involvesonlycopyingofbaseconsonants.#XUXXXU#XUXXXU &  hTheaspectualparadigmofthecausativevoiceisshownin(15).#XUXXXUD#XUXXXU 3V e( 15.#XUXXXU#XUXXXU `     h Biconsonantal#XUXXXU]#XUXXXU p   Triconsonantal#XUXXXUޭ#XUXXXU !I,  3 *qt$u ddG dd ( dd ( dd r$s(#(#q,G dd ,( dd ,( dd +   C#. ӀBase(Act.Perf.) %#/ %#XUXXXUK#XUXXXUc1vc2 #0 Мk<<wtocall %$1 %c1.c2vc3 #2 Мs.l<gtoliedown#XUXXXU# %$3 %za.Perfective %&c!4 %XUXXXUԜtr.c1vc2 &c!5 Мtr.k<<w %&O"6 %c1r.c2vc3 &c!7 Мsr.l<g#XUXXXUa#XUXXXU %&O"8 %b.Simulfactive %(#9 %#XUXXXU<#XUXXXUt.ra.c1vc2 (#: t.ra.k<<w %k)$; %c1.ra.c2vc3 (#< s.ra.l<g#XUXXXU#XUXXXU %k)$= %c.Continuative %*;&>  %#XUXXXUݳ#XUXXXUt.rc2.c1vc2 *;&? t.rw.k<<w %+''@  %c1.rc3.c2vc3 *;&A s.rg.l<g +''B 7#XUXXXU#&,(C   &OaH  XUXXXU hTheperfectiveaspect#XUXXXU#XUXXXUisformedfromthecorrespondingactiveperfectivebase(repeatedin15asthe  Base)byadditionoftheaffix/tr/,(15a)#XUXXXUl#XUXXXU.#XUXXXU<#XUXXXUThisaffixissubjecttoallomorphy,asshowninthecase  oftriconsonantalbases,whereitappearsasaninfixed/r/  9#  10      .Asintheactivevoice,thesimulfactive ` OaH  OaH  andcontinuativeareformedfromtheperfectivebase.Thesimulfactiveisagainmarkedbythevowelinfix/#XUXXXU#<XUXXXUa#XUXX<XU#XUXXXU/,(15b),andthecontinuativebycopyingofvariousbaseconsonants,(15c).#XUXXXU#XUXXXU  `  h% 3Animportantpropertyintheseparadigmsconcernsthelocusofaffixationofthesimulfactiveandcontinuativemorphemes.Inallsimulfactivepatternstheaffix/#XUXXXU#?XUXXXUa#XUXX?XU#/appearsimmediatelytotheleft   ofthemajorsyllableofthebase,asshownbytheformsenumeratedin(16a)XUXXXU.Thecontinuative   patternshaveacopiedconsonantalsoimmediatelytotheleftofthemajorsyllableofthebase,asshownin(16b).#XUXXXUb#XUXXXU H #XUXXXU>#XUXXXU16. ` a.Simulfactives:c1a.c1vc2#XUXXXU#XUXXXU  c1a.c2vc3#XUXXXU#XUXXXU p t.ra.c1vc2  #XUXXXU#XUXXXUc1.ra.c2vc3#XUXXXU#XUXXXU ,  3ƺ 3   ` b.Continuatives:c1#XUXXXUh#XUXXXUc2.c1vc2#XUXhXU#XUXXXU  c1#XUXXXU|#XUXXXUc3.c2vc3#XUXXXU#XUXXXU p t.r#XUXXXU5#XUXXXUc2.c1vc2#XUXXXU#XUXXXU  #XUXXXU#XUXXXUc1.r#XUXXXU7#XUXXXUc3.c2vc3#XUXXXU#XUXXXU #XUXXXU#XUXXXU &v    ` #XUXXXU3#XUXXXU 3 h 3Thegeneralizationthatstandsoutisthatanewsegment(/#XUXXXU#AXUXXXUa#XUXXAXUV#XUXXXU/oracopyofaconsonant)appearsinthe   Лrimepositionoftheprefinalsyllable.Thisisarobustpropertyofthelanguage,applyingtoallcontinuativeandsimulfactiveformsmd 9#  11      .Iproposetocaptureitwithanalignmentconstraint,requiring j  thattherightedgeofanaffixmustbealignedwiththeleftedgeofthemajorsyllableofthebase#XUXXXU#XUXXXU.  Thismajorsyllable,beingstressed,istheheadoftheprosodicword(PrWd)inTemiar.TheconstraintcanthenbestatedinthealignmentschemaofMcCarthy&Prince(1993b),asin(17).Thisconstraintappliesthroughouttheverbalmorphologyofthesimulfactiveandcontinuative.#XUXXXU#XUXXXU 8  #XUXXXU"# hXUXXXU17.Align(Affix,R,Head(PrWd),L)     3 3Therightedgeofanaffix#XUXXXU#XUXXXUmustbealignedwiththeleftedgeoftheprosodicheadofthebase.#XUXXXU#XUXXXU 3& 3    Ѐ(henceforth,head)   8 3D 3ӛApartfromthis,theredoesnotappeartobeanyparticularprosodicrequirementonhowtheseaffixessurfaceinthevariousoutputs.#XUXXXU#XUXXXUThesimulfactiveisrealizedwiththevowel/a/,andinthecase v  ofbiconsonantalswithacopyofaconsonantofthebaseaswell.Thecontinuative,ontheotherhand,isalwaysrealizedwithacopyofatleastonebaseconsonant.Thefollowinganalysiswillshowthatthesimulfactiveandthecontinuativeaffixesarebothreduplicativeandspecifiedasconsistingofasingle(segmental)Rootnode.TheonlydifferencebetweenthetwoisthatinthesimulfactivetheRootisphonologicallyspecifiedtobethevowel/a/,whileinthecontinuativeitisnot.Thisdifferenceisillustratedin(18a)and(18b)respectively.18.a.Simulfactiveaffix: h Root    b.Continuativeaffix:   x Root  j"    `     h |     p     #  3   `     h a     p     (nofeaturalcontent)#XUXXXU#XUXXXU2 d!$ #XUXXXU #XUXXXU hItwillbeseenthatthefactthatinthecontinuativetheRootisrealizedwithaconsonantfollowsfrom <#& theinteractionofindependentconstraintsontheprosodyofthelanguage.Infact,allotherdifferencesbetweenthesimulfactiveandcontinuativepatternswillfollowentirelyfromthe interactionoftheindependentlyestablishedprosodicregularities,expressedby*Complex,Ons, ($, and1v.#XUXXXU#XUXXXUԜ 9#  12         #XUXXXU# hXUXXXUIconcludethissectionwithadiscussionoftheproductivityoftheverbalparadigms.#XUXXXUw#XUXXXUԀNotall  basesappearinallpossibleformsofeveryparadigm.#XUXXXU#BenjaminnotesthatXUXXXUc1.c2vc3#XUXXXU#Ԁbases commonly ` lackperfectiveforms,ortheyexistonlyinthecausativeform(1976:168)andthatitishardlythecasethateveryrootexhibitsallpossibleformsofeveryvoice.Thisappearstobeageneralpropertyofthesocallednonconcatenativemorphologicalsystems.McCarthy(1979:239),onArabic,notesthatitis anidiosyncraticpropertyofanyrootwhetheritcanappearinaparticularbinyan.AronoffXUXXXU   #XUXXXU#(1994:124)makesthesameobservationabouttheHebrewBinyansystem,notingthat fewifany   rootsactuallyoccurinallfivemajorbinyanim.Finally,Prunet(1995:2)identifiesthislackoffullproductivityasoneofthemaincharacteristicsofSemiticmorphology.AtleastinTemiar,thisphenomenoncanusuallybeascribedtothesemanticincompatibilityofthestemwiththeaspectualcategory.XUXXXUBenjaminnotesthat,wherethemeaningallows,alltheseforms H areproductive,withtheexceptionofthecausativesimulfactivewhichoccursincertaincrystallizedformsonly,andeventhenusuallywithanontransparentmeaning(i.e.differentthanimpliedbytheinflectionalcategory).Eventhisform,however,frequentlyoccursinexpressives(Benjamin1976:170),andthusishighlyproductiveinordinaryconversation,storiesandsonglyrics.Itseemssafetoconcludethenthatwhenstemsdonotexemplifyallpossiblepatternsthereasonsareprimarilysemantic,andnotrelatedtothecomplexityoftheformperse.#XUXXXU#XUXXXU $X" ?TH  Tosumup,thethreebasicpropertiesofTemiarwhichwillbecrucialtotheanalysisareasfollows:everysyllablemusthaveanonset(Ons),complexsyllabicmarginsarenotallowed  (*Complex),andevery(output)wordmustcontainonlyonespecifiedvowel(1v).Finally,#XUXXXU#XUXXXUthebasic ` generalizationaboutthelocusofaffixationinthesimulfactiveandcontinuativepatternsisthattheaffixappearsalignedwiththeleftedgeofthemajorsyllableofthebase(head#XUXXXUf#XUXXXU).#XUXXXU#XUXXXU   ` #XUXXXU#XUXXXU 3 h3.2.SegmentalCopyingDerivedbyCorrespondence #XUXXXU&#XUXXXU 2  Thissubsectionpresentsafullanalysisofthesimulfactiveandcontinuativeaspects,inthatorder.#XUXXXU#XUXXXU |  Inthesimulfactiveformsof(19)therearetwovoices,activeandcausative.Eachvoiceexhibitstwopossiblepatterns,abiconsonantalandatriconsonantal.#XUXXXUx#XUXXXUCopiesofconsonantsareshowninboldface. ,| #XUXXXUn# hWXUXXXU#XUXXXU#XUXXXU19.ACTIVE   Biconsonantal󀀀   p Triconsonantal x    3E  a.Base   c1vc2 h      p c1.c2vc3  Z    `    k<<wtocall     p s.l<gtoliedown F   b.Simulfactive  c1 a.c1vc2     p c1a.c2vc3 X    `    k a.k<<w     p sa.l<g H   CAUSATIVE   Biconsonantal󀀀   p Triconsonantal  N  c.Base   tr.c1vc2     p c1r.c2vc3 b     `    tr.k<<w     p sr.l<g N! =  d.Simulfactive  t.ra.c1vc2     p c1.ra.c2vc3 ` #    `    t.ra.k<<w     p s.ra.l<g#XUXXXUF#XUXXXU L!$ #XUXXXU# 3XUXXXUA#XUXXXU#XUXXXUllsimulfactiveformshaveaprefinalsyllablewiththevowel/#XUXXXU#UXUXXXUa#XUXXUXU_#XUXXXU/,aclearviolationofthe #' constraint1v#XUXXXU#XUXXXU,expressingthefamilywidegeneralizationofMonKhmerlanguagesthatallowsonly Z% ) foronesyllablewithafullyspecifiedvowel#XUXXXU(#XUXXXU.Thisprovidesuswiththefirstrankingargumentinthe 2'"+ analysis.Letmeassumethattheinputofthesimulfactiveconsistsofthesegmentalexpressionoftheaspect,namely,thevowel/#XUXXXU#YXUXXXUa#XUXXYXU#XUXXXU/,andthebase.Forexample,inthecaseofanactivetriconsonantal *2&/ basetheinputwillbeasshownintheupperleftcorneroftableau(20)below. , (1 OaH  #XUXXXUU#XUXXXU20.Rankingargument:MaxIO>>1v 3  *wJ$K ddG dd ( dd ( dd t$u(#(#w,dd ,dd ,dd +  33  Input:#XUXXXU~#Xl XXXUԀ#XUX XXl,#nXUXXXUԀa#XUXXnXUq#FXUXXXU,#XUXXFXU#XUXXXUԀ#XUXXXU#XUXXXUc1#XUXXXU;#XUXXXU.c2vc3#XUXXXU#XUXXXU )3?33 )MaxIO 23!?3 21v 43!?"3 4a.#XUXXXU#XUXXXUԀc1#XUXXXU#XUXXXU.c2vc3#XUXXXU6#XUXXXU +!y 3 +*! 'y  " ' '3y  " ' 3b.+#XUXXXU#XUXXXUc1#XUXXXUj#pXUXXXUa#XUXXpXU#XUXXXU.c2vc3#XUXXXU#XUXXXU +! I 3 + ' I " '*" I "  "#XUXXXUK#XUXXXU 3$ hConstraint1vfavorsacandidatelike(20a),wheretheinputvowel/#XUXXXU#qXUXXXUa#XUXXqXU#XUXXXU/doesnotsurfaceinthe _  output.Thiscandidate,however,incursaviolationofMaxIOwhichrequiresthateverysegmentin 7  OaH  OaH  theinputmusthaveacorrespondentsegmentintheoutput#XUXXXU#XUXXXU.Candidate(20b)isinperfect _  correspondencewiththeinputbutincursaviolationof1v#XUXXXU#XUXXXU.Thetwoconstraintsarethusinconflict. 7 #XUXXXU@#XUXXXUTochoosethecorrectcandidate,MaxIOmustdominate1v#XUXXXU#XUXXXU,i.e.MaxIO>>1v.H 9#  13      ׀#XUXXXUN#XUXXXUTheprosodic  regularityexpressedby1v#XUXXXU #XUXXXUԀisthusviolatedunderspecificmorphologicalconditions.#XUXXXU#XUXXXUItis  neverthelessevidentintherestofthelanguageanditwillbeshowntoplayanactiveroleinthemorphologyofthecontinuativeaspect.1P 9#  14       G #XUXXXU # h8XUXXXUTheoutputc1a.c2vc3isotherwiseunrevealing.Thevowel/a/issimplyprefixedtothemajor o syllableofthebaseasrequiredbyhead.Similarly,thesimulfactiveofthecausativevoice, G! c1.ra.c2vc3in(19d)isformedfromthecorrespondingcausativebasec1r.c2vc3in#XUXXXUm#XUXXXU(19c)byaffixation # of/a/accordingtothedemandsofhead#XUXXXU&#XUXXXU.#XUXXXU#XUXXXUԀTheonlydifferencebetween#XUXXXU#@XUXXXUthe#XUXX@XUe#XUXXXUcausativec1.ra.c2vc3#XUXhXU#XUXXXU !% #XUXXXU#XUXXXUandtheactivec1a.c2vc3isthatthebaseoftheformerhasonemoreconsonantinitsminorsyllable,  i.e.causativec1r.c2vc3versusactivec1.c2vc3.Thiscausesthecausativeoutputtocontainonemore  minorsyllable,asinc1.ra.c2vc3.Analternativeoutput,c1ra.c2vc3,withacomplexonsetisexcluded ` because*Complexisundominated.Moreover,c1oranyotherconsonantofthecausativebase 8  cannotbeleftunparsedbecausethiswouldincuraviolationoftheundominatedMaxIO.Itiseasy  ` toseethattheothercausativesimulfactiveoutputpatternofbiconsonantalsin(19d),t.ra.c1vc2,is  8  similarinallrespectstoc1.ra.c2vc3.#XUXXXUk#XUXXXU   Considernowtheactivesimulfactiveofbiconsonantals c1#XUXhXU4#;XUXhXU a#XUXX;XU#XUXXXU.c1vc2#XUXXXU#XUXXXUԀin(19b)#XUXXXUg#XUXXXU.Affixationof#XUXXXU#XUXXXU/a#XUXXXU#XUXXXU/here   isaccompaniedbyacopyofabaseconsonant.Theconstraintheadwillrequirethat/#XUXXXUH#XUXXXUa#XUXXXU#XUXXXU/#XUXXXUH#XUXXXUԀbeina x prefinalsyllable#XUXXXU#XUXXXU,whichisthenrequiredtohaveanonsetbecauseconstraintOnsisundominated. P Thisthenexplainsthepresenceofthenewconsonantintheoutput.Italsoshowsthatthisconsonantisnotpartofsomeoutputtemplatespecifictothesimulfactive,ashasbeenassumedinpreviousanalysesofthesefacts(McCarthy1982,Broselow&McCarthy1983),orevenpartofsomeprosodicrequirementimposedontheshapeofthisparticularaffix.#XUXXXU#XUXXXU  Theaffixisthusonlypartiallyspecifiedintheinputas/a/and#XUXXXU #XUXXXUitssurfacerealizationisleftto  bedeterminedbythegrammarofthelanguage.#XUXXXU #XUXXXUԀItremainstobeexplainedwhytheneededonsetis `  acopyofabaseconsonant.#XUXXXUL #XUXXXUԀ 3Iproposethatthesimulfactiveaffixisreduplicativeinthesensethat 8"  thereisacorrespondencerelationbetweenitandthebase.Thiscorrespondencerelationiswhatdictatescopying.Morespecifically,constraintDepBR#XUXXXU #XUXXXUԀrequiresthattheonsetoftheprefinalsyllable %8!$ beacopyofabaseconsonant.#XUXXXUF#XUXXXUԀHadtheneededonsetbeenadefaultconsonant,asinT#XUXXXU#2XUXXXUa#XUXX2XUn#XUXXXU.c1vc2#XUXXXU#XUXXXU,it '#& wouldhavenocorrespondentsegmentinthebase,aviolationofDepBR#XUXXXU#XUXXXU.#XUXXXU#XUXXXUԀThesituationisdepicted )$( formallyintableau(21)below. p+&* Ї21.Activesimulfactiveofbiconsonantals;copyinginducedbyDEPBR#XUXXXU#XUXXXU J  3 ` x` x*w $d ddd dd dd J$K` x` xw,Mdd ,kdd ,Fdd +  33 D Input:a,c1vc2 +333 +Ons +3$3 +DepBR#XUXXXU#XUXXXU +33$ +a.a.c1vc2 +( x3 +*! .$( x  .*(a) +3( x $ +b.Ta.c1vc2 +  3 + +$   +*(a)*(T)! +3  $ + 3c.+ 򀀀c1 a.c1vc2 +( x3 + +$( x +*(a)#XUXXXU#XUXXXU&( x  $ &` G` 5#XUXXXUT#XUXXXU 3 hSince/a/oftheaffixdoesnotcorrespondtoanybasesegmentthereisaviolationof#XUXXXU#XUXXXUDepBR#XUXXXU#ԚXUXXXUԀforeach   oneofthesecandidates.Candidate(21a)hasaffix/#XUXXXU#aXUXXXU/prefixedtothemajorsyllableoftheword.No j  onsetisprovidedfortheprefinalsyllable,however,whichcausesafatalviolationofOns.Candidate B  (21b)providesanonsetbyepenthesizinganunmarkedconsonant/T/withnocorrespondentinthebase.ThiscausesasecondfatalviolationofDepBR.Finally,candidate(21c)avoidsasecond B violationofDepBR#XUXXXU#XUXXXUԀbycopyingabaseconsonant.Ons#XUXXXU#XUXXXUԀandDepBR#XUXXXU#XUXXXUԀareunrankedwithrespecttoeach  other.#XUXXXUN#XUXXXU   h 3Forcopyingoftheconsonanttotakeplace,anadditionalrankingmustbeestablished.Creatingacopyofasegmentintroducesanotherinstanceoftheoriginalsegment,#XUXXXU#XUXXXUinheritingitsmarkedness.#XUXXXU#XUXXXU R" FollowingPrince&Smolensky(1993),Iwillassumethatsegmentshavemarkednesscharacterizedprimarilybytheirplaceofarticulation.LetPL//standforasegmentwith/#XUXXXUM#XUXXXUԀplaceofarticulation. !R& TheMarkednessHierarchyin(22)directlyexpressesthefactthatcertainconsonants(usuallythecoronals)arelessmarkedthanothersbyarankingofthe*PL//#XUXXXUX#XUXXXUԀconstraints. $ * 22.MarkednessHierarchy 3 3  Somerankingof*PL/Labial,*PL/Dorsal,*PL/Coronal(usuallyinthisorder)#XUXXXUj#XUXXXU =(#0  3Hence,forcopyingtotakeplace,themarkednessviolationofthecopiedsegmentmustnever ~*%4 beseriousenoughtoblockcopying#XUXXXUz #XUXXXUԀoftheconsonant,epenthesizinganunmarked/T/instead.#XUXXXU!#XUXXXUԀInother  words,thedependencerequirementmustberankedhigherthanthemarkednessviolationofthecopiedsegment.Usingthesymbol*PL//#XUXXXU "#XUXXXUMAXforthehighestconstraint(s)inthemarkedness ` hierarchy,therankingensuringthatcopyingisneverblockedbythemarkednessofthecopiedsegmentis#XUXXXU"#XUXXXUԀDepBR#XUXXXU##XUXXXUԀ>>*PL//#XUXXXU=$#XUXXXUMAX#XUXhXU$#XUXXXUԀ(whichisinturnrankedhigherthan#XUXXXU$#XUXXXU*PL//#XUXXXU>%#XUXXXUMIN#XUXhXU%#XUXXXU,theconstraintforthe  ` markednessoftheleastmarkedsegment/T/).#XUXXXU%#XUXXXU  8  #XUXXXUn&#XUXXXUConsidernowthefactthatintherestofthesimulfactiveoutputs,c1a.c2vc3,t.ra.c1vc2,and   c1.ra.c2vc3,nocopyingtakesplace.Inpreviousanalysesthishasbeentakenasevidencethatthe   simulfactiveinvolvesnoreduplicationatallandthatthecopyingoftheconsonantin c1 a.c1vc2isthe p resultofacompletelyunrelatedmechanism,namely,LDCspreading.AcrucialtenetofOT,however,isthatconstraintsaregradientlyviolable.Inparticular,MaxBR,whichrequiresthatevery (x segmentofthebasehaveacorrespondentinthereduplicant,showsdifferentdegreesofviolation:inc1a.c2vc3nosegmentofthebaseiscopied,incurringfourviolationsofMaxBR,andonlyone ( consonantiscopiedin c1 a.c1vc2,incurringtwoviolationsofMaxBR.Someconstraint(s)mustthen  beforcingtheseviolations.#XUXXXU&#  Asdiscussedabove,copiesofsegmentsincurmarkednessviolations.Themoresegmentsarecopiedthelessoptimaltheoutputbecomes.XUXXXUԀ#XUXXXU+#XUXXXUInTemiarcopyingisminimized.Forexample,in @"   c1 #XUXXXU ,#aXUXXXU.c1vc2#XUXXXU,#XUXXXU,aconsonantisrequiredintheoutputbecauseofundominatedOns,butinc1#XUXXXU -#aXUXXXU.c2vc3#XUXXXU-#XUXXXUԀno $h" consonant#XUXXXU-#XUXXXUisrequiredbecausethebasealreadycontainsc1#XUXXXU].#XUXXXUԀthatcanservetheroleoftheneededonset. %H!$ Thusnocopyingtakesplace.Ifitdid,asinthealternativeoutputc1#XUXhXU.#XUXhXU. #XUXXXU/#XUXXXUc2#XUXhXU/#XUXhXU a#XUXXXU40#XUXXXU.c2vc3#XUXXXU}0#XUXXXU,itwouldincurthe ' #& additionalviolation*PL/c2#XUXXXU0#XUXXXU.#XUXXXU]1#XUXXXUMaxBR#XUXXXU1#XUXXXUԀmustthenberankedlowerthanthemarkednessconstraintof )%( theleastmarkedsegment,*PL//#XUXXXU1#XUXXXUMIN#XUXXXU2#XUXXXUԀ>>#XUXXXU2#XUXXXUMaxBR#XUXXXU;3#XUXXXU.Thisrankinghastheeffectofminimizingcopying +&* whichtakesplaceonlywhenthepresenceofanewconsonantisrequiredbyhigherprosodicconstraintsofthelanguage,inthiscase,#XUXXXU3#XUXXXUOns#XUXXXU4#XUXXXU.  Twocrucialranklingshavethusbeenestablished.DEPBR#XUXXXU4#XUXXXUԀ>>*PL//#XUXXXU5#XUXXXUMAX#XUXXXU5#XUXXXU,forcingcopyinginstead ` ofepenthesisofdefaultconsonants,and*PL//#XUXXXU6#XUXXXUMIN#XUXXXU6#XUXXXUԀ>>#XUXXXU 7#XUXXXUMaxBR#XUXXXUO7#XUXXXU,minimizingthenumberofcopied 8  segments.Since,bydefinition*PL//#XUXXXU7#XUXXXUMAX#XUXXXUA8#XUXXXUԀ>>*PL//#XUXXXU8#XUXXXUMIN#XUXXXU8#XUXXXU,theoverallrankingisDEPBR#XUXXXU%9#XUXXXUԀ>>*PL//#XUXXXU9#XUXXXU#XUXXXU9#XUXXXU>>  ` #XUXXXU(:#XUXXXUMaxBR#XUXXXU:#XUXXXU,where*PL//isavariableforanyconstraintofthemarkednesshierarchy 9#  15      .#XUXXXU:#XUXXXU 3Tableau(23)  8  formalizestheprecedingdiscussionintermsoftheproposedconstraints.(#XUXXXU;#XUXXXUIonlyshowtheadditional   violationofmarkednesscausedbycopyingorepenthesisundertheconstraint*PL//#XUXXXUy<#XUXXXU).   23.Activesimulfactiveofbiconsonantals#XUXXXU?=#XUXXXU ,|  3;` x` x*w.$>ddMdd kdd Fdd $` x` xw,Sdd ,dd ,Fdd ,'dd ,dd +  33 #s Input:a,c1vc2 +333 +Ons +3$3 +v=DepBR )33$ )*PL//#XUXXXU=#XUXXXU )33 )MaxBR#XUXXXUa@#XUXXXU +33 +a.a.c1vc2 +W3 +*! .$W .*uXXU(a)#XUXuA# %W$ % %W %*** +3W +b.Ta.c1vc2 +3 + +$  +*uXXU(a)#XUXuB##XUXXXU@#XUXXXU*uXXU(T)#XUXu_C#! %!$ %*uXXU(T)#XUXuC# %" %*** +3# +c.+ 򀀀c1 a.c1vc2 +W$3 + +$W% +*uXXU(a)#XUXuD# %W&$ %*uXXU(c1#uu?C#uu)#XUXuQE# %W' %**#XUXXXUE#XUXXXU +3W( +d. c1vc2#XUXXXU)F#XUXXXU .a#XUXXXUF#XUXXXU.c1vc2#XUXXXUF#XUXXXU +)3 +*! +$* +*uXXU(a)#XUXuG# .$+$ .*uXXU(c1#uuJG#uu)#XUXu6H#*uXXU(v)#XUXuH#*uXXU(#uuH#uuc2#uuNI#uu)#XUXu I##XUXXXUI#XUXXXU .$, .#XUXXXUI#XUXXXU .3- . e.+#XUXXXUnJ#XUXXXUԀ c2#XUXhXUJ#XUXXXU a#XUXXXUHK#XUXXXU.c1vc2#XUXXXUK#XUXXXU ._.3 . .$_/ .*uXXU(a)#XUXu^L# %_0$ %*uXXU(c2#uuK#uu)#XUXuL##XUXXXUM#XUXXXU %_1 %**&_2   &` >` 0>#XUXXXUzM#XUXXXU J hCandidates(23ac)havealreadybeendiscussedintableau(21).#XUXXXU-N#XUXXXUCandidate(23d)copiesthewhole J4 base,satisfyingMaxBR#XUXXXUN#XUXXXUԀcompletelybutattheexpenseofviolatinghigherrankedconstraints0@ 9#  16      . "!r6 Finally,candidate(23e)issofarpredictedtobe#XUXXXUmO#XUXXXUanalternativeoptimaloutput(atleastinthosecases "J8 where*PL/c1#XUXXXUP#XUXXXUԀ>>*PL/c2#XUXXXU1Q#XUXXXU). $" :  hcNIthusturntothequestionofwhatdeterminesthechoiceoftheconsonanttobecopied,namely,thechoicebetweenthetwopossiblecandidates c1 a.c1vc2#XUXXXUQ#XUXXXUԀand c2 a.c1vc2#XUXXXUR#XUXXXU.ApartfromDepBR#XUXXXUS#XUXXXUԀ#XUXXXUgS#XUXXXU,thecopied  consonantsinthesetwocandidatessatisfyanotherconstraintofcorrespondencetheory,whichrequiresthatcorrespondentsegmentsbefeaturallyidentical.Thisinfactisaconstraintfamily,IdentBR(F)#XUXXXUS#XUXXXU,satisfiedhereforeveryfeatureF.Theconstraints#XUXXXUT#XUXXXUIdentBR(F)#XUXXXU]U#XUXXXUԀthuscannotbeinvolved  h inthechoicebetweenthetwocandidates.#XUXXXUU#XUXXXUԀ#XUXXXUMV#XUXXXUAsidefromtheirfeaturalcomposition,however,segments  @  intheoutputcomeequippedwithotherprosodicpropertiessuchastheirsyllabicroles.Iproposethatthisisthepropertyofsegmentswhichdeterminesthechoicebetweenthetwocandidates.Puttingalternativesasideforthemoment,therelevantconstraintisSrole,#XUXXXUV#XUXXXUrequiringthatcorrespondent x segmentshaveidenticalsyllabicroles.#XUXXXU5X#XUXXXU P  3Tableau(24)showshowSrole#XUXXXUX#XUXXXUԀdeterminesthechoiceofthecopiedconsonant.#XUXXXUZY#XUXXXUԀIn(24a)the (x copied c2 isparsedasanonsetwhilec2#XUXXXUY#XUXXXUԀinthebaseisparsedasacoda.#XUXXXUiZ#XUXXXUIn(24b)bothc1#XUXXXUZ#XUXXXUԀanditscopy P areparsedasonsets#XUXXXU&[#XUXXXU.InTemiarSrole#XUXXXU[#XUXXXUԀ#XUXXXU[#XUXXXUisneverviolated,andthus#XUXXXU:\#XUXXXUԀIwillassumeitisundominated.#XUXXXU\#XUXXXU 0 24.Choiceofcopiedconsonant#XUXXXU\#XUXXXU t ` x` x*$ddSdd dd Fdd 'dd dd .$>` x` x,Sdd ,dd +  33 k  3YInput:#XUXXXU]#4XUXXXUԀa#XUXX4XU_#XUXXXU,c1vc2 )333 )Srole#XUXXXUV_#XUXXXUԀ#XUXXXU_#XUXXXU +3 3 +/]a. c2#XUXhXU"`# -XUXXXUa#XUXX-XU`#.c1vc2 %O !3 %*! .3O " . 3b.+ c10XUXhXU a#XUXX0XUa#XUXXXU.c1vc2 %!##3 %#XUXXXUa#XUXXXU&!#$   &` ]` ]#XUXXXUOb#XUXXXU 3ha 3Notethatthereisapotentialalternativetotheanalysisjustproposedintermsoftheconstraintshownin(25).#XUXXXUb#XUXXXU &e!( 25.Anchoring#XUXXXUc#XUXXXU(McCarthyandPrince1994a)#XUXXXU6d#XUXXXU (#,  3cc 3Correspondencepreservesalignmentinthefollowingsense:theleft/rightperipheralelementoftheReduplicantcorrespondstotheleft(right)peripheralelementofBase,iftheReduplicantistotheleft/rightoftheBase.#XUXXXUd#XUXXXU x+&0 e ,m'1  3dAnchoring#XUXXXUf#XUXXXUԀismeanttocaptureageneralizationaboutthecopyingandassociationrulesof  operationaltheoriesofreduplication(Marantz1982,McCarthy&Prince1986).Thegeneralizationisthatreduplicativeprefixescopybasematerialfromtheleftedgeofthebase,whilereduplicativesuffixescopyfromtherightedgeofthebase.#XUXXXUf#XUXXXUԀThecontinuativeforms,however,providethecrucial 8  evidencethatthisisnottherelevantconstraintinTemiar.#XUXXXU^h#XUXXXUWhenthecopiedconsonantisplacedin  ` onsetposition,asinthesimulfactive c1#XUXhXU'i#ZXUXhXU a#XUXXZXUi#XUXXXU.c1vc2#XUXXXUj#XUXXXU,thecopystartsfromtheleftmostsegmentofthebase.  8  Butwhenthecopiedconsonantisplacedatthecodaposition,asinthecontinuativec1#XUXXXUpj#XUXXXU c3#XUXXXU\k#XUXXXU .c2vc3#XUXXXUk#XUXXXU,itis   therightmostsegmentofthebasethatischosenforcopying.Thisthenshowsthatitisnottheedgeofthebasethatiscrucialherebuttheprosodicroleofthecopiedsegment(i.e.#XUXXXUl#XUXXXUinTemiarSrole>>  Anchoring#XUXXXUm#XUXXXU;intheSemiticpatternsdiscussedinsection6,theinverserankingwillbeatwork). X  3Turningtotheanalysisofthecontinuativeaspect,considerthefourpatterns,twofortheactiveandtwoforthecausativevoice,shownin(26).#XUXXXUm#XUXXXU X #XUXXXUn#XUXXXU#XUXXXU>o#XUXXXU26.ACTIVE   Biconsonantal󀀀   p Triconsonantal x    3 n  a.Base   c1vc2 h      p c1.c2vc3 6 ;p   `    k<<wtocall     p s.l<gtoliedown " p  b.Continuative  c1c2 .c1vc2     p c1 c3 .c2vc3 4 q   `    k w .k<<w     p s g .l<g $ t    CAUSATIVE  Biconsonantal󀀀   p Triconsonantal x  !" r  c.Base   tr.c1vc2     p c1r.c2vc3 ">$    `    tr.k<<w     p sr.l<gs #*%   d.Continuative  t.r c2 .c1vc2     p c1.r c3 .c2vc3 %!'    `    t.r w .k<<w     p s.r g .l<g &!( #XUXXXUo#u XUXXXÙItisclearfromallfourcontinuativeoutputsthatthechoiceofthecopiedconsonant(s),isdeterminedbySrole.#XUXXXUv#XUXXXUԀTherearetwootherinterestingobservationsthatcanbemadeaboutthese V*%- patterns,statedin(27). v .,~'/ ?TH  27. ` a.Onlyconsonantsarecopied(i.e.thebasevowelisnevercopied).  Sx 3   ` b.Thenumberofcopiedconsonantsvaries.Inthecaseof#XUXXXUhw#XUXXXUc1vc2#XUXXXU}y#XUXXXUԀtherearetwo b    ` consonantscopied.Inallothercasesthereisonlyoneconsonantcopied.#XUXXXUy#XUXXXU N x 3x hRegarding(27a),considerthecontinuativeoftriconsonantalsin#XUXXXUz#XUXXXUԀs g .l<g.#XUXXXUs{#XUXXXUԀRecallthatthecontinuative = affixisrequired,#XUXXXU{#XUXXXUunderhead#XUXXXUM|#XUXXXU,to#XUXXXU|#XUXXXUbeprefixedtothemajorsyllableofthebase.#XUXXXU|#XUXXXUAsinthecaseof   thesimulfactive,Iwillassumethatthecontinuativeaffixisreduplicative.Iargueherethatthecontinuativeaffixshouldnotbespecifiedforanysegmentalcontent,beingsimplyanemptyRoot?TH  ?TH  nodewhoserealizationisdeterminedbythegrammarofthelanguage.Notingthattheaffixisinvariantlyrealizedwithacopyofatleastonebaseconsonant,itcouldperhapsbespecifiedtobesomesortofconsonantalsegment 9#  17      .However,thisfactfollowsfromtheregularprosodyofthe U language.Indeed,iftheaffixwasrealizedbyacopyofavowel,#XUXXXU\}#XUXXXUasecondsyllablewithavowel - wouldbecreated,aviolationof1v.Thelanguageevadesthisviolationbyrealizingtheaffixwith  aconsonant.(Iarguebelowthattheaffixalsolacksaprosodictarget).#XUXXXU#XUXXXU   h{ 3Thesituationis#XUXXXU#XUXXXUexpressedformallyintableau(28),wherethesegmentrealizingtheaffix#XUXXXU#XUXXXUԀis e underlined,andboldfacedifitisacopyofsomeothersegment.Intheinputshowninthistableau,#XUXXXU #XUXXXUԀindicatesthecontinuativeaffix,aRootnodewithnosegmentalcontent| 9#  18      .#XUXXXUу#XUXXXUԀ e К#XUXXXU#XUXXXU28.Continuativeoftriconsonantals;1vinaction#XUXXXU #XUXXXU !"  3<` x` x 3 *fw$xddSdd dd $` x` xf,4dd ,dd ,dd ,dd +  #XUXXXU#XUXXXU 33 "$ Input:,#XUXXXU#XUXXXUc1.c2vc3#XUXXXU#XUXXXU +3#T%33 +1v#XUXXXU#XUXXXU +3$#T&3 +DepBR .3$#T'3$ .*PL//#XUXXXUZ#XUXXXU 73$#T(3$ 7)a. 򀀀 󀀀c1 v .c2vc3 4!$)3 4*! +$$* + .$$+$ .*(v#XUXXXU#XUXXXU) 73$$,$ 7b.c1#XUXXXUk#XUXXXUT.c2vc3#XUXXXU#XUXXXU 4!&X!-3 4 +$&X!. +*(T)! .$&X!/$ .*(T#XUXXXU]#XUXXXU) 73$&X!0$ 7c.+c1#XUXXXU4# XUXXXUc3 .c2vc3 4!'"13 4 +$'"2 + .$'"3$ .#XUXXXUˌ#XUXXXUԀ*(c3#XUXXXU#XUXXXU)&'"4  $ &OaH  ` ` #XUXXXU #XUXXXU 3 hCandidate(28a)realizestheaffixwithacopyofthebasevowel/#XUXXXUÎ#XUXXXUv#XUXXXUm#/,aviolationof1vXUXXXU,while#XUXXXUˏ#XUXXXUԀ(28b) D fillsitwithadefaultconsonant/T/#XUXXXU#XUXXXU,aviolationofDepBR#XUXXXU#XUXXXU.Finally,(28c)avoidsbothviolationsby  l copyingaconsonantofthebase.#XUXXXU#XUXXXU  D #XUXXXU# hXUXXXURecallthat1v#XUXXXU#XUXXXUԀisdominatedbyMaxIO#XUXXXU]#XUXXXU,andthusviolatedinallsimulfactiveoutputsas    establishedearlierintheanalysis.Inthecaseofthecontinuatives,ontheotherhand,1v#XUXXXU#XUXXXUԀplaysan   activeroleindeterminingtheoptimalcandidate.ThisdifferencearisesfromthefactthesimulfactiveOaH  OaH  ispartiallyspecifiedas/a/,whilethecontinuativehasnoprespecifiedphonologicalcontent.Itisthuslefttothegrammartodeterminethecontentoftheaffix,andhenceconstraintsdeterminingtheregularprosodyofthelanguage,like1v#XUXXXU#XUXXXU,playanactiveroleinchoosingtheoptimalcandidate.#XUXXXUו#XUXXXU T #XUXXXUQ#XUXXXUThesecondobservationnoted#XUXXXU#XUXXXUin(27b)highlightsthedifferencebetweenthetwoactive , continuativepatternsin(26b),namely,thebiconsonantal c1c2 .c1vc2andthetriconsonantalc1 c3 .c2vc3.  Asinthecorrespondingsimulfactives c1 a.c1vc2andc1a.c2vc3,thebiconsonantalscopyonemore  consonantthanthetriconsonantals.Thisisbecauseinthecaseofbiconsonantalstheaffixisrealizedas/... c2 .c1vc2/,accordingtothedemandsofheadand1v,andtheprefinalsyllablethenneeds L anonsetbecauseOnsisundominated.Thisonsetisprovidedbycopyingabaseconsonantforthe ,!|  samereasonsasinthesimulfactive,namely,DepBR.Asinthecaseofthesimulfactive#XUXXXU#XUXXXUԀ c1 #XUXhXUњ#!XUXXXUa#XUXX!XU##XUXXXU.c1vc2#XUXXXUf#XUXXXU,the #T" copiedconsonantisnotpartofsomeoutputtemplatespecifictocontinuativeformationorpartof#XUXXXU#XUXXXUsomespecificationontheprosodicshapeoftheaffixitself.ItisinsteadrequiredbyOns,an & "& Лinviolatedprosodicpropertyofthelanguage. 3Theemergenceofthebiconsonantalactivecontinuativeform c1c2 .c1vc2isexpressedintableau(29),evaluatingthemorerelevantcandidates. l*%* #XUXXXUy#XUXXXU + ', 29.Continuativeofbiconsonantals(affixrealizationsareunderlined)#XUXXXU7#XUXXXU   3B*G$Odd4dd dd dd dd w$x(#(#,Hdd ,Ddd ,dd ,edd ,dd +     +$33R +ЀInput:,c1vc2 +3R$33 +Ons#XUXXXU#XUXXXU +3$R3 +DepBR#XUXXXU#XUXXXU )3R3$ )*PL//#XUXXXU#XUXXXU %R 3 %#XUXXXU#kXUXXXU +$3  +#XUXXXUs#XUXXXUa.D c2 .c1vc2 + $3 +*! +$  + % $ %*(c2)#XUXXXU#XUXXXU % %#XUXXXUޣ#XUXXXU +$3 V +#XUXXXUE#XUXXXUb.TDT.c1vc2 + V$3 + +$ V +*!* % V$ %*(T)*(T)#XUXXXU#XUXXXU % V %#XUXXXU#XUXXXU +$3  +#XUXXXU#XUXXXUc.TD c2 .c1vc2 + $3 + +$  +*! % $ %*(T)*(c2)#XUXXXUt#XUXXXU %  %#XUXXXUv#XUXXXU +$3 Z +#XUXXXUݧ#XUXXXUd. 򀀀c1 DT.c1vc2 + Z$3 + +$ Z +*! % Z$ %*(T)*(c1)#XUXXXUJ#XUXXXU % Z %#XUXXXUQ#XUXXXU .$3  .#XUXXXU#XUXXXU 3e.+ c1 D c2 .c1vc2 + $3 + .$  . % !$ %*(c1)*(c2)#XUXXXU(#XUXXXU& "   &#XUXXXUP#XUXXXU 3G hCandidate(29a)realizestheaffixwithabaseconsonantplacedinthe#XUXXXU#XUXXXUcodapositionofanonsetless H $ syllable,causingafatalviolationofOns(thephoneticminorsyllablevowel[D]isshowninthis  & tableautomakeclearthat c2 isplacedinthecodapositionasrequiredbyhead).Candidates(29b ( d)realizeatleastoneoftheconsonantsoftheprefinalsyllablebyepenthesizingasegmentTwithnobasecorrespondent.ThiscausesatleastoneDepBRviolation.Theoptimalcandidate(29e)copies `, bothconsonantsofthebase,avoidingallDepBRviolationsj 9#  19      .#XUXXXUf#XUXXXU 8.  h#XUXXXUt#XUXXXUInthetriconsonantaloutput,c1 c3 .c2vc3,recallthattheaffixisplacedattheposition/... c3 .c2vc3/ `0 andrealizedwithaconsonantasestablishedintableau(28)above.Thebaseincludesanotherconsonant,c1,whichcanserveasanonsetoftheprefinalsyllable,i.e.c1 c3 .c2vc3,andthusno 4 additionalcopyingisnecessary.Thesameappliestotheothertwocontinuativepatternsofthecausativevoice,t.r c2 .c1vc2andc1.r c3 .c2vc3.Placementoftheaffixisdeterminedbyheadandits #8  Oa   Oa   Oa  realizationasaconsonantbytheregularprosodyofthelanguage,namely,1v.#XUXXXUݯ#  Thisconcludesthemainpartoftheanalysis.Tosumup,thesimulfactiveandcontinuativeaffixesarebothspecifiedtobeaRootnode.TheonlydifferencebetweenthetwoisthatinthesimulfactiveaffixtheRootnodeisfurtherspecifiedas/a/.XUXXXUԀMoreover,thetwoaffixesobeyacommon 8  placementconstraint#XUXXXUմ#XUXXXUhead#XUXXXUi#XUXXXU.#XUXXXU#XUXXXUThesurfaceshapeoftheaffixbasecombination#XUXXXU#XUXXXUԀemergesfromthe  ` interactionofhead#XUXXXUo#XUXXXUԀwith#XUXXXU#XUXXXUtheregularprosodyofthelanguage,thatis,mainlytheconstraintsOns#XUXXU=#XUXXU  8  and1v#XUXXXUʷ#XUXXXU.#XUXXXU1#XUXXXUԀCopyingofsegmentsisinducedbythecorrespondenceconstraintDepBR#XUXXXUt#XUXXXU.Thenumberof    Oa   Oa  copiedsegmentsisminimizedbecausethesegmentalmarkednessconstraints*PL//areranked   higherthantheotherbasiccorrespondenceconstraintMaxBR#XUXXXU#XUXXXU.Twomoreconstraints,IdentBR(F)#XUXXXU#XUXXXUԀ#XUXXXU#XUXXXUand p Srole,#XUXXXU@#XUXXXUԀrequireidentityintermsoffeaturalcompositionandsyllabicrolebetweencorrespondent H segments.#XUXXXU#XUXXXU  p #XUXXXU`#XUXXXU 3Somefurtherrefinementsintherankingofcorrespondenceconstraintsaremadepossibleby H certainregularitiesnotedby#XUXXXU#XUXXXUBenjamin.Thefirstregularityisthatavoicelessstopinthefirstposition   ofamedialconsonantalclusterundergoesnasalizationinNorthernTemiarandvoicinginSouthernTemiar#XUXXXU#XUXXXU.Theeffectsofthisregularitybecomeevidentintheformationofthecontinuative,asshown  in(30ab)below,whereabasefinalvoicelessstopiscopiedandplacedinpreconsonantalposition.#XUXXXU#XUXXXU X  30.CodaNasalization/Voicing#XUXXXU#XUXXXU "<"  3ּ   `  #XUXXXU3#XUXXXU  Verb h   Gloss   p Surfaceform 6$$ a.NorthernTemiar  yaap h   tocry p y m .yaa p  % &  3b.SouthernTemiar  <<t h   tofast p  d .<< t#XUXXXU#XUXXXU  &!'  3Iwillassumethataconstraintoncodasisinvolvedhere,requiringthatcodaconsonantsbeaboveacertainsonoritylevel,andhencecausingunvoicedstopstobecomenasalizedwhentheyareincodaposition#XUXXXU#XUXXXU:Codcond(seeIt=&Mester1993onthedetailsofhowtoformalizeconstraints ),y'. :Oa  :Oa  :Oa  likethisinOT).ThisconstraintisnowinconflictwithIdentBO(F)#XUXXXU#XUXXXUԀwhichrequirescopiedsegments  tobefeaturallyfaithfultotheircorrespondents.TherequiredrankingsmustthereforebeCodcond#XUXXXU#XUXXXU  >>IdentBR#XUXXU#XUXXU(nasal)forNorthernTemiar,andCodcond#XUXXXU#XUXXXUԀ>>IdentBR#XUXXUc#XUXXU(voice)#XUXXXU#XUXXXUԀforSouthernTemiar.#XUXXXU#XUXXXU ` Anotherregularityisthatnasalsassimilatetotheplaceofarticulationofafollowingstop.Examplesaregivenin(31ab)withthenasalstopclustersunderlined#XUXXXU\#XUXXXU.Benjaminnotesthatthereare  ` exceptionstothisregularity,asshownby(31cd).AllexamplesarefromtheNortherndialect.#XUXXXUW#XUXXXU 331.Verb `    Gloss h      p #XUXXXU5#XUXXXU  Z   3T  #XUXXXU#7XUXXXUa.kandee   we(theyandI)(emphatic) T    b.c.rd.kubna  thatlid(affix/n/withthebasecr.kub,[d]isavelarnasal) @   c.pn.pBt   tolongfor(continuativeof p pBt) ,|  3  d.gn.gr.lut   spindlyness(expressiveofgr.lut)#XUXX7XU#XUXXXU h E 3The#XUXXXU#XUXXXUinterestingfactabouttheseexceptionsisthattheyareattestedpreciselywhenthenasal W consonantisderivedbyacopyofabaseconsonant,nasalizedduetotherankingCodcond>> / IdentBR(nasal).Thenonhomorganic/n/isanasalizedcopyof/t/inthecontinuativeof(31c)and  theexpressiveof(31d)^ 9#  20      .Thuswhenacopiedsegmentisplacedinthecontextwhereassimilation  wouldnormallyoccur,thesegmentretainsitsplaceidentity,remainingfaithfulinthisrespecttoitscorrespondentinthebase.AssumingthatnasalplaceassimilationisimposedbyaconstraintNc, ?   thesefactsmotivatetherankingIdentBR(PL)>>Nc.&@ 9#  21      #XUXXXU>#XUXXXU "g" ЛAcoupleofconclusionsfollow.First,thesefactscanbeequallydescribedinarulebased #?$ approachbyorderingaruleofnasalplaceassimilationbeforethemorphologicaloperationofcontinuativeformation,whichisinturnorderedbeforearuleofcodanasalization.Whenthecontinuativeisformed,thenasalplaceassimilationrulewouldbeinapplicablealthoughthecodanasalizationrulecreatestheenvironmentforitsapplication.Hencearulebasedframework,sofar,appearstorequirenomoremachinerythantheconstraintbasedapproach,whichalsoutilizesthree#XUXXXU5#XUXXXUconstraintswiththerankingCodcond,IdentBR(PL)>>Nc.However,whenanasalinfixisinserted  8  inthebase,asinc.rd.kubnathatlid(affix/n/withbasecDrkub)of(31b)#XUXXXU#XUXXXU,thenasaldoes   assimilatetotheplaceofarticulationofthefollowingstopk 9#  22      #XUXXXU;#XUXXXU.Toaccountforthisfact,therulebased   approachneedstopostulateanadditionalorderingrelation:theruleofnasalinfixationmustprecedetheruleofnasalplaceassimilation.Intheconstraintbasedapproach,ontheotherhand,noadditionalmoveisnecessary.ThenasalinfixassimilatestothefollowingstopbecausethenasaldoesnotderivefromacopyofabasesegmentandthusIdentBR#XUXXU5#XUXXXU(PL)#XUXXXU9#XUXXXUԛdoesnotapply.#XUXXXU#XUXXXU 9#  23       H Second,furtherrefinementsintherankingofthecorrespondenceconstraintshavebeendeduced,whichinturnprovidesupportforoneofthemainanalyticalchoicesofCorrespondenceTheoryofMcCarthy&Prince(1995a).Inparticular,thefeaturalidentityrequirementmustindeedbeseenasafamilyofconstraints#XUXXXU#XUXXXUIdent#XUXXXU#XUXXXU(F)foreveryfeatureF.Thisisrequiredbecausethefeatural X  correspondenceconstraintsarerankeddifferently.Identityofplacefeaturesismoreimportantthan 0"   ?TH   ?TH   ?TH  identityofotherfeaturessuchasvoicingandnasality.#XUXXXU#XUXXXU  #XUXXXU#XUXXXUThenexttwosectionsofthepaperwillfurthermotivateandjustifythecorrespondenceapproach  tocopying,comparingitwithalternativeswhichusetheadditionalmechanismofspreadingtocreatecopiesofsegments#XUXXXU3#XUXXXU.#XUXXXUa#XUXXXUԀSection4focusesonevidencefromTemiar.Section5discussescrosslinguistic 8  evidence. 3  `  34 h  4.TemiarinPreviousAnalyses  2  Inthissection,IconsiderpreviousanalysesoftheTemiarverbalaspect,addressing% _ Ԁthesubstantivedifferencesbetweentheproposedoptimalitytheoreticsolutionandthoseearlieranalyses.#XUXXXU#XUXXXUThefocus 2 willbeonMcCarthy(1982),brieflyconsideringalsoBroselow&McCarthy(1983)andSloan(1988).#XUXXXU#XUXXXUThecommoncharacteristicofallpreviousapproachesistheiruseoftemplatesand  associationrules,whichmapunderlyingmelodicsequencestothesetemplates,parametrizedforwhetherassociationproceedslefttorightorrighttoleft.#XUXXXU#XUXXXUForbrevity,Irefertothisapproachasthe B TemplateandAssociationApproach(henceforthTA).#XUXXXU#XUXXXUԀThissectionaimstoshowthattemplatesare j notnecessaryandthatassociationrulesmisscertainsignificantgeneralizationsinthedata.#XUXXXU#XUXXXU B  hq 3TheanalysisofTemiarintheTAapproachassumesthatcopyinginthesimulfactiveandcontinuativepatterns#XUXXXUp#XUXXXUistheeffectof#XUXXXUX#XUXXXUtwoentirelydifferentmechanisms,namely, spreadingand !  reduplication,respectively.Specifically,theanalysisofthesimulfactivestipulatesaprosodictemplateCVCV(V)Cwhosefirstvocalicpositionisoccupiedbythesimulfactiveaffix/a/.In(32),Iconsiderthederivationofthebiconsonantalbasek<<w#XUXXXU#XUXXXUtocall. *'z"& Ӏ32.k<<w[kak<<w]   h a )%*  3   `     h |     p    c*%+   ProsodicTemplate:   h CVCVVC p #XUXXXUE#XUXXXU    #XUXXXU#XUXXXU O+&,    `     h ||/| p   #XUXXXUo#XUXXXU ;,'- #XUXXXUD#XUXXXU 3     #XUXXXU# XUXXXU h k<w '-w(. "Oa H  #XUXX XU#XUXXXU 3#XUXXXU#XUXXXUThebasemelodyisfirstassociatedinalefttoright(LR)directiontothepositionsofthis J template.TheemptytemplaticpositionC,whichremainsafterthisfirstassociationstep,isfilledbyspreadingofthebaseinitialconsonant/k/,aninstanceofLDCspreadingapplyingovertheinterveningvowel/a/.LDCspreadingisallowedbecauseofthegeometryoftherepresentation,namely,segregationoftheaffix/a/andthesegmentsofthebase/k<<w/totwodifferentplanes.This Z   representationisaspecialcaseofplanarsegregationrequiringthatdifferentmorphemeslieondifferentplanes,introducedinMcCarthy(1979)andextendedlatertoV/Cplanarsegregationin"Oa H  "Oa H  McCarthy(1989).#XUXXXU8#XUXXXU 2  3Thederivationofthecontinuativepatterns( c1#XUXXXU#XUXXXUc2 .c1vc2andc1#XUXXXUi#XUXXXU c3 .c2vc3)#XUXXXU#XUXXXUԀismoreinvolvedanduses   adisparatenumberofmechanisms,givenin(33).#XUXXXU9#XUXXXUԀ   333.a.Morphemictemplate:[RootRoot]  b.Prosodictemplate:CCCVC  c.ContinuativeAssociationRule:Associatethelastelementinthefirstcopyoftherootwith V    ` thesecondC-positionoftheprosodictemplate.#XUXXXU#XUXXXUԀAfterthisruleisapplied,association B    ` proceedslefttoright. 3  d.MorphologicalOpacity:Allsegmentsofthebasemustappearintheoutput.#XUXXXU# 3 3XUXXXU  & 3 3Ibrieflyillustratetheuseofthesemechanismswithanexample,thecontinuativeoftheverb#XUXXXU#XUXXXUs.l<gtoliedown,sg.l<g.Themorphemictemplate[RootRoot]in(33a)stipulatesthattheoutput !" mustconsistoftwocopiesofthebasemelody,i.e.[sl<gsl<g].Thesegmentsinthesetwocopies i#$ mustthenbeassociatedtotheprosodictemplateCCCVC,in(33b).Theassociationproceduremustbeginwithaspecialassociationstep,statedinthecontinuativeassociationrule,(33c),whichessentiallystipulatesthatthefinalsegmentofthefirstcopyofthebasemustbelinkedtothesecondCpositionintheprosodictemplate,givingC#XUXXXU#XUXXXUgC#XUXXXU_#XUXXXUVC.Associationoftherestofthesegmentsisthen *&, initiatedintheunmarkedlefttorightdirection,andanadditionalprovision,calledMorphologicalOpacityin(33d),isrequiredtoderivethecorrectoutput,shownin(34)below.  9#  24        34.sl<g[sgl<g]uXXU h #XUXud#  l  3-  ProsodicTemplate:   h   CCCVC  c   MorphemicTemplate: h   sl<gsl<g#XUXXXU#XUXXXU u    3  Output:#XUXXXU#XUXXXU    h   [sgl<g]#XUXXXU8#XUXXXU) 3  '  Theproblemwiththisaccountisthecontinuativeassociationrulein(33c).Thisruleisadhocbecauseitmakesreferencetoarbitrarytemplaticpositions,bystipulatingthatthelastelementinthe   firstcopyoftherootmelodyassociatestothesecondpositioninthetemplate.#XUXXXU#XUXXXUԀItalsomissesa _  generalizationevidentinallcontinuativepatterns.Thesecondconsonantintheoutputs c1#XUXXXU#XUXXXUc2 .c1vc2 7 andc1#XUXXXUM#XUXXXU c3 .c2vc3#XUXXXU#XUXXXUԀisacopyofthebasefinalconsonantbecauseitisinacodaposition,justlikethebase g finalconsonant.#XUXXXU0#XUXXXUThedecisionofwhichconsonanttocopydoesnotrequireanyadhocruleof G association,butfollowsfromtheuniversalcorrespondenceconstraintSrole,requiringidentity  betweenthesyllabicrolesofthecopiedsegmentanditscorrespondentinthebase. 3Consideringthesimulfactiveandcontinuativepatternstogether,theargumentagainsttheparametricrulebasedapproachcanbetightenedfurther.Thecrucialfactisthatwhenthecopiedconsonantisplacedinonsetposition,thefirstconsonantofthebaseischosen,asinthecaseofthesimulfactivein(35a)below.However,whenthecopiedconsonantisplacedinacodaposition,thefinalconsonantofthebaseischoseninstead,asinthecaseofthecontinuativein(35b).AnyTA $/ & approachneedstwodifferentrulesofassociationwithincompatibledirectionalsettingsV 9#  25      #XUXXXU#XUXXXU.This  missesthegeneralizationevidentinbothaspects,whichisdirectlycapturedbythecorrespondenceapproach,requiringthatcopiedsegmentsmusthaveidenticalsyllabicrolesastheircorrespondents.#XUXXXU!#XUXXXU ` #XUXXXUF#XUXXXU35. ` Aspect   h   Output p   DirectionofAssociation  D  3A   ` a.Simulfactive h   c1 #XUXhXU#XUXhXUa.c1vc2#XUXXXU#XUXXXU p   Lefttoright  ;   3   ` b.Continuative h   c1#XUXXXU@ #XUXXXU c3 .c2vc3#XUXXXUA #XUXXXUԀ#XUXXXU #XUXXXU p   Righttoleft S     3 Broselow&McCarthy(1982)alsouseaspecialassociationruletoderivethecontinuativeforms,#XUXXXU #XUXXXUalthoughthemechanicsoftheiranalysisareslightlydifferent.Inparticular,derivationofthe p  continuative/sgl<g/isexecutedinthefollowingway.AreduplicativeinfixCisinsertedinthe H contextC_CVC,i.e./s_l<g/.Thisinfixinducescopyingofthemelodyofthebase,and  p  autosegmentalassociationisstipulatedtoproceedfromrighttoleft,yielding/sgl<g/(Broselow H &McCarthy1982:39).Finally,Sloan(1988)analyzesthecontinuativeasinvolvingtheprefixationofaminorsyllableconsistingoftwoconsonants,CC,tothebase.Acopyofthebase#XUXXXU #XUXXXUԀmelodyisthen  createdandassociatedtotheCCtemplate.Hereagain,thefirststepofthisassociationprocedureisstipulatory,employinga SpecialAssociationPrinciplewhichassociates therightmostelementofthecopytotheaffixaltemplate(Sloan1988:321).Withoutconsideringthedetailsoftheseanalyses,itisclearthataconsiderableamountofstipulationisinvolved,missingtheimportantgeneralizationsuncoveredbytheanalysispresentedintheprevioussection.#XUXXXU#XUXXXU #0& #XUXXXU#XUXXXUFinally,anotherproblemwithTAapproachesisthestatementofthetemplatesthemselves. %!( #XUXXXUt#XUXXXUPreviousanalysesstipulatethattheoutputsofthesimulfactiveandcontinuativeaspectsarethe  templatesC#XUXXXU*#XUXXXUa#XUXXXU#XUXXXU.CVC#XUXXXU0#XUXXXUԀandCC.CVC#XUXXXUv#XUXXXUԀrespectively.Asseeninsection3,however,thepresenceofthe  firstconsonantinthesimulfactivetemplateC#XUXXXU#(XUXXXUa#XUXX(XU#XUXXXU.CVC#XUXXXU#XUXXXUԀisrequiredbecausetheprefinalsyllable,created ` byprefixing#XUXXXU#XUXXXU/#XUXXXU#XUXXXUa#XUXXXU#/tothebaseXUXXXUc1vc2#XUXXXU>#XUXXXU,musthaveanonset.#XUXXXU#XUXXXUԀ#XUXXXU#XUXXXUThisfirstCpositionofthetemplateisthusnot 8  arbitraryandshouldnotbestipulatedassuch.#XUXXXU+#XUXXXUSimilarly,inthecontinuativeoutputCC.CVC,the  ` secondCpositionistherealizationofanaffixalRootnodealignedattheleftedgeofthefinalsyllableofthebase,apropertycommontoallsimulfactiveandcontinuativepatterns.Moreover,thefirstconsonantpositioninthetemplateisrequiredbecause,asinthesimulfactive,theprefinalsyllablemusthaveanonset.#XUXXXU#XUXXXUԀTheinteractionofaffixationwithindependentprosodicrequirements p ofthelanguage,suchasOnsand1v,derivestheshapeoftemplatesthatpreviousanalyseshadto H stipulate.#XUXXXU#XUXXXUԀThisresultisanotherexampleofanatemplaticanalysis,inthesensethatnotemplatesare  p neededtostipulatetheshapeoftheoutputoroftheaffix.OtheratemplaticanalyseshavebeenpresentedinMcCarthy(1993)forpartofArabicandAkkadian,andArchangeli(1991)forYawelmani.#XUXXXU#XUXXXU (Ireturntothispropertyoftheanalysisinthetypologicalconsequencesexploredin  section6.)     5.#XUXXXUT#XUXXXUOntheNecessityofEliminatingLDCspreading  8"    h 3InthissectionIarguethattheproposedreductionofLDCspreadingtocopyingviacorrespondenceisnotonlypossible,butalsonecessary.Section5.1reexaminesthetraditionalargumentfortheneedforbothreduplicationandLDCspreadinginthetheory.Section5.2presentsindependentevidenceforeliminatingLDCspreading. 3ӹ h 3  #XUXXXU=#XUXXXU p+&*  3I[ h 5.1.TheApparentNeedforReduplicationandLDCSpreading #XUXXXU#XUXXXU   h#XUXXXUU #XUXXXU hTheonlyargumentthatIcanreconstructfortheexistenceoftwoseparatemechanismsforcreating  copiesofsegmentsisbasedonsomedataofHebrewandArabicdiscussedinMcCarthy(1979,1981) 9#  26      #XUXXXU #XUXXXU.#XUXXXU9"#XUXXXUThisargumentwillbefirstillustratedwiththeTemiardata,turningtotheoriginaldatanext.#XUXXXU}"#XUXXXU 8   h  3Considerthederivationofthesimulfactiveof#XUXXXU ##XUXXXUbiconsonantals, kak<<wfromk<<wtocall,  ` whichisarguedtoemployLDCspreading,asshownin(36a)below.Theconsonant/k/automaticallyspreadstofilltheunassociatedpositionintheCaCVCtemplate.Allpreviousanalysesofthesimulfactivehadassumedthataspreading,asopposedtoareduplicative,mechanismwasinvolvedherebecause,asshownbytheothersimulfactivepatterns(e.g.c1a.c2vc3),copyingdoesnot x alwaystakesplace.36.a.Simulfactivea   h     b.Continuative 4    `  |   h        3f# Ԓ   `  CVCVVC     p 7XXU*#XUX7'#  CCCVVC .~  ' 3+   `  ||/|     p     ||/| !  3( 3   k<w     p     k<w v  3( 3   ` (output:)[kak<<w]    (intendedoutput:)[kwk<w] k #XUXXXU##%XUXXXU h 3)Ontheotherhand,thecontinuativepattern,#XUXXXU*#XUXXXUkwk<<win(36b#XUXXXU(+#XUXXXU),#XUXXXU~+#XUXXXUԀ#XUXXXU+#XUXXXUshowsthatspreadingcannotbe  Z involvedhere,becauseitwouldcreatelinecrossing.Hence,theothermechanismdevotedtocopyingsegments,reduplication,needstobeinvoked,whichiswhythemorphemictemplate[RootRoot]ispostulatedforthecontinuative.#XUXXXU,#XUXXXUԀThistemplate#XUXXXUb-#XUXXXUstipulatesthatacopyofthewholeRootmustbe "$ created.ThesegmentsinthetwocopiesoftheRootarethenassociatedtothecontinuativeprosodictemplateCCCVCbyacomplexsetofmechanismswhosedetailswerediscussedinsection4.#XUXXXU-#XUXXXU B&!( #XUXXXU.# h*XUXXXU 3Turningnowtotheoriginaldata,asimilarsituationtothatofTemiarexistsinArabicand  Hebrew(McCarthy1979,1981).McCarthynotesthatsomequadrilateralverbsinArabicareofthepattern#XUXXXU`/#XUXXXUc1vc2c1vc2,e.g.zalzaltoshake,waswastowhisper(somesharedsemanticfeatureof ` Лrepetitivenessisclaimedtounderlythisclass).ThispatternofcopyingisnotproductiveinArabic,butthetraditionalgrammarofHebrewincludestwobinyanim,knownasthePilpelandtherelatedreflexiveHitpalpel,whichshowthesamepattern.Comparedtothefirstbinyanofabiconsonantalroot,thesebinyanimshowthepatternsin(37)below.37.&&XXUԀRoot: `  gl   h      p s T   3/  BinyanI galaltoroll(intrans.)   p saa    tobesmeared    Pilpel `  gilgeltoroll(trans.)   p siasa    tostroke t  3  Hitpalpel hitgalgeltorolloneselfalong p histaasa  toindulgeoneself#XUX&&2# J #XUXXXU0#XUXXXU[3#XUXXXU5#XUXXXUTheanalysissuggestedofMcCarthy(1979,1981)usesLDCspreadingtoderivetheformofthe )y firstbinyan(e.g.spreadingof/l/in&&XXUgalal#XUX&&6#),butwholerootreduplicationtoderivetheformofthe Q PilpelandHitpalpel(consistingoftwosteps:copyingoftherootglandthenmappingofthetwo ) copiesgl,gltothetemplates 34ӀCiCCeC,and&&XXUhit#XUX&&B8#CaCCeCrespectively).Thepointisagainthatcases  likethoseofthePilpelandHitpalpelcannotbeanalyzedusingLDCspreadingbecauselinecrossingwouldresult,whilecaseslikethatofBinyanIappeartorequireLDCspreadingbecausetriconsonantalrootsshownocopyingatall(e.g.gadaltogrow).Thishasledtotheconclusionthattwosubstantiallydistinctmechanismsareatwork:LDCspreading,inthesimulfactiveofTemiarandthefirstbinyanofHebrew,andreduplicationproper,inthe#XUXXXU-6#XUXXXUԀcontinuativeofTemiarandPilpel %9!' andHitpalpelofHebrew.ThecrucialassumptiononwhichtheaboveconclusionisbasedisthatthesimulfactiveofTemiar,andthefirstbinyanofHebrew,involveLDCspreading,becauseinbothcasestriconsonantal q+&-  =&7UXXdXXd7   =&7UXXdXXd7   =&7UXXdXXd7  roots#XUXXXU:#XUXXXUԀdonotshowanycopyingatall.AswasseenintheTemiaranalysis,however,thefactthatno  copyingtakesplaceinthesimulfactiveoftriconsonantals(c1a.c2vc3#XUXXXUD=#XUXXXU)cansimplybeseenasthe  extremecaseofgradientviolationoftheconstraintMaxBR.InTemiar,becauseMaxBR#XUXXXUQ>#XUXXXUԀisranked ` lowerthanthesegmentalmarkednessconstraints,copyingoccursonlywhenrequiredbyconstraintsrankedhigherthanthemarkednessconstraints,namely,theundominatedOns(e.g. c1 a.c2vc3#XUXXXU/?#XUXXXU)} 9#  27      .  ` Notice,moreover,thattheassumptionthattwodistinctmechanismsofcopyingareinvolvedisclearlysuspectgiventhesimplefactthattheallegedcasesofLDCspreadingaredescriptivelysimilartothecasesofreduplicationproperinimportantways.Forexample,#XUXXXUj@#XUXXXUinthe =&7XXdXXd7   =&7XXdXXd7  simulfactive,   spreadingwouldbeneededtofillanemptyCslotinthetemplate#XUXXXU"B#XUXXXUCVCVC.Inthecontinuative,the x sameneedtofillthetemplateCCCVCwouldexist.Inbothcasestherewouldbeamismatchbetweenthesizeoftheunderlyingmelody(numberofsegments)andthesizeoftheoutput(numberofprosodicpositions).However,inthecontinuative,copyingwouldnotautomaticallybyinducedbythismismatch,butwouldbearbitrarilystipulatedasapropertyoftheoutputbyuseofthemorphemictemplate[RootRoot]#XUXXXUC#XUXXXU.#XUXXXUE#XUXXXUIf,asproposedinthispaper,themechanismresponsibleforthe  effectsofLDCspreadinginthesimulfactiveisidentifiedasthesamemechanismusedinthecontinuative,thisproblemdisappears.#XUXXXUE#XUXXXUCopyingofbasesegmentsinboththecontinuativeandthe `  simulfactivefollowsfromtheunitarynotionofcorrespondenceasaresponsetothecommonneedtofillpositionsintheoutput.#XUXXXUF#XUXXXU $`" Tosummarize,IhaveshownthattheassumptionthatreduplicationcannotbeinvolvedinthesimulfactiveofTemiarandothersimilarcasesisbasedonthefactthatinsomesimulfactiveoutputs '#& nocopyingofsegmentstakesplace.ThisassumptionfailstorecognizetheidenticalconditionsunderwhichreduplicationandLDCspreadingapply.#XUXXXU H#XUXXXUAccordingtotheproposalofthispaper,thisproblem  disappearsbecauseLDCspreadingisreducedtothesameformalmechanismasinreduplication,thatis,copyinginducedbycorrespondence#XUXXXUI#XUXXXU.Asshown,thenotionofgradientviolationofconstraintsis 8  crucialinachievingtheunificationofbothinstancesofsegmentalcopying,previouslyanalyzedwithtwounrelatedmechanismsofthetheory.#XUXXXUJ#XUXXXUԀAtthesametime,byobviatingLDCspreadingthis  8  proposalsolvesanindependentsetofproblemsadmittedinthetheorywhichpermitsthiskindofspreading,apointtowhichIturninthenextsection.#XUXXXUL#-XUXXXU   #XUXXXU9M#XUXXXU#XUXXXUM##M XUXXXU h5.2.TheExceptionalStatusofLDCSpreading #XUXXXUM#XUXXXU \ Virtuallyalldiscussionsofthelocalityofautosegmentalspreadinginthefeaturegeometricresearch#XUXXXUmN#XUXXXU 4 programignoreLDCspreadingortreatitasexceptional(seeforexampleClements1985,Clements&Hume1995,NChiosin&Padgett1993#XUXXXU*O#XUXXXU).#XUXXXUP#XUXXXUԀThereasonforthisisthatthesediscussionsfocuson 4 concatenativelanguages,wherevowelsandconsonantsaregenerallyassumedtolieonthesameplane(seeSteriade1987forarguments),andthusthegeometricpremiseofLDCspreading,V/Cplanarsegregation,doesnotapply.Thissectionshowsthatinnonconcatenativelanguages,whereV/Cplanarsegregationisassumed,theexistenceofLDCspreadingisproblematic.#XUXXXULP#XUXXXU D!  h N 3InV/Cplanarsegregation,thetwoconsonantsinC1VC2areadjacent,asshownin(38a). #l! &&XXU38.a.V/Cplanarsegregation  #&&&&HR#c&&&&b./pan/[pam]    c./pad/[bad#&&&c&S#&&&&] %!%  3R 3   ` VRoot   h   VRoot p     VRoot &!' 8S   `  |   h     | p      x | p'")    ` X X  X h   X  X p X    X x X X P(#+    ` |   | h   |   p |     x  | 0)$-  3nT=U  C1Root C2Root#XUX&&S#&&XXU#&&&&"T#&&&&Ԁ[labial][coronal]     x [voice]#XUX&&W##XUXXXU'X#XUXXXU *`%/  hThisrepresentationblursthedistinctionbetweenbiconsonantalclustersandpairsofconsonants + '1 separatedbyavowel.Inbothcasestheconsonantsareadjacentaccordingtothisgeometricnotionofadjacency.Onthebasisofarepresentationlike(38a),wethusexpecttofindphonologicalinteractionbetweensuchpairsofconsonantsseparatedbythevowel,whichisofthesamecharacteraswhenC1andC2arestringadjacentinaCCcluster.Specifically,weexpectcasesoflongdistance 8  assimilation#XUXXXUX#eXUXXXUasinthetwoexamplesin(38bc):(38b)showsahypotheticalprocessofplace  ` assimilation,/pan/[pam],and(38c)anotherprocessofvoiceassimilation,/pad/[bad],both  8  applyingovertheinterveningvowel.SuchprocessesarewidelyattestedinCCclusterscrosslinguistically,butarenotattestedoverinterveningvowels,asnotedbyClements(1985:46),ineitherconcatenativeornonconcatenativelanguages[ 9#  28      .#XUXXeXU[#XUXXXU p  h Y#XUXXXU]#XUXXXUHence,V/Cplanarsegregationadmitsunfoundedexpressivepowerinthetheory,predicting H unattestedphenomenalikethoseof(38bc).Inotherwords,LDCspreadinganditsgeometricpremiseofV/Cplanarsegregationfailtoexplainwhy spreadinginthisconfigurationalwaysspreadsthewholeconsonant.#XUXXXUR^#XUXXXUԀ#XUXXXU_#XUXXXUIf,asIhaveproposedabove,putativecasesofLDCspreadinginfact   involvethesamemechanismasreduplication,thisproblemdisappears.Asinreduplication,copyingtargetsthewholeRootofthesegmentandnotitsisolatedfeatures.ThisfactisinturncapturedbyMcCarthy&Princes(1995a)restrictionofcorrespondencerelationstoholdbetweenwholesegmentsandnotbetweentheirindividualfeatures.Ifcorrespondencerelationswereallowedtoapplybetweenindividualfeaturesitwouldbepossibletoexpressreduplicationphenomenawherethereduplicantselectivelyborrowscertainfeaturesfrom %0!$ thebase.Suchphenomenaareunattested.Whatisfoundinsteadisthatidentityrequirementsapplyinparallelforallfeaturesofasegment(andotherprosodicpropertiesasclearlyshownbythefactsofTemiar).This#XUXXXU"`#XUXXXUidentitymaybesacrificedalwaysunderthedemandsofhigherrankedconstraints. ` Forexample,inMakassaresebulam[bulad]Base(seefootnote21),identityofplacebetween/m/and 8  itscorrespondentinthebase/d/isviolatedbecauseofthehigherrankedhomorganicityconstraint  ` onthecluster/mb/.#XUXXXUc#XUXXXU  8  TherestofthissectiondiscussesfurtherevidencefortheeliminationofLDCspreading.OnetypeofevidenceispresentedindetailinGafos(1996),whereIproposethatthenotionoflocalityinphonologyisdirectlydefinedintermsofadjacencyofactualarticulatorygestures 9#  29      #XUXXXUe#XUXXXU.Thebasicfact p ofinterestisthatthetwovocalicgesturesinaVCVsequencearearticulatorilycontiguous,incontrasttothetwoconsonantalgesturesinaCVCsequencewhicharenotcontiguous.Specifically,inaVCVsequence#XUXXXUQg#XUXXXUԀthevowelgesturesformacontiguousvocalicsubstrateonwhichthegestureof H theconsonantissuperimposed.AcousticallythisarticulatorycontiguityismanifestedinthefactthatthevowelformantsintheVCandCVtransitionpartsoftheVCVsequencedependnotonlyontheconsonantbutalsoonthevowelontheothersideoftheconsonant.#XUXXXUh#XUXXXUThisphenomenon,whichis  knownasVtoVcoarticulation,isauniversalpropertyoflanguage(hman1966,1967,Fowler1983,Lieberman&Blumstein1988)#XUXXXUAj#XUXXXU. 0"  #XUXXXU0k#XUXXXUOntheotherhand,itcanbeshownthattheconsonantgesturesinaCVCsequencearenot $X" contiguous.#XUXXXUk#XUXXXUԀThisfollowsfromthefundamentalasymmetrybetweenconsonantsandvowels,namely, %0!$ thepresenceofaconstrictionwiththeformerversustheabsenceofaconstrictionwiththelatter#XUXXXUBl#XUXXXU.#XUXXXUOm#XUXXXUԀIn '#& thesequenceofarticulatoryeventsC[+constriction]V[constriction]C[+constriction],theinterveningvocalic  gestureisboundtocompletelyundotheconstrictionofthefirstconsonant,andthetwoconsonantgesturescannotbecontiguous.Iftheywere,acontinuousconsonantalconstrictionwouldoverlapwiththearticulationofthevowelwhichisphysiologicallyimpossible.#XUXXXUm#XUXXXU 8  TheresultisthatvowelgesturesarecontiguousinaVCVsequencebutconsonantgesturesarenotcontiguousinaCVCsequence(VtoVcontiguityversusnoCtoCcontiguity).InGafos(1996)Iproposethatthecorrectnotionoflocalityinphonologyisarticulatorycontiguity.VtoVcontiguityiswhatmakesvowelharmonyalocalprocess.Atthesametime,#XUXXXUo#XUXXXUԀthelackofphonologicalinteraction   notedin(38bc)abovefollowsfromthelackofCtoCcontiguityintheCVCconfiguration.Nocontiguityimpliesnobasisforphonologicalinteraction,contrathepredictionofV/Cplanarsegregationandinaccordancewiththefacts.Inotherwords,ifphonologicalspreadingappliesundertheconditionofarticulatorycontiguity,itfollowsthatitcouldnotbeinvolvedbetweenthetwoconsonantsofaCVCconfiguration. 9#  30      #XUXXXUQq#XUXXXU   Otherresearchershaveproposedinvariousformsrestrictionsonspreadingthatareconsistentwiththeresultsoftheabovediscussion.NChiosin&Padgett(1993)#XUXXXUs#XUXXXUԀproposeasimilarexplanation  fortheabsenceofphonologicalprocesseslikethosein(38bc).Theirexplanationisbasedontheassumptionsthatspreadingisstrictlylocal,inthesensethatitcannotapplyoveraninterveningsegment,andthatplaceandstrictureofconsonantsformaclassoffeatures,spreadingtogether.Thus,ifspreadingwastoapplybetweenthetwoconsonantsintheCVCconfiguration,itmustgothroughthevowel.Thiswillimposeaconsonantalconstrictiononthevowelresultinginafatalviolationofsyllablestructure(thisideaisattributedtoBruceHayes).#XUXXXUt#XUXXXUThisideahasobviousaffinitieswiththe  8  aboveproposal.Also,McCarthy(1994)analyzingcasesofvowelharmonyacrosswhatareseeminglytransparentcoronalconsonantsinadialectofBedouinArabic,arguesthattheinterveningconsonantsmustparticipateintheharmony,i.e.spreadingofthevowelfeaturescannotapplyoveraninterveningconsonant,butratherthroughit.Buildingonthesameidea,#XUXXXUnw#XUXXXUPadgett(toappear) H analyzesinstancesoftranslaryngealvowelharmony,arguingthatspreadingofvocalicfeaturesmustgothroughtheinterveninglaryngealconsonants.#XUXXXUYy#XUXXXUFinally,Archangeli&Pulleyblank(1994)suggest H thatlongdistancespreadingofasegmentoveravowelshouldbeuniversallyprohibited.#XUXXXUXz#XUXXXU   Beforeconcludingthissection,IbrieflyconsideranotherargumentthathasbeenbroughtinsupportforphonologicallymotivatedV/Csegregation,whichappearsnowtobesignificantlyweakenedbyrecentworkintheliterature.Specifically,McCarthy(1989)hasarguedthatV/Csegregationistherepresentationalmanifestationofunderspecifiedlinearorderbetweenconsonantsandvowelsinlanguageswheresufficientlyrichconstraintsontheshapeoftheoutputrenderthisorderingpredictable(e.g.Semitic,Yawelmanietc.).Thecrucialassumptiononwhichthisargumentrestsisthatunderlyingrepresentationmustcontainonlyunpredictableinformation.Therehasbeenconsiderableworkunderminingthevalidityofthisassumption#XUXXXU7{#,XUXXXUespeciallyinOTandBurzios )$( constraintbasedframework#XUXXXUq~#XUXXXU(Prince&Smolensky1993:Chapter9;Burzio1994;Inkelas1994;#XUXXXU#XUXXXUIt=, h+&* Mester&Padgett1994;#XUXXXU#XUXXXUMcCarthy1995;seealso#XUXXXU#Steriade1995XUXXXU).Thebasicideaisthatproperties  oflexicalformsarenottheresultofarbitraryconditionsontheseforms,suchastheassumptionoftheunderspecificationargumentabove,butemergeoutoftheinteractionofuniversalconstraints.Theargumentsforspecificationofpredictableinformationcovertheentirerangeoflexicalspecifications,rangingfromunderlyingspecificationofpredictabledistinctivefeatures(Smolensky1993,Inkelas1994;It=,Mester&Padgett1994#XUXXXUa#XUXXXU)andprosodicinformation(Prince&Smolensky  8  1993#XUXXXU#XUXXXU,Burzio1994,Inkelas1994#XUXXXU!#XUXXXU),tospecificationofconsonantvowelorderinginlanguages   previouslythoughttobeprimecandidatesforV/Cplanarsegregation(McCarthy1995).#XUXXXU~#XUXXXU   Inaddition,manycasesoflanguagesthathavebeenassumedtoemployV/Csegregationinthepasthavebeenreanalyzedorarguedtoprovidenoevidenceforthisrepresentation#XUXXXUf#XUXXXU.#XUXXXUl#XUXXXUMostnotably, H Steriade(1987),inwhatcanbeseenasaprecursoroftheproposalinthispaper,#XUXXXU#XUXXXUԀarguesforan  p alternativeaccountoftheYawelmanifacts(Archangeli1985)thatdoesnotrelyonV/Csegregationandusesmelodycopying,insteadofspreading,toderivemultiplecopiesofsegments.BuildingonSteriade,BatEl(1994)presentsananalysisofModernHebrewsegmentalcopying#XUXXXUf#XUXXXU,usingno  spreadingorV/Csegregation.Finally,McCarthy(1995)providesareanalysisofRotumanmetathesiswithoutusingV/Csegregation#XUXXXU߇#XUXXXU.i 9#  31       X  #XUXXXU#XUXXXUTosummarizethemainpointofthissection,IhaveshownthatthetheorywhichadmitsLDC 0"  spreadinganditsgeometricpremiseofV/Cplanarsegregationfalselypredicts#XUXXXUl#XUXXXUԀtheexistenceof $X" unattestedlongdistancespreadingofindividualfeaturesbetweenthetwoconsonantsinaCVCconfiguration.ThepresentproposalthattheallegedinstancesofLDCspreadingareactuallycopying '#& resolvestheseproblems.IthusconcludethatLDCspreadingandasaresultV/Cplanarsegregationshouldbeeliminated.#XUXXXUl#XUXXXUԀThenextsectionexaminesthetypologicaldistinctionbetweenconcatenative  andnonconcatenativelanguagesinlightofthisconclusion.#XUXXXU&#ԚXUXXXU   6.TypologicalConsequences  ` #XUXXXU#*XUXXXU h Thissectionhastwogoals.First,itshowsthattheproposalofthispapertoanalyzeLDCspreading  8  ascopyingnaturallyextendstocasesofsegmentalcopyinginSemiticlanguagesaswell.Thenitarguesthatthedistinguishingcharacteristicofsegmentalcopyingfoundinnonconcatenativelanguagesshouldbeidentifiedwithaspecialmodeofmorphology,namely,atemplaticreduplicativeaffixation,wherethereduplicativeaffixisnotspecifiedforanyprosodictarget.ThisspecialtypeofreduplicationfillsatypologicalgapinthetheoryofProsodicMorphology,providingstrongsupportfortheanalysespresentedhereinandforthetheoryofProsodicMorphologyitself.uXXU#uu#uu H #XUXu##XUXXXUȐ#XUXXXU B6.1.FurtherAnalyses #XUXXXUT#XUXXXU   h hConsiderthewellknownSemitic#XUXXXUʑ#XUXXXUpatternc1v.c2#XUXXXUa#XUXXXUv c2 #XUXXXU#XUXXXU,whichisfoundinModernHebrewdenominal l formation,#XUXXXU#XUXXXUasinkodcodekidedtocodify,andalsointhefirstMeasureofthebilateralArabic L! verbs,asinsamampoisoned(byfarthemostfrequentMeasureinthelexiconofArabic).IwillfirstpresentananalysisofthispatterninModernHebrewdenominalformation,turningtothediscussionoftheArabicpatternnext.#XUXXXU# &$"%  hXUXXXUF#XUXXXUw#XUXXXUollowingBatEl(1994),Iassumethatthebaseforthedenominalformationisthe (#' correspondingnounandthattheshapeoftheoutputmustconsistoftwosyllables,#XUXXXU#XUXXXU[%#XUXXXU#XUXXXUԀ%#XUXXXU#XUXXXU]#XUXXXUC#XUXXXU.In *%) addition,Iassumethattheoutputmustendinaconsonant,duetothegeneralcanonoftheverbalstemsofSemitic,calledFinalConsonantalityinMcCarthy&Prince(1990b),henceforthFinalc.#XUXXXU#XUXXXU  Essentially,thisconstraintfurtherspecifiestheshapeofthebisyllabicoutput.̀Onedifferencebetweenthedenominaloutput,c1ic2#XUXXXU#XUXXXUe c2#XUXXXU#XUXXXU ,andthepatternsofcopyinginTemiar, 8  e.g. c1 a.c1#XUXXXU#XUXXXUvc2#XUXXXU#XUXXXU, c1c2 #XUXhXUњ#XUXhXU.c1#XUXXXU)#XUXXXUvc2#XUXXXUw#XUXXXU,#XUXXXU›#XUXXXUԀisthattheTemiarcopiedconsonant(s)appeartotheleftofthebase,while  h intheSemiticpatterntheyappeartotheright.ThisisbecausetheaspectualmorphologyofTemiarprefixessegmentalmaterialtotheprosodicheadofthebase.IntheSemiticpattern,ontheotherhand,Iassumethatthemorphologyissuffixational.Specifically,Iassumethatdenominalverbformationinvolvesaconstraintrequiringthatanaffixmustbealignedwiththerightedgeoftheprosodicoutput,Align(Affix,R,PrWd,R),henceforthAlignAFFIX#XUXXU#XUXXUR.Thisaffix,asinthecaseof X Temiarsimulfactive,willbepartiallyspecifiedforthemelody#XUXXXU#XUXXXUԀ/ie/,andwill#XUXXXU=#XUXXXUԀalsobereduplicativein 0 thesensethatthereisacorrespondencerelationbetweenitandthebase.#XUXXXU#XUXXXU X Beforediscussingthechoiceofthecopiedconsonant,considerfirstanotherinterestingfactaboutdenominalformation.Theoutputverbdoesnotcontaintheinputbasevowel,whichhasbeenreplacedbythevocalismoftheaffix.Thisphenomenon,calledMelodyOverwriting#XUXXXUH#XUXXXUԀ(McCarthy  &Prince1990a,BatEl1994),hasapparentlybeentreatedasanidiosyncracyofSemitic,#XUXXXU#XUXXXUimplementedasarulewhichliterallysubstitutesthevocalismofthebasewiththevocalismofthe @"  affix.Amoreprincipledaccountexists,however,inthecaseathand.Ifallvowelsinthe#XUXXXUq#XUXXXUinput,/o/#XUXXXU#XUXXXU $h" ofkod,and/i,e/oftheaffix,surfacedintheoutput,thentherewouldbeaviolationoftheundominatedtemplaticconstraint,requiringthattheoutputmustconsistoftwosyllables(e.g.asinc1v[c2i c1 e c2 ]).Thevowelofthebasedoesnotappearintheoutputinordertoavoidthisviolation, )$( hencetheranking[%%]>>MaxBASEIO.Thefactthatnoaffixvowelgivesitsplacetothebasevowel +&* intheoutputsimplymotivatesthefurtherrankingMaxAFFIXIO>>MaxBASEIO.#XUXXXUӣ#XUXXXU   3Turningnowtotheissueofthecopiedconsonant,recallthat#XUXXXU#XUXXXUAlignAFFIX#XUXXU&#XUXXUR#XUXXXU{#XUXXXUԀdemandsthatthe  rightedgeoftheaffixbealignedwiththerightedgeoftheoutput.Thisconstraint,togetherwithFinalc,dictatesthatacopyofaconsonantbeattherightedgeoftheoutput.Arelevantsetof 8  candidatesisconsideredin(39)below.#XUXXXU§#XUXXXUAnchoring#XUXXXU:#XUXXXU,itwillberecalled,requiresthattheright  ` peripheralelementofthereduplicativeaffixcorrespondtotherightperipheralelementofthebase.(ThefinalconsonantintheoutputmustbeacopyofsomebasesegmentbecauseofDepBR,not   showninthetableau.).#XUXXXU#XUXXXU   39.#XUXXXU #XUXXXURankingargument:Anchoring#XUXXXUf#XUXXXUԀ>>#XUXXXUɫ#XUXXXUSrole#XUXXXU#XUXXXU ,|  3#XUXXXU\#XUXXXU*E$NddHdd Ddd dd edd dd G$O(#(#,Xdd ,dd ,rdd ,rdd",'dd"+   v  +$33* +Input:#XUXXXUɬ#XUXXXUԀc1vc2#XUXXXU#XUXXXU,ie#XUXXXUӮ#XUXXXU 43!*$33 4AlignAFFIX#XUXXU#XUXXUR#XUXXXU#XUXXXU =3$**"3 =Anchoring#XUXXXU#XUXXXU 23!*"3$ 2#XUXXXUx#XUXXXUSrole %*3 %#XUXXXU#XUXXXU +$3Z +a.+#XUXXXU^#XUXXXUԀc1i.c2#XUXXXUձ#XUXXXUe c2#XUXXXU.#XUXXXU 4!Z$3 4 =$*Z" = 7-Z"$ 7#XUXXXU|#XUXXXU*#XUXXXUi# 3XUXXXU .$Z" . +$3. +b.#XUXXXU#XUXXXUc1i.c2#XUXXXUh#XUXXXUe c1#XUXXXU# XUXXXU 4!. $3 4 =$*.!" =*! 7-.""$ 7 .$.#" . +$3b$ +c.c1i. c2 #XUXXXU #XUXXXUec2#XUXXXUy#XUXXXU 4!b%$3 4*! =$*b&" = 7-b'"$ 7#XUXXXUĶ# 3 3XUXXXU* .$b(" . +$36) +d.c1i. c1 #XUXXXUط#XUXXXUec2#XUXXXU#XUXXXU 4!6*$3 4*! =$*6+" = 7-6,"$ 7#XUXXXU# 3Ʒ 3XUXXXU/%#6-"   /#XUXXXU# 3XUXXXU 3#XUXXXU#XUXXXU 3 hCandidate(39a)copiestherightmostsegmentofthebase,incurringaviolationof#XUXXXUֺ#XUXXXUSrole#XUXXXU#XUXXXU,while P/ candidate(39b)copiestheleftmostconsonant,violating#XUXXXUܻ#XUXXXUAnchor#XUXXUs#XUXXUing#XUXXXU#XUXXXU.Tochoosethecorrect ( x1 candidate,then,therankingmustbe#XUXXXU#XUXXXUԀAnchor#XUXXU#XUXXUing#XUXXXU#XUXXXUԀ>>#XUXXXU-#XUXXXUSrole#XUXXXUs#XUXXXUM 9#  32      .Candidates(39cd)violatethe "P3 #XUXXXU#XUXXXUundominatedAlignAFFIX#XUXXU#XUXXUR#XUXXXU#XUXXXUԀbecausetherightmostaffixalsegment/e/isnotalignedwiththeright #(5 edgeoftheprosodicoutputa| 9#  33      .#XUXXXU3#XUXXXU %!7  h #XUXXXUB#XUXXXUIntheaboveanalysisIhaveassumedthatthebaseofdenominalformationisthecorresponding  nounform.Asimilaranalysiscanbegivenifweassume,followingtradition,#XUXXXU#XUXXXUthattheinputto  denominalformationconsistsoftheconsonantalroot/c1c2/andthevocalicaffix/ie/.Theonly ` differencewillbethatconstraintswhichmakereferencetothebasemustnowrefer#XUXXXU#XUXXXUtothe 8  consonantalrootinstead.Forexample,Anchor#XUXXU#XUXXUing#XUXXXUx#XUXXXUԀwillnowrequirethattherightelementofthe  ` affixcorrespondtotheright#XUXXXU#XUXXXUelementoftheroot.Everythingelseremainsthesame.Suchananalysis  8  wouldberequired,forexample,forthecaseofthec1vc2v c2 patternfoundinthebilateralverbsof   theArabicfirstMeasure(sama m poisoned).h 9#  34      #XUXXXUd#XUXXXU   Itshouldbepointedout,however,thattheroleofthetraditionalconceptofaconsonantalrootlike/c1c2#XUXXXU#XUXXXU/#XUXXXU#XUXXXUԀ,apeculiarityofSemiticlanguages,hasbeensignificantlyreducedinrecentanalysesin X theframeworkofProsodicMorphology(McCarthy&Prince1986etseq.).Specifically,BatEl(1994)arguesthatModernHebrewmorphology,infact,needmakenoreferencetosuchaconcept. X AllmorphologyiswordbasedandnomappingoperationsofconsonantalrootstotemplatesareneededP 9#  35      .Also,theanalysisoftheArabicbrokenpluralsinMcCarthy&Prince(1990a)makesno  referencetoconsonantalroots.Mostimportantly,McCarthy(1993)presentsasignificantrevisionoftheearlieranalysisoftheArabicverbalmorphologyofMcCarthy(1979,1981).Inthenewanalysis,thereisonlyasingleremainingcase#XUXXXU1#XUXXXUwherereferencetoaconsonantalrootisrequired.This  ` isthecaseofMeasureIofbilateralandtrilateralverbs,samam,faal,respectively,whicharestill  8  derivedfromthetraditionallyassumedconsonantalroots/sm/and/fl/,mappedontoaniambic   template(witharequiredfinalconsonant)~ 9#  36      .McCarthyarguesthattherearenoothertemplatesand   hencenoothermappingoperationsofaconsonantal#XUXXXU#XUXXXUԀroottoatemplate.Allothersurfaceformsof p theverbderivefromthebasicMeasureIformbyaffixation,andinsomecasesbycircumscriptionofaprosodicallydefinedconstituentofthebase.Thisanalysisconstitutes,asMcCarthy(1993)notes,asignificantdeparturefromtheearlieranalysesinMcCarthy(1979,1981),whereadifferenttemplatewaspositedforeveryMeasure.#XUXXXUP#XUXXXUInanycase,however,notethatthemorphological   distinctionbetweenconsonantsandvowelsinSemitic,ifitexistsatall,neednotbeencodedrepresentationallybythegeometricnotionofV/Csegregation.Theobviousalternativeistoencodethedistinctionintermsofconsonantsandvowels,twophonologicallyseparatecategoriesofsounds.Thisseemsjustasadequate,andinfactsimpler,requiringnoadditionalgeometricdevicesunique 0"  toSemiticlanguages.Tosumup,theanalysisofSemiticcopyingrequiresnouseofV/CsegregationorLDCspreading.#XUXXXU&#XUXXXUԀThisresultfurthersecurestheconclusionthatLDCspreadingandV/Csegregation#XUXXXU#XUXXXUԀcan ` andshouldbeeliminatedasprevioussectionsofthispaperhavedemonstrated.PutativecasesofLDCspreadingareliterallycopying,thesamephenomenonfoundinthereduplicativemorphologyofmanylanguages#XUXXXUW#XUXXXUԀ(AncientGreek,Diyari,Panopeanetc.).#XUXXXU#XUXXXUԀ#XUXXXU#XUXXXUSimplyput,thedistinctionbetween  8  concatenativeandnonconcatenativelanguagesneednotandshouldnotbeencodedintermsofthespecialphonologicalmechanismsofV/CsegregationandLDCspreading.Thisresulthasthewelcomeconsequenceof#XUXXXU0#XUXXXUunifyingtherepresentationalapparatusinthephonologicalcomponentof p thetwotypesoflanguages.#XUXXXU{#uXXUuu H #uuf#uu5#XUXuG##XUXXXU#XUXXXU h6.2.AtemplaticAffixation #XUXXXU9#XUXXXU   #XUXXXU# h\XUXXXU hInthetheoryofwordformation,theprogramofProsodicMorphology(McCarthy&Prince1986et  seq.)hasestablishedasoneofitscentralclaimsthatgrammaticalcategoriesareoftenexpressedbyinvariantprosodicshapesortemplates.Thesetemplatesaremadeoutoftheunitsofprosody,namelysyllables,feet,andprosodicwords.AsMcCarthy&Prince(1995b)show,therearetwowelldocumentedspeciesoftemplaticspecificationofmorphologicalconstituents:templaticspecificationoftheaffixandtemplaticspecificationofthebase. h3 3Templaticspecificationoftheaffixisfoundinordinaryreduplication,wherethereduplicativeaffixisspecifiedforaninvariantshapecorrespondingtosomeunitoftheprosodichierarchy.Forexample,#XUXXXU#XUXXXUasshownbelow,tworeduplicativeaffixesofIlokanoarespecifiedtobealightsyllablein h)$( (40a)andaheavysyllablein(40b)respectively(#XUXXXU#XUXXXUdatadrawnfromMcCarthy&Prince1995b). @+&* Ї#XUXXXUp#XUXXXU40.a. ` Affix%󀀀  Base h   si+%+base p   Gloss( covered/filledwith)   3 3Rj   `    bu.ned  si bu bu.ned p   carryingabuneng 3 Z U Ӏb. ` Affix%  Base h   %+base p   Gloss(plural)      `    pu.sa h   pus pu.sa p   cats#XUXXXU#XUXXXU   3ExamplesoftemplaticspecificationofthebasewereseenintheanalysesoftheSemiticpatternsabove,whereabisyllabicrequirementwasimposedontheshapeoftheoutputinthecaseModernHebrewdenominals.Anothercaseofbasetemplaticismisfoundin#XUXXXU#XUXXXUYawelmani,wheretheshapeof 9   thesurfaceformisdeterminedbyasetofprosodictemplatesapplyingtosomeinitialpartofthestem(Archangeli1991),asshown#XUXXXU#XUXXXUin(41).Theinitialtemplaticallyspecifiedpartofthestemisin 9  boldfaceandthegivenformsabstractawayfromregularrulesofepenthesis,closedsyllableshortening,androundingharmony#XUXXXU #XUXXXUԀ(cisaglottalizedc.#XUXXXU"#XUXXXUԀDatadrawnfromMcCarthy&Prince  1995b).#XUXXXU#XUXXXU q #XUXXXU#XUXXXU#XUXXXUV#XUXXXU41. `    % h   [%%]Iamb p Gloss U  3uXXU  CVshape   CvvC h   CvCvv   p #XUXu# L v6  Biconsonantal   cuu m h   cumuu  p devour "r   Triconsonantal  hii wt h   hiwii t   p walkk#XUXXXU#XUXXXU   hIntheformsunder%#XUXXXU#XUXXXU,abimoraicparseisimposedonthefirstsyllableoftheoutput,whileinthe j  formsunder[%%]Iamb#XUXXXUR#XUXXXU,aniambicparseisimposedinstead(seeMcCarthy1993forother B " examplesoftemplaticspecificationofthebasefromthenounandverbalmorphologyofArabic).#XUXXXU4# "j$  hXUXXXU 3Asexpected,therearealsocaseswherethemorphologyspecifiesnotemplateatall,aspecial #B& modeofprosodicmorphology,knownasatemplaticprosodicmorphology(Archangeli1991,McCarthy&Prince1990b).IntheEthiopianSemiticlanguageChaha,forexample,verbalbasesinthemorphologicalcategorycalledjussivesurfaceintwoformsCCBCorCBCC,asshownin(42), z)$, where#XUXXXUR#XUXXXUy#XUXXXUP#XUXXXUԀisanagreementprefix(datadrawnfromMcCarthy1993). R+&. Ї#XUXXXU# XUXXXU42. ` Root   JussiveVerb    Gloss   3q  a. ` gfr   ygfBr h     release    b. ` nks   ynkBs    bite    c. ` srt   ysBrt h     cauterize    d. ` trx   ytBrx h     makeincision k #XUXX XU6#XUXXXU hThechoicebetweenthetwoformsisentirelypredictablefromtheregularsyllabificationofthelanguage.Thejussivemorphologythenspecifiesnotemplateonthesurfaceformoftheverb.OtherexamplesofatemplaticbasespecificationcanbefoundinYawelmani(McCarthy&Prince1995b),ArabicandAkkadian(McCarthy1993). hSofar,then,theliteratureonprosodicmorphologyhasdocumentedthefollowingthreecasesofprosodicspecificationorlackthereof:templaticspecificationontheaffix(asinordinaryreduplication),templaticspecificationofthebase(asinSemitic,Yawelmanietc.),andnotemplaticspecificationofthebase(asinChaha).Afourthcaseisthereforepredictedtoexist,namely,notemplaticspecificationoftheaffix.ThisispreciselythecommoncharacteristicofaffixationintheanalysesoftheTemiaraspectandoftheSemiticpatternsdiscuss% g edabove.Thereduplicativeaffixesintheseanalyseshave#XUXXXUc#XUXXXUnoprosodicrequirementontheirshape,althoughtheymaybepartially  specifiedsegmentally.Forexample,thesimulfactiveofTemiarandthedenominalaffixofModernHebrewarespecifiedas/a/,/ie/,respectively.Noprosodicconstraintisimposedontheshapeoftheseaffixes,however.Similarly,inthecontinuativeofTemiarnoprosodictemplateisimposedontheaffix,andinadditiontheaffixalsolackssegmentalspecification.Reduplicativeaffixationofthistypeisthereforeonespecialmodeofwordformationwherenoprosodictemplateisimposedontheaffix.ThisfillsatypologicalgapinthetheoryofProsodicMorphology,providingstrongsupportforthespecificanalysespresentedinthispaperandfortheprogramofProsodicMorphologyitself.Lackofprosodicspecificationonthepartofthereduplicativeaffixexplainsadescriptive s+&+ dissimilaritybetweenordinaryreduplicationandreduplicationofthesortdiscussedinthispaper.#XUXXXU#XUXXXUԀ#XUXXXU#XUXXXUIn  ordinaryreduplication,theaffixisrealizedasacontiguousstringofbasesegmentsparsedinsomeprosodicunit(syllable,foot,orprosodicword),whichisarbitrarilyspecifiedbythemorphology(e.g. ` %#XUXXXU #XUXXXU,%#XUXhXUU#XUXhXU#XUXXXU#XUXXXUԀaffixesofIlokano).#XUXXXU#XUXXXUԀOntheotherhand,thereduplicativeaffixesdiscussedinthispaperare 8  realizedwithcopiesofisolatedsegmentsofthebase,invariousshapesandquantities.Forexample,thesimulfactiveofTemiarcopiesoneornoneoftheconsonantsofthebase,whilethecontinuativeaffixcopiesoneortwoconsonantsofthebase.Theelusiverealizationoftheseaffixesissimplyaconsequenceoftheirlackofaprosodictarget.Withoutaprosodictemplateoftheirown,mattersofrealizationoftheseaffixesarelefttobedeterminedbyconstraintsregulatingtheprosodyofthelanguage(Onsand1vinTemiar),oroftheparticularmorphologicalcategoryinvolved([%%]in H Semitic).#XUXXXUD#XUXXXU  p  h   7.Conclusion #XUXXXUk#XUXXXU   Ihavearguedthatinthephonologicalcomponentofthegrammarthereisnoplaceforanoperationthatspreadsaconsonantoveravowel(LDCspreading),withitsgeometricpremiseofV/Cplanarsegregation.ThetheoryadmittingthesetwomechanismsfailstoexplainwhyLDCspreadingalwaystargetswholesegments,predictingunattestedspreadingofindividualfeaturesoveravowelundertheconfigurationofV/Cplanarsegregation.IhaveproposedtoreplaceLDCspreadingwiththesameformalmechanismusedinreduplication,whichisindependentlyneededinthetheory.#XUXXXU#XUXXXUԀWhat %0!$ wasformerlyseenasLDCspreadingisnowliterallycopying.#XUXXXU#XUXXXUCopying,asinreduplication,targets '#& thewholesegment,notitsindividualfeatures.Hence,theexcessivepowerthatthetheoryadmittingLDCspreadingandV/Cplanarsegregationwouldhaveisavoided,whileatthesametimethe h+&* obviousredundancybetweenLDCspreadingandreduplicationiseliminated.#XUXXXU.#XUXXXU   h#XUXXXU#XUXXXUTemiarisanonconcatenativelanguageforwhichbothLDCspreadingandreduplicationwere  considerednecessarytoaccountforitsintricatepatternsofcopying.IhaveshownthatusingthenotionofCorrespondenceinOptimalityTheoryitispossibletoprovideaunifiedaccountofthecopyingpatternsintheverbalmorphologyofthelanguage.#XUXXXU0#XUXXXUCopyingofsegmentsisinducedbya  ` correspondencerelationholdingbetweenthesegmentsofthebaseandthesegmentsofareduplicativeaffix.Constraintsrequiringthefeatural(Ident(F))andprosodicidentity(Srole)   betweencorrespondentsegmentsevaluatethequalityofthiscorrespondencerelation.Theinteractionoftheseconstraintswithothersexpressinggeneralprosodicregularitiesofthelanguagesufficestoaccountforthefullrangeofpatterns.uXXU H #XUXu ##XUXXXU#XUXXXUIhaveshownthatanyderivationalapproachtothefactsofTemiarisboundtomisssignificant  p generalizationsdirectlycapturedbytheOptimalitytheoreticapproach.#XUXXXU| #XUXXXUԀ#XUXXXUy #XUXXXUThechoiceofthecopied H consonantsisonesuchimportantgeneralization.Itwasshownthataderivationalapproachwouldrequiretwoseparateassociationruleswithincompatibledirectionalparametersettings.IntheOptimalitytheoreticapproach,ontheotherhand,thechoiceofthecopiedconsonant(s)isdeterminedbyasingleuniversalconstraint,Srole,andnotbylanguageparticularrulesofassociation.Inthis X  andotherrespects,thefactsofTemiarsupporttheOptimalitytheoreticconceptionofphonologymorphologyinteraction,wheretheparallelapplicationofphonologicalandmorphologicalconstraintsdeterminetheformoftheoutput.#XUXXXU #uXXUuu %0!$ #XUXu#TheproposaltoreplaceLDCspreadingwithcopyingviacorrespondencewasshowntoapply '#& tosomeofthebasicpatternsfoundinSemiticlanguagesaswell.Ineffect,segmentalcopyinginlanguageslikeTemiarandSemiticdoesnotneedtorelyonspecialphonologicalmechanismssuch h+&* asLDCspreadingandV/Cplanarsegregation.#XUXXXU#XUXXXUIarguedthatthedistinctionbetweenconcatenative  andnonconcatenativelanguagescannotbeencodedintermsofthesemechanisms.Phonologicallyspeakingthereisnothingspecialtononconcatenativelanguages.Thedescriptivedissimilaritybetweenthetypesofsegmentalcopyingfoundinnonconcatenativelanguagesandthosefoundinordinaryreduplicationisattributedtoaspecialmodeofaffixation.Nonconcatenativesegmentalcopyingissimplyatemplaticreduplicativeaffixation,wherethereduplicantisnotspecifiedforanyprosodictarget.Thistypeofreduplicativeaffixation,infact,fillsatypologicalgapinthetheoryofwordformation,whichhadsofarpredictedbutnotdocumentedthistypeofatemplaticaffixation.#XUXXXU#XUXXXU   Improvementsoverprevioustheoriesareachievedwhensimpleintuitionsreceivecoherentaccountsinformaltermsinthenewtheory.Thesuccessofthecorrespondenceapproachinachievingtheintuitivelydesirableunificationofallinstancesofsegmentalcopying,obviatingtheneedfortheproblematicmechanismsofLDCspreadingandV/Csegregationofprevioustheories,providesstrongsupportforthegeneralapproachtakeninOptimalityTheoryandCorrespondenceTheory,whereinthenotedunificationhasbeencaptured.#XUXXXU#XUXXXU   hReferences  X  0   Archangeli,D.:1985,YokutsHarmony:EvidenceforCoplanarRepresentationinNonlinearPhonology,LinguisticInquiry16,335-372#XUXXXUa#XUXXXU.#XUXXXU#XUXXXU$X"(#(# 0   Archangeli,D.:1991,SyllabificationandProsodicTemplatesinYawelmani,NaturalLanguage %0!$ andLinguisticTheory9,231283.#XUXXXU#XUXXXU'#&(#(#  h0   Archangeli,D.,andD.Pulleyblank:1994.GroundedPhonology.Cambridge,MA:MITPress.#XUXXXU#XUXXXU)$((#(# 0   Aronoff,M.:1994,MorphologybyItself:StemsandInflectionalClasses,LinguisticInquiry h+&* Monograph22,MITPress:Cambridge,MA. (#(# 0   BatEl,O.:1994,StemModificationandClusterTransferinModernHebrew,NaturalLanguage  andLinguisticTheory12,571596.#XUXXXU#XUXXXU`(#(# 0   Benjamin,G.:1976,AnOutlineofTemiarGrammar,inJenner,Thompson,andStarosta1976,129-187.#XUXXXU#XXXX#XXXX2#XUXXXU `(#(# 0   Benua,L.:1995,IdentityEffectsinMorphologicalTruncation,inJ.Beckman,S.Urbanczyk,andL.Walsh(eds.)UniversityofMassachusettsOccasionalPapersinLinguistics18:Papersin   OptimalityTheory,GraduateStudentAssociation,Amherst,MA. (#(# 0   Berkley,D.:1994,TheOCPandGradientData,StudiesintheLinguisticSciences24,114.p(#(# 0   Broselow,E.andJ.McCarthy:1983,ATheoryofInfixingReduplication,TheLinguisticReview H 3,2598. (#(# 0   0(#(#  0(#(#  Browman,C.andL.Goldstein:1989,ArticulatoryGesturesasPhonologicalUnits,Phonology6, H 201251.#XUXXXU#XUXXXU (#(# 0   Burzio,L.:1994,PrinciplesofEnglishStress,Cambridge:#XUXXXU #XUXXXUԀCambridgeUniversityPress.#XUXXXU!#XUXXXU(#(# 0   Clements,N.:1985,TheGeometryofPhonologicalFeatures,PhonologyYearbook2,225252.(#(# 0   Clements,N.andE.Hume:1995,TheInternalOrganizationofSpeechSounds,inJ.Goldsmith,(ed.)HandbookofPhonologicalTheory,Blackwell,Oxford,245306.#XUXXXU!#0" (#(# 0   Diffloth,G:1976a,Minor-SyllableVocalisminSenoicLanguages,XUXXXUinJenner,Thompson,and $X" Starosta1976#XUXXXU($#,229-248.%0!$(#(# 0   Diffloth,G.:1976b,ExpressivesinSemai,XUXXXUinJenner,Thompson,andStarosta1976#XUXXXU%#,249-264.'#&(#(# 0   Diffloth,G.:1976c,JahHut,anAustroasiaticLanguageofMalaysia,SoutheastAsianLinguistic )$( Studies2,73118.h+&*(#(# 0   Flemming,E.andM.Kenstowicz:1995,BaseIdentityandUniformExponence:AlternativestoCyclicity,XUXXXUMassachusetts#XUXXXU&#ԀInstituteofTechnology,MA,ms.(#(# XUXXXU0   Fowler,C.:1983,ConvergingSourcesofEvidenceforSpokenandPerceivedRhythmsofSpeech: ` CyclicProductionofVowelsinSequencesofMonosyllabicStressFeet,Journalof 8  ExperimentalPsychology112,384412.#XUXXXU'# `(#(# 0   Gafos,A.:1996,TheArticulatoryBasisofLocalityinPhonology.PhDdissertation(inpreparation),  8  TheJohnsHopkinsUniversity,MD,ms. (#(# 0   Gafos,A.:1995a,CorrespondenceinTemiar:NoNeedforLongDistanceSpreadingHere,inProceedingsoftheFifthMeetingoftheSoutheastAsianLinguisticsSociety,Tucson, p Arizona,May20,1995. (#(# 0   Gafos,A.:1995b.OntheNotionofParadigmStructureinTemiar,TheJohnsHopkinsUniversity,ms. (#(# 0   Greenberg,J.:1950,ThePatterningofRootMorphemesinSemitic,Word6,162181. (#(# XUXXXU0   Hulst,H.andJ.Weijer:1995,VowelHarmonyinJ.Goldsmith(ed.)HandbookofPhonological  Theory,Blackwell,Oxford,495534.(#(# 0   Inkelas,S.:1994,TheConsequencesofOptimizationforUnderspecification,UniversityofCalifornia,Berkeley,ms.#XUXXXUB,#0" (#(# 0   It=,J.:1984,MelodicDissimilationinAinu,LinguisticInquiry15,505513.$X"(#(# 0   It=,J.andA.Mester:1993,LicencedSegmentsandSafePaths,CanadianJournalofLinguistics %0!$ 38:197214. (#(# 0   Jenner,P.,L.ThompsonandS.Starosta:1976,AustroasiaticStudies,Vol.II.Honolulu:University )$( PressofHawaii.h+&*(#(# L?T:H  0   Jenner,P.:1969,AffixationinModernKhmer.UnpublishedPh.D.dissertation,Universityof  Hawaii. (#(# 0   Lieberman,P.andS.Blumstein:1988,SpeechPhysiology,SpeechPerception,andAcoustic ` Phonetics.CambridgeUniversityPress.8 (#(# 0   Lombardi,L.:1991,LaryngealFeaturesandLaryngealNeutralization,XUXXXUUnpublishedPh.D.  ` dissertation,UniversityofMassachusetts,Amherst,MA#XUXXXU62#. 8 (#(# 0   Lombardi,L.andJ.McCarthy:1991,ProsodicCircumscriptioninChoctawMorphology,Phonology8,3771. (#(# 0   Matisoff,J.:1973,TonogenesisinSoutheastAsia,inL.Hyman(ed.)ConsonantTypesandTone, p SouthernCaliforniaOccasionalPapersinLinguistics,1,7195. (#(# 0   McCarthy,J.:1979,FormalProblemsinSemiticPhonologyandMorphology,Ph.D.dissertation,  p MIT;publishedbyGarlandPublishers,NewYork. (#(# 0   McCarthy,J:1981,AProsodicTheoryofNonconcatenativeMorphology,LinguisticInquiry   12:373418. (#(# 0   McCarthy,J.:1982,ProsodicTemplates,MorphemicTemplates,andMorphemicTiers,inv.d.HulstandN.Smith(eds.)TheStructureofPhonologicalRepresentations,PartI,Foris, X  Dordrecht. (#(# 0   McCarthy,J.:1986,OCPeffects:GeminationandAntigemination,LinguisticInquiry7:207-263.$X"(#(# XUXXXU0   0(#(#  McCarthy,J.:1988,FeatureGeometryandDependency:AReview,Phonetica45,84108.#XUXXXU7#%0!$(#(# 0   McCarthy,J.:1989,LinearOrderinPhonologicalRepresentation,LinguisticInquiry20:71-99.'#&(#(# 0   McCarthy,J.:1994,OnCoronal Transparency.PaperpresentedattheannualTrilateralPhonologyWeekend,UCSantaCruz. (#(# 0   McCarthy,J.:1993,TemplateForminProsodicMorphology,inL.Stran(ed.),Papersfromthe3rd @-(, AnnualMeetingoftheFormalLinguisticSocietyofMidAmerica,187218. (#(# 0   McCarthy,J.:1995,ExtensionsofFaithfulness:RotumanRevisited,UniversityofMassachusetts,Amherst,MAm,ms. (#(# XUXXXU0   McCarthy,J.andA.Prince:1986.ProsodicMorphology.UniversityofMassachusetts,Amherst,and 8  BrandeisUniversity,ms.#XUXXXU;# `(#(# 0   McCarthy,J.andA.Prince:1990a,FootandWordinProsodicMorphology:TheArabicBrokenPlural,NaturalLanguageandLinguisticTheory8:209283. (#(# 0   McCarthy,J.andA.Prince:1990b,ProsodicMorphologyandTemplaticMorphologyinM.EidandJ.McCarthy(eds.)PerspectivesonArabicLinguistics:PapersfromtheSecond p Symposium151,Amsterdam:Benjamins,Amsterdam.XUXXXUH(#(# 0   McCarthy,J.andA.Prince.1993a.ProsodicMorphologyI:ConstraintInteractionand  p Satisfaction.UniversityofMassachusettsandRutgersUniversity,ms.H(#(# 0   McCarthy,J.andA.Prince:1993b.GeneralizedAlignment,inG.BooijandJ.v.Marle(eds.),YearbookofMorphology,Dordrecht:Kluwer,79153#XUXXXU>#XUXXXU.(#(# 0   McCarthy,J.andA.Prince:1994,TheEmergenceoftheUnmarked:OptimalityinProsodicMorphology,inM.Gonzalez(ed.)ProceedingsoftheNorthEastLinguisticsSociety24, X  Amherst,MA,GraduateLinguisticStudentAssociation,333379. (#(# 0   McCarthy,J.andA.Prince:1995a,ReduplicativeIdentity,toappearinR.Kager,H.v.dHulstandW.Zonneveld(eds.)ProceedingsoftheOTS/HILWorkshopinProsodicMorphology.%0!$(#(# #XUXXXU@#XUXXXU0   McCarthy,J.andA.Prince:1995b,ProsodicMorphology,in#XUXXXU{C#XUXXXUinJ.Goldsmith,(ed.)Handbookof '#& PhonologicalTheory,Blackwell,Oxford,318366.#XUXXXUD#XUXXXU)$((#(# D?T<H  0   NChiosin,M.andJ.Padgett:1993.InherentVplace,Reportno.LRC9309,LinguisticsResearchCenter,UCSantaCruz.#XUXXXUD#XUXXXU@-(,(#(# D?T<HH  0   Odden,D.:1987,ArgumentsAgainsttheVowelPlaneinGta#XUXXXUE#Xl XXXU/#XUX XXlF#XUXXXU,LinguisticInquiry18,523529.(#(# 0   hman,S.:1966,CoarticulationinVCVUtterances:SpectrographicMeasurements,Journalof  theAcousticalSocietyofAmerica39,151168.t(#(# 0   hman,S.:1967,NumericalModelofCoarticulation,JournaloftheAcousticalSocietyof L  America41,310320.#XUXXXUG#XUXXXU$ t(#(# 0   Padgett,J.:1991,OCPSubsidiaryFeatures,inProceedingsofNELS21,GLSA,Departmentof  L  Linguistics,UniversityofMassachusetts,MA.#XUXXXU-I#XUXXXU$ (#(# 0   Padgett,J.:(toappear),FeatureClasses,inUniversityofMassachusettsOccasionalPapers,   SpecialIssueonOptimalityTheory. (#(# 0   Pierrehumbert,J.:1992,DissimilarityintheArabicVeralRoots,inProceedingsofNELS23, \ GLSA,DepartmentofLinguistics,UniversityofMassachusetts,MA.#XUXXXUVJ#XUXXXU4(#(# 0   Paradis,C:1988,OnConstraintsandRepairStrategies,TheLinguisticReview6,7197. \(#(# 0   Purcell,E.:1979,FormantFrequencyPatternsinRussianVCVUtterances,Journalofthe 4 AcousticalSocietyofAmerica66,16911702.#XUXXXU`L#XUXXXU (#(# 0   Prunet,J.F.:1995,SomeCorePropertiesofSemiticMorphology:EvidencefromtheFarSouth,HandoutoftalkpresentedattheCurrentTrendsinPhonology:ModelsandMethods, l  Royaumont,June1921,1995. (#(# 0   Recasens,D.:1984,VoweltoVowelCoarticulationinCatalanVCVSequences,Journalofthe $l" AcousticalSocietyofAmerica76,16241635.#XUXXXUN#XUXXXU%D!$(#(# 0   Ruhlen,M.:1987,AGuidetotheWorldsLanguages,Vol.1:Classification.StanfordUniversity '#& Press. (#(# 0   Prince,A.&P.Smolensky.1993.OptimalityTheory:ConstraintInteractioninGenerative |+&* Grammar,ms.,RutgersUniversity,NewBrunswick,andUniversityofColorado,Boulder. T-(, [Toappear,MITPress.]#XUXXXUZP#XUXXXU(#(# 0   Shaw,P:1991,ConsonantHarmonySystems:TheSpecialStatusofCoronalHarmony,inC.Paradis&J.F.Prunet(eds.)PhoneticsandPhonology2,TheSpecialStatusofCoronals: ` InternalandExternalEvidence.AcademicPress,#XUXXXU{R#XUXXXU25155.#XUXXXUS#XUXXXU8 (#(# 0   Smith,N.:1985:Spreading,Reduplication,andtheDefaultOptioninMiwokNonconcatenativeMorphology,inH.v.d.HulstandN.Smith(eds.)AdvancesinNonlinearPhonology,Foris  8  Dordrecht. (#(# 0   Smolensky,P.1993,Optimality,Markedness,andUnderspecification,PaperpresentedattheRutgersOptimalityWorkshopI,NewBrunswick,NJ.#XUXXXU?T#XUXXXUp(#(# 0   Sproat,R.andO.Fujimura:1993,AllophonicVariationinEnglish/l/anditsImplicationsforPhoneticImplementation,JournalofPhonetics21,291311.#XUXXXU]V#XUXXXU p(#(# 0   Steriade,D.:1986,YokutsandtheVowelPlane.LinguisticInquiry17,1:129146.H(#(# 0   Steriade,D.:1987,LocalityConditionsandFeatureGeometry,inProceedingsoftheNorthEast   LinguisticsSociety17,Amherst,MA,GraduateLinguisticStudentAssociation,595617.#XUXXXU}W#XUXXXU(#(# 0   Steriade,D.:1988,ReduplicationandSyllableTransferinSanskritandElsewhere,Phonology5,  73155. (#(# 0   Steriade:1995,UnderspecificationandMarkedness,inJ.Goldsmith,(ed.)Handbookof 0"  PhonologicalTheory,Blackwell,Oxford,114174.#XUXXXUbY#XUXXXU$X"(#(# 0   Svantesson,J-O.:1983,KammuPhonologyandMorphology.TravauxdeL'InstitutLinguistiquede %0!$ Lund,XVIII.#XUXXXU<[#XUXXXU'#&(#(# 0   Yip,M.:1988,TemplateMorphologyandtheDirectionofAssociation,NaturalLanguageand )$( LinguisticTheory6,551577.#XUXXXU9\#h+&*(#(#  @-(, XUXXXU&&XXU#XUX&&]##XUXXXU]#