WPCL aښ?u'WgoFbT@|17^4_d=*_(_MFDYc-dXB &v# "hcw\9*B~Hťt},)MD8O"T#- ૆}_&JYߜ( Τ˓ !8I$^9— ^3.RZ7Rb4˂VV]JڳU_7yѷː8!Le-'VE9ntT?v bv-gf }eމkʉ.Nt*W'k,w: ={":ǩX\OEvu*/5^md.Qqj Ze$ag`$WV09;7MTuF*LM\%I`ڦ4S=T~P*z*0'[ 0 0 0 0L 0 0 0I 0 0 0Z 0 0y 0k 0kz 0~ 0~c 0 0 0 0 ! 0" 01# 0C$ 0V)&U:' %' 0'U., 1i, 0]Q- 0j- 0e. 0W}. 0W. 0+/ 0j/ 00 00 01 002 03 04 0 5 0-&6 0@S7 08 0N9 09 0j: 0; 0; 0b< 0%= 0= 0zA> 0> 1? 0@ }A 0=A 0TA 0PB 1oB 0CC 0U=GEG 0}G 0xH 0HfIU@I 0*CXIIL6K# LMU<PPaS D3S CS4Sf,\ B*.\ BcX\ B u\U@\a\a\f\Q\ BPeUNme e 0!effafCfUNhU@AiiUN1kU@kfkfkak 1k D+ll 0Um 0)m AMmJn 0Jo AMZoo 0)mp 0Up AQps AYgssn[s 0:+{ 1me{2{ B7~ C;~ CU~UNo~4~ B%5~faU@ zJ Bĉfaz2qBfavafafafʗa̗UN.f6a8LfPaRf B*CmUN;U@6U N$-rU@: zcUNwaU@1sqUNf24fafafsaU@fU@af a !U@@aU@/-oU@UN*U@(hb#U@+QkU NU@ JUNU@h]U@]EU@]U@?]U@]U@y Rc h{  h  h >  h  h  U@b 64U.j h h h h hb h h h h hg  m  hz$ h$$ hM(S( hQ+ hW+ h]+fc+ae+fy+a{++ h11!'3 hH5fN5aP5 hd5 hj5 hp5 h5f5a5 h55 hL9 hR9 h,X9 h99 h1;7; hR==f#>a%>%>f9>a;>;>;>fO>aQ>fe>ag>f{>a}>f>a>f>a>f>a>>>>>>>>>>>>>>>f>a>>f@a@@@@@@@@sFFFFJl83:Technical[1]Technical Document Style  23  Ԁ     .. l-(:Technical[2]Technical Document Style 23  Ԁ   .. l*$:Technical[3]Technical Document Style 23  Ԁ   .. l&$:Technical[4]Technical Document Style 23  Ԁ   .. l/:Technical[5]Technical Document Style.. 23  Ԁ. l/:Technical[6]Technical Document Style.. 23  Ԁ. l/:Technical[7]Technical Document Style.. 23  Ԁ. l/:Technical[8]Technical Document Style.. 23  Ԁ. VV!8Document[1]Document Style  @..  23  Ԁ     Ԉ V-8Document[2]Document Style 23  Ԁ   V=8Document[3]Document Style.. 23  Ԁ   V 8Document[4]Document Style.. . V 8Document[5]Document Style0..V 8Document[6]Document Style8..V& 8Document[7]Document Style0..0` ..` V& 8Document[8]Document Style0..8` ..` zU :Right Par[1]Right-Aligned Paragraph Numbers..2I.3  Ԁ..0..zh :Right Par[2]Right-Aligned Paragraph Numbers..` ..2A.3  Ԁ..0` ..` z{ :Right Par[3]Right-Aligned Paragraph Numbers..` ..`  ..P 21.3  Ԁ` ..` 0 .. z :Right Par[4]Right-Aligned Paragraph Numbers..` ..`  .. .. 2a.3  Ԁ .. 0..z :Right Par[5]Right-Aligned Paragraph Numbers..` ..`  .. ..h..2(1)3  Ԁ..0h..hz :Right Par[6]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h..2(a)3  Ԁh..h0..z :Right Par[7]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....2i)3  Ԁ..0..z :Right Par[8]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....p..2a)3  Ԁ..0p..p<6X9`("Courier NewTTgX64MACNormal            _UX` hp x (#%'0*,.8135@8:<_        #d#  :::C7D4A`ArialTTCS  $ I. A. 1. a.(1)(a) i) a)S ,   ,  .  ,     &:d& CE              _UX` hp x (#%'0*,.8135@8:<_        #:d#  :::C7D4A`ArialTTC .  ,     &:d& CE   7D4A`ArialTT$35;AGMU]c1I.A.1.a.(1)(a)i)a)20Endnote    .    B& $2Document Style0..8` ..` B $3Document Style.. . B $4Document Style8..B $5Document Style0..B-$6Document Style 23  Ԁ   B& $7Document Style0..0` ..` dS $8Right-Aligned Paragraph Numbers..23  Ԁ..0..df $9Right-Aligned Paragraph Numbers..` ..` 23  Ԁ..0` ..` D=&10Document Style.. 23  Ԁ   fy &11Right-Aligned Paragraph Numbers..` ..`  .. 23  Ԁ` ..` 0 .. f &12Right-Aligned Paragraph Numbers..` ..`  .. ..23  Ԁ .. 0..f &13Right-Aligned Paragraph Numbers..` ..`  .. ..h..h23  Ԁ..0h..hf &14Right-Aligned Paragraph Numbers..` ..`  .. ..h..h..23  Ԁh..h0..f &15Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....23  Ԁ..0..f &16Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....p..p23  Ԁ..0p..pDV!&17Document Style  @..  23  Ԁ     Ԉ X/&18Technical Document Style.. 23  Ԁ. X/&19Technical Document Style.. 23  Ԁ. X-(&20Technical Document Style 23  Ԁ   .. X*$&21Technical Document Style 23  Ԁ   .. X)$&22Technical Document Style 23  Ԁ   .. X83&23Technical Document Style  23  Ԁ     .. X/&24Technical Document Style.. 23  Ԁ. X/&25Technical Document Style.. 23  Ԁ. R& 8BibliogrphyBibliography0....jo4Tech InitInitialize Technical StyleS 9  1 .1 .1 .1 .1 .1 .1 .1 S CuyTechnicalTechnical Document Style11.11.1.11.1.1.11.1.1.1.11.1.1.1.1.11.1.1.1.1.1.11.1.1.1.1.1.1.1fp2Doc InitInitialize Document Style;S ?    I. 1. A. a.(1)(a) i) a)S<=>($0 (,!$0  0` (#(#  (($0 0 0   A_ekqwDocumentDocument StyleI.1.A.a.(1)(a)i)a)x?Su2PleadingHeader for Numbered Pleading Paper A z B(  XXXXX6VC'*d66VC'*d6HH 1HH 2HH 3HH 4HH 5HH 6HH 7HH 8HH 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28  .+(-2$ \!CEUS.,  p4Heading 2Underlined Heading Flush LeftV 4Heading 1Centered Heading@..  b* 8Bullet ListIndented Bullet List0..0` ..` \  `(Times NewRomanTT($      `$CEUS.,  XX XXgX X ,  7  X3XHX  `  SeeIt=andMester's(1996)extensionofthisapproachtoJapaneseRendaku,inwhichasimilar  'Neighborhood'constraintisproposedwhichdoesnotrequirefeaturaloverlap. B \!CEUS.,  HX XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  H3 hHX   1  HP DeskJet 500HPDSKJET,,,,0   \!CEUS.,  HX XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi    5  H3 XHX HereIamassumingthattheInputismadeupofalinearlysequencedsetofmorphemes.Itisinfactnot  crucialtotheanalysisthatthispositionbemaintained,sinceitisonlyLinearitywithintherootthatmust h beobeyed,andthereareotherwaysofrulingouttransmorphemicnasalsubstitution,suchasthroughtheuseofDisjointnessconstraints(McCarthyandPrince1995).' HA`$WP CyrillicATT  \!CEUS.,  g3 XgXH3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    2  IamgratefultoJohnKingstonandDoncaSteriadeforveryhelpfuldiscussionofthephoneticfacts,  thoughIhastentoclaimsoleresponsibilityforanyerrorsofinterpretation.SeealsoHayes(1995)forasomewhatdifferenthypothesisaboutthephoneticgroundingofH3 H3*N)Cll(.#H3 H3f#Table_C]]]]'dxd d *CEUS.,g3 XgXH3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    17  #g3  H31##gX X g3#US.,CE.,g3 XgX ԀUKUS.,Theseexamplesalsodemonstratethewellknowncomplicationthat/s/becomesapalatalnasalunder  substitution.TheapparentoddnessofthisalternationissomewhattemperedbytheindependentevidencefromaJavanesemorphemestructureconstraintthatAustronesian/s/isinfactitselfphonologicallypalatal(Mester1986).US.,UK.,Arelatedcomplicationisthatnasalsubstitutionalsooftenfailstooccurwitha/c/initialroot(/c/isvariouslydescribedasapalatalstoporanalveopalatalaffricate).Basedonthefactthatavoidanceofhomophonywith/s/initialrootsseemstodeterminewhethersubstitutionwith/c/canoccurornot,Onn(1980:62)suggeststhatresistancetosubstitutionaroseasapropertyoftheroots,ratherthanof/c/itself,whichisperhapsspreadingtoother/c/initialrootsbyanalogy(thegeneralizationsabouttheoccurrenceofsubstitutionwith/c/reportedbyOnnforJohoreMalaywereconfirmedforBahasaIndonesiabymyconsultant,ChoirulDjamhari).Lexicalexceptionssuchasthesecanbecapturedbymeansoflexicallyspecificconstraintrankings,forwhichwemightextendtheschemaforRootspecificconstraints,sothatconstraintsapplytoaspecifiedsubsetoflexicalitems.Forinstance,thesewordsmightbesubjecttoalexicallyspecifichighrankingofafeaturalIdentityconstraint,forcingallInputoral,voiceless    consonantstobefaithfullyrepresentedassuchintheOutput.SeePater1995fordevelopmentofthis h   approachtolexicalexceptionalityinthecontextofEnglishstress.&0 d d dd\  `(Times NewRomanTTTable_ATable_C Gg3 XgXCEUS.,H3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    3  #g3  H31#US.,CE.,ԀThediscussionhereabstractsfromtwootherH3  g3CEUS.,H3 H3N)Cll(#g3H3#US.,CE.,#g3 g3#Ԁeffects:nasaldevoicingandobstruentaspiration.These  processescannotbecapturedbythesimplestatementof*H3  g3CEUS.,H3 H3N)Cll(#g3H3#US.,CE.,Ԁin#g3 g3#Ԁ(3).Itisconceivablethatthearticulatory H orperceptualdifficultiesofpostnasalvoicelessnesscouldbeovercomebyenhancementwithaspirationand/orextensionofthedurationofvoicelessness.However,apropertreatmentofthesephenomenawouldforcealongdigressionfromthecentralconcernsofthispaper,sinceatleastthefollowingrathercomplexquestionswouldhavetobeanswered:Whatisthenatureoftheinteractionbetweentheseprocesses:doesdevoicingresultfromaspiration,orviceversa(Herbert1986,NurseandHinnebusch1993)?Are @  voicelessnasals[Voice],or[+Aspirated](Lombardi1991,Huffman1994)?Arethevoicelessnasalsinfactevenentirelyvoiceless(MaddiesonandLadefoged1993:262)?Relatedtothelastquestion,aretheseprocessescategoricalormoreimplementationalinnature?Therefore,forpresentpurposesIleaveg3 g3*H3g3CEUS.,N)Cll(#g3H3d#US.,CE.,Ԁin#g3 g3D# 0 itsperhapsoverlysimpleform.K Kd`6Times New Roman RegularH(jG2$ \!CEUS.,    HX&XXgX    0  H# XHX&Table_B 9 \!CEUS.,  d3 XgXe3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    14  Onecouldevenimagineahybridtheory.Featuresthatdisplayclearindependencefromsegments,most  prominentlytones,mightbesubjecttoCorrespondencerequirements,whilethosethatdonotwouldbetargetedbyIdentity.9 Z6Times New Roman Regular\  `(Times NewRomanTT  `$CEUS.,  hX XXgX X ,  13  h3XHX  `  ݀Sincethiswasfirstwritten,Lombardi(1995)hasfounda'happyresult'inonedomain,while  Aldereteetal.(1996)finda'fatalflaw'inanother.Needlesstosay,theissueisfarfromsettled.9 Z6Times New Roman Regular\  `(Times NewRomanTTTable_B CRight ParRight-Aligned Paragraph NumbersI.A.1.a.(1)(a)i)a)'dxd Level 1 Level 2 Level 3 Level 4 Level 5(-2$ \!CEUS.,  ($nopmA<< g Level 1 Level 2 Level 3 Level 4 Level 5(P$ Table  1  rosmA<< g Level 1 Level 2 Level 3 Level 4 Level 5($(-2$ \!CEUS.,  uwovmC<< g Level 1 Level 2 Level 3 Level 4 Level 5($yozmA<< g Level 1 Level 2 Level 3 Level 4 Level 5($|o}mM << ddUU3|P2<6X9`("Courier NewTTXXx6X@DQX@7D4A`ArialTTTT::}D4P :P%2A`ArialTTTT::T2PQ:P%2A`ArialTTTTXXX2PQXP\  `(Times NewRomanTT^\  P6QP\  `(Times NewRomanTTXXP\  P6QXP)4 k" `,SILManuscriptIPATTXXt4 k" PqQXP:\ Y `SymbolTTmanTTTXXt\  P[AXP' HA`$WP CyrillicATTTTb H PQP)4 k" `,SILManuscriptIPATT4 k" PqQP*{2A`8WP MultinationalARomanTTXXx{2PQXP(\$CEUS.,# e37=CIQYag1.a.i.(1)(a)(i)1)a) ( \!CEUS.,  3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    12  Millsdoesnotcommentonnasal/s/clusters,butasfarasIcantellfromPelenkahuetal.(1983),the  samerestrictionholdsasforthestops,sincetherearemanyexamplesof/ss/,butnoneof/ns/.3+0 d d dd2d9 Z6Times New Roman RegularRaW!ddd\ 1Large Dashed dTable_C CEUS.,g3 XgXH3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    10  #g3  H31##gX X g3#US.,CE.,g3 XgX ԀOnepathtoexplanationmaylieinthefactthatH3  g3CEUS.,H3 H3N)Cll(#g3H3#US.,CE.,Ԁ#g3 g3 #sequencestendtobeplaceassimilated,andthus  resistepenthesisduetosomeversionofgeminateintegrity.However,thisexplanationisdifficult,ifnotimpossibletoformalizeinOptimalityTheory(whyshouldplaceassimilationhaveprecedenceoverg3 g3*H3g3CEUS.,N)Cll(#H3 H3##g3  H3#US.,CE.,?),andfacestheempiricalchallengeg3 g3thatH3g3CEUS.,N)Cll(#g3H3#US.,CE.,Ԁeffects#g3 g3#Ԁdooccurintheabsenceofplaceassimilationin x severallanguages(see4.1).\  `(Times NewRomanTT p3 XgXCEUS.,3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    11  #3  H31#US.,CE.,ԀInTobaBatak,theobstruentsproducedbydenasalizationfailtoundergothedebuccalizationthataffects  otherobstruentsinthesameposition.Hayes(1986)attributesthistoatypeofgeminateinalterability,withthedoublelinkingofa[Voice]featurespreadfromthefollowingvoicelessconsonantinhibitingdebuccalization.Moreplausibly,thisacaseofavoidanceofneutralization.Thatis,underlyingnasalsfailtogoallthewaytoglottalssoastoavoidneutralizingthedistinctionbetweenthemandunderlyingobstruents.SeeFlemming(1995)fordiscussionoftheformalissuesinvolvedinsettingupcontrastmaintainingconstraints;seealsoMcCarthy(1993)andKirchner(1995)forotherapproachestochainshiftsinOptimalityTheory. dTable_E pg3 XgXCEUS.,H3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    11  #g3  H31#US.,CE.,ԀInTobaBatak,theobstruentsproducedbydenasalizationfailtoundergothedebuccalizationthataffects  otherobstruentsinthesameposition.Hayes(1986)attributesthistoatypeofgeminateinalterability,withthedoublelinkingofa[Voice]featurespreadfromthefollowingvoicelessconsonantinhibitingdebuccalization.Moreplausibly,thisacaseofavoidanceofneutralization.Thatis,underlyingnasalsfailtogoallthewaytoglottalssoastoavoidneutralizingthedistinctionbetweenthemandunderlyingobstruents.SeeFlemming(1995)fordiscussionoftheformalissuesinvolvedinsettingupcontrastmaintainingconstraints;seealsoMcCarthy(1993)andKirchner(1995)forotherapproachestochainshiftsinOptimalityTheory. ( \!CEUS.,  g3 XgXH3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    12  Millsdoesnotcommentonnasal/s/clusters,butasfarasIcantellfromPelenkahuetal.(1983),the  samerestrictionholdsasforthestops,sincetherearemanyexamplesof/ss/,butnoneof/ns/. 8 \!CEUS.,  g3 XgXH3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    14  Onecouldevenimagineahybridtheory.Featuresthatdisplayclearindependencefromsegments,most  prominentlytones,mightbesubjecttoCorrespondencerequirements,whilethosethatdonotwouldbetargetedbyIdentity. g3 XgXCEUS.,H3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    15  #g3  H31#US.,CE.,ԀUKUS.,Itshouldbenotedthatthesediagramsdonotrepresentautosegmentalmappings;rather,theyillustrate  thesettheoreticrelationshipbetweentheInputandOutputsetsofsegments.US.,UK.,fԀInaddition,thereisnotheoreticalstanceimplicitintherepresentationofthegeminated/t/asapairofsegments.Thisrepresentationisusedbecausedenasalizationsometimesproducesanonassimilatedsegment(KaingangandTobaBatak),andbecausetheresultsintermsofCorrespondenceandIdentityarethesameifasingle/t/isusedforageminate.DifferentresultsintermsofLinearitymightobtaindependingonwhether  geminateswereconsideredasinglesegmentwithamora,ortwosegmentswithlinkedfeatures.Table_DTable_ATable_DTable_CTable_C9 Z6Times New Roman RegularTable_A  `$CEUS.,  HX XXgX X ,  21  H3XHX  `  ݀H3 H3Seeg3  H3ԀhoweverKawasaki(toappear),inwhichredundantfeaturelicensingissupplementedbya  principleofgovernment#H3  g3#Ԁthatproducestherequiredasymmetryinnasalobstruentvoicing.Table_A g3 XgXCEUS.,H3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    22  #g3  H31#US.,CE.,ԀJayePadgett(p.c.)notesthefollowingcomplication.Thoughvoicinginteractionsareperhapssonority  sensitiveinmakingadistinctionbetweenobstruentsandsonorants,theydonotappeartodividethesonorityscaleinthesamewayasplaceinteractions.Casesinwhichonlyfricativesinteractinvoicing,forinstance,seemratherunlikely.&p d Pd vCEUS., X ,    iUX` hp x (#%'0*,.8135@8:<Xi    16  US.,CE.,Ԁg3 XgXThisleavesanotinsignificantproblemunresolved.Howdowedistinguishbetweennasalizationofthe  voicelessstop,andnasalsubstitution?Intermsoftheconstraintsconsideredthusfar,nasalsubstitutionincursalltheviolationsthatnasalizationdoes,plusaLinearityviolationthatisavoidedbynasalization. P Onepossiblykeydifferenceisthatinfusion,oneoftheunderlyingcorrespondentsoftheOutputnasalisanasal,whileinnasalizationthesecondmemberoftheclusterhasasitssolecorrespondentavoicelessobstruent.IshouldalsonoteherethatKonjonasalizationissubjecttoconsiderablemorphologicalconditioning.Infact,theprefixthatcausesnasalizationhasahomophonouscounterpartthatdiffersonlyinthatitfailstonasalizethefollowingvoicelessobstruent.(  Z 6Times New Roman Regular  `$CEUS.,  HX XXgX X ,  4  H3XHX  `  InusingLinearitytoblockfusion,IadoptasuggestionofJohnMcCarthy's(p.c.).McCarthyand  Prince(1995,thisvolume)invokeUniformityforsuchcases.IhaveretainedLinearitybecauseitisstill H notentirelyclearthataseparateUniformityconstraintisneeded,andbecauseLinearityhassome 0 interestingpotentialextensionsinthefeaturaldomain,whicharenotedbelowinthetext.\  `(Times NewRomanTT CEUS.,3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    10  #3  H31##X X g3#US.,CE.,3 XgX ԀOnepathtoexplanationmaylieinthefactthat3  g3CEUS.,3 H3N)Cll(#3H3#US.,CE.,Ԁ#3 g3 #sequencestendtobeplaceassimilated,andthus  resistepenthesisduetosomeversionofgeminateintegrity.However,thisexplanationisdifficult,ifnotimpossibletoformalizeinOptimalityTheory(whyshouldplaceassimilationhaveprecedenceover3 g3*3g3CEUS.,N)Cll(#3 H3##3  H3#US.,CE.,?),andfacestheempiricalchallenge3 g3that3g3CEUS.,N)Cll(#3H3#US.,CE.,Ԁeffects#3 g3#Ԁdooccurintheabsenceofplaceassimilationin x severallanguages(see4.1).9 Z6Times New Roman Regular # `$CEUS.,  HX XXgX X ,  6  H3XHX  `  g3H3USCE.,Itshouldbenotedthat#H3g3#CE.,US.,UK.,CE.,Ԁfusionisnotfreetooccurbetweenanytwomorphemes.CE.,UK.,Rg3H3USCE.,Boththeprefix+prefix  CE.,US.,USCE.,#H3g3#g3H3androot+suffixboundariesareimpermeabletonasalsubstitution(e.g./mBN+pBr+besar/[mBmpBrbesar] H toenlarge#H3g3e#g3H3CE.,US.,0USCE.,Ԁand/mBN+yakin+kan/[mByakinkan]toconvinceCE.,US.,|USCE.,#H3g3C#g3H3).Toencodethissortofmorphological 0 conditioning,constraintsareneededtorenderparticularmorphemeboundariesopaquetofusion.Inparticular,McCarthyandPrince's(1995)CE.,US.,USCE.,ԀDisjointnessconstraints,whichrequirethatthesetsofcorrespondents(or ` exponents)ofmorphemesbenonoverlapping,couldberecruitedforthispurpose.#H3g3#CE.,US.,  `$CEUS.,  HX XXgX X ,  7  H3XHX  `  SeeIt=andMester's(1996)extensionofthisapproachtoJapaneseRendaku,inwhichasimilar  'Neighborhood'constraintisproposedwhichdoesnotrequirefeaturaloverlap.\  `(Times NewRomanTT  `CEUS.,  HX XXgX X ,  18  H3XHX  `  H3 H3ԀTostemanysuspicionthatdeletionbeforethefricativesismotivatedsolelybythemarkednessof  nasal/fricativeclusters(seePadgett1994),IshouldnotethatvoicedfricativesundergopostnasalhardeninginKhehe(OddenandOdden1985:598).Thisshowsthat*#H3 H3#N)Cll(ԀiH3 H3sneededfordeletioninanasal/voiceless 0 fricativesequence,sinceonewouldotherwisepredictthat/ns/shouldsurfaceas[nt].  \!CEUS.,  3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    9  IndiscussingtheseAfricanlanguagesIfollow,foreaseofexposition,Herbert(1986)andPadgett(1994)  intreatingderivedprenasalizedstopsassegmentalsequences(cf.Piggott1992,andSteriade1993forotherviewsonprenasalization).Itshouldbeemphasized,though,that'segment'inCorrespondencetheorymightwellbeunderstoodastheequivalenttowhatinfeaturegeometrictermsistherootnodeandeverythingitdominates(i.e.amelodicelement).TworootnodetheoriesofprenasalizedstopshavebeenproposedbyPiggott(1988),Rosenthall(1989),Trigo(1993),andtosomeextent,Steriade(1993),andPiggott(1995).  `$CEUS.,  HX XXgX X ,  13  H3XHX  `  ݀Sincethiswasfirstwritten,Lombardi(1995)hasfounda'happyresult'inonedomain,while  Aldereteetal.(1996)finda'fatalflaw'inanother.Needlesstosay,theissueisfarfromsettled.9 Z6Times New Roman Regular"2U222P2P2P2E2P2E2P2P2P2P2P2P2E2P2U22E22E22E2222[22[2D2P2E2\  `(Times NewRomanTT i \!CEUS.,  3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    8  Swahilinasaldeletionishistoricallyprecededbyaspirationofthefollowingvoicelessconsonant,which  spreadtothenasal,thoughthereisnoevidenceforthisintermediatestageintheotherlanguagescitedhere(seeHerbert1986:252,NurseandHinnebusch1993:168).9 Z6Times New Roman Regular  `$CEUS.,  HX XXgX X ,  20  H3XHX  `  ݀Inalldialects,thenasalisdeletedwithintheword(5b),andinmostdialects,includingCypriot,it  isdeletedinanarticleprecedinganoun,exceptinslow,deliberatespeech(5d).Table_FTable_E i \!CEUS.,  g3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    8  Swahilinasaldeletionishistoricallyprecededbyaspirationofthefollowingvoicelessconsonant,which  spreadtothenasal,thoughthereisnoevidenceforthisintermediatestageintheotherlanguagescitedhere(seeHerbert1986:252,NurseandHinnebusch1993:168).Table_F\  `(Times NewRomanTT  \!CEUS.,  g3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    9  IndiscussingtheseAfricanlanguagesIfollow,foreaseofexposition,Herbert(1986)andPadgett(1994)  intreatingderivedprenasalizedstopsassegmentalsequences(cf.Piggott1992,andSteriade1993forotherviewsonprenasalization).Itshouldbeemphasized,though,that'segment'inCorrespondencetheorymightwellbeunderstoodastheequivalenttowhatinfeaturegeometrictermsistherootnodeandeverythingitdominates(i.e.amelodicelement).TworootnodetheoriesofprenasalizedstopshavebeenproposedbyPiggott(1988),Rosenthall(1989),Trigo(1993),andtosomeextent,Steriade(1993),andPiggott(1995).\  `(Times NewRomanTTTable_FTable_E  `$CEUS.,  HX XXgX X ,  19  H3XHX  `  ݀ThankstoAdamantiosGafosforbringingNewton(1972)tomyattention.\  `(Times NewRomanTTTable_F  `$CEUS.,  X XXgX X ,  7  3XHX  `  SeeIt=andMester's(1996)extensionofthisapproachtoJapaneseRendaku,inwhichasimilar  'Neighborhood'constraintisproposedwhichdoesnotrequirefeaturaloverlap.\  `(Times NewRomanTT # `$CEUS.,  X XXgX X ,  6  3XHX  `  3H3USCE.,Itshouldbenotedthat#3g3#CE.,US.,UK.,CE.,Ԁfusionisnotfreetooccurbetweenanytwomorphemes.CE.,UK.,R3H3USCE.,Boththeprefix+prefix  CE.,US.,USCE.,#3g3#3H3androot+suffixboundariesareimpermeabletonasalsubstitution(e.g./mBN+pBr+besar/[mBmpBrbesar] H toenlarge#3g3e#3H3CE.,US.,0USCE.,Ԁand/mBN+yakin+kan/[mByakinkan]toconvinceCE.,US.,|USCE.,#3g3C#3H3).Toencodethissortofmorphological 0 conditioning,constraintsareneededtorenderparticularmorphemeboundariesopaquetofusion.Inparticular,McCarthyandPrince's(1995)CE.,US.,USCE.,ԀDisjointnessconstraints,whichrequirethatthesetsofcorrespondents(or ` exponents)ofmorphemesbenonoverlapping,couldberecruitedforthispurpose.#3g3#CE.,US.,\  `(Times NewRomanTT9 Z6Times New Roman Regular  \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi    5  3 XHX HereIamassumingthattheInputismadeupofalinearlysequencedsetofmorphemes.Itisinfactnot  crucialtotheanalysisthatthispositionbemaintained,sinceitisonlyLinearitywithintherootthatmust h beobeyed,andthereareotherwaysofrulingouttransmorphemicnasalsubstitution,suchasthroughtheuseofDisjointnessconstraints(McCarthyandPrince1995).\  `(Times NewRomanTT  `$CEUS.,  X XXgX X ,  4  3XHX  `  InusingLinearitytoblockfusion,IadoptasuggestionofJohnMcCarthy's(p.c.).McCarthyand  Prince(1995,thisvolume)invokeUniformityforsuchcases.IhaveretainedLinearitybecauseitisstill H notentirelyclearthataseparateUniformityconstraintisneeded,andbecauseLinearityhassome 0 interestingpotentialextensionsinthefeaturaldomain,whicharenotedbelowinthetext.- -\  `(Times NewRomanTT G3 XgXCEUS.,3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    3  #3  H31#US.,CE.,ԀThediscussionhereabstractsfromtwoother3  g3CEUS.,3 H3N)Cll(#3H3#US.,CE.,#3 g3#Ԁeffects:nasaldevoicingandobstruentaspiration.These  processescannotbecapturedbythesimplestatementof*3  g3CEUS.,3 H3N)Cll(#3H3#US.,CE.,Ԁin#3 g3#Ԁ(3).Itisconceivablethatthearticulatory H orperceptualdifficultiesofpostnasalvoicelessnesscouldbeovercomebyenhancementwithaspirationand/orextensionofthedurationofvoicelessness.However,apropertreatmentofthesephenomenawouldforcealongdigressionfromthecentralconcernsofthispaper,sinceatleastthefollowingrathercomplexquestionswouldhavetobeanswered:Whatisthenatureoftheinteractionbetweentheseprocesses:doesdevoicingresultfromaspiration,orviceversa(Herbert1986,NurseandHinnebusch1993)?Are @  voicelessnasals[Voice],or[+Aspirated](Lombardi1991,Huffman1994)?Arethevoicelessnasalsinfactevenentirelyvoiceless(MaddiesonandLadefoged1993:262)?Relatedtothelastquestion,aretheseprocessescategoricalormoreimplementationalinnature?Therefore,forpresentpurposesIleave3 g3*3g3CEUS.,N)Cll(#3H3d#US.,CE.,Ԁin#3 g3D# 0 itsperhapsoverlysimpleform.9 Z6Times New Roman Regular\  `(Times NewRomanTT  \!CEUS.,  3 XgX3  g3 X ,    iUX` hp x (#%'0*,.8135@8:<Xi    2  IamgratefultoJohnKingstonandDoncaSteriadeforveryhelpfuldiscussionofthephoneticfacts,  thoughIhastentoclaimsoleresponsibilityforanyerrorsofinterpretation.SeealsoHayes(1995)forasomewhatdifferenthypothesisaboutthephoneticgroundingof*X X H3N)C0#3 XHX g#.9 Z6Times New Roman Regular\  `(Times NewRomanTT S \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  3 hHX   1  ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnoted  otherwisethephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixis  employedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysisoftheassimilativebehaviouroftheprefix.\  `(Times NewRomanTT S \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  3 hHX   1  ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnoted  otherwisethephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixis  employedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysisoftheassimilativebehaviouroftheprefix.\  `(Times NewRomanTT S \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  3 hHX   1  ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnoted  otherwisethephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixis  employedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysisoftheassimilativebehaviouroftheprefix.\  `(Times NewRomanTT S \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  3 hHX   1  ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnoted  otherwisethephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixis  employedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysisoftheassimilativebehaviouroftheprefix.\  `(Times NewRomanTT S \!CEUS.,  X XXgX X ,    iUX` hp x (#%'0*,.8135@8:<Xi  3 hHX   1  ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnoted  otherwisethephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixis  employedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysisoftheassimilativebehaviouroftheprefix.\  `(Times NewRomanTT  `$CEUS.,  HX XXgX    1  H3XHX   H3 H3ThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnotedotherwise  thephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixisemployedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysis  oftheassimilativebehaviouroftheprefix.#H3 H3ñ# H ,          15  gXgXThereisnotheoreticalstanceimplicitintherepresentationofthegeminated/t/asapairofsegments.  Thisrepresentationisusedbecausedenasalizationsometimesproducesanonassimilatedsegment(KaingangandTobaBatak),andbecausetheresultsintermsofCorrespondenceandIdentityarethesameifasingle/t/isusedforageminate.DifferentresultsintermsofLinearitymightobtaindependingonwhethergeminateswereconsidered  asinglesegmentwithamora,ortwosegmentswithlinkedfeatures.#gXXgn#UKUS.,gXgXItshouldalsobenotedthatthesediagramsdo 0 notrepresentautosegmentalmappings;rather,US.,UK.,theUKUS.,yillustratethesettheoreticrelationshipbetweentheInputandOutputsetsofsegments.US.,UK.,#gXXg# Y `$CEUS.,  HXXXgX    *  HXHX  ݀gHgXXgUSCE.,Acknowledgments:Thispaperhasundergonesubstantialchangessinceitwasoriginally  presentedinUtrecht.ThecommentsofJohnKingston,JohnMcCarthy,andDoncaSteriadehaveparticularlyhelpedtoshapeitspresentform.ThanksalsotoAbbyCohn,EdwardFlemming,BruceHayes,JunkoIt=,TakakoKawasaki,JayePadgett,SharonRose,SuUrbanczyk,RachelWalker,andtheparticipantsinLING751,UMass,andtheRutgers/UMassJointClassMeeting,Spring1995,forusefulsuggestions,andtoHeatherGoadandGlynePiggottfortheirindispensibleguidance.IwouldalsoliketoacknowledgeChoirulDjamhariforhishelpwiththeIndonesiandata,andLisaTravisformakingitpossibleformetoworkwithhim,aswellasJanVoskuilforhishospitalityandassistanceinsecuringmanyoftheAustronesianmaterialsduringanalltooshortvisittoLeiden.ThisresearchhasbeensupportedbySSHRCCfellowshipno.752932773totheauthor,andSSHRCCgrantno.410920759toGlynePiggott.#HXXXgX##HXHX#CE.,US.,GT[ 4 `$CEUS.,  HXXXgX    2  HXHX  HHԀIamgratefultoJohnKingstonandDoncaSteriadeforveryhelpfuldiscussionofthephoneticfacts,thoughI  hastentoclaimsoleresponsibilityforanyerrorsofinterpretation.SeealsoHayes(1995)forasomewhatdifferenthypothesisaboutthephoneticgroundingof#HH߱#*N)Cll(.ai  `$CEUS.,  XTXXgX    4   XHXT  ݀InusingLinearitytoblockfusion,IadoptasuggestionofJohnMcCarthy's(p.c.).WhileMcCarthyandPrince  (1995,thisvolume)havesubsequentlyproposedaseparateUniformityconstraintforsuchcases(seealsoGnanadesikan H 1995),IhaveretainedLinearitybecauseitisstillnotentirelyclearthataseparateconstraintisinfactneeded,andbecause 0 Linearityhassomeinterestingpotentialextensionsinthefeaturaldomain,whicharenotedbelowinthetext.\  `(Times NewRomanTT- -       , `$CEUS.,  HXXXgX    3  HXHX  ݀gHggUSCE.,Thediscussionhereabstractsfromtwoother#Hg߲#CE.,US.,#HH#N)Cll(gHUSCE.,ggԀeffects:nasaldevoicingandobstruentaspiration.These  processescannotbecapturedbythesimplestatementof*#Hgߪ#CE.,US.,#HH#N)Cll(gHUSCE.,ԀinggԀ(3).Itisconceivablethatthearticulatoryorperceptual H difficultiesofpostnasalvoicelessnesscouldbeovercomebyenhancementwithaspirationand/orextensionofthedurationofvoicelessness.However,apropertreatmentofthesephenomenawouldforcealongdigressionfromthecentralconcernsofthispaper,sinceatleastthefollowingrathercomplexquestionswouldhavetobeanswered:Whatisthenatureoftheinteractionbetweentheseprocesses:doesdevoicingresultfromaspiration,orviceversa(Herbert1986,NurseandHinnebusch1993)?Arevoicelessnasals[Voice],or[+Aspirated](Lombardi1991,Huffman1994)?Arethevoicelessnasalsinfactevenentirelyvoiceless(MaddiesonandLadefoged1993:262)?Relatedtothelastquestion,aretheseprocessescategoricalormoreimplementationalinnature?Therefore,forpresentpurposesIleave#gg0#*#Hg#CE.,US.,N)Cll(gHUSCE.,ԀinggԀitsperhapsoverlysimple p  form.#HgE##HHy#CE.,US.,d\  `Tms Rmnj  `$CEUS.,  HXXXgX    5  HXHX  ݀HHHereIamassumingthattheInputismadeupofalinearlysequencedsetofmorphemes.Itisnotcrucialtothe  analysisthatthispositionbemaintained,sinceitisonlyLinearitywithintherootthatmustbeobeyed,andthereareother H waysofrulingouttransmorphemicnasalsubstitution,suchasthroughtheuseofDisjointnessconstraints(McCarthyand 0 Prince1995;seethefollowingnote).#HH߲#- -  `$CEUS.,  HXXXgX    16  HXHX  ݀gHggUSCE.,Thisleavesanotinsignificantproblemunresolved.Howdowedistinguishbetweennasalizationofthevoiceless  stop,andnasalsubstitution?Intermsoftheconstraintsconsideredthusfar,nasalsubstitutionincursalltheviolationsthatnasalizationdoes,plusaLinearityviolationthatisavoidedbynasalization.Onepossiblykeydifferenceisthatinfusion, 0 oneoftheunderlyingcorrespondentsoftheOutputnasalisanasal,whileinnasalizationthesecondmemberoftheclusterhasasitssolecorrespondentavoicelessobstruent.IshouldalsonoteherethatKonjonasalizationissubjecttoconsiderablemorphologicalconditioning.Infact,theprefixthatcausesnasalizationhasahomophonouscounterpartthatdiffersonlyinthatitfailstonasalizethefollowingvoicelessobstruent.#Hg߳##HH#CE.,US.,  `$CEUS.,  HXXXgX    1  HXHX  HHThoughthedialectsofMalayspokeninMalaysiaandIndonesiaaredistinctinsomeways,unlessnotedotherwise  thephenomenadiscussedherearecommontobothBahasaIndonesiaasdescribedinLapoliwa(1981),andCohnandMcCarthy(1994),amongstothers,andtheJohoredialectofMalaydescribedinOnn(1980)andTeoh(1988).TheIndonesiandatacitedareallfromLapoliwa(1981).BothChamorroandMalagasyalsodisplayessentiallythesamepatternasthatin(1),asdoanumberofotherlanguagesspokenintheIndonesianarchipelago.Theunspecifiednasalintheunderlyingformofthe/mBN/prefixisemployedonlyasamatterofconvention,anddoesnotimplyanyparticularanalysis  oftheassimilativebehaviouroftheprefix.#HH߱#  `$CEUS.,  HXXXgX    17  HXHX  ݀gHggUKCE.,Theseexamplesalsodemonstratethewellknowncomplicationthat/s/becomesapalatalnasalunder  substitution.TheapparentoddnessofthisalternationissomewhattemperedbytheindependentevidencefromaJavanesemorphemestructureconstraintthatAustronesian/s/isinfactitselfphonologicallypalatal(Mester1986).CE.,UK.,US.,CE.,Arelatedcomplicationisthatnasalsubstitutionoftenfailstooccurwitha/c/initialroot(/c/isvariouslydescribedasapalatalstoporanalveopalatalaffricate);seeOnn(1980:62)fordiscussion.#Hg߳##HH#CE.,US.,x  `$CEUS.,  HXXXgX    4  HXHX  ݀InusingLinearitytoblockfusion,IamadoptingasuggestionofJohnMcCarthy's(p.c.).WhileMcCarthyand  Prince(1995,thisvolume)havesubsequentlyproposedaseparateUniformityconstraintforsuchcases(seealso H Gnanadesikan1995),IhaveretainedLinearitybecauseitisstillnotentirelyclearthataseparateconstraintisinfact 0 needed,andbecauseLinearityhassomeinterestingpotentialextensionsinthefeaturaldomain,whicharenotedbelowin x thetext. Table_ATable_B  `$CEUS.,  HXXXgX    6  HXHX  ݀gHUSCE.,Itshouldbenotedthat#Hgܲ#CE.,US.,UK.,CE.,Ԁfusionisnotfreetooccurbetweenanytwomorphemes.CE.,UK.,7gHUSCE.,Boththeprefix+prefixCE.,US.,USCE.,#Hgܟ#gHand  root+suffixboundariesareimpermeabletonasalsubstitution(e.g./mBN+pBr+besar/[mBmpBrbesar]toenlarge#Hg6#gHCE.,US.,USCE.,Ԁand H /mBN+yakin+kan/[mByakinkan]toconvinceCE.,US.,MUSCE.,#Hg#gH).Toencodethissortofmorphologicalconditioning,constraintsareneeded 0 torenderparticularmorphemeboundariesopaquetofusion.Inparticular,McCarthyandPrince's(1995)CE.,US.,USCE.,ԀDisjointness x constraints,whichrequirethatthesetsofcorrespondents(orexponents)ofmorphemesbenonoverlapping,couldberecruitedforthispurpose.#Hg#CE.,US., v `$CEUS.,  HXXXgX    7  HXHX  ݀SeeIt=andMester's(1996)extensionofthisapproachtoJapaneseRendaku,inwhichasimilar'Neighborhood'  constraintisproposedwhichdoesnotrequirefeaturaloverlap.  `$CEUS.,  HXXXgX    8  HXHX  ݀HHSwahilinasaldeletionishistoricallyprecededbyaspirationofthefollowingvoicelessconsonant,whichspread  tothenasal,butthereisnoevidenceforthisintermediatestageintheotherlanguagescitedhere(seeHerbert1986:252,NurseandHinnebusch1993:168).#HH߲#Table_B),,0Table_C22  `$CEUS.,  HXXXgX    9  HXHX  ݀HHIndiscussingtheseAfricanlanguagesIfollow,foreaseofexposition,Herbert(1986)andPadgett(1994)in  treatingderivedprenasalizedstopsassegmentalsequences(cf.Piggott1992,andSteriade1993forotherviewsonprenasalization).Itshouldbeemphasized,though,that'segment'inCorrespondencetheorymightwellbeunderstoodastheequivalenttowhatinfeaturegeometrictermsistherootnodeandeverythingitdominates(i.e.amelodicelement).TworootnodetheoriesofprenasalizedstopshavebeenproposedbyPiggott(1988),Rosenthall(1989),Trigo(1993),andtosomeextent,Steriade(1993),andPiggott(1995).#HH߲#46888::  `$CEUS.,  HXXXgX    13  HXHX  ݀Sincethiswasfirstwritten,Lombardi(1995)hasfounda'happyresult'inonedomain,whileAldereteet  al.(1996)finda'fatalflaw'inanother.Needlesstosay,theissueisfarfromsettled.J  \!CEUS.,  ,    iUX` hp x (#%'0*,.8135@8:<Xi      14  gXgXԀOnecouldevenimagineahybridtheory.Featuresthatdisplayclearindependencefromsegments,most  prominentlytones,mightbesubjecttoCorrespondencerequirements,#gXXg#ԀgXgXwhilethosethatdonotwouldbetargetedby H Identity.#gXXg#MMMMMTable_DTable_ETable_CTable_CTable_ATable_DTable_CTable_CTable_A  `$CEUS.,  HXXXgX    18  HXHX  ݀HHSeegHԀhoweverKawasaki(1995),inwhichredundantfeaturelicensingissupplementedbyaprincipleof  government#Hg#Ԁthatproducestherequiredasymmetryinnasalobstruentvoicing.Table_A  `$CEUS.,  HXXXgX    10  HXHX  ݀gHggUSCE.,Onepathtoexplanationmaylieinthefactthat#Hg߳#CE.,US.,#HH#N)Cll(gHUSCE.,Ԁggsequencestendtobeplaceassimilated,andthusresist  epenthesisduetosomeversionofgeminateintegrity.However,thisexplanationisdifficult,ifnotimpossibletoformalizeinOptimalityTheory(whyshouldplaceassimilationhaveprecedenceover#gg#*#Hgܯ#CE.,US.,N)Cll(HHgHUSCE.,?),andfacestheempiricalchallenge#ggߎ#that 0 #Hgܭ#CE.,US.,N)Cll(gHUSCE.,ԀeffectsggԀdooccurintheabsenceofplaceassimilationinseverallanguages.#Hgߘ##HH#CE.,US., i `$CEUS.,  HXXXgX    11  HXHX  ݀gHggUSCE.,InTobaBatak,theobstruentsproducedbydenasalizationfailtoundergothedebuccalizationthataffectsother  obstruentsinthesameposition.Hayes(1986)attributesthistoatypeofgeminateinalterability,withthedoublelinkingofa[Voice]featurespreadfromthefollowingvoicelessconsonantinhibitingdebuccalization.Moreplausibly,thisacaseofavoidanceofneutralization.Thatis,underlyingnasalsfailtogoallthewaytoglottalssoastoavoidneutralizingthedistinctionbetweenthemandunderlyingobstruents.SeeFlemming(1995)fordiscussionoftheformalissuesinvolvedinsettingupcontrastmaintainingconstraints;seealsoMcCarthy(1993)andKirchner(1995)forotherapproachestochainshiftsinOptimalityTheory.#Hg߳##HH#CE.,US.,  `$CEUS.,  HXXXgX    12  HXHX  ݀HHԀMillsdoesnotcommentonnasal/s/clusters,butasfarasIcantellfromPelenkahuetal.(1983),thesame  restrictionholdsasforthestops,sincetherearemanyexamplesof/ss/,butnoneof/ns/.#HH߳# \!CEUS.,  TRX3'3'3' Letter3'T  2  22g[XgX2HXXXgX2 27^^%%dd7222    222220  X 222FinalVersion:222 220  22 2220 ` 22ToappearinKager,Ren222e222,HarryvanderHulst,andWimZonneveld(eds.),2222The  ProsodyMorph222p222ologyInter222ca222face2222.CambridgeUniversityPress22H[XHX22 2 (# (# 2Z2AustronesianNasalSubstitutionandother N)C4Ԁeffect*\!s#g[H[#H[g[#HXXH[\#H[XHX I  *       d   JoePater,McGillUniversity  @   November,1996  h   #g[H[a##gXXg[#USCE.,HXXXgXCE.,US.,H[XHX#g[H[#H[g[2 2#HXXH["#2H[XHX2#gXXXHXd#HXXXgX22#HXXH[y#222  22  22 ` 22Currentaddress:222 h 2222 h 222 h 22DepartmentofLinguistics, T 22  22 ` 22 22  22 h 22UniversityofBritishColumbia22  22 ` 22 22  22 h 22C3691866MainMall22  22 ` 22 22  22 h 22Vancouver,BC22  22 ` 22 22  22 h 22CanadaV6T1Z12222  22 ` 222 22222  22 ` 2email:2  2jpater@cortex.psych.ubc.caH[XHX  , 2]2#HXXH[: #  *\!Introduction (  NasalsubstitutionoccursinAustronesianlanguagesasfarflungasChamorro(Topping1969,1973),andMalagasy(Dziwirek1989),aswellasinseveralAfricanlanguages(Rosenthall1989:50).However,itismostfamousforitsappearanceintheIndonesianmH& &XHXB#HXX &H&&#N-prefixationparadigm(seee.g. `  HalleandClements1983:125). I  1      ׀Nasalsubstitutionreferstothereplacementofaroot-initialvoiceless 8  obstruentbyahomorganicnasal(1a).Iftheobstruentisvoiced,ahomorganicclusterresultsinstead(1b).Asillustratedbythedatain(1c),N)C0Ԁ(nasal/voicelessobstruent)clustersarepermittedroot 8  internally:  ,(1)  a. ` /mH& &XHXB#HXX &H&ȼ#N+pilih/   h mH& &XHXB#HXX &H&##milih  'tochoose,tovote'     ` /mH& &XHXB#HXX &H&Ⱦ#N+tulis/   h mH& &XHXB#HXX &H&%#nulis  'towrite'   p    ` /mH& &XHXB#HXX &H&#N+kasih/   h mH& &XHXB#HXX &H&)#asih  'togive' \   b. ` /mH& &XHXB#HXX &H&ȿ#N+bH& &XHXB#HXX &H&#li/   h mH& &XHXB#HXX &H&j#mbH& &XHXB#HXX &H&Ȳ#li  'tobuy' 4    ` /mH& &XHXB#HXX &H&f#N+dapat/   h mH& &XHXB#HXX &H&#ndapat  'toget,toreceive'  p    ` /mH& &XHXB#HXX &H&i#N+ganti/   h mH& &XHXB#HXX &H&#ganti  'tochange'  \ :  c. ` H& &XHXB#HXX &H&ȏ#mpat'four'   h untuk'for'   p mukin'possible' 4 c  Thoughfamiliartomoststudentsofphonology,Austronesiannasalsubstitutionhasnotengenderedmuchtheoreticaldiscussion.Thestandardanalysisinvokestwoorderedrulestogeneratethesinglenasalfromtheunderlyingpairofsegments:nasalassimilation,followedbyaruleofrootinitial,post-nasal,voicelessconsonantdeletion(e.g.Topping1973:49;Onn1980:15;Herbert1986:252; 0&!$ Teoh1988:156;thoughcf.Lapoliwa1981:111,Uhrbach1987:72).  Inthispaper,Ireanalyzenasalsubstitutionasfusionofthenasalandvoicelessobstruent,drivenbyaphoneticallymotivatedconstraintthatdisallowsnasal/voicelessobstruentclusters(*N)C0UKCE.,)CE.,UK.,. ` ThisanalysisiscastintheframeworkofOptimalityTheory,asdevelopedinPrinceandSmolensky(1993),andMcCarthyandPrince(1993a,b,1994a,b,1995,thisvolume).Inparticular,aspectsofCorrespondenceTheory,andthetheoryofmorphologyphonologyinteractionexpoundedinMcCarthyandPrince(1994b,1995,thisvolume),playacentralrole.  NasalsubstitutionisjustoneofarangeofprocessesthatlanguagesmakeuseoftoridthemselvesofN)C0Ԁclusters,whichalsoincludeUKCE.,postnasalvoicing,nasaldeletion,anddenasalizationCE.,UK.,. p PermutationoftheconstraintrankingspositedfornasalsubstitutionisallthatisneededtoprovideaunifiedaccountoftheseN)C0Ԁeffects.Nasalsubstitutionoccurswhentheantifusionconstraint  p Linearityisdominatedby*N)C0ԀandtheotherFaithfulnessconstraints.EachoftheotherN)C0Ԁeffects H issimilarlygeneratedwhentheFaithfulnessconstraintthatitviolatesfallstothebottomofthehierarchy.EspeciallystrongmotivationforaunifiedtreatmentoftheN)C0Ԁeffects#gXXXHX #HXXXgXԀcomesfromthe  existenceoflanguagesinwhichtwooftheprocessesactina'conspiracy'(Kisseberth1970)toeliminateN)C0Ԁclusters.InthispaperIintroduceconspiraciesbetweennasalsubstitutionandeachof X  nasaldeletionandpostnasalvoicing(seePater1996forothers)#gXXXHX#HXXXgX.Sinceneitherthestandard 0"  rule-basedanalysesofnasalsubstitutionorpostnasalvoicing,norIt=,Mester,andPadgett's(1995)recentanalysisofpostnasalvoicing#gXXXHX!#HXXXgXԀextendtothefullrangeoftheseprocesses,theyfailtoyieldan %0!$ accountoftheconspiraciesbetweenthem.  Theanalysisofnasalsubstitution,andtheotherN)C0Ԁeffects,appearsin1through3.Section )$( 1.1introducesthe*N)C0Ԁconstraint.In#gXXXHX"#HXXXgX1.2,IdiscussthesegmentalviolationsofInputOutput h+&* Faithfulnessthatsatisfy*N)C0Ԁ(e.g.fusionanddeletion),andprovideanaccountofthemorphological  restrictionsonIndonesiannasalsubstitution.Section3isconcernedwiththeInputOutputmismatchesinthefeaturalmakeupofN)C0Ԁsequences(e.g.denasalizationandpostnasalvoicing),and ` containsamodificationtotheformulationofMcCarthyandPrince's(1995)FeaturalIdentity,whichisnecessitatedbytheIdentityviolationsincurredbyfusion.Section4focusesontheOshiKwanyama#gXXXHX'$#HXXXgXconspiracybetweennasalsubstitutionandpostnasalvoicing,andonIt=,Mester,andPadgett's  8  (1995)redundantfeaturelicensingapproachto#gXXXHX&#HXXXgXpostnasalvoicing.Theresultsaresummarizedinthe   finalsection,withdirectionsforfurtherresearch.̀ 1* N)C0UKCE.,HXHX#gH'#Ng H #NXrZXN(##gXXXNXrZ)#HXXXgXInawidevarietyoflanguages,CE.,UK.,(ԀN)C0UKCE.,Ԁclustersseemtobedisfavoured.Thatis,InputN)C0Ԁ(nasal/voiced (x obstruent)sequencesarerepresentedfaithfullyintheOutput,whileCE.,UK.,)ԀN)C0UKCE.,saresomehowaltered.The P usualresultisfortheobstruenttobevoiced,thoughthereareotherpossibilities,asenumeratedintheIntroduction,andbelow.  ThefactthattheseCE.,UK.,*N)C0UK.,CE.,Ԁeffects,inparticularpostnasalvoicing,occurwithsuchfrequencyhas  longbeenassumedtostemfromtheeaseofarticulationofN)C0ԀclustersrelativetoCE.,UK., ,ԀN)C0ԀUKCE.,(see `  KenstowiczandKisseberth1979:37,Herbert1986)CE.,UK.,"-UKCE.,,butwithoutaspecifichypothesisaboutthearticulatorydifficultyinherentinN)C0Ԁbeingproposed.However,Huffman's(1993:310)observation $`" thattheraisingofthevelumoccursverygraduallyduringavoicedstopfollowinganasalsegment,withnasalairflowonlyreturningtoavaluetypicalofplainobstruentsduringthereleasephase,suggestsanarticulatorybasisfora*CE.,UK.,-N)C0UK.,CE.,Ԁconstraint,sinceanN)C0Ԁclusterallowsamoreleisurelyraising )$( ofthevelumthananCE.,UK.,/ԀN)C0UKCE.,.Putanotherway,anCE.,UK.,Y0ԀN)C0UKCE.,Ԁclusterrequiresanunnaturallyquickvelarclosure. p+&* Thefactthatthisconstraintisasymmetrical(i.e.*CE.,UK.,0N)C0UKCE.,,andnot*CE.,UK.,|1)C0UK.,CE.,Nseethediscussioninsection  5),canthenbeunderstoodinlightofCE.,UK.,1Zuckermans(1972)findingthat'thevelumcanbeloweredmorequicklyandwithgreaterprecisionthanitcanberaised'(Herbert1986:195). I  2      ׀OhalaandOhala ` (1991:213citedinOhalaandOhala1993:239)providethefollowingcomplementaryperceptuallyorientedexplanationfornasaldeletionintheN)C0Ԁconfiguration:  ` &  (2)8  Amongtheauditorycuesforavoicedstoptheremustbeaspectralandamplitude $  discontinuitywithrespecttoneighbouringsonorants(ifany),lowamplitudevoicingduringitsclosure,andterminationinaburst;theserequirementsarestillmetevenwithvelicleakageduringthefirstpartofthestopaslongasthevelic'lD4valveisclosedjustbeforethereleaseandpressureisallowedtobuildupbehindtheclosure.However,voicelessstopshavelesstoleranceforsuchleakagebecauseanynasalsoundvoicedorvoicelesswouldundercuteithertheirstoportheirvoicelesscharacter.    AdditionalevidenceforthemarkednessofN)C0ԀclusterscomesUKCE.,ԀfromSmiths(1973:53)observation  \ thattheyemergedconsiderablylaterthanN)C0'sinhisson'sspeech,withthenasalconsonantofadult 4 CE.,UK.,X7N)C0UK.,CE.,'sbeingdeletedinthechild'sproduction.Thispatternhasalsobeenobservedinthespeechof   learnersofGreek(DrachmanandMalikoutiDrachman1973)andSpanish(Vogel1976).CE.,UK.,p8Thus,datafromtypology,phonetics,andacquisitionallconvergeontheexistenceofauniversal,butviolable,*N)C0Ԁconstraint: D! (3)  *N)C0 $X"   Nonasal/voicelessobstruentsequences:OneoftheprimarystrengthsofaconstraintbasedtheorylikeOptimalityTheoryisthatphonetically '#& groundedcontextualmarkednessstatementslike*N)C0Ԁcanbedirectlyincorporatedintothephonology  (Mohanan1993:98,PrinceandSmolensky1993:5,ArchangeliandPulleyblank1995;seeFlemming1995,Hayes1995,Jun1995,Kirchner1995andSteriade1995bforextensivedevelopmentofthissortofapproachwithinOptimalityTheory).Inwhatfollows,Idemonstratehowtheinteractionbetween*N)C0ԀandconstraintsonInput-OutputCorrespondencecreatesgrammarsthatgeneratenasal  ` substitution,aswellastheotherN)C0Ԁeffects.UKCE., I  3      CE.,UK.,=UKCE.,  8   2*CE.,UK.,1>N)C,UK.,CE.,ԀandSegmentalCorrespondence    2.1SegmentalFusion p Ratherthanpositingdiscretestepsofnasalassimilationandvoicelessconsonantdeletion,orof H completeassimilationofthevoicelessconsonanttothenasalanddegemination(Uhrbach1987:72;cf.Herbert1986:252)CE.,UK.,>UK.,CE.,#gXXXHXz)#HXXXgX,IassumethattherelationshipbetweenInputmH& &XHXB#HXX &H&ȟ@#N+pilihandOuputmH& &XHXB#HXX &H&@#milih H ismediatedbyfusion,orcoalescenceofsegments(Lapoliwa1981:111).Partofthemotivationforthisassumptionisspecifictothemodelofphonologybeingassumedhere-afusionalanalysisallowsnasalsubstitiontobetreatedasaonestepInputOutputmapping,withouttheintermediatederivationalstagethatassimilation+deletionrequires.Thereare,however,tworelativelytheoryneutralargumentsforfusion:oneisfromtypology,theotherisinternaltothephonologyofIndonesian. $X" &    'iCInarguingforfusionbasedanalysesofotherprocesses,CE.,UK.,@UK.,CE.,#gXXXHXM@#HXXXgXԀStahlke(1976)makesthepointthat  anorderedruleaccountpredictsthateachoftherulesshouldbeindependentlyobserved.Whileplaceassimilationofnasalsisofcourseextremelycommon,postnasalvoicelessconsonantdeletionseemsnevertoapplywithoutthepriorassimilationofthenasal.Aswewillseebelow,thereareexamplesofotherCE.,UK.,CԀN)C0UK.,CE.,Ԁeffectsapplyingwithoutplaceassimilation,suchasZoquepostnasalvoicing(Wonderly  ` 1951,KenstowiczandKisseberth1979:36,Padgett1994),anddenasalizationinbothTobaBatak(Hayes1986)andKaingang(Henry1948,Piggott1995).Byusingfusionratherthanorderedrules,weavoidthe'falsestep'ofvoicelessconsonantdeletion.CE.,UK.,E#gXXXHXD#HXXXgXUK.,CE.,     ThereisalsoevidencefromwithinthephonologyofIndonesianforthefusionalanalysis.Lapoliwa(1981:110)notesthatreduplicationcopiesasubstitutednasal(4a),whileprefixalnasalsprecedingavoicedobstruent(4b),oravowel(4c),failtobecopied:(4)  a./mH& &XHXB#HXX &H& I#N+kata+RED+i/ h mH& &XHXB#HXX &H&ȀI#ataatai   p 'tospeakillaboutsomeone'  4   b./mH& &XHXB#HXX &H&;J#N+gerak+RED/ h mH& &XHXB#HXX &H&ȚJ#gerakgerak p 'tomovesomethingrepeatedly     c./mH& &XHXB#HXX &H&=K#N+H& &XHXB#HXX &H&ȅK#lu+RED+kan/ h mH& &XHXB#HXX &H&K#H& &XHXB#HXX &H&-L#luH& &XHXB#HXX &H&uL#lukan   p 'topraise'   HLapoliwaformulatestheruleofnasalsubstitutionasoneofphonologicalandmorphologicalcoalescence,sothatthesubstitutednasalin(4a)becomespartofthemorphologicalstem,unliketheunassimilatednasalin(4c).BuildingonworkbyUhrbach(1987),CohnandMcCarthy(1994)proposeanentirelyprosodicapproachtothesefacts,inwhichtheprefixfinalnasalin(4a)becomesinitialtotheprosodicword,whiletheonein(4c)endsupincodapositionoutsideoftheprosodicword.ThedifferingprosodicpositionoftheseconsonantsisduetoanAlignWordconstraint,which '#& forcescoincidenceoftheedgesoftherootandprosodicword.Iftherootinitialconsonantsimplydeleted,thisanalysiswouldbedifficult,ifnotimpossibletomaintain. h+&* Ѐ  ToformalizethefusionalInputOutputmapping,IdrawonMcCarthyandPrince's(1994b,  1995,thisvolume)proposalthattherelationshipbetweenInputandOutputisdirectlyassessedbyconstraintsonCorrespondence.ThisapproachcontrastswiththeindirectmethodofusingpurelyOutputbasedconstraints,andstipulatingthatthephonologicalandmorphologicalpropertiesoftheInputmustbecontainedintheOutput,bytheprinciplesofContainmentandConsistencyofExponence(PrinceandSmolensky1993,McCarthyandPrince1993a&b).IntheContainmentapproachtoInputOutputFaithfulness,theconstraintParseSegmentforcestherealizationof   underlyingsegments(unpronouncedInputsegmentsarepresentintheOutput,butunparsed).TheequivalentinCorrespondencetermsisaMaxCE.,UK.,GUK.,CE.,#gXXXHXG#HXXXgXԀconstraintthatdemandsthateverysegmentintheInput p maptoasegmentintheOutput,inotherwords,thateveryInputsegmenthaveanOutputcorrespondent.ThereplacementofParseSegmentwithMaxallowsaninterpretationoffusionas  p atwotoonemappingfromInputtoOutput:twoInputsegmentsstandincorrespondencewithasingleOutputsegment(McCarthyandPrince1995;seealsoGnanadesikan1995andLamontagneandRice1995).ThisresultsinthesatisfactionofMax,thoughunderastrictinterpretationof  Containment,ParseSegmentwouldbeviolatedinthissituation(McCarthyandPrince1993a:163,  Myers1994,Russell1995).IillustratethedifferencebetweenInputandOutputin(5),wheresubscriptingisusedtoindicatethecrucialcorrespondencerelationship:(5)  mH& &XHXB#HXX &H&ȋW#N1+p2ilih(Input)  mH& &XHXB#HXX &H&W#m1,2ilih(Output)   %0!$   Eventhoughfusiondoesnotinvolvedeletion,andsosatisfiesMax,itdoesincurviolations )$( ofotherconstraints.Atthefeaturallevel,fusionbetweennonidenticalsegmentsviolatesconstraints h+&* demandingIdentitybetweenInputandOutputsegments(seesection3belowforelaborationofIdentityconstraints,andforanexampleinwhichCE.,UK.,SN)C0UK.,CE.,ԀfusionisoverruledbyaFeaturalIdentity  constraint).BecausefusionincursviolationsofFeaturalIdentity,ittendstooccurbetweensegmentsthatareidentical,ornearlyso(cf.McCarthyandPrince1993a:163,wherefusionisrestrictedtoidenticalelements).However,evenfusionbetweenidenticalsegmentsisnotautomaticoruniversal,soitmustviolateatleastoneconstraintotherthanFeaturalIdentity.OnesuchconstraintisLinearity,whichisindependentlyneededinCorrespondenceTheorytomilitateagainstmetathesis.2 222 I  4         McCarthyandPrinces(1995)formulationofLinearityisasin(6),whereS1andS2refertoInput   andOutputstrings(oranyotherstringofcorrespondentsegments,suchasBaseandReduplicant):&  (6)  Linearity  p   S1reflectstheprecedencestructureofS2,andviceversa.    \ '^^InthefusionalI,Orelationshipdepictedin(5),/N/precedes/p/intheInput,butnotintheOutput,soLinearityisviolated. I  5      ׀TocommandaviolationofLinearity,*CE.,UK.,@ZN)C0UK.,CE.,Ԁmustberankedabovethe  Faithfulnessconstraint,asillustratedinthetableauin(7).Acheckmarkindicatesagrammaticalform,andexclamationmarksshowwhereothercandidatesfail.Solidlinesbetweenconstraintsareusedwhentheconstraintsareranked,anddashedlineswhenthereisnoevidencefortheirranking.Unlessnotedotherwise,allofthefollowingtableauxapplytoIndonesian. $X" (7)&  Ԁ Nasalsubstitution:*CE.,UK.,6`N)C,UK.,CE.,>>Lin   *q  ddd Xdd Xdd X(#(#q,Sc ,c ,c +  ,UUUU T bb,Input:mH& &XHXB#HXX &H&c#N1+p2ilih 5UU$fUUUU 5CE.,UK.,bN)C0UK.,CE., @UVUU-f"UU @Lin OUU<*f" bb  UVUU BB O  a.mH& &XHXB#HXX &H&Ie#m1,2ilih3 7-F UU 7 @VU-F" @HbXHX*#HXXHbf# LVUUU9*F " BB  VU BBL  b.mH& &XHXB#HXX &H&ȸf#m1p2ilih >VU- * VUUU >HbXHX*!#HXXHbPg# @VVUU- * "VU @:0. * " BB   VVUU :' ,bWiththerankingreversed,thecandidatewithoutsubstitution(7b)wouldbeoptimal.SucharankingcharacterizeslanguagesthattolerateCE.,UK.,dN)C0UK.,CE.,Ԁclusters. ,|  2.2Morphologicalconditionsonfusion , ThefactthatfusionviolatesLinearityleadstoastraightforwardaccountofthelackofroot-internal  nasalsubstitutioninIndonesian.McCarthyandPrince(1994b),andUrbanczyk(1996)showthatalargenumberofdisparatephonologicalphenomena,reduplicativeandotherwise,resultfromstricterFaithfulnessrequirementswithintherootthanelsewhereintheword,thatis,fromtherelativemarkednessofroots.Thegreatermarkednessofrootsisnodoubtdrivenbytheneedtomaintainmorecontrastsbetweenrootsthanbetweenaffixes.McCarthyandPrinceformalizethisdifferenceinmarkednessbyproposingageneralrankingschemainwhichrootspecificversionsofFaithfulnessconstraintsareintrinsicallyrankedhigherthanthegeneral,oraffix-specificversionoftheconstraints.Ifnasalsubstitutionweretoapplywithintheroot,massiveneutralizationwouldresult.Aroot-specificrankingofLinearity(RootLin)above*CE.,UK.,hN)C0UK.,CE.,Ԁstopsthisfromhappening.Atableau L&!( illustratingtheblockingofsubstitutionwithintherootappearsin(8): Ӏ )L%, n&   (8) Rootinternal CE.,UK.,tm N)C,UK.,CE.,Ԁtolerance:RootLin>>*CE.,UK.,nN)C, UK.,CE.,  *q ddSc c c (#(#q,c ,8Sc ,c ,9c +  ,UUUU  bb,Input:H& &XHXB#HXX &H&Ȁp#m1p2at 4UUV!RUUUU 4 -RootLin >UU-R"UUV >*CE.,UK.,;oN)C0UK.,CE., @UVUU-R"UU @Lin OUU<*R" bb  UVUU BB Oa.H& &XHXB#HXX &H&Lr#m1,2at 7V$ UU  7HbXHX*!#HXXHbr# 7-" V 7 @VU-" @HbXHX*#HXXHbs# LVUUU9* " BB  VU BBL  b.H& &XHXB#HXX &H&/t#m1p2at3 =VUV*v  VUUU =ln >VU-v  "VUV >HbXHX*#HXXHbu# @VVUU-v  "VU @:0.v  " BB   VVUU :' VnRootLinrulesoutfusionwithintherootbecausefusiondestroystheprecedencerelationship  @ betweenInputrootsegments/m/and/p/(8a).Sincethenasalin/meN+pilih/isnotpartoftheroot,   nasalsubstitutionacrossthemorphemeboundarydoesnotdisturbtheprecedencestructureofrootelements,andRootLinisobeyed.CE.,UK.,q I  6      ׀UK.,CE.,ԀCE.,UK.,wUSCE.,HHX#gHS# x  Hg#HXXH#x#CE.,US.,xԀ  RootLiniseffectiveinblockingsubstitutionwithintherootbecauseitisaconstraintonthe P relationshipbetweenInputandOutputstrings,ratherthanbetweenindividualInputandOutputsegments,orfeatures.IfweattemptedtoruleoutrootinternalfusionwitharootspecificconstraintonIdentitybetweenInputandOutputcorrespondents,substitutioninthemiddleoftheroot,andatthebeginningofitwouldbeassessedequally,sincebothwouldturnavoicelessobstruentbelongingtotherootintoanOutputnasal.AsDoncaSteriade(p.c.)haspointedout,itisnotatallclearhowatheorywithFaithfulnessconstraintsdemandingonlyfaithfulsegmentalandfeaturalparsingwouldhandletheseandothersegmental'derivedenvironment'effects(seeKiparsky1993forrecentdiscussion).ThemaindifferencebetweenIndonesiannasalsubstitution,andmorecommonlydiscussedcasessuchastheSanskritRukiruleandFinnishassibilation,isthatthelatterinvolvesegmentalchange,ratherthansegmentalfusion.However,iflinearityisgeneralizedtosubsegmental &"+ elements,suchthatitforcestheirunderlyingprecedencerelationshiptobemaintained,andifthesecasescanallbeanalyzedasinvolvingpartialsegmentaloverlap,thenrootspecificrankingsofsubsegmentallinearitywouldgeneratenonderivedenvironmentblockingeffects. I  7      ׀Clearly,agreatdeal ` ofworkneedstobedonetodeterminetheempiricalcoverageofroot-specificLinearity 8  constraints,butitseemsplausiblethattherankingofmorphemespecificFaithfulnessconstraintsabovephonotacticconstraintsisthesourceofthissortofphenomenon.UK.,CE.,  8  2.3SegmentalDeletionandInsertion   SofarwehaveonlyconsideredcandidateswithandwithoutCE.,UK.,N)C0UK.,CE.,Ԁfusion.Deletion,andepenthesis p couldalsosatisfy*CE.,UK.,N)C0UK.,CE.,,withoutincurringviolationsofLinearity.ThismeansthatinIndonesian, H theconstraintsMax,andDep,whichareviolatedbydeletionandepenthesisrespectively(McCarthy  p andPrince1995),mustberankedaboveLinearity.Infact,theseconstraintsmustbeplacedeven H higherinthehierarchy,above*CE.,UK.,xN)C0UK.,CE.,,sinceneitherdeletionnorepenthesisisusedtoresolve*CE.,UK.,GN)C0UK.,CE.,   violationsrootinternally,wherefusionisruledoutbyRootLin:  & F (9) DeletionandepenthesisblockedbyMax,Dep>>* CE.,UK.,΃N)C0UK.,CE., l * ddc 8Sc c 9c "",dc ,dcc ,dSc ,dc +  ,UUUU  $ bb,Input:H& &XHXB#HXX &H&V#mpat 7UU$6! UUUU 7Max >UU-6!!"UU >Dep @UVUU-6!""UU @*CE.,UK.,N)C0UK.,CE., OUU<*6!#" bb  UVUU BB O  a.H& &XHXB#HXX &H& #mpat3 @-"$ UU  @ 7-"%"  7 @VU-"&" @HbXHX*#HXXHb# OVUU<*"'" BB  VU BBO  #gXXXHXzx#b.g& &XgXB#gXX &g&׉#pHXXXgXat @V-$(VUU  @HbXHX*!#HXXHb# :V-$)" V  : @VVU-$*" V @ LVUUU9*$+" BB  VVU BbL  iU3 ; CKS!#&[(* -c/14k68;s=?Xic.H& &XHXB#HXX &H&#mH& &XHXB#HXX &H&I#rUX` hp x (#%'0*,.8135@8:<X ; Xrpat @VU-&!,VUUU @ >VU-&!-"VU >HbXHX*!#HXXHb# @VVUU-&!."VU @:0.&!/" Bb   VVUU :'F X"z P(#/ IfMax,orDepwererankedbeneath*CE.,UK.,N)C0UK.,CE.,,deletion(9b),orepenthesis(9c)wouldbewrongly  preferredovertheoptimalcandidate(9a).  ThoughneitherdeletionnorepenthesisisresortedtoinIndonesiantoavoid*CE.,UK.,N)C0UK.,CE.,Ԁviolations, ` permutationoftherankingsoftheseconstraints(PrinceandSmolensky1993:6)predictstheexistenceofotherlanguagesinwhichMaxandDeparedominatedby*CE.,UK.,N)C0ԀUK.,CE.,andtheotherFaithfulness  ` constraints,producingCE.,UK.,N)C0UK.,CE.,ԀdeletionandCE.,UK.,N)C0UK.,CE.,Ԁepenthesis.  8    ExamplesofsegmentaldeletionintheCE.,UK.,ԀN)C0UK.,CE.,Ԁconfigurationincludetheaforementionedcases   ofchildEnglish(Smith1973:53),childGreek(DrachmanandMalikoutiDrachman1973),andchildSpanish(Vogel1976).AmongsttheadultlanguageswithCE.,UK.,N)C0UK.,CE.,ԀdeletionistheKelantandialectof p Malay,whichdiffersfromstandardJohoreMalayinthatitlacksnasalsbeforevoicelessobstruents,thoughitpermitshomorganicN)C0Ԁclusters(Teoh1988).ThispatternisreplicatedinAfrican  p languagessuchasVenda(Ziervogel,Wetzel,andMakuya1972:citedinRosenthall1989:47),Swahili I  8      ׀andMaore(NurseandHinnebusch1993:168),aswellasseveralCE.,UK.,ϓotherscitedbyOhalaand   Ohala(1993:239).UK.,CE., I  9      CE.,UK.,^  UK.,CE.,  Whatunitesalloftheseexamplesisthatthenasal,ratherthantheobstruentisdeleted.Thisparallelsthenasal/fricativeclustereffectsdetailedinPadgett(1994),whichsometimesinvolvenasal,butneverfricative,deletion.Theconstraintspositedthusfarassessobstruentandnasaldeletionequally,asviolationsofMax.Howtoformalizenasalobstruentasymmetriesindeletion,aswellas $X" inassimilation,remainsunaddressedinOptimalityTheory(andmoregenerally,inphonology:seeMohanan1993).OnepossibilityistointroduceintrinsicrankingsoftheFaithfulnessconstraints.Forexample,thefactthatnasalstendtoassimilateinplacetoobstruents,ratherthantheotherwayaround,couldbecapturedbyafixedrankingofObsPlaceIdent>>NasPlaceIdent(i.e.the 8  identityrequirementbetweenanobstruentanditsunderlyingcorrespondentisintrinsicallyhigherrankedthanthatbetweenanasalanditscorrespondent;seeJun1995fordevelopmentofthistypeofapproach).Fordeletion,arankingofanobstruentspecificMaxconstraint(ObsMax)abovethe   nasalspecificNasMaxachievesthedesiredresult.Establishingthephoneticbasis,andtypological   correctnessofthispresumedfixedrankingisbeyondthepurviewofthisstudy,butitcanbenotedthatitsuniveralityissupportedbytheobservationthatafewlanguageslacknasals,butnonearewithoutoralsegments(Maddieson1984,citedinMcCarthyandPrince1994a,whoprovideadifferentexplanationforthisgeneralization). H Ѐ  Thetableauin(10)demonstrateshowan/CE.,UK.,NT/UK.,CE.,Ԁclusterwouldbetreatedinalanguagesuch   asKelantanMalay,inwhich*CE.,UK.,N)C0ԀUK.,CE.,dominatesMax(notethatallotherFaithfulnessconstraints,  includingLinearity,arealsorankedaboveMax):    X  & l  (10) TableauforKelantanlikelanguages   * dddc dcc dSc dc "",c ,Fcc ,Sc ,'c +  /UUUU h !bb/    Input:CE.,UK.,N1TUK.,CE.,2 .UUzUUUU .%*CE.,UK.,[N)C0UK.,CE., >UU-z UU >ObsMax @UVUU-z"UU @NasMax OUU<*z" bb ! UVUU BBO  CE.,UK.,N1TUK.,CE.,2 7$ ZUU 7  HbHX*!#HXXHby# 7- Z  7 @VU- Z " @ OVUVU<* Z " BB  VU BBO  N1 7VV$ > VUVU 7   >VV- >  VV >HbHX*!#HXXHb# @VVVU- > "VV @ OUUU<* >" BB  VVVU BbO  T2󀀀3 7U$ "UUU 7   >U- " U > @VUU- ""U @HbHX*#HXXHbզ#:0. "" Bb   VUU :Infuturetableaux,IwillmergethetwoMaxconstraints,andshowonlythecandidatewiththe l  deletednasal.'l  Forsomereason,languagesseemnottomakeuseofepenthesistoresolve*CE.,UK.,N)C0UK.,CE.,Ԁviolations. l OnemightstipulatethatDepuniversallydominates*CE.,UK.,N)C0UK.,CE.,,butwithoutanyindependentmotivation D forthisfixedranking,suchaformalizationwouldremainintherealmofdescription,ratherthanexplanation.Q I  10      ׀WiththispotentialgapinthetypologyofCE.,UK.,5N)C0UK.,CE.,Ԁeffectsdulynoted,Iwillnowturntothe  featuralchangesthatcanbeusedtosatisfy*CE.,UK.,N)C0UK.,CE.,,andproposeconstraintstorulethemoutin |  Indonesian.Intheseinstances,wewillseethepredictedfactorialtypologyisindeedfulfilled.̀   ,|$  &  3*CE.,UK.,eN)C,UK.,CE.,ԀandFeaturalFaithfulness   T& 3.1Denasalization' $D !,( Insteadofcompletelydeletingthenasal,anotherwaytomeetthe*CE.,UK.,N)C0UK.,CE.,Ԁrequirementistochangethe #*  underlyingnasalintoanobstruent.Thereareatleastthreelanguagesthattakethisroute:Toba % , BatakV I  11      ׀(Hayes1986),Kaingang(Henry1948;cf.Piggott1995),andMandar(Mills1975).Mandar,   alanguagespokeninSouthSulawesi,isparticularlyinterestingbecauseithasaprefixationparadigmthatdiffersminimallyfromthatofIndonesian.Ahomorganicnasalappearsbeforevoicedobstruents(11a),butinsteadofnasalsubstitutionwiththevoicelessones,thereisgemination(11b)(inTobaBatakandKaingang,theresultingobstruentretainsitsplacespecification,andcanbeheterorganicwiththefollowingconsonant).(11) MandarmaNprefixation      a. ` /maN+dundu/   h mandundu  'todrink'     b. ` /maN+tunu/   h mattunu  'toburn'   հInMandar,unlikeIndonesian,theprohibitionagainstCE.,UK.,N)C0UK.,CE.,Ԁextendsthroughoutthelanguage: L (12)8  & D NowhereinmymaterialnorinPelenkahu'sextensivelistsofminimalpairsistherea ` singleinstanceofnasalplusvoicelessstop.Z I  12      ׀Wheresuchaclusterwouldbe L expected,becauseofcognateitemsoratcertainmorphemeboundaries,thereisinvariablyageminatevoicelessstop.Inthisrespect,[Mandar]isfarmoreconsistentthan[Buginese];perhapsitreflectsgreaterfreedomfromoutsideinfluence(Mills1975:82).'D    Therearenumberofpotentialconstraints,orsetsofconstraintsthatcouldruleoutdenasalizationinIndonesian,aswellasinlanguageslikeKelantanMalaythathavenasaldeletion.Beforeturningtothem,ashortdiscussionoffeaturalFaithfulnesswithinCorrespondencetheoryisinorder.  #p!   ToreplacethecontainmentbasedParseFeature(seee.g.It=,Mester,andPadgett1995)  inCorrespondenceTheory,McCarthyandPrince(1994a,1995)outlinetwoapproaches.OneistoextendCorrespondenceintothefeaturaldomain,andrequiremappingsbetweeninstancesoffeaturessuchas[voice]intheInputandOutput.Alesselaboratetheory,andtheonethatMcCarthyandPrinceadopt,invokesasetofidentityrequirementsbetweensegmentalcorrespondents.Ageneralformulationforsuchconstraintsisgivenin(13):&  (13)  FeaturalIdentityIdent(F)     CorrespondentsareidenticalintheirspecificationforF   'ݸ˸Formulatedinthisway,featuralFaithfulnessisnotviolatedifasegmentisdeleted,sinceifanInputsegmenthasnoOutputcorrespondent,Identityconstraintsdonotcomeintoforce.Ontheotherhand,iftherewereawholesetofCorrespondenceconstraintsthatexaminedfeatures,theneverytimeanunderlyingsegmentfailedtoberealizedintheOutput,alloftheapplicableFeaturalCorrespondenceconstraintswouldbeviolated.ThiswouldforcealloftheFeaturalCorrespondenceconstraintstobedominatedbywhateverconstraintfavoureddeletion.Whetherthisisafatalflaw,orahappyresult,$ I  13      ׀canonlybeassessedthroughcarefulstudyoftherelationshipbetweensegmental l deletionandfeaturechangingprocesses,butitisevidentthatFeaturalIdentityhastheadvantageofanalyticconvenience,especiallywhenconsideringreduplication,whichofteninvolveslongstringsofCorrespondenceviolations.& I  14       $D #   Incasesoffusion,however,thesimplestatementofFeaturalIdentitygivenin(13)doesleadtosomecomplications.ConsidertheInputOutputmappingsin(14): I  15      #gXXXHX#HXXXgXCE.,UK.,~UK.,CE.,  &  (14)#gXXXHXJ#X!XXgX  Inputa.nt   h b.nt   8     `  \/   h |  | , |   Output n   h tt    P #gXX!XX#HXXXgX'<Nasalsubstitutionisrepresentedin(14a),anddenasalizationin(14b).OneconsequenceofthesymmetricalnatureofIdentityisthatIdent[Nas]isviolatedtothesamedegreein(14a)and(14b),   sinceinbothinstancesanasalandavoicelessobstruentstandincorrespondencewithoneanother.NasalsubstitutionalsoviolatesLinearity,sointermsoftheconstraintsconsideredthusfar,itis 4 impossibleforalanguagetoprefer(14a)over(14b),sincetheFaithfulnessviolationsincurredby(14b)areasubsetofthosefor(14a).  Onemightconsiderrulingout(14b)withconstraintsagainstcodaobstruents,and/orgemination.Byusingasyllablestructureconstrainttoruleoutdenasalization,however,theresultingpredictionshouldbethatlanguagesthatdisplaynasalsubstitutionhavetightrestrictionsonpossiblecodas.Tosomeextent,thisisborneout.However,Chamorro,whichhasnasalsubstitutioninman D andfanprefixation,alsohasgeminatesandcodaobstruents(Topping1973:3649),evenin !l prefixes,suchashat,chat,andtak(Topping1973:66).Thus,nasalsubstitutiondoesnotappear "D! tobedrivenbyadesiretoavoidcodaobstruents,orgemination. $ #   AnotherresponsetothisproblemistoelaborateIdentitysomewhat,sothatwehaveawayofstatingthatinnasalsubstitutionanInputnasalmapstoanOutputone,whileindenasalizationanInputnasalmapstoanobstruent.WiththisshiftawayfromsymmetrythetheoryoffeaturalFaithfulnessbeginstolookmorelikesegmentalCorrespondence,whichhasseparateMaxandDep 8  constraints.However,IwillpreservetheanalyticadvantageofIdentitynotedabovebystatingtheconstraintinsuchawaythatfeaturalFaithfulnessisnotviolatedincasesofdeletion:(15)  IdentIO[F]   Ѐ  AnycorrespondentofanInputsegmentspecifiedasFmustbeF   YNasalsubstitutiondoesnotviolateIdentIO[Nas],whiledenasalizationdoes.CE.,UK.,[Nas]herewould \ refertothefeature[Nasal]inmonovalentfeaturetheory,or[+Nasal]ifbivalentfeatureswereassumed.UK.,CE.,ԀThechoiceisnotcrucial,butsincethefeature[Nasal]seemsnottobeactiveinanyphonologicalprocess,Iwillassumethereisbutasinglemonovalentfeature[Nasal](Piggott1993,Rice1993,Steriade1993,Trigo1993,cf.Cohn1993).#gXXXHXT#CE.,UK.,$USCE.,Notethatifbivalentfeatureswereused,and   FeaturalIdentitywerestatedwithoutanyreferencetothevalueofthefeature(i.e.'anycorrespondentofInputsegmentXmustbeidenticaltoXinitsspecificationforF'),thentheeffectsofthisconstraintwouldremainsymmetrical,andtheproblemofdifferentiatingIOandOIIdentity D! wouldremain.HXXXgXCE.,US.,jUK.,CE.,Ԁ #l!   ForalanguagelikeMandar,IdentIO[Nas]isrankedbeneath*CE.,UK.,FN)C0UK.,CE.,Ԁandtherestofthe $D # Faithfulnessconstraints.InIndonesian,IdentIO[Nas]isrankedaboveLinearity,sothatfusion &"% ispreferredoverdenasalization.AtableauforMandarisgivenin(16):  |*%) &   (16) Mandardenasalization:*CE.,UK.,N)C,UK.,CE., >> IdentIO[Nas]  p   *)* ddc Fcc Sc 'c "",c ,dcc ,gSc ,gSc ,Sc ,fc +  ,UUUU  bb,Input:maN1+t2unu .UU>UUUU .  Dep =UU*R UU =Max =UU*R"UU =Linearity >UU-R"UU >*CE.,UK.,N)C0UK.,CE., @UVUU-R"UU @IdentIO R [Nas] OUU<*> " bb  UVUU mBB O  a.man1,2unu 7$~  UU 7  @-~     @ @-~ "   @*! 7-~ "  7 @VU-~" @ LVUVU9*~" BB m VU mBBL  b.man1t2unu 7VV$ >VUVU 7  =VV* > VV = =VV* >"VV = >VV- >"VV >*! @VVVU- >"VV @ OUU<* >" BB m VVVU mBB Oc.mat1t2unu3 .^  UU . 7$^    7 @-^ "   @ 7-^ "  7 @VU-^ " @* LVUVU9*^ " BB m VU mBBL  d.mat2unu 7VV$ VUVU 7  =VV*  VV =*! =VV* "VV = >VV- "VV > @VVVU-  "VV @ OUUU<* !" BB m VVVU BbO  e.ma1at2unu 7U$> "UUU 7  *! @U-> # U @ @U-> $"U @ >U-> %"U > @VUU-> &"U @:0.> '" Bb   VUU :' 7#gXXXHX #SomefurthermotivationfortherecognitionofseparateHXXXgXIdentIO[Nas]andIdentOI[Nas] d ) constraintscomesfromthefactthatthereisatleastonelanguageinwhichageminatenasalis <+ createdtoavoida*CE.,UK.,N)C0UK.,CE.,Ԁviolation(theSouthSulawesilanguageKonjoFribergandFriberg1991: d- 88).TodistinguishKonjofromitsnearneighbourMandar,IdentOI[Nas]canberankedbeneath </ IdentIO[Nas],sothathavinganOutputnasalincorrespondencewithanInputobstruent(i.e.NT 1 NN)isabetterresolutionof*CE.,UK.,QN)C0UK.,CE.,ԀthanhavinganInputnasalincorrespondencewithanOutput 3 obstruent(i.e.NTTT).InMandar,ofcourse,therankingbetweentheseconstraintswouldbe t5 reversed. I  16       L7 3.2Postnasalvoicing !L; Themostcommon,andmostwidelydiscussedCE.,UK.,N)C0UK.,CE.,Ԁeffectispostnasalvoicing.Aparticularly #$= relevant,andperhapslessfamiliarexampleisthatofthePuyoPungodialectofQuechua(Orr1962, % ? Rice1993).Asshownin(17),postnasalvoicingonlyaffectsaffixalconsonants.Rootinternally,postnasalconsonantscanremainvoiceless.Ӏ& J (17)  PuyoPungoQuechua  L    a. Rootinternal CE.,UK., N)C,UK.,CE.,:      #gXXXHX.#ikiHXXXgX ` 'soot'   cuntina'tostirthefire' p pampalyina'skirt'  l  Ѐ  b. Suffixalalternations:  h  ~   r  sinikpa 'porcupine's' h   kamba p 'yours' 6    sacapi 'inthejungle' h   hatumbi p 'thebigone' "    wasita 'thehouse' h   wakinda p 'theothers'   x    'J L wObviously,postnasalvoicingsatisfies*CE.,UK.,-N)C0UK.,CE.,.Again,thequestionofwhatitviolatesisnotas  straightforwardasitmightatfirstseem.ComparetheI,Ocorrespondencesfornasalsubstitutionandpostnasalvoicing:(18)#gXXXHX#X!XXgX  Inputa.nt   h b.nt   F    `  \/   h |  | :   Output n   h nd  #gXX!XX#HXXXgXIfweassumefullspecificationofthetraditionalsetoffeatures(i.e.thoseofChomskyandHalle1968),Ident[Voice]istheonlyconstraintviolatedin(18b),yetitisalsoviolatedin(18a)since B   Input/t/correspondstoOutput/n/.NasalsubstitutionviolatesLinearity,whilepostnasalvoicing "j" doesnot,soagain,thereissomedifficultyinestablishinghowIndonesiancouldprefer(18a)over(18b).  Inthiscase,itispointlesstoconsiderconstraintsthatwouldruleouttheN)C0Ԁconfiguration '"( itself,sincethisdoesoccurinIndonesianastheOutputofanunderlyingN)C0Ԁsequence.Nordoesthe z)$* problemlieinthesymmetryofIdentity,sinceinbothcasesavoicelessInputsegmentstandsin R+&, correspondencewithavoicedOutputsegment.Rather,itisduetothemistakenassumptionthat[voice]onasonorant,andonanobstruent,areequivalent(seeChomskyandHalle1968:300,Lombardi1991,RiceandAvery1989,Piggott1992,Rice1993,andSteriade1995afordiscussionfromavarietyofperspectives).Becausetheexactmethodadoptedforcapturingthenonequivalencyofsonorantandobstruent[voice]isofnoparticularconsequenceinthepresentcontext,IwillsimplyinvokeanIdentityconstraintthatspecificallytargetsobstruent[voice].ThereisnoneedtospecifytheconstraintasapplyingfromItoOorO-to-I:(19)  ObstruentVoiceIdentityIdent[ObsVce] p   Correspondentobstruentsareidenticalintheirspecificationfor[voice] H Asitappliesonlytoobstruentsincorrespondence,thisconstraintisnotviolatedbynasalsubstitution,inwhichanobstruentisincorrespondencewithanasal.ForIndonesian,wecanthusblockpostnasalvoicingbyrankingIdent[ObsVce]above*CE.,UK.,N)C0UK.,CE.,.InPuyoPungoQuechua,aroot  specificversionofIdent[ObsVce]ranksabove*CE.,UK.,KN)C0UK.,CE.,,andthegeneralIdent[ObsVce]ranksbelow  it,thusproducingaffixalpostnasalvoicingonly.  Asthiscompletestheanalysisofnasalsubstitution,itisappropriatetoprovideanillustrativetableau:  %0!$ &   (20) Finaltableaufornasalsubstitution     O*,- ddc dcc gSc gSc Sc fc )*"",rc ,cc ,9Sc ,Sc ,Sc ,Sc ,cc ,Xc +  ,UUUU h bb,Input:/mH& &XHXB#HXX &H&ȃ#N1+p2ilih/ .UUfUUUU .  Dep =UU*z UU =IdentIO f [Nas] =UU*R"UU =Max @UU-z "UU @Root z  Lin >UU-f "UU >#gXXHX/#CE.,UK.,Ident z  [ObsVce]UK.,CE.,HXXgX >UVU-f "UU >*CE.,UK.,N)C0UK.,CE., @UVUU-z"UVU @Lin OUU<*z" bb  UVUU BB O  a.mH& &XHXB#HXX &H&7#m1,2ilih3 7$ 2 UU 7  @- 2    @ @- 2"   @ @- 2"  @ 7- 2" 7 <V- 2" < @VU- 2"V @HbXHX*#HXXHbG# LVUVU9* 2" BB  VU BBL  b.mH& &XHXB#HXX &H&#m1p2ilih 7VV$ VUVU 7  =VV*  VV = =VV* "VV = @VV- "VV @ >VV- "VV > >VVV- "VV >HbXHX*!#HXXHb# @VVVU- "VVV @ OUU<* " BB  VVVU BB Oc.mH& &XHXB#HXX &H&ȑ#p1p2ilih .   UU . 7$ !   7HbXHX*!#HXXHbP# @- ""   @ @- #"  @ 7- $" 7 <V- %" < @VU- &"V @ LVUVU9* '" BB  VU BBLd.mH& &XHXB#HXX &H&#m1b2ilih .VV (VUVU . 4VV! )VV 4 =VV* *"VV = @VV- +"VV @ >VV- ,"VV >HbXHX*!#HXXHb# >VVV- -"VV > @VVVU- ."VVV @ OUU<* /" BB  VVVU BB O  e.mH& &XHXB#HXX &H&ȱ#p2ilih 7$r 0 UU 7  @-r 1    @ @-r 2"   @HbXHX*!#HXXHb# @-r 3"  @ 7-r 4" 7 <V-r 5" < @VU-r 6"V @ LVUUU9*r 7" BB  VU BbL  f.mH& &XHXB#HXX &H&Z#1H& &hHXB#HXX &H&Ȩ#p2ilih 7VU$V8VUUU 7  HbXHX*!#HXXHb@# =VU*V9 VU = =VU*V:"VU = @VU-V;"VU @ >VU-V<"VU > >VVU-V="VU > @VVUU-V>"VVU @:0.V?" Bb   VVUU :'9Noteworthyinthistableauisthefactthatallofthenonoptimalcandidates,withtheexceptionoftheepenthetic(20f),doturnupasoptimalinotherlanguages,andthateachofthesecasescanbegeneratedsimplybyhavingoneoftheconstraintsfallbeneathalltheothers.Candidate(20b)isgeneratedif*CE.,UK.,yN)C0UK.,CE.,ԀranksbeneaththeFaithfulnessconstraints,asinlanguagesthatpermitCE.,UK., N)C0UK.,CE.,Ԁclusters. xG WithIdentIO[Nas]atthebottomofthishierarchy,candidate(20c)ismadeoptimal,asinwehave PI seeninMandar.Candidate(20d)ispreferredwhen#gXXHX#CE.,UK.,] Ident[ObsVce]HXXgXUK.,CE.,Ԁislowestranked,asinPuyo ( xK PungoQuechua.Finally,candidate(20e)winswithMaxdominatedbytheothers,asinKelantan "PM Malay.  WiththeintroductionofconstraintssuchasRootLinthatdisallowoneoftheCE.,UK., N)C0UK.,CE.,Ԁeffectsin %!Q aparticularenvironment,wewouldalsoexpecttoseecaseswhereanalternateprocesstakesplaceintheenvironmentinwhichtheusualoneisruledout.SuchconspiraciesbetweenCE.,UK.,N)C0UK.,CE.,Ԁeffectscan `)$U bemodeledsimplybyhavingbothoftherelevantFaithfulnessconstraintsrankedbeneath*CE.,UK.,0N)C0UK.,CE.,.It 8+&W isapowerfulargumentforthisapproachthatthisexpectationCE.,UK.,UK.,CE.,#gXXXHX #HXXXgXԀisindeedfulfilled.  3.3CE.,UK.,}N)C0UK.,CE.,ԀfusionoverruledbyFeaturalIdentity ` Inthissection,IshowhowahighrankingFeaturalIdentityconstraintcandisallowfusionbetweenparticularsegments.Thisdiscussionalsoservestointroduceevidenceofaconspiracybetweennasalsubstitutionandnasaldeletion.ThedatatobeaccountedforinvolveaparametricdifferencebetweenAustronesianandAfricannasalsubstitution.InalltheAustronesianexamplesofwhichIamaware,thefricative/s/undergoessubstitution: I  17      ׀   (21)  /mH& &XHXB#HXX &H&#N+sapu/ [mH& &XHXB#HXX &H&5#capu] h 'tosweep'  (Indonesian) \   /man+saga/ [macaga] h 'stay'    (Chamorro:Topping1973:50) H   /N+sambu/ [cambu] h 'toconnect'  (Javanese:Poedjosoedarmo1982:51) 4 Africanlanguageswithnasalsubstitutiondemonstrateasplitinbehaviourbetweenstopsandfricatives,asinthefollowingexamplescitedbyRosenthall(1989:49)(seealsoOddenandOdden1985onKhehe):CE.,UK.,<UK.,CE.,&  (22)  a./N+tuma/ [numa] h 'Isend'     l   /N+seva/ [seva]   h 'Icook'  (Umbundu:Schadeberg1982) X  ' $   b./N+tabi/ [nabi]   h 'prince' 0"    /N+supa/ [supa]   h 'soup'    (SiLuyana:Giv;n1970) #l! CE.,UK.,Tostemanysuspicionthatdeletionbeforethefricativesismotivatedsolelybythemarkednessof %0!$ nasal/fricativeclusters(seePadgett1994),notethatvoicedfricativesundergopostnasalhardeninginKhehe(OddenandOdden1985:598).Thisshowsthat*N)C0Ԁisneededfordeletionina  nasal/voicelessfricativesequence,sinceonewouldotherwisepredictthat/ns/shouldsurfaceas[nt].UK.,CE.,#gXXXHX#HXXXgX `   AsinIndonesian,fusionwiththevoicelessstopscanbeattributedtotherankingofLinearitybeneath*CE.,UK.,N)C0UK.,CE.,ԀandtherestoftheFaithfulnessconstraints,includingMax.However,  ` unlikeIndonesian,deletionoccurswithrootinitialvoicelessfricativesinsteadoffusion.ThisindicatesthatpreservationofInputcontinuancyismorehighlyvaluedthanpreservationoftheInputnasalsegmentintheselanguages,inotherwords,thatIdentIO[Cont]dominatesMax.Thefact   thatdeletiondoesoccurratherthana*CE.,UK.,N)C0UK.,CE.,Ԁviolationplaces*CE.,UK.,<N)C0UK.,CE.,ԀaboveMax.Combiningthese p rankings,weget*CE.,UK.,N)C0UK.,CE.,,IdentIO[Cont]>>Max>>Linearity.Thefollowingtableauxshowhow H thishiearchygeneratesthedifferentresponsesto*CE.,UK.,- N)C0UK.,CE.,Ԁviolationsinfricativeinitialandstopinitial  p roots:&  (23) Fusionwithstops    !*01 ddrc cc 9Sc Sc Sc Sc cc Xc ,-"",Fc ,cc ,FSc ,Sc ,ec +  /UUUU $t !bb/    Input:CE.,UK.,!!N1+t2abiUK.,CE., .UUrUUUU .#*CE.,UK.,#N)C0UK.,CE., >UU- UU >IdentIO  [Cont] @UVUV-r "UU @Max @UUU-!"UVUV @Lin OUU<*"" bb ! UUU BBO  CE.,UK.,[$a.n1t2abiUK.,CE., 7$ R#UU 7  HbHX*!#HXXHbT&# 7- R$  7 @VV- R%" @ @U- R&"VV @ OVUVU<* R'" BB  U BBO  b.n1,2abi3 7VV$!6(VUVU 7   >VV-!6) VV > @VVVV-!6*"VV @ @VVU-!6+"VVVV @HbHX*#HXXHb(# OUUU<*!6," BB  VVU BbO  c.t2abi 7U$#-UUU 7   >U-#. U > @VUV-#/"U @HbHX*!#HXXHb=*# @UU-#0"VUV @:0.#1" Bb   UU :' !Withastopinitialroot,Ident[Cont]issatisfiedinfusion,soMaxisfreetochoosefusion(23b) &>"3 overdeletion(23c)asthebestalternativetoa*CE.,UK.,%N)C0UK.,CE.,Ԁviolation(23a).2 ($5   2Whentherootbeginswithfricative,asin(24),fusioncreatesaviolationofIdentIO[Cont], *%7 sinceanInputfricativehasastopasanOutputcorrespondent(assuminganundominatedconstraint  againstnasalfricativesinalltheselanguages-seeCohn1993,Padgett1994).WithIdentIO[Cont]  >>Max,thecandidatewithdeletion(24c)becomesoptimalinthisinstance: ` &  (24) Deletionwithfricatives  $ t *23 ddFc cc FSc Sc ec 01"",c ,ecc ,Sc ,Sc ,c +  /UUUU  d !bb/    Input:CE.,UK.,!,N1+s2upaUK.,CE., .UUb  UUUU .0*CE.,UK.,S0N)C0#gXXXHX;#HXXXgXUK.,CE., >UU-v   UU >IdentIO v   [Cont] @UVUV-b  "UU @Max @UUU-v "UVUV @Lin OUU<*v " bb ! UUU BBO  CE.,UK.,1a.n1s2upaUK.,CE., 7$B UU 7  HbHX*!#HXXHb3# 7-B   7 @VV-B " @ @U-B "VV @ OVUVU<*B " BB  U BBO  b.n1,2upa 7VV$& VUVU 7   >VV-&  VV >HbHX*!#HXXHb5# @VVVV-& "VV @ @VVU-& "VVVV @HbHX*#HXXHb5# OUUU<*& " BB  VVU BbO  c.s2upa3 7U$ UUU 7   >U-  U > @VUV- "U @HbHX*#HXXHbF7# @UU- "VUV @:0. " Bb   UU :' 8 n.AustronesiannasalsubstitutionevincestheoppositerankingMax>>Ident[Cont],sincelossof T  Inputcontinuancy,asin(24b),ispreferredtodeletion.CE.,UK.,2  AsKisseberth(1970)originallypointedout,caseslikethisinwhichtwoprocessesconspiretoavoidasingleconfigurationprovidestrongmotivationfortheformalrecognitionofoutputconstraints.Underapurelyrulebasedanalysisofnasalsubstitution,suchasthestandardoneofnasalassimilationfollowedbyvoicelessconsonantdeletion,thefunctionalconnectionbetweennasalsubstitutionandnasaldeletionwouldhavetobestatedindependentlyoftherulesthemselves;theirsharedpropertyofeliminatingN)C0Ԁclustersisonlyobliquelyretrievablefromtheruleformulation#gXXXHX1#HXXXgX. <#. ThiscontrastswiththepresentOptimalityTheoreticanalysisofAfricannasalsubstitutionandnasaldeletion,inwhichthefunctionalmotivationfortheseprocessesisdirectlyincorporatedintotheformalexplanation,thusallowingforaperspicuousaccountoftheconspiracybetweenthem.UK.,CE.,  *%6  &   4CE.,UK.,<*N)C,#gXXXHX;#HXXXgXԀvs.redundantfeaturelicensingUK.,CE.,#gXXXHX=#HXXXgX  ` ItisofcoursenotthecasethatsimplybeingframedwithinOptimalityTheoryCE.,UK.,=UK.,CE.,#gXXXHX>#HXXXgXԀautomaticallyendows 8  ananalysisofoneoftheCE.,UK.,>N)C0#gXXXHX>#HXXXgXUK.,CE.,Ԁeffectswiththepowertoextendtothewholeset.AcaseinpointisIt=,  ` Mester,andPadgett's(1995)accountofpostnasalvoicing,whichingeneouslyreducesthephenomenontowhatappeartobemorebasicandgeneralconstraints,butCE.,UK.,?UK.,CE.,#gXXXHX?#HXXXgXԀfailstocopewithnasal   substitutionCE.,UK.,@#gXXXHX&A#UK.,CE.,HXXXgX,andalsostraightforwardlygeneratesanunattestedpatternofnasalobstruentvoicing.#gXXXHXA#HXXXgXCE.,UK.,AUK.,CE.,Ԁ   TheexistenceofaconspiracybetweenpostnasalvoicingandnasalsubstitutionCE.,UK.,B#gXXXHXlB#UK.,CE.,HXXXgXԀintheBantu p languageOshikwanyama,aswellasthenonexistenceofprenasalvoicing,arguefortheuseofarelativelyparochial,locallymotivatedconstraintlike*NCE.,UK.,]C)C0UK.,CE.,#gXXXHXsC#HXXXgX,whichbyhuggingthephoneticground,  p staysclosertotheattestedfacts.' '=CE.,UK.,DUK.,CE.,4.1Postnasalvoicing  #gXXXHXD#CE.,UK.,hEThebasicpremiseofIt=,Mester,andPadgettsanalysisisthatbecause[voice]isredundantin  sonorants,itcannotbelicensedbysonorants.Withthisrestriction,anasalspecifiedfor[voice]violatestheconstraintLicense[Voice],asinthefirstcandidateinthetableauin(25): 0"    $X" &   (25) Postnasalvoicingasredundantfeaturelicensing   *45 ddc ecc Sc Sc c 23"", ,rdd ,Ftt ,d +  5UUUU"  Mb5Input:NT +UUV$tUUUU +  License[Voice] 5UU$$t UUV 5  SonVoi @UVUU-$t UU @I  Faith QUU>*$t" b M UVUU M2Q Ja.NT̀|gR|'XgXԀ#gXX'gR|J#gR|'XgX[VOICE]#gXX'gR|J# +V  UU +  gbXgX*!#gXXgbOK# .$&  V . .VU&  . LVUU9!&  d2 M VU Mt2Lb.NT +VV (VUU + (V <VV (  gbXgX*!#gXXgbL# 7VVU$ < V 7 LVUUU9! < t2 M VVU MtbLc.ND̀|/̀gR|'XgXԀ[VOICE]#gXX'gR|yM#ԀHXXXgXUK.,CE.,Ԁ3CE.,UK.,M#gXXXHXM# +VUVB VUUU + ,VUj VUV , .VVUUj VU .  gbXgX*:0.j   tb M  VVUU :I  #gXXgbN#  '_Go'hAscanbeseenin(25b),thealternativeofleavingthenasalunspecifiedfor[voice]runsafouloftheimplicationalconstraintSonVoi,whichdemandsthatsonorantsmustbespecifiedfor[voice].The  finalcandidatemanagestosatisfybothLicense[Voice]andSonVoibyhavingasingle[voice]  featurelinkedtoboththenasalandtheobstruent,thelatterofwhichisabletolicenseit.ThiscandidateisoptimalwhentheFaithfulnessconstraintthatisviolatedbynonidentitybetweenthevoicingspecificationonInputandOutputobstruentsisrankedbeneathLicense[Voice]and _# SonVoi.IhavelabelledthisFaithfulnessconstraint'Faith'soastoabstractfromirrelevant 7% differencesinformulationbetweenIt=,Mester,andPadgett(1995)andthepresentanalysis.4.2Nasalsubstitution? o#+ USCE.,Tounderstandwhyredundantfeaturelicensingcannotdealwithnasalsubstitution,considerthetablein(26): `  'o"/    ([#0 S&   (26) NasalsubstitutionandredundantfeaturelicensingCE.,US.,OS   *67 dd rdd Ftt d 45"", ,zdd ,' ,'d +  5UUUU"  Mb5Input:NT 2UU!$tUUUU 2  License[Voice] 4UU!$t UU 4  SonVoi =UUU*$t UU =V  Linearity QUU>*$t  b M UUU M2QVa.Ǹ|gR|'XgXԀ#gXX'gR|wW#gR|'XgXԀ[VOICE]#gXX'gR|W# +!  UU +gbXgX  * '&   ' 4U!&  4  *#gXXgb1X# UVUUB*&   d2 M U Mt2Ub.N .V! < VUU .gbXgX +V <V +  * =VU* < V =  *#gXXgboY# UVUUB* <  t2 M VU Mt2UZc.NT̀|gR|'XgXԀ[VOICE]#gXX'gR|Z# .V! VUU .  gbXgX*#gXXgb,[#gbXgX 4V! B  V 4  =VU* B  V =[#gXXgbo[# LVUUU9! B  t2 M VU MtbLd.NT 2VU!X VUUU 2gbXgX  4VU!X  VU 4#gXXgb\#  gbXgX*#gXXgbO]#gbXgX =VUU*X  VU =C]  #gXXgb]#U:0.X   tb M  VUU :USCE.,'7TInalanguagewithnasalsubstitution,either(26a)or(26b)mustbeoptimal.However,theviolationsincurredbyeachofthosecandidatesareasupersetofthoseofoneofthefaithfulones,(~227~2~226~2c)and(22722262d)respectively.Therefore,fusioncouldnotbetheresultofanyrankingofthissetofconstraints.  Intuitively,onemightthinkthatnasalsubstitutionandpostnasalvoicingareinsomewayrelated,sincebothacttogetridofHXXXgXCE.,US.,}^N)C0USCE.,#gXXXHX`#Ԁsequences.Thisintuitionisborneoutbythefactsof R$ OshiKwanyama,awesternBantulanguagediscussedbySteinbergs(1985),whichdemonstratesaconspiracybetweennasalsubstitutionandpostnasalvoicing.Whiletherearenoalternations,rootinternalpostnasalvoicingisevidencedbythecomplemantarydistributionof[k]and[g][k]appearswordinitiallyandintervocalically,while[g]occursafternasals.Furthermore,loanwordsaremodifiedbyvoicingthepostnasalobstruent.ThefollowingareborrowingsfromEnglish:&  (22822272) PostnasalvoicinginOshiKwanyamaloanwords  'N#1 Wc  [sitamba] stamp f)$3   [pelenda] print   R*%4   [oinga] ink >+&5 ''ic *,z'6 Rootinitially,nasalsubstitution,ratherthanpostnasalvoicing,occurstoresolveunderlyingHXXXgXCE.,US.,`N)C0#gXXXHXWe#USCE.,  sequences(nasal/voicedobstruentclustersremainintact,thoughSteinbergsprovidesnoexamples):Ӏ(22922230222282) RootinitialnasalsubstitutioninOshiKwanyama  L  lf  /e:N+pati/ [e:mati] h ribs     /oN+pote/ [omote] h goodfornothing     /oN+tana/ [onana] h calf    AstraightforwardanalysisofOshiKwanyamaisobtainedundertheassumptionsofthepresentstudy.AsinIndonesian,rootinternalnasalsubstitutioncanberuledoutbyaRootspecificrankingofLinearityabove*HXXXgXCE.,US.,eN)C0USCE.,#gXXXHX i#,whilerootinitialsubstitutionispermittedbecausethegeneralLinearity P constraintisdominatedby*HXXXgXCE.,US.,aiN)C0#gXXXHXj#USCE.,.However,unlikeIndonesian,HXXgXCE.,US.,jUK.,CE.,Ԁ#gXXHXj#CE.,UK.,jIdent[ObsVce]HXXXgXUK.,CE.,Ԁisalsoranked ( beneath*CE.,UK.,kN)C0UK.,CE.,,sothatpostnasalvoicingoccursrootinternallyCE.,UK.,kUSCE.,#gXXXHX`k#.CE.,US.,ZlUSCE.,Alsocrucialhereistherankingof  CE.,US.,lIdent[ObsVce]>>Lin,sinceUSCE.,thereverserankingwouldresultinpostnasalvoicingeverywhere,as  canbeverifiedinthefollowingtableaubycomparingtheviolationsincurredbycandidates(23022292b)and(23022292c):CE.,US.,7mUSCE.,CE.,US.,UnUSCE.,& x (23022292) Rootinitialnasalsubstitution  8 CE.,US.,n*89 dd zdd ' 'd 67"", ,dd ,t ,Ft ,Qt +  5UUUU"  ( Mb5Input:N1#T2 .UUl! UUUU .  RootLin 7UUV$l!! UU 7  *HXXXgXN)C0#gXXXHX q# 5UU$l!" UUV 5q  Ident l!# [ObsVce] 7UVUU$X"$ UU 7q  Lin QUU>*l!%  b M UVUU M2Q-ra.N1#T2 . $Z&UU .  7V$ $Z'  7  gbXgX*!#gXXgb1s# .$ $Z( V .%s  7VU$ $Z)  7s :VUVU'! $Z* d2 M VU 2:b.N1#D2 .VV&P!+VUVU .  7VVV$&P!, VV 7tn  5VV$&P!- VVV 5t  gbXgX*!#gXXgbIu# 7VVVU$&P!. VV 7=u  AUUU.%&P!/  d2 VVVU 2Ac.#N1,23 .U'F#0UUU .u  7UV$'F#1 U 7sv  5U$'F#2 UV 5v  7VUU$'F#3 U 7w  gbXgX*#gXXgbvw#5+)'F#4  d2  VUU 5jw'x n  )$4 (23122302) Rootinternalpostnasalvoicing   *:; dd dd t Ft Qt 89"", ,dd ,t ,Ft ,Qt +  /UUUU  Mb/Input:N1T2 .UU4UUUU .  RootLin 7UUV$4 UU 7  *HXXXgXN)C0#gXXXHXz# 5UU$4 UUV 5vz  Ident 4 [ObsVce] 7UVUU$ p UU 7%{  Lin KUU8*4  b M UVUU M2K{a.N1T2 ."UU .  7V$"   7  gbXgX*!#gXXgb|# .$"  V .|  7VU$"   7#} =VUVU*!"  d2 M VU 2=b.N1D23 .VV  VUVU .&x  7VVV$  VV 7~  5VV$  VVV 5b~  gbXgX*#gXXgb~# 7VVVU$  VV 7~  AUUU.%   d2 VVVU 2Ac.N1,2 .U UUU .J  gbXgX*!#gXXgb# 7UV$  U 7  5U$  UV 5{  7VUU$  U 7̀  gbXgX*#gXXgb+#5+)   d2  VUU 5USCE.,CE.,US.,  USCE.,Sinceredundantfeaturelicensingcannotgeneratenasalsubstitution,itcannotexpresstheOshiKwanyamaconspiracy.Thismustbecountedasseriousinadequacy,especiallywithinOptimalityTheory,inwhichoutputconstraintsplaysuchacentralrole.ForfurtherevidenceofconspiraciesbetweenNHXXXgXCE.,US.,)C0#gXXXHX(#USCE.,ԀeffectsCE.,US.,USCE.,,drawnfromNewton's(1972)studyofGreekdialects,CE.,US.,҃USCE.,Ԁwhichposesimilar  problemsforredundantfeaturelicensingCE.,US.,3USCE.,,seePater(1996).CE.,US.,USCE.,̀CE.,US.,USCE.,4.3Prenasalvoicing 8$ AtleastasproblematicastheinabilityofredundantfeaturelicensingtogeneratenasalsubstitutionisitsabilitytogenerateCE.,US.,prenasalvoicing.TheresultofsupplyinganInput/TN/clustertoexactlythesamehierarchythatproducespostnasalvoicingisillustratedin(23222312):  ", &   (23222312) Prenasalvoicingasredundantfeaturelicensing   *JK dd dd t Ft Qt :;"", ,zdd ,tt ,rd +  5UUUU" h Mb5Input:TN +UUVUUUU +  License[Voice] 5UU$ UUV 5  SonVoi @UVUU- UU @E  Faith QUU>*" b M UVUU M2Qa.TǸ|gR|'XgXԀ[VOICE]#gXX'gR|1# +V6  UU +  gbXgX*! .$J   V . .VU^  . LVUU9!^  d2 M VU Mt2L#gXXgb#b.TN t   +VV` VUU + (Vt VV (  gbXgX*!#gXXgb# 7VVU$t  V 7 LVUUU9!t  t2 M VVU MtbLЀc.DǸ\|̀gR|'XgX[VOICE]#gXX'gR|#ԀgX(XgX3#gXX(gX># +VUV VUUU + ,VU B VUV , .VVUU B VU .  gbXgX*#gXXgb#' :0..   tb M  VVUU :Withjustthethreeconstraintsdiscussedthusfar,allsonorantswouldbe[voice]linkedtoadjacentobstruents.It=,Mester,andPadgettsingleoutnasalsastheonlysonoranttriggersof[voice]spreadbyintroducingasetofconstraintsthathavetheeffectofprohibitinglinkagebetweenobstruentsandsegmentsthataremoresonorousthannasals(theNoLinkconstraints).However,boththissolution, ! andthealternativeofchangingSonVoitoNasVoi(seeIt=,Mester,andPadgett1993,andthe x# discussioninIt=,Mester,andPadgett1995)wouldequallylimitpresonorantvoicingtonasals.Thoughpostnasalvoicingisextremelywidespread,therearenoreportedcasesofregressivevoicingtriggeredbynasalsonly.Theprogressivenatureofnasalobstruentvoicingisparticularlystrikingsincemoregeneralformsofvoicingassimilationtendtoberegressive(Anderson1979,Lombardi1991,Mohanan1993).Thisdirectionalasymmetry,whichisafundamentalpropertyofpostnasalvoicing(hencethename),completelyescapestheredundantfeaturelicensinganalysis.L I  18       % / HXXXgXUK.,CE.,  Itisworthnotingthattheasymmetryofnasalobstruentvoicingalsomilitatesagainstaview `'"1 ofpostnasalvoicingasautosegmentalspreadingof[voice](orcopyingofSonorantVoice;seeRice 8)$3 1993).Ifnasal[voice]canspreadright,thenwhycoulditnotspreadleft?#gXXXHX#CE.,UK.,ԀOneanswermightinvolve  claimingthatrulesonlyapplytorepairanillformedconfiguration,andthatHXXXgX*N)C0#gXXXHX˕#,butnotHXXXgX*)C0#gXXXHX3#N,  definesarepresentationinneedofrepair.However,ifspreadingisitselfnotthemotiveforce,butisonlyaresponsetoanindependentconstraint,thisessentiallyconcedesthepointthatthelocusoftypologicalexplanationhereliesintheconstraintsystem,ratherthanintheruleformalism.UK.,CE.,HXXXgX  ` & 8 4.4LymansLawandredundantfeaturelicensing   Whileredundantfeaturelicensingfailstogeneralizetonasalsubstitution(ortheotherCE.,UK.,ԗN)C0UK.,CE.,#gXXXHX#HXXXgXԀeffects;   seePater1996),itdoesgeneratethesonorant[voice]underspecificationrequiredforanOCPaccountofLyman'sLawinYamatoJapanese,andovercomestheorderingparadoxbetweenLyman'sLawandpostnasalvoicingfirstnotedbyIt=andMester(1986).#gXXXHX@#CE.,UK.,'8 "HereIwillbrieflydiscusswhether  p theLyman'sLawfactsbearatallonanunderstandingofpostnasalvoicing.  Lyman'sLawisacooccurrenceconstraintthatallowsonlyonevoicedobstruentperroot.ItcanbeanalyzedintermsofaOCPbasedrestrictionagainstadjacent[voice]features,providedthatsonorantsareunspecifiedfor[voice]whenthisrestrictionapplies.Ifpostnasalvoicingisviewedasthetransmissionofthenasal's[voice]featuretotheobstruent,thenLyman'sLawmustderivationallyprecedepostnasalvoicing.TheorderingparadoxarisesbecausethepostnasalvoicedobstruentisatargetforLyman'sLaw,whichwouldleadonetobelievethatpostnasalvoicingoccursbefore,ratherthanafter,Lyman'sLaw.  Redundantfeaturelicensingresolvesthisparadoxbysupplyinga[voice]featuretosonorantsonlyintheNCcontext.ThisisdonebyrankingLicense[Voice]aboveSonVoi,sothatwhenthere )$( isnoadjacentobstruentlicenserthatwouldallowthesatisfactionofbothconstraints,thesatisfaction h+&* ofLicense[Voice]takesprecedence:  Ӏy&  (23322322) UnderspecificationofnonNCsonorants  t *MN dd zdd tt rd JK"", ,dd ,tt +  ,UUUU T  Mb,Input:NV +UUV 8UUUU +  License[Voice] 7UUU$ 8 UUV 7  SonVoi HUU5* 8  b M UUU M2Ha.NV̀|gR|'XgXԀ#gXX'gR|-#gR|'XgX[VOICE]#gXX'gR|p# +Vr  UU +  gbXgX*! 7U$   V 7 CVUUU0!   d2 M U Mt2C#gXXgb#b.NV3    +VUV VUUU + .VUU VUV .  gbXgX*#gXXgb4#'t:0.   t2 M  VUU :Withouttheadjacentobstruentashostfortheparasiticlicensingof[voice],thenasalwithout[voice]isoptimal.USCE.,  A*HXXXgXCE.,US.,ON)C0Ԅbasedanalysisofpostnasalvoicing#gXXXHXq#HXXXgXUK.,CE.,,incontrast,issilentaboutthepresenceor B absenceof[voice]onsonorants.OneresultofthisisthattheOCP+underspecificationaccountofLyman'sLawcouldbemaintainedbyunderspecifyingallsonorantsfor[voice],includingnasalsinthe#gXXXHX #CE.,UK.,,HXXXgXN)C0UK.,CE.,#gXXXHX#HXXXgXԀ#gXXXHX#CE.,UK.,Ƨconfiguration,sinceUSCE.,*HXXXgXCE.,US.,qN)C0Ԁwouldcontinuetodemandapostnasalvoicedobstruent,evenif  thenasalitselflacked[voice].Whenpost-nasalvoicingisattributedto#gXXXHX#USCE.,asubstantiveoutputconstraint  like*HXXXgXCE.,US.,N)C0,ratherthantoautosegmentalfeaturepropagation,theorderingparadoxthusquietly z! vanishes.  Ontheotherhand,becausethe#gXXXHX#USCE.,*HXXXgXCE.,US.,N)C0#gXXXHX#HXXXgXԀanalysisofpostnasalvoicingiscompletelyindependent *#z% ofsonorant[voice]underspecification,wearefreetocontemplatealternativeaccountsofLyman'sLaw.Iftemporaryunderspecificationofnon-contrastivefeatureslikesonorant[voice]wereatypologicallyproductivewayofdealingwithcooccurrenceconditionsandotherphonologicalregularities,thenthestandardanalysiswouldbesecure.However,asSteriade(1995a)notes,no *%- casesbesidesthatofsonorant[voice]appeartoexist.Itisthuswellworthconsideringalternativesthatgeneralizetootherphenomena,andavoidtheproliferationofderivationalstagesthattemporaryunderspecificationrequires.ExtantaccountsofLyman'sLawwhichmakenoappealtotemporaryunderspecificationcanbefoundinRice(1993),Lombardi(1995),Steriade(1995a),andUKCE.,It=andMesterCE.,UK.,#gXXXHX#HXXXgXԀ(1996);discussionoftheirrelativegeneralizabilitywouldunfortunatelytakeusfartooafield.  ` Thecrucialpoint#gXXXHXh#HXXXgXԀhereisthattheirveryexistenceshowsthatdealingwitheachofLyman'sLawand  8  postnasalvoicingcan,andprobablyshould,beaseparateundertaking.#gXXXHX2#     Insum,theredundantfeaturelicensingand*HXXXgXN)C0#gXXXHXu#Ԁanalysesofpostnasalvoicingextendto   differentphenomena:sonorant[voice]underspecification,andtheHXXXgXN)C0#gXXXHXR#Ԁeffectsrespectively.Whilethe p conspiraciesexaminedhereandinPater(1996)firmlyestablishtheneedforaunifiedtreatmentoftheHXXXgXN)C0#gXXXHXL#Ԁeffects,neitherempiricalnortheoreticalexigenciesforceananalyticconsolidationofLyman's  p Lawandpostnasalvoicing.HXXXgXUK.,CE., 5Conclusions   Ihavearguedthatnasalsubstitutionisbestanalyzedasfusionofanasalandvoicelessobstruent,  drivenbyaphonotacticconstraintagainstthissequence,*CE.,UK.,[N)C0UK.,CE.,,whichcanalsobesatisfiedbynasal X  deletion,denasalization,andpostnasalvoicing.Thetraditionalanalysisofnasalsubstitution,andtherecentanalysisofpostnasalvoicinginIt=,Mester,andPadgett(1995),wereshowntocapturebothtoomuch,andtoolittle,whencrosslinguisticpossibilitiesaretakenintoconsideration.Incontrast,thefactorialtypologypredictedbythepermutationoftherankingof*CE.,UK.,N)C0UK.,CE.,ԀandtheFaithfulness '#& constraintsisnearlycompletelyfulfilled.  Thefactthatlanguagesexercisearangeofoptionsindealingwith*CE.,UK.,N)C0UK.,CE.,Ԁviolations,alongwith h+&* theexistenceofconspiraciesbetweentheseCE.,UK.,N)C0UK.,CE.,Ԁeffects,providesstrongsupportfortheOptimality  TheoreticprogramofdecouplingphonotacticconstraintsfromFaithfulnessconstraints,andallowingthemtobefreelyrankedwithrespecttooneanother.However,theapparentlackofCE.,UK.,'N)C0UK.,CE.,Ԁepenthesis ` raisesanintriguingquestionforfutureresearch:IsitthecasethateveryphonotacticconstraintissatisfiedinallofthewayspredictedbythepermutationoftherankingsbetweenitandtheFaithfulnessconstraints?Gapsinfactorialtypologiesoftenserveasindicationsthatconstraintsmustbereformulated,butpersistentlinksbetweenmarkedconfigurationsandtheprocessesusedtorepairthemwouldseemtoforceamorefundamentalshiftintheoreticalassumptions.Eitherthat,orwecouldsettleforatheoryofgrammarthatisinsomerespectsonly'exegeticallyadequate',asopposedto'explanatorilyadequate',thatis,wecouldrestcontentwithhaving'madesomeprogressinunderstandingthefactsastheyare,thoughnotinthesenseofshowingthattheycouldnotbeotherwise'(Anderson1979:18).Suchresignationwouldbedisappointingthough,inlightofthestridesthatOptimalityTheoryhasmadetowardpredictiveexplanatoryadequacyinmanyareasofphonology.#gXXXHX<#CE.,UK.,sHXXXgXUKCE.,  Finally,Iwouldliketoconcludebycommentingonanissuethatbearsmoredirectlyonthe  mainconcernofthisvolume,thatis,thenatureoftheinteractionbetweenphonologyandmorphology.TheprimaryroleofmorphemespecificFaithfulnessinMcCarthyandPrince(1994b),andUrbanczyk(1996)istoexplainprosodicinfluencesonmorphologythatwereformerlyattributedtotemplates.Inthepresentpaper,arootspecificconstraintisusedtoaccountforaninfluenceintheoppositedirection:amorphologicalrestrictiononthephonotacticallymotivatedprocessofnasalsubstitution.Bykeepingthephonotacticconstraintgeneral,andemployingmorphologicallyconditionedFaithfulnessconstraints,weareabletostraightforwardlycapturetheOshiKwanyama h+&* conspiracy,inwhichthewaythat*CE.,UK.,N)C0UKCE.,#gXXXHXȾ#HXXXgXԀissatisfieddependsonthemorphologicalcontext.Thisis  countertotheusualapproachtothemorphologicalsensitivityofOCPeffects,inwhichthemorphologicaldomainofthephonotacticconstraintitselfisstipulated(McCarthy1986,Myers1994).Significantly,caseslikeOshiKwanyama,inwhichtherearedifferentresponsestoaphonotacticconstraintdependingonthemorphologicalenvironment,cannotbedealtwithinOptimalityTheorybyproliferatingdomainspecificphonotacticconstraints.Whatevertherankingofsuchspecificphonotacticconstraintsmightbe,thelowestrankedFaithfulnessconstraintwillalwaysbetheonethatisviolated.Itistobeexpectedthatcontinuedexaminationofthedifferencesinempiricalscopebetweenthese,aswellasotherapproachestomorphologicalinfluencesonphonology,shouldyieldaclearerunderstandingoftheprinciplesunderlyingmorphophonologicalprocesses.#gXXXHX#CE.,UK., 2  H 222222HXXXgXUSCE., H[XHX  7XXd^%%d7r P  References#HXXH[{#   #gXXXHXG#XXXgXCE.,US.,fUSCE.,CE.,US.,0   Alderete,John,JillBeckman,LauraBenua,AmaliaGnanadesikan,JohnMcCarthy,andSuzanneUrbanczyk.1996.ReduplicationandSegmentalUnmarkedness.Ms.,UniversityofMassachusetts,Amherst. "" 0   USCE.,#gXXXX#HXXXgX "" 0   Anderson,StephenR.1979.OntheSubsequentDevelopmentofthe'StandardTheory'inPhonology.InDanielA.Dinnsen(ed.)CurrentApproachestoPhonologicalTheory. L  Bloomington:IndianaUniversityPress. "" #gXXXHX6#XXXgXCE.,US.,USCE.,0   Archangeli,DianaandDouglasPulleyblank.1995.GroundedPhonology.Cambridge,MA:MIT `  Press.#gXXXX#HXXXgXCE.,US.,USCE.,L "" CE.,US.,USCE.,#gXXXHX#HXXXgX0   Beckman,Jill,CE.,US.,(USCE.,#gXXXHX]#HXXXgXԀLauraWalsh,andSuzanneUrbanczyk(eds.)1995.PapersinOptimalityTheory: $  UniversityofMassachusettsOccasionalPapers18CE.,US.,USCE.,#gXXXHX#HXXXgX.UniversityofMassachusetts,Amherst:   GLSA. "" CE.,US.,USCE.,0   #gXXXHX#CE.,US.,USCE.,HXXXgX0""  Chomsky,Noam,andMorrisHalle.1968.ThesoundpatternofEnglish.NewYork:Harperand  Row. "" #gXXXHX#CE.,US.,USCE.,HXXXgX0   Cohn,AbigailC.1993.TheStatusofNasalizedContinuants.InHuffmanandKrakow,329368. "" 0   Cohn,AbigailC.CE.,US.,USCE.,,andJohnJ.McCarthy.1994.AlignmentandparallelisminIndonesianphonology.#gXXXHX%#XXXgXMs.,CornellUniversity,Ithaca,N.Y.,andUniversityofMassachusetts,Amherst.[Revised  \ versiontoappearinLinguisticInquiry]#gXXXX#HXXXgXCE.,US.,UKCE.,.H"" #gXXXHXv#XXXgXCE.,UK.,#gXXXX#HXXXgXUSCE.,0   Drachman,Gaberell,andAngelikiMalikoutiDrachman.1973.StudiesintheAcquisitionofGreek   asaNativeLanguage:I.SomePreliminaryFindingsonPhonology.OhioStateWorking   PapersinLinguistics15.99114."" #gXXXHXa#XXXgXCE.,US.,0   USCE.,.CE.,US.,@USCE.,#gXXXX#HXXXgX0""  Dziwirek,Katarzyna.1989.MalagasyPhonologyandMorphology.LinguisticNotesfromLaJolla   15.130. "" #gXXXHX#XXXgXCE.,US.,mFlemming,Edward.1995.PerceptualFeaturesinPhonology.Ph.D.dissertation,UCLA. D#! USCE.,#gXXXX#HXXXgX0   Friberg,TimothyandBarbaraFriberg.1991.NotesonKonjoPhonology.InJ.N.Sneddon(ed.)StudiesinSulawesiLinguisticsPart2.NUSALinguisticStudies,Jakarta,Indonesia.&X!$"" #gXXXHXw#XXXgXCE.,US.,B#gXXXX#HXXXgXUSCE.,0   Giv;n,Talmy.1970.TheSiLuyanalanguage.UniversityofZambia,InstituteforSocialResearch, '0#& Communication6. "" #gXXXHX#CE.,US.,#USCE.,HXXXgX0   Gnanadesikan,AmaliaE.1995.MarkednessandFaithfulnessConstraintsinChildPhonology.Ms.,#gXXXHX;#HXXXgXCE.,US.,)USCE., *%) UniversityofMassachusetts,Amhert.+&*"" #gXXXHX#XXXgXCE.,US.,-#gXXXX#HXXXgXUKCE.,0   0""  Goldsmith,John.(ed.)1995.AHandbookofPhonologicalTheory.Cambridge,MA:Basil  Blackwell.#gXXXHX#CE.,UK.,6HXXXgXUKCE.,"" #gXXXHX6#HXXXgXCE.,UK.,UUSCE.,0   CE.,US.,#gXXXHX#USCE.,HXXXgXHalle,MorrisandGeorgeN.Clements.1983.ProblemBookinPhonology.Cambridge:MITPress.`"" CE.,US.,JUKCE.,CE.,UK.,#gXXXHX\#XXXgXUSCE.,0   Hayes,Bruce.1986.AssimilationasspreadinginTobaBatak.LinguisticInquiry17:467499.CE.,US.,UKCE.,8 "" 0   Hayes,Bruce.1995.APhoneticallyDriven,OptimalityTheoreticAccountofPostNasalVoicing.HandoutfromMITcolloquium,4/28/95.CE.,UK.,+#gXXXXf#HXXXgXUSCE., L "" 0   Henry,Jules.TheKaingangLanguage.InternationalJournalofAmericanLinguistics14.194204.$ "" 0   Herbert,RobertK.1986.LanguageUniversals,MarkednessTheory,andNaturalPhonetic   Processes.Berlin:MoutondeGruyter. "" #gXXXHX7#XXXgXCE.,US.,V0   USCE.,#gXXXX#HXXXgX0""  Huffman,MarieK.1993.PhoneticPatternsofNasalizationandImplicationsforFeature p Specification.InHuffmanandKrakow,303327. "" 0   Huffman,MarieK.1994.Laryngealspecificationsfortheimplementationofnasals.HandoutofpaperpresentedattheUniversityofDeleware,November14. "" 0   0""  Huffman,MarieK.CE.,US.,eUSCE.,,andRenaA.Krakow(eds.)1993.PhoneticsandPhonology5:Nasals, H Nasalization,andtheVelum.SanDiego:AcademicPress.4"" CE.,US.,UKCE.,CE.,UK.,#gXXXHX#USCE.,HXXXgX0   It=,JunkoandArminMester.1986.ThePhonologyofVoicinginJapanese.LinguisticInquiry17.   4973. "" 0   It=,JunkoandArminMester.1996.StructuralEconomyandOCPInteractionsinLocalDomains.InvitedLecture,WesternConferenceonLinguistics.CE.,US.,USCE.,#gXXXHX#HXXXgXl"" CE.,US.,#gXXXHX#HXXXgXUKCE.,#gXXXHXA#CE.,UK.,`HXXXgXUSCE.,0   0""  It=,JunkoandArminMesterCE.,US.,USCE.,,andJayePadgett.1993.LicensingandRedundancy: D! UnderspecificationTheoryinOptimalityTheory.Reportno.LRC9307,LinguisticsResearchCenter,UCSantaCruz. "" 0   0""  It=,JunkoandArminMesterCE.,US.,QUSCE.,,andJayePadgett.CE.,US.,USCE.,Ԁ1995.LicensingandUnderspecificationinOptimalityTheory.LinguisticInquiry26:571614.%0!$"" #gXXXHX#XXXgXCE.,US.,UKCE.,#gXXXX#HXXXgX0   Jun,Jongho.1995.PerceptualandArticulatoryFactorsinPlaceAssimilation:AnOptimality '#& TheoreticApproach.Ph.D.dissertation,UCLA.#gXXXHX#XXXgXCE.,UK.,UKCE.,(#'"" CE.,UK.,F#gXXXX #HXXXgX0   Kawasaki,Takako.1995.VoicingandCodaConstraints.McGillWorkingPapersinLinguistics |*%) Volume1,2245.h+&*"" USCE.,#gXXXHX#CE.,US.,HXXXgXUSCE.,0   Kenstowicz,MichaelCE.,US.,USCE.,,andCharlesKisseberth.1979.GenerativePhonology:Descriptionand  theory.NewYork:AcademicPress."" CE.,US.,\UKCE.,CE.,UK.,)#gXXXHX#USCE.,HXXXgX0   Kiparsky,PaulCE.,US.,yUSCE.,.1993.Blockinginnonderivedenvironments.PhoneticsandPhonology4:Studies ` inLexicalPhonology,ed.bySharonHargusandEllenKaisse,277313.SanDiego: L  AcademicPress. "" 0   Kirchner,Robert.1995.Contrastivenessisanepiphenomenonofconstraintranking.InProceedings  ` oftheBerkeleyLinguisticsSociety. L "" #gXXXHX#XXXgXCE.,US.,&  Kisseberth,Charles.1970.Onthefunctionalunityofphonologicalrules.LI1:291306.0 x $  " " '&USCE.,#gXXXX#HXXXgX22220   Lamontagne,Greg,andKerenRice1995.ACorrespondenceAccountofCoalescence.InBeckmanetal.,211223. "" 0   Lapoliwa,Hans.1981.AGenerativeApproachtothePhonologyofBahasaIndonesia.Canberra: p PacificLinguisticsD34. "" #gXXXHX#XXXgXCE.,US.,USCE.,#gXXXX#HXXXgX0   Lombardi,Linda.1991.Laryngealfeaturesandlaryngealneutralization.PhDdissertation,University 4 ofMassachusetts,Amherst. "" #gXXXHX.#XXXgXCE.,US.,0   Lombardi,Linda.1995.WhyPlaceandVoiceareDifferent:ConstraintInteractionsandFeaturalFaithfulnessinOptimalityTheory.Ms,UniversityofMaryland. "" USCE.,#gXXXX2#HXXXgXMaddieson,Ian.1984.PatternsofSounds.CambridgeUniversityPress.   0   Maddieson,IanCE.,US.,9USCE.,,andPeterLadefoged.1993.PhoneticsofPartiallyNasalConsonants.InHuffmanandKrakow,251301. "" 0   McCarthy,John.J.1986.OCPEffects:GeminationandAntigemination.LinguisticInquiry17.207 X  263. "" #gXXXHXn#XXXgXCE.,US.,.USCE.,0     CE.,US.,UKCE.,#gXXXX#HXXXgXMcCarthy,John.1993.TheParallelAdvantage:Containment,Consistency,andAlignment.#gXXXHXL#XXXgXCE.,UK.,USCE.,Handout #l! oftalkpresentedatROW-1,RutgersUniversity,NewBrunswick,N.J. "" CE.,US.,USCE.,222222CE.,US.,USCE.,CE.,US., USCE.,#gXXXX#HXXXgX0   McCarthy,JohnCE.,US.,8USCE.,,andAlanPrince.1993a.ProsodicMorphologyI:ConstraintInteractionand %0!$ Satisfaction.UniversityofMassachusetts,Amherst,andRutgersUniversity,NewBrunswick, &"% NJ,ms.[ForthcomingMITPress].'#&"" 0   McCarthy,JohnCE.,US.,USCE.,,andAlanPrince.CE.,US.,9USCE.,Ԁ1993b.GeneralizedAlignment.YearbookofMorphology1993 )$( ed.byGeertBooijandJaapvanMarle,79-153.Dordrecht:Kluwer. ""  h+&* 0   McCarthy,JohnCE.,US.,wUSCE.,,andAlanPrince.CE.,US.,USCE.,Ԁ1994a.TheEmergenceoftheUnmarked.ProceedingsofNELS  24.333379. "" #gXXXHXm#XXXgXCE.,US.,USCE.,0   McCarthy,JohnandAlanPrince.1994b.ProsodicMorphology:anoverview.TalkspresentedattheOTS/HILWorkshoponProsodicMorphology,UniversityofUtrecht.0x""  CE.,US.,USCE.,#gXXXXw#HXXXgX`x"x" 22220   McCarthy,JohnCE.,US.,USCE.,,andAlanPrince.CE.,US.,|USCE.,Ԁ1995.ReduplicativeIdentity.InBeckmanetal.,249384. "" &  #gXXXHX#XXXgXCE.,US.,#gXXXXL#HXXXgXUSCE.,0   McCarthy,JohnCE.,US.,USCE.,,andAlanPrince.CE.,US.,USCE.,ԀThisvolume.#gXXXHX#XXXgXCE.,US.,XԀFaithfulnessandIdentityinProsodicMorphology. `"" USCE.,#gXXXX#HXXXgX0   Mills,RogerF.1975.ProtoSouthSulawesiandProtoAustronesianPhonology.Universityof  8  Michigandoctoraldissertation.'$  "" 0   Mohanan,K.P.1991.OntheBasesofUnderspecification.NaturalLanguageandLinguisticTheory   9.285325. "" 0   Mohanan,K.P.CE.,US.,%USCE.,1993.FieldsofAttractioninPhonology.TheLastPhonologicalRule,ed.byJohn p Goldsmith,61116.UniversityofChicagoPress. "" Myers,Scott.1994.OCPEffectsinOptimalityTheory.Ms,UniversityofTexas.0   Newton,Brian.1972.TheGenerativeInterpretationofDialect.AStudyofModernGreek  \ Phonology.CambridgeUniversityPress.0""  H"" #gXXXHXZ#XXXgXCE.,US.,N USCE.,CE.,US., USCE.,#gXXXXs #HXXXgX0   Nurse,Derek,andThomasJ.Hinnebusch.1993.SwahiliandSabaki:alinguistichistory.Berkeley:   UniversityofLosAngelesPress. "" 0   Odden,David,andMaryOdden.1985.OrderedReduplicationinKhehe.LinguisticInquiry16.  497503. "" #gXXXHX #XXXgXCE.,US., USCE.,CE.,US.,USCE.,Ohala,JohnJ.andManjariOhala.1991.Replytocommentators.Phonetica48.271274. X  0   Ohala,JohnJ.andManjariOhala.CE.,US.,/USCE.,Ԁ1993.Thephoneticsofnasalphonology:theoremsanddata.InHuffmanandKrakow,225249. "" 0   Onn,FaridM.1980.AspectsofMalayPhonologyandMorphology.Bangi:UniversitiKebangsaan $D # Malaysia.[Publishedversionof1976UniversityofIllinoisdissertation]. "" 0   Orr,Carolyn.1962.EcuadorQuichuaPhonology.InStudiesinEcuadorianIndianLanguages,ed. '#& byBenjaminElson,6077.Norman,Oklahoma:SummerInstituteinLinguistics. "" CE.,US.,USCE.,0   0""  Padgett,Jaye.CE.,US.,USCE.,ԀCE.,US.,USCE.,1995b.FeatureClassesII.HandoutoftalkpresentedattheHollandInstituteofLinguisticsPhonologyConference,UniversityofAmsterdam,January1995.h+&*"" CE.,US.,,2220   Pater,Joe.1995.OnthenonuniformityofweighttostressandstresspreservationeffectsinEnglish.Ms,McGillUniversity.0h""  USCE.,h"h" CE.,US.,222#gXXXX#HXXXgXUSCE.,222222Pater,Joe.1996.*NC.ProceedingsofNELSCE.,US.,qUSCE.,#gXXXHXR#HXXXgXԀ26.0   0""  "" 0    "" 0   Pelankahu,R.A.,AbdulMuthalib,andM.ZainSangi.1983.StrukturBahasaMandar.Jakarta: t PusatPembinaandanPengembanganBahasa.CE.,US.,USCE., "" &  0   Pierrehumbert,Janet.1993.DissimilarityintheArabicverbalroots.ProceedingsofNELS23,367 8  381.'L  "" 0   Piggott,GlyneL.CE.,US.,USCE.,#gXXXHX!#HXXXgX1988.PrenasalizationandFeatureGeometry.ProceedingsofNELS19,345352.CE.,US.,USCE.,#gXXXHX #HXXXgX L "" Piggott,GlyneL.1992.VariabilityinFeatureDependency:TheCaseofNasality.Natural $  0  LanguageandLinguisticTheory10.3377. "" CE.,US.,USCE.,Piggott,GlyneL.CE.,US.,USCE.,Ԁ1993.Thegeometryofsonorantfeatures.Ms.CE.,US.,+USCE.,#gXXXHX#HXXXgX,McGillUniversity.   CE.,US.,USCE.,&  0   CE.,US.,'.&  USCE.,0""  Piggott,GlyneL.CE.,US.,USCE.,Ԁ1995.FeatureDependencyinOptimalityTheory:OptimizingthePhonologyofSonorants.Ms.,#gXXXHX#HXXXgXCE.,US.,USCE.,ԀMcGillUniversity.\"" 'p#gXXXHXy#CE.,US.,USCE.,HXXXgX0   Poedjosoedarmo,Soepomo.1982.JavaneseInfluenceonIndonesian.Canberra:PacificLinguistics 4 D38. "" CE.,US.,GUKCE.,CE.,UK.,?#gXXXHXY#USCE.,HXXXgX0   Prince,AlanS.andPaulSmolensky.1993.OptimalityTheory:ConstraintInteractioninGenerative H Grammar.MsCE.,US.,USCE.,#gXXXHX#HXXXgX,RutgersUniversityandUniversityofColoradoatBoulder.[Forthcoming 4 MITPress]. "" CE.,US., UKCE.,#gXXXHX #CE.,UK.,m!USCE.,HXXXgX&  0   Rice,Keren.1993.Areexaminationofthefeature[sonorant]:Thestatusof'sonorantobstruents'.  Language69.30844."" #gXXXHX!#CE.,US.,!HXXXgX'!USCE.,CE.,US.,0#USCE.,#gXXXHX"#HXXXgX0   Rice,KerenCE.,US.,]#USCE.,,andPeterAvery.1989.OntheInteractionbetweenSonorancyandVoicing.Toronto&   l WorkingPapersinLinguistics10.6582.'lU$X "" 0   Rosenthall,Samuel.1989.ThePhonologyofNasalObstruentSequences.Master'sThesis,McGill 0"  University. "" CE.,US.,#UKCE.,CE.,UK.,%#gXXXHX##HXXXgXUSCE.,0   Russell,Kevin.1995.MorphemesandcandidatesinOptimalityTheory.UniversityofManitoba,ms.$D #"" 0   Schadeberg,T.1982.NasalizationinUmbundu.JournalofAfricanLanguagesandLinguistics4. &"% 109132. "" #gXXXHX%#XXXgXCE.,US.,&USCE.,#gXXXX'#HXXXgX0   Selkirk,Elisabeth.1988.Dependency,Place,andtheNotion Tier.UniversityofMassachusetts, )$( Amherst,ms.|*%)"" CE.,US.,'USCE., h+&* CE.,US.,(#gXXXHX(#HXXXgX0   Smith,NeilsonV.1973.Theacquisitionofphonology:acasestudy.CambridgeUniversityPress."" 0   Stahlke,Herbert.1976.SegmentSequencesandSegmentalFusion.StudiesinAfricanLinguistics  7.4463.t"" #gXXXHXa)#XXXgX#gXXXX*#HXXXgX0   Steinbergs,Aleksandra.1985.TheRoleofMSCsinOshiKwanyamaLoanPhonology.Studiesin L  AfricanLinguistics16.89101.8 "" #gXXXHX;+#X*XXgX+*XX#g,+J,##gXX,gi,#HXXXgX0   Steriade,Donca.1993.Closure,Release,andNasalContours.InHuffmanandKrakow,401470.USCE.,( x"" CE.,US.,d-0   Steriade,Donca.USCE.,1995a.UnderspecificationandMarkedness.InJohnGoldsmith(ed.)Handbookof P  Phonology,114174.Cambridge:BasilBlackwell.#gXXXHX,#XXXgXCE.,US.,-0p""< p"p" USCE.,#gXXXX.#HXXXgX&  Steriade,Donca.1995b.Positionalneutralization.Ms,UCLA.   Teoh,BoonSeong.1988.AspectsofMalayPhonologyANonLinearApproach.Urbana:   0  UniversityofIllinoisatUrbanaChampaigndoctoraldissertation. "" Ѐ'f/&  #gXXXHXG/#CE.,US.,/USCE.,HXXXgX0   Topping,DonaldM.1969.ARestatementofChamorroPhonology.AnthropologicalLinguistics ` 11.6277.L"" 0   Topping,DonaldM.1973.ChamorroReferenceGrammar.Honolulu:UniversityofHawaiiPress.'`0$t"" &  CE.,US.,0#gXXXHX1#XXXgXUSCE.,#gXXXX2#HXXXgX0   Trigo,R.Lorenza.1993.TheInherentStructureofNasalSegments.InHuffmanandKrakow,369 L 400.#gXXXHX$3#XXXgXCE.,US.,28"" USCE.,#gXXXX3#HXXXgX'2&  0   Uhrbach,Amy.1987.AFormalAnalysisofReduplicationanditsInteractionwithPhonological  andMorphologicalProcesses.PhDdissertation,UniversityofTexasatAustin.'4"" CE.,US.,N4#gXXXHX4#XXXgXUSCE.,CE.,US.,+6#gXXXX 6#USCE.,HXXXgX0   Urbanczyk,Suzanne.1996.PatternsofReduplicationinLushootseed.Ph.D.dissertation,University  ofMassachusetts,Amherst. "" 0   Vogel,Irene.1976.Nasalsandnasalassimilationpatterns,intheacquisitionofChicanoSpanish.PapersandReportsinChildLanguageDevelopment,201214.4#!"" &  CE.,US.,{6UKCE.,CE.,UK.,w8#gXXXHX6#USCE.,HXXXgX' $N80   Wonderly,William.1951.ZoqueII.InternationalJournalofAmericanLinguistics17.105123. %\ #"" &  '%9CE.,US.,8USCE.,0   Ziervogel,D.,P.J.Wetzel,andT.N.Makuya.1972.AHandbookoftheVendaLanguage.Pretoria: &4"% UniversityofSouthAfricaPress. "" 0   Zuckerman,Suzanne.1972.Themigratoryhabitsofnasalconsonants:Astudyinphonological )$( universals.OhioStateUniversityM.A.thesis.CE.,US.,9Ԁ#gXXXHX8#XXXgX*%)"" #gXXXX;#2