FCT1.WPDTEXTdosaMVVgWPC9 M^e$N,b~٨,tryku-:1WRU!je.7"d$ϹZr68uM.zYcyy"#` :.Ur}n;*6,^Qr :xSa/(KZRJO*֕9{QN+Zpp0*O*r;`fk;o^sγDwDHپQ)vMJ,ZW6G;U%ޭY߆SN 4@?yͱt#!X} X4o2-XmȀ!bL}Wul5c&SB25N=K#  UN % 0 U>@ h4 h: 1m@ 0tw@!4au 0d6DU.TUN}#U@! h! h! h?! h/"/" B3H"H"H"H"H"H" 0#e"e"e"e"e"e"e"e"e"e"e"e"e"e"e"e" 1k### 0#%r& 0CK(e((-....22266666666f6a666f6a6 B'*6f7a7f%7a'7;7":2:gB:B:=DD=GKG^KLO\OjQzQQQXUXUhUxUUUXqX\r\)_bbbbbbbLhU@kk Cl A +l+l 0#l D+m@4 no >np,^r AYO 0Fnnnnnnnnnnn@4op'@4op(@4o p@4o pQ+@4Rop+@4<.o p.pK6@4oՁp$Չ@4ˮo p4@4o pQ @4ZZo p60  B fafa+Apple LaserWriter II NTPSCRIPT0 (9 Z 6Times New Roman RegularX(7$XucXXX  C 6 $#C--  \  `(Times NewRomanTTNR #)e37=CIQYag1.a.i.(1)(a)(i)1)a)(Lm$2(  1  )3  0  DDDDD~? (;3|x  z$XDXXXC 6 $$6C--  o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  1    ݀TherankingschemaC>>Constraint(Class)oversimplifiestheactualGXDX~)a``0#~XXD#Ԁsituation.InPadgett  (1994),thealignmentconstraintpenalizingassimilationacrossthemorphemeboundaryconflictsforallfeatures(andnotjustapropersubset)withasubordinateenforcingplaceassimilation.Otherhighrankingconstraintsrequiringthatnasalsreceivesomelicensedplacespecificationensurethatpartialassimilation t obtains,minimallyviolatingalignment.(!2z$ 7!X\XXX  C 6 $# C--   o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XX\Ԁ  0  shintoshinto1<6X9`(Courier 6Times New Roman Regularshinto2  z$XDXXXC 6 $$6C--  o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  4    ݀IthasbeenarguedthatvowelspreadinginTurkishmustbelefttoright(e.g.,Anderson1980).  Besidesthesuggestivelylicensedstatusoftheinitialvowel,evidencecomesfromdisharmonicforms,wheresomenoninitialvowelsareexceptionallyspecifiedforcolorfeatures,astateofaffairsthatcanleadtoanunspecifiedvowelbeingsurroundedoneachsideby[back]or[round].Suchvowelsalwaysreceivethesefeaturesfromtheleft.ForproposalsinOTterms,seeKirchner(1993)andPolgardi(1994).  z$XDXXXC 6 $$6C--  o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  2    ݀ThisaccountdrawsontheworkofLees(1961),Haiman(1972),ClementsandSezer(1982),van  derHulstandvandeWeijer(1991),Kirchner(1993),andreferencestherein.Forourpurposestheharmonizingfeaturesareassumedtobethefamiliar[back]and[round].Level 1Level 2Level 3Level 4Level 5mindy8MINDY12  z$XDXXXC 6 $$6C--  o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  4    ݀IthasbeenarguedthatvowelspreadinginTurkishmustbelefttoright(e.g.,Anderson1980).  Besidesthesuggestivelylicensedstatusoftheinitialvowel,evidencecomesfromdisharmonicforms,wheresomenoninitialvowelsareexceptionallyspecifiedforcolorfeatures,astateofaffairsthatcanleadtoanunspecifiedvowelbeingsurroundedoneachsideby[back]or[round].Suchvowelsalwaysreceivethesefeaturesfromtheleft.ForproposalsinOTterms,seeKirchner(1993)andPolgardi(1994).\  `(Times NewRomanTTUW d(3$ 7!X\XXX  C 6 $# C--   $35=CIOW_e4(1)A.1.a.(1)(a)i)a)  9'XDXXXC 6 $$6C--    `     h ~XXDFeatureClasses   p     @    385  A  (3$ 7!X\XXX  C 6 $# C--    h &(XDXXXC 6 $$6C--   ~XXD@    385  A  Ԁ `     h JayePadgett( $     [ 7!XDXXX@@ @ C 6 $$6C--   o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  *  ThisworkhasbenefittedfromdiscussionswithaudiencesatthesecondTrilateralPhonology  WeekendatStanford,UCLA,theUniversityofBritishColumbia,RutgersUniversity,theUniversityofMassachusetts,Amherst,theJohnsHopkinsUniversity,theUniversityofMaryland,theUniversityofPennsylvania,andinseminarsatUCSCandtheGironaInternationalSummerSchoolinLinguistics.IamverygratefultothelinguisticsdepartmentatRutgersUniversityforprovidingtheenvironmentinwhichthisworkcoulddevelop.Iwouldalsoliketoespeciallythankthefollowingpeopleforvaluablecomments:AkinAkinlabi,JillBeckman,JaneGrimshaw,MarkHewitt,HarryvanderHulst,JunkoIt=,JohnKingston,LindaLombardi,JohnMcCarthy,ArminMester,MireNChiosin,AlanPrince,DougPulleyblank,LisaSelkirk,PaulSmolensky,CherylZoll.SomeoftheideasherewerealsoairedinNChiosinandPadgett(1994).NONCONST  7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  15  Representationslikethesewithlinkingtoaconsonantraisethequestionofthephonetic  interpretationofthatconsonant.Underthecommonassumptionthatvowelsandglidesarenotfeaturallydistinguished,theseformsmightseemtoimplyyandpy.However,asMcCarthy(p.c.)notes,these   representationsimplymerelythattheinterveningconsonanthassome(inthiscase)palatalcharacter,thoughitneednotbephonologizedinanyway.Thisreasoningimpliesthataplainversuspalatalizedcontrastshouldnotbemaintainedinalanguagewhere[-back]linkslocallyasshownhere,i.e.noipi \ versusipyicontrastinsuchalanguage.MINDY1MINDY2  7!XDXXX@@ @ C 6 $$6C--   o[X4 hp x (#%'0*,.8135@8:<H?ACXo~XXDԀ  3   F ԀThisnotiongetsasettheoreticinterpretationinsection3.Ofcourse,wemightchoosetolabel  featuresfortherelevantproperties,e.g.,[back]CindicatesColor,LabialPPlace,andsoon.(Theretention H ofinertclassnodesseeninHalle1993,whereindividualfeaturespreadingisadvocated,isalabelingstrategyaswell).Sincethefeatureclassesarelargelyfamiliar,andtheirnumbersmall,thislinguist'saidisusuallyunnecessary.MINDYXXMINDY4MINDY5MINDY8Table_ATable_B&0 d dTable_CTable_D  7!XDXXX@@ @ C 6 $$6C--   o[X4` " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  5  Thecarefulspeech,lexicalcontrastisintendedhere(casualspeechpronunciationslike  i[]competentarepossible).Nasalplaceassimilationtoavelarwilloccurinothercontexts,including H inmonomorphemicwordsandfootinternally.SeeBorowsky(1986).MINDY10MINDY11 ] 7!XDXXX@@ @ C 6 $$6C--   o[X4` " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  6  ThesesetscarryovertheindeterminacyintheFGrepresentationabove.Itshouldbeclearthatthe  centralissueathandisnotthequestionofwhatclassesexist,andwhattheyconsistofprecisely(apartfromtheclaimsconcerningColor).Rather,thequestioninvolveshowwecharacterizeclasses.Aswe   haveseen,however,answeringthelatterquestionhasimplicationsfortheformer.  7!XDXXX@@ @ C 6 $$6C--   o[X4` " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  7  Oneargumentforthesefeaturaldependenciescomesfromtheentailment:ifthedependentfeature  ispresent,thensoisthedominatingfeature.Thusan[ant]or[dist]specificationrequiresthepresenceofCoronal,andthespecificationofanyfeaturerequires[son],thelatterhavingtheresultthat[son]isspecifiedforeverysegment(seeMcCarthy1988forareviewofthesepoints).Thisisessentiallyanissueoffeaturecooccurrence,andingeneralFGhasnotprovedanappropriatetheoryofthisareaconstraintsonfeaturecooccurrencesuchas*[Dorsal,+nasal],*[+asp,+son](andexamplescanbemultiplied)targetcombinationsthathavenoplausibleclassstatusongroundsofassimilationorneutralization(thelattergroundsformingtheempiricalnucleusofFG).Wemightthencomparetheentailmentsabovewithpositivelyformulatedconstraintslike'if[ATR]then[+high]'positedbyArchangeliandPulleyblank(1994a).Thoughtheformerareperhapsinviolable,incontrasttothe[ATR]constraint,thismerelyputsthematthetopofamarkednesshierarchyofconstraints,fromrelativelyweakoneslike*[Dorsal,+nasal]tovirtuallyundominatedoneslike*[+nasal,+cont,+cons].Thesemusingsstilldonotaddresstherequirementthat[ant]spreadwhenCoronaldoes,andsoon.MINDY 13  7!XDXXX@@ @ C 6 $$6C--   o[X4` " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  8  IwouldliketothankDonkaFarkas,WilliamLadusaw,andespeciallyPaulPortnerforveryhelpful  discussionofunfamiliarnotionsandnotationshere.Theyshouldnotbeassociatedwithanycontinuingerrorsonmypart.FORMAL  7!XDXXX@@ @ C 6 $$6C--   o[X4 " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  9  Sinceformsvaryinthenumberoffeaturesfromagivenclasstheycontain,thenotion'Ntrue'relates  tothenumberofviolationsonlyindirectly:twoformsmaybothscoreas'3true',yetifonehas4membersinarelevantrepresentationandtheother6,thefirstrequiresone'*'andthesecond3'***'.Alldiscussionhereignoresmultipleviolationsasaresultofdistancefromalignmentdestination,awellknownsourceofgradientviolationofalignment(seeMcCarthyandPrince1993a,bonthislatterformofgradience).Alignmentconstraintsareactuallydoublygradient,then:violationsareincurrednotonlyforeachunalignedfeature,butalsoforeachsyllable(orotherappropriateunit)bywhichalignmentfails. T 7!XDXXX@@ @ C 6 $$6C--   o[X4 " N z Dpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  10  M.Liberman(p.c.)notesaconceptualsimilaritywiththequestionposedinthetransitionto  GovernmentandBindingsyntaxabouttheburdenofexplanationforpatternsofwhmovement.General H formulationslikeAlign(Color)arelikeMoveinthattheexplanationofactualdifferencesinitseffect  mustnecessarilyliewithindependentconstraints.MINDY15  7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  11  ThankstoMarkHewittforfirstbringingtheimportanceofthisquestiontomyattention.MINDY16MINDY14DAMNFOOL  7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  13  GeometriesemployinganodeVPlace(seereferencesattheoutsetofsection3)havethesame  accountavailabletothem,asClements(1991)notes.FGisfreetowriterulesspreadingeitherPlaceor L VPlace.Iftheformeroccurs,totalvowelplaceassimilationistranslaryngeal.Ifthelatter,assimilation  spansallconsonanttypes.Onthislatterkindoftransparencywithoutclassnodes,seePadgett(1994, X 1995).MINDY17MINDY20MINDY22MINDY21  7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  14  Thisisaphonologicalrenderingoftheobservedphoneticindependenceoflaryngealsfromoral  articulations(Keating1988).Perhapsamoreimportantconnectionshouldbemadetotheabsenceofformanttransitionsimposedonvowelsbylaryngeals,incontrasttoallotherconsonanttypes.MINDY23 g 7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  12  Asinalanguagewhere[back]and[round]aremutuallydependent.YawelmaniYokuts,saidtohave  aruleof[round]harmony(andnonefor[back]),forexample,hassuchaninventory.Yettheruleconverts[a]intoo,leavinganinterpretationintermsof[back]lessthanobvious(seeArchangeli1984and   referencesthereinonYokuts,where[a]~~Ԁistreatedas[+back]).MINDY24 h 7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  16  Presumablyanaccountbroadlyalongtheselinescanbeextendedtocasesofgutturaltransparency  (seeMcCarthy1991forfacts,andanaccountwithinFGsimilarinstructuretoSteriade's1987accountoftranslaryngealcases).However,sincePharyngealplaceisitselfrathermarked,thisclassofcasesmightsuggestthattheviewhereoftherelationbetweensimplexandcomplexsegmentmarkednessrequiresfurtherthought.  7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  17  AratherdirectapproachtotheArboreproblemispossible:adaptingideasinMcCarthy(1994),we  mightemployaconstraintdisfavoringtheplacefeaturecombinationformidfrontvowels,*[-back,-low,τhigh].AsMcCarthyshows,suchmarkednessconstraintscanthemselvesdrivemultiplevowelplacelinkage,withouttheneedtopositalignmentatall.MINDY25MINDY27 u 7!XDXXX@@ @ C 6 $$6C--   o[X4` hpx (#%'0*,.8135@8:<H?ACXo~XXDԀ  18  Tunicatranslaryngealcolorharmonytargetsonlythelowvowela.TheoutputislowfrontEorlow  backO,regardlessoftheheightofthetriggeringvowel(high,midorlow).Onemightattemptto L reanalyzethiscaseasoneoftotalvowelplaceassimilationinprinciple,byrequiringthatthetarget's[+low]valueremainparsed.Thiswouldhavetheeffectofpreventingthe(inconsistent)heightfeaturesofmidandhighvowelsfromlinking,givingpartialclassspreading.However,ifColorisalegitimate  featureclass,itisnotclearhowadirecttranslaryngealColorharmonycanbe(elegantly)ruledoutin d FCT.Inanycase,giventhegenerallyrestrictedoccurrenceofharmonies,andcolorharmonies,inlanguages,itisnotclearwhatwecanconcludefromthispaucityofattestedtranslaryngealcolorharmonies,andthisissuewillremainopenhere. { z$}XDXXXC 6 $$6C--  o[X4 hp x (#%'0*,.8135@8:<H?ACXo}~XXDԀ  1    ݀TherankingschemaC>>Constraint(Class)oversimplifiestheactualGasituation.InPadgett  (1994),thealignmentconstraintpenalizingassimilationacrossthemorphemeboundaryconflictsforallfeatures(andnotjustapropersubset)withasubordinateenforcingplaceassimilation.Otherhighrankingconstraintsrequiringthatnasalsreceivesomelicensedplacespecificationensurethatpartialassimilation L obtains,minimallyviolatingalignment.\  `(Times NewRomanTTd&OLE Box<<=8C(3$ 7!X\XXX  C 6 $# C--   'dxdOLE Generated ObjectWPC! }jm*3',,,,51+'"!t8xcX 6Times New Roman Regularc @XXX[+nas]""t84 * cX 6Times New Roman Regularc @XXXRootRoot"#pNG JcX 6Times New Roman Regularc @XXXԀDorsLab+'1"( [cfM") O%X,)"* cJD)(6nhhJ6 %9<F s>3{B6 O&4BMN"&޲ <sD@@ vVSQdUAWPWin 6.0/OLE 1.0 Prefix Information Marker WPGraphic21 WPC! }jm*3',,,,51+'"!t8xcX 6Times New Roman Regularc @XXX[+nas]""t84 * cX 6Times New Roman Regularc @XXXRootRoot"#pNG JcX 6Times New Roman Regularc @XXXԀDorsLab+'1"( [cfM") O%X,)"* cJD)(6nhhJ6 %9<F s>3{B6 O&4BMN"&޲ <sD@@ vVSQdUA METAFILEPICTH($ F X  HTimes New Romank-- .   --$ < < / / 0 0 < < <---% < < / / 0 0 < < <---$ 8554400445588---% 8554400445588--- !2 !2 !3 !3 !3 !4 !4 !4 !5 !5 !6 !6 !7 !7 !8 !8 !9 !9---.%422334589999999853344--(%489999999985333224--- !2! !3 !3" !4 !4" !5 !5" !5" !6 !6" !7 !7" !8 !8" !8" !8% !9 !9! !9" !9$---R%'"8"9!9 99877 6!6"5"5"3!3 3 4 4 55543 3!2"2#3#4#5#7#9$9$9$9%8%9#9#9"9"8--% "8"5!6 6 7 9!9!9"8---R$')2)5(5(3&3&3%4&4&5&5(6)7)8(9'9%9%9%9%9%7%7%8&8&9'9(9(8(7'7%5%5%4%3&3'2'3(3(3)2---R%')2)5(5(3&3&3%4&4&5&5(6)7)8(9'9%9%9%9%9%7%7%8&8&9'9(9(8(7'7%5%5%4%3&3'2'3(3(3)2---$ */./.<*<*<,<,0*0*/---% */./.<*<*<,<,0*0*/-"System--Times New Romank-- .  -- ! 8 ! ; !8 !< !8 != !8 != !8 !< !8 !9 !; !8 !; !8 !< !8 !< !8 != !7 !>---F%!A>:::99::;;77788888 7 7 7 : < =>>>=<>@A--% 9:;<===; : :9--- !C !D !B !E !B !F !A !F !A !F !A !F !B !F !C---&%DFGGGFEDCBAAABBCD--% DDCBBBDFFFEED--- !J !K !I !L !I !M !H !M !H !M !H !M !I !M !J---&%KMNNNMLKJIHHHIIJK--% KKJIIIKMMMLLK---2$QQRRQQQRRSSRQPPPOOOPPPQ---2%QQRRQQQRRSSRQPPPOOOPPPQ--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---H%"    --%  --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---&%--% --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---&%--% ---2$---2%---Times New Romank-- .  -- !1s !2t !2y !3t !3z !4t !4{ !5t !5{ !6t !6{ !7t !7{ !8t !8{ !9t !9z !:t !:y !;t !;w--- %s;s;s;--$%s;t;t9t3t2s2s2s1w1y2z2}5}6{:z;s;--% u:v;w;y:{8{6{5z3y2w2v2u3u:--- !4 !5 !5 !6~ !6 !7~ !7 !8~ !8 !9~ !9 !: !: !; !;---&%46789:;;;:~9~8~66544--"%556679;;:987655---B$465445665567::;;;;;;;::76655544---B%465445665567::;;;;;;;::76655544---X$*4776555667779:;;;;;;;99;;:9988776654445554---X%*4776555667779:;;;;;;;99;;:9988776654445554---<$89;;;;;;93222112239:;;;;::98---<%89;;;;;;93222112239:;;;;::98--- !4 !5 !5 !6 !6 !7 !7 !7 !8 !8 !9 !9 !: !: !: !; !;---T%(:;;;;;;:987766555566776644455679::::;;;:--% :78889:::--- !1 !2- !3- !4 !5 !5 !5 !6 !6 !6 !7 !7 !8 !8 !9 !9 !: !: !; !;---(%645789:;;;43222116--% 6:;;;:987556--66----,G-- %11--- !F !H !J !R ! U !!W !#^ !$_ !%a !(j !)l !*n !-v !.y !/{--3?M << deUULevel 1Level 2Level 3Level 4Level 5(n$ (  1  ) OLE Generated ObjectWPCr0 u ))3',,,,519V+' !c 6Times New Roman RegularcXXX @[+nas]Place "boc 6Times New Roman RegularcXXX @RootRoot # c 6Times New Roman RegularcXXX @DorsLab'1 $$ %P W q m & / x) ' .)'1 (dc 6Times New Roman RegularcXXX @?@ obuAt@3:Q 'c2_!& ufP"  3[DFגPbWV4 ^dI(JB P$D(9ifhWPCub wؘqyaP K#C / !"QErD?0*$a ԉV`P" FWW/    " 4 E+K tF1**  HʠLC^䱉D5ǁk).;DHnT҅ hLx#w\`'b||0&A aLZX2*W~ ؉VڋڒC tv?4πS/Mw v؂*&G Y[zp򣅋=veEB?]Iwgp P2\DDbxvԗ"~1L;۟荆zPCFXV]4]7_&&>@g%&w>;tAwAeeQHMNHdžv*z$L%G`-!-G&9M5|#4CG_D"&Ϙya?O} ?Z%. ȉkxjQK%nd/um&9Gs$nh?b ph9r &C'fuA+4-D>!r8 ,vN,  R; z  t'8%Q)u(y5O/ӻ0WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@pzOle 9WPG20CompObj!^  "%&'()*+,-./0123456789:;<=>?@ABCDEFGHI)F! WPWin61[7]ࡱWPCr0 u ))3',,,,519V+' !c 6Times New Roman RegularcXXX @[+nas]Place "boc 6Times New Roman RegularcXXX @RootRoot # c 6Times New Roman RegularcXXX @DorsLab'1 $$ %P W q m & / x) ' .)'1 (dc 6Times New Roman RegularcXXX @?@ obuAt@3:Q 'c2_!& ufP"  3[DFגPbWV4 ^dI(JB P$D(9ifhWPCub wؘqyaP K#C / !"QErD?0*$a ԉV`P" FWW/    " 4 E+K tF1**  HʠLC^䱉D5ǁk).;DHnT҅ hLx#w\`'b||0&A aLZX2*W~ ؉VڋڒC tv?4πS/Mw v؂*&G Y[zp򣅋=veEB?]Iwgp P2\DDbxvԗ"~1L;۟荆zPCFXV]4]7_&&>@g%&w>;tAwAeeQHMNHdžv*z$L%G`-!-G&9M5|#4CG_D"&Ϙya?O} ?Z%. ȉkxjQK%nd/um&9Gs$nh?b ph9r &C'fuA+4-D>!r8 ,vN,  R; z  t'8%Q)u(y5O/ӻ0 `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; ࡱ; d     d3f3333f333ff3fffff3f3f̙BOlePart#OlePres000$H 3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--C2 M ([+nas] Place  M -- .  --.2 ARoot Root  -- .  --2 x Dors Lab  x--?!P-- %lRj--A--- !@ !? !> !< !; !9 !8 !7 !5 !4--- .  -- 2 .w?.~-ࡱ; ࡱ;  METAFILEPICTd ( d     d3f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--C2 M ([+nas] Place  M -- .  --.2 ARoot Root  -- .  --2 x Dors Lab  x--?!P-- %lRj--A--- !@ !? !> !< !; !9 !8 !7 !5 !4--- .  -- 2 .w?.~-OLE Generated ObjectWPC{'63',,,,519V+'")yc 6Times New Roman RegularcXXX @a.*sondon"+ c 6Times New Roman RegularcXXX @b.*sondAn",c 6Times New Roman RegularcXXX @Color"- Jc 6Times New Roman RegularcXXX @Color".!c 6Times New Roman RegularcXXX @[back][round]"/4 Vc 6Times New Roman RegularcXXX @[back][round]'1"0O"1 W "2"3  )"5WD\$Jf 0SS$@ B!:S <  8 J: :FpM P    f`P  $fHS$8  U Y W8 V d`WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@p1X{Ole 9WPG20CompObj!^  "%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLM)F! WPWin61[8]ࡱWPC{'63',,,,519V+'")yc 6Times New Roman RegularcXXX @a.*sondon"+ c 6Times New Roman RegularcXXX @b.*sondAn",c 6Times New Roman RegularcXXX @Color"- Jc 6Times New Roman RegularcXXX @Color".!c 6Times New Roman RegularcXXX @[back][round]"/4 Vc 6Times New Roman RegularcXXX @[back][round]'1"0O"1 W "2"3  )"5WD\$Jf 0SS$@ B!:S <  8 J: :FpM P    f`P  $fHS$8  U Y W8 V d` `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; wKraࡱ; 6" @      6"3f3333f333ff3fffff3f3f̙BOlePart#OlePres000$h 3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  bLTimes New Roman$-- .  "System--%2 a. *s o n d o n| -- .  --%2 b. *s o n d A n 8-- .  --2 :8Color :\-- .  --2 6Color 6-- .  --2 N'[back] [round]N-- .  --2 I[back] [round]IA--!I.I-- *-- %8DD<X@-- %>:<--- !.K !-M !,O !)U !(W !'Y !%_ !$` !#b ! i--ࡱࡱ;  METAFILEPICT6"H 6"      6"3f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  bLTimes New Roman$-- .  "System--%2 a. *s o n d o n| -- .  --%2 b. *s o n d A n 8-- .  --2 :8Color :\-- .  --2 6Color 6-- .  --2 N'[back] [round]N-- .  --2 I[back] [round]IA--!I.I-- *-- %8DD<X@-- %>:<--- !.K !-M !,O !)U !(W !'Y !%_ !$` !#b ! i--OLE Generated ObjectWPC !  z!%3',,,,51+' ! cX 6Times New Roman Regularc @XXXsonun "@;cX 6Times New Roman Regularc @XXXbr'1 #x ) $x  %  tqFa܏ 0 u(F PHEt!'l'ttYp3=ċyQBQ.QQQcQN\iD a&G*ƝJyFD H-(p-~tK;^F|;Nw?;V|:f r3ч]GVW&M?kvr!A!9WT|GU[j;u"D%Bܥ JH޲khdFPSF}M FuFk4& 1Fw>#>i@_Z*t D\MkO,u^ɌF{ 뉈M_:~~uGr=WPWin 6.0/OLE 1.0 Prefix Information Marker  WPGraphic21 WPC !  z!%3',,,,51+' ! cX 6Times New Roman Regularc @XXXsonun "@;cX 6Times New Roman Regularc @XXXbr'1 #x ) $x  %  tqFa܏ 0 u(F PHEt!'l'ttYp3=ċyQBQ.QQQcQN\iD a&G*ƝJyFD H-(p-~tK;^F|;Nw?;V|:f r3ч]GVW&M?kvr!A!9WT|GU[j;u"D%Bܥ JH޲khdFPSF}M FuFk4& 1Fw>#>i@_Z*t D\MkO,u^ɌF{ 뉈M_:~~uGr= METAFILEPICTGgE R  1!Times New Roman-- .   --R$'---R%'--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---"%--"%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---.%--(%---@$---@%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---*%--,%-"System--Times New Roman-- .  -- ! !- !- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---.%--% --->$--->%---- %--- ! ! ! ! ! ! ! ! ! ! !---- ! ! ! ! ! ! ! !------4 d OLE Generated ObjectWPCZx?] A3',,,,51+'"!tK8#cX 6Times New Roman Regularc @XXX[son]""t !cX 6Times New Roman Regularc @ @XXXLaryngeal"#tt cX 6Times New Roman Regularc @XXX[nasal]"$t cX 6Times New Roman Regularc @XXX[voice][asp][glot]"%txcX 6Times New Roman Regularc @ @XXXPlace"&tjcX 6Times New Roman Regularc @ @XXXPharyngeal"'tcX 6Times New Roman Regularc @ @XXXOral"(t=cX 6Times New Roman Regularc @XXXPhar")tcX 6Times New Roman Regularc @XXXLabCorDors"*t cX 6Times New Roman Regularc @ @XXXVPlace"+tBcX 6Times New Roman Regularc @XXX[ant][dist]",tcX 6Times New Roman Regularc @ @XXXHeightColor"-tcX 6Times New Roman Regularc @XXX[lo][hi][bk][rd]'1".t"!"0tf "1t"E!"2t "#"4t,9M"5tBx"8t.q":t#!:,";t:My"= -m  -">m C"? 8 0 ;C "@,&WZNe;=i8< CW6tWPWin 6.0/OLE 1.0 Prefix Information Marker  WPGraphic21 WPCZx?] A3',,,,51+'"!tK8#cX 6Times New Roman Regularc @XXX[son]""t !cX 6Times New Roman Regularc @ @XXXLaryngeal"#tt cX 6Times New Roman Regularc @XXX[nasal]"$t cX 6Times New Roman Regularc @XXX[voice][asp][glot]"%txcX 6Times New Roman Regularc @ @XXXPlace"&tjcX 6Times New Roman Regularc @ @XXXPharyngeal"'tcX 6Times New Roman Regularc @ @XXXOral"(t=cX 6Times New Roman Regularc @XXXPhar")tcX 6Times New Roman Regularc @XXXLabCorDors"*t cX 6Times New Roman Regularc @ @XXXVPlace"+tBcX 6Times New Roman Regularc @XXX[ant][dist]",tcX 6Times New Roman Regularc @ @XXXHeightColor"-tcX 6Times New Roman Regularc @XXX[lo][hi][bk][rd]'1".t"!"0tf "1t"E!"2t "#"4t,9M"5tBx"8t.q":t#!:,";t:My"= -m  -">m C"? 8 0 ;C "@,&WZNe;=i8< CW6t METAFILEPICTZ?KZ?%, =_h  )Times New Romanf-- .   --$ (((((---% (((((---V$)!!  !!#$%%%%%%%##%%$$##""!! ---V%)!!  !!#$%%%%%%%##%%$$##""!! --- ! ! ! ! ! !! !! !" !" !# !# !$ !$ !% !%---% "$%%%$#"!--%  !#%%$"! --- ! ! ! ! ! ! ! ! !! !! !" !" !# !# !$ !$ !% !%---2% !$%%%%%%%%%$$! --,% $%%%%%%%%%$! ---$ ((((---% ((((-"System--Times New Romanf-- .  Times New Roman-BTimes New Roman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aKaL`L`L`M_N^O]O\P\P\P^Q_Q_R`R`S_S_T\TYTYSYRZR[Q[Q[P\O\O[N[M[K^K^KaK--% _L_L^K]K\L\N\O]O]O^O^N_M_L-- %[Q[RZRZSZS[T\T^T_S_R^R]R\Q[Q--- !Kc !Ke !Lb !Lf !Mb !Me !Na !Nb !Oa !Of !Pa !Pd !Qb !Qd---8%bNbObPcQdQdPePfOfPeQcQbQaPaOaNbMbLcKeKgKgLgLfMeNdNbN--%bNdNfLfLeKdKbMbN--- !Kj !Km !Kn !Lj !Lm !Lm !Mi !Mm !Nh !Nm !Oh !Ol !Ol !Ph !Pk !Pl !Pn !Qh !Qj !Ql !Qn---@%oKmPmPmQmQmQmQnQnPnPnQmQlQlPlOjQjQiQiQhQhPhOiMjLjKlKmKmLnKoK--% lKkKjLjMiOiPiPjQlOmNmLmKlK---.$sGpPpPpQpQqQqPrPqQpQpQoQoPqIrHrHqHqHpHpGsG---.%sGpPpPpQpQqQqPrPqQpQpQoQoPqIrHrHqHqHpHpGsG-----Times New Romanf-- .  --$ VVIIIIUUV---% VVIIIIUUV---X$*MLLLMNQRRRSSRRRRRQNMMMNQRRRSSRRRRQOMMMMLLM---X%*MLLLMNQRRRSSRRRRRQNMMMNQRRRSSRRRRQOMMMMLLM--- !L !L !M !M !N !N !O !O !P !P !Q !Q !Q !R !R !R !R !S !S---R%'RSSSRRQPPOONNMLLMMNNNNMMLLLMNPRRRRQRSSR--% QOOPQQRRQ--- !L !L !M !M !N !N !O !P !P !Q !Q !R !R !R !S !S---P%&LNNMMLLLMNNOQRRSSSRSSPPRRRRQQPPONMLLLL--LL--- !L !L !M !M !N !N !O !O !P !P !Q !Q !Q !R !R !R !R !S !S---P%&RSSSRRQPPOONNMLLMMNNNNMMLLLMNPRRRQRSSR--% QOOPQQRRQ---"$HQRRSSRRQKIIIHH---"%HQRRSSRRQKIIIHH---$ I I VVUUIII---% I I VVUUIII---Times New Romanf-- .  --$ }}pppp||}---% }}pppp||}--- !s !s! !t !t! !u !u! !v !v !w !w !x !x - !y----0%ssssstu x!t"t!t!s!s#s#s"t"u zzuts--ss--- !s% !s' !t$ !t' !u$ !u( !v# !v( !w# !w( !x$ !x( !y$ !y( !z%---(%&s(s)t*v*w)x)y(z&z%z$y$x#w#v$u$t%s&s--$%&s&t%t%u$v%w%x&y'y'y(y(w(v't's&s--- !o,- !p,- !s-- !t,-- !y,----% -o-p.p.q-q,q,p,p,o-o--(%-s-x-y.y.z.z+z+z,z,x,v,t,t,t+t+t-s-s---B$5w5x3z2z1z0y0x/w0v0t1s2s3s4t5t5u4u3u3u3t2s1t1u0v1w1x2y3y4y4x5w---B%5w5x3z2z1z0y0x/w0v0t1s2s3s4t5t5u4u3u3u3t2s1t1u0v1w1x2y3y4y4x5w--- !s8 !s9 !t7 !t: !u7 !u: !v7 !w6 !w< !x7 !x; !y7 !y; !z8---4%7v7w8x9y:y;y;x<w<w<x:z9z8z7y7x6w7v7t8s9s:s<u;v7v--%7u:u:u:t9t9s7u---$ <p?p?}<}<|>|>p<p<p---% <p?p?}<}<|>|>p<p<p---$ I}F}FpIpIpGpG|I|I}---% I}F}FpIpIpGpG|I|I}--- !sL !sM !tK !tN !uK !uN !vM !vN !wK !wN !xJ !xN !yK !yN !yN !zK !zN---N%%NyMzLzKzKyJxJxKwKwLwNuNuNtMsLtLtLuLuKuKuKtLsMsOsOtOuOxOyPyPyPyQyQyPzOzNzNy--% NyNvMvMwLwKxLyMyMyNy---V$)UsUuTuTtStRsRtQtQuQuRuRvSvUwUxUyTyTzSzQzQzQzQzQxQxQySySyTyTxSwQwQvQuQtRsSsSsTsTsUs---V%)UsUuTuTtStRsRtQtQuQuRuRvSvUwUxUyTyTzSzQzQzQzQzQxQxQySySyTyTxSwQwQvQuQtRsSsSsTsTsUs--- !sX !sY !tW !tY !t[ !uW !uX !u[ !vW !v\ !wW !w\ !xW !x\ !yW !yX !y[ !zW !zY- !{W----@%VtXsXsXuYtYsZs[s\t]u]v]x\y[zZzYzYzXyX{X}Y}Y}V}V}V}W}W{WuWtVt--tVtV-- %XuXwXyZy[y\x\w\v[u[tZtYtYtXu---$ ^papa}^}^|`|`p^p^p---% ^papa}^}^|`|`p^p^p---$ j}g}gpjpjphph|j|j}---% j}g}gpjpjphph|j|j}--- !sm !so !tl !to !ul !up !vl !vp !wm !wp !xl !yl !zl !zq !{k !{r !|l !|q !}m---X%*mwlwlultmsnspsrsrtrtrtrtqtptquqvoxnxnxmxmymynypyqyrzr{q|p}n}m}l|k|k{l{lzlzlylylxmxmw--% nsnsmtmumvnwnwpwpvpuotosns--% mzmzl{l|m|o|q|r{qzpzmz---$$uouxuzvzvzrzrzszsxsrsqsqrqrptouo---$%uouxuzvzvzrzrzszsxsrsqsqrqrptouo--- !sy !s{ !tx !t| !ux !u| !vx !v| !ww !w| !xx !x| !yx !y| !zy---(%{s|s}t~v~w~x}x}y|zzzyzxyxxwwxvxtys{s--% zsztytyvyxzy{y|y|w|t{t{szs---6$qssttxyyyyyyzzzzxt~t~s~ssrrq---6%qssttxyyyyyyzzzzxt~t~s~ssrrq---$ pp}}||ppp---% pp}}||ppp---Times New Romanf-- .  Times New Roman-BTimes New Roman- -- ! ! ! ! ! ! ! ! ! !- !-- !- !---6%--% ---($---(%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---<%-- %---D$ ---D% --- ! ! ! ! ! ! ! !- !- ! ! ! !-----0%--% -----Times New Romanf-- .  Times New Roman-BTimes New Roman- -- ! ! ! ! ! ! ! ! ! ! ! ! !- !-- !----6%--% ---^$----^%---- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---J%#--"%---@$---@%---`$.---`%.---X$*---X%*--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---R%'-- %--% --- ! ! ! ! ! ! ! !- !- ! ! ! !-----0%--% --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---F%!--"%---,$---,%-----Times New Romanf-- .  Times New Roman-BTimes New Roman- -- !  ! !  ! !  ! !  ! !  ! !  ! !  ! !  ! !  ! !  ! !  !---0%        --$%         --->$--->%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---B%--"%---.$  ---.%  -----Times New Romanf-- .  -- ! ! ! ! ! ! ! ! ! ! !- !----<%--% ---P$&---P%&--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---P%&--% ---<$---<%---Times New Romanf-- .  --:$---:%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---P%&--% --- ! - ! - !  !  !  !  !  !  !  ! !  ! !  ! !  !  ! ---,%              --%          ---<$---<%--- !" !$ !" !% !! !& !! !& !! !& !! !& !! !% !"---(%$%&'''&&%%$#!!!""$-- %##"""#$%&&&%$#--->$++,,---,,+++++++(()))))))((*+--->%++,,---,,+++++++(()))))))((*+--- !5 !9 !6 !; !6 !< !6 != !6 != !6 != !6 != !6 != !6 !< !6 !; !5---0%44556666655448<>>>=<:4--% 79;===<;9877--- !@ !B !@ !C !? !D !? !D !? !D !? !D !? !C !@---(%BCDEEEDDCCBA???@@B--% B@@@ABCDDDCB--->$IIJJKKKJJIIIIIJJFFGGGGGGGFFHI--->%IIJJKKKJJIIIIIJJFFGGGGGGGFFHI---T$(PPPOONMMMNOPPPONMLLLLLLNOOONMMLLLMNNOOOP---T%(PPPOONMMMNOPPPONMLLLLLLNOOONMMLLLMNNOOOP---Times New Romanf-- .  Times New Roman-BTimes New Roman- --:$---:%--- ! ! ! ! ! ! ! ! ! !- !-- !----.%--% ---.$---.%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---D% --"%---D$ ---D% --- ! ! ! ! ! ! ! !- !- ! ! ! !-----<%--% -----Times New Romanf-- .  --$   ---%   --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---R%'--% ---V$)---V%)---6$!!##!!!""###"!!     !!---6%!!##!!!""###"!!     !!---$ #&&##%%##---% #&&##%%##---$ 0--00..00---% 0--00..00--- !7- !6- !3 !5 !6 !2 !6 !6 !2 !6 !2 !6 !1 !6 !2 !6 !2 !6 !6 !3 !6---:%654322122345666557777888666--$%6666554433233466--- !9- !9- !:- !9----% ::;;:9999:--&%;;;<<8899999988:;---V$)BBAA@??>>??ABBBAA@>>>>>>>@@AAA@??>>>@@AAB---V%)BBAA@??>>??ABBBAA@>>>>>>>@@AAA@??>>>@@AAB---6$DDFFDDDEEFFFEDDCCCBBBCDDD---6%DDFFDDDEEFFFEDDCCCBBBCDDD---$ HKKHHJJHH---% HKKHHJJHH---Times New Romanf-- .  Times New Roman-BTimes New Roman- --h$2mrrsssrrvvuttrrrssoopppqqrmlllmmiijjjkkmmmllpponnm---h%2mrrsssrrvvuttrrrssoopppqqrmlllmmiijjjkkmmmllpponnm--- !w !z !v !z !v !y !u !v !u !z !u !y !v---6%vvvxxyzzywvuuuvvwy{{{zyxv--%vxzzyxvv---- !}- !} !| !| !{- !{- !} !{ !}---%~~~}}~--*%~|||}}}~}|{{|}}|||~--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---J%#--% -- %---V$)---V%)---0$---0%---J$#---J%#--- ! ! ! ! ! ! ! ! ! ! ! ! ! !--- %--% ---.$---.%--- ! ! ! ! ! ! ! ! ! ! ! ! ! !--- %--% ---:$---:%-----Times New Romanf-- .  --$ gddggeegg---% gddggeegg---&$jj jkkhhhii iihhhjj---&%jj jkkhhhii iihhhjj--- !n !n !p ! l ! p ! l ! q ! l ! q ! l ! q ! m ! p !n !p---$%oqr r r qonm l l l m nno--"%onn m m m opp q q q p po---$ twwttvvtt---% twwttvvtt---$ ---% ---V$)            ---V%)            ---- !- !- !----%--(%     ---$ ---% ---$ ---% --- !- !- ! ! ! ! !  !  !  !  !  !  !  !  !  !  !  ! !---(%     --%         ---P$&          ---P%&          ---$ ---% ---$ ---% ---B$             ---B%             --- !- !- ! ! ! ! ! !  !  !  !  !  !  !  !  !  !  !  ! !---4%         -- %           ---$ ---% ----*EU--YQlq--*E--*|-- %---- %9-- %-- %lz-- %-- %A-- %SlQY.k-- %-888488OLE Generated ObjectWPCOh"z!43',,,,519V+' !z c 6Times New Roman RegularcXXX @a.sondAn""(c 6Times New Roman RegularcXXX @br # c 6Times New Roman RegularcXXX @b.*sondon"$(#c 6Times New Roman RegularcXXX @br %` c 6Times New Roman RegularcXXX @sondan'1"&X  'M   *Mx 9  +)[ `$ ) , "-C  )"/P"  0( ! ! 1?! ! 3 !  d`,fD\|rX+~p++112$6*<62TJ:KBKJhMMQ` W"D"DE gA@ W8BhF WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawing ࡱ;  Root Entry`.@pa{Ole 9WPG20CompObj!^  "%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQ)F! WPWin61[0]ࡱWPCOh"z!43',,,,519V+' !z c 6Times New Roman RegularcXXX @a.sondAn""(c 6Times New Roman RegularcXXX @br # c 6Times New Roman RegularcXXX @b.*sondon"$(#c 6Times New Roman RegularcXXX @br %` c 6Times New Roman RegularcXXX @sondan'1"&X  'M   *Mx 9  +)[ `$ ) , "-C  )"/P"  0( ! ! 1?! ! 3 !  d`,fD\|rX+~p++112$6*<62TJ:KBKJhMMQ` W"D"DE gA@ W8BhF  `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; [iVZࡱ; O%M    M O%3f3333f333ff3fffff3f3f̙BOlePart#OlePres000$H 3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  [jTimes New Roman-- .  "System--!2 "a. s o n d A n "q -- .  -- 2 F/b rFD-- .  --"2 b. *s o n d o n -- .  -- 2 Cb rC-- .  --2 -sondanZ-- %299&>8-- %6#7#--(---*/--- !:0 !92 !84 !5: !4< !3> !0D !/F !.H !+N !*P !)R !&X---- !9C !8D !7E !6G !1M !0O !/P !.Q !*W !)X !(Y !'Z---6#-- -- --  -"&X ࡱ; ࡱ;  METAFILEPICTO%( O%M    M O%3f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  [jTimes New Roman-- .  "System--!2 "a. s o n d A n "q -- .  -- 2 F/b rFD-- .  --"2 b. *s o n d o n -- .  -- 2 Cb rC-- .  --2 -sondanZ-- %299&>8-- %6#7#--(---*/--- !:0 !92 !84 !5: !4< !3> !0D !/F !.H !+N !*P !)R !&X---- !9C !8D !7E !6G !1M !0O !/P !.Q !*W !)X !(Y !'Z---6#-- -- --  -"&X OLE Generated ObjectWPCDQ~8,3',,,,51+' !xcX 6Times New Roman Regularc @XXXRoot "cX 6Times New Roman Regularc @XXXRoot #NcX 6Times New Roman Regularc @XXX[+nas] $NcX 6Times New Roman Regularc @XXXPlace %cX 6Times New Roman Regularc @XXXVPlace"&+@cX 6Times New Roman Regularc @XXXColor 'cX 6Times New Roman Regularc @XXXVPlace'1 (\ ) c  * \ ) +U   , W\P>MMtvI up#)ַuA&v8ovv]5T!o'tvt53X ;K}Y"xp&on&$ } /Ň{ŴK}ŴŶ^Ŷ Y馌ĨWUz64G>qnqqqqlr *ͺհ}#yUT0*G!)qBWJ#qwq)o  8&)£]h=,HFNx Ě>?u@@&9Gpv+7WPWin 6.0/OLE 1.0 Prefix Information Marker WPGraphic21 WPCDQ~8,3',,,,51+' !xcX 6Times New Roman Regularc @XXXRoot "cX 6Times New Roman Regularc @XXXRoot #NcX 6Times New Roman Regularc @XXX[+nas] $NcX 6Times New Roman Regularc @XXXPlace %cX 6Times New Roman Regularc @XXXVPlace"&+@cX 6Times New Roman Regularc @XXXColor 'cX 6Times New Roman Regularc @XXXVPlace'1 (\ ) c  * \ ) +U   , W\P>MMtvI up#)ַuA&v8ovv]5T!o'tvt53X ;K}Y"xp&on&$ } /Ň{ŴK}ŴŶ^Ŷ Y馌ĨWUz64G>qnqqqqlr *ͺհ}#yUT0*G!)qBWJ#qwq)o  8&)£]h=,HFNx Ě>?u@@&9Gpv+7 METAFILEPICTGhBg6  !V  1!Times New Roman-- .   -- !y !} !y !~ !y !~ !y !~ !y !} !y !| !y !} !y !~ !y !~ !y !---D% |{{{{{|||xxyyyyyyxx|}--{{--% {}~~~}}|{{--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---$%--% --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---$%--"%---4$---4%-"System--Times New Roman-- .  -- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---J%#----% --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !--- %--$%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---"%-- %---6$---6%---Times New Roman-- .  --$ khhkkiikk---% khhkkiikk---$ oollooppttppo---% oollooppttppo--- !v !x !t !w !x !u !v !y !u !y !u !y !u !y !u !x !u !x---4%vxyyzzzzz{{xxxxxyyyxxwwv--.%vvvvwwttttuuuuttttvvv--- !~ !} ! !| ! !} ! ! !} ! !| ! !| ! ! !| !---T%(~~}}|||}~~}}}}}||~--% ~~}}~---V$)---V%)---$ ---% ---Times New Roman-- .  -- ! ! ! ! ! ! ! ! ! ! !- !---->%--% ---$$---$%--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---R%'--% ---@$---@%--- ! ! ! ! ! ! !- !- ! ! ! !---.%--% ---Times New Roman-- .  -- !" !) !" !) !" !) !# !) !# !( !$ !( !$ !( !% !' !% !'- !&----++--6%++**&&"""!!%%$$$'))))(((+--- !- !/ !- !1 !- !1 !- !2 !- !1 !- !/- !-----:%////00,,,-------,,,/233300/--% //011222110//---$$6667774455554466---$%6667774455554466--- !9 !; !9 !< !9 !< !9 !< !8 !< !8 !< !9 !< !< !9 !<---P%&<;:99888899;<<<;;::9::999;<====>>>>=<<--% <<::99:<<---B$EECBA??@@ABDEEDCCCCBAA@AABCCDDE---B%EECBA??@@ABDEEDCCCCBAA@AABCCDDE--- !G !H !F !I !E !I- !E- !K !E !J !F !J !G---0%FFGHIJJKKKIHGEEEGHIKJF--%FIIIIHF---Times New Roman-- .  --L$$,,,,+)''&&&()*+,,-+)'%$$$%&&()*++,,,---L%$,,,,+)''&&&()*+,,-+)'%$$$%&&()*++,,,--- !1 !0 !2 !. !2 !. !3 !. !3 !. !3 !/ !3 !0 !2--- %13444310/.../1-- %100///12333221---$$7778855566665577---$%7778855566665577--- !< !; != !9 != !9 !> !9 !> !9 !> !: !> !; !=--- %<>???><;:999:<-- %<;;:::<=>>>==<---B$CCEEFFEEDDDCCCCCDDAAABBBBAAAACC---B%CCEEFFEEDDDCCCCCDDAAABBBBAAAACC---Times New Roman-- .  -- !i !q !j !q !j !q !k !p !k !p !l !o !l !o !m !n !m !n- !n----ss--<%srrrqnnjjiihhlllkklnqqqqppps--- !t !w !u !y !u !y !u !y !u !y !u !v- !u- !t---:%vvvwwwssttuuuuuttsswy{{{zxv--% vvwxyyyxwwvv---*$~~~||||}}}}||||~~---*%~~~||||}}}}||||~~--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---P%&--% ---B$---B%--- ! ! ! ! ! !- !- ! ! ! ! ! !---0%--% ----z----66--- ! ! ! ! ! ! ! ! ! !---- !---- !6 !9 !B !E !N !Q !Z !] !f !i !r !u--\& 0 OLE Generated ObjectWPCb+3',,,,519V+'"! !c 6Times New Roman RegularcXXX @[son]""r:c 6Times New Roman RegularcXXX @[voice]LabPhar[hi][glot]Cor[back][lo][nasal]Dors[round][asp]"# Vc 6Times New Roman RegularcXXX @[ant][dist]'1"$kr !"% !"&  ! %"'  !:""( !")D !""*.  $  _« j@ obuAt@3:Q 'c2_!& ufP"  3[DFגPbWV4 ^dI(JB P$D(9ifhWPCub wؘqyaP K#C / !"QErD?0*$a ԉV`P" FWW/    " 4 E+K tF1**  HʠLC^䱉D5ǁk).;DHnT҅ hLx#w\`'b||0&A aLZX2*W~ ؉VڋڒC tv?4πS/Mw v؂*&G Y[zp򣅋=veEB?]Iwgp P2\DDbxvԗ"~1L;۟荆zPCFXV]4]7_&&>@g%&w>;tAwAeeQHMNHdžv*z$L%G`-!-G&9M5|#4CG_D"&Ϙya?O} ?Z%. ȉkxjQK%nd/um&9Gs$nh?b ph9r &C'fuA+4-D>!r8 ,vN,  R; z  t'8%Q)u(y5O/ӻ0WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@p?sOle 9WPG20CompObj!^  "%&'()*+,-./0123456789:;<=>?@ABCDEF)F! WPWin61[2]ࡱWPCb+3',,,,519V+'"! !c 6Times New Roman RegularcXXX @[son]""r:c 6Times New Roman RegularcXXX @[voice]LabPhar[hi][glot]Cor[back][lo][nasal]Dors[round][asp]"# Vc 6Times New Roman RegularcXXX @[ant][dist]'1"$kr !"% !"&  ! %"'  !:""( !")D !""*.  $  _« j@ obuAt@3:Q 'c2_!& ufP"  3[DFגPbWV4 ^dI(JB P$D(9ifhWPCub wؘqyaP K#C / !"QErD?0*$a ԉV`P" FWW/    " 4 E+K tF1**  HʠLC^䱉D5ǁk).;DHnT҅ hLx#w\`'b||0&A aLZX2*W~ ؉VڋڒC tv?4πS/Mw v؂*&G Y[zp򣅋=veEB?]Iwgp P2\DDbxvԗ"~1L;۟荆zPCFXV]4]7_&&>@g%&w>;tAwAeeQHMNHdžv*z$L%G`-!-G&9M5|#4CG_D"&Ϙya?O} ?Z%. ȉkxjQK%nd/um&9Gs$nh?b ph9r &C'fuA+4-D>!r8 ,vN,  R; z  t'8%Q)u(y5O/ӻ0 `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; ࡱ; 2`   23f3333f333ff3fffff3f3f̙BOlePart#OlePres000$3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--2 $[son]$ -- .  --i2 oA[voice]LabPhar [hi][glot]Cor [back] [lo][nasal]Dors [round][ asp]    o-- .  --2 [ant] [dist]-- %-a*Oa-- %pa+_-- %`+^-- %a, `-- %*b*TdTc-- %wc*a-- %t-ࡱ; ࡱ;  METAFILEPICT2h2   23f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--2 $[son]$ -- .  --i2 oA[voice]LabPhar [hi][glot]Cor [back] [lo][nasal]Dors [round][ asp]    o-- .  --2 [ant] [dist]-- %-a*Oa-- %pa+_-- %`+^-- %a, `-- %*b*TdTc-- %wc*a-- %t-OLE Generated ObjectWPCd  z 13',,,,519V+'"!1a c 6Times New Roman RegularcXXX @a.Acrossalaryngeal""1i c 6Times New Roman RegularcXXX @RootRootRoot"#1r Dc 6Times New Roman RegularcXXX @ကV?V"$1c 6Times New Roman RegularcXXX @Place'1"%1jOjP)"&1% OjZ)'1"'1 c 6Times New Roman RegularcXXX @b.Acrossasupralaryngeal"(1c 6Times New Roman RegularcXXX @RootRootRoot")1!4c 6Times New Roman RegularcXXX @ကVpV"*Vc 6Times New Roman RegularcXXX @PlacePlace"+1uc 6Times New Roman RegularcXXX @Lab'1",1Mg"-1".>M=)"/1Hf)'1"01<c 6Times New Roman RegularcXXX @xrJ3 4 f*# NG  M; / Qb" Hth$> 4 (8* "vu{MJ!Hb H! N`j36J͍*UtGtR#eqo*c$Äl*]PNQ\NȉXhFG5 i& fG ڲV X%09@N \Vr"\5)g&C38uf GA@u|B-Nx}LT? sy= r/S+:,,-2LdFQ+ tt.z% n̛MAu:G9_ts)LN1TwWPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawing&ࡱ;   Root Entry`.@pwuOle 9WPG20 CompObj1^  !"#$%&'()*+,-./0256789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklm)F! WPWin61[3]ࡱWPCd  z 13',,,,519V+'"!1a c 6Times New Roman RegularcXXX @a.Acrossalaryngeal""1i c 6Times New Roman RegularcXXX @RootRootRoot"#1r Dc 6Times New Roman RegularcXXX @ကV?V"$1c 6Times New Roman RegularcXXX @Place'1"%1jOjP)"&1% OjZ)'1"'1 c 6Times New Roman RegularcXXX @b.Acrossasupralaryngeal"(1c 6Times New Roman RegularcXXX @RootRootRoot")1!4c 6Times New Roman RegularcXXX @ကVpV"*Vc 6Times New Roman RegularcXXX @PlacePlace"+1uc 6Times New Roman RegularcXXX @Lab'1",1Mg"-1".>M=)"/1Hf)'1"01<c 6Times New Roman RegularcXXX @xrJ3 4 f*# NG  M; / Qb" Hth$> 4 (8* "vu{MJ!Hb H! N`j36J͍*UtGtR#eqo*c$Äl*]PNQ\NȉXhFG5 i& fG ڲV X%09@N \Vr"\5)g&C38uf GA@u|B-Nx}LT? sy= r/S+:,,-2LdFQ+ tt.z% n̛MAu:G9_ts)LN1Tw `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; ࡱ; z-f    fz-3f3333f333ff3fffff3f3f̙BOlePart3OlePres0004H3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--(2 "a. Across a laryngeal " -- .  --'2 R)Root Root Root   R-- .  --+2 C* V ? V  C-- .  --2 wPlace w--Vk--- !V8 !W; !YD !ZH ![P !\Q !]U !^\ !_^ !ah !bk !dt !ex !f !g !h !i !j--- .  --02 #b. Across a supralaryngeal #-- .  --'2 RRoot Root Root   R-- .  --+2 D V p V  D-- .  --2 xGPlace Place  x-- .  -- 2 ILab b--WRiR--{SS--Vk--- !l} !kz !ip !hn !gg !fc !eb--- .  -- 2 eYxea-Bࡱ; ࡱ;  METAFILEPICTz-(z-f   fz-3f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--(2 "a. Across a laryngeal " -- .  --'2 R)Root Root Root   R-- .  --+2 C* V ? V  C-- .  --2 wPlace w--Vk--- !V8 !W; !YD !ZH ![P !\Q !]U !^\ !_^ !ah !bk !dt !ex !f !g !h !i !j--- .  --02 #b. Across a supralaryngeal #-- .  --'2 RRoot Root Root   R-- .  --+2 D V p V  D-- .  --2 xGPlace Place  x-- .  -- 2 ILab b--WRiR--{SS--Vk--- !l} !kz !ip !hn !gg !fc !eb--- .  -- 2 eYxea-BOLE Generated ObjectWPC8573',,,,51+'"!'a cX 6Times New Roman Regularc @XXXa.Acrossalaryngeal""'i cX 6Times New Roman Regularc @XXXRootRootRoot"#'r DcX 6Times New Roman Regularc @XXXԀV?i"'' cX 6Times New Roman Regularc @XXXb.Acrossasupralaryngeal"('cX 6Times New Roman Regularc @XXXRootRootRoot")'!4cX 6Times New Roman Regularc @XXXԀVpi"+'ucX 6Times New Roman Regularc @XXXLab 1cX 6Times New Roman Regularc @XXX[-bk][+hi]'1 2Z) 3|Z) 4n|n'1"5?9cX 6Times New Roman Regularc @XXX[-bk][+hi]'1 6| ) 7w|WPWin 6.0/OLE 1.0 Prefix Information Marker WPGraphic21 WPC8573',,,,51+'"!'a cX 6Times New Roman Regularc @XXXa.Acrossalaryngeal""'i cX 6Times New Roman Regularc @XXXRootRootRoot"#'r DcX 6Times New Roman Regularc @XXXԀV?i"'' cX 6Times New Roman Regularc @XXXb.Acrossasupralaryngeal"('cX 6Times New Roman Regularc @XXXRootRootRoot")'!4cX 6Times New Roman Regularc @XXXԀVpi"+'ucX 6Times New Roman Regularc @XXXLab 1cX 6Times New Roman Regularc @XXX[-bk][+hi]'1 2Z) 3|Z) 4n|n'1"5?9cX 6Times New Roman Regularc @XXX[-bk][+hi]'1 6| ) 7w| METAFILEPICT7F76T UZ  Times New Romanf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``!`"a"a#^#^"^"_"_!__^^^^``---(%``!`"a"a#^#^"^"_"_!__^^^^``--- !c !e !b !f !b !f !c !f ! b ! f !!b !!f !"b !"e !"f !#c !#f---V%)f"e#d#d#c#c#b"b!b!b b cdfffeeddccccbbbcdffgg g"g"h"h"g#f#f#f"--% f!fdd c c!c"d"e"f!---:$kklmmmllkkk!k"k"l"l#h#h"i"i"i"i!iihhjk---:%kklmmmllkkk!k"k"l"l#h#h"i"i"i"i!iihhjk---R$'mppppoppq!ssssrrtttssq$o&n&n&n%m%n%n$n$o$o$p$p#q"nnnnm---R%'mppppoppq!ssssrrtttssq$o&n&n&n%m%n%n$n$o$o$p$p#q"nnnnm---V$)wxyyzz{{!{"{"|"|#x#x"x"y"y"y!yyxxww!w"w"x"x#t#t"t"u"u!uuuttvww---V%)wxyyzz{{!{"{"|"|#x#x"x"y"y"y!yyxxww!w"w"x"x#t#t"t"u"u!uuuttvww--- !~ ! !} ! !| ! !| ! ! } ! ~ ! !!| !"| !#| !# !$| !$ !%| !% !&}---(%~ } ||~  ~ --8%~ ~!}!}"~"~""""#%&&~&|%|%|$|$|$|#}#|"|"|!}!~ --% ~~~  --% }#}$}%%%$$##}#--- ! ! ! ! ! !- !- ! !! !! !" !" !#---4% !"""!! !"###! --% --- ! ! ! ! ! ! ! ! ! ! !! !! !" !" !" !# !#---T%("#####"!!  """"###"--% ! !"""!---&$!""##"""!---&%!""##"""!-"System--Times New Romanf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a !L` !Lc !M_ !Mc !N_ !Nd !O_ !Od !P_ !Pd !Q_ !Qc !R` !Rc---$%bKcLdLeNeOePdPdQcRbRaR_Q_O_MaKbK--"%bLaL`M`N`QaRbRcRcQdPdOdNcMcLbL--- !Ki !Lh !Lk !Mg !Mk !Ng !Nl !Og !Ol !Pg !Pl !Qg !Qk !Rh !Rk---$%jKkLlLmNmOmPlPlQkRjRiRgQgOgMiKjK--"%jLiLhMhNhQiRjRkRkQlPlOlNkMkLjL---2$pIpKqKqLpLpPpQpQqQqQqRpRnRnQnLmLmLnLnKoKoJoJpI---2%pIpKqKqLpLpPpQpQqQqQqRpRnRnQnLmLmLnLnKoKoJoJpI--- !H !I !I !J !J !K !K !L !L !M !M !M !M !N !N !N !O !O !P !P !Q !Q !R !R---L%$RRMNNNMPRRRRRRRRRPJIIHHHHHIIKLMMPQRR--MM--% MMLLKJIIIIM--- !K !L !L !M !M !N !N !O !O !P !P !Q !Q !R !R--- %KLLNOPRRRQOMKK--$%LLMNPQRRRQPONMLL--- !K !L !L !M !M !N !N !O !O !P !P !Q !Q !R !R--- %KLLNOPRRRQOMKK--$%LLMNPQRRRQPONMLL---0$IKKLLPQQQRRRQLLLLKKJJI---0%IKKLLPQQQRRRQLLLLKKJJI---Times New Romanf-- .  -- !91 !98 !:2 !:: !;2 !;9 !<3 !<9 !=3 !=8 !>4 !>8 !?4 !?8 !@5 !@7 !A5 !A7- !B6----9;9;--4%;9;::::;7C6C2;2:2:191959595:4:4;7A9;::8:8:8989;9--- !9c !:b !:e !;b !;f !<b- !<f- !>e !?d- !@d-- !Bd----AdAd--<%dAd?e>f=f<f;e:d:c:b;c;c<c<b<b<b;b:c9d9e9f:g;g=f>e?e@d@dA--%dBeBeCeCdCdCdB---- !9- !< !=- !>-- !B----% 99:::::99--(%<BCCCCCCBB?>>===<<---Times New Romanf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bbcddeeeddcccbb"b#b#c#c#`#`#`#a#a#a"aa```bb---D% bbcddeeeddcccbb"b#b#c#c#`#`#`#a#a#a"aa```bb--- !f !i !f !i !f !i !i ! g ! i !!f !!i !"e !"i !"l !#f !#h !#i !#k !$f !$j---P%&i#h#g$f$f#e"e"f!f!g h iiihggggfffffhjjjj!j#k#k#k#l"l#k$j$i#--% i"ih f"g"g#h#h#i"---"$nn"n#n#k#k#l#l"lllkkmn---"%nn"n#n#k#k#l#l"lllkkmn--- !q !t !q !t !q !t !t ! r ! t !!q !!t !"p !"t !"w !#q !#s !#t !#v !$q !$u---P%&t#s#r$q$q#p"p"q!q!r s tttsrrrrqqqqqsuuuu!u#v#v#v#w"w#v$u$t#--% t"ts r!r"r#s#s#t"---B$xxyzz{{{zzyyyxx"x#x#y#y#v#v#v#w#w#w"wwvvxx---B%xxyzz{{{zzyyyxx"x#x#y#y#v#v#v#w#w#w"wwvvxx--- !| ! !| ! !| ! ! } !  !!} !! !"~ !"- !#~- !%{ !&{---N%%{~~~}}}"~$~%}&|&|&{&{%|%|%}%}%~%~$~#|||{--{{--- ! ! ! ! ! ! ! ! !  !  !! !! !" !" !# !#---,%"#######"--(%"########"--- ! ! ! ! ! ! !  !  !! !! !! !" !# !# !$ !$ !% !% !& !&---.%!!  !!!!--.%!""###$%&&&&%%%$$##"!--"% !!!!!-- %#$%%&&&%%$$$$#--- ! ! ! ! ! !- ! - !! !" !" !# !$---,%!"""!!"$$$# --% --- ! ! ! ! ! ! ! !  !  !! !! !" !" !" !# !# !# !# !$ !$---N%%##$#"!!  !###"#$$##--% "  !"###"---$$"######"---$%"######"---Times New Romanf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`L`L_L_Q_R`R`QaQ_S^S^Q^L\L\L^L^K^J_J---,%_J_L`L`L_L_Q_R`R`QaQ_S^S^Q^L\L\L^L^K^J_J--- !It !Iw !Jt !Jx !Kt !Ky !Lt !Ly !Mt !Mx !Nt !Nu !Nw !Ot !Ow !Pt !Px !Qt !Qx !Rs !Ry !Ss !Sz---H%"|SzSvNvNvNuNuQuRvRwRwSsSsRsRtRtRtQtJtJtIsIsIsIvIyIzJzKzMyNxNzQ{R|R|S--% uMuNvNwNyLyKwIvIvIuJuM--- !L~ !L !M~ !M !N} !N !O} !O !P} !P !Q} !Q !R~ !R !S---$%LLMNOQSSS~R}Q}O}N~M~LL--% LLM~M~O~PRRQPOML--- !L !L !M !M !N !N !O !O !P !P !Q !Q !R !R !S---$%LLMNOQSSSRQONMLL--% LLMMOPRRQPOML---.$JLLLLQRRQQSSQLLLLLKJJ---.%JLLLLQRRQQSSQLLLLLKJJ---Times New Romanf-- .  -- !;! !;) !<" !<) !=# !=( !># !>( !?$ !?' !@$ !@' !A% !A& !B% !B&- !C&----;+;+--6%+;*;)<&D&D"<"<!;!; ; ;$;$;$;#;#<$=&B(=)<);(;(;(;+;--- !>L !>P !?M !?O !?Q !@M !@R !AM !AR !BM !BR !CM !CR !DM !DO- !EM----@%L>N>O>O?O?P>Q>R>S@SASCRDPDODODOFOGPGPGLGLGLGMGMGMFM@M?M?L?L>--"%O@OBOCODPDQDRCRBR@Q?Q?P?P?O?O@---- !:z- !>y- !?z----%z:{:{;{<z<z<z;z:--(%{>{C{D|D|DyDyDyDzDzCz@z?z?y?y?y>{>{>---Times New Romanf-- .  --<$QQPIIIJJJJJJJIIIMMLLLLLMNPPQ---<%QQPIIIJJJJJJJIIIMMLLLLLMNPPQ--- !S !V !S !V !S !V !T !V !R !V !R !V !X !S !V !V !X !S !V---R%'VUTSSRRRSTUVVVUTTTTTSSSSSUVWWWWXXXXXWVV--% VVTTSTTVV--- ![- !Z- !\ !Z ![ !^ !Z !^ !Z !_ !Z !_ !Z !_ ![ !^ ![---,%[\]^``_^]\[[ZZZYY[[[--% [[]^__^^]\[[---Times New Romanf-- .  --$ ~{{~~||~~---% ~{{~~||~~---$---%--- !- !- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---(%--% ---P$&---P%&---$ ---% ---$ ---% ---$ ---% ---Z$+---Z%+---- !- !- !----% --&%---$ ---% ---- %U--- ! ! ! ! ! !~ !} !| !u !~t !}r !yl !xk !wi !vh !rb !qa !p_ !o^ !kX !jV !iU !hT !dM !cL !bJ !]C !\B ![@ !W: !V9---- !V9 !W; !ZB ![C !\E !_L !`M !aO !dV !eX !fZ !i` !jb !kd !nj !ol !pn !rt !su !tw !w~ !x !y !| !} !~ ! ! ! ! ! ! ! ! !---VPP-Times New Romanf-- .  --$ soossqqss---% soossqqss---$ttwwt---%ttwwt--- !z- !y- !| !y !{ !} !y !z !} !y !~ !y !~ !y !~ !y !} !z !|---*%z|}~}|zyyyyxxzzz-- %zz{|~~~}}||{{z---V$)---V%)---$ ---% ---$ ---% ---$ ---% ---V$)---V%)--- ! ! !- !----%--(%---$ ---% ---- %}V--- !} !{ !y !s !q !p !o !~i !}g !|f !{e !x_ !w^ !v\ !u[ !rU !qT !pR !oQ !kJ !jH !iG !e@ !d? !c= !_6 !^5 !]3 !Y- !X+ !W)---- !V( !Y/ !Z1 ![3 !^; !_< !`? !cG !dH !eJ !hS !iT !jV !kX !n_ !oa !pd !sk !tm !uo !xw !yy !z{ !} !~ ! ! ! !--DDDODOLE Generated ObjectWPC$ v;3',,,,519V+'"!Pa c 6Times New Roman RegularcXXX @a.notwithalaryngeal""Pi c 6Times New Roman RegularcXXX @RootRootRoot"#Pr Dc 6Times New Roman RegularcXXX @ကV?i"'P c 6Times New Roman RegularcXXX @b.butwithasupralaryngeal"(Pc 6Times New Roman RegularcXXX @RootRootRoot")P!4c 6Times New Roman RegularcXXX @ကVpi"+uc 6Times New Roman RegularcXXX @Lab"1c 6Times New Roman RegularcXXX @[-bk][+hi]"5 ?9c 6Times New Roman RegularcXXX @[-bk][+hi]'12"8C"95rr.":C~h-WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawing$ࡱ;   Root Entry`.@pvOle 9WPG20 CompObj1^  !"#$%&'()*+,-./0256789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abc)F! WPWin61[5]ࡱWPC$ v;3',,,,519V+'"!Pa c 6Times New Roman RegularcXXX @a.notwithalaryngeal""Pi c 6Times New Roman RegularcXXX @RootRootRoot"#Pr Dc 6Times New Roman RegularcXXX @ကV?i"'P c 6Times New Roman RegularcXXX @b.butwithasupralaryngeal"(Pc 6Times New Roman RegularcXXX @RootRootRoot")P!4c 6Times New Roman RegularcXXX @ကVpi"+uc 6Times New Roman RegularcXXX @Lab"1c 6Times New Roman RegularcXXX @[-bk][+hi]"5 ?9c 6Times New Roman RegularcXXX @[-bk][+hi]'12"8C"95rr.":C~h- `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; ࡱ; h.    h.3f3333f333ff3fffff3f3f̙BOlePart3OlePres0004 3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--+2 #a. not with a laryngeal # -- .  --'2 R*Root Root Root   R-- .  --+2 C* V ? i C-- .  --32 $b. but with a supralaryngeal $-- .  --'2 SRoot Root Root   S-- .  --+2 E V p i E-- .  -- 2 yHLab ya-- .  --2 G [-bk] [+hi] -- .  --2 9 [-bk] [+hi] --%U>XxYUiWzz--WUlU--%K/ZbWKUTe-ࡱ; ࡱ;  METAFILEPICTh.1 h.   h.3f3333f333ff3fffff3f3f̙3f3333f3333333333f3333333f3f33ff3f3f3f3333f3333333f3̙333333f333ff3ffffff3f33f3ff3f3f3ffff3fffffffffff3fffffff3fff̙ffff3fffff???__ߟ߿3f3333f333ff3fffff3f3f̙3f3f3333f333ff3fffff̙̙3̙f̙̙̙3f̙3f3f3333f333ff3fffff3f3f̙3f45  Times New Roman-- .  "System--+2 #a. not with a laryngeal # -- .  --'2 R*Root Root Root   R-- .  --+2 C* V ? i C-- .  --32 $b. but with a supralaryngeal $-- .  --'2 SRoot Root Root   S-- .  --+2 E V p i E-- .  -- 2 yHLab ya-- .  --2 G [-bk] [+hi] -- .  --2 9 [-bk] [+hi] --%U>XxYUiWzz--WUlU--%K/ZbWKUTe-dengiMindy501 dTABLE ATABLE Ahowdie 7!XDXXX  C 6 $# C--   ݛ8:  o[X4 hp x (#%'0*,.8135@8:<H?ACXo@XXD[FromPapersinOptimalityTheory,UniversityofMassachusettsOccasionalPapers18,J.Beckman,S.Urbanczyk  andL.Walsh(eds.),GLSA,UMass,Amherst,1995.UMOPpagenumberingretained.]#XDX@Բ#  h XXD FeatureClasses #XDXh #*{!< ;   *      *7!       JayePadgett    UniversityofCalifornia,SantaCruz  1.Introduction      4 InworkonFeatureGeometry,classnodeshavethestatusofthesinequanon.Anexampleofthewellknownphonologicalstrategywherebyweseekrepresentationalsolutionstoproblems,classnodesliterallyembodytheunderlyinginsightoffeature 8 classes.Thispaperattemptsafreshlookatthisarea,andarguesforadisembodied $ alternative,calledFeatureClassTheory,inwhichsuchclassesareseenasnonrepresentationalpostulatesofthetheory.Muchofthegeometryisremovedfroma  theorywithoutclassnodes,andfeatureslikeLabial,Coronal,DorsalandPharyngeal,forexample,areunitedinsharingthepropertyof'placeness'ratherthanincohabiting X underanodePlace.Thatis,featureclasstheorypostulatesasetPlace,ofwhichthese D featuresaremembers.ConstraintscanmentionaclasslikePlace!hencetheimportant 0 ideaoffeatureclassesremains!buttheytherebytargettheindividualmembersofthatset. 4 Theoccasionalresultofsuchindividualfeaturetargeting,andacentralargumentforthetheory,isacertaintypeofpartialclassbehavior,firstbroachedbySagey H (1987)as'nonconstituent'behavior,andillustratedherewithanexampleinvolvingconsonantalplacefeatures.InG)a``0ԛ(NigerKordofanian),nasalconsonantsassimilate ! ! inplacetoafollowingconsonant.Withinmorphemes,theassimilationiscompletebeforecomplexsegmentsgb/kp,e.g.,mkpai'libation'.Acrossamorphemeboundary, |##  however,theapparentlyrecalcitrantLabialarticulatordeclinestospread,andremains h$ $  linkedonlytothesourcesegment,e.g.,N+kpaikpai'mycheeks'(Ryder1987,  Padgett1994,1995),givingthepartiallinkingscenarioin NONCONST(1)  . x[X4 hp x (#%'0*,.8135@8:<H?ACXX4 Xx2(  1  )3   >  4   PartialclassbehaviorinFeatureClassTheoryG)a``0Ԁnasalplaceassimilation  Л~[X4 N h*p x (#%'0*,.8135@8:<H?ACX4 X~ 4   d0 `H<p1 E! z! z d  PartialclassbehaviorlikethisisunexpectedinFeatureGeometryTheory(henceforthFG):sinceconstraintstargetclassnodes,featureclassbehaviorisanallornothing %  affair.WhenaruletargetsPlaceasshownbelow,eitherbothLabialandDorsalwill   spread,or(iftherulefails)neitheronewill.̜2(  2  )3  Ԁ  TotalclassbehaviorinFG   Л 4   d0 `H<p  E   *d   [X4 N h*p x (#%'0*,.8135@8:<H?AC X4 Xh*pX 4 Partialclassbehaviorasin NONCONST(1)  Հhasreceivedlittlementionbyphonologists.(TherathercommonfailuretoassimilatecompletelytocomplexsegmentsexemplifiedbyϜGa)a``0wasnotobservedinSagey's1986study,forexample;seePadgett1994,1995). 2" Thisisinpartduetofactualoversight,butthisisnotthemostinterestingreason.Rather,featuregeometrytheory,withitsincarnationoffeatureclassesasnodes,makesitdifficulttoseecertainlegitimategeneralizationsbecausetheyinvolvepartial  classbehavior.ThispaperwillfocusontheproposedclassColor,unitingthevowel  placefeatures[back]and[round](Odden1991,Selkirk1991a).Thisgroupingderivesagreatdealofsupportfrompatternsofvowelharmony,oncewecountenancetheexistenceofpartialclassbehaviorlikethatillustratedhere.Thus,Iargue,featureclasstheorybroadenstheempiricalandexplanatorydomainoffeatureclassestoareaswherefeaturegeometrycannotgo. 4 Crucialtoanunderstandingofpartialclassbehaviorarethenotionsofconstraintrankingandviolability,andtheworkheredrawsheavilyonOptimalityTheory(henceforthOT,PrinceandSmolensky1993).Acentralclaimhereisthatpartialclassbehaviorshouldbeunderstoodasthegradientviolabilityofconstraintstargetinga ' " featureclass,oftenresultingfromtheconstraintconfigurationC>> (" Constraint(Class),whereCisanyconstraint,Constraint(Class)isaconstraint (# targetinganyclassoffeatures,andthetwoconstraintsconflictforsomepropersubsetofthesetClass.IntheexampleinvolvingGa)a``0 above,theclassisPlace,whichis *%  assimilatedbynasals,andCisplausiblyaconstraintenforcingalignment(McCarthy +&!  andPrince1993c)offeaturesandmorphemeedges(seePadgett1994,1995).Inthe ,'" remainderofthispaper,however,thefocuscentersmostlyoncaseswhereCprohibitsfeaturecooccurrences. | ;   1      כ $   4 Withtheabandonmentofclassnodesasameansofcapturingfeatureclassbehavior,thequestionnaturallyariseswhetherclassnodes(quarepresentationalentity)haveservedanyotherpurposeinthetheory.Explicitargumentationtothiseffecthasbeensurprisinglyscarce,thoughitturnsoutthattheyhavebeeninvokedinaccountsofsegmenttransparencytofeaturespreading,andinexplanationsoffeatureinteractionmoregenerally.Herewewillexamineoneoftherelevantcases,involvingtranslaryngealharmonies,(seePadgett1994,1995,forothers),andseethatanumberofcriterialeadusawayfromanodebasedunderstandingoftransparency.Thealternativeexplanationpursuedhere,whichhaspotentiallyfarreachingimplicationsforthetheory,harnessesindependentlymotivatednotionsinvolvinglocalityandfeaturecooccurrence. 4 ThispapercanbeconstruedascontributingtoabroaderthemeemerginginworkwithinOT,thatofthereducedroleoffeaturalrepresentation.Inthelasttwentyyearsorsoofphonologicalhistory,arelianceonrepresentationbasedsolutionstoproblemshasbeenanoftnotedandproductivestrategy,andhasbeenresponsibleforawellknownshiftawayfromattentiontocomplexrulesandtheirinteractions.Whilethisstrategyisunlikelytodisappearfromthescene,onecanalreadyperceivepropertiesofOptimalityTheoreticthinkingthattendtoundermineittosomedegree.Onesuchpropertyinvolvesthedirectharnessingbygrammarsofmarkednessrelations,capturedbyuniversalconstraintrankingsalongsomedimension,e.g.,sonority,placeofarticulation(PrinceandSmolensky1993),apropertyresponsibleformuchoftheappealofthetheory.ThusPrinceandSmolensky(1993),Smolensky(1993)andMcCarthy(1994)havearguedthatthespecialpropertiesofcoronalsarearesultoftheirunmarkednessinthissense,andnotofanyrepresentationalpropertytheyevince(underspecification).AnotherpropertyofOTrelevanttothisthemeisthecentraloneofconstraintviolability,andofmostconcernhere,gradientviolability.Whilegradient  violabilityisnotinconsistentwiththenotion'representation'broadlyconstrued,itisneverthelesstruethatsomeproposedrepresentationalentitiesandtheirattendantpropertieshaveanallornothingcharacterthatmakesthematbestirrelevanttotheexplanationofpartialconstraintviolation(andatworstarealhindrancetoit),atleastastheyaretraditionallyconceived.Classnodesfallintothiscategory,aswewillsee.̀ 4 Section2exemplifiespartialclassbehaviorwithfactsofvowelharmony, h&0!$ motivatingtheclassColor,andintroducesFeatureClassTheory(henceforthFCT). T'"%  Section3articulatesthepropertiesofFCTmorefully,anddiscussesvarious @(#& implications.Insection4wetakeupanissueinvolvingsegmenttransparencyandassimilation.Section5concludesthepaper. ̛ 2. 4 Partialclassbehavior:vowelcolor    4 Webeginwithawellknowngeneralization:acrosslanguages,certainfeaturespatterntogetherrecurrentlyinphonologicalprocesses;thesefeaturesfallintoclassesthataretypicallycharacterizedwithreferencetocertainphoneticdimensions,e.g.,Place,Laryngeal.Giventhisrelativelytheoryneutralobservation,howshouldwe \ L  characterizetheseclassoffeaturesinourtheory?McCarthy(1988:87)articulatesthreepotentialapproaches:̜2(  3  )3    ԛ 4   Waystocapturefeatureclasses     4   a.0 OrganizetherelevantfeaturesunderabstractclassnodesPlace,Laryngeal,   etc.Constraintstargetthesenodes.(Clements1985).    4   b.0 naryfeaturesfortherelevantgroupings,e.g.,[nplace]etc.   4   c.0 Postulatetherelevantclassesnonrepresentationally.  ̜ shinto(3)  aisthecoreofFeatureGeometryTheory(FG),asolutionquitewellknownandwidelyadopted,duetoClements'(1985)influentialstudy. shinto(3)  bisdifficulttodistinguishempiricallyincertainrespectsfromboth shinto(3)  aand shinto(3)  c,andwewillnotpursueithere.(SeeMcCarthy1988).Thispaperwilldevelopanexplicitunderstandingof shinto(3)  ccalledFeatureClassTheory(FCT).Beforegivingitmorecontent,letusturntothefactsinvolvingvowelcolorthatwillbetheprimaryimpetusinthispaperforFCT. 2.1  Colorharmonies    4 InmanyTurkiclanguageswefindpatternsofharmonyinvolvingbothroundnessandbacknessinvowels.PerhapsthemostfamiliarexampleisthatofTurkish,thefactsofwhicharesummarizedhere. ;   2      כVowelsinTurkishwordsagreeinbacknesswitha "  voweltotheleft;iftheyarehigh,theyalsoagreeinroundness.Inmostaccounts,spreadingoccursrightwardfromtheinitialvoweloftheroot. shinto1(4)  aillustratesharmonyofbothfeatureswithexamplesinvolvingtargetvowelsinrootsandinsuffixes.Thedifferentbehaviorofnonhightargetvowels!theydonotacquireroundnessfromaprecedingvowel!isshownin shinto1(4)  b.  (#& М& L  2(  4  )3    ԛ 4   Turkishvowelharmony 8 T( X4 N h2X X4 Xh*pXT 4   a. [back]and[round]spreadfrominitialvowel   4    somun N   'loaf' h   iyi 2 'good'   4    sonun N   'end,gen.'  ipin 2 'rope,gen.'    4    kz N   'ox' h   ad5m 2 'step'    4    yzn N   'face,gen.'  sap5n 2 'stalk,gen.'  t 'L83 4   b. Only[back]spreadsto[high]vowel  L   4    kpek N   'dog' \$   4    yzden  'face,abl.' H   4    uzak N   'far' 4   4    sondan  'end,abl.'    ЛX" X4 Xh*pX X4 N h2X3XTheTurkishvowelinventoryisgivenbelow.Sincethepatternofvowelharmonyrequiresreferenceonlytotheheightdistinction[+high]versus[high],nomoredistinctionsareindicatedhere.̜2(  5  )3  ԛ 4   Turkishvowelinventory p HX4 XX X4 Xh*pXH 4     N   [back] V [+back] H  4    [+high]  i   V 5  u X   4    [high] N   e   V a  o 0  4     N   [rnd]  [+rnd] V [rnd]  [+rnd]  L" X4 Xh*pXX4 XX:L̛ 4 Thepointofinteresthereisthecoexistenceofback/roundharmonywithinasinglelanguage.ThisisnotanoddityofTurkish,butratherpervadestheTurkicgroup(seeforexampleKorn's1969surveyofapproximately15languages),anditisafeatureofAltaicandUralicmorebroadly(Hungarianbeingonewellknownexamplefromthelattergroup,Vago1980).Witheffortonecanfindcandidatecasesoutsideofthisgroup;thus,vanderHulstandSmith(1985)argueforthespreadingofbothfrontnessandlabialityintwolanguagesofAustralia,DjingiliandNyangumarda,andOdden(1991)(arguingfora[back]/[round]constituentinFG)motivatescasesfromseveralotherlanguagefamilies.Inverymanyofthesecases,roundnessharmonyisrestrictedinsomewayincontrasttobacknessharmony,apointtowhichwewillreturn.Thenotableconcentrationofback/roundharmonyamongtheUralAltaiclanguagesisacorollaryofthenearconfinementtothisgroupof[back]harmoniesingeneral.Thedivergingpatternsofback/roundharmonyfoundthere,whichdonotresembletherandomdisintegrationofacoincidentalparentalcopatterning,butratherreflectseveralplausiblenotionsofmarkednessinvowelandharmonysystems(seeKaun1994,andbelow),testifytotheindividualviabilityofthesystems.Itseemsclearthat .|)- otherimaginableharmoniesinvolvingarbitraryfeaturepairs,say[back]and[low],or[back]and[nasal],donotsharetheprivilegedstatusinthetheoryof[back]and[round].Itisthejobofphonologicaltheorytoprovideanaccountforthecopatterningof[back]and[round]. 4 Attentiontothisfeatureclass,defininga'horizontal'rangeofvowelqualitiesoftenopposedtothe'vertical'axisofheight,hasalonghistoryinphonologicaltheory.Trubetzkoy(1939)opposestimbre,andJakobsonandHalle(1956)tonality,tothe  t sonority/aperturedimension,motivatingthisbifurcationbasedonfactsofacoustics,inventorymarkednessandalternation.(Anexamplefrominventories:nonlowvowelstendtocontrastinthecolordimensionmorethanlowvowels,givingthefamiliarandcommontriangularityofvowelsystems).Relatedtothephonologicalinterdependenceof[back]and[round](onthisinterdependenceseeSchane1972,aprecursortothe    A/I/UparticleviewofSchane1984)isamuchdiscussedphoneticenhancement    relationbetweenthefeatures(Stevensetal.1986,StevensandKeyser1989).Selkirk(1991a)notesthesignificanceoftheharmonypatterns,andOdden(1991)notablyarguesforafeaturegeometricalnode[back]/[round]frommorerestricted(nonword  spanning)examplesofvowelplaceassimilation.EwenandvanderHulst(1987)andvanderHulst(1988)positagroupingoftheelementsIandUinavariantof  DependencyPhonologytheory(seeAndersonandEwen1987andreferencestherein),basedonconsiderationsofinventories,andvanderHulstandSmith(1987)andvanderHulst(1988)arguefortheroleofthisgroupinginaccountingfor[back]/[round]harmonies.Mester(1986)analyzesKirghiz[back]and[round]harmonyintermsofadependencyrelationbetweenthesefeatures,thusmakingofthemaphonologicalconstituentforeshadowingOdden's[back]/[round].LikeSelkirkandothers,Iwilluse 0  thetermColortodesignatethisfeatureclass.    4 InspiteofafairprecedentforaclassColor,virtuallyallresearchersinthe  generativetraditionaddressingTurkishvowelharmonyhaveassumedtwoseparaterules/constraintsofharmony,implicitlyorexplicitly:oneharmonyof[back],andoneof[round],astateofaffairsrenderingcolorharmonyaslikelyseemingasacopatterningof[back]and[nasal].ItisparticularlystrikingthatnooneworkingwithinFGhasmarshaledthefactsofTurkishorsimilarharmonies,where[round]spreadingisrestrictedincomparisonto[back],asevidenceforaColornodethoughitisnot |#l! difficulttoseewhy.FGembodieswhatmightbecalledasourgrapesunderstandingofassimilation:eitherallfeaturesmustspread,ornonewill.ConsiderthecaseofspreadingtoanonhighvowelinTurkish,asinsondan,  TURK(4)  babove.Assuming @&0!$ harmonyasColornodespreading,wewronglyspread[round]tothenonhighvowel, ,'"% whereitcannotoccur(seediscussionbelow),asshownin MINDY1(6)  a.IfColorspreading (#& failsforthisreason,thenwewronglyfailtospread[back]aswell,leavingthetargetvowelunderdeterminedforthisfeature, MINDY1(6)  b.Wecannothopetobesavedbyanyimagineddefault[back]valueingeneral,sincethesamevowelinyzden,forexample, *%) evincestheoppositevaluefor[back];[back]arrivesbyharmony.TheessentialpointisthattheFGrepresentationitselfmakesnoprovisionforanintermediatepossibility, thespreadingof[back]butnot[round],apartialclassspreading. -(, &   2(  6  )3   C  4   Sourgrapesspreading 8 Л[X4 " N z Dpx (#%'0*,.8135@8:<H?AC X4 Xh*pX 4   d0 `H<p  E$''\Pd'8Q $  4 ConsiderinthislightanamendmenttoFGtheory:totheconventionof MINDY2(7)  aweadd MINDY2(7)  b,a'lastresort'mechanismtoallowforcasesofpartialclassspreading(class   standsfor,e.g.,place,etc.):   2(  7  )3   D  4   a.0 Tospreadasetoffeaturesclass,spreadthenodeClass;[   Л 4    or,ifspreadingofClassfails, G   4   b.0 SpreadasmanyfeaturesdominatedbyClassaspossiblek3   Itseemsobviousthatonceweaddthisamendment,thestandard MINDY2(7)  aitselfisnotrequiredinthetheory:spreadingisalwaysofasmanyfeaturesaspossible.Further, /  theclassnodequarepresentationaldevicehasbeenseriouslyundermined;inatheoryofonly MINDY2(7)  b(compareHalle1993)theclassnodeisdemotedtothestatusofmerefeatureclasslabel.Belowitwillbeexcisedentirely.AsforFGproper, MINDY2(7)  aalone,itsinabilitytoilluminatepartialclassspreadingpresumablyexplainswhynoFGresearcherhasfingeredTurkishandsimilarharmoniesasevidencefortheexistenceofColor(seePadgett1994,1995,forcasesinvolvingtheclassPlaceinconsonants).  Ifthisfeatureclassisreal,thenbyobscuringitFGisfailingonitsownterms. 2.2FeatureClassTheory  {C  4 Ourproposedalternative,FeatureClassTheory,fullyexploitsthebasicprinciplesofOptimalityTheory(PrinceandSmolensky1993).OurcentralargumentisthatconstraintsmentioningfeatureclasseslikePlace,Color,etc.,aregradientlyviolable + inthesenseestablishedbyOT:partialclassspreadingisonegradientlyviolatingresult.InordertoteaseouttherelativerolesofOTandtherepresentationalsimplificationitself,itisworthwhiletocontinueinaderivationalistveinforthemoment,pursuingonlytherepresentationalissue. 4 Deferringamoreformaldiscussiontosection3,letusregard'placeness','laryngeality','color',etc.,simplyaspropertiesoffeatures.Inthewaythatanobject %g  maybeblue,ortall,afeaturemaybe'color'or'place'.E ;   3      ׀Topreservethecentralinsight &S!  ofFG,wemustallowrulestocapitalizeontheseproperties.ThereisnoColornode; w'?"! insteadtherelevantfeaturesaretargeteddirectlyandindividually,byvirtueofbeingcolorfeatures.Thus'SpreadColor'means'everyfeaturethatisacolorfeatureshould O)$# bespread'.Theeffectofsuchastatementisdepictedbelow(hereandbelow[back]and[round]arebandrrespectively).  &  2(  8  )3  Ԁ  SpreadColor,withoutnode   4    d0 `H<p E . X?d  '_ 4 Individualfeaturespreading,asaresultofarulementioningaclassoffeatures,isapossibilityraisedbySagey(1987)undertherubric'nonconstituentspreading',thoughSageyregardssuchspreadingasamarkedoption,adeviationfromtheusualFGscenarioofclassnodespreading.(Theterm'nonconstituentspreading'isinappropriateforus,sinceconstituencyitselfasthekeytofeatureclassesisforsaken).Selkirk(1991a,b)andPadgett(1991)alsoadvocateindividualfeaturetargeting,butastheonlymeansofclassbehavior(abandoningtheclassasnode),asarguedforhere.Thoughsuperficiallydistancedbyadifferenceofnotation,Hayes'(1990)understandingof'diphthongizationparadoxes'isessentiallyindividualfeaturebehavioraswell(seethediscussionofthisissueinGoldsmith1990:2958,wherethealternativeviewofclassmembershipaspropertiesoffeaturesisconsidered).Halle(1993)revivesSagey'sargument,fromspreadinginBarraGaelic,thoughNChiosin(toappear)providesamoreconvincinganalysisoftheBarrafactsthatrecaststheissue.Herewearerootingtheargumentinnewempiricalterritory,involvingvariouslanguagesshowingdiversemeansofspreadingColoronlypartially(withthefocuson B2 Turkish),andthuswefindrobustevidenceforagenuineclasswhereFGcannot.Aswewillsee,theideaacquiresavirtualconceptualnecessitywithinOptimalityTheory. 4 Considernowthecaseofsondanagain.Thefirststepinourprovisional  derivationspreadstheclassColor,asin MINDY4(9)  a.Thismoveestablishesanillicitlink  between[round]andthesecondvowel,anerrorrepairedin MINDY4(9)  b.Thissecondsteprequiresthedelinkingof[round]only,apossibilitythatwouldbeobscuredbythepresenceofanodeColormediatingtheconnection(see MINDY1(6)  Հabove,andcf.Hayes   1990on'diphthongizationparadoxes').WemightneverthelessretainFGandstilleffecttheneeded(moredrastic)relinkings,sincetheycansurelybeviewedascompelledbytheconditionforcingdelinking.Indeedthemostimportantpointaddressedheretranscendstherepresentationalissue:featureclassbehaviordisplaysgradientviolability,andseeingthisfactallowsustobroadentheexplanatoryscopeofthefeatureclassidea.However,itisathemehereandinsection4thatfeatureclassesquanodescontributelittletotheexplicationofthisidea,andinfactmakeit '"# difficulttoseecertainrelevantgeneralizations,asnotedearlier.̜& K 2(  9  )3   G ԛ 4   SpreadColorwithrepair )$& [X4 " N z Dpx (#%'0*,.8135@8:<H?AC X4,,xX 4   d0 `H<p` E*``)2d'K)k *%'  4 OurtransitiontoOptimalityTheorybeginswiththeacknowledgmentthatthereisnointermediaterepresentation*sondonatwhich'SpreadColor'hasbeenfully $ obeyed.Rather,partialclassspreadingispreciselythe(minimal)violationofthe  spreadingimperative.TheessentialaccountcapitalizesonthecentralpropertiesofOT(PrinceandSmolensky1993):[X4` " N z Dpx (#%'0*,.8135@8:<H?AC X4,,xX2 (  10  )3   H Ԁ ` CoreOTaccount    4  ` a.0 " Constraintrankingandviolability.[round]nevergetstospreadin MINDY4(9)  a,  ` becauseofthehigherrankingprohibitionagainst[high,+round].'SpreadColor'isviolated.p8 " "   4  ` b.0 " Minimal,gradientviolation.Failureof'SpreadColor'isnotsourgrapes H  failure:theconstraintisminimallyviolated([back]doesspread).4 " "   4 Theprohibitionreferredtoin MINDY5(10)  ainvolvesawellknownmarkednessgeneralizationnotedalready(Trubetzkoy1939):colorfeaturecontrastsaremorefavoredinhighervowels,lesssoinlowervowels,afactunderpinningtheunmarkedtriangularvowelsystem.Further,environmentsofvowelreductiontendtodisplaylessmarkedinventories.Haiman(1972)andvanderHulstandvandeWeijer(1991)notethatthereducedvowelinventoryfoundinnoninitialsyllablesofTurkish,shownin MINDY6(11)  ,shouldbeviewedaslessmarkedinthissense.2 (  11  )3  Ԁ ` Turkishnoninitialvowelsystem l4  4  `  i N 5u D   4  `  e "  N a  ̛Inpractice,thisreductionhassometimesbeenenforcedthroughastipulationonroundnessharmonyitself(e.g.,Lees1961,ClementsandSezer1982).Yettheindependentneedinthetheoryforamarkednessconstraintagainstnonhighroundvowels,*[high,+round],responsibleforpatternsinlanguagesinvolvingbothunderlyingandreductioninventories,arguesinsteadforfactoringtheconstraintoutoftheharmonygeneralizationaltogether,asadvocatedbyHaiman(1972).OTprovidesthetheoreticalcontextinwhichthisstrategycanbepursuedtothelimit.Thefactthattheeffectofsuchaconstraintisrestrictedtononinitialsyllables(notanaturaldomain)suggestsaspecialstatusforthefirstsyllableasapositionofprominenceinsomesense.Iassumehere,followingSteriade(1994),thattheinitialsyllablelicensesthefullrangeofvowelcontrasts,whilethisstatusisnotgrantedto ,)#' [high,+round]elsewhere(seeSteriadeonthisroleforthewordinitialpositionacrosslanguages,andIt=1986,Goldsmith1990,andIt=andMester1993onthenotion'license').ForconvenienceIwillrefertoaconstraintLico: +&* 2 (  12  )3  Ԁ ` Lico:Nonhighroundvowelsarelicensedonlywordinitially -(, М .|)- Ї 4 FollowingKirchner(1993),Smolensky(1993),Pulleyblank(1994)andothers,IviewharmonyastheresultofalignmentinOT,buildingonGeneralizedAlignmentTheory(McCarthyandPrince1993a,b,extendingideasofPrinceandSmolensky1993).Theconstraintin shinto2(13)  ՀtargetsthefeatureclassColor,thepointofinteresthere.  Thebidirectionalalignmenttoprosodicwordedgesisaconvenientlysimpleformulationforourpurposes,abstractingawayfromsomeissues. ;   4         2 (  13  )3    Ԁ ` Align:Align(Color,Edgex,Pwd,Edgex)(bidirectional)  t  4  ` 'Alignallcolorfeaturestobothedgesofaprosodicword' 4 AfulleranalysisofTurkishthanIcanprovidehererequiresgrapplingwiththeperennialdisharmonicrootsandsuffixes(e.g.,peron'railwayplatform',disharmonic 4$  forboth[back]and[round],mezat'auction',disharmonicfor[back],andthe    progressivesuffixiyor,whosesecondvoweldoesnotharmonize).Theissues    involvingdisharmony,patternsofdisharmony,underlyingforms,andthestatusofharmonywithinrootsarecomplexandwouldtakeustoofarafieldhere(forproposalsandinterestingdiscussionseeClementsandSezer1982,vanderHulstandvandeWeijer1991,Kirchner1993,Polgardi1994andInkelas,OrgunandZoll1994).Theseissuesdonotappeartointeractinanycrucialwaywithourclaimthat[back]and[round]formaclass:thiscopatterningisproductivelyevidentinmostsuffixedforms. 4 ThesourceofpartialclassspreadinginTurkishistherankingbetweenourtwoconstraints,Lico>>Align.Considerthefollowingconstrainttableau(capitals XH denotevowelsunspecifiedforcolorfeatures,A=[high],I=[+high]).Iassumefor D4 discussionthatcolorfeaturesassociatedwitharootareunderlyinglyunassociated,thoughtheymustbeparsedbylinkagetoasegment(ClementsandSezer1982).Fortypographiceasethelinksof[back]and[round]arenotationallymergedwherethesefeatureslinktothesamevowel,thoughtheyareinfactlinkingindividually.Candidate mindy8(14)  aevincescolorfeaturesfullyalignedtotherightandleftedgesoftheprosodicword,andsoisnotchargedforanyalignmentviolation.However,thelinkbetween[round]andthesecondvowelsegmentviolateshigherrankingLico.Sinceonlythis   candidateviolatesthisconstraint,thedecisionisbetween LMINDY8(14)  band LMINDY8(14)  c.Bothviolatealignment,buthereiswhereminimalviolation, MINDY5(10)  babove,becomessignificant.In MINDY8(14)  bbothfeaturesaremisaligned,thoughnothingforcesmisalignmentof[back]inTurkish.SatisfactionofLicorequiresonlythemisalignmentof[round], h$X" andminimalviolationpicks MINDY8(14)  c.Align(Color)isthusgradientlyviolable,and T%D # partialclassspreadingistheminimalviolationofsuchaconstraint.Putanotherway,thereasonwhyspreadingisonlypartialisthatLicoconflictswithAlignwith ,'"% respecttoonlyapropersubsetofthecolorfeatures(namely,one).̜ )#' 2 (  14  )3   L *qNO ddd Xdd Xdd X  q,Bdd ,td ,td +  ,"   ,  Input: %  p %sAndAǹr,b 0 !\  01'%p    1*qRS ddBdd td td NO  q,d ,d","+  1 "  T 1Candidate 9 T* "  9Lico 4 T!   T 4Align HT 9!      T sHa.sondoǹ\/r/b 9TT*X 3T  9*! -T T lTT - J 9!   s T T c Jb.sondAn|r/b ; T * 3  ; 4 T!  T  4**! J 9!  c  T Sx Jc.+sondan p8 |\/rb ;T *H3  ; 4 T!p8T  4*1'%\$ x S   T 1 4 Itisclearfromthetableauthatthereverseranking,Align>>Lico,picks @ candidate MINDY8(14)  ainsteadthemorefamiliarcase(popularizedbyFGtheory)oftotalclassspreading.SuchspreadinghasnospecialstatusinFCTnexttopartialclassspreading(andrestsonnoassumptionsaboutfeaturerepresentation,aswehaveseen),butmerelyrepresentsthecasewherenohigherrankingconstraintinterfereswithAlign(orwhateverconstraintinducesclassspreadinginaparticulargrammar # andconfiguration).FullColorspreadinginfactexistsinTurkishinformswithhigh $ noninitialvowels,asinsonun'end,gen.'HereLicoisnotviolatedbytotal |% alignment,sinceinitialsyllablesarelicensers,andfollowing[+high]vowelsarenot subjecttotheconstraint.Inthiscaseitisofcoursetotalalignmentthatisoptimal: T' &   2(  15  )3  *qUV ddd d""RS  q,Bdd ,td ,td +  ,"   ,  Input: %   %sAnIǹr,b 0 !p  01'%    1*qWX ddBdd td td UV  q,d ,d","+  1 " x h 1Candidate 9 T* "  9Lico 4 T!  T 4Align HT 9!      T sHa.+sonun    \/r/b 9TT*| l 3T  9 4T T!  TT 4 J 9!   s T T c Jb.sonIn|r/b ; T * 3  ; 4 T!  T  4*!* J 9!  c  T Sx Jc.son5n|\/rb ;T *4$3  ; 4 T!\L T  4*!1'%H8 x S   T 1'Ҕ 4 Summingupsofar:incontrasttoFGwithitsvirtuallyexclusiveconcernforcasesofuniformfeaturebehavior,FCTencodesfeatureclassesnonrepresentationally,andviewsaconstraintmentioningafeatureclassasgradientlyviolable.ThismovehasthegreatadvantageofallowingustoextendthefeatureclassideatoareaswhereFGhasfailedtoventure(seePadgett1995onavarietyofcasesinvolvingconsonantalcomplexsegments,foranextensionofthisidea).Beforecontinuinginourcharacterizationofthetheory,letusconsidersomegeneralissuesrelevantatthispoint. 4 Partialclassbehaviorisanotionalreadyimmanentinthetheory.Consideramoremundaneexample,thefailureofnasalplaceassimilationtovelarsinEnglishdisplayedbytheprefixin,e.g.,i mp lacable,i nd elibleversusi[ n ] c ompetent(seeBorowsky P @' 1986andreferencestherein).Y ;   5      ׀Thisisafamiliarsortofstructurepreservationeffect, @!0( involvingtheassertionofsegmentwellformednessoveranimperativeofassimilation(Kiparsky1985,ArchangeliandPulleyblank1986).Suchcasesarelegion,andawiderangeofworkinOTalreadycapitalizesontherelevantrankingsinthiscase,somethinglike*>>Nassim(thelatteracoverhereforsomeaccountofnasalplace $, assimilation).ThiscaseisidenticalintheessentialformalrespectstothatofTurkishcolorharmony:here,asinTurkish,aconstraintdemandingassimilationofsomeclassisviolatedwhenahigherrankingsegmentalmarkednessconstraintisatstake.Violationisminimal(obviouslynotinfectingcaseslikeindelible),andtheoverall (#0 resultispartialspreadingoftheclassPlace.WithTurkishwehavemerelyshiftedour )|$1 attentiontoacasewheremorethanonefeatureofaclassmayhappentooccupyone x*h%2 andthesamesegment,andsoshouldspreadjointlyfromasinglesource.WhileFG 8 ispresentedwithnodifficultiesbypartialclassspreadingacrossformsasinEnglish,Turkishhasbeenmoreelusive;yetsurelytheidealtheorywillrecognizewhattheseexampleshaveincommon. 4 Thisaccountchallengesthecommonstrategywherebyrulesarejudgedtobedifferentduetodifferentconditionsplacedonthem.Accordingtothisstrategy,[back]harmonyand[round]harmonyinTurkishareclearlydifferentrules(andaccordinglyhavebeentreatedassuch).Odden(1991)isaparticularlyrelevantworkinthisregard:inordertoestablishtheunityof[back]and[round]spreadinginEasternCheremis,forexample,Oddeniscarefultoarguethatthetwofeaturespreadingsaresubjecttoidenticalconditions.(Cf.Goad1993,whoisnotpersuadedofaneedforaclassColor,butwhosharestheassumptionaboutconditionsonrules).Odden'scase H  fortheclassColorisinfactmaderatherchallengingbythisstrategy,becausecolor 4  harmoniesveryoftendisplaymorerestrictedbehaviorbythefeature[round](seebelow).Totheextentthatconditionsonrulesaretrulyidiosyncratic(perhapsinthesenseeitherofphoneticallyunnatural,morphologicallyrestricted,orboth),thisstrategyforequatingordistinguishingrulesmaybequitevaluable.However,totheextentthattheymanifestindependentlymotivateddimensionsofwellformedness,astheyoftendo,ourrealimperative,inrecentphonologicaltheoryandmostclearlyinOptimalityTheory,istofactortheconditionsawayfromtherules(constraints)altogether.Arestrictionon[-high,+round]occurrencesissurelyofthelattercategory,andtheupshotofrecentphonologicaltheory,surprisingly,isthatwehavenobasisonwhichtodistinguish[back]and[round]spreadingconstraintsinTurkish.Infact,giventhecrosslinguisticcopatterningofthesefeaturesweponderedattheoutset,ourgoalmustbetounitethem. 4 Though[round]spreadingisoftenrestrictedincomparisonto[back]spreadingincolorharmonies,thediversepatternsofrestrictionfoundarefarfromidiosyncraticorrandom,butrathercanbeapproachedandunderstoodintermsofphonologicalwellformedness,oftenarguablygroundedinphoneticconstraints(infact,thespecialstatusof[round]over[back]inbeingsingledoutforrestrictioninitselfarguesforthisview,suggestingmarkednessissuesatwork,ratherthanrandomlyworkingconditions).Kaun(1994),buildingonKorn's(1969)surveyofTurkicandSteriade's(1981)earlierdiscussion,amongothers,notesthefollowinggeneralizations(Turkishisacaseof MINDY10(16)  a):2(  16  )3   Z Ԁ ` Roundingharmonycanfailif T'"%  4  ` a.0 " itwouldcreateanonhighroundvowel;or,)#'" "   4  ` b. " itwouldinvolvelinkingbetweenvowelsofdifferentheight;or *$(  4  ` c. " thetriggerishigh;or +%)  4  ` d. " thetriggerisback +&* KaunproposesasetofconstraintsintendedtocapturethesepatternsinanOTframework.Forourpurposes,theimportantpointisthatthepatternsofharmony .|)- evinceconditionsthatareplausiblyviewedasreflectinggeneralconstraintsonwellformedness,andsomustbeapproachedjustaswehaveapproached[back]and[round]harmonyinTurkish.FactoringouttherightwellformednessgeneralizationswillleaveuswiththebarepresenceofColorharmony.   3.Capturingfeatureclasses     4 Inordertoexplicatethestatusoffeatureclassesfurther,webeginwiththeFGrepresentationshownin MINDY11(17)  ,withclassnodesitalicized.ThisrepresentationbuildsonClements(1985),Sagey(1986)andMcCarthy(1988)initsincorporationofPlace H8  andLaryngealclasses,withtheformerdominatingprivativearticularfeatures(onthe 4$  laryngealfeaturesshown,seeLombardi1991).SincethefocushereislargelyontheclassColor,andinthefollowingsectiononthebehaviorofplacefeaturesmore    generally,Iwillfocusmoreonthisareaofthegeometry. MINDY11(17)  ՀincorporatesMcCarthy's(1991)proposalforaseparationofconsonantalplacefeaturesintoOral   andPharyngeal,andfurtherassumesaclassVPlace(seeSteriade1987,Prince1987,  Clements1991,NChiosin1994,Odden1991,Hume1992,andNChiosinandPadgett1993,amongothers).OpposedtoourclassColoristhatofHeight(see  Hyman1988,Clements1991,Odden1991,Selkirk1991a,Goad1993andothers).Thisgeometryremainsunderdeterminedinvariousrespects(notindicatingthelocationof[ATR]feature(s),otherfeaturessubsumedbyPharyngeal,andvarious l\ mannerfeatures,forexample),andassumesfordiscussionthetraditionalvowelplacefeatures[high],[back],etc.(Whereprivativity,andaunifiedfeaturestance,mightberelevantbelow,theyareexplicitlyaddressed).However,thislevelofdetailissufficientforourpurposes.& c 2(  17  )3   [ 0 ` Featuregeometryrepresentation` `   4 d0 `H<pf E4ZxZxld'ca   4 Inhiswidelyinfluentialstudy,Clements(1985)rootstheexplanationoffeatureclassbehaviorintwohypotheses:first,phonologicalrulesmayrefertoanynodein C.3) therepresentation,includingtheorganizationalclassnodes.Second,rulestargetsinglenodes.Withoutthislatterrequirement,theexplanatorybasisoffeature $ geometrywoulddisappear:ifrulescouldspreadmorethanonenodecostfree,thenwewouldpredictnopatterns.Instead,ifrulesarelimitedtosinglenodes,agroupoffeaturescanbehavetogetheronlybyvirtueofthedominantclassnode.Suchrulesincludetheexampleformulationsshownbelow('Root'isusedfor[son],forthesakeoffamiliarity):[X4` " N Dpx (#%'0*,.8135@8:<H?AC X4` ,,xX2(  18  )3  Ԁ ` Rulestargetingclassnodes  `  4  ` d0 `H<p: E` :: 4d  L   4 FeatureClassTheoryapproachesthenotionoffeatureclassesfromamore'semantic'andless'syntactic'pointofview.Whileclassesareequallyapartofthetheoryofphonologyhere,theyarenotincarnatedasnodes,butinsteadacquirestatusassetsoffeaturespostulatedinthetheory.WithFG,FCTassumesthatthese   postulatedclassesarefixedanduniversal.Assumingthevalidityoftheclassesrepresentedin MINDY11(17)  ,wethereforehavethefollowing:\ ;   6        & 8 2(  19  )3  Ԁ ` Featureclassesassetsoffeatures c  4  ` Laryngeal:  {voice,asp,glo}   s;  4  ` Place: N   {Lab,Cor,Dors,Phar,ant,dist,hi,lo,back,round...} _'  4  ` Pharyngeal:  {Phar,...} K  4  ` Oral: N   {Lab,Cor,Dors,ant,dist,hi,lo,back,round...} 7  4  ` VPlace: N   {hi,lo,back,round...} #  4  ` Height N   {hi,lo}   4  ` Color: N   {back,round}  '8 ?Inthecontextofrecentphonologicalhistorywithitsemphasisonrepresentationbasedexplanation,thenotionoffeatureclassesaspostulatedsetscanatfirstbedisconcerting.Theclaimcanperhapsmosteasilybeunderstoodtomeanthatfeatureshavevariouspropertiesattributedtothem,namely,'placeness','laryngeality','color',andsoforth(seefootnoteMINDYXX3  ),justastheproperty'red'canbeunderstoodbysemanticistsasthesetofallredobjects.Naturally,ifthismoveistoretainthebasicinsightofFG,thenconstraintsmustbeallowedtorefertotheseproperties,andhencetotherelevantfeatures,amatterweturntomomentarily. 4 Withclassesnowdisembodied,ourrepresentationbecomesmuchflatter,shownbelow.(Toemphasizethepointthatclasshoodhasnographtheoreticimplications,  *$% thefeaturesarescatteredrandomly).Infact,thisstructureisaclosecousintothe'bottlebrush'theory(seeHayes1990onthistypeofstructure),thoughFCTincorporatesthefeatureclassinsightofFG,unlikethattheory.&  2(  20  )3    Ԁ ` FCTrepresentation  j6&"` \p E @-L j   ' ^Thehierarchyremainingisthatrelatedtotruefeaturaldependencies(Mester1986, gW  Selkirk1988),thatis,dependenciesamongfeaturesthatbearrealphoneticcontent.Inotherwords, MINDY12(20)  ՀistheminimaldeparturefromtheFGrepresentationthatisconsistentwithFCT,andassumesthattotalassimilation(via[son]=Root),andthespreadingof[ant]/[dist]whenCoronalspreads,remainrepresentationalissues.Itmaybethatthisassumptiondeservesfurtherscrutinyaswell,butforreasonsoutsidethescopeofthispaper.^ ;   7          4 ThetaskistodesignFCTsothatclassesmaybepickedoutjustastheyareinFG.TheintuitiontobecapturedisthataconstraintlikeAlign(Color,Pwd,R)requires  ofeverymemberofthesetColorthatitbealignedwiththerightedgeoftheprosodic  word.Thatis,inFCT,constraintscanmentionfeatureclasses,buttheytherebytarget { therelevantindividualfeatures.Theconstraintrepeatedin MINDY 13(21)  awillthereforetargettheconfigurationin MINDY 13(21)  bcontaining[back],ortheonein MINDY 13(21)  ccontaining[round].Wemustalsoensurethatthisconstrainttargetsanynumberofcolorfeaturesinthe O? rightconfiguration.  '! [X4` " N z px (#%'0*,.8135@8:<H?AC X4` ,,xX&   2(  21  )3   _ Ԁ ` a. " Align(Color,Pwd,R)willtarget 8  4  ` b. " [..........]Pwd  c.  [..........]Pwd  Ѐ 4  `   "  N  z      p    4  `   " bk z     rd  [X4` " N z Dpx (#%'0*,.8135@8:<H?AC X4` ,pxX'8= 4 Therearetwogeneralapproachestothistaskthatonemightentertain,heredubbedthe'variable'approachandthe'metarule'approach.` ;   8      ׀Thevariable  t understandingispresentedfirst.[X4 " N z Dpx (#%'0*,.8135@8:<H?AC X4` ,,xX& d 2(  22  )3   a 0  Variableapproachtofeatureclasstargetingp8    0 4 04 4 LetClassdenoteasetoffeatures{F,G,H...},xrangeoverfeatures,Rbeany H  representation,andConstraintbeanyphonologicalconstraint.4     4 0  Then :0"   Constraint(Class)istrueofRiffforeveryxClass,Constraint    istrueofRifwelettheinterpretationofClassbex." "  'dLAppliedtoouralignmentexample,thisdefinitionmeansthefollowing:Align(Color)  istrueofarepresentationifandonlyifalignmentof[back]istruethere,andalignmentof[round]isalsotruethere.Sincealignmentistriviallysatisfiedbyaconfigurationinwhichthetargetedfeatureisabsentfromtherepresentation,theeffectissimplythatanyinstancesof[back]or[round]presentwillhavetobealigned.Asitstands,though,thisformulationisinappropriatelyabsolutealignmentiseithersatisfiedornot,andpartialclassalignmentwillberegardedastotalfailure.TobringthedefinitionmoreinlinewithOT'snotionofgradientviolability,werequiresomethingliketheidea'Ntrue'truthtosomedegreeN.Thedefinitionisaccordingly 4 modifiedasshownbelow.Inordertomaketheconnectiontonumberofviolationsmosttransparent,'Ntrue'isrecastas'Nviolated'.b ;   9         2(  23  )3  ԀVariableapproach:amendedconsequent 4   Then :0 " Constraint(Class)isNviolatedforRiffforNelementsxClass, "  ConstraintisviolatedforRifwelettheinterpretationofClassbex.#t!" "   $`"  4 Itmightbehelpfultomentionafewconfigurationsconcretelyandratethemwithrespecttothealignmentconstraint.First,wordswithnocolorfeaturesspecifiedwillsatisfyAligncompletely;thatis,forboth[back]and[round],alignmentistrivially  satisfied.Intheconfiguration[......b]Pwd([back]alignedfullyrightward),theconstraint  islikewisefullyobeyed.In[...r...]Pwd([round]anchoredtoanonfinalsyllable)Align  isviolatedby[round],thoughtriviallysatisfiedby[back].In[...r..b]Pwd[back]satisfies   alignmentinthemoreinterestingsense,though[round]doesnot,andwehavepartialclassalignment(seeTurkishtableauxinsection2.2). 4 AninterestingpropertyofthevariableapproachisitsrelianceonaunitaryformulationoftherelevantconstraintthereissimplyoneconstraintAlign(Color), H8  anditseffectistomakedemandsofseveralfeaturetypes.Wecancontrastthistothesecondapproachtofeatureclasstargeting,involvingmetarules'metaconstraints'wouldbemoreapthere,butthenameandideaarelooselyinspiredbyworkinGeneralizedPhraseStructureGrammar(Gazdaretal.1985).Theideabehindtheformulationisthis:givenametaconstraintmentioningaclass,e.g.,Color,the   grammargeneratesanumberofsubconstraints,oneforeachmemberoftherelevantclass,hereeachcolorfeature.Thisantecedentconsequentrelationisindicatedhere:2(  24  )3    Metaruleapproachtofeatureclasstargeting   4   Wherea/bmeanssubstituteaforb, l\  4    Constraint(Class),xClass D4 ЀConstraint(x/Class) 0  Therefore,frome.g.,Align(Color)wederivetwoconstraints:Align(back)and  Align(round).Theroleofthemetaconstraintinthetheoryissimplyoneof  characterizinganumberofrelatedconstraints,ineffect.Itisthesesubconstraintsthatareactuallyarrayedinthehierarchyofconstraintsthatconstitutesthegrammarofalanguage,andsoitisthesubconstraintsthatdotherealwork.Ineffect,themetaruleapproachcreatesafamilyofconstraints(seePrinceandSmolensky1993onafamily ! ofsonorityconstraints,andIt=,MesterandPadgetttoappearononeinvolvingantiinteractionconstraints).Thisapproachcanobviouslyleadtoagrammarwhereboth[back]and[round]arealigned,butitdifferssignificantlyfromthevariableapproach,inthatthesubconstraintsneednotberankedequivalentlyinthegrammar.ShouldsomegeneralconstraintdisfavoringalignmentcomebetweenAlign(back)and @&0!$ Align(round),theeffectsoftheseconstraintswillbesharplydistinguished.In ,'"% contrast,aunitaryconstraintAlign(Color)withvariableinterpretationcanleadto (#& distinctbehavioronthepartof[back]versus[round]onlyifsome(higherranking)constraintinthegrammarconflictswiththealignmentofoneofthesefeaturesspecifically.TheconstraintLicoplayedthisroleinthepreviousaccountofTurkish, *%) restrictingalignmentof[round]. 4 Thequestionofwhichapproachtoadoptisinpartaquestionaboutwheretheburdenofexplanationshouldbeplacedwhenfeaturesofaclassdonotbehave .|)- identically:isitbecausethegrammarcontainsdifferentalignmentsubconstraintsfortherelevantfeatures(whoseunityresidesonlyinthemetaconstraint),withsomegeneralconstraintdisfavoringalignmentintervening?Orisitbecausetheconstraintsdisfavoringalignmentarefeaturespecific,andanyoneofthemcanleadtothegradientviolationofaunitaryalignmentconstraint?c ;   10      ׀Inmuchoftheforegoing  (includingthediscussionofTurkish)theassumptionwasthataunitaryconstraintexists.Thoughtherelevantissuesaresubtle,thereisreasontomaintainthisunitaryvariableunderstandingoffeatureclasses. 4 OptimalityTheorymakesthestrongclaimthatgrammarsaresimplyrankingsofuniversalconstraints,animplicationbeingthatanygrammarcontainsallconstraints.Inthislight,considerthescenarioinwhichFCTallowsconstraintstomentioneitherasingleclass,asin MINDY15(25)  a,orsinglefeatures,asin MINDY15(25)  bandc.Thisscenariomightbeviewedasanullhypothesis,sinceitcarriesthispropertyoverfromFG.[X4` hpx (#%'0*,.8135@8:<H?AC X4,,xX2(  25  )3   d Ԁ ` Generalandspecificconstraints     4  ` a. Align(Color) h   b.  Align(back)   4  `      h   c.  Align(round)  Yetifallgrammarscarryallconstraints,isitobviousthatwerequirealloftheseformulations?e ;   11      ׀Ifnot,thereareinprincipletwopossiblemeansofreduction: \ eliminatethegeneral,oreliminatethespecific.FollowingasuggestionofPrince(p.c.),letuspursueeliminationofthespecific.Itishelpfulfirsttoconsidertheoppositesupposition:retentionofthespecificalone. 4 Thereasonthatthetheorycannotmakedowith MINDY15(25)  bcaloneisthattheirmerepresenceingrammarsinitselfcannotguaranteefeatureclassgeneralizations.Toseethis,considerjustthefollowing:alongwithconstraintsaligningcolorfeatures,wepresumablywouldhaveotherstargetingheightfeatures,i.e.,Align(high)and  Align(low)(nottomentionconstraintsforotherfeaturespotentiallysubjectto   alignment).Givenotherconstraintsthatcandominateoneormorealignmentconstraints,forcingthelattertobeviolated,wepredictvariouspossibleharmonysystems.Assumefordiscussionthatalignmentispreventedbyahigherrankingconstraintpenalizingassociationlinesnotpresentintheinput,termedFillLinkby $X" It=,MesterandPadgetttoappear(oralternativelyaconstraintagainstmultiplelinkingitself,NoLink,op.cit.).Giventhepossibilitiesofcrosslinguisticreranking, h&0!$ justtwoofthepredictedgrammarsareshownbelow: @(#& &  2(  26  )3   f Ԁ ` Freelyrankedspecificconstraints   4 a. ` Align(back),Align(round)>>FillLink>>Align(high),Align(low)   4 b. ` Align(back),Align(high)>>FillLink>>Align(round),Align(low)  'Theproblemisthatgrammarscannowachievecolorharmony,asin MINDY16(26)  a,buttheyjustaseasilyachieveanyrandomharmonypattern,includingtheonein MINDY16(26)  b.BarringgrammarextrinsicconstraintsdemandingthatcertainconstraintssayAlign(back) p ` andAlign(round)󀄀remainnearenoughtoeachotherontheconstrainthierarchy \ L  (ifwecouldevenpindownwhatthismeans),thistheoryactuallyhasnocharacterizationofthereal,recurrentfeatureclasspatterningsinlanguages.Thesamepointcouldbemadeforallfeatureclasses.Whateverourgeneraltheoryofharmonies,assimilations,neutralizations,etc.,turnsouttobe,itwillhavetosomehowsingleout,    e.g.,consonantplacefeatures,colorfeatures,laryngealfeatures,asclassesevincingspecialandunitarybehavior. 4 Thisbringsusbacktothemetaruleunderstandingoffeatureclasses.Thoughitprovidesbydecreeaunitarycharacterizationofclassesintermsoffamilymembership,thereisnothinginthisnotionthatpreventsthesamerandomrankingpatternsingrammarsthemselves.ConstraintfamilyprecedentslikePrinceandSmolensky's(1993)sonorityconstraintfamilies,andtheNoLinkfamilyofIt=,Mester l\ andPadgett(toappear),imposeanintrinsicanduniversalorderingontherelevantconstraints,havingapresumedbasisinphoneticconsiderations.Ontheotherhand,itisnotobviousthattherearemarkednessimplicationsinvolvedhere(roundnessharmonycouldbesaidtoimplybacknessharmonyinlanguagesonlyifcasesofapparent[round]harmonyalonecouldbeunderstoodasinvolving[back])q ;   12      ,andthe  intrinsicorderingofconstraintswithinafamilywouldinanycasefailtopreventthescenarioin MINDY16(26)  b. 4 Theintriguingalternativeistopursuethevariableapproach,involvingunitaryconstraintformulationslikeAlign(Color)tocaptureclassbehavior,andtocastit ! inthestrongestpossiblesense:thereinfactarenoconstraintsAlign(back), "  Align(round)availabletothetheory,norconstraintsaligninganysinglefeature. |#l! Moregenerally,suchconstraintstargetclassesonly.Beforefurtherconsideringsuch h$X" amove,itisworthwhiletoturntosomerelevantfacts.  ,'"%  &    4.Transparency,interactionandsegmentmarkedness  8  4 Recallthedifferentpossibleapproachestofeatureclassgeneralizationsnotedby '8McCarthy(1988):FGisoneofthem,andFCTisadevelopmentofanother,thepostulationofclassesinanonrepresentationalway.Ifbothtypesoftheorycapturethebasicfeatureclassidea,whyhastherepresentationalonelargelybeentheparadigmpursued?ThelikelyansweristhatFGfollowsastrategywellknowntomodernphonologicaltheory:positrepresentationalelements,andinprincipleinterestingunexpectedthingsmayfollowfromthem.FCTmayruncountertothisstrategy,buttwopointsareworthnoteinthislight.First,FCTfollowsanotherstrategythatprovesequallyfruitful:positrankingsamongsimpleconstraints,andagaininterestingunexpectedthingsmayfollow(PrinceandSmolensky1993).Theprecedingsectionsmakethispoint,showingthatFCTactuallybroadenstheexplanatoryscopeofthefeatureclassidea,illuminatingpartialclassbehavior.Second,classnodes,apartfromtheircentralroleincapturingfeatureclassbehavior,haveinfactengenderedveryfewresultsofinterest.Thatis,oncewetakeawaytheroleforclassnodesofexplainingfeatureclasses,thereseemstobelittleleftforthemtodo.However,asurveyoftheliteraturerevealsafewlimitedcaseswhereclassnodeshavebeenarguedtoplayafurtherrole;thesecasesinvolveissuesoftransparencyversusblocking,andmoregenerallytheissueoffeatureinteraction.Padgett(1994,1995)discussesonegeneralcaserelatingtoconsonantandvowelplaceinteraction,andrecaststheissueintermsofanaccountappealingtosegmentsimilarity.Hereweexamineadifferentsortofcase,involvingthetransparencyoflaryngealsegmentstovowelplaceassimilation.Thoughunsolvedissueswillremain,theinitialresultsheregoaconsiderabledistancetowardsreconcilingthesecasestotheFCTperspective,andraiseintriguingquestionsoftheirown. 4 PerhapsthebestknownandmostinterestinguseofclassnodestoachieveresultsoutsideoffeatureclassbehaviorisduetoSteriade(1987),inadiscussionofwhatshecalls'translaryngealharmony'.Toseethesortoffactsintended,considerthedatabelowfromKashayaPomo(Buckley1994).Withinamorpheme,anysequencesV?V   andVhVmustexhibitidenticalvowels(C'isglottalized): ! 2(  27  )3   g 0 ` TranslaryngealharmonyinKashayaPomo#l!` `   4  ` a. s'i?i  'flesh' h     nihn p 'tooneself' |%D #  4  `  he?n  'how' h     behe p 'baynut' h&0!$  4  `  ma?a  'food;eat'    ?aha p 'mouth' T'"%  4  `  q'o?o  'dance;song;sing'  soh;y p 'seal' @(#&  4  `  hu?l  'awhileago'    yuhu p 'pinole' ,)#'  4  `  cf.*se?o,*q'u?i,etc.    cf.*yeha,*?ohu,etc. +%) Thisconstraintdoesnotholdacrosssupralaryngeallyarticulatedconsonants(Tisalveolar): .|)- & ` 2(  28  )3   h 0 ` Noassimilationacrosssupralaryngealconsonants` `   4  `  ?imo  'hole'   4  `  hoya  'scoringsticks'   4  `  baco:  'boat'   4  `  buTaq  'bear'    4  `  phaT'e:  'burr'   '` 4 FollowingSteriade,letususetheterm'translaryngealharmony'tomeannotjustassimilationacrosslaryngeals MINDY14(27)  ,butassimilationacrosslaryngealsonly DAMNFOOL(28)  .The \ L  goal,asSteriadeargues,istoprovideaprincipledaccountforthispatternonethatdoesnotstipulatearbitrarilythatonlylaryngealsmayintervene.Steriadeproposestorelatethiseffecttoadifferenceinrepresentationbetweenlaryngealsandotherconsonants:sincelaryngealslackplaceofarticulationfeatures(Clements1985,McCarthy1988),theylackanodePlace,incontrasttosupralaryngeals.Iftotalvowel   placeassimilationinvolvesspreadingofthisnode,thenonlylaryngealswillbetransparentspreadingasin  dengi(29)  bwouldentailalinecrossingviolation.[X4` z h (#%'0*,.8135@8:<H?AC X4,XpxX&  2(  29  )3    0 ` Placeasablocker` `   4  ` d0 `H<pz E` zz ; d'   AsSteriadenotes,thisunderstandingoftranslaryngealharmonyhastheimportantvirtueofsimultaneouslyexplaininganotherinterestingfactaboutsuchharmonies:theyareneverofasinglefeature,butalwaysinvolvetotalvowelplaceassimilation. ! Sincetranslaryngealharmony(asdefinedabove)canoccuronlyifspreadinginvolvesthenodePlace,totalassimilationisexactlywhatshouldoccur.i ;   13      ׀Initsroleasa " blockerofassimilation,Placeachievesagenuinestatushereasrepresentationalentity, # independentofitsroleinexplainingfeatureclassbehavior. 4 ThequestionforFCTiswhetherthereisanotheraccountavailableforthesepropertiesoftranslaryngealharmony.ItshouldbeclearthatthereisnonodeasblockerexplanationinFCT.Inneither MINDY17(30)  anor MINDY17(30)  bisthereanylinecrossingissue.Thispointisindependentofwhatvowelfeaturesspread,andindependentofthe b)R$ choicewemakeofwhatvowelplacefeaturestoassume.Thus,evenifspreadingwerefromu,andassumingthefeatureLabialtocharacterizeroundnessinvowels $ (Clements1991,Selkirk1991a,1993,cf.Schane1984,AndersonandEwen1987,amongothers),wecannotcountoninterferencefromtheconsonantp,becauselabials  aremosttypicallytransparenttoroundnessharmony,andsomustbetreatedassuchsomehowbyanytheory.[X4` z h (#%'0*,.8135@8:<H?AC X4,z h, X&  2(  30  )3   j 0 ` Nolinecrossingissue t` `   4  ` d0 `H<pz E` AA m d'  $  `  4 TheanswertothispuzzleisalsosuggestedbySteriade(1987),thoughnotadopted:assumethatlinkageissubjecttoaconstraintrequiringstrictlocality,called `(  NoGaphere(Kiparsky1981,Levergood1984,ArchangeliandPulleyblank1994a, L  amongothers,andcf.Smolensky1993on*Embed;seeNChiosinandPadgett 8  1993andIt=,MesterandPadgetttoappearforrecentapplications).2(  31  )3   ` NoGap  XģXXXD 4  ` * x Ѐ \/ x@ ЀF#XDXXXģu(# z  h whereisalegitimateanchorforF @ Ananchorforanyfeatureislegitimateonlyifitrespectstheintegrityofthefeature   hierarchy.Inourterms,thismeansthat[sonorant]isalegitimateanchorforanyfeature,except[anterior]and[distributed](onlyCoronalprovidesaplaceforthesetwofeatures).Theimportantresulthereisthatmultiplelinkageofvowelplacefeaturescannotskipanysegment.ThisinterpretationofNoGapisthereforemore # demandingthanthatofArchangeliandPulleyblank(1994a),forexample,whotakeonlymoraicsegmentstobevowelplaceanchors. 4 ThestrategyhereistofollowMcCarthy(1994)inunderstandingeffectslikethatofKashayaascasesnotofsegmenttransparencytolinkage,butratherasinstancesofsegmentparticipationinlinkage.McCarthyexploresindetailthefactsofcoronal P)$ sonoranttransparencytototalvowelplacelinkageinadialectofBedouinArabic.PursuingideasofPrinceandSmolensky(1993),andSmolensky(1993),heassumesauniversalmarkednessrankingamongconsonantalplacefeaturesthatplacesCoronalplaceatthebottom,i.e.,*Dors,*Lab>>*Cor.McCarthyalsocallsonauniversal -'# harmonicorderingfavoringsimplexconsonantsovercomplexconsonants,adoptedhereintheformbelow: .)% Ї2(  32  )3   l Ԁ ` SimplexwComplex h wherecomplex=morethanoneplacefeature  Puttingthesetwoscalestogether,andgivingpriorityto MINDY20(32)  ,theconstraintrankingshownbelowisderived,ignoringDorsal,whichforourpurposescanbesubstitutedforLabial(Xisanyplacefeature,vocalicorconsonantal):  2 (  33  )3   m Ԁ0 ` *[Lab,X]>>*[Cor,X]>>*Lab>>*Cor ` `   4 Fromthisperspective,coronaltransparencyeffectsarenotduetoanyunderspecificationofCoronalplace(paceParadisandPrunet1989).Insteadtheyareinadirectsensetheresultofcoronalunmarkednessitself.AsMcCarthyshows,coronaltransparencyfollowsinthescenarioin MINDY21(34)  ,whereCisaconstraintwhose 4$  effectistosolicitlinkagebetweenvowels(whichhappentobeseparatedbyaconsonant).Aswewillseebelow,thisreasoningdependsonthehighrankofNoGap,    whichrequiresthatthemultiplelinkingaffectconsonantsaswell,potentiallycreatingtherelevantcomplexsegments.Segment'transparency'isthusamatteroffeaturecooccurrenceinageneralsense.&  2!(  34  )3   n Ԁ ` Coronaltransparencyfromcoronalunmarkedness   4  ` *[Lab,X]>>C>>*[Cor,X] p '4 4 Ofparticularrelevancetousistheplacelessstatusoflaryngeals.Wearenowinapositiontocapitalizeonthisstatus,asSteriadedoes,inordertogiveaprincipledaccountoflaryngealtransparency.However,theissueisnowreformulatedasoneofsegmentmarkedness,anotionobviouslyquitegeneralandwideranginginthetheory,ratherthanasamatterofclassnodeblockage.ExtendingMcCarthy'sideas,theimportantresultisthatlaryngealsegmentsdonotevenappearonthescalein MINDY22(33)  ,  ortherankingconfiguration MINDY21(34)  ,becauselaryngealshavenoplace.Theywillthereforebetheideallinkingpartnerforvowels.o ;   14      ׀Assumingalignmentastherelevant  harmonyconstraint,andhigherrankingNoGap,thenlaryngealtransparency,and   supralaryngealblockage,followfromtherankinggivenbelow,aswewillsee:2"(  35  )3   p Ԁ ` Laryngealtransparencyasamarkednesseffect |#l!  4  ` *[Lab,X]>>*[Cor,X]>>Align T%D # Giventhisranking,harmonybyalignmentcannotresultintheformationofcomplexsegments,consonantsspecifiedforextraplacefeatures.WhenNoGapishigh (#& ranking,thesefeaturecooccurrencestatementsareinevitablyatstake,asthefollowingcomparisonshows: *%) [X4` z h (#%'0*,.8135@8:<H?AC X4,z h, X&  2#(  36  )3   r 0 ` Strictlylocallinkagecancreatecomplexsegments? ;   15      8` `   4  ` d0 `H<pz E` $\{o d'8< $ Giventherankingin MINDY23(35)  ,theformin MINDY24(36)  bcannotbeoptimal(seebelow),norcananyformwithaninterveningsupralaryngealconsonant.Thereisnomarkednessprobleminthecaseof MINDY24(36)  a,however,andthegeneralpointisthatvowelplaceassimilationacrossalaryngealcanneverfailforreasonsofplacecomplexity.Ofcourseitcanfailifotherconstraintsaremilitatingagainstlinkagefaithfulnessconstraints,forexample,orantilinkageconstraintssuchasNoLink(It=,Mesterand f Padgetttoappear).s ;   16       R  4 Inordertoillustratethereasoningwithconstrainttableaux,letusfirstcollapseconstraintsoftheform*[CPlace,X]intooneblockforconvenience,called N  *Compseg,anddefineitasfollows: :  2$(  37  )3  Ԁ ` *Compseg: z Consonantsmaynothave>1placespecification  By'consonant'wemeanasegmentwithanyconsonantalconstrictionintheoralcavity.Theunderstandinghereexcludeslaryngeals(whetherornottheyare[+consonantal]themselves),becausetheylackanyoralconsonantalconstrictionandsoarenotrelevanttoanyconstraint*[CPlace,X].Theconstraintsmakingup v *Compsegruleoutsegmentslikegb,ty,kw,etc.,whileallowingsimplek,p,etc.,as b wellasanygroupingofvocalicconstrictions,e.g.,w,o(assumingthesevocoidshave  N complexplacespecifications).Forthesakeofdiscussion,letusalsoassumearightwardspreadingtranslaryngealharmonydeterminedbyalignment: J# 2%(  38  )3  Ԁ ` AlignR: z AlignR(VPlace,Pwd)  Therankingsthatareofchiefconcernherearegivenbelow.2&(  39  )3  Ԁ ` Translaryngealharmonyrankings   4  ` a. *Compseg>>AlignR    4  ` b. NoGap>>AlignR  t XDXXXD 4 ThefirsttableauconsiderstheKashayainput/#XDXXXD"H#imo/'hole',inwhicha \ L  supralaryngeallyarticulatedconsonantseparatesthetwovowels:2'(  40  )3    Ԁ ` Input/XDXXXDimo/'hole'    *q ddd d""WX  q,Ldd ,Ldd",Ldd",Ldd"+  TT   Candidate +T\L TT +NoGap )T\L T )*Compseg )T\L T )Align +T\L T +a.#XDXXXDsI#XDXXXDԀimi  Ѐ\/̀vpl +T +*! - - - - +T +b.#XDXXXDK#ԀXDXXXDԀimi 4$ Ѐ\ /   Ѐvpl + T + "4$ "*!    +T4$  +c.+#XDXXXDL#XDXXXDԀimo |! Ѐ   xh" Ѐvplvpl +dT#T + "|$ "  |% * xh'  Alignmentrequiresthatallvowelplacefeaturesberightalignedintheword(hereforconveniencetheyarecollapsedtogetherandabbreviatedas'vpl').Thoughalignmentisentirelysatisfiedinboth howdie(40)  aand howdie(40)  b,thefirstcomesattheexpenseofafatal D 4* NoGapviolation,sincemisskipped,andthesecondfailsbecauseoflinkagetom, 0! + violating*Compseg.Thoughalignmentisviolatedin howdie(40)  c,thisformisoptimal(the " , numberofAlignRviolationsherewillactuallydependonthenumberofVPlace #- featuresatissue). 4 Comparenowtheform/nihin/'tooneself': #XDXXXD-N# &!1 & 8  2((  41  )3  Ԁ ` Input/nihin/'tooneself' 8 * ddLdd Ldd"Ldd"Ldd"   ,Ldd ,Ldd",Ldd",Ldd"+  TT  Candidate +Tt<TT +NoGap )Tt<T )*Compseg )Tt<T )Align +Tt<T +a.XDXXXDԀnihin  Ѐ\/̀vpl +  T +*! -   - -  - +T  +b.+nihin L  Ѐ\ / 8  Ѐvpl +$T + "L  "  8   T L  c.#XDXXXDU#Ԁnihin l  Ѐ   X  Ѐvplvpl +|D T + "l  "  l  *! X   '88RThecrucialdifferencehereisthatnoamountofvowelplacefeaturelinkingtohcan L violate*Compseg,sincehhasnoprimaryconsonantalplace,andsotheresultisnot p8 complexintherightsense.Withinterveninglaryngeals,therefore,thereisnoobstacletoalignment. 4 AlthoughSteriade(1987)consideredanaccountfortranslaryngealharmonyinvokingstrictlylocalspreadinginderivationalterms,thisapproachwasnotadopted.Steriadepresentedtwochallengesforsuchanaccount.First,somecasesoftranslaryngealharmonyspecifythetargetvowel.InArbore,accordingtoSteriade,thetargetvowelmustbemidfront.Theproblemisthatlaryngealsarebynolikelycriteriathemselvesmidfront,andsocannotbeviewedastargetsofassimilation.Ifspreadingisstrictlylocal,thenspreadingmustfirstoccurtoaninterveninglaryngeal,beforepassingontothefollowingvowel,andthisisnotpossibleinArbore. 4 However,inthenonderivationalsettingofOptimalityTheory,thingslookdifferent,becausetheobjectsbeingevaluatedinvolvemultiplelinkingattheoutset.Thelaryngealisneveranexclusivetargetatsomeintermediatestageofthederivation,sincetherearenosuchstages.Rather,linkageisdrivenbyalignmentoveralegitimatedomainliketheprosodicword,asseenabove.InconcertwithNoGap,alignment $. itselfiswhatensuresthatlinkagewillspanover,andinclude,anywordinternalconsonantssolongasnohigherrankingmarkednessconstraintsinterfere.Itremainsanopenquestionhowweoughttoencodespecialconditionsontargetvowels,asinArbore.t ;   17      ׀Buttheexistenceofsuchconditionsisnolongeranobstacletothefeature '"2 cooccurrenceapproachtosegment'transparency',apointverymuchinfavorofOT'snonderivationalperspective. )l$4 Ї 4 AmuchmorepressingconcernisSteriade'ssecondchallenge:recallthattranslaryngealharmoniesareobservedtoinvolvetotalvowelplaceassimilation,andnevertheassimilationofasinglefeature.Thisgeneralizationfollowsfromthenodeasblockerviewofthefacts,sincetranslaryngealharmonycanoccurinanycaseonlyifspreadingisofthenodePlace.Itisimportantheretorememberthedefinitionof  translaryngealharmony:harmonyacrosslaryngeals,andnotacrosssupralaryngeals.   Ifspreadingwereofanylessernodeinthehierarchy,Placewouldnotblock,and   harmonywouldspansupralaryngealconsonantsaswell. 4 Topursuethepoint,recalltheconjecturereachedinsection3(Prince,p.c.):supposealignmentcannevertargetindividualfeatureslike[back]or[high],butonlytargetsfeatureclasses.Thisnotionisinterestingnotonlyforreasonsgiveninsection 4$  3,butbecause,giventheresourcesofOTalreadyavailable,itissufficientforordinary    casesofsinglefeaturevowelharmony:Finnish[back]harmony,forexample,canalreadybemadetofollowfromaconstraintAlign(Color),becauseofthe   independentlyrequiredpossibilityofpartialclassbehavior.Iffaithfulnessconstraints,forexample,canbefeaturespecific,assuggestedbyvariousresearchers(Kirchner1993,ArchangeliandPulleyblank1994b,amongothers),thensuchsinglefeatureharmonieswillinevitablyresultfromrankingssuchas MINDY25(42)  aand MINDY25(42)  b.Colorharmonyismadepossiblebyranking MINDY25(42)  c,and MINDY25(42)  disthecaseofnoharmony.2)(  42  )3   u Ԁ ` SingleanddoublefeatureharmoniesfromAlign(Color) l\  4  ` a. FillLink(round)>>Align(Color)>>FillLink(back) D4  4  ` b. FillLink(back)>>Align(Color)>>FillLink(round) 0   4  ` c. Align(Color)>>FillLink(round),FillLink(back)    4  ` d. FillLink(back),FillLink(round)>>Align(Color)   4 Assumingthisunderstandingofalignmentconstraints,themostdirectattempttoconstructan(illicit)caseoftranslaryngeal[back]harmony,say,willinvolvetheconstraintAlign(Color).Giventhisconstrainttogetherwith*Compsegandthetwo   relevantFillLinkconstraints,thereare24rankingspossible.Sometelling ! subrankingscanbeabstractedfromthesepossibilities;notethefollowinggeneralizationfirst:2*(  43  )3   v Ԁ ` Align(Color)harmonytypes T%D #  4  ` a. Align(Color)>>*Compseg0  ,'"%    4  ` 0 Anyharmonyexisting(subjecttoFillLink)spansanyconsonant.)#'  &   4  ` b.*Compseg>>Align(Color) *%)  4  ` 0 Anyharmonyexistingistranslaryngeal    '*wm -(, Restrictingourattentiontocasesof MINDY27(43)  b,anyvariationwithinthisclasswilldepend onthepositioningofthefaithfulnessconstraints,giving12rankingsinall.Inhalfofthese,bothFillLinkconstraintsoutrankalignment,meaningthatthereisno  harmonyatall.Therestgivethreetranslaryngealharmonytypes,tworankingsforeach:&  2+(  44  )3  Ԁ ` Translaryngealharmonytypespredicted    4  ` a. *Compseg>>Align(Color)>>FillLink(back),FillLink(round)  `  4  `   Translaryngealcolorharmony' p 4  ` b. *Compseg,FillLink(round)>>Align(Color)>>FillLink(back) H   4  `   Translaryngeal[back]harmony 4  ` c. *Compseg,FillLink(back)>>Align(Color)>>FillLink(round)   4  `   Translaryngeal[round]harmony 4 Thusevenassumingthatalignmentinitselftargetsonlyfeatureclasses,wehavenotyetruledouttheunattestedsinglefeaturetranslaryngealharmonies(thoughwehaveobviouslyengenderedfewersuchrankingsthanatheorycontainingbothclassalignmentandsinglefeaturealignmentconstraints).WhethertranslaryngealharmoniesofColorexistisamatterrequiringfurtherinvestigation,butOdden(1991)provides X  acandidatecase.w ;   18      ׀ D   4 Asamatterofinterest,itisthedecompositionoffaithfulnessintovariousfeaturespecificconstraintsFillLink(back),FillLink(round),andsoonthatleadsto  thisstateofaffairs.Thisdecompositionseemsrequired,though:assumingthatalignmenttargetsonlyfeatureclasses,itisthedecompositionoffaithfulnessthatallowsforthepossibilityofsinglefeaturespreadingatall,asnotedin mindy25(42)  .However,consideradifferentstrategyforgleaningsinglefeatureharmoniesfromconstraintslikeAlign(Color).Insteadofdecomposingfaithfulness,weunpack*CompSegonce #l! again,allowingspecificreferencetotherelevantvowelfeaturesinthismanner:*[CPlace,round],*[CPlace,back],etc.Suchconstraintspenalizecomplex |%D # segmentsinvolvingCPlace(anyoralconsonantalconstriction)andaparticularvowel h&0!$ placefeature.SubstitutedforthecorrespondingFillLinkconstraintsin MINDY25(42)  ,the  effectswillbeintriguinglydifferent:2,(  45  )3    Ԁ ` Align(Color)harmonytypes,reconsidered   4  ` a. *[CPlace,round]>>Align(Color)>>*[CPlace,back]    4  ` b. *[CPlace,back]>>Align(Color)>>*[CPlace,round]    4  ` c. Align(Color)>>*[CPlace,round],*[CPlace,back]  t  4  ` d. *[CPlace,back],*[CPlace,round]>>Align(Color) p ` Assumingthatfaithfulness(nowgenerallyconstrued)andotherconstraintspermit,thepatternspredictedby Mindy501(45)  Հarethefollowing: Mindy501(45)  distranslaryngealcolorharmony,fromtheconfigurationalreadyfamiliar; Mindy501(45)  cderivesgeneralcolorharmony,asinTurkish.Mostinterestingaretheremainingcases. Mindy501(45)  bforces[round]harmonythroughanyconsonant,butallows[back]harmonyonlyifalaryngeal(ornothing)intervenes.Thesurfaceeffectofthisconfigurationisalanguagedisplayingtranslaryngealcolorharmony,aswellasageneral[round]harmony. Mindy501(45)  agivesthecomplementpattern:translaryngealcolorharmony,withgeneralharmonyof[back].Singlefeaturetranslaryngealharmonyisnotderivedbytheserankings. 4 Whetherthesepredictionsthemselvesaredesirableisnotentirelyclear;yetinrulingoutsinglefeaturetranslaryngealharmoniesthisapproachearnsalegitimateclaimtoourinterest.Shoulditturnouttobeontherighttrack,itimpliesanunderstandingofindividualfeaturebehaviorthatmotivatesitnotbyvirtueofproliferatingfeaturefaithfulnessconstraints,butratherbyvirtueofconstraintsonfeaturemarkedness.Constraintslike*[CPlace,X],afterall,aresimplymembersof   alargerassortmentofconstraintsdisfavoringfeaturecooccurrences,ormoregenerallydisfavoringvarioussegmenttypes.TheconstraintinformallydubbedLico  intheaccountofTurkishabovebelongstothisclassaswell.Giventhatconstraintsonsegmentalmarkedness(howeverstated)areanineradicablepartofthetheory,wheretheyalreadyplayalargeroleinconstrainingprocesseslikefeaturespreading,perhapsthemostinterestingquestionraisedbythisdiscussioniswhethertheymightactuallysufficeforsinglingoutfeatureswhereneeded,allowingustomaintainrathermonolithicfaithfulnessconstraintsontheorderofParseFeature,FillLink,etc. |#l!  4 Asforthematterathand,furtherdiscussionwillbedifficultuntilmoreisknownaboutthepropertiesoftranslaryngealharmonycases.Questionsalsoloomconcerningtherangeofexistingsegmentalmarkednessconstraints,andtheformstheymaytake;theiranswerswillthemselvesbearontheissueofpossiblegrammarsconcerningusabove. 5.Conclusion  +&*  4 ThispaperfindsitsprimarysupportforFeatureClassTheoryinfactsofcolorharmony,anarearichinpartialclassbehavior.Padgett(1995)locatesargumentsfor .|)- FCTinotherareasobligatorycontourprincipleeffectsoverplace,andpartialassimilationbynasalstocomplexsegmentsandexplorestheimplicationsofFCTforaccountsofthe(non)interactionofplacefeaturesofconsonantsandvowels. 4 PresentedwiththeargumentforaclassColorfromTurkish,where[back]and  [round]spreadingdonotevencoincideallofthetime,theskepticmightwonderhowonecanreliablyinferaclassfromdataatalltheinferenceinvolvesafairdegreeofabstractionfromthedata.Inclosing,itmightbeusefultorecallthatthisissuealreadydeeplypervadesallworkonFG.Inhisworkintroducingthattheory,Clements(1985)consideredasimilarquestioninrelationtosomefactsofKikuyu.Inthislanguage,whenanasalconsonantprecedesafricative,thefricativeishardenedtoastop(seealsoPadgett1991formoredetaileddiscussion):&  2-(  46  )3  Ԁ ` FricativehardeninginKikuyu 4   4  ` mbureetD z 'lopoff' h   cf.ura     4  ` ndeheetD z 'pay' h   cf.reha   4  ` goreetD z 'buy' h   cf.Oora  '4fWhenanasalprecedesavoicelessobstruent,theobstruentisvoiced:2.(  47  )3  Ԁ ` ObstruentvoicinginKikuyu \  4  ` ndomeetD z 'send' h   cf.toma l4  4  ` n)j ``0ineetD z 'burn' h   cf.cina X   4  ` gereetD z 'cross' h   cf.kera D  AsClementsnotes,thesefactsonthesurfaceresemblecounterevidencetotheclaimthat[voice]and[continuant]occupyindependentareasoftheFGhierarchy:theappearanceisofauniformprocessofobstruenthardening/voicing.Yetwedonotrushtopositagroupingof[voice]and[continuant],forvariousmutuallyreinforcingreasons.Thus,thesefeaturesrepresentindependentphoneticdimensions,andmoreimportant,theyareknowntobehavequiteindependentlywhenthelargerpictureofcrosslinguisticpatterningisviewedthechancecooccurrenceofhardeningandvoicingafternasals,bothwidelyattestedphenomenaoccurringalone,isexpectedinthislargerpicture.Moregenerally,hypothesesaboutfeatureclassescannotbebasedonsurfacescanningsofasinglelanguage.Rather,theymustbeinformedbythelargercontextbothofthetheory,andofcrosslinguisticpatterning.ItisjustsuchconsiderationsthatmotivatedtheclassColorinthiswork:Colorderivessupportfrom @(#& phoneticconsiderations,frompatternsofvowelinventories,andfromthecrosslinguisticcopatterningof[back]and[round]invowelharmonysystems.  +&* [X` hp x (#%'0*,.8135@8:<H?A X4,z h, X &    References   Ք 0 `  Anderson,JohnandColinEwen.'ɔ(1987).PrinciplesofDependencyPhonology.   CambridgeUniversityPress,Cambridge. ` `  0 `  Anderson,StephenR.(1980).'ProblemsandPerspectivesintheDescriptionofVowelHarmony'.InRobertM.Vago(ed.),IssuesinVowelHarmony.John   Benjamins,Amsterdam,pp.148. ` `  0 `  Archangeli,Diana.(1984).UnderspecificationinYawelmaniPhonologyand  t Morphology.DoctoralDissertation,MIT,Cambridge,MA.p `` `  0 `  Archangeli,DianaandDouglasPulleyblank.(1986).TheContentandStructureof \ L  PhonologicalRepresentations.Ms.,UniversityofArizonaandUniversityof H8  Ottawa. ` `  0 `  Archangeli,DianaandDouglasPulleyblank.(1994a).GroundedPhonology.MIT    Press,Cambridge,MA. ` `  0 `  0` ` `   Archangeli,DianaandDouglasPulleyblank.(1994b).'KinandeVowelHarmony:Domains,GroundedConditions,andOneSidedAlignment'.Ms.,UniversityofArizonaandUniversityofBritishColumbia. ` `  Borowsky,Toni.(1986).TopicsintheLexicalPhonologyofEnglish.Doctoral  dissertation,UniversityofMassachusetts,Amherst.0 `  Buckley,Eugene.(1994).TheoreticalAspectsofKashayaPhonologyand  Morphology.CSLIPublications,Stanford,CA.p` `  0 `  Clements,G.N.(1985).'TheGeometryofPhonologicalFeatures'.Phonology l\ Yearbook2,22552.XH` `  0 `  Clements,G.N.(1991).'PlaceofArticulationinConsonantsandVowels:aUnifiedTheory'.Ms.,version1.1,CornellUniversity. ` `  0 `  Clements,G.N.andEnginSezer.(1982).'VowelandConsonantDisharmonyinTurkish'.InHarryv.d.HulstandNorvalSmith,eds.,TheStructureof  PhonologicalRepresentations,PartII,Foris,Dordrecht,pp.213255.` `  0 `  Ewen,ColinandHarryv.d.Hulst.(1988).'[high],[low]and[back]or[I],[A]and[U]?'InP.CoopmansandA.Hulk(eds.),LinguisticsintheNetherlands.  Foris,Dordrecht,pp.4958. ` `  0 `  Gazdar,Gerald,EwanKlein,GeoffreyPullumandIvanSag.(1985).Generalized ! PhraseStructureGrammar.HarvardUniversityPress,Cambridge,MA." ` `  0 `  Goad,Heather.(1993).OntheConfigurationofHeightFeatures.Doctoral |#l! dissertation,UniversityofSouthernCalifornia. ` `  0 `  Goldsmith,John.(1990).AutosegmentalandMetricalPhonology.BasilBlackwell, T%D # Cambridge,MA. ` `  0 `  Haiman,John.(1972).'PhonologicalTargetsandUnmarkedStructures'.Language ,'"% 48.2,pp.365377. ` `  0 `  Halle,Morris.(1993).'FeatureGeometryandFeatureSpreading'.Ms.,MIT,Cambridge,MA. ` `  0 `  Hayes,Bruce.(1990).'DiphthongisationandCoindexing'.Phonology7.1,pp.3171.*%)` `  0 `  Hulst,Harryv.d.(1988).'TheGeometryofVocalicFeatures'.InHarryv.d.HulstandNorvalSmith,eds.,Features,SegmentalStructureandHarmonyProcesses ,'+  PartII,Foris,Dordrecht,pp.77125.-(,` `  0 `  Hulst,Harryv.d.andNorvalSmith.(1985).'VowelFeaturesandUmlautinDjingili, NyangumardaandWarlpiri'.PhonologyYearbook2,pp.277303.$` `  0 `  Hulst,Harryv.d.andNorvalSmith.(1987).'VowelHarmonyinKhalkhaandBuriat(EastMongolian)'.InF.BeukemaandP.Coopmans(eds.),Linguisticsinthe  Netherlands.Foris,Dordrecht,pp.8191.` `  0 `  Hulst,Harryv.d.andJeroenv.d.Weijer.(1991).'TopicsinTurkishPhonology'.InHendrikBoeschotenandLudoVerhoeven,eds.,TurkishLinguisticsToday,   E.J.Brill,Leiden. ` `  0 `  Hume,E.(1992).FrontVowels,CoronalConsonantsandtheirInteractioninNon  ` LinearPhonology.Doctoraldissertation,CornellUniversity,Ithaca,NY. L ` `  0 `  Hyman,Larry.(1988).'UnderspecificationandVowelHeightTransferinEsimbi'.Phonology5,pp.255273.\$ ` `  0 `  Inkelas,Sharon,OrhanOrgunandCherylZoll.(1994).'SubregularitiesasCogrammars:theTheoreticalStatusofNonproductivePatternsinGrammar'.Ms.,UCBerkeley. ` `  0 `  It=,Junko.(1986).SyllableTheoryinProsodicPhonology.DoctoralDissertation,    UniversityofMassachusetts,Amherst. ` `  0 `  It=,JunkoandR.ArminMester.(1993).'LicensedSegmentsandSafePaths'.InC.ParadisandD.LaCharit(eds.),ConstraintBasedTheoriesinMultilinear  Phonology,specialissueoftheCanadianJournalofLinguistics.` `  0 `  It=,Junko,R.ArminMesterandJayePadgett.(toappear).'NC:LicensingandUnderspecificationinOptimalityTheory'.LinguisticInquiry.\` `  0 `  Jakobson,R.andM.Halle.(1956).FundamentalsofLanguage.Mouton,The H Hague. ` `  0 `  Kaun,Abigail.(1994).'AnOptimalityTheoreticAccountofRoundingHarmonyTypology'.PaperpresentedattheWestCoastConferenceonFormalLinguistics. ` `  0 `  Keating,Patricia.(1988).'UnderspecificationinPhonetics'.Phonology5,pp.275  292. ` `  0 `  Kiparsky,Paul.(1981).'VowelHarmony'.Ms.,MIT,Cambridge,MA. ` `  0 `  Kiparsky,Paul.(1985).'SomeConsequencesofLexicalPhonology'.Phonology2,   85138. ` `  0 `  Kirchner,Robert.(1993).'TurkishVowelHarmonyandDisharmony:anOptimalityTheoreticAccount'.Ms.,UCLA,oftalkpresentedattheRutgersOptimalityWorkshopI,NewBrunswick,NJ. ` `  0 `  Korn,David.(1969).'TypesofLabialVowelHarmonyintheTurkicLanguages'.AnthropologicalLinguistics11.3,pp.98106.h&0!$` `  0 `  Lees,Robert.(1961).ThePhonologyofModernStandardTurkish.UralicandAltaic T'"% Series6,IndianaUniversity,Bloomington. ` `  0 `  Levergood,Barbara.(1984).'RuleGovernedVowelHarmonyandtheStrictCycle'.InProceedingsofNELS14,GLSA,UniversityofMassachusetts,Amherst, *$( pp.275293. ` `  0 `  Lombardi,Linda.(1991).LaryngealFeaturesandLaryngealNeutralization. +&* Doctoraldissertation,UniversityofMassachusetts,Amherst. ` `  0 `  McCarthy,JohnJ.(1988).'FeatureGeometryandDependency:AReview'.Phonetica -(, 43,pp.84108..|)-` `  0 `  McCarthy,JohnJ.(1991).'OnGutturalTransparency'.PaperpresentedattheWorkshoponDistinctiveFeatures,LSASummerInstitute,SantaCruz,CA. ` `  0 `  McCarthy,JohnJ.(1994).'OnCoronal"Transparency"'.PaperpresentedattheannualTrilateralPhonologyWeekend,UCSantaCruz. ` `  0 `  McCarthy,JohnJ.andAlanPrince.(1993a).'ProsodicMorphologyI:ConstraintInteractionandSatisfaction'.Ms.,UniversityofMassachusetts,Amherst,andRutgersUniversity,NewBrunswick. ` `  0 `  McCarthy,JohnJ.andAlanPrince.(1993b).'OptimalityinProsodicMorphology:theEmergenceoftheUnmarked'.Paperpresentedatthe24thannualNELSconference,Amherst,MA. ` `  0 `  McCarthy,JohnJ.andAlanPrince.(1993c).'GeneralizedAlignment'.Ms.,UniversityofMassachusetts,Amherst,andRutgersUniversity,NewBrunswick. ` `  0 `  Mester,R.Armin.(1986).StudiesinTierStructure.Doctoraldissertation,University    ofMassachusetts,Amherst,published1988,byGarlandPublishers,NewYork. ` `  0 `  NChiosin,Mire.(1994).'IrishPalatalisationandtheRepresentationofPlaceFeatures'.Phonology11.1.` `  0 `  NChiosin,Mire.(toappear).'BarraGaelicVowelCopyand(Non)ConstituentSpreading'.ProceedingsofWCCFLXIII. ` `  0 `  NChiosin,MireandJayePadgett.(1993).'InherentVPlace'.Reportno.LRC9309,LinguisticsResearchCenter,UCSantaCruz. ` `  0 `  NChiosin,MireandJayePadgett.(1994).'FeatureClassesandSpreading'.PaperpresentedattheSecondPhonologyWorkshop:fromCognitiontoRomance.UniversityofManchester. ` `  0 `  Odden,David.(1991).'VowelGeometry'.Phonology8,pp.26190.0 ` `  0 `  Padgett,Jaye.(1991).StrictureinFeatureGeometry.Doctoraldissertation,   UniversityofMassachusetts,Amherst,toappearCSLIPublications,Stanford,CA. ` `  0 `  Padgett,Jaye.(1994).'FeatureClasses'.PaperpresentedattheUniversityofMassachusetts,Amherst,October21. ` `  0 `  Padgett,Jaye.(1995).'FeatureClassesII'.PaperpresentedattheSecondHollandInstituteofLinguisticsPhonologyConference,UniversityofAmsterdam,January1995.[Tobepublishedinproceedings]. ` `  0 `  Paradis,CaroleandJ.F.Prunet.(1989).'OnCoronalTransparency'.Phonology6.2, |#l! pp.317348. ` `  0 `  Polgrdi,Krisztina.(1994).'TurkishVowelHarmonyinOptimalityTheory'.Ms.,HIL/LeidenUniversity. ` `  0 `  Prince,Alan.(1987).'PlanesandCopying'.LinguisticInquiry18.3,pp.491509.,'"%` `  0 `  Prince,AlanandPaulSmolensky.(1993).'OptimalityTheory:ConstraintInteractioninGenerativeGrammar'.Ms.,RutgersUniversity,NewBrunswick,andUniversityofColorado,Boulder. ` `  0 `  Pulleyblank,Douglas.(1994).'VowelHarmonyinOptimalityTheory'.Toappear,ProceedingsoftheWorkshoponPhonology,UniversityofCoimbra,Portugal, +&*  September1993.,'+` `  0 `  Ryder,MaryEllen.(1987).'AnAutosegmentalTreatmentofAssimilationtoLabiovelars'.InproceedingsofCLS23PartTwo:ParasessiononAutosegmentalandMetricalPhonology,ChicagoLinguisticSociety. ` `   0 `  Sagey,Elizabeth.(1986).TheRepresentationofFeaturesandRelationsinNon  LinearPhonology.Doctoraldissertation,MIT,Cambridge,MA.` `  0 `  Sagey,Elizabeth.(1987).'NonConstituentSpreadinginBarraGaelic'.Ms.,SchoolofSocialSciences,UCIrvine. ` `  0 `  Schane,S.(1972).'[back]and[round]'.InS.R.AndersonandP.Kiparsky(eds.),A  t FestschriftforMorrisHalle.Holt,RinehartandWinston,Inc.,NewYork,  ` pp.174184. ` `  0 `  Schane,S.(1984).'TheFundamentalsofParticlePhonology'.PhonologyYearbook p8  1,pp.129155. ` `  0 `  Selkirk,Elisabeth.(1988).'Dependency,Place,andtheNotion"Tier"'.Ms.,UniversityofMassachusetts,Amherst. ` `  0 `  Selkirk,Elisabeth.(1991a).'MajorPlaceintheVowelSpace.VowelHeight'.PaperpresentedattheWorkshoponFeatureOrganization,LinguisticSocietyofAmericaInstitute,SantaCruz,CA.[manuscriptUniversityofMassachusetts,Amherst]. ` `  0 `  Selkirk,Elisabeth.(1991b).'ATwoRootTheoryofLength'.InE.DunlapandJ.Padgett,eds.,UniversityofMassachusettsOccasionalPapersinLinguistics14:PapersinPhonology.GLSA,UMass,Amherst. ` `  0 `  Selkirk,Elisabeth.(1993).'LabialRelations'.Ms.,UniversityofMassachusetts,Amherst. ` `  0 `  Smolensky,Paul.(1993).'Optimality,Markedness,andUnderspecification'.PaperpresentedattheRutgersOptimalityWorkshopI,NewBrunswick,NJ. ` `  0 `  Steriade,Donca.(1981).'ParametersofMetricalVowelHarmonyRules:theUnderlyingRepresentationofHarmonicVowels'.Ms.,MIT. ` `  0 `  Steriade,Donca.(1987).'LocalityConditionsandFeatureGeometry'.InproceedingsofNELS17,GLSA,DepartmentofLinguistics,UniversityofMassachusetts,Amherst,pp.595617. ` `  0 `  Steriade,Donca.(1994).'UnderspecificationandMarkedness'.InJohnGoldsmith,ed.,HandbookofPhonologicalTheory,Blackwell,Oxford,pp.114174.!` `  0 `  Stevens,KennethandSamuelJ.Keyser.(1989).'PrimaryFeaturesandtheirEnhancementinConsonants'.Language65,pp.81106.#l!` `  0 `  Stevens,Kenneth,SamuelJ.KeyserandH.Kawasaki.(1986).'TowardsaPhoneticandPhonologicalTheoryofRedundantFeatures'.InSymposiumon |%D # InvarianceofSpeechProcesses,J.PerkellandD.H.Klatt,eds.,Lawrence h&0!$ Erlbaum,Hillsdale,NJ. ` `  0 `  Trubetzkoy,N.S.(1939).GrundzgederPhonologie.VandenhoeckandReprecht, @(#& Gttingen.[SecondeditiontranslatedintoRussianbyA.A.Kholodovich.OsnovyFonologii,Izdatel'stvoInostrannojLiteratury,Moscow,1960].*$(` `  0 `  Vago,RobertM.(1980).TheSoundPatternofHungarian.GeorgetownUniversity +%) Press,Washington,D.C. ` `    ,'+ &   DepartmentofLinguisticsStevensonCollegeUCSCSantaCruz,CA95064padgett@ling.ucsc.edu'