FCT2.WPDTEXTdosa'WIWIo1WPC> '%+6SwVDs`<237e?L䒜tJqkF6TviZ?,G݈-ۈ$:<^o "c45FAezɘ}t㙵K==mV]LaA=+XǟPH%Acp9^j8=E1@]=;߁ʷ)5U=OY,ܣwR-!{ʮVܢ!hMT9'FQf;H255Cy16[挺˝Vjd_} mEb.x4 6;7]6bW&)cA!wD v4]Lp 1Ir<H4U<4-Oq_v,cħÏw3w;CiUU]/_21=2Zh[V@or|]DXΓrh(&trX}#LsKݙ̩ 0p 0G 0 0 0{ 0 0 0a 0 0 0h 0y  0k 0k 0~[ 0~UG0W % 0 0N 0 0 0@ 0 0X 0 0p 0) 0 0gA! 0Y! 0Y" 0lZ" 0l" 02# 0$ 0$ 0% 0& 0.' 0A)) 0Tj* 0z+ 08, 1- 0. . 0= / 0TH/ 0P/ 1/ 0061 04U5E5 0}5 0xZ6 06 02_7 07ej@ 0k@ 0|E 0(K 0(K 0(K L 1mLwLiNUNgQ Q CQXQ1V B AV C^VfxV zV B4*VUNVfVaV W W(W 9YfCY EYUNQY BY YKY ]U>] []@g]o}]pV}a}nX} 0; 1m AX 0 D+U NUN8 UN4ނU@T2 h 1 0D; C  2U> h! h#  h hhɍ h11111 AY7 0FVVVVVVVVVV@4 А@ڐopv f pppppppppppFzUNUNo\p\@4RRo pV%# 0 0D AW 0K@Kw@aop .UN.@...o//p3UNHUNTHUNHH@JUNJUNJ@EKo[Kp[OoqdpqhG\ȇ $0$ 2 >J  !f-/f˒f͒aϒaa ݓi/= = IaU 0ci 0w̘ 0C 0͙ 0m 0! 0 0Ü =I  Ĥ Ф ܤ a6l83:Technical[1]Technical Document Style  23  Ԁ     .. l-(:Technical[2]Technical Document Style 23  Ԁ   .. l*$:Technical[3]Technical Document Style 23  Ԁ   .. l&$:Technical[4]Technical Document Style 23  Ԁ   .. l/:Technical[5]Technical Document Style.. 23  Ԁ. l/:Technical[6]Technical Document Style.. 23  Ԁ. l/:Technical[7]Technical Document Style.. 23  Ԁ. l/:Technical[8]Technical Document Style.. 23  Ԁ. VV!8Document[1]Document Style  @..  23  Ԁ     Ԉ V-8Document[2]Document Style 23  Ԁ   V=8Document[3]Document Style.. 23  Ԁ   V 8Document[4]Document Style.. . V 8Document[5]Document Style0..V 8Document[6]Document Style8..V& 8Document[7]Document Style0..0` ..` V& 8Document[8]Document Style0..8` ..` ) p`CG TimesXV83$1Technical Document Style  23  Ԁ     .. V-($2Technical Document Style 23  Ԁ   .. V*$$3Technical Document Style 23  Ԁ   .. V)$$4Technical Document Style 23  Ԁ   .. V/$5Technical Document Style.. 23  Ԁ. V/$6Technical Document Style.. 23  Ԁ. V/$7Technical Document Style.. 23  Ԁ. V/$8Technical Document Style.. 23  Ԁ. BV!$9Document Style  @..  23  Ԁ     Ԉ D-&10Document Style 23  Ԁ   D=&11Document Style.. 23  Ԁ   D &12Document Style.. . D &13Document Style0..D &14Document Style8..D& &15Document Style0..0` ..` D& &16Document Style0..8` ..` zS :Right Par[1]Right-Aligned Paragraph Numbers..23  Ԁ..0..zf :Right Par[2]Right-Aligned Paragraph Numbers..` ..` 23  Ԁ..0` ..` zy :Right Par[3]Right-Aligned Paragraph Numbers..` ..`  .. 23  Ԁ` ..` 0 .. z :Right Par[4]Right-Aligned Paragraph Numbers..` ..`  .. ..23  Ԁ .. 0..z :Right Par[5]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h23  Ԁ..0h..hz :Right Par[6]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h..23  Ԁh..h0..z :Right Par[7]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....23  Ԁ..0..z :Right Par[8]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....p..p23  Ԁ..0p..pR& 8BibliogrphyBibliography0....jo4Tech InitInitialize Technical StyleS -  1 .1 .1 .1 .1 .1 .1 .1 S CuyTechnicalTechnical Document Style11.11.1.11.1.1.11.1.1.1.11.1.1.1.1.11.1.1.1.1.1.11.1.1.1.1.1.1.1fp2Doc InitInitialize Document Style/S 3    I. 1. A. a.(1)(a) i) a)S012($0 (,!$0  0` (#(#  (($0 0 0   A_ekqwDocumentDocument StyleI.1.A.a.(1)(a)i)a)x?t2PleadingHeader for Numbered Pleading Paper 5  6(  oXXXXoXXo7XoXXXoXX6M8'*d66M8'*d6\\1\\2\\3\\4\\5\\6\\7\\8\\910111213141516171819202122232425262728  .+(2 $ g8!    C E $#hC--  \ `p4Heading 2Underlined Heading Flush LeftV 4Heading 1Centered Heading@..  b* 8Bullet ListIndented Bullet List0..0` ..` 20Exampler:Squib ReviewRequest for review of squibXXXE) p`CG TimesE,|..LinguisticInquiry..SquibsandDiscussion..Editors:..|..|ChisatoKitagawa..|..|JohnMcCarthy(mccarthy@cs.umass.edu)..|..|MargaretSpeasDepartmentofLinguisticsUniversityofMassachusettsAmherst,MA01003USAXXXE) p`CG TimesE..|..|"US ,  > January22,1995!  8.,4 <DL84..4TheenclosedmanuscripthasbeensubmittedforpublicationintheSquibsandDiscussionsectionofLinguisticInquiry.Wewouldappreciateitifyoucouldlookoverthemanuscriptandsendusabriefreviewofit.Inordertoensurethatwewillbeabletorespondtotheauthorwithinareasonableamountoftime,werequestthatyousendusyourreviewwithinthreeweeks.Ifyouarenotabletoreviewitwithinthreeweeks,pleasereturnitimmediatelyand,ifpossible,suggestanalternatereviewer.4..4Thankyouinadvanceforyourhelp.Forourconvenience,pleaseenclosethisletterwithyourreview.4..4____Publishasis.4..4____Donotpublish.4..4____Publishwithminorrevisions.4..4____Publishonlywithmajorrevisions.4..4____Inappropriateasasquib;subjectmatterrequiresgreaterlength.4..4Pleaseencloseadditionalcommentsonaseparatesheet.4..4 ..  .. <..<....D..DSincerely,4..4 ..  .. <..<....D..DJohnJ.McCarthyXXXE) p`CG TimesE<:Squib LetterXXXE) p`CG TimesE,|..LinguisticInquiry..SquibsandDiscussion..Editors:..|..|ChisatoKitagawa..|..|JohnMcCarthy(mccarthy@cs.umass.edu)..|..|MargaretSpeasDepartmentofLinguisticsUniversityofMassachusettsAmherst,MA01003USAXXXE) p`CG TimesE..|..|"US ,  > January22,1995!  8.,4 <DL84..4 ..  .. <..<....D..DSincerely,4..4 ..  .. <..<....D..DJohnJ.McCarthyXXXE) p`CG TimesET.LetterLinguistics LetterXXXE) p`CG TimesE   =    XXXE) p`CG TimesEUNIVERSITYOFMASSACHUSETTS(#..(#XXXE) p`CG TimesEDepartmentofLinguisticsXXXE) p`CG TimesEATAMHERSTXXXE) p`CG TimesESouthCollege(#..(#mccarthy@cs.umass.eduAmherst,MA01003(413)5456830XXXE) p`CG TimesE       (#..(#"US ,  > January22,1995!  Ԉ  4..4 ..  .. <..<....D..D..Sincerely,4..4 ..  .. <..<....D..D..JohnJ.McCarthy4..4 ..  .. <..<....D..D..ProfessorofLinguisticsXXXE) p`CG TimesE    (&17(&18(&19jamie2$57?EKQYag20(1)A.1.a.(1)(a)i)a) m $  C E $$EC--    4    ݀ XXItisacommonperceptionthatFGmakesthispartitioningseemmorenatural;wearepresumablynotallowed  tolink[labial],say,tobothPlaceandLaryngealinatree.TheimplicittreetheoreticnotionhereistheSingleMother  Condition.Intruth,suchaconditionmusteitherbestipulatedorderivedinFG.#X߀X # _ $  C E $$EC--    2    ݀ XXTheclassesareassumedtobefixedanduniversalaswell,intheusualway.ThemakeupofLaryngealand  especiallyPlacein  odw1(4)  Հisobviouslynotintendedtobeeitherexhaustiveordefinitivelyaccurate;thesesubstantiveissues  crosscutthechoiceoftheory.#X߀X #(9 6Times New Roman Regularodw4 d N $  C E $$EC--    12    ݀ XXThoughMcCarthyandPrincethemselvesclearlynotethepossibilityoffeaturalcorrespondence,featural  faithfulnessisgenerallyhandledinthatworkbymeansofseparateIdentityconstraints,and(inarelatedmove)features  aregenerallyviewedasattributesofsegmentsratherthanentitiesenteringintocorrespondencerelationsthemselves.OnepossibleimplicationofthefeatureasattributeviewisalossofthechestnutautosegmentaldistinctionbetweentwosegmentssharingafeaturevalueFandtwosegmentseachseparatelyspecifiedforF(withconcomitantimplicationsinvolvinggeminateintegrityorinalterabilityeffects).Nothingintheanalyseshereseemstodependonthemoreconservativedecisiontoemployfeaturalcorrespondence(andsoclearlypreservesuchdistinctions)#X߀X #.nonoise ddodw9&0 d d(9 Z 6Times New Roman RegularTABLE Bodw10odw5  $  C E $$EC--    3    ݀ XXPadgett(1995a)attemptsageneraldefinitionofConstraint(Class)thatseemstowronglyusurpthe  freedomofaconstrainttochooseitsownquantifiersandvariablesinthisway.Thediscussionhereismeanttosupersedethatattempt.#X߀X #min2odw35 6Times New Roman Regular dmin3 A $  C E $$EC--    *    ݀ XXForveryhelpfuldiscussionleadingtoimprovementsofthispaperIwouldliketothankDonkaFarkas,Junko  It=,EwanKlein,LindaLombardi,ArminMester,GeoffPullum,participantsattheArizonaPhonologyConference,andthemembersofmySpring1995phonologyseminaratUCSC.ThepaperalsobenefittedfromdiscussionduringtalkspresentedatUCBerkeley,StanfordUniversityandtheCurrentTrendsinPhonologyConferenceinRoyaumont,France.IamespeciallyindebtedtoJillBeckmanandMireNChiosinforaclosereadingofthefirstdraft.#X߀X #noise*'Z&IPAKiel Regularodw13cd*OLE*WPCh@j_'pXXXX,,,,519V+'"!0'@ j c 6Times New Roman RegularcXXX @RootRoot""0p j7 c 6Times New Roman RegularcXXX @[nas][-cont]"#mj5 c 6Times New Roman RegularcXXX @Place'1"$0R 4 "%0GE ^@ ^h )"&04 ^ Q+N&O Fk GJz](]]^u%0f $ L*Ht,HIWPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@pPWf Ole 9WPG20CompObj^  !"#$%&'()*+)F! WPWin61[8]ࡱWPCh@j_'pXXXX,,,,519V+'"!0'@ j c 6Times New Roman RegularcXXX @RootRoot""0p j7 c 6Times New Roman RegularcXXX @[nas][-cont]"#mj5 c 6Times New Roman RegularcXXX @Place'1"$0R 4 "%0GE ^@ ^h )"&04 ^ Q+N&O Fk GJz](]]^u%0f $ L*Ht,HI `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; PPࡱ; ";  ?  ; "  QTimes New Roman-- .  "SystemBOlePartOlePres000--"2 !,Root Root  ! -- .  --?2 F%[nas] [-cont]F-- .  --2 ruPlace r--8&<-- %9%_--- !&; !'< !(= !)? !.E !/F !0H !1I !5O !6P !7Q !8R !9S !=Y !>Z !?[ !@\ !Ec !Fd !Ge !Hg !Mm !Nn !Op !Pq !Tw !Ux !Vy !Wz !X{ !\ !]--ࡱࡱ;  METAFILEPICT"";  ?  ; "  QTimes New Roman-- .  "System--"2 !,Root Root  ! -- .  --?2 F%[nas] [-cont]F-- .  --2 ruPlace r--8&<-- %9%_--- !&; !'< !(= !)? !.E !/F !0H !1I !5O !6P !7Q !8R !9S !=Y !>Z !?[ !@\ !Ec !Fd !Ge !Hg !Mm !Nn !Op !Pq !Tw !Ux !Vy !Wz !X{ !\ !]--3|:;\ `imesXXP\  P6QXP<6X`(imesXXx6X@JQX@(8$    C E $$EC--#2,37=CIQYag1.a.i.(1)(a)(i)1)a)kl&OLE Box<<=8Cmin1(3o$ g8!    C E $#hC--  'dxd(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regularodw8(9 Z 6Times New Roman Regular~? (\  `(Times NewRomanTT J $  C E $$EC--    6    ݀ XXSomeworkinwhatisbroadlycalledUnificationBasedgrammardispensesaltogetherwiththedistinction  between'constraint'and'representation'.InScobbie(1991)forexample(andcf.Bird1995),[labial],[coronal]etc.havethestatusofvaluesofalargerfeaturecategoryPlace,allembeddedinatheorybuiltonfeaturestructuresandtheirunification.Thisgeneralparadigmfavorsconstraintsthatarestrictlyspeakingtrueofsurfaceforms(andsonotviolable),andsoseemstorequirethe'hard'understandingoffeatureclassgeneralizationsaswell.#X߀X #egg1OOOOOOOOOOOOOOOOOOOOOTABLE AOmin36b$Triangle0 dodw14odw16 ( $  C E $$EC--    19    ݀ XXArizonaPhonologyConferencehandout.#X߀X #*'Z&IPAKiel RegularS SSSSSSSSSSSTABLE BSYYYYYYYYYYYYYYYYYYYYTABLE AYApple LaserWriter II NTPSCRIPT0 TABLE AklM << deUULevel 1Level 2Level 3Level 4Level 5(n$ (  1  ) OLE Generated Objectodw1*OLE*WPChS1jG(pXXXX,,,,519V+'"!C j c 6Times New Roman RegularcXXX @[son]""C1P j c 6Times New Roman RegularcXXX @[lab][voiced][c.g.][dors][cor][asp]'1"#Cp ^ > "$C^W j > "'C W j J  @ B!````@ `@ ``@ďď뺯??{J@J@ (b```@ `@ ``@?J@J@ (b```@ `@WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ; Root Entry`.@p?89_@Ole 9WPG20CompObj^  )F! WPWin61[7]ࡱWPChS1jG(pXXXX,,,,519V+'"!C j c 6Times New Roman RegularcXXX @[son]""C1P j c 6Times New Roman RegularcXXX @[lab][voiced][c.g.][dors][cor][asp]'1"#Cp ^ > "$C^W j > "'C W j J  @ B!````@ `@ ``@ďď뺯??{J@J@ (b```@ `@ ``@?J@J@ (b```@ `@ `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; PPmmࡱ; "D  ^R  D "  nQTimes New Roman[-- .  "SystemBOlePartOlePres000--2 s[son] -- .  --R2 ]2[lab] [voiced] [c.g.] [dors] [cor] [asp]]-- %JKN-- %LN-- %LM-HMove the focus to thࡱ; ࡱ;  METAFILEPICT""D  ^R  D "  nQTimes New Roman[-- .  "System--2 s[son] -- .  --R2 ]2[lab] [voiced] [c.g.] [dors] [cor] [asp]]-- %JKN-- %LN-- %LM-HMove the focus to thodw7odw3 < $  C E $$EC--    5    ݀ XXThislogicofdependencyfailsuswhenconfrontedwithfeatureclassgeneralizationsinvolvingphenomena  otherthanassimilationorpositionalneutralization,providingargumentsforFCTconsideredinPadgett(1995b).TheseincludeOCPeffectsandconstraintslikeHavePlaceseenearlier.#X߀X #(9 Z 6Times New Roman Regular(9 Z 6Times New Roman RegularWPC ' a3{| +3',,,,519LDhF4+'"!f c 6Times New Roman RegularcXXX @a.yGzGn'face(gen.)'""\ c 6Times New Roman RegularcXXX @b.yGzden'face(abl.)'"# c 6Times New Roman RegularcXXX @-br"$!c c 6Times New Roman RegularcXXX @r-b+'1 (!! *y0y"  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`.@p9y[ Ole 9WPG20CompObj!^  "%&'()*+,-./0123456)F! WPWin61[4]ࡱWPC ' a3{| +3',,,,519LDhF4+'"!f c 6Times New Roman RegularcXXX @a.yGzGn'face(gen.)'""\ c 6Times New Roman RegularcXXX @b.yGzden'face(abl.)'"# c 6Times New Roman RegularcXXX @-br"$!c c 6Times New Roman RegularcXXX @r-b+'1 (!! *y0y"  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; ` T    "Times New Roman+-- .  "SystemBOlePart#OlePres000$--T2 03a. y z n 'face (gen.)'( -- .  --R2 n22b. y z d e n 'face (abl.)'n'-- .  --2 I_-b rIz-- .  --2 \r -b--%`h;v;`-- %q~`q`-8ࡱ; ࡱ;  METAFILEPICTTh T    "Times New Roman+-- .  "System--T2 03a. y z n 'face (gen.)'( -- .  --R2 n22b. y z d e n 'face (abl.)'n'-- .  --2 I_-b rIz-- .  --2 \r -b--%`h;v;`-- %q~`q`-8OLE Generated ObjectWPC} 2! pn!'3',,,,51+' ! cX 6Times New Roman Regularc @XXXyGzden " cX 6Times New Roman Regularc @XXXColor #B cX 6Times New Roman Regularc @XXX-br'1 $ > I 7 % ) & %)'1"'z cX 6Times New Roman Regularc @XXX?WusssssKuststssuVfFf~t'^&GW t&?u8Ƌ&_ff1jhU5S $V&6ɏRfvRP Ł4lPH=pu NM>$)523^%d2ɲh h|Xjj6G~~dž0+P5fhfo/ N^ ?______RFnSSn eiP&otFc:PPSB0c2 &n tl:PPSB0j- ot&iej!!TPbx08F7u~ ~ PCancel-0 P&Attributes...B0 P&Help Retrieve Text Save Text As Import TextIMPORT FROM: SampleExit Text EditorTopStaff)Can'tWPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawing WPC} 2! pn!'3',,,,51+' ! cX 6Times New Roman Regularc @XXXyGzden " cX 6Times New Roman Regularc @XXXColor #B cX 6Times New Roman Regularc @XXX-br'1 $ > I 7 % ) & %)'1"'z cX 6Times New Roman Regularc @XXX?WusssssKuststssuVfFf~t'^&GW t&?u8Ƌ&_ff1jhU5S $V&6ɏRfvRP Ł4lPH=pu NM>$)523^%d2ɲh h|Xjj6G~~dž0+P5fhfo/ N^ ?______RFnSSn eiP&otFc:PPSB0c2 &n tl:PPSB0j- ot&iej!!TPbx08F7u~ ~ PCancel-0 P&Attributes...B0 P&Help Retrieve Text Save Text As Import TextIMPORT FROM: SampleExit Text EditorTopStaff)Can't METAFILEPICTGg >T  1!Times New Romanf-- .   --T$(---T%(--- ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !--->%      --%--%---($---(%--- !,- !+- !) !* !+ !( !+ !+ !' !+ !' !+ !' !+ !' !+ !( !+ !+ !) !+---6%++))'''()*++++++,----.,++--% ++++**)(((*+--- !< != !; !> !: !>- !:- !? !: !? !; !> !<---2%;;<=>??@??>=<:::;<=>@?;--% ;>>>==<;;--- !L !N !K !M !O !L !M !P !L !P !L !P !L !P !K !O !K !O---6%MNOPPQQQQQRROOOOOPPPPONNM--.%MMMMNNKKKKKLLLLKKKLMM-"System--Times New Romanf-- .  --D$ ---D% --- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---&%-- %---$$           ---$%           --- ! !  ! !  ! !  ! !  ! !  ! !  ! ! !---$%     --%    --->$--->%---Times New Romanf-- .  --$---%--- !- !- ! ! ! ! ! ! ! ! ! ! ! ! ! ! !---,%-- %--->$--->%---- %----- ! ! ! ! ! ! !!--Times New Romanf-- .  -- !' !& !) !% !) !% !) !% !) !) !(- !(-- !'----((--:%(()))''&&&&%%%'()*+++**))((--% (()))(''(---vj -Hfv fvv  $  C E $$EC--    1    ݀HX߀XXXH XHX߀Aninterestingquestionthenfollows:doclassnodesquarepresentationalentitiesplayanyroleinthetheory,  otherthanthatofcapturingfeatureclassbehavior?Assertionsthattheydoareuncommonintheliterature,thoughclassnodeshavebeencalledoninordertoexplainfactsofsegmenttransparencyandopacity,andtoilluminatefactsofplacefeatureinteractionmoregenerally.SeePadgett(1995a,b)forrespectivecriticaldiscussion.#HX߀XH ##X߀XXHX߀#((3$ g8!    C E $#hC--      0  (#$  0  kl&OLE 2.0 Box <=8C HKKKK( $ Figure  1  *OLE* ;;;g;d;D;;D;;;WPC hm CiL)pXXXX,,,,519DD<+'"!] i c 6Times New Roman RegularcXXX @[son]""]&iq c 6Times New Roman RegularcXXX @LaryngealPlace"#ZGkWc 6Times New Roman RegularcXXX @[asp][voiced][c.g.][lab][cor][dor]'1"$]~w~"%]r"&]   "'] 4  "(]4 [ YfmV&Times New Roman Regular Arial RegularACourier WP Regular96iZK F(G Ykt " 1]ėFl bM .e'Rm 4& t"h})RcG9G0Us36+Y!. pW4@Phh Xh~BJAm:D/D]&FP#F[KFFCF=*6uU=фL`r`$)`{ `v.F!ENrVr<2Rk%DaC9iu Gk(7KA_ AxmHH=J{.)  Xq.eH-h x{7['66^x.1=;9 o u)u .c)Uc$7 K:.u/*o vCu#]*Ep<]HKG)ޡL)u)0%dBh$::Ru v T#WX5)Vk PJŦfEucG}ujj ̋<>2GAPx#C1r1_1#kvLrЀ Y,,Q,^fK^(H"'''T(8!TA% P .,L3$4$$&$?$ @ K :L: r: 3R$j>hF**h  $  C E $$EC--    18    ݀ XXThisstatementmayoversimplify,ifthenasalinformslikeH  mkpai'libation'isasynchronicmarkerofnoun  class;inanyevent,assimilationdiffersaccordingtomorphologicaldomains.#HX߀XH #WPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@pj6_ Ole 9WPG20CompObj!^  "%&'()*+,-./0123456)F! WPWin61[6]ࡱWPC hm CiL)pXXXX,,,,519DD<+'"!] i c 6Times New Roman RegularcXXX @[son]""]&iq c 6Times New Roman RegularcXXX @LaryngealPlace"#ZGkWc 6Times New Roman RegularcXXX @[asp][voiced][c.g.][lab][cor][dor]'1"$]~w~"%]r"&]   "'] 4  "(]4 [ YfmV&Times New Roman Regular Arial RegularACourier WP Regular96iZK F(G Ykt " 1]ėFl bM .e'Rm 4& t"h})RcG9G0Us36+Y!. pW4@Phh Xh~BJAm:D/D]&FP#F[KFFCF=*6uU=фL`r`$)`{ `v.F!ENrVr<2Rk%DaC9iu Gk(7KA_ AxmHH=J{.)  Xq.eH-h x{7['66^x.1=;9 o u)u .c)Uc$7 K:.u/*o vCu#]*Ep<]HKG)ޡL)u)0%dBh$::Ru v T#WX5)Vk PJŦfEucG}ujj ̋<>2GAPx#C1r1_1#kvLrЀ Y,,Q,^fK^(H"'''T(8!TA% P .,L3$4$$&$?$ @ K :L: r: 3R$j>hF**h `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; CPࡱ; "` !U  "  QTimes New Roman[-- .  "SystemBOlePart#OlePres000$--2 &}[son]& -- .  Times New Roman-Times New RomanX ---32 k@Laryngeal Place  k--- .  --U2 4[asp] [voiced] [c.g.] [lab] [cor] [dor]"-- %]sI--r^{-- %aZ,[-- %p--p-ࡱ; ࡱ;  METAFILEPICT"h" !U  "  QTimes New Roman[-- .  "System--2 &}[son]& -- .  Times New Roman-Times New RomanX ---32 k@Laryngeal Place  k--- .  --U2 4[asp] [voiced] [c.g.] [lab] [cor] [dor]"-- %]sI--r^{-- %aZ,[-- %p--p-odw2(9 Z 6Times New Roman Regular*OLE*WPChjr$pXXXX,,,,519V+'"! jl c 6Times New Roman RegularcXXX @[+son][-son]""f jo c 6Times New Roman RegularcXXX @[lab][dor]'1&"#r " f   !p@(12Ń.n%HZQ-.h jh*TB:DG0G/1AoG W eG~W u(C"f+@XG /( ǃ] V&*_#b 0K [YpvKiD_#7FoIT$[n*jrV&0AT 5Ɂp-JzW#[%[[[2BA\@<]]]茽?]]~t_ V ~ k C|EIzkCDQ+N&O Fk GJz](]]^u%0f $ L*Ht,HIWPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ; Root Entry`.@pjdOle 9WPG20CompObj^ )F! WPWin61[0]ࡱWPChjr$pXXXX,,,,519V+'"! jl c 6Times New Roman RegularcXXX @[+son][-son]""f jo c 6Times New Roman RegularcXXX @[lab][dor]'1&"#r " f   !p@(12Ń.n%HZQ-.h jh*TB:DG0G/1AoG W eG~W u(C"f+@XG /( ǃ] V&*_#b 0K [YpvKiD_#7FoIT$[n*jrV&0AT 5Ɂp-JzW#[%[[[2BA\@<]]]茽?]]~t_ V ~ k C|EIzkCDQ+N&O Fk GJz](]]^u%0f $ L*Ht,HI `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; PPeeࡱ; "p ` %  p "  fQTimes New Roman-- .  "SystemBOlePartOlePres000--%2 ![+son] [-son]  -- .  --2 V1[lab] [dor]V--%8!@D|"qD9 -ࡱ; ࡱ;  METAFILEPICT"h"p  %  p "  fQTimes New Roman-- .  "System--%2 ![+son] [-son]  -- .  --2 V1[lab] [dor]V--%8!@D|"qD9 -(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular  $  C E $$EC--    7    ݀ XXAsaseparateissue,anytheorymustaccountforthespreadinghereof[dorsal]inparticular,ratherthan  [labial];seesection4.#X߀X #*OLE*(9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular*OLE*WPChvjr%pXXXX,,,,5198DD+'"! jl c 6Times New Roman RegularcXXX @[+son][-son]""f bo c 6Times New Roman RegularcXXX @က[dor][lab]'1 $W`~W|?WPWin 6.0/OLE 1.0 Prefix Information Marker WPDraw30.Drawingࡱ; Root Entry`.@pPdOle :WPG20CompObj^ *F! WPWin61[10]ࡱWPChvjr%pXXXX,,,,5198DD+'"! jl c 6Times New Roman RegularcXXX @[+son][-son]""f bo c 6Times New Roman RegularcXXX @က[dor][lab]'1 $W`~W|? `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; CPaeࡱ; "p  ..  p "  fQTimes New Roman-- .  "SystemBOlePartOlePres000--%2 ![+son] [-son]  -- .  --.2 W2 [dor] [lab]W-- %<!ZF|!C-ࡱࡱ;  METAFILEPICT""p  ..  p "  fQTimes New Roman-- .  "System--%2 ![+son] [-son]  -- .  --.2 W2 [dor] [lab]W-- %<!ZF|!C-WPChj*pXXXX,,,,519DD+'"!p j7 c 6Times New Roman RegularcXXX @[petroly󀀀G]""f j, c 6Times New Roman RegularcXXX @ [-b] 󀀀 [+b] [-b]"# jc 6Times New Roman RegularcXXX @က[r]'1"$6^ 6Y "%^ Y "'d  d  ( )y(TRE 0 DWPWin 6.0/OLE 1.0 Prefix Information MarkerWPDraw30.Drawingࡱ;  Root Entry`.@p#ʯe Ole 9WPG20CompObj^  !"#$%&'())F! WPWin61[3]ࡱWPChj*pXXXX,,,,519DD+'"!p j7 c 6Times New Roman RegularcXXX @[petroly󀀀G]""f j, c 6Times New Roman RegularcXXX @ [-b] 󀀀 [+b] [-b]"# jc 6Times New Roman RegularcXXX @က[r]'1"$6^ 6Y "%^ Y "'d  d  ( )y(TRE 0 D `.@pWP Draw 3.0 Drawing Embed SourceWPDraw30.Drawingࡱ; PPࡱ; "@ *  "  QTimes New Roman0-- .  "SystemBOlePartOlePres000h--*2 F6[ p e t r o lFTimes New Roman---- 2 @y@---2 F ]F -- .  Times New Roman-Times New Roman--- 2 K[-b] d---2 d |--- 2 |[+b] ---2  [-b]-- .  --2 x| [r]x--7V#V--7#-- %KgK--6$--%7-ࡱࡱ;  METAFILEPICT"H" *  "  QTimes New Roman0-- .  "System--*2 F6[ p e t r o lFTimes New Roman---- 2 @y@---2 F ]F -- .  Times New Roman-Times New Roman--- 2 K[-b] d---2 d |--- 2 |[+b] ---2  [-b]-- .  --2 x| [r]x--7V#V--7#-- %KgK--6$--%7- w $  C E $$EC--    8    ݀ XXForexample,Hallepositsaseparaterulespreading[back]fromtheinterveningconsonantinBarra;[back]  spreadingfromlyinTurkishinsteadfallsoutdirectlyfromarequirementofalignmentofColortotheprosodicword  edge,andsoisunifiedwithharmony.WhetherthisdisadvantageisintrinsictotheframeworkassumedinHalle(1995)isunclear.Asaseparatematter,Halle(1995)allowsindividualfeaturetargetingonlyforassimilations(p.20);FCTisintendedtocoverallfeatureclassbehavior,involvingdissimilations,neutralizations,andsoonaswell.#X߀X # R $  C E $$EC--    9    ݀ XXPositiveformulationsareclearlyrequiredinordertoaccountforcasesinwhichthecomplementary  unlicensedenvironmentdoesnotconstituteanysortofdomainatall.Wordandfootinitialpositionsareoftensingledoutasregionsoflicensing(seeSteriade1995a);thecomplementaryenvironmentsareundefined.#X߀X #odw18  $  C E $$EC--    10    ݀ XXSteriade(1993a,1994)instantiatesreleaseasapositionalentityintherepresentationakintoaRootnode,a  viewabandonedinfavorofdirectappealtotheperceptualbenefitsofreleaseinSteriade(1995b).#X߀X #jamie1odw17odw19  $  C E $$EC--    11    ݀HX߀XXXH XHX߀Steriade(1995b)arguesformoresubtlephoneticallybaseddiscriminationsbasedoncasesoflaryngeal  neutralization,andseeSelkirk(1982)foradetaileddiscussionofthefactsofEnglishphoneticrelease# H #.Itseemsclear  that[release],likeallfeatures,thoughtiedtophoneticfacts,mustalsoabstractawayfromthemtosomedegree.#X߀X #odw12odw20  $  C E $$EC--    13      Ԁ XXSpreadingofconsonantalplacefeatureswillbeconfinedtoaclusterofstrictlyadjacentconsonants,for  instance,giventwoassumptions:spreadingcannot'skip'avowelsegment(seeArchangeliandPulleyblank1994,It=,MesterandPadgetttoappearonlocality),andspreadingcannottargetavowelsegmenteither.Thelatterfollowsfrom ` theunityofplaceandoralstricture(Padgett1991,1994,1995c,cf.thegestureofBrowmanandGoldstein1989): (x spreadingastoparticulationtoavowelwouldillicitly'harden'thevowel(seeNChiosinandPadgett1993).#X߀X #TABLE ATABLE ATABLE Aodw25Level 1Level 2Level 3Level 4Level 5Level 1Level 2Level 3Level 4Level 5Level 1Level 2Level 3Level 4Level 5Level 1Level 2Level 3Level 4Level 5ladida  $  C E $$EC--    14    ݀ XXMoreaccurately,monosegmentalNCexistsonlyasaphoneticrealizationofvoicingorperhapssonorancyin  languageshavingnonasalvs.voicedobstruentdistinction,asubgroupwearenotconcernedwith(seePiggott1992).#X߀X #odw37odw36TABLE A(;s$22,  0  .3  0  (O{$0  22,  a  .3  0` (#(#(b$0  0` (#(#22,   .3  0 ` (#` (#(x$0  0` (#(#0 ` (#` (#22,(  0  )3  0 (# (#($0  0` (#(#0 ` (#` (#0 (# (#22,(  a  )3  0h(#(#($0  0` (#(#0 ` (#` (#0 (# (#0h(#(#22,(   )3  0h(#h(#(4$0  0` (#(#0 ` (#` (#0 (# (#0h(#(#0h(#h(#22,  0  )3  0(#(#($0  0` (#(#0 ` (#` (#0 (# (#0h(#(#0h(#h(#0(#(#22,  a  )3  0p(#(# y $  C E $$EC--    15    ݀ XXHalle(1995)arguesthatclickshaveonlyone'designatedarticulator',[coronal]inthecaseofdentalsand  palatals,[dorsal]foralveolarsandlaterals.Hespecifiesallclicks[+high]though,andgivesallnonlabialclicksan[anterior]value,thusmakingallclickscomplexinessence.Thedatabelowshowthat[dorsal]isanelementofallclicks.#X߀X #odw28  $  C E $$EC--    16    ݀ XXInbothclicksandlabiovelarsthevelararticulationprecedestheotheroneintime.Atleastinlabiovelarsthis  leadisveryslight,however,andisalesslikelysourceofexplanation#X߀X # XX߀forsimpleH  underassimilation,accordingto  OhalaandOhala.#HX߀XH #  $  C E $$EC--    17      Ԁ XXRecasensdescribest/dasapicodentalsincontrasttoMascar;'s[+dist]specification.Healsoindicates  variationbetween[+dist]and[dist]fors/z.#X߀X #jamieodw30odw31odw32odw38TABLE A , $  C E $$EC--    20    ݀ XXItispossibletoenvisionamoregeneralconstraintSpreadatworkhere,requiringfeaturesofanyclasstobe  maximallylinked,solongasotherclassspecificconstraintscancurtailtheresulttoplacelinkingasrequired.ThenspreadingisperforcelesstotalinG .#X߀X # g8!    C E $#hC--  o[Xr hp x (#%'0*,.8135@8:<H?AXo;7߀XXdd7;;  ;; XX;[Toappearin ;;ProceedingsoftheArizonaPhonologyConference:WorkshoponFeaturesinOptimalityTheory;;,  CoyoteWorkingPapers,UniversityofArizona,Tuscon.];; ; ;; ; ;;#X߀X o#; ; ; XX߀ԛ@ =PartialClassBehaviorandNasalPlaceAssimilation#X߀X# *g8!ٜ^ #  *      כ*g8!؈ (x @QJayePadgett@ZZEUniversityofCalifornia,SantaCruz̜;; 1.Introduction     HX߀XXX߀ r Thispaperhastwogoals.ThefirstistopursueandfurthermotivatesomeideasdevelopedinPadgett(1995a)onthestatusofthenotionfeatureclassinphonologicaltheory.Thesecondis   toexplorethephenomenonofplaceassimilation.Thesegoalsarebroughttogetherinanexplorationoffactsinvolvingnasalplaceassimilationtocomplexsegments:assimilationinsuchcasesisoftenpartialinthesensethatonlyonearticulatorofthecomplexsegmentspreadsunderageneralprocessofplaceassimilation.Aswewillsee,partialassimilationofthissortbearsinanimportantwayonourunderstandingoffeatureclasses. r ThenotionfeatureclassmakesgeneralreferencetofamiliarcategorieslikePlace, X Laryngeal,VPlace,andsoon,classificationsoffeaturesinstantiatedmostnotablyinworkwithin D FeatureGeometrytheory(henceforthFG,Clements1985).FGseekstoexplainrecurrentcopatterningsoffeaturesaccordingtothesecategories,bymeansoftherepresentationaldeviceclassnode:constraintscantargetanodePlace(forexample),andtherebyindirectlyaffectall  dependentfeaturesasagroup.Padgett(1995a)exploresanalternativeunderstandingoffeatureclasses,calledFeatureClassTheory(FCT),inwhichclassesaredefinedinsteadassetsoffeatures(havingnotreetheoreticinstantiation).Featurespatternintosetsaccordingtopropertiessharedbytheirmembers"placehood,laryngeality,andsoon.Constraintscanmakementionofthesesets,sothatthecentralinsightofFGispreserved,buttheytherebytargetindividuallythe T memberfeatureshavingtherightproperties. r FCTfindskeymotivationinacertainbrandofpartialclassbehaviorfirstnotedbySagey !h (1987)undertherubric'nonconstituent'behavior.ThephenomenonisillustratedbelowwithanexampleinvolvingTurkishandthefeatureclassColor={back,round}.Theexampleand "@! argumentcomefromPadgett(1995a);ontheclassColorseealsoOdden(1991)andSelkirk #," М(1991a).Turkishvowelharmonyamountstoanimperativethatthisclassoffeaturesspanaword,asshownin jamie1(1)  a.(ForthefactsofTurkishvowelharmonyandaclassicalautosegmentalanalysisseeespeciallyClementsandSezer1982.)However,harmonycannotestablishalinkbetween[round]andanonhighvowel,duetoanindependentsegmentalmarkednessconditiondisfavoring '"& suchlinkage.Resultingfromthisprohibitionareinstancesofpartialclassbehavior,inwhichonlyonememberofthetargetedclassColor"namely[back]"conformstotheharmonyimperative,  shownin jamie1(1)  b.(bandrare[back]and[round]respectively.)  & 8 2(  1  )3    Ԁ r Full(a)andpartial(b)classbehaviorTurkishandColor ` gi3#` L"apZ EL "a}(a g L   r '8 `SuchpartialclassbehaviorissurprisingfromtheperspectiveofFG,becauseofthewaythatfeatureclasseslikeColorarerepresentedandunderstood.Theyareembodiedaselementsofa   treerepresentation,theclassnodes,uponwhichrulesoperatedirectly,affectingtherelevantdependentfeaturesonlyindirectly.Thesenodeshavealwaysbeeninterpretedtofunctioninanallornothingfashion:ifcolorharmonyresultsfromthelinkageofanodeColor,asin jamie2(2)  ,thenthe \  invariantexpectationhasbeenthatboth[back]and[round]willbelinked,orfailingthat,neitheronewillbelinked(unlessaseparateruleisinvoked).Suchatheorymakesnoprovisionfortheintermediatepossibilityofpartialclassbehavior#X߀XXHX߀#HX߀XXX߀.  p  2(  2  )3   D Ԁ r TotalclassbehaviorinFG:*yzdnor*yzdAn(noharmony) H  r gi3#` LpOz Er\g 4 Apparentlyforthisreason,theFGliteraturehasfailedtocallontherobustevidenceofTurkishandother(especiallyAltaicandUralic)colorharmoniesasevidenceforaclassColor󀄀though T  thisclasshasampleprecedentinphonologicaltheory,andhasbeenarguedforwithinFGitself(Odden1991).ThoughFGcanberecastsoastoaccommodatepartialclassbehavior,therequiredalterationessentiallystripstheclassnodesoftheirfunctionofcapturingfeatureclasses,andentailssomethingliketheFCTpointofview,aswewillsee. r FCTthereforeextendstheexplanatoryroleofthefeatureclassideatoempiricalareasotherwiseuntraversed.ThecaseofcolorharmonyoutlinedaboveisexploredindetailinPadgett(1995a).Thepresentpaperinsteadconcentratesonfactsofnasalplaceassimilationbyconsonants(henceforth:NPA),andconsidersarangeofpartialassimilationdatainvolvingdiversecomplexsegmenttypes.Thefirstgoalofthepaper,then,istodemonstratethepervasivenessofpartialclassbehaviorwithreferencetoanotherwellknownphenomenon,NPA,andsofurthermotivateFCT. d+& Ї r Explorationofthisterritoryrequiressomereckoningwiththefactsofplaceassimilationverygenerally,andatheoreticaldiscussionofthisphenomenonformsthesecondgoalofthepaper.Foremostamongthethemeshereistheimportantrolegiventothenotionsegmental  release(McCawley1967,Selkirk1982,Steriade1993ab,1994),seenhereasafeaturalproperty t thatlendsgreatprominencetoasegment'sotherfeaturaldistinctions.Suchprominenceexertsitsinfluencethroughreleasesensitivefaithfulnessconstraints;thisbuildsonthemoregeneralideaofpositionorprominencesensitivefaithfulness(Selkirk1994,Beckman1995).Itemergesfromthesefindingsthatpositionalcategorieslike'onset'and'coda'areoftenonlyindirectlyrelevanttofeaturelicensing,andhenceassimilations,contrarytotheprevalentview.Further,motivationisfoundforafeaturespreadingimperativeinassimilations,asurprisingfactfromtheperspectiveofaccountsappealingsolelytofeaturallicensinginordertoachievemultiplelinkageofplace. r ThisworkisinspiredbycentraltenetsofOptimalityTheory(henceforthOT,PrinceandSmolensky1993),callinginparticularonthoseofconstraintrankingand(minimal)violability.Therepresentationalissueaside,thecentralpointofFCTisthatconstraintsmentioningclasseslikePlaceandLaryngealaregradientlyviolable,aclaimthatinformstheupcomingaccountof   partialclassbehavior.SuchbehaviorensuesmostcommonlyunderthescenarioC>> p Constraint(Class),whereCisanyconstraint,Constraint(Class)isaconstrainttargeting \ anyclassoffeatures,andthetwoconstraintsconflictforsomepropersubsetofthesetClass.In H thecaseofTurkishabove,Cprohibitstheoutputof(noninitial)nonhighroundvowels,and 4 Constraint(Class)isaconstraintrequiringharmony(treatedasalignmentinPadgett1995a,  p andcalledAlign(Color));theseconstraintsconflictwithrespecttothefeature[round](aproper  \ subsetofColor)informswhereharmonywouldeffectlinkageof[round]toanonhighvowel. H Theharmonyimperativeisthereforeviolatedbutnotfully,since[back]conforms. r OTprovidesatheoreticalcontextinwhichFCTisessentiallynecessitated.Oncewecastaninformedeyeontraditionalrulespecificstipulations,factoringthemoutbyappealtotheinteractionofindependentconstraints,wefrequentlyfindthescenarioC>>Constraint(Class)  (understoodasabove)asamatterofcourse.Insomecasesthisfactoringwillpropelustounifyprocessesoncethoughtofasseparate,aswhen[back]and[round]harmoniesareunitedinColor l harmony(seePadgett1995a).Inallcases,includingthoseexploredbelowinvolvingPlace,the X  notionclassnodeisreducedincontenttomerefeatureclasslabel,doingnorealworkinthe D! theory. #  1      כ 0"   r Section2belowexplicatestheessentialpropertiesofFCTandarguesforitonthebasisofthepartialNPAphenomenon.Insection3weturnourattentiontothetaskoffindingageneral accountofplaceassimilationsinOT.#X߀XXHX߀k#HX߀XXX߀ԛSection4providesempiricalsubstanceforunitingthese %0!$ separateexpositionalstrands,brieflysurveyingarangeofdatainvolvingpartialNPAtovariouscomplexsegmenttypes,andanalyzingacasefromG ,callingonideasjustdeveloped. &  2.FeatureClassTheory  t  2.1CapturingFeatureClasses  L   r Itisnowafamiliarobservationthatsomephonologicalfeaturespatterntogetherrecurrently't+inprocessesofassimilation,dissimilation,neutralization,andsoon.FeatureclassessimilartoLaryngealandPlaceachievedaformalstatuswithingenerativephonologyfirstintheworkof  L  Dependencyphonologists(LassandAnderson1975,Lass1976,AndersonandEwen1987),andlatermostprominentlyinFeatureGeometrytheory(Clements1985,cf.Mascar;1983,Mohanan1983).WithinFG,thefeatureclasspromiseisfulfilledbymeansofontologicalentitiesknownasclassnodes,incorporatedintoatreerepresentation;thesearedirectlytargetedbyphonologicalrules/constraints.ThefollowingFGrepresentation,apareddownadaptationfromMcCarthy(1988),servestoanchorthediscussion.& W 2(  3  )3  Ԁ r TheclassesPlaceandLaryngealinFG \ Ѐ r o;+'`|2h%g ph  ErIg gl ' o H #X߀XXHX߀)#'W \.ClassnodesinFG(likethoseitalicizedabove)arecentraltothattheory,servingaskeymediators  offeatureclassbehavior.ThusaruledemandingthespreadingofPlaceisunderstoodtotarget  therelevantclassnode;entailmentsattributedtofeaturaldependencyensurethattheactualplacefeaturesthemselveswillbeaffectedatthesametime.Theselatterfeaturesarethereforeaffectedonlyindirectly.Wewillreturntothisfactandexploreitssignificanceinsection2.2. r FCTisaless'syntactic'andmore'semantic'approachtothefeatureclassinsight:termslikePlaceandLaryngealstandforsetsoffeatures,asshownbelow.Thesesetshavetheirbasisinthe ;%  propertiesthattheirmembersshare:here,placehoodandlaryngealityrespectively.(Foraprecedentinvolvingsyntacticfeatureclasses,seeGazdaretal.1985:23.)2(  4  )3    Ԁ r TheclassesPlaceandLaryngealinFCT (;$  r Place=def {labial,coronal,dorsal...} *&   r Laryngeal=def {aspirated,voiced,constrictedglottis} +&! ЇNaturally,constraintsmustbeabletorefertotheseclasses;FCTmaintainstheimportantfeatureclassinsightofFG.G #  2      כHowever,therearenownoobjectsintherepresentation(classnodes)to  Мmediatefeatureclassbehaviorinthewayseenin Djamie2(2)  .Byhypothesiswearerevertingtoasimplifiedstructuremostreminiscentof'bottlebrush'theories(seeHayes1990onthesetheories).Intherepresentationbelow,featureorderisrandomizedtoemphasizethepoint:classesareunderstoodassetsoffeaturesthatsharetherelevantproperty.Theintuitionwearedevelopingisstraightforward:constraintsmentionclasseslikePlaceandLaryngeal,buttheytherebytarget, 8  directlyandindividually,thefeaturesintherespectiveextensionsoftheseclasses,definedin  odw1(4)  .& * 2(  5  )3  Ԁ r FCTrepresentation  L   r m9)%`|2 `hSEr hShS' m  8  '*u9 r Forthesakeofdiscussion,letusconsideraprovisionalconstraintdemandingNPA,intheformulationgivenbelow.Thisformulationisatemporaryconvenience;insection3wewillinquiremoreseriouslyintothenatureofthephenomenon.<,Xr` XXr X<2(  6  )3   I Ԁ r Npa:0 ` IneverysequenceNC,everyPlacelinkedtoCislinkedtoN,andviceversa:` (#` (# @Xr X,Xr` X;@Thisconstraintmustbeunderstoodtobearonthefeature[labial]intherepresentationin  odw2(7)  ,byvirtueofthisfeature'sstatusasaplacefeature.Moregenerally,itmustrequiredoublelinkageofeveryindividualfeatureofthesetPlaceintheright(NC)configuration.  2(  7  )3    Ԁ r [labial]targetedasaPlacefeature r  r    [+son][son] r   /\/ r   [+nas][labial]  AnunderstandingoffeatureclassbehaviormoreorlessalongtheselinesisadvocatedinSelkirk(1991a,b)andPadgett(1991);Hayes(1990)representsamoveinthesamedirection,maintainingtheclassnodesofFGwhileessentiallymotivatingindividualfeaturebehaviorinvarious'diphthongizationparadoxes'.Sagey(1987)arguesforalimiteduseofindividualfeaturespreadinginordertocaptureclassbehavior,viewingitasamarkedoptionwithinFG;Halle(1995)picksup Sagey'sargument(thoughcf.NChiosin1995),butadvocatesindividualfeaturetargetingasthe &!  onlypossibilityinthetheory,aswedohere.Differencesamongthesevariousproposalsforindividualfeaturetargettingarediscussedinthenextsection.  r Inordertoensureclassbehaviorasinformallyillustratedabove,wesimplycapitalizeonthepostulatedsetsratherthanonpostulatednodes.TermslikePlace,LaryngealandColorare ` convenientstandinsforthesesetsinconstraintformulations.However,alookatvariouspossibleconstrainttypesrevealsafactworthkeepinginmind:differentconstraintswillquantifyoverthesesetsindifferentways.HX߀XXX߀ԀThustheintentofourconstraintNpaistoquantifyuniversallyoverthe $ t placefeatures,andarenditionwiththerelevantportionformalizedmightbethefollowing(wherexrangesoverfeatures,andF(S)means'featureFiscontainedinsegmentS',i.e.,dominatedbythe  L  sameRootnode=[son]feature):#X߀XXHX߀D#HX߀XXX߀  8  2(  8  )3   _ Ԁ r Nparedux: ForeverysequenceNC,zx,xPlace#X߀XXHX߀,F#HX߀XXX߀,x(N)<x(C)   #X߀XXHX߀6G#HX߀XXX߀Itshouldstillbeclearthatthisconstraintissatisfiedbytherepresentationin  odw3(7)  .Wewillfurther   exploreitsinterpretationmomentarily.First,toseehowadifferentquantificationalforcecanarisefromanotherconstraint,considertheformulationbelowofaconstraintbanningplacelesssegments,againwiththeportionofinterestformalizedbelowit(seeIt=andMester1993,Lombardi1995a,amongothers,onsuchaconstraint).2(  9  )3    Ԁ r HavePlace: EverysegmentmusthavesomePlace  p  r    ForeverysegmentS,yx,xPlace#X߀XXHX߀G#HX߀XXX߀,x(S)  \ Thisconstraintrequiresexistentialquantificationoverthefeatureclass.Thisistheresultwewant; 4 wedonotintendthateveryRootmustdominateallplaces,forinstance.Oneneednotendorsetheparticularconstraintsjustconsideredinordertoseethepoint:constraintswilldifferinthesortofquantificationalrestrictionstheyplaceontheirvariables.TermslikePlaceandLaryngealinan  informallystatedconstraintactuallystandinconvenientlyforthesevariablesandtherelevantsets.W #  3        #X߀XXHX߀J# r ReturningtoourprovisionalNpa,letusrunitbyafewmorerepresentationsfor X  completeness.Thefollowingsummaryratesthesuccessofseveralrepresentationswithrespecttothisconstraint,andprovidesexplanatorycommentary.Therepresentationshavebeenminimized:L,C,Dstandfor[labial],[coronal]and[dorsal]respectively.  $X" ;&   ;2 (  10  )3   V Ԁ r HowNPAratesrepresentations   r   Representation  Npa h   Commentary    r a.  pagur   3 h   OneNCsequence,oneplace,doublylinked. ` ;'O; r   \/ r   D&   r b.  mpagur  3 h   IneveryNCsequence,everyplaceisdoublylinked.  `  r   \/\/ r   LD' Q&   r c.  Npagur  * h 0  InoneNCsequence,oneplaceisnotdoublylinked. (#(#  r   \/    h   (HavePlaceisalsoviolated.)    r   LD'R r d.  npagur   ** h   InoneNCsequence,twoplacesarenotdoubly p  r    \/    h   linked.(Notethebiconditionalin _ noise(8)  .) \  r CLD r e.  npamgur  **** h   IntwoNCsequences,twoplacesarenotdoubly  p  r          h   linked.  \  r CLLD r f.  pagur   3 h   Constraintvacuouslysatisfied.   Ascanbeseenfromtheviolationsmarkedandthecommentary,Npacanbeviolatedmorethan  onceinarepresentation.Theconstraintistreatedasgradient"itisworsetohavefourplacefeaturesintherelevantconfigurationfailtobedoublylinkedthantohaveonlytwoplacefeaturesfailinthisway, V odw5(10)  dvs. V odw5(10)  e,andsoon.Thatis,everyx,xPlaceisnotinterpretedinthe l moststraightforwardtruthconditionalway,suchthatone,ormany,failurestranslatessimplyas'false'.GradienceofthissortisnowfamiliarintheOTliterature;forgeneraldiscussionseePrinceandSmolensky(1993),andMcCarthyandPrince(1993a).Gradiencewillbeimportanttoourunderstandingofpartialclassbehaviorlater. r Thoughwewillsoonreviseourunderstandingofnasalplaceassimilation,thediscussionhereillustratesverygenerallyhowanyconstraintmentioningafeatureclassistobeconstructedandunderstood.Beforewemoveon,itisworthmakinganotherpointconcerningFCT.LikeFG,FCTpostulatespartitioningsofsetsoffeatures;thus,theclassesPlaceandLaryngeal,for '#&  example,donotintersect.Thismoveaddressesanempiricalclaimthatnofeaturecouldbelongto (#' bothclasses;ifcorrect,aformalstatementtothiseffectisrequiredineitherFGorFCT.F #  4      ׀Asfor   substantivequestionslikewhynonintersectionseems(atleastlargely)correct,orwhythesetsaredivideduppreciselyastheyareandnotsomeotherway,thegeneralanswerisagaintheoryneutral:featureclassesseemtohavetheirbasisinphoneticparameters,whetherarticulatoryoracoustic.(SeeforexampleClements1985,Sagey1986,McCarthy1988,Padgett1991,1995candZsiga1993onthis.)ThuswhiletheextensionofthesetPlaceisamerelist{labial,coronal, L  dorsal...},theintensionisrevealedbythephonetictermPlaceitself.Presumablyafuller 8  understandingofthisbasiswillshedsomelightontheissuesraisedhere. 2.2PartialClassBehavior   L   r SupposewecompareaccountswithinFCTandFGforastraightforwardcaseofNPA,suchasthatofSpanish(seeespeciallyHarris1984forananalysisanddiscussionoftheSpanishfacts).NPAobtainswithinwordsinSpanish,acrossallplaces,givingformslikethefollowing:2 (  11  )3  Ԁ r NPAinSpanish    r sie[m]pre 'always' h   pre[n]sa p 'press' \  r [b]fora 'amphora' h   co[;)n``0;;)n$``0;HX߀XXX߀]cha p 'shell'#X߀XXHX߀{c# H  r a[)n(``0HX߀XXX߀]#X߀XXHX߀d#dar 'towalk' h   []gel p 'angel' 4 ̛AssimilationinvolvesattheleastallofthemajorarticulatorfeaturesrequiredforSpanish,[labial],[coronal]and[dorsal],andsothegeneralizationisoverthecategoryPlace.Inmanyinstances, H includingthisone,FCTandFGwillbeempiricallyindistinguishable.Theformerwilltargeteverymemberofasetoffeatures,asoutlinedinthelastsection;theresultwillbelinkageofanysuchfeaturefoundinaparticularrepresentation,asshownin  odw7(12)  a.FGinsteadtargetsanodePlacein   therepresentation,demandinglinkageofthisentity,  odw7(12)  b;thelogicofdependencyentailsthatanyfeaturesdependentonPlacewillbeinterpretedasextendingoverbothsegmentsnowalso. #  5      כ     & `  2 (  12  )3  Ԁ r PlaceassimilationinFCTandFGtworepresentationsof[nt]     r a.[+son][son] h   b.[+son][son]  Ѐ/\//\/̀[+nas][cor][+nas]Placè  L  Ѐ[cor]'`yhOnebenefitofFGnotedbySagey(1986)involvesitsstraightforwardhandlingofassimilationtocomplexsegmentslike[gb]ofKpelle.LikeSpanish,KpelleevincesNPAacrossallplaces,e.g.,[m]bolu'myback',[n]dia'mytattoo',[pX߀XXX߀]g<<X߀XXpX߀ԛ'myfoot'(Welmers1962).Tonesareomittedfrom  8  thecitedforms;theassimilatingnasalissyllabic.AsistrueinmanyWestAfricanlanguages,assimilationto[gb]yields[nX߀XXX߀,mg,b],ahomorganicsequenceofadoublyarticulatednasalfollowed   bythedoublyarticulatedstop,e.g.,[X߀XXnX߀m]gbiX߀XXX߀'myself'.If[gb]isasinglesegmentrepresentedasin    o odw8(13)  a,thenlinkageofthenodePlacewillentailthisfactwithoutfurtherado, o odw8(13)  b.   & ` 2 (  13  )3   o Ԁ r AssimilationtocomplexsegmentinFG p  r a.[son]    h b.[+son][son] H  r       h   \/ 4  r   Place    h   Place  p  r   /\    h   /\  \  r [lab][dor]  [lab][dor] H '`poFCTretainsthisbenefit,ifNPA(howeverconceived)requireslinkageofallmembersofPlacein   theNCconfiguration,assuggestedinourprovisionalconstraintNpaabove.Inthiscasethe   relevantfeaturesaretargeteddirectly,byvirtueoftheirstatusasplacefeatures,withoutthemediationofamothernode:&  2 (  14  )3  Ԁ r IndividualfeaturetargetingviathefeatureclassPlaceinFCT l  r m9)%`|2 ! `hErX hh' m X  'ls r However,anotheroutcomeofNPAbeforecomplexsegmentsisobservedinsomelanguages.Deferringfulldiscussionofthefactstosection4,wesimplynotetheresultheretopursuethepoint:assimilationto[gb]cangivesimple[X߀XXX߀],asinHX߀XXX߀G #X߀XXHX߀Mv#,anotherlanguageofWest 'n"! Africa.Thisdifferenceiscontrastivewithinthelanguage:NPAtolabiovelarswithinmorphemesistotalinG ,e.g.,[X߀XXX߀m]kpai'libation'.Acrossamorphemeboundary,incontrast,weseethepartial (F$# assimilation,e.g.,[X߀XXX߀]kpai'mycheeks'(Kropp1966,Kotey1974,Ryder1987).Here,asin )2%$ Kpelle,theassimilatingnasalissyllabic.Theoutcome[X߀XXX߀gb]isanexampleofpartialclass *&% behavior,shownbelow. + '& Ї2(  15  )3   P ԛ r PartialNPAto[gb]inFCT  HX߀XXX߀ r m9)%`|2 ! `hErhh'Qm  AssimilationispartialhereinmuchthesamewayitispartialinthecaseoftheEnglishprefixin,   whichgivesi[m]portantandi[n]delibleetc.,butfailstoassimilatebeforevelars,asin v  i[ n ]competent(carefulspeech).InbothcasesNparequiresNCdoublelinkageofeveryplace b  featureinanNCconfiguration.ThisconstraintisviolatedinEnglishbyformssuchasϜincompetent,duetoahigherrankingmarkednessprohibitionagainst#X߀XXHX߀y#HX߀XXX߀.InG wesimilarlyseea >  violationoftheconstraint,dueagaintoreasonsofmarkedness"abanondoublyarticulated#X߀XXHX߀|#HX߀XXX߀m *z  (seebelow).InneitherlanguageisNpaviolatedcompletelyorwantonly;infactitisotherwise f  obeyed.However,whileFGhasnoproblemwithpartialclassbehaviorlikethatofEnglish,thatofG causestrouble.Thereasonissimple:inG oneneednotlookacrossformsorevendifferentNCclusterstoseethatassimilationispartial.RatherwefindpartialclassbehaviorarisingwithinasingleNCclusterin P odw9(15)  :some,butnotall,ofthePlacefeaturesofgbspread.Partialclass   behaviorinsuchcasesfindsnohomeinFG(seealsoPadgett1995aonthispoint).BecauseinFGconstraintstargetactualplacefeaturesonlyindirectly,viaamotherPlacenodeanddependency  entailments,weessentiallyexpectoneoftwooutcomes:eitherallplacefeatureswillassimilate,or,ifassimilationisimpossible,nofeatureswillassimilate.Theseoutcomesarecontrastedbelow.& ` 2(  16  )3  Ԁ r NopartialclassbehaviorinFGconstraints/rulestargetnodePlace N  r a.[+son][son]   h   b.[+son][son] &v  r   \/    h    b  r   Place=#X߀XXHX߀}#RX߀XXX߀mgbPlace=Ngb N  r   /\    h      p /\ :  r   LabDorLabDor'`Nȁ̜Incontrast,acentralclaimofFCTisthatconstraintsmentioningfeatureclassesaregradientlyviolable.FCTmakesnoformaldistinctionbetweentheEnglishandG cases,allowingtheirsimilaritytoemergenaturally.Thetableauin U odw10(17)  Հanticipatesthefullerdiscussioninsection4.Assumingthatthehigherrankedconstraintrulingoutthedoublyarticulatednasalisoneofsegmentalmarkedness,called*CompSeg('nocomplexsegments'),candidate U odw10(17)  aiseliminated ^$ fromconsideration.YetonlythiscandidatefullysatisfiesNpa,andsothislatterconstraintmust J%  beviolated.Violationisnottotal,however,asincandidate U odw10(17)  c;thistotalfailureofassimilationisunmotivated(andindeedwillnormallyviolateanotherimportantmarkednessconstraint,HavePlace).TheoptimalcandidateviolatesNpaminimally,thatis,onlytotheextentrequired (^#" bythepositedrankingofconstraints.ThisunderstandingofthefactsthereforerestsontwoimportanttenetsofOptimalityTheory(PrinceandSmolensky1993):constraintsareviolable,andviolationisminimal. +'& 2(  17  )3   U Ԁ r Npaisgradientlyviolable  *wSMTddd Xdd Xdd X(#(#w,dd ,dd ,dd +    /N+gb/ ,Q< ,*CompSeg 4*<"Q 4Npa !Q<" !a.mgb 0QQ!lQ 0М**! ?Qx*l"QQ ? +!l"Qx +b.gb #Q  <  #М* 4* < "Q 4*  < " c.Ngb #Q l   #М* 4*l  "Q 4**!"l "  "Itiseasytoseethatunderthereverseranking,Npa>>*CompSeg,thefullyassimilated t  candidate U odw10(17)  awouldemergeasoptimal.Theanalysisthereforecorrectlypredictstheexistenceofbothfullandpartialassimilation,anddoessowithaminimumofassumptions:markednessconstraintsagainstcomplexandplacelesssegmentsarerequiredbyanytheory,asissomeaccountofNPA.Partialassimilationfallsoutforfree,giventheexistenceofconstraint(re)ranking,thequitegeneralmeansoflanguagevariationinOT. r ThefailureofFGresidesinitsunderstandingoftherepresentationalentityclassnode:this 8 entityanditsinterpretationtogethertranslateessentiallyintoahard,'sourgrapes'constraint:ineffect,eitherassimilateallfeatures(ofsomecategory),orassimilatenone.Itisworthemphasizingthatitisthisinterpretationoftherepresentationthatisattheheartofthefailure,ratherthantheϜrepresentationitselfconsideredinatheoreticalvacuum,ifsuchathingispossible.Indeed,Halle(1995)arguesforindividualclassfeaturespreadingwhileretainingclassnodes,seebelow.Theargumenthereisagainsthardorsourgrapesfeatureclassgeneralizations,inwhateverguise,infavoroftheirsoftunderstandingillustratedabove.s #  6      כFGisindeedsucha'hard'theory.Presenting \ thefactsofKpelle,inwhichassimilationtolabiovelarsistotal(seeabove),Sagey(1986:101)providesatellingquote:"Giventherepresentationof/kp/...therewouldbenowayforplaceassimilationspreadingtheplacenodetospreadjustthedorsal,orjustthelabial,articulation."ClearlySageyisnotaloneinthisunderstandingofFG;inthewakeofherimportantwork,therelativelywidespreadphenomenonofjustsuchpartialassimilationshasgonelargelyunnoticed(thoughseeRyder1987);itishardtoavoidtheconclusionthattheoreticalexpectationshavethemselvesconstrainedperceptionofthedata.Infact,FGhasfailedtorevealsomegenuinefeatureclassgeneralizationsentirely,becauseofitspromotionofhardfeatureclassconstraints(seethediscussionofColorharmonyintheintroductionandinPadgett1995a).Thesefailings "& pointupthetwoedgednatureofthestrategybywhichrules(andmoregenerallyourunderstandingofphonology)followfrom(interpreted)representationsthemselves.  l% )  r Somepossibleobjectionsariseatthispoint.First,supposeFGiscorrecttoregardfeature classgeneralizationsascategoricalratherthangradientlyviolable.PerhapsNPAinG shouldbesupplementedwithanotherrule/constrainteffectinglinkageofthefeature[dorsal]injustthecontextN+gb.Thisviewmightshareoneaspectofouraccount:regularNPAisnotpossible t beforelabiovelarsforreasonsofmarkedness(thoughitturnsoutthatsomesuchlanguagesallowunderlying/m/tosurface,sothegeneralizationmighteludesomesurfacetruetheories).Either L  [dorsal]assimilationfollowsasa'repair'(Paradis1988)toavoidthesurfacingofaplacelesssegment,orelsearule/constraintof[dorsal]assimilationpreemptsNPAbytheElsewhere $ t Condition(Kiparsky1973,1982,cf.Scobbie1993).Itishardtoseetheappealofsuchanapproachoverthatadvocatedhere.Theactivityofaconstrainttargeting[dorsal]mustbeseparatelypostulated,havingnoformalconnectiontothemoregeneralconstraintthattargetsaclassPlace,HX߀XXRX߀andsoseemspurelycoincidental.#RX߀XXHX߀*#ԀAswesawaboveincontrast,aformlike[X߀XXRX߀gb]falls $  withinthescopeofageneralNpaintheFCTaccount,giventheinteractionbetweenthis   constraintandthatofmarkedness. #  7          r ConsiderinsteadanattemptwithinFGtoembracepartialclassbehaviorofthiskind,bypostulatingthefollowingprincipleguidingfeatureclassgeneralizations:rX߀XXX߀#X߀XXrX߀#rX߀XXX߀& x 2(  18  )3   t  r FGwitha'resort'strategy H  r #X߀XXrX߀#rX߀XXX߀TotargetasetoffeaturesClass,#X߀XXrX߀Ѡ#rX߀XXX߀  p <,Xr` XXr X< r   a. ` TargetthenodeClass;#X߀XXrX߀<#rX߀XXX߀ unless targetingthisnodewillfail,inwhichcase:#X߀XXrX߀#rX߀XXX߀ԛ H  r   b.0 ` TargetindividuallyasmanyfeaturesdominatedbynodeClassaspossible8` (#` (# #X߀XXrX߀#@Xr X,Xr` Xq@'x Whetherunderstoodprocessuallyornot,anyresortstrategyofthiskindhasthe'dothisunlessthatinwhichcase...'qualitythatOptimalityTheoryseekstoredress,andseemsunappealingonthefaceofit.Infact,onceweallowforrX߀XXX߀  egg1(18)  b,thetheoryhasnoplacefor  egg1(18)  a:surelya  constraintalwaystargetsasmanyfeaturesaspossible.FGrecastinthiswayisnowanungainlykintoFCT.#X߀XXrX߀# p  r Finally,wemightpreservetheFGrepresentationwhileabandoningitstraditionalinterpretation,incorporatingsomethinginthespiritofrX߀XXX߀  egg1(18)  bintothetheory,anddispensingwith#X߀XXrX߀# 4"  rX߀XXX߀  egg1(18)  a,#X߀XXrX߀;#HX߀XXX߀asinHalle(1995),whobuildsonSagey(1987)#X߀XXHX߀#rX߀XXX߀.#X߀XXrX߀#ԀThisviewshareswithFCTtheassumption  #p! thatfeatureclassconstraintstargettherelevantfeaturesdirectlyandindividuallyasamatterofcourse.Theclassnodesarenowreducedincontenttothestatusoffeatureclasslabels,havinga $H # functionsimilartothatattributedtosetsinFCT.TheversionofthistheorysuggestedbyHalle(1995)differssignificantlyfromFCT,however.First,FCTisexplicitlycommittedtotheideaofgradientviolabilityofconstraintsatthesurface,anditiscouchedthereforewithinOT.Halle(1995)isinsteadcastwithinaframeworkemployingorderedrules(asinBrombergerandHalle (#' 1989),andthetheoreticalimplicationsoftheconflictsthatinevitablyarisebetweenfeatureclassconstraintsandotherconstraintsarenotexplored.Halle'sexampleofpartialclassbehaviorcomesfromBarraGaelic,andtacitlycallsonarankingoftheLineCrossingProhibitionoverarulerequiringthatvowelplacefeatures(classedasPlacebyHalle)spread.TheanalysisofBarra t positedisnotfullysecure(seeNChiosin1995),butanotherexamplehereservestoillustratetheidea.VowelharmonyinTurkishinvolvestheclassColor={back,round},andmaybetreated L  asalignmentofthisclasstoprosodicwordedges(seePadgett1995a).Thefocushereisontherightwardspreadingrequiredoffeaturesinformslikethatshownbelow.ThereisacontrastinTurkishstemfinallybetweenplainandpalatalizedl,thelatterbearingthefeature[back]and  ` interactingwiththespreadingofthisfeaturethroughharmony.InthewordpetrolyHX߀XXX߀'petrol  L  (acc.sg.)'#X߀XXHX߀'#,thefinalsuffixalvowelreceivesits[back]specificationfromtheprecedingpalatalized  8  lateral(cf.okulu'school',withanonpalatalizedlateral);butitreceivesits[round]specification $  fromtheprecedingvowel.(SeeClementsandSezerfordiscussion.)Neitherofthepreceding   vowelsofthisdisharmonicformiscapableoftransmittingits[back]valueHX߀XXX߀(shownboldedbelow)#X߀XXHX߀F#   tothissuffixalvowel.Thisisalinecrossingorlocalityissueinonesense"the[back]valueofthelateralintervenes,andsoblocksanysuchspreading.ThusrightwardalignmentofColoris   violatedbythese[back]features.Thereisofcourseafaithfulnessissuehereaswell:harmonyisnotfullyobeyedattheexpenseofthelossofany[back]specifications.ObviouslyColorharmony \ seenasthelinkageofasuperordinatenodebearingthisnamewillencounterdifficultywith H exampleslikethisone.HX߀XXX߀&  2(  19  )3  Ԁ r Partialclassspreadingduetolinecrossing/localityTurkish  \ #X߀XXHX߀#HX߀XXX߀ r m9)%`|2o `hEr'~m H #X߀XXHX߀#HX߀XXX߀' TheanalysisofBarraGaelicduetoSageyandHalleissimilar:itpositsarulespreadingaclassof / placefeaturesfromavowel;[back]cannotspread,duetoaninterveningpalatalizedconsonant,andspreadsfromthisconsonantinstead.Itisnotclear,though,whylinecrossing(ortheequivalentoffaithfulness,protectingthedisharmonic[back]values)prevails,orwhyarulespreadingvowelplacefeaturesisallowedtosucceedonlypartially,#X߀XXHX߀#HX߀XXX߀Ԁsincenoconcreteviewson # constraintconflictareassumed.ThegeneralpointhereisthatthepredictionsofFCTdependinpartonthetenetsofOptimalityTheory;thoseofHalle(1995)willdependonhowtherelevantissuespanoutintheframeworkinwhichthattheoryisembedded. #  8      כ S&! Ї#X߀XXHX߀#mX߀XXX߀ 3.PlaceAssimilations    r ThepreviousdiscussionassumesaprovisionalconstraintNpaformulatedin I odw4(6)  / _ noise(8)  .This  sectiondelvesintotheissueofplaceassimilationmoreseriously(thoughnecessarilyleavingmanyissuesuntouched).Toprobeintoassimilationsisinterestingnotonlyinitsownright,butbecauseitcanallowustofeelsecurethatpointsmadeaboveinfavorofFCTcarryoveronceweunderstandNPAbetter.Thissectionmakesseveralclaimsaboutthenatureofplaceassimilation.First,crucialtoanaccuratecharacterizationofenvironmentsofassimilation(andneutralizationmoregenerally)isthenotionofsegmentalrelease,familiarmostnotablyfromtheworkof  ` Steriade(1993a,1994).Itisthedistinctionbetweenreleasedandunreleasedsegments,ratherthanthatbetweenonsetandcoda,thatdeterminesthepatterns,contrarytotheprevalentview.Thisunderstandingofagenerallicensingissuewillbeimplementedviareleasesensitivefaithfulnessconstraints,buildingonworkmotivatingprominencesensitivefaithfulness(Selkirk1994,Beckman1995,Lombardi1995a).Second,assimilationsresultinpartfromtheworkingsofconstraintsdirectlyrequiringthespreading(ormultiplelinkage)offeatures,groundedinfactsofperceptionandarticulation.Thatis,factsofassimilationcannotbefullyreducedtoissuesoflicensingorprominencebasedfaithfulnessalone.Atleastforthecasesconsideredhere,thisspreadingimperativedoesnotitselfincorporateanypreferencesaboutdirectionality,makingitdifferentfromAlignmentasstandardlyconceived(McCarthyandPrince1993a).Athemetyingthetwomajorclaimsabovetogetherinvolvestheappealtophoneticbasesintheformulationofconstraints,anappealthatfindsitsmotivationinthesuccessoftheconstraintsthemselves.Thisthinkingfollowsmanyothersmakingthephoneticsphonologyconnectioninrecentwork,includingmostnotablyArchangeliandPulleyblank(1994)andSteriade(1993b,1995b).#X߀XXmX߀͹#HX߀XXX߀ r Accountsofnasalplaceassimilationfallintotwobroadclassesthathaveoddlycoexisted.   First,researchersinFGhaveoftenpositedrulesofroughlytheformshownin a odw13(20)  a(Padgett1994,forexample).TheparametersoftheruledepictedherearemoreperspicuouslypresentedinaformatadaptedfromArchangeliandPulleyblank(1994),shownin a odw13(20)  b.Thesestatementstogetherrequestthatanassociationline(path)beinserted,leftwards,fromthePlaceofastopto  anasalsegment.AsMohanan(1993)pointsout,eachstipulation,totheextentthatitrepresentsacrosslinguistictendency,provokesacorrespondingquestion,asin a odw13(20)  c.& C 2(  20  )3   a Ԁ r AruleofNPA 0"    r a.mb9)%`|2  `h@E $#m $X" B" ,Xr XpXXr XB'l0";& t  ; r b.  Argument: Place   h c.  I.  WhydoesPlacefrequentlyassimilate?   r .  Function: Insert   h   ii.  Whydoesitassimilate?    r   Type: Path   h   iii.  (Sameasii.)   r   Direction: LR   h   iv.  Whyisassimilationtypicallyleftwards? t  r   Target: [nasal]   h   v.  Whyarenasalsfavoredtargets? `  r   Arg.Conds: [cont] h   vi.  Whyarestopsfavoredtriggers? L  FXr X ,Xr XpXF;'to;Thesequestionsdonotexhaustthepotentiallist.Thuscoronalconsonantsarefavoredtargetsofassimilation,andobstruentsfavoredtriggersaswell.Padgett(1991,1994)roots a odw13(20)  cviandotherfactsofNPAinthehypothesisthatarticulatorfeaturesandoralstricturefeatureslike[continuant]and[consonantal]formaunitcorrespondingtotheGestureofBrowmanandGoldstein(1986,1989),anideawemakebriefreferencetolater.Thissectionattemptstoanswertheremainingquestionsof a odw13(20)  c.AnimportantprecursortothistaskisMohanan(1993),whoformulatesthemajorquestionsaboutplaceassimilationandproposesarangeofsolutions.Theproposalshereaddressfewerquestions,thoughtheyattempttodevelopinspecificareasMohanan'sinsightthatrecurrenttendenciesinlanguagesrequireexplanationintermsoftheinteractionofsimple,universalandphoneticallygroundedconstraints.RecentideasaboutplaceassimilationdevelopedindependentlybyJun(1995)alsohavemuchincommonwiththosehere,includingtheappealtoprominencebasedfaithfulness(inabroadsenseoftheterm)inordertocaptureasymmetriesinassimilation.Junextendsthisideatoareasnotexploredhere,includingasymmetriesintargetandtriggerplace. r ThesecondmajortackonNPAemergesfromworkonprinciplesofsyllabificationduetoIt=(1986,1989),furtherdevelopedintoageneralnotionoflicensinginGoldsmith(1990)andIt= 4 andMester(1993).Putmostsimply,thisapproachcapitalizesonahypothesizedtensionbetweentwoneeds:first,Placemustbelicensedbyapathtoasyllableonset;second,segmentsmusthave   somePlacefeature.ThelatterconstraintwecallHavePlacehere,formulatedin  odw12(9)  Հabove.Call  theformerLicenseforthemoment.Thetensionbetweentheseconstraintsleadsdirectlytoplace  assimilationintheseaccounts,asmadeclearbythefollowingconstrainttableau.Thereadershouldunderstandthethirdcandidateashavingdoublylinkedplace.&  2(  21  )3   y Ԁ r AlicensingaccountofNPA D! #X߀XXHX߀#HX߀XXX߀*qM dddd dd dd SMT(#(#q,dd ,dd",dd"+  QQ #l! Input=/saNpo/ +Q\#"QQ +HavePlace )Q##Q\ )License +Q#$Q +a.saNpo +\%P %Q +*! %%P &\ % +Q%P ' +b.sanpo +\&!(Q + "&!)\ "*! Q &!* c.+sampo +\(P#+Q + "(P#,\ " (P#-  #X߀XXHX߀ #HX߀XXX߀'D!Q̜#X߀XXHX߀_#mX߀XXX߀Thenasalsegmentincandidate y odw14(21)  aisplacelessandsoviolatesHavePlace.Candidate y odw14(21)  b *X%. hasanasalsegment,inasyllablecoda,bearingitsownplacefeature.Withoutapathfromthisplacefeaturetoasyllableonset,suchaplacefeatureisunlicensed.DoublelinkingofPlaceasin +0'0  y odw14(21)  cmanagestoavoidbothproblemsinonestroke.Assumingnohigherrankedconstraintsthatsayotherwise,thiscandidateisnecessarilyoptimal;ascanbeseen,norankingofHavePlaceand  Licenseisrequired.Partofthegreatappealofthelicensingapproachtoplaceassimilationliesin  thewayitbringstogetherfactsofassimilation,neutralization,andsyllablestructure(seethereferencesabove,andIt=andMester1994,Lombardi1995bforadevelopmentoftheapproachemployingconstraintsofalignment).Belowwemodifythisconceptionoflicensing,recastingtheissueasoneinvolvingprominencesensitivefaithfulness,onlyindirectlyconnectedtoprosodicfactorsliketheonsetversuscodadistinction.However,theaccountpreservestheinsightthatfactsofassimilationandneutralizationaretiedtogetherandhavesomeconnectiontoprosodicposition.#X߀XXmX߀#HX߀XXX߀ 3.1LicensingandRelease  $  #X߀XXHX߀i#mX߀XXX߀ r ItisusefultosurveythesyllablepositionenvironmentsunderwhichNPAobtainscrosslinguistically.Allcasesinvolveanasalconsonantstrictlyadjacenttothetriggeringconsonant#X߀XXmX߀#HX߀XXX߀.The   followingexamplesillustratetherangeofsyllableaffiliationsfoundfortargetandtriggersegmentsacrosslanguages(bracketsindicatesyllableboundaries).#X߀XXHX߀#HX߀XXX߀&  2(  22  )3   z Ԁ r SyllablerelatedenvironmentsofNPA H  r a.0  N][C Japanese,Spanish    d.C][N]  German#X߀XXHX߀#haben~hab)m0 p(#(#  r b.  NC] Englishkent,hampton  e.[NC  LugandaHX߀XXX߀lugaa][ndaetc.#X߀XXHX߀#  \  r c.  [N][C KpelleHX߀XXX߀)m0#X߀XXHX߀#HX߀XXX߀bolu#X߀XXHX߀#HX߀XXX߀ H 'H̛ z odw16(22)  abhavebeenthecentralfocusofclassicallicensingaccountsandshouldbefamiliar.Wehavealreadyseenanexampleofthecommonpattern z odw16(22)  #X߀XXHX߀@#HX߀XXX߀c,involvingasyllabicnasal,inKpelle   (section2.2).TheGerman z odw16(22)  #X߀XXHX߀#HX߀XXX߀drepresentsanapparentlylesscommonreversalofKpelle,with  thesyllabicnasalfollowingthetriggeringconsonant.Finally,thenasalistautosyllabicandpartoftheonsetinanothercommonpattern z odw16(22)  #X߀XXHX߀N#HX߀XXX߀e;compensatorylengtheningoftheprecedingvowel  providesevidenceforthebisegmentalstatusofthisNCsequenceinLuganda(Clements1986,andseethediscussioninsection4). r Thediversityofcasesin z odw16(22)  Հpointsupthedifficultyoftryingtodetermineacrosslinguisticgeneralizationconcerningthepositioninwhichnasalplaceisunlicensed(thatis,inwhich #l! assimilationmustoccur).Thenasalisnotalwaysinthecoda,forinstance,butcanoccupyanysyllableposition"onset,nucleusorcoda.TheGermancasesuggeststhat'preconsonantal'isalsonottherightpositionalcharacterization.ApartfromGerman,themanycasesofwordfinalplaceneutralization,aphenomenonalsointhedomainoflicensingtheory,indicatetheinadequacyof'preconsonantal'forunderstandingunlicensedplace(e.g.,Spanish/Adam/[adan]).Infact, '#& Goldsmith(1990),MesterandIt=(1993),Lombardi(1991)andSteriade(1995a),amongothers, seekgeneralizationsinthecomplementarysphere,statingpositiveconstraintsoflicensing;the )$( rightquestionforus,inthisview,iswheredonasalsnotassimilate? #  9      ׀Theanswertothisquestion   ismuchmorestraightforward;apreliminarycharacterizationisshownbelow('R'isaliquid).#X߀XXHX߀#HX߀XXX߀2(  23  )3    Ԁ r Wherenasalsdon'tassimilate t  r a.  [NV,[NR0 Englishob][noxious(*obmoxious),Russian[mracno'gloomy'(*rracno, L  *nracno),etc.8  (# (#  r b.  [CNV Russiandno'bottom',kniga'book'(*k#X߀XXHX߀#HX߀XXX߀iga),etc. $ t ̛Thissketchy(butcrosslinguisticallyrepresentative)descriptivepicturelightlymasksonesimplegeneralization:nasalplaceispreservedwhenthenasalprecedesatautosyllabicliquidorvocoid.ThisoddstatementisstrikinglysimilartoLombardi's(1991)LaryngealConstraint,meanttocharacterizethepositioninwhichlaryngealfeaturesarelicensed.TheLaryngealConstraintineffectsaysthefollowing(thisformulationhasastraightforwardimplementationinFCT,itshouldbeclear):2(  24  )3    Ԁ r LaryngealConstraint(Lombardi1991) p  r Laryngealislicensediffitoccurs[X__[+son]Y]%#X߀XXHX߀#HX߀XXX߀ H  r Formulatedinthisway,theLaryngealConstraintseemsstrange,afactLombardialsonotes.Whythisenvironmentandnoother?Thewaytomakesenseofthesefactsinvolvingbothplaceandlaryngealfeatures,followingespeciallySteriade(1993ab,1994,1995ab)andKingstonϜ(1990),istoinvokethenotionofsegmentalrelease.Thisexpressionrefers(inphoneticterms)to 4 theoffsetphaseofaconsonantalconstriction,andtothehighlyperceptibleburstthataccompaniesit(undertherightconditions),knowntoprovideimportantacousticcuestocontrastsin(especially)placeandlaryngealfeatures(Ohala1990,andseePickettetal.1995forarecentoverviewoftheliteratureonthemoregeneralCVvs.VCtransitiondistinctionandrelateddiscussion).Itisthishighperceptibilityofreleasedfeaturecontraststhatunderwritesthepresenceofsuchfeatures.Inthegrammaticalaccountbelow,itwillbebymeansoffaithfulnessconstraints#X߀XXHX߀#HX߀XXX߀(thusphoneticallygrounded)thatwecapitalizeonthisidea. X   r Itisreleasethatrationalizesanotherwisemysteriousconstraintlike  odw17(24)  .Thisconstraintroughlysumsupthepositioninwhichreleaseisvirtuallyphoneticallyinevitable:anonsetconsonantisreleasedintoasucceedingsonorantasthesyllableopensuptothenucleus#X߀XXHX߀T#HX߀XXX߀.Dueto $X" restrictionsonsonoritysequencing,a[+sonorant]segmentfollowinganasalinthesyllableonsetwillnormallybeeitheraliquidoravocoid(glideorvowel),hencethiswordinginouroriginalstatementabove.Therestrictiontotautosyllabicityisimportant,sincenasalsfrequentlyassimilatetothesesegmentsacrossasyllableboundary,e.g.,/...N+r.../[...r.r...](seePadgett1995cfor '#& variouscases);thusreleaseisfavoredinpartbyonsetstatus.However,releaseisnotreducibleto  aconstraintonthenumberofonsetplacefeatures,contrarytotheclassicallicensingview.Whilenasalsfrequentlymustbeplacelinkedinonset[NCclusters,asintheLugandaexamplein z odw16(22)  e,theyfailtoassimilateinonset[CNclusterslikethoseofRussianshownin  odw18(23)  b.Itisreleasethatilluminatesthisdistinction:thenasaloccupiesapositionofuniversalreleaseinthelatter,butnotintheformer.Asimilarpointcanbemadeconcerninglaryngealneutralization:thedevoicingseeninRussianexampleslike/ v +plavlyaty/[ f plavlyaty]'fusein'motivatedalicensingconstraintlike 8  thatof  odw17(24)  ՀinLombardi(1991),ratherthanaprohibitionononset[voice]perse,sincethisfeatureisotherwisesanctionedinonsetobstruents.Moregenerally,Steriade(1993b,1995b)arguesthatwhenprosodybearsonlicensing,itdoessoonlyindirectly;thereallicensingactioninvolvesperceptuallymotivatedfactorsincludingrelease.Forlanguageslackingonsetclusterslikendandfpl,thedistinctionbetweenclassicalprosodybasedlicensingandanaccountbasedon (  releasedisappears:thereisonlyoneonsetconsonant,anditisnecessarilyreleased.ItisthemorecomplexscenariosengenderedbyRussianandLuganda(amongmanyothers)thatmotivatesreleaseasthepivotalnotion.#X߀XXHX߀-#HX߀XXX߀ r Steriadefollowsandexpandsonpreviousworkappealingtoreleaseinphonology,notably t McCawley(1967),Selkirk(1982)andKingston(1990).AsinSelkirk(1982),weunderstandreleasehereasafeaturalpropertyofsegments;thusninRussianknigais[+release]. #  10      ׀Asafirst L steptowarddeterminingthedistributionofthisfeature"alwaysdependentonbothprosodicandsegmentalenvironments,asseenabove"letusassumethatalloutputsegmentsprecedingatautosyllabicsonorantarenecessarily[+release],aconstraintonGENinOptimalityTheory.Nowitisrequiredoflanguageparticularphonologyonlytospecifywhereelse(ifanywhere)segmentsarereleased.Wecanisolatetworoughcategoriesofbroaderreleasedistributionthatseemnecessarytothetheory(cf.Steriade1994:2089):2(  25  )3    Ԁ r Extralanguagespecificenvironmentsofrelease   r a.  Word(orutterance)finalconsonantsare[+release]   r b.  Allconsonantsare[+release] p   odw19(25)  bcharacterizesthemostreleasepermissivelanguage,inwhichplaceandlaryngealdistinctionsarereleased"andthereforemorelikelytobepreserved"inanyconsonant,includingthosethatarewordfinalorbeforeanyotherconsonant.BellaCoolaprovidesoneexample(Hoard1978,Bagemihl1992),andFrenchanother(Selkirk1982andreferencestherein).  odw19(25)  arepresentsamoreconservativecase,alanguageinwhichreleaseisnotpossiblebeforeaconsonant(exceptofcourseatautosyllabicsonorant),butoccursattheendofaword.Englishpossiblyprovidesacaseofthissort:nasalplaceisneutralizedandassimilatedwithinmorphemes,asincamp(*canp)andrant(*ramt)#X߀XXHX߀#HX߀XXX߀.Butthiscontrastsurviveswordfinally,asincamandran. ' #& #X߀XXHX߀ #HX߀XXX߀Thephoneticbasisforthistypeisagainreasonablyclear:areleasecanbemaskedbythe (#' overlappingarticulationofafollowingconsonant;butthereisnofollowingconsonanttomaskthereleaseofaword(orutterance)finalsegment.Thegeneralnotionofreleasethereforepotentiallyexplainsadistinctionoftenassociatedwithwordfinal'exceptionality'(seeespeciallyLombardi1991onthisdevice).Yetexceptionality,unlikerelease,failstomakeanyconnectionbetweenthewordfinalandpretautosyllabicsonorantenvironments(bothoftenreleased),ortoilluminatethehierarchyofneutralizationcapturedby  odw19(25)  ՀandthestatementinGEN;nordoesitexplainwhyitispreciselyfinalconsonantsthatcanbespecial(theycanbereleasedwhen 8  preconsonantalconsonantscannot),andnotinitialconsonants.(Byrd1992makessimilarpoints.)#X߀XXHX߀ #HX߀XXX߀Thispromiseofdrawingtogetherthewordinternalandwordfinalfactsclearlymakestherelease  ` ideaworthpursuing.Whether  odw19(25)  abthemselvesrepresentconstraintsavailableforranking,orrequirefurtherscrutinyandperhapssubdivision,isunclear. #  11      כSincethefocushereison  8  assimilationwithinwords,bypreconsonantalnasalsthatareunreleased,wecanmakethefollowingsimplifyingassumptionforthelanguagesunderdiscussion:consonantsare[+release]beforetautosyllabicsonorants(necessarilybyGEN),otherwiseunreleased.#X߀XXHX߀#HX߀XXX߀ r Theideathatfeaturesunderreleaseareperceptuallymoresalientsuggestsagrammaticalimplementationviareleasesensitivefaithfulness,underthenaturalassumptionthatwhetherafeatureispreserved(aswellasotherfaithfulnessissues)candependonsuchperceptualdifferences.ThisstrategybuildsonSelkirk(1994),whoarguesforthesubdivisionoffaithfulnessconstraintsaccordingtotheprosodicpositionoftherelevantsegment"onsetvs.elsewhere,etc.Ϝ(seealsoBeckman1995andLombardi1995a).Themajorpointofthediscussionaboveisthattherightdistinctionforplaceassimilationinsteadinvolvesrelease,andsofollowsfromprosodyonlyindirectly.Thisleadstofaithfulnessdistinctionslikethefollowing:2(  26  )3    Ԁ r Featurefaithfulness,releasesensitiveandgeneral    r a.  Max(place):0  Everyinputplacefeaturehasanoutputcorrespondent.(#(# М̛ r b.  MaxRel(place):0  LetSbea[+release]outputsegment.Theneveryplacefeaturein  theinputcorrespondentofShasanoutputcorrespondentinS. (#(# М̛ r c.  Universalranking:  MaxRel(place)#X߀XXHX߀#HX߀XXX߀ԛ>>Max(place)#X߀XXHX߀f#HX߀XXX߀ԛ D! TheformulationshereassumeanunderstandingoffeaturalfaithfulnessduetoMcCarthyandPrince(1995)(whichseeforextensiveargumentation).RecastingtheproposalsofPrinceand Smolensky(1993)onfaithfulness,McCarthyandPrinceargueforanunderstandingofthisnotion $D # intermsofcorrespondence,ateitherthesegmentalorfeaturallevel.K #  12      ׀Forourpurposesitwill   alwaysbeclearwhattheinputoutputsegmentalcorrespondencesare,andsonofurthermentionofthisissuewillberequired.#X߀XXHX߀#HX߀XXX߀  odw20(26)  asimplyrequiresthatanyunderlyingplacefeaturebemaintained  intheoutput.  odw20(26)  bismorediscriminating,requiringthatasegmentthatisreleasedintheoutput t maintainitsunderlyingplacespecifications.Byhypothesisreleaseddistinctionsarebetterpreservedthanothers;hencethefixedrankingbetweenthetwoconstraintsin  odw20(26)  c.TheMax L  familyofconstraintsonlymilitatesagainstfeatureloss;anotherfamily(calledDep,seeMcCarthy 8  andPrince1995)worksagainstoutputmaterialthatisnotintheinput,featuralepenthesis.Aswewillseebelow,NPAcanleadtothelossofinputfeaturalmaterial,violatingMax(Place),as  ` wellastheestablishmentofassociationlinesnotintheinput,violatingDep(Link)(seeIt=,Mester  L  andPadgetttoappearonthefullarrayoffeaturalfaithfulnessconstraints);bothtypesofviolationoccurforexamplein/n+gb/#X߀XXHX߀#HX߀XXX߀mgb.Therewillbenoneedtodistinguishtheseviolations, $  however,andinordertosimplifythetableauxallfaithfulnessconstraintswillbelumpedtogetherintoFaithandFaithRelconstraints.    r Inordertoseethereleasesensitivefaithfulnessideaatwork,wenowneedtomotivateanaccountofassimilationormultiplelinkingitself.Thisisthejobofthefollowingsection.Theideaissimple,though:oncemultiplelinkingbetweentwosegmentsC1C2(withcompetingplace \ specifications)isrequired,thenthedirectionofassimilationwillbedeterminedbyreleasesensitivefaithfulness.IfC2#X߀XXHX߀#HX߀XXX߀ԀisreleasedandC1#X߀XXHX߀"#HX߀XXX߀Ԁisnot,thentheformer'splacefeaturewillnecessarily 4 winoutoverthelatter's. 3.2TheSpreadingImperative  H #X߀XXHX߀"#HX߀XXX߀ r Ifreleasesensitivefaithfulnessprovidestherealcontentofthenotionlicenseinthecaseof   placeassimilation,thenextquestionmightbewhatisleftofthefeaturegeometricunderstandingofNPAasaplacespreadingphenomenon.Infact,somedirectimperativeoffeaturalspreadingis  requiredbyourtheory.Toseewhy,considerfirsttotalNPAtocomplexsegments,asseeninKpelle#X߀XXHX߀##HX߀XXX߀mgbearlier.Revertingtotheclassicallicensingaccountmomentarilyforthesakeof  discussion,considerthetableaubelow.TheproblemwiththisindirectviewofspreadingisthatitpredictsonlypartialNPAtosuchcomplexsegments,nevertotalNPA(recallthatweassumeaworlddevoidoftheorganizingclassnodesofFG).License#X߀XXHX߀%#HX߀XXX߀ԀandHavePlace#X߀XXHX߀'#HX߀XXX߀Ԁtogethermerely D! requirethatthenasalreceivesomeplacefeaturefromthefollowingsegment.Anyfurther 0"  spreadingtoderive#X߀XXHX߀'#HX߀XXX߀mgb#X߀XXHX߀(#HX߀XXX߀ԛcanonlybeinviolationofmarkedness"acomplexnasalismore #l! markedthanasimpleone"andfaithfulness(iftheinputitselfisnot#X߀XXHX߀;)#HX߀XXX߀mgb#X߀XXHX߀&*#HX߀XXX߀).Candidate Lnonoise(27)  chas $X" doublylinked[dorsal],butnot[labial]. $D # Ї&  2(  27  )3   L Ԁ0 r #X߀XXHX߀n*#HX߀XXX߀LicensingpredictsonlypartialNPAr(#r(# *qM dddd dd"dd"M(#(#q,dd ,dd",dd"+  QQ  Input=/N+gb/ +Q\<QQ +HavePlace )Q<Q\ )License#X߀XXHX߀+#HX߀XXX߀ +Q<Q +a.Ngb +\lQ +*! %l\ % +Ql +b.ngb +\ < Q + " < \ "*! Q  <  c.+#X߀XXHX߀-#HX߀XXX߀gb +\l  Q + "l  \ " l   #X߀XXHX߀/#HX߀XXX߀'#+Thisproblemcarriesovertootherindirectaccountsofspreading.Spreadingcannotfollowfromsegmentalorfeaturalmarkednessalone(asinMcCarthy1994,Beckman1995),forreasonsjustgiven:thesegmentproducedbytotalassimilationismoremarked.Norcanaconstraintinthe L  spiritof*FF(barringtwoinstancesofthesamefeature,identicalornot,ArchangeliandPulleyblank1993,Kiparsky1993),ifplacefeaturesareprivativeaswidelyassumed"#X߀XXHX߀/#hasno $t  valuefor[labial]besidethatofgb.Similarargumentsforaspreadingimperativecomefromfacts `  of[voice]assimilation,asinRussian/ot+davaty/od.da.vaty'giveup'(seeHayes1984and L Kiparsky1985).Thevoicedsegmentproducedbyspreadingismoremarkedthanwhatfailedassimilationwouldproduce,andthisfeatureisagainwidely(thoughnotuniversally)viewedasprivative(MesterandIt=1989,Cho1990,Lombardi1991). r By'spreadingimperative'wemeanaconstraintrequiringofafeaturethatitbemultiplylinked(orspread),orextendedinitsdomainintheoutput.SomethingclosetothisiswidelyemployedintheOTliteraturealreadyintheformoffeaturalalignmentconstraints(Kirchner1993,Smolensky1993,Pulleyblank1994,ColeandKisseberth1994andmanyothers).Alignmentcouldbemadetoserveourpurposes,butaninterestingpointariseshere:alignmentastypicallyconstruedincorporatesdirectionalityintoconstraintstatements(seePrinceandSmolensky1993,McCarthyandPrince1993a),whileitisarguablytrue(inthefeaturaldomainatleast)thatdirectionalityofspreadingfollowsfromotherfactors.Thesourceofvowelharmony"accountingforitsapparentdirectionality"istypicallyavowelinamoreprominentposition,forexamplestressed,wordinitial,etc.,asSteriade(1995a)notes.DirectlyrelatedtothisarethefindingsaboveconcerningreleaseandNPA:inplaceassimilations,itisthereleasedsegmentwhoseplacewins.(Thereadermightlookoncemoreatthepatternslaidoutin z odw16(22)  Հtoseethis.)Giventhisfact,itwouldbeoddtostipulatethetypicallyleftwardspreadingofplaceinNPA(apointsimilarlymadebyclassicallicensingaccounts),anditwouldinfactbeincorrectforGermanhabHX߀XXX߀)m0#X߀XXHX߀9#HX߀XXX߀ #' etc.(wheretheobstruentisreleasedintothenasal,itselfunreleased).Theconstraintemployedherethereforeabstractsawayfromdirectionality;themostgeneralformulationmightbethefollowing:everyfeatureislinkedtoeverysegment(insomedomain,e.g.,Pwd).Moreformally,withxrangingoverfeaturesandyoversegments,#X߀XXHX߀9#HX߀XXX߀ D'"+ 2(  28  )3  Ԁ r Spread(x):0 zxy,x(y)(insomedomain))l$- (# (# Aswithalignment(seereferencesabove),thisformulationisverygeneral.Realoutcomeswillbereinedinbyotherconstraintsonlocality(nosegmentskipping),markedness(don'tderivebad +0'0 things),andsoon.Inwhatfollowswewillignoretheseissues;theessentialpointissimplythatsomespreadingimperativeisrequiredinordertoexplainfactsofassimilation. #  13      כVarious  researchershavesuggestedplausiblephoneticbasesforsuchaconstraint;itwillenhancetheperceptibilityoftheaffectedfeaturebyextendingit(Steriade1993b,Kaun1994,ColeandKisseberth1994),eliminatecontrastsinnonprominentlocations(Steriade1993b,1995a,b,andthediscussionofreleaseabove),andleadtofeweroverallarticulations/specifications(Mohanan1993,ColeandKisseberth1994).ItmaybethatSpreadmustultimatelybereducedtomore 8  fundamentalconstraintsinvokingsuchfunctionalnotions;suchamoveiswellbeyondthescopeofthiswork,however. r ThecoreoftheNPAaccountisnowathand.TheimpetusisprovidedbySpread(Place)  8  (zxyxPlace,yasegment,x(y#X߀XXHX߀;#HX߀XXX߀))whilereleasesensitivefaithfulnessensuresthatreleased $  segmentsarethesourceofassimilation(andunreleasedsegmentsthetargets).AsLombardiϜ(1995a)showsinaparallelexaminationof[voice]assimilationfacts,thereareonlytworankingsoftherelevantconstraintstoconsider.(Recallthattheranking#X߀XHX߀B#HX߀XX߀FaithRel#X߀XHX߀C#ԚHX߀XX߀Ԁ>>Faith#X߀XXHX߀C#HX߀XXX߀ԛisfixed.)First,   assimilationwilloccurasdescribedabovewheneverSpread(Place)outranksFaith,asshown   belowforourexample#X߀XXHX߀FD#HX߀XXX߀mgbagain.Theinputforthisformhasbeenchangedarbitrarilyto/n+gb/, p wherethenasalsegmentbearsitsown[coronal]specification,apointtowhichwereturnϜmomentarily.#X߀XXHX߀ME#HX߀XXX߀ԛ H &  2(  29  )3    Ԁ r PlaceassimilationbySpread(Place)#X߀XXHX߀YF#HX߀XXX߀Ԁ>>Faith#X߀XXHX߀BG#HX߀XXX߀ԛ  p *qM dddd dd"dd"M(#(#q,dd ,dd",dd",dd"+  QQ H Input=/n+gb/ )Q\QQ )М#X߀XHX߀G#HX߀XX߀FaithRel )\Q\Q )Spread(Place)#X߀XXHX߀0I#HX߀XXX߀ )Q\\Q )Faith#X߀XXHX߀I#HX߀XXX߀ +Q\Q +a.Ngb %,Q %М '\, '*!* -x,\ -* +Q,x +b.ngb "\Q "М \ \  *!** -x\!\ - +Q\"x +c.#X߀XXHX߀5J#HX߀XXX߀gb ",#Q "М \ ,$ *! -x,%\ -** +Q,&x +d.#X߀XXHX߀L#HX߀XXX߀nd "\'Q "М*!** \ \( #X߀XXHX߀L#HX߀XXX߀ -x\)\ -*** +Q\*x +e.+#X߀XXHX߀|M#HX߀XXX߀mgb " ,+Q "М \  ,,  -x ,-\ -***& ,.  x &' F#X߀XXHX߀'N#HX߀XXX߀Spread(Place)#X߀XXHX߀O#HX߀XXX߀Ԁisunderstoodcruciallytoquantifyuniversallyoverplacefeaturesinagradiently "4/ violableway,andsupersedestheprovisionalconstraintNpaofsection2.1.Candidate  odw25(29)  a # 0 violatesthisconstrainttwice:inthistwosegmentspan,therearefourpossibleplacefeaturesegmentlinks([labialand[dorsal]eachwitheachsegment);buttwodonotobtain.(Inwhatfollowswewillcontinuetocountviolationsonlywithintheconsonantcluster,ignoringextra &!3 violationssharedbyallcandidatesforfailuretoSpreadthroughoutanentireword;seefootnoteladida13  .)Candidate  odw25(29)  #X߀XXHX߀iO#HX߀XXX߀bviolatesthesameconstraintthreetimes,sincetherearethreepossiblebut  nonoccurringplacesegmentlinks.#X߀XXHX߀R#HX߀XXX߀Incontrast,candidate  odw25(29)  cviolatesSpreadonlyonce(gb's  [dorsal]featurehasspread),butstillfatally.#X߀XXHX߀iS#HX߀XXX߀Candidate  odw25(29)  disaninterestingattemptto t circumventSpreadviolations,withouttotallyassimilatingtogb,bylosingtheinputplace ` specificationsofthissegment;further,itpreserves(andspreads)theinput[coronal]specificationofthenasal.Nevertheless,thismoveviolatesFaithRel,sincetheinputplacefeaturesofgb󛀜have 8  nooutputcorrespondents,andgbisthereleasedsegmentintherealexampleswehaveinmind, $ t e.g.,Kpelle#X߀XXHX߀^T#HX߀XXX߀mgbi#X߀XXHX߀ W#HX߀XXX߀ԛ'myself'.  odw25(29)  ewinsbypassingbothofthesetoprankedconstraints.As  ` Lombardi(1995a)notes,theseconstraintsthereforeneednotberanked.Infact,giventheϜelsewhererelationbetweenthetwofaithfulnessconstraints,areleasedsegmentcanneverloseas  8  in  odw25(29)  d(allelseequal).Inthetableauxthatfollow,wewillthereforeomitsuchcandidatesandtheconstraintFaithRel.    r Forcompleteness,theFaithviolationsarecountedinthefollowingway:candidate  odw25(29)  a   incursoneMax(Place)violation,sincethenasalsinput[coronal]specificationislost;candidate   #X߀XXHX߀WW#HX߀XXX߀  odw25(29)  cinadditionviolatesDep(Link)once,giventheestablishmentofalinkbetweenthenasaland p [dorsal]intheoutput;in  odw25(29)  dreleasedgbhaslostbothits[labial]and[dorsal]specifications; \ further,ithasacquiredalinktothenasals[coronal]feature,hencethreeFaithRelviolations, H mirroredinthemoregeneralFaithconstraintaswell;finally,#X߀XXHX߀Z#HX߀XXX߀in  odw25(29)  ethenasalsinput[coronal] 4 specificationisagainlost,andboth[labial]and[dorsal]arelinkedunfaithfullytothatnasal.#X߀XXHX߀=]#HX߀XXX߀ԛ  p  r FailureofassimilationresultsfromthereverserankingFaith#X߀XXHX߀D^#HX߀XXX߀ԛ>>#X߀XXHX߀^#HX߀XXX߀ԀSpread(Place)#X߀XHX߀4_#HX߀XXX߀,asthe H readercanverifybystudyingtheabovetableau.Hencetheminimaltypologyofassimilatingversusnonassimilatinglanguagesisderived.Apointleftsofaruntouchedinvolvesawellknownasymmetrybetweennasalsandobstruents:theformeraremorelikelytoassimilatethanthelatter,afactleadingtocontrastswithinlanguageslikeEnglisha p titudeversus*a m titude.Theaccount  invokingreleasediscussedabove,whichofcourseextendstocovercaseswhereobstruentsare  targetsofassimilation,mustbesupplementedwitharecognitionofthisdifferencebetweenobstruentandnasalplace.Themostdirectattackonthisprobleminvolvesanotherbreakdownofplacefaithfulnesswithafixedranking:Faith(ObsPlace)>>Faith(NasPlace).The \  interventionofSpread(Place)betweentheseconstraintswillgivetheEnglishlikeoutput H! contrastweseek.Thisdistinctionbetweennasalandobstruentplacefaithfulnessonceagainfindsrootsinfactsofperception:nasalplaceismoredifficulttoperceivethanobstruentplace,sincetheformerlackssignificantburstcues.(SeeforexampleOhalaandOhala1993:2412andreferencesthereinonthisdistinction,anditsrelevancetotheplaceassimilationasymmetry.)Sinceobstruentassimilationwillnotbeaconcernofours,wecansafelybypassthisdistinctioninwhatfollows. r Asamatterofinterest,theaccountofplaceassimilationproposedheredoesnotrelyinanywayonassumptionsaboutthedegreeofinputspecificationofthetargetsegment.The[coronal]inputspecificationattributedtothenasalin  odw25(29)  Հwaspositedsimplytohelpmakethispoint.Asthereadercanverify,theoutcomeoftableau  odw25(29)  ՀisagainthesameifthenasalisassumedtobeunspecifiedN,orspecifiedforanotherplace.ThisaccountfollowsPrinceandSmolensky(1993), l+&* It=,MesterandPadgett(toappear)andothersinlimitingconstraintsotherthanthoseoffaithfulnesstostatementsaboutthewellformednessoftheoutput.Underthisviewofthings,noimperativesconcerninginputdegreeofspecificationarepossible;thepointofthereferencescitedandtheexamplehereisthattheyarealsounnecessary,giventherighttheoryofoutputconstraints.ThisthinkingalsofollowsIt=andMester(1989),Mohanan(1991),Smolensky(1993)andSteriade(1995a)inrejectingsocalled'naive'usesofunderspecificationinordertoaccountforassimilatoryfacts(achievingNPAbystipulatinganinputunspecifiednasal,forexample).Initsleastembellishedform,suchauseofunderspecificationonlybegsmanyoftheessentialquestionsaboutplaceassimilationposedattheoutset. r Toconcludethissection,thepreliminaryanswerswenowhavetothequestionsaboutNPAraisedattheoutsetofsection3areasfollows.AssimilationsoccurduetotheworkingsofafamilyofSpreadconstraintsplausiblygroundedinfactsofperceptionandarticulation;place   assimilationfallsintothisfamily.Directionalityofplaceassimilationisanartifactofreleasesensitivefaithfulness;moregenerally,directionalityeffectswilloftenbetiedtosuchprominentialasymmetries.Nasalsaremorecommonvictimsofplaceassimilationthanobstruentsgivenanotherdimensionofprominenceinvolvingnasalversusobstruentplace,withnasalplacebeingthelesssalient(andthereforefaithful).TheseideasarebroadlysimilartomanyraisedbyMohanan(1993)andJun(1995)inaddressingplaceassimilations;theseworksinadditionaddressotherquestionsconcerningplaceassimilationsnotraisedhere.Moregenerally,thesuccessofthisaccountistiedtoitsappealtoconstraintswithbasesinfactsofphonetics,placingthisapproachinagrowingtradition(seemostnotablyArchangeliandPulleyblank1994andSteriade1993b,1995b). 4.PartialAssimilations  4 #X߀XXHX߀_# r Thisfinalsectionhastwomaingoals.First,itgivesempiricalsubstancetothearguments   fromsection2infavorofFCT,bysurveyingarangeofcasesofpartialNPA,involvingvariouscomplexsegmenttypes.ThisendeavordocumentsrobustpartialclassbehaviorfortheclassPlace,andthereforeHX߀XXX߀broadenstheresultsofPadgett(1995a),wherethesameisdoneforColor.  #X߀XXHX߀|q#Second,itintegratesideasoftheprevioussectionaboutNPAintotheanalysisofpartial l assimilation,andshowsthattheargumentsforFCTstillhold. 4.1Survey  0"   r Inordertodemonstratetheexistenceofpartialnasalplaceassimilationtoacomplexsegment,oneneedstosatisfyonseveralpoints.First,isassimilationreallypartial,andphonologicallyso?Thedistinctionbetweenmgbandgb,say,isphoneticallysubtle,and %0!$ transcriptionofthelattermightbewrittenofftoerror;orifitisaccurate,thewholedistinctionmightbeattributedtophonetics,perhapsanartifactofdifferentdegreesofgesturaloverlapinspeech(BrowmanandGoldstein1989).Second,aretheNCsequencesunderconsiderationreallybisegmentalasananalysisintermsofplaceassimilationimplies,oraretheyprenasalizedsinglesegments?Third,sincethefrequentresultofpartialNPAis,asegmentassociatedwithnasal |*%) 'placelessness'bytheworkofTrigo(1988),couldHX߀XXX߀gb#X߀XXHX߀~v#ԛactuallyinvolveacompletely h+&* unassimilatednasal,phonologicallyNgb(andhenceshowingnopartialassimilationatall)?  Finally,isthetriggersegmenttrulyphonologicallycomplex?Ifgb,say,werejustthesurface  embellishmentofanunderlying/g/,thenperhapsHX߀XXX߀gb#X߀XXHX߀(x#ԛistotallyassimilatedafterall.   r Someoftheseissueswillbeaddressedaswesurveyparticularcasesbelow,butafewgeneralremarksnowwilllimitthefieldofquestionssomewhat.First,thequestionaboutprenasalizationcarrieslittleweight,becauseofarangeofcompellingargumentsagainstaphonologicaldistinctionbetweenmonoandbisegmentalNC #  14      ׀(Herbert1975,1986,Feinstein $ t 1979,Clements1986,Duanmu1990,Padgett1991,1995c,Piggott1992,Steriade1993a).Onprincipledgroundswethereforemakenodistinctioninsegmentalstatusbetweenthend󛀜sequence  L  inLu.ga.ndaandthesamesequenceinEnglishMe.lin.da"botharebisegmentalplaceassimilated  8  NC,andtheydifferonlyinsyllabificationasshown.InfactseveralofthelanguagessurveyedbelowinvolveNCsequencesthatarebisegmentalwithoutcontroversy,withanasalthatisclearlysyllabifieddifferentlyfromthefollowingC(i.e.,eithersyllabicitselforacoda). r TurningtotheinterpretationofHX߀XXX߀gb#X߀XXHX߀1}#ԛasNgb,pursuingTrigo(1988):theideaheremightbe   thatassimilationfailsentirely,duetoaprohibition(thatisassumedhereanyway)oncomplexm. p Thisidearestsonanunderstandingofasasortofphoneticrealizationofthephonological, \ debuccalizedN.YetifwetrulyhaveoutputHX߀XXX߀gb#X߀XXHX߀#"withavelarnasalwhosevelarityisgesturally H mergedwiththatofgb(HX߀XXX߀pendingdirectinvestigation,thebestwecandohereisrelyonphonetic 4 descriptionswheretheyseemmostclear#X߀XXHX߀#)"itisnecessaryinanoutputorientedtheorytoproduce  p it.(SeetherelevantdiscussioninMcCarthyandPrince1995againstarelegationofsystematicoutputsegmentalcontentto'phoneticinterpretation',asdonebytheFillconstraintofPrinceand H Smolensky1993.)Coulditbethatthegrammarindeedgeneratesadorsalnasal,yetonethatisneverthelessunassimilated?Thisinterpretationcanberejectedasageneralmoveonempiricalgrounds,puttingasideitsimplausibilityfromthestandpointoftheOCP(giventheadjacent[dorsal]specifications).Asadefaultnasalplace,isacompetitorwith[coronal]n;bothare  foundwithgreatabundanceinlanguagesasplaceneutralizednasals.Yetnoreportofngb(where  assimilationotherwiseobtains,i.e.tosimplesegments)existstothebestofmyknowledge,somethingthatwouldbeexpectedifthenasalcanresorttoitsownunassimilatedplace.Ifthenasalisindeed[dorsal],andifitsharesthisplacewithgb,thenwehavepartialplaceassimilation. X  Forreasonstobediscussed,partialassimilationtolabiovelarsindeedtypicallygivesvelarrather D! thanlabialm;butininfrequentinstances(involvingatriggerw)mresultsinstead.Theexistenceof 0"  thisalternatepatternisonlypossibleifthenasalispartiallyassimilating,since[labial]isadefaultplaceinnotheory.Finally,itshouldbepossibleinprincipletoshowforsomelanguageswith $X" beforelabiovelarsthatthedefaultnasalisactuallyn.Spanishisapossibleinstance:inmost $D # dialects,thedefaultnasaliscoronal;butpartialassimilationtowproduces,asinu𛀜weo'an %0!$ egg'(Harris1969,1984).  '#& HX߀XXX߀ r Theotherconcernsmentionedabovewillbeaddressedwhereappropriatebelow.To previewwhatfollows,fourdifferenttypesofcomplexsegmentareconsideredoneatatime.Theseareclicks(labioandcoronovelar),palatals(coronovelar),glides(labiovelarwand  coronovelarj)#X߀XXHX߀Z#HX߀XXX߀Ԁandlabiovelarstopslikegb.Foreachtype,evidenceofpartialassimilationis t found.Thesurveyisnecessarilybriefforreasonsofspace.Mostattention,andactualanalysisarerestrictedtothefinalcategoryoflabiovelars,withananalysisofG ,aKwalanguage. &  4.1.1Clicks  $ t  r Clicksarephoneticallycomplex,inthattheyrequireavelarconstrictionandsomeother'$ 6constrictionfurtherforward,coronalorlabial.BuildingonChomskyandHalle(1968)andHalle(1983),Sagey(1986)arguedforthecomplexstatusofclicksonphonologicalgroundsaswell,attemptingtoshowtherelevanceofeacharticulatortosomephonologicalprocess. #  15      ׀As   MaddiesonandLadefoged(1989)pointout,suchanargumentforthepresenceof[dorsal]inclickscomesfromNPAinBantulanguageshavingclicks.Therelevantexamplesinvolvetheprefixationofanasalfinalclassmarker,asintheZuluexamplesin  odw28(30)  .Thefinalnasaloftheclass10prefixiziNassimilatesacrossallplacesofarticulation,asshownin  odw28(30)  a.Beforethe p clicksin  odw28(30)  b(respectivelydental,palatoalveolarandlateral),assimilationproducessimplevelarϜ#X߀XXHX߀#HX߀XXX߀,asbeforek.(Doke1926:78,1931:14,523,datafromthelatter.Theclicksarealsovoiced H followingthenasal.)2(  30  )3  Ԁ r   ZuluNPA  \ H( ,Xr hXxXXr XH r a.  izi mp aph#X߀XXHX߀#DHX߀XXX߀  'feathers' h   b.  izi #X߀XXHX߀#| ezu    'slices' x   4  r   izi nt i  'sticks' h     izi )|=pp, u )|=pp, ulu    'speciesofbird(pl.)' $  r   izi k Dz<  'spoons' h     izi || a||a    'greenfrogs'  LXr X ,Xr hXxXL̛ThedescriptionbyDokeisclearonthispartialassimilation.Further,MaddiesonandLadefoged(1989)implythatitistrueofotherBantulanguageswithclicks.IfNPAisphonological,thenthesedatademonstratea[dorsal]placeinclicks.Sinceclicksaredistinguishedbytheirforwardarticulations,theymustinfactbecomplex,andassimilationtothemisthereforepartial. r NPAinZuluisnotplausiblyviewedasapurelyphoneticphenomenonontheorderofgesturaloverlap.Assimilationbringsaboutarangeofsystematicandneutralizingchangesinthequalityofthefollowingconsonantthatcouldnotfollowsimplyfromoverlap,includingdeaspiration,voicing,nasalizationandhardening(seeDokereferences).Further,MaddiesonandLadefogedpointoutthatthepurelyvelaroutcomeisnotanarticulatorynecessity.Zulucontrasts nasalizedclickstothesequencesillustratedabove;nasalizedclickshavenasalityduringboth &("% articulationsoftheclick.NoneofthisistosaythatthepropertiesofNPAinZuluseenherelackabasisinphoneticeffects;surelytheoppositeistrue.Nevertheless,NPAisphonologized.  r AgermanequestionatthispointconcernsthenatureofpartialNPAtoclicks:whyisit[dorsal]thatassimilates,andnottheotherarticulatorfeature?Aswewillsee,thisskewinresultsoccurswithpartialNPAbeforelabiovelarsalso.AccordingtoOhalaandLorentz(1977)andOhalaandOhala(1993),thephoneticmotivationforthisimbalanceismostlikelyacoustic:thespectralpropertiesofanasalwithvelarclosureHX߀XXX߀Ԁarenecessarilythoseofsimple#X߀XXHX߀#"itmakesno $ t differencewhatanyarticulatorsfurtherforwardmaybedoing.Therefore,beforeacomplexsegmentwithlargelyoverlappingplaces,oneofwhichis[dorsal],puremayoftenwinout  L  perceptually. #  16      כItismoredifficulttodeterminehowthisinsightshouldbearonourgrammatical  8  formulations.Thissectiontogetherwiththosefollowingshowthatassimilationcanbephonologicallytotalorpartialbeforecomplexsegments.(Itisusefultobearinmindthattotal   assimilationaswellaspartialassimilationisacommonpatterninlabiovelarstops;thisisagrammaticalchoice.)Thisfactmayhavephoneticunderpinningsliketheonejustmentioned(onlytendential,note,andnotdeterminant),buthowdoweexpressitsphonologicalrealization?Theassumptionhereisthattheconstraint*Compsegmakesthedeterminationbetweenfullandpartial p assimilation,asseenearlier.Oncepartialassimilationisfavored,theselectionof[dorsal]HX߀XXX߀canbe \ viewedasconformingtoperceptualdemands.Whetherthelatterassumptionshouldbephonologicallyimplementedaswell(orremainsapurelyphoneticeffect)isunclear;itisviolatedatleastinlanguageswherethenasalbecomesmbeforew(seesection4.2.4,andalsotheCatalan  p casejustbelow).Thisquestionneednotberesolvedhere,sinceforourpurposestheimportantfactissimplythatpartialassimilationoccursunderphonologicalcontrol. 4.1.2Palatals     r Thefeaturalmakeupofpalatalandalveopalatalsegmentsisnotamatterofconsistentagreement,butawidelyheldviewduetoKeating(1988)treatsthemascomplexsegments,both[coronal]and[dorsal].Inthislight,consideracaseofpartialassimilationtopalatalsfromCatalan.Catalanhasthefollowingcoronalsegments,specifiedroughly(andfully)asfollows: #  17       l   X  ;&   ;2(  31  )3  Ԁ r CoronalsofCatalan(Mascar;1976,Recasens1991,1993)  6,X XXr X6  t/d n/l/r/s/z h #X߀XXHX߀%#s/    c/_ p   j    dental alveolar h postalveolar  alveopalatal  palatal `    [cor] [cor]   h [cor]    [cor] p   [cor] L    [+ant] [+ant]   h [ant]    [ant] p   [ant] 8    [+dist] [dist]   h [+dist]    [+dist] p   [+dist] $ t       h (dorsal?)  [dorsal]  [dorsal]  `       h (back?)  [back]  [back]  L        h     [+high]  [+high]  8  :Xr X,X X?:;' ;̛ThesefeaturespecificationsfollowMascar;(1976),whospecifiespostalveolarsas[+cor,ant,+dist,back]andspecifiesthe(alveo)palatalsasadditionally[+high].TheyalsoseemconsistentwiththecarefuldescriptionsofRecasens(1991,1993)(modulofootnote jamie17  ).Accordingtothesereferences,NPAbyn,whenoccurringbefore_andjinCentralCatalan,resultsnotinc,an   identicallyspecified(forplace)phonemeofthelanguage,butinapostalveolarnasaltranscribed p [n,].Thisistheplaceofarticulationofs/,andnappearsas[n,]beforethesesegmentsaswell. \ RepresentativeexamplesofNPAbeforethesecoronalsaregivenbelow.2(  32  )3  Ԁ r   Ԝso[n,][]ermans  'theyarebrothers'  p  r so[n,][_]iures   'theyarefree'  \  r u[n,][j]ogurt   'ayogurt' H ЛAccordingtoRecasens,thispartialassimilationisobligatory.Incontrast,assimilationbeforecis   total,givinggeminatecc(perhapsbecausen,cisasequenceoftoosimilarthingsinsomesense,   cf.Steriade's1982SharedFeaturesConvention).Recasenssuggeststhatassimilationcanbetotalbeforeallofthe(alveo)palatalsinsomeotherCatalandialects.Inordertocapturethepartialassimilationseenabove,Mascar;simplyomitted[high]fromaruleofNPAemployinggreekvariablenotation.ThissimpletackbecomesimpossibleinFG,iftherelevantprocessisPlace l nodespreading.InFCTwemighthypothesizethatNPAsucceedsincausinglinkageofonlythecoronalrelatedfeatures[coronal,ant,+dist]before(alveo)palatals,dueagaintotheworkingsof*CompSeg.Whetherthisiscorrectornot,(alveo)palatalsaremorecomplexthancoronalsliket 0"  underanytheory,andassimilationhereisinsomewaypartial. r ItismoredifficulttosaywhetherthisprocessshouldbeviewedaspartofaphonologicalNPA.Mascar;(1976)(andmanysubsequentworks)includedthesepartialassimilationsunderageneralphonologicalruleofNPAinCatalan.Thisruleisobligatorywithinwords,butinCentralCatalanisoptionalacrosswords"exceptwhennappearsbeforeanothercoronal,asin  odw30(32)  Հand '#& exampleslikeso[)n(dd0][)d(dd0]os'theyaretwo'.(SeeespeciallyRecasens1993,p.190.)Thesefactsmight (#' meanthatgeneralcrosswordNPAisaphoneticprocess,and/orthatassimilationbeforecoronalsisaprocessseparatefromgeneralNPA.ThereismuchinterdialectalvariationinthedegreeofobligatorinessofNPAacrosswordsinCatalan,however,andthefactsneedfurtherstudy.Asa h+&* separatematter,itisinterestingthatpartialassimilationproducesabasically[coronal]segment,ratherthana[dorsal]one,inapparentcontradictiontoOhala'sgeneralizationnotedearlier.Perhapsthisisbecausethe[dorsal]componentof(alveo)palatalsisvocalic,andsocannotrendernasalityeffectivelyvelarasitcanwithothercomplexsegmenttypes.Further,assimilationtothestop[coronal]articulationisfavoredforreasonsofstructurepreservation(seenextsection). 4.1.3Glides  8   r Theglidewisphoneticallylabiovelar,andjiscoronovelarifthisistherightunderstanding  ` ofpalatals(Keating1988),asnotedabove.Yettheglidesraisequestionsoftheirown,andsomeritseparatemention. r Assimilationbynasalstowinlanguagesistypicallypartial.Thefollowingarejust   representativeillustrationsofthemanyexistingalternations.ManyothersarecitedinOhalaandLorentz(1977),OhalaandOhala(1993),andCahill(1995),andtheyareeasytofindbyperusinggrammars.&  2 (  33  )3  Ԁ r   NPAtow \ E% ,X,h XXr XE  0  0(#(#[labial]onlyNdali: h N+wmb p Herbert(1986)4(#(# Ѐ          " Chukchee h N+wmw p Skorik(1961)  p '\      [dorsal]onlyKihungan: h N+wgw p Clements(1987) H           " Bakweri h N+ww p Kingston(p.c.) 4 ЛThoughassimilationofthe[dorsal]articulationappearsbyfarmorecommon,examplesinvolving[labial]exist,asshownabove.wisthereforedifferentfromclicksandperhapslabiovelarstops  (seebelow)inthatthereisachoiceofwhicharticulatortospread.Weleaveopenherethequestionofwhywallowsthischoice,andwhythepreferenceneverthelessremainsfor[dorsal]  spreading.  ItseemsclearthatatleastmanycasesofNPAgivingresultsasin  odw31(33)  Հbeforewaretruly D! phonological.ConsideroneexamplefromLuganda(Ashtonetal.1954,Chesswas1963;segmentsofinterestbolded):2!(  34  )3    Ԁ  NPAinLuganda $D #     a.  mp andiika h 'Iwrite'   p cf.oku w andiika &"%         a mp ulira h 'Hehearsme' p cf.oku w ulira ' #&   (#' &       b.  mb ala "  h 'Icount'          mm o m ba h 'Iescape'   p cf.oku b omba          nd ya "  h 'Ieat'   p cf.oku l ya          nn o n da h 'Ichoose'   p cf.oku l onda  '    c.  nzib a "  h 'Isteal'   p cf.oku bb a \          nzit a "  h 'Ikill'   p cf.oku tt a L          nzis a "  h 'Ibreathe'   p cf.oku ss a <  ЛThisNPA(ofwhichonlyaselectionofformsisgiven)evincespropertiesnotlikelytofollowϜdirectlyfromphoneticaccountsinvolvingoverlapofarticulationsoracoustic/perceptualproperties.Thehardeningofwtopitselfseenin  odw32(34)  aisarguablyone.Moretellingevenare @  thoseseenin  odw32(34)  bc.IntheformerweseeNPAgivingeitherNCorNNasaresult,dependingonwhetheranasalconsonantappearsinthesecondsyllableofthestem.ThelatterillustratesaseemingsubstitutionofthesequencenziCforallstemsbeginningingeminateCC"inorderto   indicatefirstpersonsingularmarking,otherwisemarkedbyastraightforwardarchiphonemicsegmentN(seereferencescitedfordetails).ThepointisthatthisNPA,likemanyothersincluding  partialassimilationtoglides,isgrammaticalinnature.  Inwhatsenseisassimilationtowpartial?Thoughthissegmentisphoneticallycomplex,itcan P beviewedasasimple[labial]or[dorsal]segmentundertheoriesofunderspecification,sincethesepropertiesofwaremutuallydependent(seeforexampleSteriade1987onu).Thismutual (x dependenceoftheplacefeaturesisaproblemforunderspecificationinitself,asSteriade(1995a)notes,sinceitmeansthatchoiceof[labial]vs.[dorsal]asunderlyingisentirelyarbitraryinthetheory.Anotherproblemfortheunderspecificationistviewofwinvolvescaseswherethissegment < behavesasboth[labial]and[dorsal]withinalanguage,dependingontheprocessconsidered(seeOhalaandOhala1993:237forexamples).WithinOT,atheoryofoutputconstraints,formslikeϜgwareunavoidablypartiallyassimilated,sincewis(usually)both[labial]and[dorsal]inthe  output.Thoughtheresultofpartialassimilationtowcanclearlybeeludedbyinvocationof  intermediaterepresentations(partiallyspecified)orotherdevices,nothingseemsgainedbytheattempt,anditispreferabletoconfrontthefactsastheypresentthemselves.  LugandampandNdalimbofcourseavoidsurfacepartialassimilation,bysimplifyingthe L"  hardenedtriggersegmentto[labial].(HereSpreadisfullysatisfied,attheexpenseofanFaithRel 8#! Мviolation.)Resultslikegw,wandmw,ontheotherhand,showpartialassimilationinthe $$t" output.Hardeningor'obstruentization'ofw,asinmbandgw,followsapatternseengenerally %` # forcontinuanttriggersofNPA,andisduetoaninteractionoffactorsexploredinPadgett(1994,1995c).Briefly,articulatorfeaturesandoralstricturefeatureslike[continuant]and[consonantal]areviewedas'bundled'togetherinaunitcalledanarticulatorgroup,analogoustothe'gesture'ofBrowmanandGoldstein(1986etseq.).Placeassimilationthereforemeansassimilationofplaceandoralstrictureproperties.Second,nasalitycombinedwith[cons]and/or[+cont]ishighlymarkedinlanguages.NPAtoafricativeorglidethreatenstocreatesuchfeaturecooccurrences;hardeningoftheassimilated'gesture'isjustoneofmanywaystoavoidthisundesirableoutcome. +&* МHX߀XXX߀LetArtSmeana[+cons,cont]articulation,andArtVa[cons]articulation#X߀XXHX߀#HX߀XXX߀;partialassimilation  andhardeninginthecaseof#X߀XXHX߀#gwHX߀XXX߀ԛareillustratedbelow.  IXr X ,X,h XyI2"(  35  )3  Ԁ r N+w#X߀XXHX߀E#gw t  r   N +  w h     Ngw L   r      /\ h     /\/\ 8   r [+nas] LabVDorsV  [+nas]DorsSLabV $ t ̛TherepresentationderivedisarguablythecorrectoneforwHX߀XXX߀ԛaswell;hardeninginthiscaseis  L  covert(forargumentsseeSteriade1993aandPadgettreferences). r Partialassimilationtotheglidejoccursalso.ThisseemsclearestincaseslikeKihunganN+   j#X߀XXHX߀Z#gj(Clements1987),wherethenasalispurely[dorsal].(Contrastsuchcaseswiththatof   Catalandiscussedabove,where[coronal]assimilates.)Forreasonsofspaceweleavethediscussionofglideshere.SeePadgett(1995c)formorediscussion. 4.2LabiovelarsG  \  r Labiovelarslikekpandgbmakeupawellknownclassofcomplexsegments.Though 4 certaininstanceshavesometimesbeentreatedassimple[labial]or[dorsal]accordingtogapsinalanguage'sinventory(seeespeciallyAnderson1976),mostareclearlyphonologicallycomplex.InG forexample,aKwalanguagespokeninGhana,theseriesofphonemeskp,gb,mcontrast H withbothp,b,m(respectively)andk,g,.NPAbeforelabiovelarsproducesalabiovelarnasalin 4 manylanguages,asintheKpelleexamplemgbi'myself'seeninsection2.2(seeespecially   Sagey1986onsuchassimilations).However,partialassimilationoftheformgbandkphas   beenreportedforvariouscases,includingGonja(Painter1970:74),Nkonya,Efik(OhalaandLorentz1977,citingReineke1972andCook1969respectively),Birifor,Konkomba,Anufo,ϜK<nni,andotherscitedbyCahill(1995)(seereferencesthere).Workingfromhisownfieldnotes,  CahillnotestheoccurrenceofbothtypesofassimilationinK<nni:NPAispartialwithinwords l (includingcompounds)andtotalacrosswords.HereweexaminesomethingclosetothereverseinG "assimilationistotalwithinamorphemebutpartialacrossamorphemeboundary. r G)a``0ԛhasNPAacrossallplaces,asshownbelow(Berry1951,Kropp1966,Kotey1974, #l! Ryder1987).Wordinitialnasalsaresyllabic.Assimilationtothelabiovelarsproducesthecomplexnasalintheformscitedhere,asinKpelle.(Hyphenindicatesamorphemeboundary.)2#(  36  )3  Ԁ r  Z GeneralNPAinG  &"%  r )jXX0HX߀XXX߀e mb a 'character' h     b#X߀XXHX߀#D k e p   'near' (#'  r )m,Ԅb ibii 'mychildren'schildren'  )||,Ԅk lDmpD  'mybasin' )$(  r bo ns o 'whale'   h     )||,mkp ai  'libation' *%)  r )npp,Ԅt ao< 'Iwant' h     )||,mkp okua  'nipple' t+&* Ѐ r   0  (# (# ЛHowever,inothercasesassimilationtolabiovelarsisonlypartial.Therelevantdifferenceliesinthemorphologicalmakeupoftheword:withinmorphemesassimilationistotal,asseenabove;acrossamorphemeboundaryassimilationispartial. #  18      כInthefirsttwoexamplesbelow,the t assimilatingnasalisareducedformofthefirstpersonprefix/mi/(accordingtothereferencesonG thisreductionischaracteristicofnormalspeech).Inthiscaseassimilationproducesapurevelar,asshown.HX߀XXX߀ԛ 8  2$(  37  )3    Ԁ0 r PartialNPAacrossmorphemeboundary `r(#r(#  r #X߀XXHX߀[#)||,Ԅgb ekD   'mychild'  8   r )||,Ԅkp ai   'mycheeks' (   r taa kp ee   'sisal'    r Adistinctionlikethisbasedonmorphologicalboundariesordomainsinvitesvariouspossibleanalyses.Earlierversionsofthiswork{@ #  19      HX߀XXX߀Ԁaccountedforpartialassimilationacrossthe   morphemeboundarybyrelyingonaconstraintoffeaturestemalignment(or'crispness',seeIt=andMester1994)thatmilitatedagainstcrossmorphemelinkage.#X߀XXHX߀g#HX߀XXX߀Asithappens,therelianceon h alignmentinthiscaseleadstocomplicationsintheaccountofNPA.Asimpleranalysisispossiblebycapitalizingontheassumptionthatconstraintscanberelativizeddirectlytomorphologicaldomainorcategory.Inparticular,supposethatsegmentalmarkednessiscomputeddifferentlyinaffixes(atleastthoseconsideredhere)thaninstems.AdaptingMcCarthyandPrinces(1995)suggestionthataffixesarephonologicallylessmarkedasaclass,letthemarkednessconstraint*CompSegbearticulatedintotwosubconstraints,onefortheaffixes,*CompSegA,andthemore @ general*CompSeg.Ifaffixesarelessmarked,thentheexpectedrankingbetweenthese , constraintsis*CompSegA#X߀XXHX߀#HX߀XXX߀Ԁ>>*CompSeg#X߀XXHX߀#HX߀XXX߀Ԁ(complexsegmentsarelesstolerableinaffixes).  Buckley(1995)(citingCassimjee1994)arguesfortaggingconstraintswith(ultimately)morphologicaldomainsinasimilarfashion.Thereareinprincipleyetotherwaysofcapturingmorphologicalrelevance,includingincorporationofLexicalPhonologysnotionlevel(Kiparsky1985,Mohanan1986,Pulleyblank1986)#X߀XXHX߀\#HX߀XXX߀ԀintoOptimalityTheory(seeespeciallyMcCarthyand x Prince1993b,andcf.Orgun1994),andthedefinitionofsubgrammarsbasedon(re)rankingsofconstraints(It=andMester1995).Thepreciseanalyticalchoicemadehereistoalargeextentindependentofthebasicpointbeingmadeaboutpartialclassbehavior. r RecallthataconstraintSpread(Place)compelsmultiplelinking;whatleadstopartial $d" assimilationisthemarkednessofaderivedcomplexsegment,*CompSeg.Inthestemdomain %P # NPAistotal(seetheexampleswithcomplexsegmentsin Z odw35(36)  #X߀XXHX߀##HX߀XXX߀).Inorderforthistobepossible, %>*CompSeg.Inaffixesontheotherhand,#X߀XXHX߀#HX߀XXX߀ԀNPAisonlypartial &("% beforecomplexsegments(see  odw37(37)  #X߀XXHX߀#HX߀XXX߀);hencewehave*CompSegA>>Spread.Thebasicstem  affixdichotomythereforeliesinthecentralranking#X߀XXHX߀#HX߀XXX߀*CompSegA>>Spread#X߀XXHX߀k#ԚHX߀XXX߀Ԁ>>*CompSeg#X߀XXHX߀#HX߀XXX߀.   r Theroleoffaithfulnessintheanalysismustalsobeconsidered.Inorderforspreadingtooccuratall,werequireSpread>>Faith,asnotedinearliersections.Now,sinceunderlying ` complexsegments(including#X߀XXHX߀)#m)mustbeallowedtosurfacewhentheyoccurbeforeavowel,as L  inmDlD'bell',faithfulnessinsomeformmustdominatethetwo*CompSegconstraints.The 8  constraintFaithRelHX߀XXX߀Ԁdoesthisjob:thesesegmentsareinapositionofrelease.Thereforethereisno $ t crucialrankingbetween*CompSegandthegeneralFaithconstraint.ForourpurposesFaithRel#X߀XXHX߀#HX߀XXX߀  ` isanundominatedconstraint;tosimplifythings,weomititfromthefollowingtableaux,andcandidatesviolatingitwillnotbeconsidered.Thefullrankingtobeconsideredisthereforeimplementedbelow.Thenasalisarbitrarilygivenitsown[labial]specificationunderlyinglysimplytoillustratethenecessarysubordinationoffaithfulnesstoSpread(seesection3).   &  2%(  38  )3    Ԁ0 r TotalNPAwithinmorpheme:#X߀XXHX߀#mkpai'libation' r(#r(# *YM dddd dd"dd"dd"M(#(#,dd ,dd",dd",dd",dd"+  QQ p Input=/mkpai/ )Q$QQ )*CompSegA )Q$Q )Spread +Q\$Q +М*Compseg )Q$Q\ )МFaith +Q$Q +a.+mkpai T Ѐ\/̀L/D +!,|Q +Л 'T" ' 2\xT" 2М** ?x*T"\x ?М*** 4Q!T"x 4b.Nkpaì|̀L/D .$Q .Л 7- " 7*!* ?\x*!" ?М* ?x*""\x ?М* 4Q!#"x 4c.kpai T$ Ѐ\/\̀DL +!,&Q +Л 'T'" '*! 2\xT(" 2М* ?x*T)"\x ?М** 4Q!T*"x 4d.mkpaì||\̀LDL .$4!-Q .Л 7-\." 7*!** ?\x*\/" ?М* ?x*\0"\x ?/%#\1"  x /'̜Sincewearenotdealingyetwithcasesinvolvingtherelevantaffixes,thetoprankingconstraint*CompSegAHX߀XXX߀Ԁdoesnoworkhere.Candidates  odw36(38)  bdallviolateSpreadtosomedegree,andsoin ($x3 thissimplecase  odw36(38)  awins,thoughitviolates*CompSeg(andfaithfulness)morethantheother %d 4 candidatesdo.ThreeviolationsofFaitharecountedin  odw36(38)  asincethenasalbothlosesitsinput &P!5 [labial]specificationandgainslinkstothefollowing[labial]and[dorsal]features.Spreadis &<"6 violatedthreetimesin  odw36(38)  dsincetherearethreepossiblebutunachievedfeaturesegmentlinksinthetwosegmentconsonantalsequence,theonlysegmentswecounthere(seesection3formoredetailsontheseassumptions).  *%:  r Thingsaredifferentwhenweconsidercasesinvolvingtherelevantaffixes,asdemonstrated inthenexttableau,fortheexample/mi+kpai/#X߀XXHX߀#kpai'mycheeks'.Herethenasalisgenuinely  [labial]underlyingly;forsimplicitythevowellossandconcomitantsyllabicityofthisnasalareassumedhere.&  2&(  39  )3    Ԁ r PartialNPAacrossmorphemeboundary:kpaimycheeks L  *OM dddd dd"dd"dd"dd"YM(#(#,dd ,dd",dd",dd",dd"+  QQ $ t Input=/mkpai/ )Q QQ )*CompSegA )Q  Q )МSpread +Q\  Q +М*CompSegHX߀XX߀ )Q  Q\ )Faith +Q  Q +a.#X߀XXHX߀l#mkpai  X  Ѐ\/̀L/D .$0 Q .*! 7- X" 7 ?\x* X" ?М** ?x* X"\x ?*** 4Q! X"x 4b.Nkpaì|̀L/D .$8 Q .Л 7-` " 7М**! ?\x*` " ?М* ?x*` "\x ?* 4Q!` "x 4c.+HX߀XXX߀kpai  Ѐ\/\̀DL .$Q .Л 7-" 7М* ?\x*" ?* ?x* "\x ?** 4Q!!"x 4d.mkpaì||\̀LDL .$8$Q .Л 7-`%" 7М**!* ?\x*`&" ?М* ?x*`'"\x ?/%#`("  x /'L   odw38(39)  athistimeisleastharmonic,sinceitviolatesthetopranked*CompSegA.Oftheremaining @) threecandidates,  odw38(39)  cisthebest,sinceitrespectsSpreadtotheextentallowedby*CompSegA; ,* theextraviolationsseenin  odw38(39)  band  odw38(39)  darenotforcedandthereforenotpossible. r TheG caseservestodemonstratethattheargumentsforFCTseeninsection2.2stillholdgiventhenewassumptionsaboutNPA.LiketheearlierhypotheticalconstraintNpa, . Spread(Place)requiresmultiplelinkingofallPlacefeatures,anditisdemonstrablygradiently x / violable. #  20      ׀#X߀XXHX߀# d!0 @  R References  <#2 0 r  Anderson,JohnandColinEwen.(1987).PrinciplesofDependencyPhonology.Cambridge %d 4 UniversityPress,Cambridge. r(#r(# 0 r  Anderson,StephenR.(1976).'OntheDescriptionofMultiplyArticulatedConsonants'.Journal &<"6 ofPhonetics4,1727.'(#7r(#r(# 0 r  Archangeli,DianaandDouglasPulleyblank.(1993).'Optimality,GroundingTheory,andRuleParameters'.RutgersOptimalityWorkshop,RutgersUniversity,NewBrunswick,NJ. r(#r(# 0 r  Archangeli,DianaandDouglasPulleyblank.(1994).GroundedPhonology.MITPress, &P!5 Cambridge,MA. r(#r(# 0 r  Ashton,E.O.,E.M.K.Mulira,E.G.M.NdawulaandA.N.Tucker.(1954).ALugandaGrammar. '(#7 Longmans,GreenandCo.,NewYork. r(#r(# 0 r  Bagemihl,Bruce.(1991).'SyllableStructureinBellaCoola'.LI22.4,589646.)%9r(#r(# 0 r  Beckman,JillN.(1995).'ShonaHeightHarmony:MarkednessandPositionalIdentity'.Ms.,UniversityofMassachusetts,Amherst. r(#r(# 0 r  Berry,J.(1951).ThePronunciationofGa.Heffer,Cambridge.r(#r(# 0 r  Bird,Steven.(1995).ComputationalPhonology:aConstraintBasedApproach.Cambridge t UniversityPress,Cambridge. r(#r(# 0 r  Bromberger,SylvainandMorrisHalle.(1989).'WhyPhonologyisDifferent'.LI20.1,5170.L r(#r(# 0 r  Browman,CatherineP.andLouisGoldstein.(1986).'TowardsanArticulatoryPhonology'.PhonologyYearbook3,219252.$ tr(#r(# 0 r  Browman,CatherineP.andLouisGoldstein.(1989).'ArticulatoryGesturesasPhonologicalUnits'.Phonology6.2,201251. L r(#r(# 0 r  Buckley,Eugene.(1995).'ConstraintDomainsinOptimalityTheory'.PaperpresentedattheGLOWWorkshop,UniversityofTroms. r(#r(# 0 r  Byrd,Dani.(1992).''.JournalofPhonetics r(#r(# 0 r  Cahill,Mike.(1995).'NasalAssimilationandLabiovelarGeometry'.PaperpresentedatACAL26.[Alsoms.,SILandOhioStateUniversity,1994]. r(#r(# 0 r  Cassimjee,Farida.(1994).'IsiXhosaTonology:anOptimalDomainsTheoryAnalysis'.Ms.,UniversityofIllinois. r(#r(# Chesswas,J.D.(1963).TheEssentialsofLuganda.OxfordUniversityPress,London. \ 0 r  Cho,YoungMeeYu.(1990).'ATypologyofVoicingAssimilation'.InAaronHalpern(ed.),ProceedingsoftheNinthAnnualWestCoastConferenceonFormalLinguistics,CSLI, 4 Stanford,141155. r(#r(# 0 r  Chomsky,NoamandMorrisHalle.(1968).TheSoundPatternofEnglish.HarperandRow,New  \ York. r(#r(# Л0 r  Clements,G.N.(1985).'TheGeometryofPhonologicalFeatures'.PhonologyYearbook2,225 4 52. r(#r(# 0 r  Clements,G.N.(1986).'CompensatoryLengtheningandConsonantGeminationinLuGanda'.InL.WetzelsandE.Sezer,eds.,StudiesinCompensatoryLengthening.Foris,Dordrecht.r(#r(# 0 r  Clements,G.N.(1987).'PhonologicalFeatureRepresentationandtheRepresentationofIntrusiveStops'.CLS23,PartII,ChicagoLinguisticSociety,2950.r(#r(# 0 r  Clements,G.N.andEnginSezer.(1982).'VowelandConsonantDisharmonyinTurkish'.InHarryv.d.HulstandNorvalSmith,eds.,TheStructureofPhonologicalRepresentations, X  PartII,Foris,Dordrecht,213255. r(#r(# 0 r  Cole,JenniferandCharlesW.Kisseberth.(1994).'AnOptimalDomainsTheoryofHarmony'.CognitiveScienceTechnicalReportUIUCBICS9402,UniversityofIllinois,UrbanaChampaign. r(#r(# 0 r  Cook,T.L.(1969).ThePronunciationofEfikforSpeakersofEnglish.AfricanStudiesProgram, $D # Bloomington. r(#r(# 0 r  Doke,ClementM.(1926).ThePhoneticsoftheZuluLanguage.UniversityoftheWitwatersrand &"% Press,Johannesburg. r(#r(# 0 r  Doke,ClementM.(1931).TextBookofZuluGrammar.Longmans,GreenandCo.,London.(#'r(#r(# 0 r  Duanmu,San.(1990).AFormalStudyofSyllable,Tone,StressandDomaininChinese )$( Languages.DoctoralDissertation,MIT,Cambridge,MA.|*%)r(#r(# mX߁XXX߁0 r  Feinstein,Mark.(1979).'PrenasalizationandSyllableStructure'.LI󛀜10.2,245278.#X߁XXmX߁.#rX߁XXX߁ԛh+&*r(#r(# 0 r  Gazdar,Gerald,EwenKlein,GeoffreyPullumandIvanSag.(1985).GeneralizedPhrase T,'+ StructureGrammar.HarvardUniversityPress,Cambridge,MA.@-(,r(#r(# #X߁XXrX߁I/#rX߁XXX߁0 r  Goldsmith,John.(1990).AutosegmentalandMetricalPhonology.BasilBlackwell,Cambridge,  МMA. r(#r(# 0 r  Halle,Morris.(1983).'OnDistinctiveFeaturesandTheirArticulatoryImplementation'.NLLT1,  91105. r(#r(# 0 r  Halle,Morris.(1995).'FeatureGeometryandFeatureSpreading'.LI26.1,146.`r(#r(# 0 r  Harris,JamesW.(1969).SpanishPhonology.MITPress,Cambridge,MA.L r(#r(# 0 r  Harris,JamesW.(1984).'AutosegmentalPhonology,LexicalPhonology,andSpanishNasals'.InMarkAronoffandRichardT.Oehrle(eds.),LanguageSoundStructure,MITPress, $ t Cambridge,MA,6782. r(#r(# 0 r  Hayes,Bruce.(1984).'ThePhoneticsandPhonologyofRussianVoicingAssimilation'.InMarkAronoffandRichardT.Oehrle(eds.),LanguageSoundStructure,MITPress,Cambridge,  8  MA,#X߁XXrX߁0#rX߁XXX߁318328.$ r(#r(# 0 r  Hayes,Bruce.(1990).'DiphthongisationandCoindexing'.Phonology7.1,3171.#X߁XXrX߁+5#rX߁XXX߁ԛ r(#r(# 0 r  #X߁XXrX߁6#mX߁XXX߁0rr(#r(#  Herbert,R.K.(1975).'ReanalyzingPrenasalizedConsonants'.StudiesinAfricanLinguistics6,   М105123. r(#r(# 0 r  Herbert,R.K.(1986).LanguageUniversals,MarkednessTheory,andNaturalPhonetic   Processes.MoutondeGruyter,NewYork.pr(#r(# 0 r  Hoard,J.E.(1978).'SyllabificationinNorthwestIndianLanguages,withRemarksontheNatureofSyllabicStopsandAffricates'.InA.BellandJ.B.Hooper(eds.),SyllablesandSegments, H NorthHolland,Amsterdam.#X߁XXmX߁|6#rX߁XXX߁ԛ4r(#r(# 0 r  It=,Junko.(1986).SyllableTheoryinProsodicPhonology.DoctoralDissertation,Universityof  p Massachusetts,Amherst.#X߁XXrX߁N9#rX߁XXX߁ \r(#r(# 0 r  It=,Junko.(1989).'AProsodicTheoryofEpenthesis'.NLLT7.2,217259.#X߁XXrX߁\:#rX߁XXX߁ԛHr(#r(# 0 r  It=,JunkoandR.ArminMester.(1993).'LicensedSegmentsandSafePaths'.InC.ParadisandD.LaCharit(eds.),ConstraintBasedTheoriesinMultilinearPhonology,specialissueof   theCanadianJournalofLinguistics. r(#r(# 0 r  It=,JunkoandR.ArminMester.(1994).'ReflectionsonCodaCondandAlignment'.InJasonMerchant,RachelWalkerandJayePadgett,eds.,PhonologyatSantaCruz,Linguistics  ResearchCenter,UCSC,SantaCruz,CA. r(#r(# 0 r  It=,JunkoandR.ArminMester.(1995).'TheCorePeripheryStructureoftheLexiconandConstraintsonReranking'.InPapersinOptimalityTheory,UniversityofMassachusetts X  OccasionalPapersinLinguistics(UMOP)18,GLSA,UMass,Amherst. r(#r(# 0 r  It=,Junko,R.ArminMesterandJayePadgett.(toappear).'NC:LicensingandUnderspecificationinOptimalityTheory'.LI.#l!r(#r(# 0 r  Jun,Jongho.(1995).'PlaceAssimilationastheResultofConflictingPerceptualandArticulatoryConstraints'.PaperpresentedattheFourteenthAnnualWestCoastConferenceonFormalLinguistics. r(#r(# 0 r  Kaun,Abigail.(1994).'AnOptimalityTheoreticAccountofRoundingHarmonyTypology'.PaperpresentedattheWestCoastConferenceonFormalLinguistics. r(#r(# 0 r  Keating,Patricia.(1988).'PalatalsasComplexSegments:XrayEvidence'.UCLAWorking (#' PapersinPhonetics69,7791.#X߁XXrX߁*;#rX߁XXX߁ԛ)$(r(#r(# ;;#X߁XXrX߁GB#0 r  Kingston,John.(1990).'ArticulatoryBinding'.InJohnKingstonandMaryE.Beckman(eds), |*%) PapersinLaboratoryPhonologyI.CambridgeUniversityPress,406434.h+&*r(#r(# 0 r  Kiparsky,Paul.(1973).'"Elsewhere"inPhonology'.InStephenR.AndersonandPaulKiparsky(eds.),AFestschriftforMorrisHalle,Holt,RinehartandWinston,NewYork,93106.@-(,r(#r(# 0 r  Kiparsky,Paul.(1982).'LexicalPhonologyandMorphology'.InI.S.Yang(ed.),Linguisticsin  theMorningCalm,vol.2,Hanshin,Seoul,391.r(#r(# 0 r  Kiparsky,Paul.(1985).'SomeConsequencesofLexicalPhonology'.Phonology2,85138.r(#r(# 0 r  Kiparsky,Paul.(1993).'RemarksonMarkedness'.SecondTrilateralPhonologyWeekend,UCSantaCruz. r(#r(# 0 r  Kirchner,Robert.(1993).'TurkishVowelHarmonyandDisharmony:anOptimalityTheoreticAccount'.Ms.,UCLA,oftalkpresentedattheRutgersOptimalityWorkshopI,NewBrunswick,NJ. r(#r(# 0 r  Kotey,PaulF.Amon.(1974).'ConsonantalLabializationandConsonantClustersinG:)a``0:'.The  ` JournalofWestAfricanLanguages.9.1,4956. L r(#r(# 0 r  Kropp,M.E.(1966).'TheMorphologyofthe:::Ga|X XXX߁&#X߁X X|XV#rX߁XXX߁ԛ:::;;HX߁XXX߁;;;G)a``0;;;#X߁XXHX߁lW#;;;;;;;;;;;;AspectSystem'.JournalofAfricanLangauges  8  5.2,121127. r(#r(# 0 r  Lass,Roger.(1976).EnglishPhonologyandPhonologyTheory.CambridgeStudiesin   Linguistics14,CambridgeUniversityPress. r(#r(# 0 r  Lass,RogerandJohnAnderson.(1975).OldEnglishPhonology.CambridgeStudiesin   Linguistics17,CambridgeUniversityPress. r(#r(# М:#X߁XXrX߁;W#:rX߁XXX߁ԛ0 r  Lombardi,Linda.(1991).LaryngealFeaturesandLaryngealNeutralization.Doctoral p МDissertation,UniversityofMassachusetts,Amherst.[PublishedbyGarlandPress,1994]. r(#r(# 0 r  Lombardi,Linda.(1995a).'PositionalFaithfulnessandthePhonologyofVoicinginOptimalityTheory'.Ms.,UniversityofMaryland,CollegePark. r(#r(# 0 r  Lombardi,Linda.(1995b).'LaryngealNeutralizationandAlignment'.InPapersinOptimality  p Theory,UniversityofMassachusettsOccasionalPapersinLinguistics(UMOP)18,GLSA,  \ UMass,Amherst. r(#r(# 0 r  McCarthy,JohnJ.(1988).'FeatureGeometryandDependency:AReview'.Phonetica43,84 4 108. r(#r(# 0 r  McCarthy,JohnJ.(1994).'OnCoronal"Transparency"'.PaperpresentedattheannualTrilateralPhonologyWeekend,UCSantaCruz. r(#r(# М0 r  McCarthy,JohnJ.andAlanPrince.(1993a).'GeneralizedAlignment'.Ms.,UniversityofMassachusetts,Amherst,andRutgersUniversity,NewBrunswick. r(#r(# #X߁XXrX߁;L#rX߁XXX߁ԛ0 r  McCarthy,JohnJ.andAlanPrince.(1993b).'ProsodicMorphologyI:ConstraintInteractionand l Satisfaction'.Ms.,UniversityofMassachusetts,Amherst,andRutgersUniversity,NewϜBrunswick. r(#r(# 0 r  McCarthy,JohnJ.andAlanPrince.(1995).'FaithfulnessandReduplicativeIdentity'.Ms.,UniversityofMassachusetts,AmherstandRutgersUniversity,NewBrunswick,NJ. r(#r(# 0 r  McCawley,James.(1967).'TheRoleofaPhonologicalFeatureSysteminaTheoryofLanguage'.Languages8,112123.$D #r(#r(# 0 r  Maddieson,IanandPeterLadefoged.(1989).'MultiplyArticulatedSegmentsandtheFeatureHierarchy'.InUCLAWorkingPapersinPhonetics72,116138.&"%r(#r(# 0 r  Mascar;,Joan.(1976).CatalanPhonologyandthePhonologicalCycle.DoctoralDissertation, '#& MIT,Cambridge,MA. r(#r(# 0 r  Mascar;,Joan.(1983).'PhonologicalLevelsandAssimilatoryProcesses'.Ms.,UniversitatAutAnomadeBarcelona. r(#r(# 0 r  Mester,R.ArminandJunkoIt=.(1989).'FeaturePredictabilityandUnderspecification:PalatalProsodyinJapaneseMimetics'.Lg65,258293.T,'+r(#r(# 0 r  Mohanan,K.P.(1983).'TheStructureoftheMelody'.Ms.,MIT,Cambridge,MA.@-(,r(#r(# 0 r  Mohanan,K.P.(#X߁XXrX߁qQ#rX߁XXX߁1986).TheTheoryofLexicalPhonology.Reidel,Dordrecht.r(#r(# 0 r  Mohanan,K.P.(#X߁XXrX߁W#rX߁XXX߁1991).'OntheBasesofRadicalUnderspecification'.NLLT9,285325.r(#r(# 0 r  Mohanan,K.P.(#X߁XXrX߁X#rX߁XXX߁1993).'FieldsofAttractioninPhonology'.InJohnGoldsmith(ed.),TheLast  PhonologicalRule:ReflectionsonConstraintsandDerivations.UniversityofChicago t Press,61116. r(#r(# 0 r  NChiosin,Mire.(1995).'BarraGaelicVowelCopyand(Non)ConstituentSpreading'.InProceedingsofWCCFL13.8 r(#r(# 0 r  NChiosin,MireandJayePadgett.(1993).'InherentVPlace'.Reportno.LRC9309,LinguisticsResearchCenter,UCSantaCruz. r(#r(# 0 r  Odden,David.(1991).'VowelGeometry'.Phonology8,26190. L r(#r(# 0 r  Ohala,JohnJ.(1990).'ThePhoneticsandPhonologyofAspectsofAssimilation'.InJohnKingstonandMaryE.Beckman(eds),PapersinLaboratoryPhonologyI.Cambridge $  UniversityPress,#X߁XXrX߁Y#rX߁XXX߁258275. r(#r(# 0 r  Ohala,JohnJ.andJamesLorentz.(1977).'TheStoryof[w]:anExerciseinthePhoneticExplanationforSoundPatterns'.InProceedingsofthe3rdAnnualMeetingoftheBerkeley   LinguisticsSociety,BLS,UCBerkeley. r(#r(# 0 r  Ohala,JohnJ.andManjariOhala.(1993).'ThePhoneticsofNasalPhonology:TheoremsandData'.InHuffmanandKrakow(1993),225249. r(#r(# 0 r  Orgun,Orhan.(1994).'MonotonicCyclicityandOptimalityTheory'.InproceedingsofNELS24,461474. r(#r(# 0 r  Padgett,Jaye.(1991).StrictureinFeatureGeometry.DoctoralDissertation,Universityof  p Massachusetts,Amherst.#X߁XXrX߁]#mX߁XXX߁ԛ \r(#r(# #X߁XXmX߁a#rX߁XXX߁0 r  Padgett,Jaye.(1994).'StrictureandNasalPlaceAssimilation'.NLLT12,465513.#X߁XXrX߁wa#rX߁XXX߁ԛHr(#r(# 0 r  Padgett,Jaye.(1995a).'FeatureClasses'.InPapersinOptimalityTheory,Universityof 4 MassachusettsOccasionalPapers(UMOP)18,GLSA,UMass,Amherst. r(#r(# 0 r  Padgett,Jaye.(1995b).'FeatureClassesII'.PaperpresentedattheHollandInstituteofLinguisticsPhonologyConference,Amsterdam.[Ms.inprogress]. r(#r(# 0 r  Padgett,Jaye(1995c).StrictureinFeatureGeometry.CSLIPublications,Stanford,CA.  [RevisionofPadgett1991]. r(#r(# 0 r  Painter,Colin.(1970).Gonja:aPhonologicalandGrammaticalStudy.IndianaUniversityPress, l МBloomington. r(#r(# 0 r  Paradis,Carole.(1988).'OnConstraintsandRepairStrategies'.LinguisticReview6,7197.D!r(#r(# 0 r  Pickett,J.M.,H.TimothyBunnellandSallyG.Revoile.(1995).'PhoneticsofIntervocalicConsonantPerception:RetrospectandProspect'.Phonetica52,140.#l!r(#r(# #X߁XXrX߁,b#mX߁XXX߁0 r  Piggott,G.L.(1992).'VariabilityinFeatureDependency:theCaseofNasality'.NLLT10.1,33 $X" 77.#X߁XXmX߁%g#ԛ$D #r(#r(# 0 r  Prince,AlanandPaulSmolensky.(1993).'OptimalityTheory:ConstraintInteractioninGenerativeGrammar'.Ms.,RutgersUniversity,NewBrunswick,andUniversityofColorado,Boulder. r(#r(# 0 r  Pulleyblank,Douglas.(1986).ToneinLexicalPhonology.Reidel,Dordrecht.(#'r(#r(# 0 r  Pulleyblank,Douglas.(1994).'VowelHarmonyinOptimalityTheory'.Toappear,Proceedingsof )$( theWorkshoponPhonology,UniversityofCoimbra,Portugal,September1993.|*%)r(#r(# 0 r  Recasens,Daniel.(1991).FonticaDescriptivadelCatal.InstitutD'EstudisCatalans,Biblioteca h+&*  FilolAgicaXXI,Barcelona.T,'+r(#r(# 0 r  Recasens,Daniel.(1993).FonticaiFonologia.BibliotecaUniversitria18,Enciclopdia   Catalana,Barcelona. r(#r(# 0 r  Reinecke,B.(1972).TheStructureoftheNkonyaLanguage.VEBVerlagEnzyklopdie,Leipzig.r(#r(# 0 r  Ryder,MaryEllen.(1987).'AnAutosegmentalTreatmentofAssimilationtoLabiovelars'.InproceedingsofCLS23PartTwo:ParasessiononAutosegmentalandMetricalPhonology,ChicagoLinguisticSociety. r(#r(# 0 r  Sagey,Elizabeth.(1986).TheRepresentationofFeaturesandRelationsinNonLinear 8  Phonology.DoctoralDissertation,MIT,Cambridge,MA.$ tr(#r(# 0 r  Sagey,Elizabeth.(1987).'NonConstituentSpreadinginBarraGaelic'.Ms.,SchoolofSocialSciences,UCIrvine. r(#r(# 0 r  Scobbie,JamesM.(1991).AttributeValuePhonology.DoctoralDissertation,Universityof  8  Edinburgh. r(#r(# 0 r  Scobbie,JamesM.(1993).'ConstraintViolationandConflictfromthePerspectiveofDeclarativePhonology'.CanadianJournalofLinguistics38,155167. r(#r(# 0 r  Selkirk,Elisabeth.(1982).'TheSyllable'.InHarryvanderHulstandNorvalSmith(eds),The   StructureofPhonologicalRepresentationsvol.2.Foris,Dordrecht,337384. r(#r(# 0 r  Selkirk,Elisabeth.(1991a).'MajorPlaceintheVowelSpace.VowelHeight'.PaperpresentedattheWorkshoponFeatureOrganization,LinguisticSocietyofAmericaInstitute,SantaCruz,CA.[manuscriptUniversityofMassachusetts,Amherst]. r(#r(# 0 r  Selkirk,Elisabeth.(1991b).'ATwoRootTheoryofLength'.InE.DunlapandJ.Padgett,eds.,UniversityofMassachusettsOccasionalPapersinLinguistics14:PapersinPhonology.GLSA,UMass,Amherst. r(#r(# 0 r  Selkirk,Elisabeth.(1994).Classnotes,Linguistics730,UniversityofMassachusetts,Amherst. r(#r(# 0 r  Skorik,P.1961.GrammatikabukotskogoJazyka.AkademijaNauk,MoscowLeningrad.4r(#r(# 0 r  Smolensky,Paul.(1993).'Optimality,Markedness,andUnderspecification'.PaperpresentedattheRutgersOptimalityWorkshopI,NewBrunswick,NJ. r(#r(# mX߁XXX߁0 r  Steriade,Donca.(1982).GreekProsodiesandtheNatureofSyllabification.Doctoral  Dissertation,MIT,Cambridge,MA. r(#r(# #X߁XXmX߁u#HX߁XXX߁0 r  Steriade,Donca.(#X߁XXHX߁v#rX߁XXX߁1987).'RedundantValues'.InAnnaBosch,BarbaraNeedandEricSchiller  (eds.),Papersfromthe23rdAnnualRegionalMeetingoftheChicagoLinguisticSociety, l PartII,ChicagoLinguisticSociety,339362. r(#r(# #X߁XXrX߁)w#mX߁XXX߁0 r  Steriade,Donca.(1993a).'Closure,ReleaseandNasalContours'.inHuffmanandKrakow(1993), D! М401470.#X߁XXmX߁x#rX߁XXX߁ԛ0" r(#r(# 0 r  Steriade,Donca.(1993b).'PositionalNeutralization'.PaperpresentedatNELS24.#l!r(#r(# #X߁XXrX߁Yy#HX߁XXX߁0 r  Steriade,Donca.(#X߁XXHX߁Mz#rX߁XXX߁1994).'ComplexOnsetsasSingleSegments:theMazatecoPattern'.InJennifer $X" ColeandCharlesKisseberth(eds.),PerspectivesinPhonology,CSLIPublications, $D # Stanford,CA,203291. r(#r(# 0 r  Steriade,Donca.(1995a).'UnderspecificationandMarkedness'.InJohnGoldsmith,ed.,HandbookofPhonologicalTheory,Blackwell,Oxford.'#&r(#r(# #X߁XXrX߁z#HX߁XXX߁0 r  Steriade,Donca.(#X߁XXHX߁|#rX߁XXX߁1995b).'LaryngealNeutralizationandLaryngealFeatures'.Paperpresentedat (#' theArizonaPhonologyConference,UniversityofArizona,Tuscon. r(#r(# 0 r  Trigo,RosariaLorenza.(1988).OnthePhonologicalDerivationandBehaviorofNasalGlides. |*%) DoctoralDissertation,MIT,Cambridge,MA. r(#r(#  0 r  Welmers,W.E.(1962).'ThePhonologyofKpelle',JournalofAfricanLanguages1,6993.T,'+r(#r(# 0 r  Zsiga,Elizabeth.(1993).Features,Gestures,andtheTemporalAspectsofPhonological   Organization.DoctoralDissertation,YaleUniversity,NewHaven,CT.r(#r(# DepartmentofLinguisticsStevensonCollegeUCSCSantaCruz,CA95064padgett@ling.ucsc.edu#X߁XXrX߁-}#