finnic stop gradation as an effect of mora ......finnic stop gradation as an effect of mora sharing...

KÜLLI PRILLOP (Tartu)

FINNIC STOP GRADATIONAS AN EFFECT OF MORA SHARING

Abstract. Finnic radical stop gradation affected stop consonants in the onset posi-tion of unstressed syllables. The stops were weakened before closed syllables,and remained unweakened before open syllables. Stop gradation, conditioned bywhether the following syllable was open or closed, is cross-linguistically extremelyrare. In this paper, gradation is argued to be a phonetically natural intervocalicweakening process, which was blocked only in the onsets of open unstressedsyllables due to restrictions on moraic structure. Under the proposed account,differences between Finnic languages in terms of gradation depend on the extentto which a specific language tolerated heavy unstressed syllables.

Keywords: Finnic, gradation, stops, geminate gradation, single stop gradation,Optimality Theory, mora sharing.

Introduction

The weakening of obstruents in sonorous environments, especially betweenvowels, is a very common phenomenon in languages all over the world.Stop gradation, historically conditioned by the presence or absence of acoda, appears to occur only in Uralic languages, being a characteristic featureof most Finnic and Saami languages, and a few Samoyedic languages.According to the view shared by most researchers, stop gradation devel-oped independently in Finno-Saamic languages and in Samoyedic languagesover three thousand years ago (Itkonen 1969; Laanest 1982; Gordon 1997and others). However, since the gradation systems in Finnic and Saamilanguages differ considerably in detail, several researchers have suggesteda strong likelihood of their independent development (Sammallahti 1998;Bye 2001).

Since gradation is a rare phenomenon in languages, providing an expla-nation of its causes is a challenge to any universalist linguistic theory. Inthe present paper, I will use moraic theory to outline the conditions whichhave triggered the development of stop gradation, and which may accountfor differences between Finnic languages in terms of its realization. Thepaper will cover the Finnic gradation of single stops and of geminates, butnot quantity gradation, which emerged much later and only in Estonian.

161

LINGUISTICA URALICA XLVII 2011 3 doi:10.3176/lu.2011.3.01

1 Linguistica Uralica 3 2011

The first chapter of the paper provides an overview of earlier theoriesof the development of gradation in Finnic languages. The second chapterdemonstrates that a simple explanation for the emergence of stop grada-tion can be offered by mora sharing, and highlights similarities betweenFinnic gradation and other similar phenomena observed cross-linguistically.The third chapter presents an Optimality Theoretic analysis of gradation.I will also provide a brief description of different types of gradation.

1. Earlier descriptions of the emergence of gradation

Originally, consonant gradation was not a phonological phenomenon. Theoccurrence of a strong or weak stop word-medially depended on the soundssurrounding it, on stress conditions and on syllable structure. The condi-tions giving rise to stop gradation were first described by Matthias Alek-santeri Castrén already in the middle of the 19th century (1839 : 15—16).Weak stops occurred (a) after a stressed syllable at the beginning of a closedsyllable between two vowels or a sonorant and a vowel (radical gradation),and (b) after an unstressed syllable intervocalically (suffixal gradation). Else-where, the stops were strong. Similarly to single stops, geminates alter-nated as well, at least when preceded by stressed syllables (Viitso 2003 :165).

The following examples are from the Estonian language. Periods marksyllable boundaries. Segment durations which bear no relevance to stopgradation are not indicated.

(1) Weak grade Strong grade

a. *jal.∑kan (> jala) ’foot, gen.sg’ *jal.ka (> jalg) ’foot, nom.sg’*ri.∑ttan (>� rea)� ’row,� gen.sg’ *ri.ta (>� rida)� ’row,� nom.sg’

b. *val.ke.∑tta.ta (>� valget)� ’white,� part.sg’c. *kät.∑ ten (>� kä ∑tte)� ’hand,� gen.pl’ *kät.tä (>� kä ªt�)� ’hand,� part.sg’

In�the�Finnic�period,�gradation�was�merely�a�quantitative�phenomenon,i.e.� it� only� affected� the� duration� of� stops.� Voicing� and� spirantization� ofstops�emerged�later,�taking�different�forms�in�different�languages�(see�Laa�-nest� 1982� :� 107;� Viitso� 1981� :� 176—177;� 2008� :� 120—121).� Languages� thatfeature� similar� differences� in� the� quantity� of� stops,� and� in� which� otherproperties� such� as� voicedness,� intensity� or� presence� of� aspiration� are� ofsecondary�importance,�can�also�be�found�in�other�language�families.�Perhapsthe� best� known� example� is� Swiss� German� (Kraehenmann� 2001).

A� controversial� question� is� whether� gradation� originally� involved� allconsonants,� as� it� does� in� Saami� languages,� or� only� stops,� as� in� Finniclanguages.� The� dominant� theory� at� the� beginning� of� the� previous� centurywas� that� of� Eemil� Nestor� Setälä,� according� to� whom� stop� gradation� origi-nally�involved�all�consonants.�Setälä’s�theory�was�subsequently�revived�byMatthew�Gordon,�who�describes�gradation�in�Finnic�and�Saami�languagesas� a� lengthening� of� consonants� which� serves� to� balance� the� foot� (Gordon1997).�According�to�the�predominating�view,�however,�the�only�consonantsoriginally� involved� in� gradation� were� stops,� and� in� Saami� languages� thegradation�of�other�consonants�developed�on�the�basis�of�analogy�(Pikamäe1957;� Ravila� 1960;� Korhonen� 1988� and� others).

Külli Prillop

162

Finnic Stop Gradation as an Effect of Mora Sharing

163

The� development� of� stop� gradation� can� be� viewed� either� as� a� processof� fortition� or� a� process� of� lenition.� In� both� cases� it� can� be� analysed� as� aphenomenon�that�was�triggered�by�specific�contexts,�or�one�which�involvedall� stops,� but� was� blocked� in� certain� environments.� In� the� case� of� Finniclanguages,�it�is�generally�considered�to�have�been�a�process�of�lenition,�notof� fortition� (Posti�1953;�Laanest�1982� :�106;�Korhonen�1988);�or�primarily�aprocess� of� lenition,� later� complemented� in� some� Finnic� languages� by� thegemination� of� single� consonants� which� were� preceded� by� a� short� stressedvowel� (Viitso� 1981� :� 178;� 2008� :� 122).

The�majority�of�accounts�treat�gradation�as�primarily�a�lenition�process.A� possible� trigger� of� the� mutation� could� have� been� an� excessively� longsyllable�in�a�weak�(i.e.�unstressed)�position.�Another�possibility�is�to�accountfor�the�lenition�of�closed�syllable�onsets�by�referring�to�foot�isochrony�(Viit�-so� 2003� :� 162;� Eek,� Meister� 2004� :� 345).� According� to Mikko Korhonen(1988),� radical� gradation� was� caused� by� the� need� for� greater� articulatoryenergy� in�closed�syllables� (compared� to�open�ones)� following� the�primarystressed� syllable.� Greater� articulatory� energy� causes� a� drop� in� intensity� atthe� �preceding� syllable� juncture� and� a� weakening� of� the� preceding� conso-nant� (Korhonen� 1988� :� 275).

Arvo� Eek� and� Toomas� Help� (1986)� describe� early� consonant� gradationas� a� result� of� an� increase� in� the� prominence� of� primary� stress.� In� Indo-European�languages�primary�stress�can�fall�on�different�syllables,��requiringspecial�attention.�According� to�Eek�and�Help,�under� Indo-European� influ-ence,� Finnic� peoples� tried� to� increase� the� prominence� of� primary� stressed(i.e.�initial)�syllables�in�their�language,�too.�In�order�to�retain�phonetic�coher-ence�in�the�word,�every�non-initial�syllable�had�to�be�weaker�than�the�initialone.� Excessively� strong� syllables� following� the� primary� stress� had� toweaken.� This� primarily� concerned� closed� syllables� which� started� with� astop�consonant�and�were�therefore�too�strong.�At�the�same�time,�open�sylla-bles� were� not� considered� as� strong� as� to� upset� the� foot.� Thus,� only� theonsets� of� closed� syllables� shortened� (Eek,� Help� 1986� :� 11—14).

Kalevi� Wiik� (1997)� has� also� considered� the� appearance� of� excessivelyheavy�syllables�in�an�unstressed�position�as�the�cause�of�the�Proto-Germanicconsonant� change� described� by� Verner’s� law.� According� to� that� law,� voice-less�word-medial�obstruents� in�a�voiced�environment�became�voiced�whenthey�preceded�word�stress�(later,�the�stress�shifted�to�the�first�syllable),�e.g.Sanskrit�pitár-�>>�Gothic�faðar ’father’,�Sanskrit�snussá >>�Old�English�snoru‘daughter-in-law’� (Campbell� 1999� :� 143—145).� Kalevi� Wiik� has� suggestedthat� Verner’s� law� is� connected� with� the� shift� of� stress� to� the� first� syllable.Long� syllables� starting� with� a� voiceless� obstruent� were� too� strong� in� anunstressed� position,� since� voiceless� obstruents� are� relatively� long� and� theytend�to�raise�the�fundamental�frequency�of�the�following�vowel,�making�thelistener� more� likely� to� perceive� such� syllables� as� stressed.� To� ensure� thatthe�listener�would�perceive�the�stress� in�its�correct�new�position,�the�voice-less� obstruent� in� the� onset� had� to� undergo� voicing� (Wiik� 1997� :� 273—274).

Theories� which� explain� Finnic� gradation� as� a� process� of� weakening� ofexcessively�strong�non-initial�syllables�are�problematic�in�many�ways.�First,they� do� not� explain� why� it� was� the� syllable� onset� that� shortened,� and� notthe�nucleus�or�coda.�The�supposedly�too�strong�contrast�between�the�obstru-ent� and� the� following� vowel� does� not� explain� the� different� development

1*

of� open� and� closed� syllables.� If� the� second� syllable� of� the� form� *jal.kan’foot,�gen.sg’�is�regarded�as�too�strong,�while�that�of�the�form�*jal.ka ’foot,nom.sg’� is� not,� then� the� reason� behind� the� excessive� strength� was� obvi-ously� not� only� the� stop� consonant� in� the� syllable� onset,� but� the� coda� aswell.�Syllables�without�a�coda,�even�when�they�began�with�a�stop,�did�notweaken� (unlike� Verner’s� law).

Secondly,� initial� stops� of� open� heavy� syllables� containing� diphthongsdid� not� shorten.� A� diphthong� was� created� in� the� non-initial� syllable� byadding� markers� starting� with� i to� stems� which� ended� with� a� vowel� (e.g.jalka +�i +�ta →�*jal.koi.ta ’foot,�part.pl’).�It�is�true�that�there�are�languageswhere�certain�diphthongs�are�regarded�as�monomoraic.�Still,�those�moderndialects,� where� instead� of� a� Vi-diphthong� there� is� a� long� vowel� (e.g.� theFinnish� South� Pohjanmaa� dialect:� annoin >� annoon ’I� gave’,� antoi >� antoo’s/he� gave’),� which� has� not� affected� consonant� gradation,� do� not� supportthe�hypothesis�of�light�Vi-diphthongs�in�Proto-Finnic�dialects.�A�long�vowelis�bimoraic�by�definition�and�the�addition�of�a�mora�in�an�unstressed��syllablewould� be� difficult� to� justify.

The�southwestern�dialects�of�Finnish�contain�inflected�forms�where�thestop� consonant� has� weakened� before� a� diphthong,� such� as� *an.toi > anno’s/he�gave’� (cf.� the�Estonian�*an.tan > annan ’I�give’�and�standard�Finnishantoi ’s/he� gave’)� (Laanest� 1982� :� 110).� The� fact� that� the� same� change� didnot� occur� in� standard� Finnish� —� apparently� because� syllables� consistingof�the�same�sounds�were�regarded�as�light�ones�—�raises�a�legitimate�ques-tion:�why�did�the�moraic�i disappear�from h e a v y diphthongs�in�south-western�dialects�(*an.toi > anno),�whereas�it�was�preserved�in l i g h t diph-thongs� in� standard� Finnish� (an.toi, not� *an.to)?� After� all,� a� simpler� expla-nation� would� be� that� in� southwestern� dialects� the� i was� a� glide� whichfunctioned� as� a� syllable-closing� consonant� (see� Laanest� 1982� :� 110).

Matthew�Gordon�(1997)�has�described�consonant�gradation�as�a�processof�fortition.�According�to�his�hypothesis,�what�happened�first�was�the�length-ening� of� the� open� vowel� at� the� end� of� the� foot.� In� order� to� balance� thefoot,� this� was� followed� by� the� lengthening/gemination� of� the� stop� conso-nant� preceding� the� lengthened� vowel.� According� to� Gordon,� the� vowel� ofa� closed� syllable� did� not� lengthen� (cf.� the� half-long� vowels� at� the� end� ofcertain�words�in�modern�Estonian,�e.g.�lugù ’story’),�consequently�the�conso-nant� in� the� onset� of� the� closed� syllable� did� not� lengthen� either.� Gordonadmits� that�such�a�scheme�of�development�would�presuppose� the� lack� (orinfrequent�occurrence)�of�long�stressed�vowels�in�the�language,�since�a�longvowel� would� be� sufficiently� heavy� to� block� the� foot� balancing� process(Gordon� 1997� :� 55—58).� At� the� same� time,� Gordon� does� not� explain� whya� word� with� a� lengthened� stop� consonant� (e.g.� *[lɑp ‚�ɑ ‚�]1 ’flat’)� should� beregarded� as� acceptable,� whereas� a� word� with� a� geminate� consonant� (e.g.*[pɑppɑ ‚�]� ’priest’)� is� not,� such� that� in� the� latter� the� stop� consonant� had� tolengthen� even� more� due� to� the� lengthened� a,� resulting� in� *[pɑpp ‚�ɑ ‚�].

The�examples�that�Gordon�(1997�:�59—60)�gives�of�consonant��lengtheningin�different�languages�due�to�the�length�of�the�following�vowel�in�fact�onlyconcern� single� consonants� preceded� by� a� short� stressed� vowel:� the� Votic

Külli Prillop

164

1 I have not changed Gordon’s (1997) transcriptions here. ‚� indicates a segmentintermediate in duration between a short segment and a long one. Long segmentis indicated by :.

ellää ’s/he�lives’,�the�Ingrian�tulloo ’s/he�comes’,�the�Savo kalloo ’fish, part.pl’,the South Estonian kannaa2 ’chicken, part.sg’.�All�these�examples�representsubsequent� developments� in� Finnic� languages� and� relate� to� the� period� inwhich� long� vowels� had� appeared� in� non-initial� syllables,� because� the� inter-vocalic�*d had�been�lost.�Theoretically,�it�cannot�be�ruled�out�that�similar�length-ening� took� place� in� front� of� long� vowels� also� during� the� �preceding� period(Finnic-Saami),� yet� it� is� only� common� when� preceded� by� a� CV� �syllable,� anddoes�not�explain�weakening�after�a�heavy� initial� �syllable,�e.g.� *lau.∑t�an >�Stan-dard� Estonian� laua,� Standard� Finnish� laudan ’board,� gen.sg’.

For�instance,�measurements�of�consonant�duration�in�the�Soikkola�dialectof�Ingrian�(Gordon�2009�:�93—96)�showed�that�a�lengthened�consonant�onlyhas� the�same�duration�as�a� lexical�geminate�when� it� is�preceded�by�a� shortvowel:�the�lengthened�l in�the�form�kalaa →�kallaa ’fish,�part.sg’�has�a�dura-tion�that�is�approximately�double�the�duration�of�the�l in�the�form�kala ’fish,nom.sg’.� A� lengthened� consonant� preceded� by� a� closed� syllable� or� one� thatincludes� a� long� vowel� has� a� duration� inferior� to� that� of� a� lexical� geminateand� may� actually� have� the� same� duration� as� a� consonant� in� the� same� posi-tion�which�has�not�undergone�lengthening.�For�instance,�the�duration�of�thestop� in� the� forms� riipuu → riippuu ’kerchief,� part.sg’� and� riipu is� the� same(Gordon�2009� :�93—96).�The�data�collected�by�Arvo�Laanest�(1986� :�33,�57—59)� concerning� the� lengthening� potential� of� single� consonants� in� Ingriandialects� are� slightly� at� variance� with� those� of� Gordon,� yet� they� too� indicatethat� the� gemination� of� single� stops� occurs� consistently� only� when� the� stopis�preceded�by�a�stressed�CV�syllable,�and�that�the�variation�of�the�durationof� the�geminates�following�such�syllables�only�appears�as�a�rare�fluctuationwhich� Laanest� does� not� consider� worthy� of� a� more� detailed� discussion.

The� other� problem� with� an� account� based� on� the� balancing� of� foot� isthe� inevitable� gemination� of� intervocalic� consonants,� which� in� Finniclanguages,� unlike� the� Saami� ones,� has� not� taken� place� due� to� early� stopgradation.� Geminates� are� considered� to� straddle� the� syllable� juncture.Regardless� of� whether� the� lengthening� of� open� final� syllables� meant� addi-tion�of�moras�or�some�other�change,�compensation�for�that�lengthening�hadto� manifest� in� the� short� initial� syllable� and� not� in� the� onset� of� the� alreadylong�second�syllable.�Since,�according�to�Gordon�(1997),� �additional�weightwas� attributed� to� the� intervocalic� consonant� and� not� to� the� vowel� of� theinitial�syllable,�the�consonant�had�to�geminate.�Word-medial�geminates�haveone� link� to� the� coda� of� the� preceding� syllable� and� another� to� the� onset� ofthe�following�one�(see�Hayes�1989�:�259).�The�structure�in�Figure�1b�is�unac-ceptable� because� moraic� onsets� are� prohibited� in� moraic� theory,� and� itstretches�the�long�second�syllable�even�more.�The�structure�1c�is��empiricallyflawed,� because� languages� prefer� onsetful� syllables� over� onsetless� ones.In many cases, an analogy to the account advanced by Gordon has been

used to explain gemination. For instance, in Livonian, in which gradationdeveloped later than in other Finnic languages and took a slightly differentroute, long vowels were already present in medial syllables by the timegradation appeared, because the intervocalic *d had� been� lost,� cf.� *tubada>>� *tubā� ’room,� part.sg’.� The� stop� on� the� boundary� between� the� first� andthe�second�syllable�geminated�as�a�counterbalance�to�the�long�vowel�in�the


165

2 I have not changed Gordon’s (1997) transcriptions here.

second�syllable,�cf.�*tubā� >�*tu’bba� >>�tu’bbõ (see�Viitso�2007�:�57—59).�Suchbalancing� may� be� interpreted� as� an� avoidance� of� éLH� feet,� as� has� beensuggested� in� respect� of� Livonian� by� Paul� Kiparsky� (2006� :� 11).� In� additionto� Ingrian�and�Livonian,�similar�geminates,� formed�due� to� the� influence�oflong�medial�syllables,�are�very�common�also�in�the�Võru�dialect�of��Estonian(Iva�2010�:�162).�Since�gemination�does�not�represent�the�same��phenomenonas�Finnic�stop�gradation�and�is�thus�irrelevant�to�the�discussion�here,�I�willnot� deal� with� alternative� accounts� of� gemination� any� further.

In� the� approaches� to� consonant� gradation� discussed� above,� change� isseen� as� resulting� from� the� context� in� which� the� consonants� appear.� Theapproaches�that�focus�on�lenition�see�it�as�resulting�from�excessive��syllableweight.�Approaches�focusing�on�fortition�regard�open�syllables�as�triggers.In� the� following� sections,� I� will� present� an� analysis� showing� that� contextmay� facilitate� the� blocking� of� stop� lenition� without� at� the� same� time�becoming� a� direct� cause� of� fortition.

2. Gradation — sharing moras to avoid lenition

2.1. Single stops

In�moraic� theory,�distinguishing�between�short�and� long�stops� in�a� �syllableonset� is�only�possible� if�we�allow�the�syllable�onset� to�be� linked� to�a�mora.Moraic� theory� does� not� allow� syllable� onset� to� be� linked� to� its� own� mora,since�syllable�onsets�play�no�role� in�assessing�the�syllable�as�heavy�or� light,i.e.� determining� the� location� of� stress� (Hyman� 1985� :� 15—16;� Hayes� 1995� :51—53),�and�the�loss�of�the�onset�does�not�trigger�compensatory��lengthening(Hayes�1989).�However,�a�syllable�onset�can�share�a�mora�with�the�nucleus.

There�are�three�common�ways�to�represent�monomoraic�CVC�syllables:(a)� the� coda� is� not� linked� to� any� mora� (Figure� 2a;� Zec� 2007� :� 175),� (b)� thecoda�shares�a�mora�with�the�preceding�vowel�(Figure�2b;�Hayes�1989�:�254;Kager� 1999� :� 147),� or� (c)� the� syllable� onset� is� linked� to� the� same� mora� asthe� syllable� nucleus� (Figure� 2c;� Hayes� 1995� :� 53;� de� Lacy� 2002� :� 14).� LarryM.� Hyman� (1985� :� 17)� links� all� segments� to� moras.� Researchers� usuallyselect� one� of� these� representations� and� use� it� exclusively.

Külli Prillop

166

!"

"

" " " "

#$" %&'()#*(+)"

!" " !" " " !" " !" " " " " " "

!" ! !" !"""

!" !! !" !" " " " " "

," !!"! ," " " ," !!"! ," " " " " " "

" " " " -$" ./)-#0#)1&2"3++*" 1$" .4)5&*0&55"5600#-0&"

" " " " " !" " !" " " !" " !" "

" " " " " !" !! !" !" " !" !! !" !"

" " " " " ," !!"! ," " " ," !!"! ," "

#$%&'(!)*"7)*&89+1#0(1"0&):*;&)():"*+"-#0#)1&"*;&"3++*"

7)"'#)6"1#5&5<"#)"#)#0+:6"*+"*;&"#11+=)*"#29#)1&2"-6"%+82+)";#5"-&&)"=5&2"*+"&>?0#()"

:&'()#*(+)$" @+8" ()5*#)1&<" ()" A(9+)(#)<" ()" B;(1;" :8#2#*(+)" 2&9&0+?&2" 0#*&8" *;#)" ()" +*;&8" @())(1"

0#):=#:&5"#)2"*++C"#"50(:;*06"2(33&8&)*"8+=*&<"0+):"9+B&05"B&8&"#08&#26"?8&5&)*"()"'&2(#0"5600#-0&5"

-6"*;&"*('&":8#2#*(+)"#??&#8&2<"-&1#=5&"*;&"()*&89+1#0(1".!";#2"-&&)"0+5*<"13$"."#$%!%"DD".!"#$"

"#$$%<"5:$&'#(")*+,-*.($&*$/*(,-"-+=)2#86"-&*B&&)"*;&"3(85*"#)2"*;&"5&1+)2"5600#-0&":&'()#*&2"#5"

#"1+=)*&8-#0#)1&"*+"*;&"0+):"9+B&0"()"*;&"5&1+)2"5600#-0&<"13$".!"#$"D".!"%##$"DD"!"%##&'0.--*122(.$*

34456* 75879:)* ;<=,* >'?'/=2/@* %'A* >-* 2/(-#&#-(-B* '.* '/* 'C$2B'/=-* $D* EF* D--(<" #5" ;#5" -&&)"

5=::&5*&2"()"8&5?&1*"+3"A(9+)(#)"-6"E#=0"F(?#85C6"GHIIJK"LLM$"7)"#22(*(+)"*+"7):8(#)"#)2"A(9+)(#)<"

5('(0#8":&'()#*&5<"3+8'&2"2=&"*+"*;&"()30=&)1&"+3"0+):"'&2(#0"5600#-0&5<"#8&"9&86"1+''+)"#05+"()"

*;&" ,N8=" 2(#0&1*" +3" O5*+)(#)" G79#" HILIK" LJHM$" P()1&" :&'()#*(+)" 2+&5" )+*" 8&?8&5&)*" *;&" 5#'&"

?;&)+'&)+)"#5"@())(1"5*+?":8#2#*(+)"#)2" (5" *;=5" (88&0&9#)*" *+"*;&"2(51=55(+)"2&9&0+?&2";&8&<" 7"

B(00")+*"2&#0"B(*;"#0*&8)#*(9&"#11+=)*5"+3":&'()#*(+)"#)6"3=8*;&8$"

7)" *;&" #??8+#1;&5" *+" 1+)5+)#)*" :8#2#*(+)" 2(51=55&2" #-+9&<" 1;#):&" (5" 5&&)" #5" 8&5=0*():"

38+'"*;&"1+)*&>*"()"B;(1;"*;&"1+)5+)#)*5"#??&#8$"Q;&"#??8+#1;&5"*;#*"3+1=5"+)"0&)(*(+)"5&&"(*"#5"

8&5=0*():"38+'"&>1&55(9&"5600#-0&"B&(:;*$"R??8+#1;&5"3+1=5():"+)"3+8*(*(+)"8&:#82"+?&)"5600#-0&5"#5"

*8(::&85$"7)"*;&"3+00+B():"5&1*(+)5<"7"B(00"?8&5&)*"#)"#)#065(5"5;+B():"*;#*"1+)*&>*"'#6"3#1(0(*#*&"*;&"

-0+1C():"+3"5*+?"0&)(*(+)"B(*;+=*"#*"*;&"5#'&"*('&"-&1+'():"#"2(8&1*"1#=5&"+3"3+8*(*(+)$"

Figure 1. Intervocalic lengthening to balance the foot.

Figure 2. Monomoraic closed syllables.

The�need�for�mora�sharing�in�describing�phonological�systems�has�beenhighlighted�by�scholars�such�as� Ian�Maddieson�(1993),�Bruce�Hayes� (1995� :337—338),� Janet� C.� E.� Watson� (2007),� among� others.� The� need� for� mora�sharing�becomes�especially�pronounced�in�the�case�of�so-called�short�gemi-nates,� whose� appearance� does� not� result� in� the� preceding� syllable� beingregarded� as� heavy� (Sprouse� 1996; Broselow, Chen,� Huffman 1997). Theconcept of mora sharing has been applied by Patrik Bye (2005) in his descrip-tion of Saami languages, and by Arvo Eek, Einar Meister (1997 : 94) andMartin Ehala (2003 : 68) in their descriptions of the Estonian quantity system.

The� mora� is� not� a� purely� abstract� unit� of� structure.� Moraic� structure� isreflected� in� segment� duration.� Since� Ian� Maddieson� (1993),� several� meas-urements�have�confirmed�that�there�are�languages�where�a�mora�shared�bytwo� segments� affects� the� duration� of� both� segments.� The� shortening� of� avowel� in� a� closed� syllable� is� a� very� common� phenomenon� in� languages.Broselow,� Chen� and� Huffman� (1997)� ascribe� this� to� the� syllable� structuredescribed� in� Figure� 2b:� the� mora� shared� by� the� nucleus� and� coda� causesthe�vowel�in�the�nucleus�to�shorten�and�the�consonant�in�the�coda�to�lengthen.

For�example,�in�Levantine�Arabic,�the�coda�consonants�following�a�longvowel�share�a�mora�with�the�syllable�nucleus,�but�the�ones�following�a�shortvowel� do� not.� � This� has� also� been� confirmed� by� measurements� (Table� 1).Since�mora�sharing�affects� the�duration�of�segments,� long�vowels� in�closedsyllables� are� shorter� than� in� open� syllables,� whereas� the� duration� of� shortvowels� is� not� affected� by� the� syllable� being� either� closed� or� open.� At� thesame�time,�a�consonant�following�a�long�vowel�is�also�shorter�than�a�conso-nant� that� follows� a� short� vowel� and� that� is� linked� to� an� unshared� mora.The� effect� of� mora� sharing� on� the� duration� of� a� segment� was� also� evidentin� Hindi� and� Malayalam� (Broselow,� Chen,� Huffman� 1997� :� 55—60).

Table 1Dependence of V and C durations on moraic structure (average durations in

milliseconds, standard deviation in brackets) (Broselow, Chen, Huffman 1997 : 59)


167

!"

"

!"#$%&'&()*+#!#,-&%)+.#/*%&,#(*#&0*)'#12+)()*+#

!"3"#4)+.12#,(*5,#

#$"%&'()*"+,-&'./"0)1+)$23)1,)$2"4-+5--$"1,&'+"($0"6&$2"1+&71" )$"("1.66(46-"&$1-+" )1"&$6."

7&11)46-" )8" 5-" (66&5" +,-" 1.66(46-" &$1-+" +&" 4-" 6)$9-0" +&" ("%&'(:";&'()*" +,-&'." 0&-1" $&+" (66&5"

1.66(46-" &$1-+" +&" 4-" 6)$9-0" +&" )+1" &5$"%&'(/" 1)$*-" 1.66(46-" &$1-+1" 76(." $&" '&6-" )$" (11-11)$2" +,-"

1.66(46-"(1",-(<."&'"6)2,+/"):-:"0-+-'%)$)$2"+,-"6&*(+)&$"&8"1+'-11"=>.%($"?@!AB"?A!?C/">(.-1"?@@AB"

A?!ADE/" ($0" +,-" 6&11" &8" +,-" &$1-+" 0&-1" $&+" +')22-'" *&%7-$1(+&'." 6-$2+,-$)$2" =>(.-1" ?@!@E:"

>&5-<-'/"("1.66(46-"&$1-+"*($"1,('-"("%&'("5)+,"+,-"$3*6-31:""

F,-'-"('-"+,'--"*&%%&$"5(.1"+&"'-7'-1-$+"%&$&%&'()*"GHG"1.66(46-1B"=(E"+,-"*&0(")1"$&+"

6)$9-0" +&" ($."%&'(" =I)23'-" J(K" L-*" JMMNB" ?NAE/" =4E" +,-" *&0(" 1,('-1" ("%&'("5)+," +,-" 7'-*-0)$2"

<&5-6"=I)23'-"J4K">(.-1"?@!@B"JAO/"P(2-'"?@@@B"?ONE/"&'"=*E"+,-"1.66(46-"&$1-+")1"6)$9-0"+&"+,-"1(%-"

%&'("(1"+,-"1.66(46-"$3*6-31"=I)23'-"J*K">(.-1"?@@AB"AD/"0-"Q(*."JMMJB"?OE:"Q(''.";:">.%($"=?@!AB"

?NE"6)$91"(66"1-2%-$+1"+&"%&'(1:"R-1-('*,-'1"313(66."1-6-*+"&$-"&8"+,-1-"'-7'-1-$+(+)&$1"($0"31-")+"

-S*631)<-6.:"

(:" " !" " 4:" " !" " *:" " !" "

" " " " " " " " " " " "

" "

"" "

" "

"" "

" "

"" "

" G" H" G" " G" H" G" " G" H" G"

!"#$%&'()";&$&%&'()*"*6&1-0"1.66(46-1"

F,-" $--0" 8&'"%&'(" 1,(')$2" )$" 0-1*')4)$2" 7,&$&6&2)*(6" 1.1+-%1" ,(1" 4--$" ,)2,6)2,+-0" 4."

1*,&6('1"13*,"(1"#($";(00)-1&$"=?@@DE/"T'3*-">(.-1"=?@@AB"DDN!DD!E/"U($-+"G:"V:"W(+1&$"=JMMNE/"

(%&$2"&+,-'1:"F,-"$--0"8&'"%&'("1,(')$2"4-*&%-1"-17-*)(66."7'&$&3$*-0")$"+,-"*(1-"&8"1&X*(66-0"

1,&'+"2-%)$(+-1/"5,&1-"(77-('($*-"0&-1"$&+" '-136+" )$" +,-"7'-*-0)$2"1.66(46-"4-)$2"'-2('0-0"(1"

,-(<." =Y7'&31-" ?@@C/" T'&1-6&5"-+" (6:" ?@@NE:" F,-" *&$*-7+"&8"%&'(" 1,(')$2" ,(1"4--$" (776)-0"4."

Z(+')9"T.-"=JMMAE")$",)1"0-1*')7+)&$"&8"Y((%)"6($23(2-1/"($0"4."['<&"V-9/"V)$('";-)1+-'"=?@@NB"@OE"

($0";('+)$"V,(6("=JMMDB"C!E")$"+,-)'"0-1*')7+)&$1"&8"+,-"V1+&$)($"\3($+)+."1.1+-%:""

F,-" %&'(" )1" $&+" (" 73'-6." (41+'(*+" 3$)+" &8" 1+'3*+3'-:" ;&'()*" 1+'3*+3'-" )1" '-86-*+-0" )$"

1-2%-$+"03'(+)&$:"Y)$*-"#($";(00)-1&$"=?@@DE/"1-<-'(6"%-(13'-%-$+1",(<-"*&$8)'%-0"+,(+"+,-'-"

('-"6($23(2-1"5,-'-"("%&'("1,('-0"4."+5&"1-2%-$+1"(88-*+1"+,-"03'(+)&$"&8"4&+,"1-2%-$+1:"F,-"

VV�� dduurraattiioonn CC�� dduurraattiioonn SSaammppllee

1μ80.2� (9.0) ʔin.na.bi ’the� prophet’

1μ79.9� (6.6)

1μ88.4� (10.3) �i.nab.hum ’their� grape’

1,5μ131.6� (5.5)

0.5μ67.6� (9.4) ki.taab.hum ’their� book’

2μ161.0� (9.9) ki.taa.bi ’my� book’

!"

"

#$%&'()*)+"%,"-".%/(0"*)"-"10%#(2"#300-40("*#"-".(&3"1%55%)"6$()%5()%)"*)"0-)+7-+(#8"9&%#(0%/:"

;$()"-)2"<7,,5-)"=>!!?@"-#1&*4("'$*#"'%"'$("#300-40("#'&71'7&("2(#1&*4(2"*)"A*+7&("B4C"'$("5%&-"

#$-&(2"43"'$(")710(7#"-)2"1%2-"1-7#(#"'$(".%/(0"*)"'$(")710(7#"'%"#$%&'()"-)2"'$("1%)#%)-)'"*)"

'$("1%2-"'%"0()+'$()8""

A%&" (D-560(:" *)" E(.-)'*)(" F&-4*1:" '$(" 1%2-" 1%)#%)-)'#" ,%00%/*)+" -" 0%)+" .%/(0" #$-&(" -"

5%&-"/*'$"'$("#300-40(")710(7#:"47'"'$("%)(#",%00%/*)+"-"#$%&'".%/(0"2%")%'8""G$*#"$-#"4(()"-0#%"

1%),*&5(2"43"5(-#7&(5()'#"=G-40(">@8"H*)1("5%&-"#$-&*)+"-,,(1'#"'$("27&-'*%)"%,"#(+5()'#:"0%)+"

.%/(0#"*)"10%#(2"#300-40(#"-&("#$%&'(&"'$-)"*)"%6()"#300-40(#:"/$(&(-#"'$("27&-'*%)"%,"#$%&'".%/(0#"

*#" )%'" -,,(1'(2" 43" '$(" #300-40(" 4(*)+" (*'$(&" 10%#(2" %&" %6()8" F'" '$(" #-5(" '*5(:" -" 1%)#%)-)'"

,%00%/*)+" -" 0%)+" .%/(0" *#" -0#%" #$%&'(&" '$-)" -" 1%)#%)-)'" '$-'" ,%00%/#" -" #$%&'" .%/(0" -)2" '$-'" *#"

0*)I(2"'%"-)"7)#$-&(2"5%&-8"G$("(,,(1'"%,"5%&-"#$-&*)+"%)"'$("27&-'*%)"%,"-"#(+5()'"/-#"-0#%"

(.*2()'"*)"<*)2*"-)2"J-0-3-0-58"=9&%#(0%/"('"-08">!!?C"KK!LM@""

!"#$%&'("N(6()2()1("%,"O"-)2";"27&-'*%)#"%)"5%&-*1"#'&71'7&(""

=-.(&-+("27&-'*%)#"*)"5*00*#(1%)2#:"#'-)2-&2"2(.*-'*%)"*)"4&-1I('#@"=9&%#(0%/:"('"-08">!!?C"K!@"

""

O"

"

J-''$(/"S%&2%)"=BMMB@"2%(#")%'"#$-&("'$(".*(/"%,"9&%#(0%/"('"-08" =>!!?@"'$-'"6$%)('*1"

27&-'*%)" 2(6()2#" 2*&(1'03" %)" '$(" 5%&-*1*'3" %," '$(" #(+5()'8" <(" &(,(&#" '%" T$-0I$-" J%)+%0*-):"

/$(&(";O;"#300-40(#"-&(" 0*+$'" *)" '(&5#"%," #'&(##"!" *8(8:" '$(")710(7#"-)2" '$("1%2-"#$%702" #$-&("-"

5%&-:"3('"'$("5(-#7&(5()'#"*)2*1-'("'$-'"'$(".%/(0"*#"%,"'$("#-5("27&-'*%)"*)"4%'$";O"-)2";O;"

98::04:(9) ;<,+=,>) 6336?) @ABC)<,+=,>%9) /,0:) .9) &,) (9&04:.9') &'() *',>(&.D) =.EE(+(>D(9) ,E) :./'&) 0>=)

$(-.3";O;"#300-40(#:"/$*1$:"%,"1%7&#(:")((2")%'"1%)#*#'"%)03"*)"2*,,(&()'"&-'*%#"%,"'$("27&-'*%)"%,"

'$(")710(7#"-)2"'$("1%2-:"#*)1("*)"4%'$"0*+$'"-)2"$(-.3";O;"#300-40(#"'$(".%/(0"1-)"(*'$(&"#$-&("-"

!"

"

#$%&'()*)+"%,"-".%/(0"*)"-"10%#(2"#300-40("*#"-".(&3"1%55%)"6$()%5()%)"*)"0-)+7-+(#8"9&%#(0%/:"

;$()"-)2"<7,,5-)"=>!!?@"-#1&*4("'$*#"'%"'$("#300-40("#'&71'7&("2(#1&*4(2"*)"A*+7&("B4C"'$("5%&-"

#$-&(2"43"'$(")710(7#"-)2"1%2-"1-7#(#"'$(".%/(0"*)"'$(")710(7#"'%"#$%&'()"-)2"'$("1%)#%)-)'"*)"

'$("1%2-"'%"0()+'$()8""

A%&" (D-560(:" *)" E(.-)'*)(" F&-4*1:" '$(" 1%2-" 1%)#%)-)'#" ,%00%/*)+" -" 0%)+" .%/(0" #$-&(" -"

5%&-"/*'$"'$("#300-40(")710(7#:"47'"'$("%)(#",%00%/*)+"-"#$%&'".%/(0"2%")%'8""G$*#"$-#"4(()"-0#%"

1%),*&5(2"43"5(-#7&(5()'#"=G-40(">@8"H*)1("5%&-"#$-&*)+"-,,(1'#"'$("27&-'*%)"%,"#(+5()'#:"0%)+"

.%/(0#"*)"10%#(2"#300-40(#"-&("#$%&'(&"'$-)"*)"%6()"#300-40(#:"/$(&(-#"'$("27&-'*%)"%,"#$%&'".%/(0#"

*#" )%'" -,,(1'(2" 43" '$(" #300-40(" 4(*)+" (*'$(&" 10%#(2" %&" %6()8" F'" '$(" #-5(" '*5(:" -" 1%)#%)-)'"

,%00%/*)+" -" 0%)+" .%/(0" *#" -0#%" #$%&'(&" '$-)" -" 1%)#%)-)'" '$-'" ,%00%/#" -" #$%&'" .%/(0" -)2" '$-'" *#"

0*)I(2"'%"-)"7)#$-&(2"5%&-8"G$("(,,(1'"%,"5%&-"#$-&*)+"%)"'$("27&-'*%)"%,"-"#(+5()'"/-#"-0#%"

(.*2()'"*)"<*)2*"-)2"J-0-3-0-58"=9&%#(0%/"('"-08">!!?C"KK!LM@""

!"#$%&'("N(6()2()1("%,"O"-)2";"27&-'*%)#"%)"5%&-*1"#'&71'7&(""

=-.(&-+("27&-'*%)#"*)"5*00*#(1%)2#:"#'-)2-&2"2(.*-'*%)"*)"4&-1I('#@"=9&%#(0%/:"('"-08">!!?C"K!@"

""""

O";"

"

J-''$(/"S%&2%)"=BMMB@"2%(#")%'"#$-&("'$(".*(/"%,"9&%#(0%/"('"-08" =>!!?@"'$-'"6$%)('*1"

27&-'*%)" 2(6()2#" 2*&(1'03" %)" '$(" 5%&-*1*'3" %," '$(" #(+5()'8" <(" &(,(&#" '%" T$-0I$-" J%)+%0*-):"

/$(&(";O;"#300-40(#"-&(" 0*+$'" *)" '(&5#"%," #'&(##"!" *8(8:" '$(")710(7#"-)2" '$("1%2-"#$%702" #$-&("-"

5%&-:"3('"'$("5(-#7&(5()'#"*)2*1-'("'$-'"'$(".%/(0"*#"%,"'$("#-5("27&-'*%)"*)"4%'$";O"-)2";O;"

98::04:(9) ;<,+=,>) 6336?) @ABC)<,+=,>%9) /,0:) .9) &,) (9&04:.9') &'() *',>(&.D) =.EE(+(>D(9) ,E) :./'&) 0>=)

$(-.3";O;"#300-40(#:"/$*1$:"%,"1%7&#(:")((2")%'"1%)#*#'"%)03"*)"2*,,(&()'"&-'*%#"%,"'$("27&-'*%)"%,"

'$(")710(7#"-)2"'$("1%2-:"#*)1("*)"4%'$"0*+$'"-)2"$(-.3";O;"#300-40(#"'$(".%/(0"1-)"(*'$(&"#$-&("-"

!"

"

#$%&'()*)+"%,"-".%/(0"*)"-"10%#(2"#300-40("*#"-".(&3"1%55%)"6$()%5()%)"*)"0-)+7-+(#8"9&%#(0%/:"

;$()"-)2"<7,,5-)"=>!!?@"-#1&*4("'$*#"'%"'$("#300-40("#'&71'7&("2(#1&*4(2"*)"A*+7&("B4C"'$("5%&-"

#$-&(2"43"'$(")710(7#"-)2"1%2-"1-7#(#"'$(".%/(0"*)"'$(")710(7#"'%"#$%&'()"-)2"'$("1%)#%)-)'"*)"

'$("1%2-"'%"0()+'$()8""

A%&" (D-560(:" *)" E(.-)'*)(" F&-4*1:" '$(" 1%2-" 1%)#%)-)'#" ,%00%/*)+" -" 0%)+" .%/(0" #$-&(" -"

5%&-"/*'$"'$("#300-40(")710(7#:"47'"'$("%)(#",%00%/*)+"-"#$%&'".%/(0"2%")%'8""G$*#"$-#"4(()"-0#%"

1%),*&5(2"43"5(-#7&(5()'#"=G-40(">@8"H*)1("5%&-"#$-&*)+"-,,(1'#"'$("27&-'*%)"%,"#(+5()'#:"0%)+"

.%/(0#"*)"10%#(2"#300-40(#"-&("#$%&'(&"'$-)"*)"%6()"#300-40(#:"/$(&(-#"'$("27&-'*%)"%,"#$%&'".%/(0#"

*#" )%'" -,,(1'(2" 43" '$(" #300-40(" 4(*)+" (*'$(&" 10%#(2" %&" %6()8" F'" '$(" #-5(" '*5(:" -" 1%)#%)-)'"

,%00%/*)+" -" 0%)+" .%/(0" *#" -0#%" #$%&'(&" '$-)" -" 1%)#%)-)'" '$-'" ,%00%/#" -" #$%&'" .%/(0" -)2" '$-'" *#"

0*)I(2"'%"-)"7)#$-&(2"5%&-8"G$("(,,(1'"%,"5%&-"#$-&*)+"%)"'$("27&-'*%)"%,"-"#(+5()'"/-#"-0#%"

(.*2()'"*)"<*)2*"-)2"J-0-3-0-58"=9&%#(0%/"('"-08">!!?C"KK!LM@""

!"#$%&'("N(6()2()1("%,"O"-)2";"27&-'*%)#"%)"5%&-*1"#'&71'7&(""

=-.(&-+("27&-'*%)#"*)"5*00*#(1%)2#:"#'-)2-&2"2(.*-'*%)"*)"4&-1I('#@"=9&%#(0%/:"('"-08">!!?C"K!@"

""""

O"O";"

"

J-''$(/"S%&2%)"=BMMB@"2%(#")%'"#$-&("'$(".*(/"%,"9&%#(0%/"('"-08" =>!!?@"'$-'"6$%)('*1"

27&-'*%)" 2(6()2#" 2*&(1'03" %)" '$(" 5%&-*1*'3" %," '$(" #(+5()'8" <(" &(,(&#" '%" T$-0I$-" J%)+%0*-):"

/$(&(";O;"#300-40(#"-&(" 0*+$'" *)" '(&5#"%," #'&(##"!" *8(8:" '$(")710(7#"-)2" '$("1%2-"#$%702" #$-&("-"

5%&-:"3('"'$("5(-#7&(5()'#"*)2*1-'("'$-'"'$(".%/(0"*#"%,"'$("#-5("27&-'*%)"*)"4%'$";O"-)2";O;"

98::04:(9) ;<,+=,>) 6336?) @ABC)<,+=,>%9) /,0:) .9) &,) (9&04:.9') &'() *',>(&.D) =.EE(+(>D(9) ,E) :./'&) 0>=)

$(-.3";O;"#300-40(#:"/$*1$:"%,"1%7&#(:")((2")%'"1%)#*#'"%)03"*)"2*,,(&()'"&-'*%#"%,"'$("27&-'*%)"%,"

'$(")710(7#"-)2"'$("1%2-:"#*)1("*)"4%'$"0*+$'"-)2"$(-.3";O;"#300-40(#"'$(".%/(0"1-)"(*'$(&"#$-&("-"

!"

"

#$%&'()*)+"%,"-".%/(0"*)"-"10%#(2"#300-40("*#"-".(&3"1%55%)"6$()%5()%)"*)"0-)+7-+(#8"9&%#(0%/:"

;$()"-)2"<7,,5-)"=>!!?@"-#1&*4("'$*#"'%"'$("#300-40("#'&71'7&("2(#1&*4(2"*)"A*+7&("B4C"'$("5%&-"

#$-&(2"43"'$(")710(7#"-)2"1%2-"1-7#(#"'$(".%/(0"*)"'$(")710(7#"'%"#$%&'()"-)2"'$("1%)#%)-)'"*)"

'$("1%2-"'%"0()+'$()8""

A%&" (D-560(:" *)" E(.-)'*)(" F&-4*1:" '$(" 1%2-" 1%)#%)-)'#" ,%00%/*)+" -" 0%)+" .%/(0" #$-&(" -"

5%&-"/*'$"'$("#300-40(")710(7#:"47'"'$("%)(#",%00%/*)+"-"#$%&'".%/(0"2%")%'8""G$*#"$-#"4(()"-0#%"

1%),*&5(2"43"5(-#7&(5()'#"=G-40(">@8"H*)1("5%&-"#$-&*)+"-,,(1'#"'$("27&-'*%)"%,"#(+5()'#:"0%)+"

.%/(0#"*)"10%#(2"#300-40(#"-&("#$%&'(&"'$-)"*)"%6()"#300-40(#:"/$(&(-#"'$("27&-'*%)"%,"#$%&'".%/(0#"

*#" )%'" -,,(1'(2" 43" '$(" #300-40(" 4(*)+" (*'$(&" 10%#(2" %&" %6()8" F'" '$(" #-5(" '*5(:" -" 1%)#%)-)'"

,%00%/*)+" -" 0%)+" .%/(0" *#" -0#%" #$%&'(&" '$-)" -" 1%)#%)-)'" '$-'" ,%00%/#" -" #$%&'" .%/(0" -)2" '$-'" *#"

0*)I(2"'%"-)"7)#$-&(2"5%&-8"G$("(,,(1'"%,"5%&-"#$-&*)+"%)"'$("27&-'*%)"%,"-"#(+5()'"/-#"-0#%"

(.*2()'"*)"<*)2*"-)2"J-0-3-0-58"=9&%#(0%/"('"-08">!!?C"KK!LM@""

!"#$%&'("N(6()2()1("%,"O"-)2";"27&-'*%)#"%)"5%&-*1"#'&71'7&(""

=-.(&-+("27&-'*%)#"*)"5*00*#(1%)2#:"#'-)2-&2"2(.*-'*%)"*)"4&-1I('#@"=9&%#(0%/:"('"-08">!!?C"K!@"

""""

O"O"

"

J-''$(/"S%&2%)"=BMMB@"2%(#")%'"#$-&("'$(".*(/"%,"9&%#(0%/"('"-08" =>!!?@"'$-'"6$%)('*1"

27&-'*%)" 2(6()2#" 2*&(1'03" %)" '$(" 5%&-*1*'3" %," '$(" #(+5()'8" <(" &(,(&#" '%" T$-0I$-" J%)+%0*-):"

/$(&(";O;"#300-40(#"-&(" 0*+$'" *)" '(&5#"%," #'&(##"!" *8(8:" '$(")710(7#"-)2" '$("1%2-"#$%702" #$-&("-"

5%&-:"3('"'$("5(-#7&(5()'#"*)2*1-'("'$-'"'$(".%/(0"*#"%,"'$("#-5("27&-'*%)"*)"4%'$";O"-)2";O;"

98::04:(9) ;<,+=,>) 6336?) @ABC)<,+=,>%9) /,0:) .9) &,) (9&04:.9') &'() *',>(&.D) =.EE(+(>D(9) ,E) :./'&) 0>=)

$(-.3";O;"#300-40(#:"/$*1$:"%,"1%7&#(:")((2")%'"1%)#*#'"%)03"*)"2*,,(&()'"&-'*%#"%,"'$("27&-'*%)"%,"

'$(")710(7#"-)2"'$("1%2-:"#*)1("*)"4%'$"0*+$'"-)2"$(-.3";O;"#300-40(#"'$(".%/(0"1-)"(*'$(&"#$-&("-"

Matthew�Gordon�(2002)�does�not�share�the�view�of�Broselow,�Chen�andHuffman� (1997)� that� phonetic� duration� depends� directly� on� the� moraicityof� the� segment.� He� refers� to� Khalkha� Mongolian,� where� CVC� syllables� arelight� in� terms� of� stress� —� i.e.,� the� nucleus� and� the� coda� should� share� amora,� yet� the� measurements� indicate� that� the� vowel� is� of� the� same� dura-tion� in�both�CV�and�CVC�syllables� (Gordon�2002� :�69).�Gordon’s�goal� is� toestablish�the�phonetic�differences�of� light�and�heavy�CVC�syllables,�which,of� course,� need� not� consist� only� in� different� ratios� of� the� duration� of� thenucleus�and�the�coda,�since�in�both�light�and�heavy�CVC�syllables�the�vowelcan� either� share� a� mora� or� not.� Khalkha� Mongolian� may� have� the� syllablestructure� described� in� Figure� 2a.� Thus,� Gordon’s� study� does� not� refute� theclaims� of� Broselow,� Chen� and� Huffman (1997), but complements them.

If�we�are�to�expand�on�the�ideas�of�Broselow,�Chen�and�Huffman�(1997),assuming� that� mora� sharing� also� influences� the� duration� of� the� onset,� yetnot�disputing�the�principle�that�the�onset�cannot�be�linked�to�its�own�mora,short� CVC� syllables� can� have� many� possible� realisations:� each� structuredepicted�in�Figure�2�can�exist�in�a�set�of�languages.�Hence,�there�is�no�needto�attempt�to�determine�which�of�these�is�universally�right.�What�the�differ-ence� of� CVC� syllables� from� CV� syllables� implies� for� stop� gradation� is� thatboth� the�coda�and� the�onset� ”crave”� to�be� linked� to�a�mora,�yet� the�syllablehas� to� stay� monomoraic� and� priority� is� given� to� the� coda� (Figure� 3b).� Thesharing�of�a�mora�between�three�or�more�segments�would�shorten�the�nucleusvowel� too� extensively� and� supposedly� was� not� an� acceptable� alternative.

The� question� arises� whether� Finnic� languages� are� the� only� languageswhose�description�requires�the�syllable�onset�to�be�linked�to�a�shared�mora.If�so,� the�reliability�of�this�analysis�would�be�significantly�reduced.�Unfor-tunately,�to�my�best�knowledge�currently�no�such�studies�exist,�since�moraictheory� does� not� take� an� interest� in� syllable� onsets.� It� is� conceivable,however,� that� shared� onset� moras� could� be� pertinent� in� descriptions� oflanguages� where� word-initial� geminates� occur.

The� effect� of� the� syllable� onset� on� the� duration� of� the� following� vowelhas� received� little� treatment� in� phonetic� literature.� The� relations� betweenthe�sonority�of�initial�stops�and�the�duration�of�following�vowels�have�beenthe�best�studied.�Typically,�the�vowel�which�follows�a�voiced�stop�is�longerthan�the�vowel�which�follows�a�voiceless�stop�(Allen,�Miller�1999;�see�alsothe�references�provided�by�them).�The�effect�of�other�consonants�has�beenstudied,�for�example,�in�Japanese,�where�sonority�is�also�an�important�factor:the�vowel�which�follows�a�voiced�consonant�is�longer�than�the�vowel�whichfollows�a�voiceless�consonant�(Vendetti,�van�Santen�1998�:�2045).�In�English,the� vowels� which� follow� certain� consonant� clusters� are� shorter� than� thevowels� which� follow� single� consonants� (van� Santen� 1992).

Moras� are� preferably� linked� to� voiced� segments� (Morén� 2003� :� 289).� Inthe� syllable� onset,� the� relation� between� a� mora� and� sonority� seems� to� be

Külli Prillop

168

!"#

#

$%&'# %&# (%)*# +,'-.,'#/%(0%-1'(#$'2# ,'34# ),4# 52--'6-4# 5)&78)7&4# 9458&1649# 1(# :107&4# ;'*# <,75=#

!"#$"%&'(')*$+($",'(%")(#,-*)4#),4#8-'1$5#%>#?&%54-%@#4)#'-*#A!BBCD=#67)#8%$E-4$4()5#),4$*##

F>#@4#'&4#)%#4GE'(9#%(#),4# 194'5#%>#?&%54-%@#4)#'-*# A!BBCD=#'557$1(0#),')#$%&'#5,'&1(0#

'-5%#1(>-74(845#),4#97&')1%(#%>#),4#%(54)=#24)#(%)#915E7)1(0#),4#E&1(81E-4#),')#),4#%(54)#8'((%)#64#

-1(.49# )%# 1)5#%@(#$%&'=# 5,%&)#HIH#52--'6-45# 8'(#,'34#$'(2#E%5516-4# &4'-15')1%(5J#4'8,#5)&78)7&4#

94E18)49# %(# :107&4# ;# 8'(# 4G15)# 1(# '# 54)# %># -'(07'045*# K4(84=# ),4&4# 15# (%# (449# )%# '))4$E)# )%#

94)4&$1(4# @,18,# %># ),454# 15# 7(134&5'--2# &10,)*# L,')# ),4# 91>>4&4(84# %># HIH# 52--'6-45# >&%$# HI#

'+../0.,'(123.1,'(-"#(')"3(4#/$/)1"%(1'()5/)(0")5()5,(6"$/(/%$()5,("%',)(76#/8,9()"(0,(.1%:,$()"(/(

$%&'=#24)#),4#52--'6-4#,'5#)%#5)'2#$%(%$%&'18#'(9#E&1%&1)2# 15#0134(#)%#),4#8%9'#A:107&4#M6D*#<,4#

5,'&1(0# %># '# $%&'# 64)@44(# ),&44# %&# $%&4# 540$4()5# @%7-9# 5,%&)4(# ),4# (78-475# 3%@4-# )%%#

4G)4(5134-2#'(9#57EE%549-2#@'5#(%)#'(#'884E)'6-4#'-)4&(')134*#

'*#N%(0#%(54)##

A5)&%(0#0&'94D#

# 6*#O,%&)#%(54)##

A@4'.#0&'94D#

# # !# # # # # !# #

# # ;# # # # # ;# #

# +# I# # # # +# I# ##H#

!"#$%&'()#O)%E#0&'9')1%(#

<,4#P745)1%(#'&1545#@,4),4&#:1((18# -'(07'045#'&4# ),4#%(-2# -'(07'045#@,%54#9458&1E)1%(#

&4P71&45#),4#52--'6-4#%(54)#)%#64#-1(.49#)%#'#5,'&49#$%&'*#F>#5%=#),4#&4-1'61-1)2#%>#),15#'('-2515#@%7-9#

64# 510(1>18'()-2# &497849*# Q(>%&)7(')4-2=# )%#$2# 645)# .(%@-4904# 87&&4()-2# (%# 578,# 5)79145# 4G15)=#

51(84#$%&'18#),4%&2#9%45#(%)#)'.4#'(# 1()4&45)# 1(#52--'6-4#%(54)5*# F)# 15#8%(8413'6-4=#,%@434&=#),')#

5,'&49#%(54)#$%&'5#8%7-9#64#E4&)1(4()#1(#9458&1E)1%(5#%>#-'(07'045#@,4&4#@%&9R1(1)1'-#04$1(')45#

%887&*#

<,4#4>>48)#%>#),4#52--'6-4#%(54)#%(#),4#97&')1%(#%>#),4#>%--%@1(0#3%@4-#,'5#&4841349#-1))-4#

)&4')$4()# 1(# E,%(4)185# -1)4&')7&4*# <,4# &4-')1%(5# 64)@44(# ),4# 5%(%&1)2# %># 1(1)1'-# 5)%E5# '(9# ),4#

97&')1%(# %># >%--%@1(0# 3%@4-5# ,'34# 644(# ),4# 645)# 5)79149*# <2E18'--2=# ),4# 3%@4-#@,18,# >%--%@5# '#

3%1849#5)%E# 15# -%(04&#),'(#),4#3%@4-#@,18,#>%--%@5#'#3%184-455#5)%E#AS--4(=#/1--4&#!BBBT#544#'-5%#

),4#&4>4&4(845#E&%31949#62#),4$D*#<,4#4>>48)#%>#%),4&#8%(5%('()5#,'5#644(#5)79149=#>%&#4G'$E-4=#

1(# U'E'(454=# @,4&4# 5%(%&1)2# 15# '-5%# '(# 1$E%&)'()# >'8)%&J# ),4# 3%@4-# @,18,# >%--%@5# '# 3%1849#

Figure 3. Stop gradation.

reversed:�it�is�rather�voiceless�consonants�than�voiced�consonants�which�sharea� mora� with� the� following� vowel.� According� to� Nina� Topintzi,� who� isconvinced� that� syllable� onsets� should� be� allowed� to� be� linked� to� their� ownmoras,�a�voiceless�stop�is�indeed�a�typical�moraic�onset�(Topintzi�2010�:�15).The moraicity of voiceless stops in the syllable onset may be induced by

the need to distinguish them from respective voiced stops. Short stops in asonorous environment have a tendency to become voiced (LaVoie 2001 : 30—32; Smith 2008 : 524 and others), long stops are preferably voiceless (Hayes,Steriade 2004 : 7—8). If the phonetic similarity of positional allophones isimportant in a language, the speakers may try to lengthen the stops insonorous environments, so as to avoid their becoming perceptibly voiced andthus significantly different from the same stops in non-sonorous environments.The contextually voiced part of the stop becomes relatively short as the stopitself lengthens. According to another explanation, the linking of the syllableonset to the mora of the following vowel would be mandatory. In the caseof voiceless stops, the resulting difference in duration would be perceptible,while in the case of voiced sounds this would not be the case necessarily.Several phonological systems avoid sonorous geminates, since these do notdiffer sufficiently from single sounds (Kawahara,� Pangilinan,� Garvey� 2011).

The� lengthening� of� an� intervocalic� stop� is� in� most� cases� caused� by� therequirement� that�a�stressed�syllable�must�be�bimoraic.�However,� there�arelanguages�where�gemination�indeed�occurs�to�preserve�the�sonority�contrast.For�example,�in�Malayalam�English,�voiceless�stops�between�vowels�are�alwaysgeminated� (packet [pæækket]),� voiced� stops� are� not� (baboon [bææbuun])(Mohanan,� Mohanan� 2003� :� 16).

The� influence� of� the� syllable� coda� on� the� syllable� onset� is� a� rare� (or� atleast� little� studied)� phenomenon,� yet� still� not� entirely� unheard� of.� In� BritishEnglish,�measurements�have�shown�that�the�onset�l is�longer�in�syllables�witha� voiced� coda� than� in� those� with� a� voiceless� coda� (Hawkins,� Nguyen� 2004).Mora�sharing�would�provide�an�explanation�to�the�relationship�between�codaand�onset� in�a� language�such�as�Huichol,�where�a�complex�onset� is�possibleonly� in� an� open� syllable.� Huichol� permits� syllable� structures� such� as� CVCand� CCV,� but� not� CCVC,� a� closed� syllable� with� a� complex� onset� (McIntosh1945�:�32).�Each�sound�in�a�syllable�in�Huichol�must�be�linked�to�a�mora�butthe�sharing�of�coda�mora�is�not�allowed.�Systematic�links�between�onsets�andcodas� play� a� role� in� word� recognition� (Hawkins,� Nguyen� 2004� :� 202).

The� likely� original� cause� of� gradation� is� the� attempt� to� increase� theduration� of� stops� to� avoid� their� becoming� voiced� in� a� sonorous� environ-ment.� Giving� longer� duration� to� stops� was� impossible� in� closed� syllables,where� the� nucleus� vowel� had� to� share� a� mora� with� the� coda.� At� the� sametime,� pressure� to� preserve� the� voicelessness� of� stops� was� insufficient� tocause� mora� insertion� in� unstressed� syllables.� The� syllables� which� wereunderlyingly� light� had� to� remain� monomoraic� also� on� the� surface� level.

2.2. Geminates

If� we� can� interpret� gradation� as� stop� lengthening� in� order� to� keep� stopsunvoiced,� we� would� have� to� find� a� different� reason� for� the� gradation� ofgeminates,� since� geminates� are� long� enough� not� to� become� perceptiblyvoiced.� Geminates� are� also� preferably� voiceless� between� vowels� (Kirchner


169

2004� :�326)�and�there� is�no�need�to� link� them�to�additional�moras� in�orderto� avoid� their� weakening.In moraic theory, a geminate is defined as a consonant underlyingly

linked to a mora. In the surface representation the geminate is syllabifiedin a configuration consisting of a moraic-coda and a non-moraic onset (seeFigure 4). If a moraic consonant belonged entirely to the preceding syllable,the following syllable would lack an onset (yet, languages prefer syllableswith onset over onsetless ones). If a moraic consonant belonged entirelyto the following syllable, the onset of that syllable would have its ownmora, which is prohibited in moraic theory (Hayes 1989 : 257).

A� moraic� geminate� (as� in� Figure� 4)� will� cause� the� preceding� syllable� tobe�heavy�in�any�case.�However,�it�is�known�that�in�Selkup�and�in��Malayalamall�closed�syllables,�even�those�closed�by�the�first�part�of�a�geminate,�are�light.In�Malayalam,�syllable�weight� is�determined�entirely�by�vowel� length.�Sylla-bles�with�long�vowels�attract�stress,�but�syllables�closed�by�geminates�do�not(e.g.�pÉa.ra.ti ’searched’,�pa.rÉaa.ti ’complaint’,�pÉa.rat.ti ’scattered’).�Thus,�gemi-nates�in�Malayalam�should�be�treated�as�non-moraic�(Tranel�1991).�This�fact,as�well�as�the�existence�of�syllable-initial�geminates,�has�forced�the�researchersto� accept� that� there� are� other� possible� realizations� of� geminates� in� additionto� the� one� depicted� in� Figure� 4.

Just�as�a�single�consonant,�a�surface�geminate�may�also�acquire�part�ofits�length�from�sharing�a�mora�with�neighbouring�vowels.�Broselow,�Chenand� Huffman (1997� :� 68—70)� demonstrate� that� geminates� with� a� structuredepicted� in�Figure�5a�solve� the�problem�for�such� languages�as�Malayalamwhere� syllables� ending� with� a� geminate� are� light.

The�scheme�depicted�in�Figure�5b�would�not�be�possible�in�a�light�closedsyllable�if�moras�cannot�be�shared�between�three�sounds.�Thus,�the�grada-tion� of� geminates� may� be� triggered� by� the� prohibition� of� linking� a� voice-less� sound� to� an� unshared� mora.� In� an� attempt� to� preserve� the� differencein� the� duration� of� geminates� and� single� stops,� the� moras� of� neighbouringvowels� will� be� linked� to� the� geminate� stop� if� possible.

If� it� is� prohibited� in� a� language� to� link� voiceless� stops� to� unsharedmoras,� then� we� may� doubt� whether� the� structure� depicted� in� Figure� 5b� isa�better�representation�than�the�ordinary�geminate�in�Figure�4.�If�two�half-moras� add� up� to� one� full� mora,� then,� in� either� case,� the� geminate� will� bemonomoraic�in�the�end.�Nevertheless,�in�the�case�of�an�ordinary�geminate,

Külli Prillop

170

!"#

#

# !# # !# #

# !# !# !# #

# $# ##%# $# #

!"#$%&'()#&#'()*+,-.(#/*0/*'*1-,-.21#23#4*+.1,-*'#

&#+2/,.(# 4*+.1,-*# 5,'# .1#6.47/*#89#:.;;# (,7'*# -)*#0/*(*<.14# '=;;,>;*# -2#>*#)*,?=# .1# ,1=#

(,'*@# A2:*?*/B# .-# .'# C12:1# -),-# .1# D*;C70# ,1<# .1#E,;,=,;,+# 5,# F/,?.<.,1# ;,147,4*9# ,;;# (;2'*<#

'=;;,>;*'B#*?*1#-)2'*#(;2'*<#>=#-)*#3./'-#0,/-#23#,#4*+.1,-*B#,/*#;.4)-@#G1#E,;,=,;,+B#'=;;,>;*#:*.4)-#

.'# <*-*/+.1*<# *1-./*;=# >=# ?2:*;# ;*14-)@# D=;;,>;*'# :.-)# ;214# ?2:*;'# ,--/,(-# '-/*''B# >7-# '=;;,>;*'#

(;2'*<#>=#4*+.1,-*'#<2#12-# 5*@4@B#!"!"#!$%# "#$%&'($)*+#&##$"#!$%# "',-./%012*+3!"!"#$!$%# "#'%22$&$)*453

H)7'B#4*+.1,-*'#.1#E,;,=,;,+#')27;<#>*#-/*,-*<#,'#121I+2/,.(@#5H/,1*;#!JJ!9#H).'#3,(-B#,'#:*;;#,'#

-)*# *K.'-*1(*# 23# '=;;,>;*I.1.-.,;# 4*+.1,-*'B# ),'# 32/(*<# -)*# /*'*,/()*/'# -2# ,((*0-# -),-# -)*/*# ,/*#

2-)*/#02''.>;*#/*,;.L,-.21'#23#4*+.1,-*'#.1#,<<.-.21#-2#-)*#21*#<*0.(-*<#.1#6.47/*#8@#

M7'-# ,'# ,# '.14;*# (21'21,1-B# ,# '7/3,(*#4*+.1,-*#+,=#,;'2#,(N7./*#0,/-#23# .-'# ;*14-)# 3/2+#

'),/.14# ,# +2/,# :.-)# 1*.4)>27/.14# ?2:*;'@# O/2'*;2:# *-# ,;@# 5!JJPQ# RS6PT9# <*+21'-/,-*# -),-#

4*+.1,-*'# :.-)# ,# '-/7(-7/*# <*0.(-*<# .1# 6.47/*# U,# '2;?*# -)*# 0/2>;*+# 32/# '7()# ;,147,4*'# ,'#

E,;,=,;,+#:)*/*#'=;;,>;*'#*1<.14#:.-)#,#4*+.1,-*#,/*#;.4)-@##

!"#$%&'*)'W/,<,-.21#23#4*+.1,-*'#

H)*#'()*+*#<*0.(-*<#.1#6.47/*#U>#:27;<#12-#>*#02''.>;*#.1#,#;.4)-#(;2'*<#'=;;,>;*#.3#+2/,'#

(,112-#>*#'),/*<#>*-:**1#-)/**#'271<'@#H)7'B#-)*#4/,<,-.21#23#4*+.1,-*'#+,=#>*#-/.44*/*<#>=#

-)*#0/2).>.-.21#23# ;.1C.14#,#?2.(*;*''#'271<#-2#,1#71'),/*<#+2/,@# G1#,1#,--*+0-#-2#0/*'*/?*#-)*#

<.33*/*1(*#.1#-)*#<7/,-.21#23#4*+.1,-*'#,1<#'.14;*#'-20'B#-)*#+2/,'#23#1*.4)>27/.14#?2:*;'#:.;;#>*#

;.1C*<#-2#-)*#4*+.1,-*#'-20#.3#02''.>;*@##

G3# .-# .'#0/2).>.-*<# .1#,# ;,147,4*# -2# ;.1C#?2.(*;*''# '-20'# -2#71'),/*<#+2/,'B# -)*1#:*#+,=#

<27>-#:)*-)*/# -)*# '-/7(-7/*#<*0.(-*<# .1#6.47/*#U># .'#,#>*--*/# /*0/*'*1-,-.21# -),1# -)*#2/<.1,/=#

!"#

#

!"#$%&'()#&#'()*+,-.(#/*0/*'*1-,-.21#23#4*+.1,-*'#

&#+2/,.(# 4*+.1,-*# 5,'# .1#6.47/*#89#:.;;# (,7'*# -)*#0/*(*<.14# '=;;,>;*# -2#>*#)*,?=# .1# ,1=#

(,'*@# A2:*?*/B# .-# .'# C12:1# -),-# .1# D*;C70# ,1<# .1#E,;,=,;,+# 5,# F/,?.<.,1# ;,147,4*9# ,;;# (;2'*<#

'=;;,>;*'B#*?*1#-)2'*#(;2'*<#>=#-)*#3./'-#0,/-#23#,#4*+.1,-*B#,/*#;.4)-@#G1#E,;,=,;,+B#'=;;,>;*#:*.4)-#

.'# <*-*/+.1*<# *1-./*;=# >=# ?2:*;# ;*14-)@# D=;;,>;*'# :.-)# ;214# ?2:*;'# ,--/,(-# '-/*''B# >7-# '=;;,>;*'#

(;2'*<#>=#4*+.1,-*'#<2#12-# 5*@4@B#!"!"#!$%# "#$%&'($)*+#&##$"#!$%# "',-./%012*+3!"!"#$!$%# "#'%22$&$)*453

H)7'B#4*+.1,-*'#.1#E,;,=,;,+#')27;<#>*#-/*,-*<#,'#121I+2/,.(@#5H/,1*;#!JJ!9#H).'#3,(-B#,'#:*;;#,'#

-)*# *K.'-*1(*# 23# '=;;,>;*I.1.-.,;# 4*+.1,-*'B# ),'# 32/(*<# -)*# /*'*,/()*/'# -2# ,((*0-# -),-# -)*/*# ,/*#

2-)*/#02''.>;*#/*,;.L,-.21'#23#4*+.1,-*'#.1#,<<.-.21#-2#-)*#21*#<*0.(-*<#.1#6.47/*#8@#

M7'-# ,'# ,# '.14;*# (21'21,1-B# ,# '7/3,(*#4*+.1,-*#+,=#,;'2#,(N7./*#0,/-#23# .-'# ;*14-)# 3/2+#

'),/.14# ,# +2/,# :.-)# 1*.4)>27/.14# ?2:*;'@# O/2'*;2:# *-# ,;@# 5!JJPQ# RS6PT9# <*+21'-/,-*# -),-#

4*+.1,-*'# :.-)# ,# '-/7(-7/*# <*0.(-*<# .1# 6.47/*# U,# '2;?*# -)*# 0/2>;*+# 32/# '7()# ;,147,4*'# ,'#

E,;,=,;,+#:)*/*#'=;;,>;*'#*1<.14#:.-)#,#4*+.1,-*#,/*#;.4)-@##

,@#D)2/-#4*+.1,-*# >@#V214#4*+.1,-*#

# !# # !# # # !# # !# #

# !# # !# # # !# # !# #

# $# %# $# # # $# ##%# $# #

!"#$%&'*)'W/,<,-.21#23#4*+.1,-*'#

H)*#'()*+*#<*0.(-*<#.1#6.47/*#U>#:27;<#12-#>*#02''.>;*#.1#,#;.4)-#(;2'*<#'=;;,>;*#.3#+2/,'#

(,112-#>*#'),/*<#>*-:**1#-)/**#'271<'@#H)7'B#-)*#4/,<,-.21#23#4*+.1,-*'#+,=#>*#-/.44*/*<#>=#

-)*#0/2).>.-.21#23# ;.1C.14#,#?2.(*;*''#'271<#-2#,1#71'),/*<#+2/,@# G1#,1#,--*+0-#-2#0/*'*/?*#-)*#

<.33*/*1(*#.1#-)*#<7/,-.21#23#4*+.1,-*'#,1<#'.14;*#'-20'B#-)*#+2/,'#23#1*.4)>27/.14#?2:*;'#:.;;#>*#

;.1C*<#-2#-)*#4*+.1,-*#'-20#.3#02''.>;*@##

G3# .-# .'#0/2).>.-*<# .1#,# ;,147,4*# -2# ;.1C#?2.(*;*''# '-20'# -2#71'),/*<#+2/,'B# -)*1#:*#+,=#

<27>-#:)*-)*/# -)*# '-/7(-7/*#<*0.(-*<# .1#6.47/*#U># .'#,#>*--*/# /*0/*'*1-,-.21# -),1# -)*#2/<.1,/=#

Figure 4. A schematic representation of geminates.

Figure 5. Gradation of geminates.

the�entire�weight�is�on�one�syllable,�whereas�a�geminate�with�shared�morasdistributes� its� weight� over� two� syllables.If we allow syllable onset to be linked to the mora of the following

vowel, we will have another interesting possibility of surface-level real-ization of underlying moraic consonants. In closed syllables, if the codaand the nucleus are linked to the same mora, the nucleus vowel, in orderto preserve its length, may share another mora with the onset (see Figure6). Such an option is not possible in open syllables, if changes in vowelduration are prohibited: the vowel in a monomoraic CV syllable with ashared mora shortens, and the vowel in a bimoraic CV syllable lengthens.

Thus,� we� have� one� more� option� to� describe� the� gradation� of� geminates:ordinary�geminates�(as�seen�in�Figure�4)�may�alternate�with�lengthened�singlestops� (as�seen� in�Figure�6).� Indeed,� in�most�Finnic� languages� the�geminatesalternate� with� single� stops� (e.g.� Finnish� and� Karelian� tuk.ka :� tu.kan ’hair,nom.sg/gen.sg.’).�The�main�cause�for�the�emergence�of�consonant�gradationis�still�the�same�as�in�the�first�case�—�the�closed�syllable.�Patrik�Bye�has�usedstructures� similar� to� the� one� depicted� in� Figure� 6� (but� without� a� moraiconset)� to� describe� the� overlength� in� Saami� languages� (e.g.� Bye� 2001� :� 165).

3. An Optimality Theory analysis of gradation

Optimality� Theory� (below,� OT;� Prince,� Smolensky� 1993/2004;� McCarthy,Prince�1993;�1995)�postulates�that�any�surface�form�is�the�result�of�an�inter-action� between� universal� constraints� that� are� violable� and� that� are� rankedvis-à-vis� one� another� in� a� manner� particular� to� the� language� in� question.The�generation�of�surface�forms�in�OT�involves�two�functions,�gen and�Eval.gen takes�an�input�(underlying�form)�and�generates�a�set�of�output�(surface)candidates.� Eval chooses� the� candidate� that� best� satisfies� the� set� of� rankedconstraints.�This�optimal�candidate�becomes�the�output�(surface�form).�Theoperation� of� Eval can� be� viewed� as� choosing� the� subset� of� candidates� thatbest�satisfies�the�top-ranked�constraint,�and,�of�this�subset,�selecting�the�sub-subset� that� best� satisfies� the� second-ranked� constraint,� and� so� on.

There�are�two�types�of�constraints.�The�function�of�markedness�constraintsis�to�prohibit�the�marked,�i.e.�complex,�hard-to-pronounce�or�hard-to-perceivesurface� structures.� Faithfulness� constraints� ensure� similarity� between� inputand� output� forms.

Table�2� illustrates�output�selection� in�OT.�The� first�column�of� the� tablelists� a� series� of� output� candidates� in� random� order.� Constraints� are� listedhorizontally,� in� a� descending� ranking� from� left� to� right.� Violations� ofconstraints� are� marked� by� asterisks� in� table� cells.� If� a� violation� eliminatesthe� candidate,� the� asterisk� is� followed� by� an� exclamation� point� and� thefollowing� cells� are� shaded.� The� optimal� candidate� is� indicated� by� �.

The� voicing� of� stops� in� a� sonorous� environment� depends� on� the� orderof�the�constraints�*D�and�*{V,N}TV.�Both�constraints�are�required�in�widely�


171

!"#

#

$%&'()*%#'(#+'$,-%#".#/0#*12#3)405&2-)6#)77#,8#*2#2(%#0,44#&2-)9#*3%(9#'(#%'*3%-#:)6%9#*3%#$%&'()*%#

1'44# ;%#&2(2&2-)':# '(# *3%# %(7.# <%=%-*3%4%669# '(# *3%# :)6%# 20# )(# 2-7'()-># $%&'()*%9# *3%# %(*'-%#

1%'$3*#'6#2(#2(%#6>44);4%9#13%-%)6#)#$%&'()*%#1'*3#63)-%7#&2-)6#7'6*-';,*%6#'*6#1%'$3*#2=%-#*12#

6>44);4%6.#

/0#1%#)4421#6>44);4%#2(6%*#*2#;%# 4'(?%7#*2# *3%#&2-)#20# *3%# 024421'($#=21%49#1%#1'44#3)=%#

)(2*3%-# '(*%-%6*'($# 8266';'4'*># 20# 6,-0):%54%=%4# -%)4'@)*'2(# 20# ,(7%-4>'($# &2-)':# :2(62()(*6.# /(#

:426%7#6>44);4%69# '0#*3%#:27)#)(7#*3%#(,:4%,6#)-%# 4'(?%7#*2#*3%#6)&%#&2-)9#*3%#(,:4%,6#=21%49# '(#

2-7%-# *2# 8-%6%-=%# '*6# 4%($*39# &)># 63)-%# )(2*3%-#&2-)# 1'*3# *3%# 2(6%*# A6%%# +'$,-%# BC.# D,:3# )(#

28*'2(#'6#(2*#8266';4%#'(#28%(#6>44);4%69#'0#:3)($%6#'(#=21%4#7,-)*'2(#)-%#8-23';'*%7E#*3%#=21%4#'(#

&2(2&2-)':# FG# 6>44);4%# 1'*3# )# 63)-%7# &2-)# 632-*%(69# )(7# *3%# =21%4# '(# ;'&2-)':# FG# 6>44);4%#

4%($*3%(6.#!

#!#

#!#

#

# !# !# !# #

# G# F# G# F#

"#$%&'!()#H2-)58-%6%-='($#7%$%&'()*'2(#

I3,69# 1%# 3)=%# 2(%# &2-%# 28*'2(# *2# 7%6:-';%# *3%# $-)7)*'2(# 20# $%&'()*%6E# 2-7'()->#

$%&'()*%6#A)6#6%%(#'(#+'$,-%#"C#&)>#)4*%-()*%#1'*3#4%($*3%(%7#6'($4%#6*286#A)6#6%%(#'(#+'$,-%#BC.#

/(7%%79# '(# &26*# +'((':# 4)($,)$%6# *3%# $%&'()*%6# )4*%-()*%# 1'*3# 6'($4%# 6*286# A%.$.# +'(('63# )(7#

J)-%4')(#!"#$#%&E#!"$#%'&&"#$%&9#(2&.#6$K$%(.#6$."'()*#+),$%-).$/0+)12&)3#+)+,+&4+-.+)21).2-02-$-3)

$-)7)*'2(#'6#6*'44#*3%#6)&%#)6#'(#*3%#0'-6*#:)6%#5#*3%#:426%7#6>44);4%.#L)*-'?#M>%#3)6#,6%7#6*-,:*,-%6#

6'&'4)-#*2#*3%#2(%#7%8':*%7#'(#+'$,-%#B#A;,*#1'*32,*#)#&2-)':#2(6%*C#*2#7%6:-';%#*3%#2=%-4%($*3#'(#

D))&'#4)($,)$%6#A%.$.#M>%#NOO!E#!BPC.#

!"#$%#&'()*+,)(-#./012-#+%+,-3)3#14#52+6+()1%#

Q8*'&)4'*># I3%2-># A;%4219# QIR# L-'(:%9# D&24%(6?># !SSTKNOO"9# H:F)-*3>9# L-'(:%# !SST9#

!SSPC# 826*,4)*%6# *3)*# )(># 6,-0):%# 02-&# '6# *3%# -%6,4*# 20# )(# '(*%-):*'2(# ;%*1%%(# ,('=%-6)4#

:2(6*-)'(*6#*3)*#)-%#='24);4%#)(7#*3)*#)-%#-)(?%7#='65U5='6#2(%#)(2*3%-# '(#)#&)((%-#8)-*':,4)-#*2#

*3%#4)($,)$%#'(#V,%6*'2(.#I3%#$%(%-)*'2(#20#6,-0):%#02-&6#'(#QI#'(=24=%6#*12#0,(:*'2(69#()'#)(7#

*+%,.#()'# *)?%6# )(# '(8,*# A,(7%-4>'($# 02-&C# )(7# $%(%-)*%6# )# 6%*#20# 2,*8,*# A6,-0):%C# :)(7'7)*%6.#

Figure 6. Mora-preserving degemination.

different� languages,� see� for� instance,� the� typologies� in� Wetzels,� Mascaró2001;� Kirchner� 2004;� Smith� 2008.� If� constraint� *D� outranks� *{V,N}TV,� allintervocalic� stops�will� be�voiceless,�while� in� the�case�of� the� predominanceof�constraint�*{V,N}TV,�such�stops�will�be�voiced.�This,�however,�will�onlybe� true� if� the� lengthening� of� the� onset,� i.e.� its� link-up� with� the� mora,� isprohibited�(that�is,�if�constraints�*SHARED-μ�and�*[son]/μ�are�present�at�thetop� of� the� hierarchy).� Table� 3� describes� a� language� in� which� stops� in� asonorous� environment� are� voiceless.� If� the� lowest-placed� constraint� were*{V,N}TV,� the� stops� would� also� be� voiceless,� although� depending� on� therate�of�speech� listeners�might�perceive� them�as�voiced.�Should�the� lowest-level� constraint� be� *[-son]/μ,� intervocalic� stops� would� geminate.

(2) a. *{V,N}TVNo� nonmoraic� stops� in� the� context� V_V� or� N_V(see� Kirchner� 2004� :� 329—331)

b. *DNo� voiced� stops� (see� Kager� 1999� :� 40)

c. *SHARED-μMoras�are� linked� to�single�segments(Broselow,�Chen,�Huffman�1997� :�65)

d. *[-son]/μThe� head� of� a� mora� must� be� a� sonorant,� i.e.,� each� mora� that� is� linkedto� a� non-sonorant� should� also� be� linked� to� a� sonorant(cf.� Morén� 2003� :� 289)

Differences� (depending� on� whether� the� syllable� is� closed� or� not)� in� theduration�of�syllable-initial�stops�are�an�inevitable�consequence�of�a�system�inwhich� the� coda� consonants� must� be� linked� to� a� mora� (constraint� CODA-μ).

Külli Prillop

172

Table 2A schematic OT table

Table 3A language that avoids voiced stops

/input/ C1 C2 C3 C4a. candidate a *!b. candidate b *! **c. � candidate c ** **d. candidate d *!

/VTV/ *D *[-son]/µ *{V,N}TV *SHARED-µa.

*!

b.*!

c.*!

d.� � **

!"#

#

$%# &'#

# ()#*)+,-.#/0)1/#2/--#345-6#!7778#9:;%#

#

,%# &<=>?@'A!#

# B)64/#46-#C+DE-.#0)#/+D5C-#/-5F-D0/#2G6)/-C)H#-0#4C%#!77I8#"J;%#

#

.%# &KA/)DLM!#

NO-#O-4.#)P#4#F)64#FQ/0#$-#4#/)D)64D0R##

+%-%R#-4,O#F)64#0O40#+/#C+DE-.#0)#4#D)DA/)D)64D0R#/O)QC.#4C/)#$-#C+DE-.#0)#4#/)D)64D0#

2,P%#B)6SD#T::U8#TV7;%##

!"#$%&'(#>#C4D5Q45-#0O40#4*)+./#*)+,-.#/0)1/#

#######Z#

W#N#W#

'+PP-6-D,-/# 2.-1-D.+D5# )D# HO-0O-6# 0O-# /\CC4$C-# +/# ,C)/-.# )6# D)0;# +D# 0O-# .Q640+)D# )P#

/\CC4$C-A+D+0+4C# /0)1/# 46-# 4D# +D-*+04$C-# ,)D/-]Q-D,-# )P# 4# /\/0-F# +D# HO+,O# 0O-# ,).4# ,)D/)D4D0/#

FQ/0# $-# C+DE-.# 0)# 4# F)64# 2,)D/064+D0# ^_'>AZ;%# N4$C-# 9# 16)*+.-/# 4D# +CCQ/0640+)D# )P# 0O-# D)DA

/06-D50O-D+D5#)P#/0)1/#+D#,C)/-.#/\CC4$C-/%#`-#/O)QC.#4C/)#,)D/+.-6#0O40#0O-#+D/-60+)D#)P#F)64/#+D0)#

QD/06-//-.# /\CC4$C-/# +/# 16)O+$+0-.%# aD# 0O-# ,)Q6/-# )P# PQ60O-6# .-*-C)1F-D0# )P# 0O-# C4D5Q45-# 0O-#

,)D/064+D0/# &'# b4# &XWR(YNW# /H411-.# 1C4,-# +D# 0O-# O+-646,O\R# HO+,O# 6-/QC0-.# +D# /\CC4$C-A+D+0+4C#

,)D/)D4D0/#)P#,C)/-.#/\CC4$C-/#$-,)F+D5#*)+,-.#2&!"#$%!"#c#&#$%&!"'!"#$%!!"'"#$%&#'()*+*&,-.%##

2U;! ^_'>A!##

>CC#,).4#,)D/)D4D0/#46-#.)F+D40-.#$\#4#F)64#2G6)/-C)H#-0#4C%#!77I8#"9;%#

aD#,C)/-.#/\CC4$C-/R#6-.Q,0+)D#411C+-/#0)#.+1O0O)D5/#16-,+/-C\#4/# +0#.)-/#0)#/+D5C-#*)H-C/R#

$-,4Q/-#0O-#/-,)D.#,)F1)D-D0#)P#4#.+1O0O)D5#/O46-/# +0/#F)64#H+0O#0O-#/\CC4$C-#,).4%#NO-# P+6/0#

,)F1)D-D0#)P# 0O-#.+1O0O)D5R#)D#0O-#)0O-6#O4D.R#,4DD)0#/O46-#4#F)64#H+0O#0O-# +D+0+4C# /0)1#)P#4#

!"#

#

$%# &'#

# ()#*)+,-.#/0)1/#2/--#345-6#!7778#9:;%#

#

,%# &<=>?@'A!#

# B)64/#46-#C+DE-.#0)#/+D5C-#/-5F-D0/#2G6)/-C)H#-0#4C%#!77I8#"J;%#

#

.%# &KA/)DLM!#

NO-#O-4.#)P#4#F)64#FQ/0#$-#4#/)D)64D0R##


2,P%#B)6SD#T::U8#TV7;%##

!"#$%&'(#>#C4D5Q45-#0O40#4*)+./#*)+,-.#/0)1/#

########Z#

W#'#W#







,)D/)D4D0/#)P#,C)/-.#/\CC4$C-/#$-,)F+D5#*)+,-.#2&!"#$%!"#c#&#$%&!"'!"#$%!!"'"#$%&#'()*+*&,-.%##

2U;! ^_'>A!##

>CC#,).4#,)D/)D4D0/#46-#.)F+D40-.#$\#4#F)64#2G6)/-C)H#-0#4C%#!77I8#"9;%#




!"#

#

$%# &'#

# ()#*)+,-.#/0)1/#2/--#345-6#!7778#9:;%#

#

,%# &<=>?@'A!#

# B)64/#46-#C+DE-.#0)#/+D5C-#/-5F-D0/#2G6)/-C)H#-0#4C%#!77I8#"J;%#

#

.%# &KA/)DLM!#

NO-#O-4.#)P#4#F)64#FQ/0#$-#4#/)D)64D0R##


2,P%#B)6SD#T::U8#TV7;%##

!"#$%&'(#>#C4D5Q45-#0O40#4*)+./#*)+,-.#/0)1/#

####Z##Z#

W#N#W#







,)D/)D4D0/#)P#,C)/-.#/\CC4$C-/#$-,)F+D5#*)+,-.#2&!"#$%!"#c#&#$%&!"'!"#$%!!"'"#$%&#'()*+*&,-.%##

2U;! ^_'>A!##

>CC#,).4#,)D/)D4D0/#46-#.)F+D40-.#$\#4#F)64#2G6)/-C)H#-0#4C%#!77I8#"9;%#




!"#

#

$%# &'#

# ()#*)+,-.#/0)1/#2/--#345-6#!7778#9:;%#

#

,%# &<=>?@'A!#

# B)64/#46-#C+DE-.#0)#/+D5C-#/-5F-D0/#2G6)/-C)H#-0#4C%#!77I8#"J;%#

#

.%# &KA/)DLM!#

NO-#O-4.#)P#4#F)64#FQ/0#$-#4#/)D)64D0R##


2,P%#B)6SD#T::U8#TV7;%##

!"#$%&'(#>#C4D5Q45-#0O40#4*)+./#*)+,-.#/0)1/#

#######Z###

W#N#W#







,)D/)D4D0/#)P#,C)/-.#/\CC4$C-/#$-,)F+D5#*)+,-.#2&!"#$%!"#c#&#$%&!"'!"#$%!!"'"#$%&#'()*+*&,-.%##

2U;! ^_'>A!##

>CC#,).4#,)D/)D4D0/#46-#.)F+D40-.#$\#4#F)64#2G6)/-C)H#-0#4C%#!77I8#"9;%#




Table�4�provides�an�illustration�of�the�non-strengthening�of�stops�in�closedsyllables.�We�should�also�consider�that�the�insertion�of�moras�into�unstressedsyllables�is�prohibited.�In�the�course�of�further�development�of�the�languagethe� constraints� *D� ja� *{V,N}TV� swapped� places� in� the� hierarchy,� whichresulted� in� syllable-initial� consonants� of� closed� syllables� becoming� voiced(*jo.�� en >� *jo.gen ~� jo.γen ’river,� gen.sg’).

(3) CODA-μ�All� coda� consonants� are� dominated� by� a� mora(Broselow,� Chen,� �Huffman� 1997� :� 64)

In closed syllables, reduction applies to diphthongs precisely as it doesto single vowels, because the second component of a diphthong shares itsmora with the syllable coda. The first component of the diphthong, on theother hand, cannot share a mora with the initial stop of a syllable, becausethis would leave the diphthong too short (half the mora plus another half,see Table 5).

(4) *SHORTVVNo� monomoraic� diphtongs� (Yip� 1996� :� 768)

Table�4�does�not�include�the�possibility�of�a�single�intervocalic�stop�takingthe�mora�that� it�needs� in�order� to� lengthen�from�the�preceding�as�opposedto�the�following�vowel,�thus�becoming�a�short�geminate�(see�Figure�5a�above).Such� gemination� cannot� be� blocked� by� a� subsequent� closed� syllable.� Thereare�several�solutions�to�this�problem.�The�simplest�way�is�to�prohibit�segmentdivision�between�syllables�(constraint�ONESYLL).�This�does�not�prevent�stopswhich� are� linked� to� a� mora� in� the� input� from� being� divided� between� twosyllables� in� the� output,� because� faithfulness� constraints� (MAx-μ-LINK andIDENT-C-μ)�require�duration�contrast�to�be�maintained�in�the�output.�The�migra-�tion� of� a� moraic� consonant� into� the� second� syllable� is� prevented� by� WSP.


173

Table 4Single stop gradation

/joken/ ’river,�gen.sg’ CODA-µ *D *[-son]/µ� *{V,N}TV *SHARED-µ�

a.� �jo � e n

*! **

b.jo g e n

*! **

c.jo k e n

*! **

d.jo k e n

*!

/joki/ ’river,� nom.sg’ CODA-µ *D *[-son]/µ� *{V,N}TV *SHARED-µ�

a.jo � i

*!

b.jo g i

*!

c.� �jo k i

**

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

########S#

./#01#)#5#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

#########S#

M1#6#)#5#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

#########S#

M1#N#)#5#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

#####S##S#

M1#N#)#5#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

#######S#

./#01#/#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

########S#

M1#6#/#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

########S#

M1#N#/#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

(5) a. ONESYLLA� segment� is� a� member� of� exactly� one� syllable� (Green� 1997� :� 137)

b. MAx-μ-LINKA�segment�that�is�linked�to�a�mora�in�the�input�is�also�linked�to�a�morain� the� output� (Morén� 2003� :� 294)

c. IDENT-C-μA�consonant�in�the�output�must�not�be�shorter�than�it�is�in�the�input

d. WSPHeavy� syllables� are� stressed� (Kager� 1999� :� 155)

In OT, the output structure of geminates depends on the hierarchy ofconstraints that defines the language. Exactly as in the case of single stopgradation, the link between coda and mora is more important than reten-tion of the full underlying length of the stop, which results in geminatespreceding a closed syllable being shortened (see Table 6). In the appro-priate part of the table, the constraints ONESYLL and WSP have beeninserted to show that they do not preclude the division of input geminatesbetween two syllables in the output.

Where,� instead� of� the� constraint� *[-son]/μ,� the� top� of� the� hierarchy� isoccupied� by� a� positional� markedness� constraint� which� prohibits� stressedvowels�from�shortening�when�they�share�a�mora�(*SHARED-μ/ éσ),�and�requiresconservation�of�vowel�length�(IDENT-V-μ)�a�different�type�of�geminate�alter-nation� occurs:� geminates� before� open� syllables� alternate� with� lengthenedsingletons� before� closed� syllables.� Closed� syllables� of� the� aforementionedtype�are�bimoraic�(see�Figure�6).� In�the�case�of�this�type�of�geminate�alter-nation,�too,�the�insertion�of�moras�into�unstressed�syllables�must�be�prohib-ited� if� gradation� is� to� appear.

Külli Prillop

174

Table 5Stops in the onset of syllables with long vowels

/antoi/ ’s/he� gave’ CODA-µ *SHORTVV *D *{V,N}TV *SHARED-µ

a.a n � o i

*! ** **

b.a n � o i

*!

c.a n d o i

*!

d.� �a n t o i

**

/antoin/ ’I� gave’ CODA-µ *SHORTVV *D *{V,N}TV *SHARED-µ

a.a n � o i n

*! **!!

b.a n t o i n

*! ****

c.� �a n � o i n

*! **

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

#########S####

'#+#67#1#/#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

##########S#S#

5'+#67#1#/#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

################S#S#

'#5#2#1#/#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

##########S#S#

'#5#-#1#/#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

##########S#S#

5'+#67#1#/#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

#########S##S###

'#5#-#1#/#5#

!"#

#

$%&&'(&)*#()+',$)#-./$#01,&2#&)'3)#-.)#2/4.-.156#-11#$.17-#8.'&9#-.)#:17'#4&,$#'51-.)7#.'&9*#$))#

;'(&)#<=>##

8?=! @ABCD;EE#

F1#:151:17'/+#2/4.-156$#8G/4#!HHIJ#"IK=#

!"#$%&'(#A/56&)#$-14#67'2'-/15#

!"#$%&)(#A-14$#/5#-.)#15$)-#19#$%&&'(&)$#0/-.#&156#310)&$#

###########S#S####

5'+#67#1#/#5#

(6) a. SHARED-μ/ éσThere� are� no� shared� moras� in� a� stressed� syllable

b.� IDENT-V-μThe� vowel� in� the� output� is� not� shorter� than� in� the� input


175

Table 6Alternation of geminate duration I

/vaμkμaμ/’bushel, nom.sg’

CODA-µ*[-son]/µMAX-µ-LINK WSP IDENT-C-µ ONESYLL *{V,N}TV*SHARED-µ

a.

v a k a*! **

b.

v a ka*! ** **

c.

v a k a*! ** **

d.� �v a k a

** ****

/vaμkμaμn/’bushel,�gen.sg’


a.

v a k a n*! ** **

b.� �v a k a n

** ** ****

c.

v a k a n*! ** ****

d.

v a ka n*! ****

/vaμkaμn/’pious,�gen.sg’


a.� � **

b.v a k a n

*!

c.v a k a n

*!

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!####

#####!#:##!####

4#*#5#*###

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!####

#####!#####!####

4#*#5#*###

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!####

#####!#####!####

4#*#5#*###

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!####

#####!#####!####

4#*#5#*###

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!#

#####!#:##!#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!#

#####!######!#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

####!#####!#

####!#####!#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

#####!#####!#

#####!#:##!#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

##########:#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

###:#####:#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

v a k a n

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#2#

##########:#

4#*#5#*#)#

#

TUV! *P# EN$OG89!345#

# HW'('#*('#),#=W*('0#/,(*=#+)#*#=&('=='0#=X%%*R%'P#

RP##28GBH9K9:#

# HW'#4,Y'%#+)#&W'#,1&Z1&#+=#),#=W,(&'(#&W*)#+)#&W'#+)Z1&P#

There is no simple way of determining whether different types of gemi-nate gradation evolved side by side or whether either one can be tracedback to the other one. Most probably, the choice of the type of gradationdepends on the extent to which the language tolerated heavy unstressedsyllables. In contemporary Standard Finnish, the diphthongs of medial sylla-bles have been preserved and geminates alternate with single stops; inStandard Estonian, unstressed diphthongs have been lost and long gemi-nates alternate with short ones.

Tiit-Rein�Viitso�(1981�:�178;�2008�:�122)�distinguishes�two�principal�typesof� gradation� in� Finnic� languages� —� the� type� in� which� the� weak� grade� ofthe�geminate� is�phonologically�similar� to� the�strong�grade�of�a�single�stop(or� obstruent),� cf.� Votic,� Ingrian,� Finnish,� North� Karelian,� and� the� type� inwhich�this�is�not�the�case,�cf.�South�Karelian,�Lude.�Estonian�could�be�clas-sified� as� belonging� to� the� first� group,� if� we� consider� alternations� of� thetype� va�.ka :� va�� .kas ’bushel’� as� quantity� gradation� and� look.ka :� loo.gas’shaft� bow’� as� stop� gradation.

Külli Prillop

176

Table 7Alternation of geminate duration II

/vaμkμaμ/ *SHARED-µ/éσ� *{V,N}TV IDENT-V-µ

a.� �v a k a

b.v a k a

*!

c.v a k a

*! ****

/vaμkμaμn/ *SHARED-µ/éσ� *{V,N}TV IDENT-V-µ

a.v a k a n

*!

c.v a k a n

*! ****

d.� �v a k a n

/vaμkaμ/ *SHARED-µ/éσ� *{V,N}TV IDENT-V-µ

a.� �v a ka

**

b.v a k a

*!

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

####!########!#

#####!###!###!#

4#*#5#*#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#

'4,%4'0#I+0'#EL#I+0'#,(#KH'&H'(#'+&H'(#,)'#G*)#E'#&(*G'0#E*G5#&,#&H'#,&H'(#,)'D#M,I&#J(,E*E%L?#

&H'#GH,+G'#,-#&H'#&LJ'#,-#.(*0*&+,)#0'J')0I#,)#&H'#'N&')&#&,#KH+GH#&H'#%*).1*.'#&,%'(*&'0#H'*4L#

1)I&('II'0#IL%%*E%'ID# 2)#G,)&'/J,(*(L#7&*)0*(0#O+))+IH?# &H'#0+JH&H,).I#,-#/'0+*%#IL%%*E%'I#H*4'#

E'')# J('I'(4'0# *)0# .'/+)*&'I# *%&'()*&'# K+&H# I+).%'# I&,JIP# +)# 7&*)0*(0# :I&,)+*)?# 1)I&('II'0#

0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##

B++&<9'+)#>++&I,#Q!""R#STURTVW#T!!X#0+I&+).1+IH'I#&K,#J(+)G+J*%#&LJ'I#,-#.(*0*&+,)#+)#O+))+G#

%*).1*.'I#%# &H'# &LJ'# +)#KH+GH# &H'#K'*5# .(*0'#,-# &H'# .'/+)*&'# +I# JH,),%,.+G*%%L# I+/+%*(# &,# &H'#

I&(,).#.(*0'#,-#*#I+).%'#I&,J#Q,(#,EI&(1')&X?#G-D#>,&+G?#2).(+*)?#O+))+IH?#@,(&H#Y*('%+*)?#*)0#&H'#&LJ'#

+)#KH+GH#&H+I#+I#),&#&H'#G*I'?#G-D#7,1&H#Y*('%+*)?#Z10'D#:I&,)+*)#G,1%0#E'#G%*II+-+'0#*I#E'%,).+).#&,#

&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

#####!########!#

#####!########!#

4#*#5#*#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

#####!#######!#

#####!#######!#

4#*#5#*#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

####!########!#

####!###C###!#

4#*#5#*#)#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

#####!#######!#

&&&&&!########!#

4#*#5#*#)#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

#####!#######!#

#####!###C###!#

4#*#5#*#)#

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

######!######!#

######!#######!#

4#*#5#*##

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

!"#

#

!"#$%&'(#$%&'()*&+,)#,-#.'/+)*&'#01(*&+,)#22#

#####!#######!#

#####!#######!#

4#*#5#*##

BH'('# +I# ),# I+/J%'# K*L# ,-# 0'&'(/+)+).# KH'&H'(# 0+--'(')&# &LJ'I# ,-# .'/+)*&'# .(*0*&+,)#





0+JH&H,).I#H*4'#E'')#%,I&#*)0#%,).#.'/+)*&'I#*%&'()*&'#K+&H#IH,(&#,)'ID##





&H'# -+(I&# .(,1J?# +-# K'# G,)I+0'(# *%&'()*&+,)I# ,-# &H'# &LJ'# !"#$!"## W# !"#%!"#$# &'()*+,-. *I# [1*)&+&L#

.(*0*&+,)#*)0#%&&"!"##W#%&&!'#$#&/0,,12.'03-.1).I&,J#.(*0*&+,)D##

Leaving� aside� the� dependence� of� the� weak� grade� of� geminates� on� thesounds� in� the� previous� syllable� (these� distinctions� have� been� discussed� indetail�by�Viitso�1981;�2008),�the�differences�between�types�can�be�adequatelydescribed�by�the�six�constraints�*[-son]/μ,�*SHARED-μ/éσ,�IDENT-C-μ,�IDENT-V-μ,*{V,N}TV�and�WSP.�The�following�analysis�was�performed�using�the�appli-cation� OTSoft� (Hayes,� Tesar,� Zuraw� 2003).

A� language� in� which� the� weak� grade� of� a� geminate� is� the� same� as� thestrong� grade� of� a� single� stop� (see� Example� 8)� would� result� from� the�following�hierarchies�(the�constraints�whose�relative�order�is�free�are�sepa-rated� by� a� comma):

(7) a. *SHARED-μ/ éσ >� *{V,N}TV� >� IDENT-V-μ� >� IDENT-C-μ,� *[-son]/μ,� WSPb. *[-son]/μ� >� IDENT-C-μ� >� *SHARED-μ/ éσ >� *{V,N}TV,� WSP� >� IDENT-V-μ

In� the� first� case� (7a),� the� strong� grade� of� the� geminate� appears� as� anordinary� geminate� possessing� an� independent� mora� (see� Figure� 4� above),in�the�second�case�(7b)�it�takes�the�form�of�a�geminate�that�shares�its�moras(see�Figure�5b�above).�As�a�development�of�the�initial�type,�the�weak�gradeof� single� stops,� which� were� not� linked� to� a� mora� (see� Figure� 3a� above),became�voiced.�Then,�the�strong�grade�of�the�single�stop�and�the�weak�gradeof� the�geminate�could�be�released�from�the� link�with�mora,�and�to�becomehalf-voiced� (as� in� Ingrian),� while� in� the� case� of� geminates,� abandoning� themora� link� may� have� at� first� been� prohibited� (as� in� South� Karelian).

(8) North� Karelian tuk.ka : tu.kan ’hair,� nom.sg/gen.sg’,šu.ka : šu.van ’comb,� nom.sg/gen.sg’

(see� Viitso� 1981;� 2008 :� 126)Ingrian� (Soikkola) säk.ki : sä.gid ’sack,� nom.sg/nom.pl’

kodi : koin ’house,� nom.sg/gen.sg’(Laanest� 1986� :� 33)

South� Karelian tuk.ka : tu.kan ’hair,� nom.sg/gen.sg’šu.ga : šu.van ’comb,� nom.sg/gen.sg’

(see� Viitso� 1981;� 2008 :� 126)

Languages� in� which� the� weak� grade� of� geminates� did� not� degeminate(see� (9)),� were� characterised� by� hierarchy� (10),� whose� only� difference� fromhierarchy�(7)�consists� in� the�presence�of� the�WSP�constraint.� If�WSP�wouldbe� a� rank� lower,� alternation� of� long� and� short� geminates� would� not� be�possible.

(9) Estonian kot.ti : ko∑t�.tid ’bag,� part.sg/nom.pl’tu.ba : toad ’room,� nom.sg/nom.pl’

(10) *[-son]/μ� >� WSP,� IDENT-C-μ� >� *SHARED-μ/ éσ� >� *{V,N}TV� >� IDENT-V-μ

The�Estonian�type�of�gradation�is�less�common.�One�could�not�give�up�thistype�of�gradation,�since�were�Estonian� to�use�hierarchy� (7),� the�contemporaryshort�stops�(such�as� in�ko∑t�.tid)�would�have�to�have�arisen�as�a�result�of�gemi-nation.� In�genitive� forms,�a�gemination� like� this�could�be�explained�by�postu-lating�the�shortening�of�the�long�vowel�resulting�from�the�dropping�of�the�geni-tive�ending�-n.�In�all�other�cases,�one�would�have�to�assume�an�analogous�path.

Almost� all� of� the� other� possible� hierarchies� neutralise� gradation� to� alarger�or�smaller�extent.�For�instance,�the�South�Veps�variant,�in�which�single


1772 Linguistica Uralica 3 2011

stops� have� become� voiced� and� geminates� have� been� replaced� by� singlestops� (see� (11)),� gives� rise� to� hierarchy� (12).

(11) Veps uk (< *ukkoi) : u.kod ’old�man,�nom.sg/nom.pl’raud : rau.dan ’iron,�nom.sg/gen.sg’ (Laanest�1982�:�135)

(12) *SHARED-μ/ éσ,� *[-son]/μ� >� WSP,� IDENT-C-μ,� IDENT-V-μ� >� *{V,N}TV

The�factorial�typology�frequently�used�in�OT�analyses�(see�Prince,��Smolensky1993/2004� :� 105—118)� will� not� provide� any� useful� information� in� the� caseof�closely�related�languages.�Overgeneration�may�be�caused�by�similaritiesin� the� history� of� the� related� languages� and� not� by� a� mistaken� selection� ofconstraints.� Still,� preliminary� data� suggest� that� the� six� constraints� listedabove�only�generate�one�combination�that�is�not�encountered�in�any�Finniclanguage�—�that�of�the�alternation�of�a�geminated�single�stop�with�a�shortsingle� stop.� Precisely� this� type� of� gradation� is� found� in� Saami� languages,e.g.� lop.pe :� lo.best ’permission,� nom.sg/elat.sg’� (see� Gordon� 1997).�

4. Conclusions

The lenition of stops between sonorants is a process that is relativelycommon cross-linguistically. Gradation in Finnic languages can be shownto represent a mirror image of that process — it can be analysed as anattempt to prevent lenition. In order to do that, the stop in the syllableonset was linked to the mora of the following vowel, i.e. the stop waslengthened at the expense of the vowel. It is easier to maintain the voice-lessness of long stops than it is of short ones. Linking syllable onset to avowel’s mora was not an option in closed syllables, in which the vowelmora had to be linked to the coda. For the same reason, the full extent ofthe duration of geminates could not be preserved in front of closed syllables.A schematic representation of the hierarchy of Optimality Theory constraints

related to consonant gradation in Finnic languages is shown in Figure 7. Solidarrows denote the original type of consonant gradation which suits all Finniclanguages. In such a hierarchy, long and short geminates will alternate, asthey presumably did in Estonian before it developed quantity gradation, cf.*kot.ti :� ko∑t�.tid ’bag,� part.sg/nom.pl’.� Arrowheads� on� dotted� lines� point� tochanges�that�lead�to�what�may�be�the�other�original�type�of�consonant�grada-tion� in� which� geminates� alternate� with� single� stops,� as� they� do� in� most� ofthe� other� Finnic� languages,� such� as� the� Finnish� and� Karelian� tuk.ka :� tu.kan’hair,�nom.sg/gen.sg’.�The�principal�difference�between�these�two�types�is�thelocation�of�the�WSP�constraint�—�that�is,�whether�the�language�allowed�heavyunstressed� syllables� or� not.� It� is� remarkable� that,� for� instance� in� the� case� ofFinnish,� which� to� date� preserves� its� original� Vi-diphthongs� of� unstressed�syllables,� the� suitable� type� of� consonant� gradation� is� one� in� which� WSP� islocated�at�a�relatively�low�level�of�the�hierarchy,�whereas�for��Estonian,�whichhas�lost�its�Vi-diphthongs,�the�suitable�type�of�consonant�gradation�is�one�inwhich� WSP� is� located� at� the� top� of� the� hierarchy.

The�possible�hierarchies�of�constraints�used�in�the�analysis,�the��gradationsystems� that� correspond� to� them� and� their� suitability� to� Finnic� languagesand�dialects�as�well�as�to�more�distant�kin�languages�represent�an��interestingsubject� for� further� investigation.

Külli Prillop

178

The� analysis� of� consonant� gradation� presented� here� turns� directly� onthe� closedness� or� openness� of� syllables,� as� does� its� traditional� definition.There� is� no� need� to� regard� closed� unstressed� syllables� as� too� heavy,� or� tolengthen�the�vowels�of�open�non-initial�syllables,�which�would�trigger� thefoot� balancing� process.� Gradation� was� not� simply� lenition� or� fortition� —it� involved� both� aspects� at� the� same� time.� In� open� syllables,� the� durationof�stops� increased,�while� in�closed�syllables� the�stops�whose�duration�hadnot� increased� started� to� shorten� and� eventually� disappeared� altogether.

Address

Külli PrillopUniversity� of� TartuE-mail:� [email protected]

Abbreviations

C — consonant, elat — elative, gen — genitive, H — heavy syllable, K — stop, L —light syllable, nom — nominative, part — partitive, pl — plural, sg — singular, V —vowel, σ — syllable, µ — mora.

R E F E R E N C E S

A l l e n, S. J., M i l l e r, J. L. 1999,� � Effects� of� Syllable-Initial� Voicing� and�Speaking�Rate�on�the�Temporal�Characteristics�of� �Monosyllabic�Words.�—The� Journal� of� the� Acoustical� Society� of� America� 106,� 2031—2039.

B r o s e l o w, E., C h e n, S., H u f f m a n, M. 1997,��Syllable�Weight.�Conver-gence� of� Phonology� and� Phonetics.� —� Phonology� 14,� 47—82.

B y e, P. 2001,� � Virtual� Phonology.� Rule� Sandwiching� and� Multiple� Opacity� inNorth�Saami,�Tromsø�(PhD�dissertation�(manuscript),�University�of�Tromsø).

—— 2005,�Coda�Maximisation�in�Northwest�Saamic.�—�Nordic�Journal�of�Linguis-tics� 28,� 189—221.

C a m p b e l l, L. 1999,� � Historical� Linguistics,� Cambridge,� MA.C a s t r ée n, M. A. 1838,��De�affinitate�declinationum�in�lingua�fennica,��esthonica

et� lapponica,� Helsingforsiae.


179

!"#

#

$%&&'(&)#*+$),#-'$#&.+/)0#,*#,1)#2*3'#*4#,1)#4*&&*-.+5#6*-)&7#.8)8#,1)#$,*9#-'$#&)+5,1)+)0#',#,1)#

):9)+$)#*4# ,1)#6*-)&8# ;,# .$#)'$.)3# ,*#2'.+,'.+# ,1)#6*.<)&)$$+)$$#*4# &*+5#$,*9$# ,1'+# .,# .$#*4# $1*3,#

!"#$%& '(")("*& $+,,-.,#& !"$#/& /!& -& 0!1#,2$&3!4-&1-$& "!/& -"& !9,.*+# .+# <&*$)0# $%&&'(&)$7# .+#-1.<1#

6*-)&#2*3'#1'0#,*#()#&.+/)0#,*#,1)#<*0'8#=*3#,1)#$'2)#3)'$*+7#,1)#4>&&#):,)+,#*4#,1)#0>3',.*+#*4#

5)2.+',)$#<*>&0#+*,#()#93)$)36)0#.+#43*+,#*4#<&*$)0#$%&&'(&)$8#

?#$<1)2',.<# 3)93)$)+,',.*+#*4# ,1)#1.)3'3<1%#*4#@9,.2'&.,%#A1)*3%# <*+$,3'.+,$# 3)&',)0#,*#

<*+$*+'+,# 53'0',.*+# .+# =.++.<# &'+5>'5)$# .$# $1*-+# .+# =.5>3)# B8# C*&.0# '33*-$# 0)+*,)# ,1)# *3.5.+'&#

,%9)#*4#<*+$*+'+,#53'0',.*+#-1.<1#$>.,$#'&&#=.++.<#&'+5>'5)$8#;+#$><1#'#1.)3'3<1%7#&*+5#'+0#$1*3,#

5)2.+',)$# -.&&# '&,)3+',)7# '$# ,1)%# 93)$>2'(&%# 0.0# .+# D$,*+.'+# ()4*3)# .,# 0)6)&*9)0# E>'+,.,%#

53'0',.*+7#<48#F!"#$#%&G#!"#$$#%'& 2.-*5&$*%6-4/76,%"!32%&844!19#-:$&!"&:!//#:&,("#$&6!("/&/!&;9-"*#$&

,1',# &)'0# ,*# -1',#2'%# ()# ,1)# *,1)3# *3.5.+'&# ,%9)# *4# <*+$*+'+,# 53'0',.*+# .+# -1.<1# 5)2.+',)$#

'&,)3+',)#-.,1#$.+5&)#$,*9$7#'$#,1)%#0*#.+#2*$,#*4#,1)#*,1)3#=.++.<#&'+5>'5)$7#$><1#'$#,1)#=.++.$1#

'+0#H'3)&.'+##(!$!)&G##($!)*&&29-(47#+*28#$5I5)+8#$582%&<9#&64(";(6-,&:(==#4#";#&.#/1##"&/9#$#&/1!&

,%9)$# .$# ,1)# &*<',.*+# *4# ,1)# JCK# <*+$,3'.+,# ># ,1',# .$7# -1),1)3# ,1)# &'+5>'5)# '&&*-)0# 1)'6%#

>+$,3)$$)0#$%&&'(&)$#*3#+*,8#;,#.$#3)2'3/'(&)#,1',7#4*3#.+$,'+<)#.+#,1)#<'$)#*4#=.++.$17#-1.<1#,*#0',)#

93)$)36)$# .,$# *3.5.+'&# +%L0.91,1*+5$# *4# >+$,3)$$)0# $%&&'(&)$7# ,1)# $>.,'(&)# ,%9)# *4# <*+$*+'+,#

53'0',.*+# .$#*+)# .+#-1.<1#JCK# .$# &*<',)0#',# '# 3)&',.6)&%# &*-# &)6)&#*4# ,1)#1.)3'3<1%7#-1)3)'$# 4*3#

D$,*+.'+7# -1.<1# 1'$# &*$,# .,$# +%L0.91,1*+5$7# ,1)# $>.,'(&)# ,%9)# *4# <*+$*+'+,# 53'0',.*+# .$# *+)# .+#

-1.<1#JCK#.$#&*<',)0#',#,1)#,*9#*4#,1)#1.)3'3<1%8#

#FML$*+NI?#####O@P?L?##

#Q?RL?LS;TH#####;PDTALOL?#####JCK##

#FCU?VDPL?7@A##

#FCU@VAWW####@TDCXSS#####FP####

#FYW7TZAW##

#FCU?VDPL?#####;PDTALWL?##

!"#$%&'()#?#1.)3'3<1%#*4#<*+$,3'.+,$#&)'0.+5#,*#=.++.<#<*+$*+'+,#53'0',.*+#Figure 7. A hierarchy of constraints leading to Finnic consonant gradation.

2*

d e L a c y, P. 2002,��The�Formal�Expression�of�Markedness,�Amherst�(PhD�disser-tation� (manuscript),� University� of� Massachusetts).

E e k, A., H e l p, T. 1986,� � Rütminihked� eesti� keele� kujunemisloos,� Tallinn.E e k, A.,��M e i s t e r, E. 1997,��Simple�Perception�Experiments�on��Estonian�Word

Prosody:�Foot�Structure�vs.�Segmental�Quantity.�—�Estonian�Prosody:�Papersfrom� a� Symposium,� Tallinn,� 71—99.

—— 2004,� Foneetilisi� katseid� ja� arutlusi� kvantiteedi� alalt.� Takt,� silp� ja� välde.� —KK,� 251—271,� 336—357.

E h a l a, M.,��2003,��Estonian�Quantity.�Implications�for�Moraic�Theory.�—��GenerativeApproaches� to� Finnic� and� Saami� Linguistics,� Stanford,� 51—80.

G o r d o n, M. 1997,��A�Fortition-Based�Approach�to�Balto-Fennic-Sámi��ConsonantGradation.� —� Folia� Linguistica� Historica� 18,� 49—79.

—— 2002,��A�Phonetically-Driven�Account�of�Syllable�Weight.�—�Language 78,�51—80.

—— 2009,� � Prominence� and� Gemination� in� Ingrian.� —� LU� xLV,� 81—100.G r e e n, A. D. 1997,� � The� Prosodic� Structure� of� Irish,� Scots� Gaelic,� and� Manx

(PhD� Dissertation� (manuscript),� Cornell� University).H a w k i n s, S, N g u y e n, N. N. 2004,� � Influence�of�Syllable-Coda�Voicing

on� the� Acoustic� Properties� of� Syllable-Onset� /l/� in� English.� —� Journal� ofPhonetics� 32, 199—231.

H a y e s, B. 1989,��Compensatory�Lenghtening�on�Moraic�Phonology.�—��LinguisticInquiry� 20,� 253—306�

—— 1995,��Metrical�Stress�Theory.�Principles�and�Case�Studies,�Chicago—London.H a y e s, B., T e s a r, B., Z u r a w, K. 2003,� � OTSoft� 2.1� (Software� �package).

http://www.linguistics.ucla.edu/people/hayes/otsoft/.H a y e s, B., S t e r i a d e, D. 2004,� � Introduction.� The� Phonetic� Bases� of� Phono-

logical�Markedness.�—�Phonetically�Based�Phonology,�Cambridge,�1—33.H y m a n, L. M. 1985,� � A� Theory� of� Phonological� Weight,� Dordrecht.I t k o n e n, T. 1969,��Zur�Wertung�der�finnisch-ugrischen�Lautforschung.�—�UAJb.

41,� 76—111.I v a, S. 2010,� �Grade�Alternation� in�Võro�South�Estonian.�—�LU�xLVI,�161—174.K a g e r, R. 1999,� � Optimality� Theory,� Cambridge.K a w a h a r a, S., P a n g i l i n a n M., G a r v e y K. 2011,� � Spectral� Conti-

nuity�and�the�Perception�of�Duration:�Implications�for�Phonological�Patternsof�Sonorant�Geminates�(Submitted�manuscript).�http://roa.rutgers.edu/files/1132-0111/1132-KAWAHARA-0-0.PDF.

K i p a r s k y, P.� 2006,� Livonian� Stød� (Manuscript).� http://www.stanford.edu/~kiparsky/Papers/livonian.pdf.

K i r c h n e r, R. M. 2004,� �Consonant�Lenition.�—�Phonetically�Based��Phonolo��gy,Cambridge,� � 313—345.�

K o r h o n e n, M. 1988,��The�History�of�the�Lapp�Language.�—�The�Uralic�Languages.Description,�History,�and�Foreign� Influences,�Leiden,�264—287.

K r a e h e n m a n n, A. 2001, Swiss German Stops. Geminates all over the Word.— Phonology 18, 109—145.

L a a n e s t, A. 1982, Einführung in die ostseefinnischen Sprachen, Hamburg.—— 1986,� � Isuri� keele� ajalooline� foneetika� ja� morfoloogia,� Tallinn.�L a V o i e, L. 2001, Consonant Strength. Phonological Patterns and Phonetic

Manifestations, New York. M a d d i e s o n, I. 1993,��Splitting�the�Mora.�—�UCLA�Working�Papers�in��Phonetics

83,� California,� 9—18.� http://escholarship.org/uc/search?entity=uclaling_wpp.

M c C a r t h y, J. 2000,� �The�Prosody�of�Phase� in�Rotuman.�—�Natural�Languageand� Linguistic� Theory� 18,� 147—197.

M c C a r t h y, J., P r i n c e, A. 1993,� � Generalized� Alignment.� —� Yearbook� ofMorphology,� Dordrecht,� 79—153.

—— 1995,� � Faithfulness� and� Reduplicative� Identity.� —� University� of� Massachu-setts� Occasional� Papers� in� Linguistics� 18.� Papers� in� Optimality� Theory,Amherst,� 249—384.

M c I n t o s h, J. B. 1945,��Huichol�Phonemes.�—�International�Journal�of��AmericanLinguisticsl� 11,� 31—35.

Külli Prillop

180

M o h a n a n, T., M o h a n a n K. P. 2003,�Towards�a�Theory�of�Constraints� inOT.�Emergence�of�the�Not-so-unmarked�in�Malayalee�English�(Manuscript).http://roa.rutgers.edu/files/601-0503/601-0503-MOHANAN-0-0.PDF

M o r é n, B. 2003,� �Weight�Typology.�An�Optimality�Theoretic�Approach.�—�TheLinguistic� Review� 20,� 281—304.

P i k a m ä e, A.� � 1957,� � Tüveline� astmevaheldus� läänemeresoome� keeltes� ja� lapikeeles,� Tartu� (TRÜT� 50).

P o s t i, L. 1953,� � From� Pre-Finnic� to� Late� Proto-Finnic.� Studies� on� the� Develop-ment� of� the� Consonant� System.� —� FUF� xxxI,� 1—91.

P r i n c e, A., S m o l e n s k y, P. 1993/2004,� Optimality� Theory.� ConstraintInteraction� in� Generative� Grammar,� Malden,� MA.

R a v i l a, P. 1960,��Probleme�des�Stufenwechsels�im�Lappischen.�—�FUF�33,�285—325.

S a m m a l l a h t i, P. 1998,��The�Saami�Languages.�An�Introduction,�Káráshjohka.�S m i t h, J. 2008,� � Lenition� and� Fortition� in� Optimality� Theory.� —� Lenition� and

Fortition,� Berlin—New� York,� 519—560.S p r o u s e, R. 1996,� �Vowels�That�Borrow�Moras.�Geminates�and�Weight� in�OT.

—� Proceedings� of� the� Northeast� Linguistic� Society� 26,� 392—408.T o p i n t z i, N. 2010,��Onsets.�Suprasegmental�and�Prosodic�Behavior,�Cambridge.T r a n e l, B. 1991,��CVC�Light�Syllables,�Geminates�and�Moraic�Theory.�—��Phono�-

�lo�gy� 8,� 291—302.v a n S a n t e n, J. P. H. 1992,��Contextual�Effects�on�Vowel�Duration.�—�Speech

Communication� 11,� 513—546V e n d e t t i, J. J., v a n S a n t e n, J. P. H. 1998,� � Modeling� Vowel� Dura-

tion�for�Japanese�Text-to-Speech�Synthesis.�—�Proceedings�of�the�5th�Inter-national� Conference� on� Spoken� Language� Processing,� Sydney,� 2043—2046.

V i i t s o, T.-R. 1981,�Finnic�Gradation:�Types�and�Genesis.�—�СFU xVII,�176—185.—— 2003,��Rise�and�Development�of�the�Estonian�Language.�—��Estonian�Language,

Tallinn� (Linguistica� Uralica.� Supplementary� Series� /� Volume� 1),� 130—230.—— 2007,� � Livonian� Gradation:� Types� and� Genesis.� —� LU� xLIII,� 45—62.—— 2008,� � Läänemeresoome� astmevaheldus:� tüübid� ja� tekkimine.� —� Liivi� keel

ja� läänemeresoome� keelemaastikud,� Tartu—Tallinn,� 120—127.W a t s o n, J. C. E. 2007,� � Syllabification� Patterns� in� Arabic� Dialects.� Long

Segments� and� Mora� Sharing.� —� Phonology� 24,� 335—356.W e t z e l s, W. L., M a s c a r ó, J. 2001,��The�Typology�of�Voicing�and��Devoicing.

—� Language� 77,� 207—244.W i i k, K. 1997,� � The� Uralic� and� Finno-Ugric� Phonetic� Substratum� in� Proto-

Germanic.� —� LU� xxxIII,� 258—281.Y i p, M. 1996,��Lexicon�Optimization�in�Languages�without�Alternations.�—�Current

Trends� in� Phonology.� Models� and� Methods,� Vol.� 2,� Salford,� 757—788.Z e c, D. 2007, The Syllable. — The Cambridge Handbook of Phonology,

Cambridge, 161—194.

KŒLLI PRILLOP (Tartu)

ÄEREDOVANIE STUPENEJ SMYÄNYH KAK REZULXTAT DELENIQMOR V PRIBALTIJSKO-FINSKIH QZYKAH

Oslablenie smyänyh v pribaltijsko-finskih qzykah, kotoroe zavisit ot ot -krytosti ili zakrytosti sleduœYego sloga, kak äeredovanie stupenej pred-stavlqet soboj redkoe qvlenie v qzykah mira. Mowet imetx mesto vse we fone-tiäeski estestvennoe oslablenie mewdu glasnymi, kotoromu prepqtstvuet ilisposobstvuet raznaq morovaq struktura slogov. V takom sluäae razliäiq v pri -baltijsko-finskom äeredovanii stupenej zavisit ot togo, dopustimy li vqzyke tqwelye bezudarnye slogi.


181

finnic stop gradation as an effect of mora ......finnic stop gradation as an effect of mora sharing...

Documents