CAS LX 500 Topics in Linguistics Language Universals (1) Sets of syllabic segments Fall 2000 October 24, 2000 a. Lithuanian, Bulgarian Paul Hagstrom Week 8a: and sonority b. vowels, English, Gonja c. vowels, sonorants, Imdlawn Tashlhiyt Berber The typology of heavy syllables and sonority (2) σσσ Clements, G. N. (1990). The role of the sonority cycle in core syllabification. In Kingston and Beckman RgU RgU RgU (eds.), Papers in Laboratory Phonology I. Cambridge University Press. µµµ Zec, D. (1995). Sonority constraints on structure. Phonology 12:85Ð129. 111 Kenstowicz, M. (1994). Phonology in generative grammar. Blackwell. (p.292) ald

Jespersen: In every group of sounds there are just as many syllables as there are clear Berber Berber Berber relative peaks of sonority. English English Bulgarian Sonority: 1. voiceless stops ¥ A similar thing can be said about the segments which contribute to voiceless obstruents (those which add a second mora). Languages make a cut on the , 2. voiced stops allowing everything higher (more sonorous) as the second mora. 3. voiced fricatives 4. voiced nasals (3) Sets of moraic segments voiced laterals sonorants (m, n, r, l) a. vowels Khalkha Mongolian, Yindiµ 5. voiced r-sounds b. vowels, sonorants Lithuanian, Tiv 6. voiced high vowels vowels c. vowels, sonorants, obstruents English, , Cairene 7. voiced mid vowels 8. voiced low vowels (4) σσσ Rgu Rgu Rgu Sonority Sequencing Principle: µµ µµ µµ Between any member of a syllable and the syllable peak, only sounds of higher sonority 11 11 11 rank are permitted. ai al ad

Syllable Contact Law: English English English Lithuanian Lithuanian In any sequence Ca.Cb, there is a preference for Ca to exceed Cb in sonority. Khalkha Mongolian

Syllables and weight: So, it looks like there is a set of language parameters: ¥ Moraicity constraint (what sonority is required to be a mora) σσ ¥ Syllabicity constraint (what sonority is required to be a syllable nucleus) 13 µµµ It turns out: Light σ Heavy σ ¥ In any given language, the set of segments that act as syllable nuclei may or may not belong ¥ More sonorous segments are more likely to serve as syllable nuclei. to the same class as those that contribute to weight. ¥ If the two sets are not coextensive, the sonority class of the moraic set includes that of the syllabic set. ¥ mar.in ‘hand (NON-FUT)’ mar.ur. ‘hand (FUT)’ mar..ta ‘hand (uninfl.)’ That is: The sonority requirement on being a mora is never more stringent ¥ pee‰in ‘ti-tree (non-fut)’ pee‰ur. ‘ti-tree (fut)’ pee‰ ‘ti-tree (uninfl.)’ than the sonority requirement on being a syllable nucleus. Point: CVV (pee‰) satisfies minimal word requirement, has two moras. This follows from our account, though— CVC (wik) does not satisfy the minimal word requirement, has one mora. • A syllable nucleus has to also be a mora… CVS (mar.) does not satisfy the minimal word requirement, has one mora. • … we can’t have something which can be a syllable nucleus but can’t be a mora.

Some examples… Lithuanian: CVS is heavy (like CVV), CVO is light (like CV) Cairene Arabic CVC is heavy (like CVV) (9) Osthoff’s Law V  Ø / V — [l r m n] C (5) a. kátabu L«LL ‘they wrote’ b. mára÷a L«LL ‘broth’ “Long vowels become short when followed by a tautosyllabic .” c. wálad-i L«LL ‘my boy’ (relevant for Homeric Greek, but true of Lithuanian too).

(6) a. qaahíra HL«L (CVV) ‘Cairo’ (10) Syllable types in Lithuanian: a. CV, CVV b. ¿aalámu HL«L ‘his world’ b. CVO, CVVO c.. taalíba HL«L ‘student (fem.)’ c. CVS, *CVVS (7) a. falsáfa HL«L (CVC) ‘philosophy’ b. maktába ‘library’ Bimoraicity Constraint c. bus..tági ‘mailman’ A syllable may have at most two moras

(8) a. fas.s.áru HL«L (CVG) ‘they explained’ ¥ l r m n (sonorants) have moras, obstruents don’t (long vowels have two moras). b. s.al.láh.it ‘she repaired’ Lithuanian accent: Two relevant kinds acute and circumflex. Lardil CVC is light (like CV) (11) a. cvv káimas ‘village’ acute To tell:¥ Accusative/non-future form of nouns add Ðn. ¥ Future forms add Ðr.. (12) a. CVV vii˜nas ‘wine’, pei˜lis ‘knife’, zui˜kis ‘rabbit’ circumflex b. CVL gar˜sas ‘sound’, baL˜sas ‘’ c. CVN sˇim˜ tas ‘hundred’, lan˜kas ‘rainbow’ Hence: yalulun ‘flame (ACC)’ yalulur ‘flame (FUT)’: root: yalulu. . d. CVO [no cases of circumflex accent or accute on CVO]. kelan ‘beach (NON-FUT)’ kelar. ‘beach (FUT)’: root: kela e. CV [no cases of cirucmflex accent or accute on CV] wit.en ‘inside (NON-FUT)’ witer. ‘inside (FUT)’: root: wit.e So CVS seem to act like CVV and CVO seems to act like CV. ¥ But the uninflected forms are: yalul ‘flame’ kela ‘beach’ wit.e ‘inside’

¥ There is a minimal word constraint:Words must have two moras. The uninflected form loses its final only if it doesn’t result in a subminimal word.

¥ wikin ‘shade (non-fut)’ wikur. ‘shade (fut)’ wika ‘shade (uninfl.)’ [a inserted] Typology and examples (assuming the compressed sonority hierarchy) Some of my own work on the accentual systems of Mohawk and Passamaquoddy [Ðcons] vowels [+son] vowels and sonorants Mohawk, a Northern Iroquoian language, spoken mainly in New York, Ontario, and — all segments (including ) Québec. Passamaquoddy is an Eastern Algonquian language spoken in parts of Maine

Syllabicity constraint Moraicity constraint (15) A METRICALLY INVISIBLE vowel is a vowel ignored by metrical processes. a. σ µ Lithuanian, Kwakwala, Tiv 1 1 (16) Penultimate stress (normally) (MOHAWK) [Ðcons] [+son] a. r á k . w a s b. σ Cairene Arabic 2 1 1 [Ðcons] a. rákwas /hra-kw-as/ ‘He picks it’ b. wakashé:tu /wak-ashet-u/ ‘I have counted it’ c. σ English c. wakharatatuhátye /wak-haratat-u-hatye/ ‘I go along lifting up’ 1 [+son] (17) Stress skips epenthetic vowels (e below) (MOHAWK) µ d. Khalkha Mongolian, a. t é k e r i k s 1 Yindiµ, Lardil 2 1 [Ðcons] a. tékeriks /te-k-rik-s/ ‘I put them together’ e. µ Gonja b. wákeras /w-akra-s/ ‘it smells’ 1 c. t4«kerike÷ /t-4-k-rik-÷/ ‘I’ll put together side by side’ [+son] d. wa÷tkatátenake÷ /wa÷-t-k-atat-nak-÷/ ‘I scratched myself’ f. — — Imdlawn Tashlhiyt Berber (18) Unskipped epenthetic vowels (e below) (MOHAWK)

Imdlawn Tashlhiyt Berber a. w a k é n y a k s 2 1

(13) ts§.ti ‘she selected’ s§ a. wakényaks /wak-nyak-s/ ‘I get married’ u.tx§k ‘I struck you’ b. tekahsutérha÷ /te-k-ahsutr-ha÷/ ‘I splice it’ tf§.tk§t ‘you suffered a sprain.’ c. sasáhket /sa-s-ahkt/ ‘go back!’

(14) a. tr§gl§tb.tr§glas ‘lock’ (19) WHAT YOU WILL COME TO BELIEVE: ts§kr§tts§kras ‘do’ i. A metrically invisible vowel does not head a syllable. (Sec. 1-2) tx§zn§ttx§znas ‘store’ ii. Metrical invisibility is dissociated from epenthesis. (Sec. 3) tzdmttzdmas ‘gather wood’ § § § iii. Syllables headed by “weak” vowels are avoided. (Sec. 4) tl§bzˇ§ttl§bzˇas ‘step onto’ tr§ks§ttr§ksas ‘hide’ iv. “Weak” vowels have deficient underlying prosodification (Sec. 5) tn§Sf§ttn§Sfas ‘graze (skin)’ tm§sx§ttm§sxas ‘transform’ 1. Syllable-sensitive phenomena in Mohawk (26) Epenthetic e breaking up -sonorant sequence (25a) 4 4 4 4 (20) Mohawk generalizations (diagnosing syllable structure) a. « k e r ÷ / - k-r - -÷/ ‘I will put it into a container’ a. Stress falls on the penultimate syllable. (16) 4 4 4 4 b. A stressed syllable must be heavy (CV:, CVC). (21-22) a. «ker ÷ / -k-r- -÷/ ‘I will put it into a container’ c. A word must consist of two syllables. (24) b. tékeriks /te-k-rik-s/ ‘I put them together’ c. t4kahsúter4÷ /t-4-k-ahsutr-4÷/ ‘I will splice it’ (21) Vowel is long in open stressed syllable d. wákeras /w-akra-s/ ‘it smells’ e. wa÷tkatátenake÷ /wa÷-t-k-atat-nak-÷/ ‘I scratched myself’ a. w a . k a s . h é: . t u (27) And it is also invisible for minimal word requirement. a. wakashé:tu /wak-ashet-u/ ‘I have counted it’ b. khyá:tus /k-hyatu-s/ ‘I write’ a. í s e r i ht /hs-ri-ht-¿/ ‘cook!’ c. khará:tats /k-haratat-s/ ‘I am lifting it up a little (with a lever)’ a. íseriht /hs-ri-ht-¿/ ‘cook!’

(22) Vowel is not long in closed stressed syllable ¥ e is invisible for stress (20a) (antepenultimate (26a-d) preantepenultimate(26e) stress) a. w a k . h a . r a . t a . t u . h á t . y e ¥ e is invisible for syllabification (20b) (e can’t take onset ⇒ stressed σ is closed ⇒ no tonic lengthening) ¥ e is invisible for minimal word requirement (20c) a. wakharatatuhátye /wak-haratat-u-hatye/ ‘I go along lifting up’ σ ⇒ b. rákwas /hra-kw-as/ ‘He picks it’ (e doesn’t count as a i-prothesis) c. kó÷kwats /k-o÷kwat-s/ ‘I dig’ d. tékya÷ks /te-k-ya÷k-s/ ‘I break it in two’ (28) Epenthetic e breaking up word final C÷ sequence (25b) a. 4ká:ra t e ÷ /4-k-ara t-÷ / ‘I lay myself down’ (23) Conclusion: CVC and CV: are both heavy with respect to (20b). a. 4ká:rate÷ /4-k-arat-÷/ ‘I lay myself down’ (24) Prothetic i inserted to ensure a bisyllabic minimal word. b. rokú:tote÷ /ro-kut-ot-÷/ ‘he has a bump on his nose’ c. wa÷tkatátenake÷ /wa÷-t-k-atat-nak-÷/ ‘I scratched myself’ b. í : . k e k s *kéks d. t4«kerike÷ /t-4-k-rik-÷/ ‘I’ll put together side by side’ e. ónerahte÷ /o-nraht-÷/ ‘leaf’ a. íky4s /k-y4-s/ ‘I put it’ f. t4«:kehkwe÷ /t-4-k-hkw-÷/ ‘I’ll lift it’ b. í:keks /k-ek-s/ ‘I eat’ c. íkya÷ks /k-ya÷k-s/ ‘I cut it’ (29) And it is also invisible for minimal word requirement

2. Metrical invisibility: Epenthetic e in Mohawk a. í t e n e h r e ÷ /t-n-ehr-÷/ ‘you and I want’

(25) Contexts in which epenthetic e is inserted: (MOHAWK) a. ítenehre÷ /t-n-ehr-÷/ ‘you and I want’ a. Between a consonant and a single sonorant (Cen, Cer, Cew) (26) b. ítewehre÷ /t-wa-ehr-÷/ ‘you and I want to’ b. Between a consonant and a word-final glottal stop. (28) c. After a consonant when followed by a (30) ¥ e is invisible for stress (20a) (except hC and sC). (antepenultimate (28a-b) preantepenultimate (28c-f) stress) ¥ e is invisible for syllabification (20b) (e can’t take onset ⇒ stressed σ is closed ⇒ no tonic lengthening) ¥ e is invisible for minimal word requirement (20c) (e doesn’t count as a σ ⇒ i-prothesis) (30) Epenthetic e breaking up triconsonantal clusters (25c) (32) Invisible %, disrupting stress a. wa k é ny aks /wa k-ny ak-s/ ‘I get married’ a. s ó k - % l a n *sók-%«lan 2 1 a. wakényaks /wak-nyak-s/ ‘I get married’ b. sérhos /s-rho-s/ ‘you coat it with something’ a. sók-%lan ‘it pours (rain)’ c. tekahsutérha÷ /te-k-ahsutr-ha÷/ ‘I splice it’ b. w%«l-%po ‘he sits nicely, comfortably; he is well off’ d. skáhkets /s-k-ahkt-s/ ‘I got back’ c. nís-ek-%pí-si-t ‘ghost’ e. sasáhket /sa-s-ahkt/ ‘go back!’ d. tót-%po ‘he sits a long way off, he is far along’ f. tekhsà:kets /te-k-hsa÷kt-s/ ‘I bend it’ e. %pì-n ‘sit (sg.)!’

¥ e is visible for stress (33) Visible %, after a cluster, same morphemes! (always penultimate, stress on e (30a-c) or before e (30d-e)) ¥ e is visible for syllabification a. p í s k - %« l a n (e can take onset ⇒ tonic lengthening (30f), prior cluster (30d-e)) F 2 1 ¥ e is visible for minimal word requirement (20c) (e counts as a σ ⇒ no i-prothesis in (30bd) *íserhos, *ískahkets) a. písk-%«lan ‘it rains so hard that it is dark or hard to see’ b. étót-apsk-%«la-k‘it is raining very big drops’ (245) c. tékk-%«pi-t ‘as far away as he sits’ (82) Conclusion: % e inserted for syllabification ⇒ visible to all syllable-sensitive processes. d. sp-«po ‘he sits up high’ (81) e inserted for other reasons ⇒ invisible to all syllable-sensitive processes. (34) Alternating visibility of % (not after a cluster):

DEDUCTION: There’s a syllable in the first case and not in the second. a. á s % w - % c % k - %« p o F 1 2 1

% a. ás%w-%c%k-%«po ‘it (an.) is flopped over to one side’ 3. Severing invisibility from epenthesis: Passamaquoddy b. ht-%«l-%t%m-%«n-%l ‘he is eating them (inan.)’ (92) c. h-pásk-%c%k-%«n-a ‘you (sg.) break him/it (an., squishy) with your hand’ (31) Passamaquoddy stress generalization: First syllable, and even syllables counting from right to left, are stressed. Conclusion 1: Passamaquoddy % shows contextual metrical invisibility. Like Mohawk epenthetic e, % is invisible when not associated to a σ. d. s é h t á y - e w é s t o (Evidence from (32) vs. (33) and from (34)). F 2 1 2 1 Argument 1 against epenthesis: Clusters interrupted by % occur elsewhere.

a. wás-is ‘child’ (35) a. kt%«k%m%«lp%n ‘we (exc.) hit you (sg. or pl.)’ cf. (32b) w%«l%po b. l-éwésto ‘he speaks’ b. t%«lpáyo ‘he is scared’ c. wík-ewésto ‘he likes to talk’ c. kt%makp%ket ‘it is a weak liquid’ cf. (32c) nísek%písit d. séhtáy-ewésto ‘he speaks while walking backwards’ d. kp%«cále ‘he is hoarse’ e. wìkp ‘black ash’ f. piskitpohk%t ‘it is a dark night’ cf. (32d) tót%po g. tp%lokemo ‘he gossips’ Argument 2 against epenthesis: % in initial position

(36) a. %pì-n ‘sit (sg.)!’ b. épi-t ‘when he sits’ (433) (“Initial change” [Conjunct]) Argument 3 against epenthesis: %-initials pattern with underlying V-initials (44) Word-initial % deletes before a sonorant (a/b) or an (c/d) (42a) a. ht-%«lamí-ptin ‘the palm of his hand’ Step one: (putative) %-initial -%po patterns with vowel-initial ewesto b. lámí-ptin ‘palm of the hand’ c. ht-%«t%lí-na-n ‘he is dying (subordinative)’ (37) Connective i is inserted between /kis-/ ‘finished, past’ and C-initial finals d. t%«lí-ne ‘he is dying’ a. kís-ewésto ‘he talked’

b. kísí-ko ‘he is full grown’ Syllable contact constraint: ...Cσ][σR... (suggested by LeSourd (p.c.))

(38) /-%pi-/ ‘sit’ patterns with vowel-initial final (37a) (45) The pattern: A “weak vowel” will delete unless kís-%po ‘it (an.) is finished’ ¥ It is needed for syllabification (i.e. it is metrically visible) ¥ It is needed to satisfy the syllable contact constraint Step two: Connective insertion is sensitive to underlying distinctions Suppose STRAY ERASURE / PROSODIC LICENSING holds. ∴ Unprosodified material does not surface. (39) Motivating initial a of final /-ahte-/ ‘be located’ a. sákh-áhte ‘it protrudes into view’ Suppose this is responsible for syncope; b. íht%l-áhte ‘it is always there’ ∴ Weak vowels are preferentially unprosodified. % Section 2: Metrically invisible weak vowels are not dominated by a syllable. (40) Syncope of a in /-ahte-/ (when not after a cluster (39a) or invisible (39b)) ∴ a. émék-te ‘it is down below’ < / emehk-ahte-w / If prosodified, they’re dominated by a mora (weak layering) b. nís-ék-te ‘it has two layers’ < / nis-ek-ahte-w / Why the preference? ¥ Weak vowels are not underlyingly moraic (41) i insertion is sensitive to underlying form: ¥ For a weak vowel to surface, a mora must be added to the representation. kís-te ‘it is finished’ < / kis-ahte-w / (*FILL-µ discourages such additions, prefers syncope) (46) Certain (lexically marked) instances of % resist syncope before an obstruent. Conclusion 2: Passamaquoddy % is not epenthetic (at least sometimes) a. pét-ék-%po ‘it (an.) comes to be located here.’ b. w%«l-%po ‘he sits nicely, comfortably; he is well off’ c. nís-ek-%pí-si-t ‘ghost’ DEDUCTION: A vowel need not be epenthetic to show metrical invisibility. d. tót-%po ‘he sits a long way off, he is far along’ e. %pì-n ‘sit (sg.)!’ 4. Avoidance of “weak vowels”: Syncope in Passamaquoddy (47) “Syncope-resistant” % projects a mora. (42) Most %’s in Passamaquoddy only surface when they are... a. not word-initial 5. The nature of “weak” vowels and ( b. metrically visible or c. preceding a sonorant ) (48) Three behaviors of the Passamaquoddy % a. Full vowel (43) % before an obstruent, surfacing if metrically visible (a/c) else not (42b) % % b. Weak vowel which resists syncope a. áps- kíhq n ‘it is small’ c. Weak vowel which can delete before obstruents b. kín-kíhq%n ‘it is big’ (49) Underlying representations of Passamaquoddy % c. h-q%«ták%n ‘his throat’ (169) σ % % a. V—µ— Full vowel d. píl wi-qták n-e ‘his voice is changing’ (169) b. V—µ Weak vowel which resists syncope c. V Weak vowel which can delete before obstruents cf. (32a) sók-%lan ‘it pours (rain)’ (42c) Consonant preceding % acts like a syllable coda rather than an onset: (50) *STRUC There can be no prosodic structure in the output which is not contained in the input. (Prince & Smolensky 1993: 25, Zoll (54) a. /h/, /˜χ/, and diphthong + /r/ do not occur either before % or syllable-finally. 1993) They may occur before full vowels. b. /˜/ may occur before % and may occur syllable-finally. It does not occur before A lot of underlying prosodification? full vowels. c. Obstruent + clusters do not occur before %. Consider the learner in the context of OT and Lexicon Optimization (Inkelas 1994). d. Syllable final consonant clusters and consonant clusters before % undergo optional epenthesis of a very short %-like vowel. e. In Dutch dialects where /sp/ is metathesized at the end of a syllable, it is also 6. Summary metathesized before %, but not before full vowels. Mohawk data and discussion: Alderete (1995), Beatty (1974), Bonvillain (1973), Broselow (1982), WHAT YOU HAVE COME TO BELIEVE: Ikawa (1995), Michelson (1988, 1989), Piggott (1995), Pizer (1996), Postal (1969), Potter (1994). (19) i. A metrically invisible vowel does not head a syllable. (Sec. 1-2) Passamaquoddy data and discussion: LeSourd (1995), Sherwood (1986), Stowell (1979). Most ii. Metrical invisibility is dissociated from epenthesis. (Sec. 3) precise source locations are given in Hagstrom (1997). iii. Syllables headed by “weak” vowels are avoided. (Sec. 4) iv. “Weak” vowels have deficient underlying prosodification (Sec. 5) References Alderete, J. (1995). “Faithfulness to Prosodic Heads,” Ms., University of Massachusetts at Amherst. [Available via Dutch? Rutgers Optimality Archive]. Beatty, J. (1974). Mohawk Morphology. Ph. D. dissertation, City University of New York (1972). Published by Occasional Publications in Anthropology Linguistics Series no. 2, Greeley, CO: Museum of Anthropology, Dutch distinguishes short vowels (å, ´, ø, Ï, ˆ) from long vowels (a, e, o, ¿, i, y, u) and University of Northern Colorado. diphthongs (åu, ´i, Ïy) in open syllables. Bonvillain, N. (1973). A Grammar of Akwesasne Mohawk. Ethnology Division Mercury Series no. 8, Ottawa: National Museum of Man, National Museums of Canada. Short vowels cannot occur in open syllables. Broselow, Ellen. 1982. On predicting the interaction of stress and epenthesis. Glossa 16(2), 115-132. Hagstrom, Paul. Contextual metrical invisibility. In Bruening, B., Y. Kang, and M. McGinnis (eds.), PF: Papers at the interface (MITWPL 30). Cambridge, MA: MIT Working Papers in Linguistics. (51) a. *tåksi Ikawa, H. (1995). “On Stress Assignment, Vowel-Lengthening, and Epenthetic Vowels in Mohawk: Some Theoretical b. *sl´ Implications,” Handout from SWOT. c. taksi ‘taxi’ Inkelas, Sharon. 1994. The consequences of optimization for underspecification. Ms., University of California at d. sle ‘sledge’ Berkeley. [Available via Rutgers Optimality Archive]. e. b4y ‘rain shower’ Itô, J. and R. A. Mester (1992). “Weak Layering and Word Binarity,” UCSC ms. f. χal i ‘galley’ LeSourd, Philip. 1993. Accent and syllable structure in Passamaquoddy. New York: Garland. ´ Michelson, Karin. 1988. A comparative study of Lake-Iroquoian accent, Dordrecht: Kluwer. Michelson, Karin. 1989. Invisibility: Vowels without a timing slot in Mohawk. Theoretical perspectives on Native (52) BIMORAIC CONSTRAINT Syllables dominate at least two moras. American languages, ed. by Donna Gerdts and Karin Michelson. Albany: SUNY Press. Piggott, Glynne. 1995. Epenthesis and syllable weight. Natural Language and Linguistic Theory 13:283-326. % is different: phonetically a short vowel, but exempt from bimoriacity constraint. Pizer, K. (1996). “Epenthesis and (non)Moraicity: the case of Mohawk Weightless [e],” paper presented at LSA 1997. Postal, P. (1969). “Mohawk vowel doubling,” International Journal of American Linguistics 35(4):291-298. (53) a. *mikå Potter, Brian. 1994. Serial optimality in Mohawk . CLS 30. Prince, A. and P. Smolensky (1993). . Forthcoming from Cambridge, MA: MIT Press. b. *hˆndˆ Selkirk, E. (1984). Phonology and Syntax, Cambridge, MA: MIT Press c. mika ‘mica’ Sherwood, David. 1986. Maliseet-Passamaquoddy verb morphology. Canadian Ethnology Service Paper no. 105. d. hˆndi ‘Hindi’ Ottawa: Canadian Museum of Civilization e. mik% (name) Sherwood, D. (1986). Maliseet-Passamaquoddy Verb Morphology. Canadian Ethnology Service Paper no. 105. f. hˆnd% ‘hind’ Ottawa: Canadian Museum of Civilization Steriade, D. (1982). Greek Prosodies and the Nature of Syllabification, Ph.D. dissertation, MIT. Stowell, T. (1979). “Stress systems of the world, unite!” MIT Working Papers in Linguistics 1. Cambridge, MA: MIT Working Papers in Linguistics. Zoll, Cheryl (1993). Directionless syllabification and ghosts in Yawelmani. ROW-1 [available on ROA].