Phonological Degrees of Labiality: Evidence from Karata Jeremy C. Pasquereau Department of Linguistics, University of Massachusetts

INTRODUCTION ASSUMPTIONS ANALYSIS PREDICTIONS

Karata (k'irʟ̄ī mac̄'i, Russian karatinskij jazyk): aperture Constraints Rounding harmony Nakh-Daghestanian family, ca. 10,000 speakers, 8 • V height is defined in terms of degrees of aperture • Anti-mismatch constraint • Parasitic rounding harmony villages, North-West Daghestan, Russia. *apertureα (Clements and Hume 1995) Assign a violation for every output rootn- Anti-mismatch >> Extend(labialx) • `Aperture' refers to the degree to which the oral cavity x ode which is linked to an aperture node labial¬α and a labial feature which disagree in • (Yokutsan): high-high low-low Background information is open/close value. • • u Kachin Khakass (Turkic): high-high 86 phonemic C (46 non-labialized C & 40 labialized C) • constraint aperture 3 i u vocalic • 5 V (with long x nasal counterparts) labial degrees are contrastive aperture 2 e o * ON Assign a violation for every output onset- • V-place aperture3 nucleus sequence if their V-place nodes • Phonotactic constraints: VαV¬α Cross-height rounding harmony w aperture 1 a agrees 1 no [round V + labialized C] sequence (*uC ) are not the same. w w [dorsal] [labial3] 2 no consecutive labialized onsets (*C1 C2 ) Extend(labialx) >> Anti-mismatch • *FLOAT Assign a violation for every feature that • The UR of the verb stem is apparent with the singular Labiality degrees and scale is floating in the output (Wolf 2007). • Turkish: low/high-high • • Neutral Class Marker b- (This is the only CM that does There are degrees of rounding too. • MAX[labial3] >> MAX[labial2] Yakut (Turkic): low-low/high high-high not neutralize the following V contrast). • In this (and others), if a V has a [labial] labial degrees are not contrastive • MAX[labx] Assign a violation for every [labialx] fea- Underlyingly non-labialized C's undergo a process of feature, it agrees in degree with the aperture node of ture that is not present in the output. (data and Extend from Kaun 1997) the same V. • Anti-labialization constraint • Vowel Coalescence Height and backness = rounding enhancement IdentOI[labx] Assign a violation for every link between labialization a [labialx] feature and a segment in the • +round labial3 In Yoruba (Niger-Congo), under coalescence, the • output that is not in the input. In Karata verbs, a stem C must be labialized when both >> rounding of a high vowel is preserved but not that of a following conditions are met: mid-vowel: /u + i/ → [u] BUT /o + e/ → [e] 1 the preceding V is underlyingly round and high labial2 Analysis (simplified) (Casali 1996) ¬ ¬ ✓b-uʧãa b-oχ̄aa /j+uč / *OαN αP LACE Max[labial3] *Float *apα/labial α IdentIO[labialx] Max[labial2] >> [lab3] Aperture/labial Mismatches N-wash N-appear w (Lionnet 2013) -round labial1 a j ič * 2 this V becomes unround as a result of assimilation to a [lab3] • In where *apα/lab¬α is low ranked, b j ič *W L preceding CM: Fem.CM j-, H.pl.CM b(a)- or nH.pl.CM r(a)- • [lab3] mismatches are allowed (e.g. cross-height rounding w The roundedness scale maps vowel height to labiality j-iʧ ãa j-eχ̄aa c j ič *W L degree. [lab3] harmony) F-wash F-appear d juč *W L • • `The faithfulness constraints are naturally ranked so [lab3] Is the gesture different from a round vowel with ¬ ¬ that a more prominent member of the scale demands /j+oč / *OαN αP LACE Max[labial3] *Float *apα/labial α IdentIO[labialx] Max[labial2] matching aperture and labiality degrees? [lab2] w more faithfulness.' (Gnanadesikan 1997) a ječ *W L CM Verb stem [lab2] b ječ *W L [lab2] c ječ * [+high] [lab2] CONCLUSION EVIDENCE FOR LABIALITY DEGREES d joč *W L j [V1 C2V2 ] [lab2] = • Articulatory evidence (Linker 1982): the rounding Conclusion [palatal] [labial] gesture of a high vowel is more extreme than the • Karata consonant labialization rounding gesture of a non-high vowel. 1 [labialx] features agree in value with V aperture ANTICIPATING QUESTIONS 2 there is more faithfulness to keep [labial3] in the output • Perceptual evidence (Terbeek 1977): high rounded • • high i u are perceived as more rounded than non-high Phonology can access gradient subphonemic Karata has phonologized distinctions that are mid e o rounded vowels. information traditionally considered purely phonetic • • low a • Typology (Maddieson 1984): the lower the rounded Is this information phonetic or phonological? This helps understand phenomena of parasitic vowel, the rarer it is. rounding harmony → • (1) high/round V1 C labialization • Contrastive labiality gestures: Claim: it is phonological. Future research: → w 1 a. /j + uʧãa/ jiʧ ãŋa `wash' 1 Swedish: 2 high front round V [yː, ʉ], [ʉ] is not as protruded Karata and languages with parasitic rounding harmony Is the labial gesture on a V whose rounding results from b. /j + uʧ'ãa/ → jiʧ'wãŋa `open' (Linker 1982) have phonologized phonetic reflexes. harmony different from the gesture on a V which is underlyingly → w 2 Assamese: 2 low back round V [ɑ,ɒ̹] (Ladefoged et al. 1996) round? c. /j + utaa/ jit aa `untie' 2 Do processes of V coalescence bear out the fixed-ranking of → w d. /j + uʟ̥̄'ãa/ jiʟ̥̄' ãŋa `share' • The degree of rounding is predictable from V height: is faithfulness to [labialx]? it necessary to have degrees of rounding on [labial]? (2) No labialization 1 → On the surface, rounding is not on V anymore so we can't refer Acknowledgements a. /j + oχ̄aːa/ jeχ̄aːa `thrust' SKETCH OF THE ANALYSIS to V height I thank my Karata informants (esp. Rashidat Khalidova). Research b. /j + oq̄aa/ → jeq̄aa `remove' 2 If at an intermediate level [labial] is linked to both V C, then reported here was supported in part by the National Science Founda- • c. /j + oχ̄aa/ → jeχ̄aa `appear' When [labial3] and [labial2] are delinked as a result of it's a case of opaque spreading tion under Grant Number BCS 0745522 to Alice Harris. d. /j + oʟ̥̄'aa/ → jeʟ̥̄'aa `warm up' assimilation, there is markedness pressure for delinked • Why not just stipulate that [+high] V trigger opaque features to delete. spreading? • ̄ • 1 At an intermediate level of the derivation, we would need C = fortis consonant But there is more pressure to keep [labial3] than the w /uC / to win Contact Information • w fortis = used across all art. places other [labial x] features. 2 But this won't happen because *uC is undominated • • • It is better to re-associate [labial3] to another segment Web: http://blogs.umass.edu/jpasquer/ phonetic realization contingent on the type of C that (see Pasquereau 2013 for more details) realizes it than to delete it. • Email: [email protected]