What is an ‘onsetless’ syllable? Geoff Schwartz - geoff@ifa.amu.edu.pl UAM - Poznań HAVE YOU EVER WONDERED. . . • why some vowel-initial syllables are prosodically inert while others are not? • why onsetless vowels in KiKirewe (Odden 1995, 1996) cannot bear tone or undergo compensatory lengthening, but are still copied into reduplicated forms? • how we can reconcile VC syllabification in Eastern Arrernte (Breen and Pensalfini 1999) with a seemingly universal constraint for consonantal onsets? • how constituent boundaries may best be represented? • if we can get rid of ONSET and ALIGN constraints? If the answer to any of these questions is YES, then read on . . . Sound Bites • The Onset Prominence environment offers a unified explanation of diverse prosodic processes • Segments and prosodic constituents are built from the same hierarchical structure • Some initial vowels contain a built-in ‘onset’ (the VO node), some do not • Well-formed syllables (e.g. initial vowels in Kikirewe) resist CL – no reason for compensation • VO parameters capture inherent phonological ambiguities 1. The problems  In most cases, the presence or absence of a consonantal “onset” is irrelevant for prosody.  Occasionally, this generalization does not hold. Vowel-initial syllables may be invisible for….  Stress assignment (e.g. Topintzi 2010)  Reduplication (e.g. McCarthy and Prince 1993)  Tone (e.g. Downing 1998)  Compensatory lengthening (e.g. Odden 1995)  ONSET is claimed to be universal, but . . .  It’s unclear how it should be formulated (Smith 2010)  Eastern Arrernte (Breen and Pensalfini 1999) appears to show only VC syllabification  Pre-vocalic glides show ambiguous behaviour with regard to Onset-Rhyme distinction (Yip 2003)  Word and phrase-initial positions are notorious for ONSET violations, necessitating Alignment constraints in OT 2. The Onset Prominence framework • Prosodic constituents and segmental representations emerge from a single hierarchical structure • The hierarchy represents the sequence of phonetic events observable in initial stop-vowel sequences • What is a segment’s underlying level in the ‘onset’ hierarchy?  Stops and nasals at the Closure level  Fricatives at the Noise level  Most sonorants at Vocalic Onset (lots of ambiguity and variation here)  Vowels at the bottom under the Rhyme • Place (and laryngeal) specifications assigned at terminal nodes • To form constituents, lower level structures are absorbed into higher level trees. Here we see quick (the final /k/ is submerged (see below)) • The Vocalic Onset level (VO) is a source of representational ambiguity 3. VO and minimal constituents • The VO node is ambiguous with respect to segmental affiliation • It may attach to higher level obstruents (C-place cues) • It may be joined with rhymal vowels (a ‘built-in onset’) • This creates parameters for the representation of initial vowels • The Rhymal vowel, without VO specification, is ill-formed • Instead of ONSET , a MINIMALCONSTITUENT (MC) constraint • MC - A prosodically active constituent must contain VO in addition to Rhymal layers of structure • In Timugon Murut (McCarthy and Prince 1993) onsetless syllables are not copied into the reduplicated forms. First MC is copied limo > li-limo 'five/ about five' abalan > a-ba-balan 'bathes/often bathes‘ • ONSET and ALIGN are superfluous in this environment • The same analysis works for IsiXhosa reduplication (Slide 1a) • For onset-sensitive stress (e.g. Topintzi 2010) there is a handout 4. To lengthen or not to lengthen? KiKirewe (Odden 1995, 1996) sends apparently conflicting messages about the status of onsetless syllables 1. Only initial onsetless vowels allowed. Glides at vowel hiatus ebi-tooke ‘bananas’ ebi-a−la > eby-aala ‘fingers’ omu-tima ‘heart’ omu-aga > omw-aaga ‘compulsion’ 2. Onsetless vowels copied in reduplicated forms udan udan-udɛn ‘rain’ kumat kumat-kumɛt ‘have a relapse’ 3. Compensatory lengthening accompanies glide formation, except when the first vowel is onsetless tu-tká ‘we cook (hab.)’ tw-aa-tká ‘we just cooked’ o-tká ‘you sg. cook (hab.)’ wa-tká ‘you sg. just cooked’ 4. Onsetless vowels cannot bear high tone (Slide 4a) From the OP perspective, there is no conflict • With VO-specified initial vowels, absorption occurs normally (left) with no lengthening. No reason to compensate for a well-formed constituent • Glide forms when the first vowel’s melody is raised to VO (left and right) • Lengthening results from submersion (center and right) 2a. OP Phonotactics • The OP hierarchy captures sonority-based generalizations without sonority (Ohala 1992). Below, constituent formation in English scrape • Lower-level structures are absorbed The /r/ and the /eɪ/ in scrape • Stranded trees may be submerged (the final /p/) or adjoined (the /s/) 1a. Ambiguity in prosodic constituency •In IsiXhosa (Downing 1998), onsetless vowels are invisible for reduplication but optionally invisible for tone assignment (Slides 3 and 4a). Why? • Tone expected on antepenult in 4 syllable stems, otherwise on penult ndi-y-onwabsa ~ ndi-y-onwbisa ‘make happy’ ndi-y-o-nwabi-nwabsa ~ ndi-y-o-nwabi-nwbisa •In KiKirewe (Odden 1995, 1996) onsetless vowels do not undergo compensatory lengthening, but do get copied into reduplicated forms (Slide 4). Why? •In Tashlhiyt Berber versification (Dell and Elmedlaoui 1988), falling sonority syllables (e.g. /nb/) are onsetless and heavy in phrase-initial position, but contain onsets and are light elsewhere. Why? •In English (Barlow 2001) Mandarin (Yip 2003) and Korean (Smith 2010) prevocalic glides sometimes behave as consonantal onsets, and sometimes pattern with rhymes. Why? 1b. Why a new theory of representation? The examples in 1a (and many others) show that phonology, like the speech signal, is full of ambiguities How can we represent such ambiguities in a principled way? I claim that any theory in which prosodic constituents translate into a linear string of segments is ill-equipped for this task But what if segments themselves are made from the same hierarchical structures as prosodic constituents? This idea is not new (Steriade 1993, Pöchtrager 2006) but its implications warrant more thorough investigation 2b. Restrictions on absorption • The OP environment also captures other constraints on ‘onset’ clusters • Place-based restrictions serve to block absorption to avoid ambiguity within a constituent. Both /t/ and /l/ can show a high F3 (Stevens 1998) so /tl/ is ambiguous. Below we see a prohibited *tl cluster, something akin to an OCP violation • A more restrictive constraint, something akin to *COMPLEXONSET , bans absorption and consonant clusters • Functionally, such a restriction serves to lessen the amount of linguistic information at the VO level of a single constituent, reducing ambiguity. In other words, we see that the OCP and *COMPLEX are functionally equivalent • These restrictions may be the source of syllabic consonants in Tashlhiyt Berber (Schwartz, under review) • VC syllabification in Eastern Arrernte has similar motivation (see 5) 3a. VO and perception • A classic issue in speech perception is the lack of linearity in speech • a single portion of the acoustic signal may contain cues to multiple segments • a single segment’s cues may span an extended portion of the signal • OP incorporates these ambiguities into phonological representations • The VO node may be claimed by consonants, vowels, or both • VO houses formant CV formant transitions for consonant identification • Consonant representations split into ‘internal’ and ‘transitional’ cues (Wright 2004) • Vowel intrusion restores consonantal structure and perceptibility • VO-specified initial vowels are both ‘onsets’ and ‘peaks’ . • Initial amplitude rise may produce ‘onset boost’, then auditory saturation later in the vowel (e.g. Wright 2004:44) • The VO layer may be fortified in many different ways • Glottal stop insertion, laryngealization, gliding, F0 effects 4a. VO, tone, and boundaries • In IsiXhosa, onsetless syllables may or may not be counted in the tone assignment algorithm (Slide 1a). The reduplication facts suggest these vowels lack VO • In KiKirewe, onsetless syllables cannot bear high tone, yet we claim they are well-formed VO-specified constituents Why should high tone be blocked from well-formed constituents? Initial vowels with VO constitute a constituent boundary and should be candidates for markers such as glottalization (Dilley et al 1996) and lowered F0 (Hillenbrand and Houlde1996) Tone looks at rhymes, not boundaries 3b. VO parameters and morphology • Representational ambiguity at the VO level also helps define the domains of phonological and morphological processes. • Consider the following data from Polish kot ‘cat’ kotek ‘little cat’ kotka ‘little cat (gen)’ kotɕe (orth. kocie) ‘cat (loc-voc)’ • The locative ending –e (left) contains VO (Schwartz, under review) and mutates the /t/ • The diminutive –ek (right) does not affect the /t/. It is Rhymal. It also alternates with zero, and word-initial vowel-zero alternations are impossible. References Barlow. J. 2001. Individual differences in the production of initial consonant sequences in Pig Latin. Lingua 111. 667-696. Breen, G. and R. Pensalfini. 1999. Arrernte – a language with no syllable onsets. Linguistic Inquiry 30. 1-25 Clements, G.N. 1997. Berber syllabification: derivations or constraints. In: I. Roca (ed). Derivations and constraints in phonology. 289-330. Dell, F. and M. Elmedlaoui. 1988. Syllabic consonants in Berber: some new evidence. Journal of African Languages and Linguistics 7. 105-130. Dilley, L. Shattuck-Hufnagel, S. and Ostendorf, M. 1996. Glottalization of word-initial vowels as a function of prosodic structure. JPhon 24. 423-444. Dogil, G. The phonetic manifestation of word stress in Polish, Lithuanian, Spanish and German. In H. van der Hulst (ed): Word Prosodic Systems in the Languages of Europe. Mouton. 273-311. Downing, L. 1998. On the prosodic misalignment of onsetless syllables. NLLT 16. 1-52. Hillenbrand, J. and R. Houlde. 1996. The role of F0 and amplitude in the perception of intervocalic glottal stops. Journal of Speech and Hearing Research. 39. 1182-1190. Ladefoged, P. and I. Maddieson. 1996. The Sounds of the World’s Languages. Blackwell Publishers. Levi, S. 2008. Phonemic vs. derived glides. Lingua 118. 1956-1978. McCarthy, J. and A. Prince. 1993. Prosodic Morphology – Constraint interaction and satisfaction. Ms. U. of Massachusetts at Amherst. Odden, D. 1995. Onsetless syllables in KiKirewe. OSUWPL 47. 89-110. Odden, D. 1996. Patterns of reduplication in KiKirewe. OSUWPL 48. 111-149. Ohala, J. 1990. The phonetics and phonology of assimilation. In J.Kingston & M. Beckman, (eds). Papers in Laboratory Phonology 1. 258-275. Ohala, J. 1992. Alternatives to the sonority hierarchy for explaining the shape of morphemes. Papers from the parasession on the syllable. CLS. 319-338. Pöchtrager. M. 2006. The structure of length. Ph.D. dissertation. University of Vienna. Rubach, J. and G. Booij. 1990. Syllable structure assignment in Polish. Phonology 7. 121-158. Scheer. T. 2008. Why the Prosodic Hierarchy is a diacritic and why the Interface must be Direct. http://www.unice.fr/dsl/tobweb/papers.htm. Schwartz, G. 2010. Auditory representations and the structures of GP 2.0. Acta Linguistic Hungarica. Schwartz, G. under review. Tashlhiyt Berber syllabification within the Onset Prominence representational environment. Schwartz, G. under review. Polish palatalization within the Onset Prominence representational environment. Schwartz, G. in press. Initial glottalization and final devoicing in Polish English. Smith 2010. The formal definition of the Onset constraint and implications for Korean syllable structure. http://www.unc.edu/~jlsmith/home/pubs.html Steriade, D. 1993. Closure, release, and nasal contours. In. M. Huffman and R. Krakow (eds) Nasals, nasalization, and the velum. San Diego: Academic Press. Stevens, K. 1998. Acoustic Phonetics. MIT Press. Tabain, M., G. Breen, and A. Butcher. 2004. VC vs. CV syllables – a comparison of Aboriginal languages with English. JIPA 34. 175-200. Topintzi, N. 2010. Onsets – suprasegmental and prosodic behaviour. Cambridge University Press. Wright, R. 2004. Perceptual cue robustness and phonotactic constraints. In Hayes et al (eds.) Phonetically Based Phonology. CUP. 34-57. Yip. M. 2003. Casting doubt on the Onset-Rime distinction. Lingua 113. 779-816. How to read this poster! – If you want a quick version, just look at the top half, proceeding across the BABY BLUE slides. – If you want the whole story, go vertically down each column, from BABY BLUE to SALMON to GREEN. 5. Built-in ‘onsets’ and syllabification 1. Tashlhiyt Berber requires consonantal onsets except in phrase-initial position The need to allow initial ONSET violations leads to a faulty analysis of syllabification falling sonority syllables (Clements 1997) We suggest that the VO layer constitutes something like a ‘built-in onset’. In other words, initial vowels act as both ‘onset’ and ‘peak’ Here is Tashlhiyt i.xng ‘he has strangled’ 2. The ‘built-in onset’ can also explain VC syllabification in Eastern Arrernte EA is unusual in that the standard absorption mechanism is disallowed If both vowels and consonants are specified for VO, the Arrernte pattern can be attributed to a constraint on the merger of two VO nodes (cf. *COMPLEXONS) Below is the string of segmental structures for itirem ‘thinking’ 5b. Initial /i/ vs /ji/ • Assuming a built-in onset on VO-specified initial vowels, how do we represent a contrast between initial /i/ and /ji/? • This contrast is rare for perceptual reasons (e.g. Ohala 1990) • It should be attributed to ‘underlying’ rather than ‘derived’ glides (Levi 2008). • To represent the contrast, we need one of two possible annotations on VO • /j/ specified for transition from a more peripheral position than /i/ • /i/ specified for more rapid amplitude rise than /j/ 4b. VO as sandhi-blocker • VO-specification on onsetless vowels defines a boundary • Initial vowels without VO are absorbed into preceding consonants (English liaison, French enchaînement) • Initial vowels with VO are likely more targets for glottalization, blocking these sandhi phenomena • Polish learners of English tend to glottalize initial vowels in sequences such as George often (Schwartz, In press) • The conclusion: Polish initial vowels are specified with VO, and are more likely to be glottalized • Some corollaries: no resyllabification across word boundaries in Polish (Rubach and Booij 1990). Secondary stress on initial syllables in Polish (Dogil 1999 5a. VO promotion • A “syllabified” representation of EA itirem is given below • The segments in each syllable are adjoined at a higher level of structure • Vocalic structures are promoted to the level of stop Closures • This raises VC transitions in the Rhyme to a higher level of structure • Phonetic evidence (Tabain et al 2004) suggests that in Aboriginal langauges, VC transitions are on a par with CV transitions in terms of acoustic stability • This is why Arrernte and related languages can support such rich place contrasts (Ladefoged and Maddieson 1996) • The representation above captures this insight • A similar promotion mechanism may be responsible for glottalization (slide 4b) and rich cluster phonotactics in Polish /rʃq/ *P/t ONSET E r .ʃq *! * F* r.ʃq **