CV Metathesis in Kwara’ae: Implications for a Theory of Stress Jeffrey Heinz University of California, Los Angeles jheinz@humnet.ucla.edu In metrical theory, the asymmetries observed between iambic languages and trochaic languages– such as iambic lengthening versus trochaic shortening– are derived from the optimal foot types for each kind of system (Prince 1992, Hayes 1994). In iambic languages, the second L in a LL sequence is often lengthened because LH is the optimal iamb. Similarly, in trochaic languages, the H in a HL sequence is often shortened because LL is the optimal trochee. On the other hand, iambic shortening and trochaic lengthening are predicted not to exist. Although there are no iambic shortening languages, there are some languages which lengthen stressed syllables in trochees. Trochaic lengthening however is not as regular nor as robust as iambic lengthening (Hayes 1994), and researchers have attributed this to other factors (Prince 1992, Hayes 1994). It comes as a surprise then in Kwara’ae, an Austronesian language with trochaic feet spoken in the Solomon Islands, possible ["C1 V1 C2 V2 ] sequences are regularly dispreferred to ["C1 V1 V2 C2 ] sequences. Although a single heavy syllable is a possible foot in trochaic languages, it is not optimal so the current theory fails to explain why a single heavy syllable is preferred to two adjacent light syllables. The analysis proposed here is that the Stress to Weight Principle supplies the motivation; this solution brings a new perspective to how we can understand the asymmetries between iambic and trochaic languages. In Kwara’ae, words have as many as three allomorphs, which are related by CV metathesis. Two of these allomorphs occur in different speech registers, the Citation and Normal forms; the third is discussed further below. There are often multiple occurrences of metathesis in the Normal form as shown in (1), with regular exceptions in CVV syllables as shown in (2). (1) Normal "Neal ‘child’ "keat.lauk ‘my height’ r.Pa.niEd "dao ‘to share them’ "raeP.raeP.na;P ‘incline, slope’ (2) Citation Normal "mae.la.ku "mae.lauk ‘my death’ "ta.Pi.hau "teiP.hau ‘all’ li."mau.mu.lu "li.mau.mu;l ‘your (pl) hands’ Blevins and Garrett’s (1998) diachronic study hypothesized that CV metathesis results from extreme vowel to vowel coarticulation, followed by lenition of the unstressed vowel "C1 V1 C2 V2 > "C1 V1 V2 C2 V2 > "C1 V1 V2 C2 . I present new data to support the view that metathesis is a synchronic process which targets unstressed syllables to make stressed syllables heavy. Additionally, previous synchronic accounts (Sohn 1980, Norquest 2001, Baird 2002) do not address the range of data covered here. There are three reasons why a synchronic account of metathesis in Kwara’ae is necessary. First, every word in the language has both a Citation form and a Normal form, including morphologically related words. (3) Citation "su.li "Po.so Normal "suil "PO;s Citation su."li.ku Po."so.Pa Citation "Ne.la "ke.ta.la.ku da."ro.Pa.ni.da "ra.Pe.ra.Pe.na.Pa
Normal ‘bone’ "su.liuk ‘my bone’ ‘a lie’ "Po.soaP ‘guile’ Second, loanwords also have undergone this transformation, an indicator of productivity.
Normal Citation Normal (4) "biEt ‘Peter’ "ha.re "haer ‘Harry’ "reas ‘razor’ "be.ba "beab ‘paper’ Third, C1 V1 C2 V2 sequences are absent in Normal surface forms. This fact must be accounted for by the grammar since C1 V1 C2 V2 sequences are legal surface forms in the Citation register, not to mention most other languages. One way to account for this is via the principle of a rich base; the grammar must transform underlying forms like /C1 V1 C2 V2 / into legal surface forms. I propose that metathesis in the Normal form is driven by the Stress to Weight Principle (SWP), which, along with Max, outranks Linearity. This means that ["Neal] *["Ne.la] because the grammar prefers changing the underlying order to having a stressed light syllable on the surface. This markedness constraint does, however, overestimate the loci of metathesis. x x In words with an odd number of moras, stress follows an initial dactyl pattern, xx x xx, which can result in a stressed light syllable followed by an unstressed heavy syllable; i.e. ["li.mau.mul] ‘your (pl) hands’, not *["liE.mu.mul], which is predicted by the current constraint ranking. Since it appears that CV metathesis may create VV clusters, but not destroy them, I propose VVContig which requires that contiguous vowels in the input be contiguous in the output. The ranking VV-Contig SWP Linearity, in addition to the stress pattern, predict exactly the loci of metathesis. An analysis along these lines predicts that a different stress pattern should yield a different metathesis pattern. Confirmation comes from recent work with a native speaker which has revealed a third allomorph in Normal discourse, found at the right boundary of a clefted phrase. (5) ki.ra so.N> leaP [na "Pih.t> ]. eIP > eIP 3p make well the bed ‘They skillfully built the bed.’ Examples of these allomorphs are given below. (7) Citation "Pi.hi.te.Pi "ku.lu "si.na "bu.lu.bu.lu Normal "Pi;h.teiP "ku;l "siEn "bu;l.bu;l Normal]clef t Pih.tei."Pi ku;."lu siE."na bul.bu;."lu (6) [na Pih.t> ] nEP kj@r so.N> leaP an. eI."Pi eIP > the bed that they make well P ‘It is the bed that they skillfully built.’
Citation "bi.ta "re.sa
‘bed’ ‘heavy’ ‘sun’ ‘star’
The third allomorph, whose surface form resembles Blevins and Garrett’s middle step, is also straightforwardly accounted for. There is prosodic prominence (i.e. phrasal stress) on the final mora. I propose FinalStress which assigns a violation if a vowel with primary stress does not immediately precede the right focal boundary. Output to Output faithfulness with the Normal form OO-VV-Contiguity (>> Integrity) ensures that underlying /Pifite1 Pi2 ]f ocus / returns [Pih.te1 i2 ."Pi2 ] and not *[Pih.te1 ."Pi2 ]. Syncope in Tonkawa has been analyzed with the Stress to Weight Principle as well; i.e. ["C1 V1 C2 V2 ] is less harmonic than ["C1 V1 C2 ] so that stressed syllables can be heavy (Gouskova 2003). The connection made here is that Kwara’ae and Tonkawa respond in different ways to the same marked conditions; i.e. CV metathesis differs from vowel deletion only in that it is information-preserving because no segment is lost. The Stress to Weight Principle not only highlights the similarities between these two languages but also relates them to cases of iambic and trochaic lengthening. Under this approach, no longer is trochaic lengthening an anomaly; rather it belongs to the class of expected changes that included iambic lengthening, syncope in Tonkawa, and CV metathesis in Kwara’ae. What remains to be explained is why trochaic lengthening is less common than iambic lengthening. One reason might be that HL is more marked than the other foot types.
References Baird, Paula. 2002. A Re-Analysis of Metathesis in Kwara’ae. MIT Working Papers in Linguistics 44, 17-32. Blevins, Juliett and Garett, Andrew. 1998. The Origins of Consonant-Vowel Metathesis. Language 74, no. 3. 508-556. Gouskova, Maria. 2003. Deriving Economy: Syncope in Optimality Theory. University of Massachusetts, Amherst, dissertation. Hayes, Bruce. 1994. Metrical Stress Theory. Chicago University Press. Norquest, Peter. 2001. The Collapse of the Foot in Oceanic. Proceedings from the Western Conference of Linguistics (WECOL). Prince, Alan. 1992. Quantitative Consequences of Rhythmic Organization. CLS 26:355-398. Parasession of the Syllable in Phonetics and Phonology. Sohn, Ho-Min. 1980. Metathesis in Kwara’ae. Lingua 52. 305-323.