Does Turkish diss harmony?*
Markus A. Pöchtrager
Boğaziçi Üniversitesi, İstanbul, Turkey
Background. In Government Phonology (GP; Kaye, Lowenstamm & Vergnaud 1985, 1990; Kaye 1990), Turkish Vowel Harmony (TVH) has received considerable interest (Charette & Göksel 1994, 1996; Denwood 1997; Ploch 1998). TVH is usually understood as an element from the harmonic head spreading into a recessive position, cf. (1). Recessive positions (in the stem or suffixes) contain either nothing or the element A underlyingly. I and U spread to the right, but U does not combine with A. Unlicensed empty nuclei are realised as ı [i].
(1)
il-ler ‘country-PL’ A I>>> |
inek-ler ‘cow-PL’ A A I>>>>> |
ok-u ‘arrow-ACC’ A U>> |
ok-lar ‘arrow-PL’ A A U |
Problem. The Minimality Principle of GP states that processes apply whenever their conditions are met (Kaye 1992: 141). Thus, any element I or U must spread to the right. Furthermore, since recessive positions only contain A or nothing, any I/U in them must come from the left. Disharmonic words (DWs) pose a problem: In (3a), the harmonic head contains an I/U failing to spread. In (3b), the second nucleus contains an I/U not coming from the left.
(3) a. bira ‘beer’, mühim ‘important’ b. hafif ‘light’, arzu ‘wish’
Kaye (p.c.) proposes that DWs have a compound structure, i.e. [[bi][ra]]. This explains the existence of DWs as well as irregular stress in bira, whose (initial) stress pattern is like that of compounds. Attractive as that account might seem, it makes three problematic predictions: (P1) The individual parts of such dummy compounds should always be big enough to be domains in their own right, (P2) stress should be irregular for all DWs and (P3) DWs should allow for any vowel combination. None of those predictions is borne out, however.
Solution. We leave aside P2 for now (stress not being an issue for my proposal) and begin with P3: Close inspection of the occurring DWs reveals that empty nuclei are always harmonised. They never lead to disharmony, i.e. ı (the realisation of an empty nucleus) never follows any other vowel but a or ı (which is harmonic). This provides the crucial clue.
Instead of claiming that DWs are compounds, I propose that (C1) recessive positions can contain any combination of elements (instead of only nothing or A) and that (C2) spreading only occurs into empty-headed positions. Claim C1 immediately explains why we find words like hafif, arzu etc. It also makes compound morphology unnecessary and thus avoids P1. C2 explains why empty nuclei are always harmonised, i.e. why harmony never disrespects ı (avoiding P3): ı is the realisation of an empty nucleus (empty-headed by definition), and if there is an I or U in the harmonic head, they will spread into that empty position.
C2 seems problematic for bira vs. il-ler. In bira, the second nucleus is lexically filled (with A) and thus, no material can spread into it. The nucleus of -lar also contains A, yet here we do have spreading (il-ler, *il-lar). Crucially, the a in bira and the a in -lar are phonetically different (usually not mentioned in accounts of TVH). I submit that the a in bira is a headed by A, while -lar has an empty-headed A. The empty nucleus thus forms a natural class with unheaded A: both are headless expressions and thus, both undergo TVH by C2. Harmonic and disharmonic words find a unified explanation.
I conclude with a discussion of differences between recessive positions in stems and suffixes (improving on C1), the constraints generating the Turkish vowel system and the interpretation of all this within GP 2.0 (this workshop).
-----------------------------------
* Thanks to hip hop music for creating the slang clipping to dis(s) from to disrespect.