Hungarian has a phonetically voiced and a ‘plain’ voiceless series of obstruents. The two sets behave in a perfectly symmetrical fashion in voicing assimilation: in an obstruent cluster, all members assimilate to the rightmost obstruent regardless of whether it is voiced or not. Furthermore, there is no positional (syllable-final or word-final) neutralisation of voicing in the language. It has therefore been assumed (for instance by Szigetvári 1998a,b) that the voicing contrast in Hungarian is marked by the element L: voiced obstruents contain this element while voiceless ones do not.

 

The behaviour of the sounds /j/ and /h/ deviates from the pattern described above. /j/ is realised as a palatal approximant except when preceded by a consonant and followed by a pause, in which case it surfaces as a palatal fricative. When the consonant preceding the palatal fricative is an obstruent, /j/ undergoes progressive assimilation: it is voiced after voiced obstruents (dobj ‘throw Imp.’) and voiceless after voiceless ones (kapj ‘get Imp.’). When preceded by a sonorant, the palatal fricative is voiced (férj ‘husband’).

 

/h/ is realised as a velar fricative in codas, and it remains voiceless even if followed by a voiced obstruent, creating otherwise unattested clusters of obstruents not agreeing in voicing (fa[xb]a ‘into the pigeon hole’). Additionally, the glottal allophone of /h/ causes the devoicing of preceding obstruents in some dialects (a[d] ‘give’ – a[th]at ‘can give’) but not in others (a[dh]at ‘can give’).

 

I present a unified analysis of the apparently irregular cases of voicing assimilation described above. My solution crucially relies on the following assumptions:

1. Laryngeal elements are linked to the element h in obstruents, not directly to the skeletal slot (Szigetvári 1998b).

2. Voicing assimilation involves the spreading of h (and its dependants), not the spreading of laryngeal elements.

3. The voicing contrast in Hungarian is marked by the element H, not L. Thus, voiceless obstruents contain an extra H compared to their voiced counterparts.

 

I show that GP lacks the necessary theoretical apparatus to account for the data. I propose an analysis in a combined model of GP and OT, following the work of Polgárdi (1998), Rowicka (1999) and others.

 

My talk also addresses the question of the phonetic interpretation of elements. While lip service is paid to the idea of substance-free phonology in GP, most analyses assume a one-to-one mapping between phonetic and phonological objects. I argue that the interpretation of phonological primes should be based on their place within the system and their phonological behaviour rather then their phonetic characteristics.

 

References

 

Polgárdi, Krisztina 1998. Vowel Harmony. An account in terms of government and optimality. Ph.D. dissertation, Universiteit Leiden. Holland Academic Graphics, The Hague.

 

Rowicka, Grażyna 1999. On Ghost Vowels. Ph.D. dissertation, Universiteit Leiden. Holland Academic Graphics, The Hague.

 

Szigetvári, Péter 1998a. ‘Voice assimilation in Hungarian: the hitches’. The Even Yearbook. ELTE SEAS Working Papers in Linguistics 4. 223-236.

 

Szigetvári, Péter 1998b. ‘Why [h] is not voiced’. In Eugeniusz Cyran (ed.) Structure and interpretation. Studies in Phonology. PASE Studies & Monographs 4. Lublin: Wydawnictwo Folium. 287-301.