There are currently two approaches to positional neutralizations within Optimality Theory: Positional Faithfulness (Beckman 1997, Alderete 1995) and Licensing (Steriade 1994a,b, Zoll 1997). In Positional Faithfulness, special faithfulness constraints specific to "strong positions" are allowed: Stressed syllables might be subject to faithfulness requirements not affecting unstressed syllables. In Licensing, special markedness constraints specific to "weak positions" are allowed: The marked vowel [e] might be more marked when unstressed, or conversely, less marked (or licensed) when stressed. Zoll (1997) argues that Licensing is a superior approach, since it handles all phenomena covered by Positional Faithfulness, plus other cases not covered by Positional Faithfulness. We present additional evidence for this claim from the Orlec dialect of Serbo-Croatian (Houtzagers 1985, 1982). Since this dialect displays both shortening of unaccented syllables, and lengthening of certain accented syllables, only the Licensing approach accounts for all the data.
In Orlec, vowel length is contrastive only in accented syllables. All unaccented vowels are short, and if accent moves off a long vowel, it concomitantly shortens. We propose the following constraints:
LIC-V:/ACCENT: Long vowels are marked in unaccented syllables. MAX-mora: Do not delete underlying moras.
Ranking LIC-V:/ACCENT above MAX-mora generates the correct pattern, as shown in the following tableaux:
/kó:nj/ 'horse' (nom.)
LIC-V:/ACCENT MAX-mora -> kó:nj kónj *!
/ko:njá/ 'horse' (gen.)
LIC-V:/ACCENT MAX-mora -> konjá * ko:njá *! *
Based only on this phenomenon, an alternative Positional Faithfulness analysis is possible: The positional faithfulness constraint MAX-mora-ACCENT ("accented syllables may not lose moras"), would simply dominate *V:. However, the presence in Orlec of two lengthening phenomena makes this alternative untenable. In Presonorant Lengthening, a vowel becomes long if it is followed by a coda sonorant. In Nonhigh Lengthening, [-hi] vowels lengthen. Neither phenomenon affects unaccented vowels. In our analysis, the constraints motivating lengthening are simply ranked below LIC-V:, thus blocking lengthening in unaccented syllables. This is demonstrated below for Presonorant Lengthening:
/púnta/ 'stitch' (gen.)
LIC-V:/ACCENT Preson.Length. -> pú:nta púnta *!
/pumpét/ 'pump' (nom.)
LIC-V:/ACCENT Preson.Length. -> pumpét * pu:mpét *!
This approach cannot be extended to Positional Faithfulness. Recall that Positional Faithfulness requires the ranking MAX-mora-ACCENT above *V:. If Presonorant Lengthening applies at all, it must outrank *V:. However, this predicts that Presonorant Lengthening will affect all vowels, including unaccented ones, as shown below. (The notation *> indicates a form incorrectly chosen as the winner.) No re-ranking of these constraints will produce the result that Presonorant Lengthening is blocked only in unaccented syllables.
MAX-mora-ACCENT Preson.Length. *V:
/púnta/ -> pú:nta * púnta *!
/pumpét/ *> pu:mpét * pumpét *!
Since the Positional Faithfulness approach cannot account for both the systematic shortenings and lengthenings found in Orlec Croatian, this dialect provides strong support the Licensing approach to positional neutralizations in Optimality Theory.
References:
Alderete, John. 1995. Faithfulness to Prosodic Heads. ms., University of Massachusetts, Amherst.
Beckman, Jill. 1998. Positional Faithfulness. Ph.D. dissertation, University of Massachussetts, Amherst.
Houtzagers, H. P. 1982. "Accentuation in a Few Dialects of the Island of Cres," Studies in Slavic and General Linguistics 2.
Houtzagers, H. P. 1985. The Chakavian Dialect of Orlec on the Island of Cres, Rodopi: Utrecht.
Steriade, Donca. 1994a. Licensing by Cue. ms., UCLA.
Steriade, Donca. 1994b. Positional Neutralization and the Expression of Contrast. ms, UCLA.
Zoll, Cheryl. 1998. Positional Asymmetries and Licensing. ms., MIT