Meinhof's Law and the representation of nasality

The oral-nasal contrast is present in many languages (Hockett 1955, Ferguson 1966, Maddieson 1986). The conventional representation of this contrast is in terms of the binary feature [+/-nasal], and this still seems to be the prevailing view (see Oements and Hume, to appear). However, the assumption that nasals are phonologically marked as [+nasal] and contrasting oral segments are [-nasal] leads to analyses of the distribution of nasality that are transparently ad hoc. The core .of the problem is the apparent inability to provide a principled account of the conditions that determine when oral segments are specified as [-nasal]. For example, nasal harmony systems in languages like Warao and Malay seem to require that liquids and obsttuents be specified as [-nasal] to explain why they are opaque to nasal spreading (van der Hulst and Smith 1982, Piggott 1989). In contrast, the corresponding consonants in Guarani and Southern Barasano cannot be specified as [-nasal], because they are either targeted by nasal spreading or are transparent to the process. The specification of the nasal-oral contrast in languages with superficially similar consonant systems (i.e., Warao and Guarani) seems to be completely arbitrary.Nasal harmony patterns like those in Guarani and Southern Barasano in which there are no opaque consonants pose another problem for the conventional specification of the nasal-oral contrast. In a system without underlying [-nasal] consonants, the feature [nasal] is functionally monovalent. This feature appears to have an equivalent status in expressing underlying nasal-oral contrast in vowel systems. Cases in which [-nasal] vowels pattern consistently with [-nasal] consonants are unattested. If orality can be the phonetic instantiation of the absence of specification for [nasal] in vowel systems and some consonant systems, one has to question whether the specification [-nasal] everidentifies a well-defined phonological unit. Further indication that the [-nasal] specification is not the appropriate description of non-nasal segments comes from the observation that [+nasal] segments never lose their nasality simply by assimilating [-nasal] from an adjacent consonant.The resolution of the problems identified in the above paragraphs requires a radical rethinking of the conventional representation of the nasal-oral contrast. Such a rethinking is reflected in the proposals by Piggott (1992). Assuming the model of feature dependency, the so-called feature geometry, proposed by Clements (1985), Piggott proposes that the feature [nasal] is variably a dependent ofa Soft Palate (SP) node or a Sonorant (SV) node. Thefrrst part of the hypothesis agrees with Sagey (1986), the second part with Rice and Avery (1989) and Rice (1993). The variable dependency hypothesis allows for two distinct patterns of nasal-oral contrast. In one pattern the contrast is relevant to sonorants only, while the other reflects an opposition within a wider class of consonants that also includes obstruents. Piggott demonstrates that these patterns of contrast account for the distribution of transparent, opaque and targeted segments in nasal harmony systems. The present paper provides additional support for the hypothesis of the variable dependency of nasality by showing that it leads to an explanation of important features of the assimilation process referred to as Meinhofs Law or Ganda Law, which occurs in a number of Banm languages (Herbert 1986). The analysis makes use of some of the significant insights of Optimality Theory (Prince and Smolensky 1993, McCarthy and Prince 1993a, b)