Measuring the phonological (un)naturalness of selected alternation patterns in Polish

Abstract

In most generative research phonological naturalness/markedness has served as a synchronic bias that can explain the predominance of certain patterns in the world's languages. In this paper, on the basis of one language, Polish, it is shown that unnatural patterns are far from rare and, therefore, phonological theories need to accommodate them. The two patterns under scrutiny, consonant mutations and progressive devoicing, are to a large degree unnatural but fully productive. Consonant mutations are subjected to a thorough examination using data from 604 outputs of the concatenation of 27 mutation-triggering suffixes, both vowel- and consonant-initial, in order to assess the role of various predictors of palatality. As there is no observable effect of naturalness but a strong influence of specific suffix-initial segments, base-final consonants and individual suffixes on the palatality of the output, the data provide support for a framework that does not incorporate phonological naturalness as an active bias in models of grammar. The morphophonological patterns, formalized as source-oriented schemas, are morpheme specific. This discussion provides evidence for a model of grammar comprising multiple morpheme-specific cophonologies.