Opacity, variation and the exponence of Polish virile de- clensions Sławomir Zdziebko John Paul II Catholic University of Lublin 1. Introduction The phonological status of Polish palatalizations has been repeatedly questioned by scholars of various theoretical inclinations (see Dressler 1985; Spencer 1985; Gussmann 2007, among others). The most often invoked argument against treat- ing palatalizations as phonological phenomena is the fact that the supposed trig- gers of palatalizations do not form a natural class: palatalizations are not ob- served before all, and only, front vowels. The derivational techniques employed to make sure that only front objects trigger palatalizations have often been criti- cized as being exceedingly abstract. 1 Additionally, the input-output mappings involved in palatalizations are difficult to express unless abstract underlying and intermediate representations are employed. One of the most common palatalizations in Polish changes / /, // and // into //, // and / / respectively. Unless some segments that do not surface in the native vocabulary of standard Polish are claimed to be involved in the rel- evant derivations, we face a paradox: an incoherent set of distinctive features is affected in the environment of a coherent set of affixes. 2 One of the possible strategies to get round this paradox is to follow Gussmann’s (2007) assumption that palatalizations in Polish involve replacements of entire stem-final segments with their palatalized congeners rather than operations manipulating distinctive features. This path will be followed in this article. Under this interpretation, pal- atalizations will be viewed as the translation of morpho-syntactic features into phonological segments that overwrite and replace the underlying stem-final segments. The primary challenge for this approach is that the analysis of palataliza- tions must be compatible with the account of the exponence of Polish morpho- logical categories. It also allows to account for the apparent cases of counter- feeding observed between some palatalizations: the replacements of the rhotic // 1 For instance, the claim made by Rubach (1984) that certain types of // that do not trigger the expected palatalizations are the back vowel // at the relevant stage of derivation has been rejected by the majority of scholars (see e.g. Spencer 1985 and Szpyra 1989). 2 It is generally assumed in the literature (e.g. Gussmann 1980) that / / is palatalized to //, which undergoes an unconditioned shift to //. The affrication of a palatalized // is usually hardly commented upon. The labio-velar // is assumed to be the phonetic guise of the dark lateral //, which does not appear in the relevant varieties of Polish.