Spot the pigeon. Between representation and derivation in Polish nouns with alternative virile/non-virile declensions
Grzegorz Michalski (Adam Mickiewicz University, Poznań)
This paper deals with a semantically-motivated split in declension paradigms of certain nouns in Polish whose morpho-phonological behaviour is different when they are meant to denote people's last names (male surnames will be in focus here) than when they denote animate or inanimate referents. Besides taking different suffixes, the split is manifested by a difference in whether or not the last vowel of the stem alternates, and whether or not the stem-final consonant palatalizes, the latter phenomenon applying disjunctively to the former in some cases. It is claimed here that the alternations are phonologically-grounded and productive, but some stimulation from morphology is needed for them to apply or not. The question is: how much of this split can be encoded in the representation alone, and how much needs to be derived? The presentation shows that for the representation to handle the split alone, a peculiar class of stem-intruding suffixes (infixing suffixes?) would be necessary, and that without any extra-phonological information at hand, phonology alone would not yield the desired output.
Polish has a number of nouns for which concurrent forms exist in their declensions. For example, tramwaj /tramvaj/ ('tram'; Masculine Inanimate) is attested as having two forms in the plural genitive, tramwajów /tramvajuv/, and tramwaji /tramvaji/. Both forms are deemed well-formed, through suffixation, leaving the stem intact, a trivial cyclic operation.
This is not the case with goł±b /gɔwɔ̃b/ ('pigeon'; Masculine Animate), whose paradigm, apart from suffixation, shows alternations in the stem. For example, the singular genitive for goł±b is gołębia /gɔwɛ̃bʲa/, with an umlauted (or raised) vowel in the stem, and a palatalized stem-final consonant. Interestingly, if the same lexeme is called upon to denote a gentleman's last name (Goł±b; Masculine Virile), the same structural case is Goł±ba /gɔwɔ̃ba/, with no umlauting or palatalization in the stem.
This paper advocates the view that phonology alone knows nothing about the goł±b/Goł±b split, and it is through derivation that it can produce the desired output.
For Lexical Phonology and (Derivational) Optimality Theory, the split can be managed either through rule ordering at the word level (cyclic before post-cyclic) or through constraint reordering at subsequent levels of derivation. In case no purely phonological justification can be found for a given output, recourse can be made to morphology taking over, either fiddling with rule application at the word level or forcing morphologically-bound faithfulness constraints.
This approach does not seem applicable, let alone convincing, for non-derivational frameworks, such as Government Phonology (including the more radical faction, CVCV; cf. Scheer 2004), where phonological representation alone is supposed to handle the alternations.
Should phonology be oblivious to non-phonological objects, ignoring syntactic information of the kind [Animate] and [Virile], and communicating with other (higher?) modules through postcards and a translator's office (cf. Scheer 2005), then how can the same representation (/gɔwɔ̃b/, that is) yield two strikingly different declensions? (Or is it two representations?)
This presentation shows a path between the extremities of alleged non-derivational parallel representationalism, and three-level serial derivationalism.
2004 A Lateral Theory of Phonology. Vol 1: What is CVCV and why should it be? Berlin: Mouton de Gruyter.
2005 "Postcard theory: how higher level modules communicate with phonology through a translator's office." Paper presented at the 36 th Poznań Linguistic Meeting (PLM 2005), Poznań.