This dissertation examines three sociolinguistic variables in Polish based on data collected in two speech communities located 30 miles apart in the Mazowsze region in central Poland: Warszawa (population 1.7 million) and Popowo Koscielne (population 850).
The first variable examined is variation between antepenultimate and penultimate stress in nouns primarily of Latin and Greek origin, e.g., matematyka 'mathematics'. The variable, which operates above the level of social awareness, was found to display all the defining characteristics of a stable sociolinguistic variable, including regular social and stylistic stratification, age and sex effects, and social evaluation consistent with the linguistic production data.
The remaining two variables involve the behavior of person/number agreement marking: its variable attachment (chapter 5), and its interaction with stress assignment when it attaches to verbs (chapter 6). The results contradict common stereotypes and popular characterizations in the literature according to which variable attachment has almost completely disappeared from the language, and that it is a rural feature. The feature is robustly present in the data in both communities, especially so in Warszawa where it has acquired a new positive social evaluation: it is particularly used by young well-educated speakers from the highest social class. With respect to interaction with stress, the sociolinguistic patterns discovered in the analysis bear a strong resemblance not to variable attachment, but to the noun stress variation examined in chapter 4.
The findings presented in the dissertation are a strong argument in favor of the non-lexicalist analysis of Polish person/number agreement marking. At the same time, however, they confirm the analysis according to which speakers of Polish have to posit competing grammars for the feature. Such an analysis is also justified by historical developments in Polish syntax and morphology.
