New York University, USA
TP187: Usage-based and rule-based approaches to phonological variation
The “variable rule” (VR) model of sociolinguistic variation grew out of generative phonological, rule-based formalisms, and continues to incorporate many assumptions about the architecture and operation of phonology that are associated with such approaches, including abstract representation and a separation between phonology and lexicon. However, the VR model resolves many of the limitations of these formalisms through probabilistic quantification; this strategy permits the treatment of variation, gradience, continuous variables, and multidimensionality. At the same time, the VR approach preserves many advantages of the rule-based models, including the capacity to represent categorical processes and abstract manipulation of phonological objects.
This paper will cite evidence from variation in speech style, child language, and language change across the adult life span showing that speakers have both discrete abstract analyses and nondiscrete, variable treatments of their phonological elements. The variable process of coronal stop deletion in English (west side~wes’ side) provides one illustration of this dualistic representation. The process is highly variable: all English speakers do it some of the time, but nobody does it all of the time. It shows complex conditioning by phonological and morphological factors, including the morphemic status of the deletable final coronal stop: regular past tense forms like missed, bowled, whose final stop is a reflex of the tense marker, undergo deletion less often than underived words like mist, bold. Words must therefore have a mental representation that reflects internal morphological structure. Significantly, this structural representation is subject to abstract reanalysis, with quantitative consequences. Guy & Boyd 1990 show that speakers treat irregular verbs of the left, told, slept class as if they were underived forms (with higher deletion rates) in adolescence and young adulthood, but in middle age tend to reanalyze the forms as involving a tense marker, with lower deletion rates. This suggests abstract mental reflection on types and categories, rather than movement driven by usage and examples.
An important consequence of the VR approach is that variable processes provide a non-deterministic but recoverable link between alternate representations. Speakers possessed of variable rules in their mental grammars know that there are linguistic items that assume several different overt forms in a probabilistically regular way, while retaining an underlying unity. This explains various mysterious phenomena, such as ‘near mergers’. A vowel merger advancing by a variable rule would reach a point in time where speakers vary between merged (i.e., phonetically overlapping) and unmerged articulations. At such a point, speakers could treat the units as both same and different.
One strength of usage-based models that is not captured in ‘rule-based’ formalisms is the treatment of idiosyncrasies of individual lexical items, such as frequency. We argue that the VR model can accommodate such properties through lexical representation. ‘Lexical exceptions’ to variable processes take the form of exceptionally high or low rates of application of the process in certain lexical items. They can be treated in VR as incorporating distinctive lexical entries with quantitative consequences for phonological variation.
Session: Themed Panel (part 2)
Usage-based and rule based approaches to phonological variation
Friday, April 4, 2008, 15:45-17:15