Stress on I: Debunking unitary contrast accounts

Much previous work on stress describes its function as being that of marking contrast. While some evidence has been adduced in experimental studies, work on spontaneous speech data has been plagued by a lack of operational definitions. To address this, we examine approximately 1,500 tokens of the English first singular subject pronoun in a corpus of conversational American English. Independently motivated operationalizations of contrast fail to support an overarching contrastive function of stress on I. Rather, examining co-occurrence patterns through multivariate analysis, we find that, besides chunked units (including discourse formulae as delimited by frequency and positioning), patterns of stress are subject to context-dependent discourse factors: accessibility (measured in distance from the previous mention), in tandem with coreferential priming (a tendency to repeat a preceding coreferential stressed I), as well as turn taking (an initial-position effect), and contrast in a semantic sense (manifested in higher rates of stress under negative polarity).