How discourse context shapes the lexicon: Explaining the distribution of Spanish f- / h- words

Using a corpus of Medieval Spanish text, we examine factors affecting the Modern Standard Spanish outcome of the initial /f/ in Latin FV-words. Regression analyses reveal that the frequency of a word's use in extralexical phonetic reducing environments and lexical stress patterns significantly predict the modern distribution of f-([f]) and h-(Ø) in the Spanish lexicon of FV-words. Quantification of extralexical phonetic context of use has not previously been incorporated in studies of diachronic phonology. We find no effect of word frequency, lexical phonology, word class, or word transmission history. The results suggest that rather than frequency of use, it is more specifically a word's likelihood of use in contexts favoring reduction that promotes phonological change. The failure to find a significant effect of transmission history highlights the relative importance of language internal sources of change. Results are consistent with usage-based approaches; contextual variation creates differential articulatory pressures among words, yielding variable pronunciations that, when registered in memory, promote diachronic change. 1. Introduction Word frequency has long been a linguistic variable implicated in phonological variation and change (Schuchardt 1885, Zipf 1929). Word frequency effects have been reported in a wide range of studies (Labov 1994, Bybee 2001, Phillips 2006, Gries & Divjak forthcoming), with more frequent words typically showing increased rates of reduction compared to lexical items of lower word frequency. Even after controlling such factors as word length, word class, prosodic and lexical stress, phonetic form, speech rate, sociolinguistic factors, and probabilistic measures of language use, many studies have found word frequency to be a predictor of pronunciation variation by at least some measures, especially measures of reduction. The amount of scrutiny and interest devoted to the effects of word frequency reflects, in part, the theoretical ramifications of evidence corroborating or contradicting such effects. Word frequency effects have been essential components in theories of lexical representation (Bybee 2001, Pierrehumbert 2001) and in theories of language variation (Labov 1981) and language change (Phillips 2006). In fact, so central to contemporary linguistic arguments is word frequency that researchers continue to modify methods of quantification and analysis in order to assure the reliability of reported results, although disagreements continue over what methods are appropriate in assessing frequency effects (File-Muriel 2010, Clark & Trousdale 2009). As Clark & Trousdale (2009) noted, it is imperative not to examine " the effects of lexical frequency on variation