An Analysis of Frequency as a Factor Contributing to the Diffusion of Variable Phenomena: Evidence from Spanish Data Manuel Díaz-Campos and Michael Gradoville Indiana University 1. Introduction The role of frequency 1 effects in the diffusion of variable linguistic phenomena has been the subject of recent study in the linguistic literature. Bybee (2002, 2003) argues that physiologically- driven phonetic change is subject to a pattern of lexical diffusion, tending to occur in high frequency lexical items first and then spreading to less frequent ones. In the case of Spanish, analyses of syllable-final /ɾ/ (Díaz-Campos 2005, 2006, Ruiz-Sánchez 2007, Díaz-Campos and Ruiz-Sánchez 2008), vowel coalescence in New Mexican Spanish (Alba 2006), intervocalic /d/ in New Mexican Spanish (Bybee 2002), and syllable-final /s/ (Brown 2009, File-Muriel 2009, Minnick-Fox 2006) have been proposed. However, the recent surge in research has lead to considerable debate regarding the role of frequency of usage in language change. For example, while Bybee (2002) and Jurafsky, Bell, Gregory, & Raymond (2001) have found a significant effect for frequency of use as a factor affecting /t,d/ deletion in Chicano English and amongst monolingual American English speakers, respectively; Bayley, Loudermilk, and Li (2006) find that frequency does not have an effect in regularly inflected past tense/past participle forms in Mexican American English as would be predicted by Bybee’s model. Bayley, Loudermilk and Li’s, results show that monomorphemes such as just, first and most are affected by frequency, while regular past tense (e.g. walked) and participial forms are not. This finding, they argued, presents a problem for an exemplar model since this model assumes that frequency effects should be found in both types of forms 2 . Bayley, Loudermilk and Li propose a rule-based account that assumes that regular forms would be rule-derived and would not show frequency effects because they do not belong in the lexicon 3 1 For this paper, frequency refers to the usage of lexical items or patterns (e.g. morphemes, phonotactic units) relative to a reference corpus and that of the corpus of study. For the Caracas data the Reference Corpus for Contemporary Spanish (CREA), while the Habla Culta corpus from Buenos Aires was used for the Buenos Aires data. . This analysis is similar in a sense to the one presented by Guy (1991a, 1991b). Guy makes a proposal, using the theory of lexical phonology, according to which /t,d/ deletion can apply at a fixed rate in different levels of grammar. For instance, monomorphemes (e.g. just, must) have three chances to undergo /t,d/ deletion, while semi-weak past tense (e.g. left) have two chances to undergo deletion and regular past tense (e.g. talked) only have one chance. In his account, it is explained that monomorphermes will show more deletion than regular past tense due to the multiple applications of the rule at different levels. Bayley, Loudermilk, and Li present an analysis within a traditional model, without addressing how current models handle frequency. It also is important to 2 Bybee (2002) argues that regular past tense and participles tend to occur before words that start with a vowel, a context that disfavors reduction. 3 Bybee (2001: 20) explains "structuralist frameworks placed great emphasis on the systematicity of language, and it was thought appropriate to reduce the enormous complexity of language by extracting regularities that could be capture in general statements (i.e., rules), thereby only representing truly idiosyncratic material in a list (i.e,, the lexicon). © 2011 Manuel Díaz-Campos and Michael Gradoville. Selected Proceedings of the 13th Hispanic Linguistics Symposium, ed. Luis A. Ortiz-López, 224-238. Somerville, MA: Cascadilla Proceedings Project.