Variable voicing in Argentine spanish /z/

This chapter reconciles recent approaches to the devoicing of /?/ in Argentine Spanish. Using a sample of nine sociolinguistic interviews of Spanish speakers from the northern neighborhoods of Buenos Aires, this study finds that only two of them are best classified as voicers. Subsequent analysis of word-initial /?/ shows that, in addition to the previously identified effect whereby post-pausal contexts favor devoicing, there is a significant effect for bigram frequency in which high-frequency bigrams disfavor the devoicing process, regardless of whether the speaker is a voicer or devoicer. Although word-medial /?/ shows no corresponding frequency effect, voicers show evidence that pretonic occurrences of /?/ are more advanced in the devoicing process than stressed and postonic occurrences. Even though the devoicing may be classified as complete for some speakers, the occurrence of voiced variants in the speech of these speakers is not random, but rather is a consequence of high-frequency sequences.