Published online by Cambridge University Press: 20 February 2015
Many languages have restrictions on word-final segments, such as a requirementthat any word-final obstruent be voiceless. There is a phonetic basis for suchrestrictions at the ends of utterances, but not the ends of words. Historicallinguists have long noted this mismatch, and have attributed it to an analogicalgeneralisation of such restrictions from utterance-final to word-final position.To test whether language learners actually generalise in this way, twoartificial language learning experiments were conducted. Participants heardnonsense utterances in which there was a restriction on utterance-finalobstruents, but in which no information was available about word-finalutterance-medial obstruents. They were then tested on utterances that includedobstruents in both positions. They learned the pattern and generalised it toword-final utterance-medial position, confirming that learners are biased towardword-based distributional patterns.
We would like to thank audiences at the annual meeting of the LinguisticSociety of America (Pittsburgh, 2011), UC Berkeley Phorum, StanfordP-interest and UCSC Phlunch, as well as Megan Crowhurst (including forher help in creating stimuli), Maria Gouskova, Junko Ito, Danny Law,Grant McGuire, Richard Meier, Armin Mester, Deniz Rudin, Kristine Yu andthe anonymous reviewers for helpful suggestions and stimulatingdiscussion of this paper. We are also grateful to Angela Aiello,Kimberly Cooper, Tommy Denby, Saralynn Emery, Scott Gaudinier, ElannaGrossman, Sean Hayes, Jules Lacour, Jessica Magallan, Melissa Ottele,Tatiana Puente, Rémy Ullman, Aileen Villapudua, BlakeWatkins, Stephen Welch, Nicholas Whittier and Akari Yamamura for theirwork in support of this project.