Proceedings of the Annual Meeting of the Cognitive Science Society

Speakers often face choices as to how to structure their in-tended message into an utterance. Here we investigate the in-fluence of contextual predictability on the encoding of linguis-tic content manifested by speaker choice in a classifier lan-guage, Mandarin Chinese. In Mandarin, modifying a nounwith a numeral obligatorily requires the use of a classifier.While different nouns are compatible with different SPECIFICclassifiers, there is a GENERAL classifier that can be used withmost nouns. When the upcoming noun is less predictable,using a more specific classifier would reduce the noun’s sur-prisal, potentially facilitating comprehension (predicted to bepreferred under Uniform Information Density, Levy & Jaeger,2007), but the specific classifier may be dispreferred from aproduction standpoint if the general classifier is more easilyavailable (predicted by Availability-Based Production; Bock,1987; Ferreira & Dell, 2000). Here we report a picture-namingexperiment confirming two distinctive predictions made byAvailability-Based Production.