Phonotactics, or the constraints on sequencing of speech sounds, forms an important part of what we know about a particular language. For example, a fluent English speaker knows that can end with ng (king) but do not begin with that sound (ngik is not likely to be a word of English); in contrast, a native speakers of Nenets knows that words can start but not end with ng. When learning a second language, we therefore have to learn that different talkers have different phonotactic constraints so that we can rapidly and efficiently process speech by speakers of our native language as well as speakers of non-native languages (Weber & Cutler, 2006). Over the past two decades, laboratory paradigms have been developed that allow us to examine the initial phases of such adaptation. Previous research shows that in the lab speakers rapidly adapt to novel phonotactics (e.g. Warker and Dell, 2006). But this provides an incomplete model of learning, as learning that different talkers have different constraints (e.g. “Matt doesn’t end his syllables in fricatives; Tommy doesn’t end his syllables in stops”) is difficult or impossible (Onishi, Chambers, and Fisher, 2002). We propose that this puzzling result reflects speakers’ past experience with phonotactic variation. As learners, we know that speakers of different languages produce different phonotactics, while two speakers of the same language produce similar phonotactics. As such, speakers should only adapt to talker-specific constraints when those talkers have different language backgrounds. We explore this hypothesis in 3 experiments: in study 1, listeners are exposed to two talkers, each with a different phonotactic constraint. Preliminary results suggest that these talker-specific constraints are learnable when talkers differ in language background, but not when they share a language background. Study 2 explores the role of interlocutor language background on adaptation to phonotactics in speech production. Study 3 will then investigate the structure and granularity of listeners’ prior knowledge. Can listeners hold models of only a native vs. non-native language, or can they maintain models for two separate non-native languages?
|Effective start/end date||8/1/17 → 7/31/19|
- National Science Foundation (BCS-1728173)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.