Change in the adult phonological processing system by learning non-adjacent phonotactic constraints from brief experience: An experimental and computational study

Recent studies show that the adult phonological processing system constantly changes as a result of word processing experience; adult speakers can learn new sound patterns from brief experience processing words that exhibit the sound patterns, and how they process words changes as a result of learning. But how malleable is the phonological processing system and what is the mechanism underlying the adaptation of the system to recent processing experience? This dissertation presents experiments and computational models that investigate whether adult speakers can learn non-adjacent phonotactic constraints, and how their perception and grammaticality judgment behavior change as a result of learning.; The experiments show that adults can learn phonotactic constraints that are nonexistent in their language and which restrict co-occurrence of two non-adjacent phonemes with one intervening phoneme. The results further document evidence of the malleability of the adult phonological processing system, and extend the range of learnable sound patterns since non-adjacent phonological dependencies are assumed to be difficult to learn.; As a result of learning, the speakers judge phonotactically legal novel words to be more grammatical than phonotactically illegal novel words. They also perceive the legal ones more quickly and accurately than the illegal ones. In addition, the experiments show that the effect of learning on perception is greater when the learned phonotactic constraint restricts co-occurrence of more confusable phonemes. This subtle effect of learning on perception is expressed as the Perceptual Facilitation Hypothesis, which provides a more detailed account of how the phonotactic knowledge functions in the adult phonological processing system to change its perceptual behavior. The experimental results are simulated with two computational models that demonstrate how the adult phonological processing system adapts to recent experience: how it comes to perceive legal sound sequences better than illegal ones after repeatedly processing sequences embodying non-adjacent phonotactic constraints, and how it learns the constraints from observing the perceptual output and computes the probability of the perceived phonological structure in judging its grammaticality. The models suggest possible mechanisms that underlie the adaptation of the adult phonological processing system and guide the direction of future research by providing falsifiable predictions.