Implicit learning of phonotactics: an example of pure associative learning?

phonotactics;associative learning;implicit learning;speech production;speech errors;phonology

Anderson, Nathaniel David

关键词: phonotactics;    associative learning;    implicit learning;    speech production;    speech errors;    phonology;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/78458/ANDERSON-THESIS-2015.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

In order to study the processes of implicit learning in speech production, research participants were exposed to sequences of nonsense syllables which obeyed artificial phonotactic constraints. For example, /f/ showed up only at the beginning of syllables for some participants (whereas in English, it appears both at the beginning and end of syllables). Participants spoke sequences of these nonsense syllables quickly in time with a metronome. Learning was measured by investigating the patterns of phonological speech errors they produced. Learning occurred very rapidly in the first block; participants' slips involving the experimentally-constrained phonemes nearly always obeyed the experimental position constraints (i.e. if the participant said an /f/ in error, it nearly always showed up at the beginning of a syllable). One third of the way through the experiment, the constraints were reversed, so that for example, /f/ now only showed at the end of syllables. Participants were much slower to learn the reversed constraint, consistent with a pattern of unlearning the initial constraint. Two-thirds of the way through the experiment, the constraint was reversed again, to the original rule (e.g. /f/ is an onset only). Results for the second reversal were mixed, with some subjects learning the reversed rule and others failing to learn it. Results appeared to be most consistent overall with a purely associative account of phonotactic learning; learning the reversed rule appeared to require unlearning the initial rule, and re-learning the initial rule appeared to require unlearning the reversed rule. This is contrasted with the reversal shift effect observed in discrimination learning tasks, in which repeated reversals of a rule lead to increasingly rapid learning. Overall, the results are consistent with a view of the speech production system as continually adapting to the changing distribution of sound patterns produced. The results are compared to a computer simulation of the task using an artificial neural network model.