Implicit and Explicit Processes in Phonotactic Learning

Recent years have seen a proliferation of adult phonological-learning studies (“artificial-language” experiments) employing a wide array of experimental tasks, instructions, and materials (reviewed in Moreton & Pater 2012a,b), in the hope of gaining experimental access to the inductive processes underlying firstor secondlanguage acquisition. But there has been little investigation into what is actually going on in these experiments. Do different experimental situations engage different learning processes? If so, do those processes have different inductive biases? How are they related to the processes involved in L1 and L2 acquisition? Answers to these questions have implications for both methodology in particular, and cognitive science in general. Studies of non-linguistic (mainly visual) pattern learning have led psychologists to hypothesize two concurrent learning processes that have different properties and that are facilitated by different experimental conditions (Ashby et al., 1998; Love, 2002; Maddox & Ashby, 2004; Smith et al., 2012). Here we will call them the explicit system and the implicit system. The explicit system is effortful, conscious, abrupt, and rule-based (e.g., it can be modelled as serial testing of featurally-simple hypotheses); it demands focused attention and working memory, and its use is facilitated by training with right/wrong feedback, instructions to seek a rule, and the use of easily verbalizable stimulus features. The implicit system is effortless, unconscious, gradual, and cue-based (i.e., it can be modelled as weight update on an array of property detectors); it does not need attention or working memory, and its use is facilitated by training without feedback, instructions that do not mention rules, and non-verbalizable stimulus features. This proposal is one manifestation of a more general idea in psychology (reviewed by Osman 2004; Evans 2008; Newell et al. 2011). The two systems also differ in sensitivity to different pattern structures (see below). This paper presents two experiments. Experiment 1 asks whether implicit and explicit processes are available for phonological learning, and, if so, whether they are facilitated by the same conditions as in non-linguistic pattern learning. Experiment 2 asks whether the two processes differ in sensitivity to different pattern types in the same way as in non-linguistic learning. This study contributes to a larger program, the comparative study of inductive learning across domains (Pertsova, 2012; Pater & Moreton, 2012; Moreton, 2012; Moreton et al., in press).