The Development of Explicit Rule-Learning

Implicit and explicit learning were originally distinguished in terms of accessibility to verbal report. We identify evidence for the proposal that the im-plicit/explicit contrast corresponds to a divide between connectionist and symbolic representations. We show that explicit learning shows marked improvement between 4 and 8 years of age. This nding contrasts against very early implicit learning abilities, and concurs with other evidence on the progressive development of symbolic reasoning abilities. The identiication and study of human learning mechanisms is a central concern of psychology and cognitive science. An important contemporary debate in this area concerns the distinction between implicit and explicit learning. Generally, implicit and explicit learning mechanisms have been distinguished in terms of the accessibility of the knowledge acquired to conscious awareness, as assessed by verbal report (e.g., Reber, 1967). At rst this division appears to be relatively convincing: In implicit learning, by deenition, participants' verbally reported knowledge is insuucient to account for their performance on some task. For example, participants who memorise strings generated by a nite state grammar are subsequently able to classify strings as obeying or violating the grammar to a signiicant degree, despite being unable to report the rules of the grammar verbally (e. By implication, explicit learning is deened as those cases where participants are able to verbally report suf-cient knowledge to account for their performance. For example, if a person can play a legal game of chess, and can also report the rules of chess, then one might conclude that their ability to play a legal chess game was based on this explicit knowledge. However, the claim that dissociations between verbal report and performance mark the boundary between two distinct learning mechanisms, diiering in the accessibility of their knowledge to conscious awareness, has proved problematic. The strongest critics, Shanks and St. John (1994), argue that the insensitivity of verbal report , and the problems of relating participants' reports to the knowledge representations underlying their performance render apparent dissociations between performance and verbal report suspect. Shank and St. John do however endorse the view that learning mechanisms can be distinguished in terms of the representational form of the knowledge acquired, contrasting exemplar or instance-based learning mechanisms, for which con-nectionism provides a natural model, against processes of hypothesis testing and rule-discovery, best described by symbolic mechanisms. In connectionist learning mechanisms, knowledge and cognition are embodied in patterns of activation of many simple units, and in …