Experiments in implicit learning

• Main Author: Kelly, Stephen William
• Published: University of Glasgow 1997
• Subjects: 150; BF Psychology
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.241897

This thesis examines two paradigms from the area of implicit learning in detail. The literature suggests that the invariance detection paradigm of McGeorge and Burton (1990) gives rise to unconscious knowledge held at a conceptual level with the decision process served by a 'nearest-neighbour' similarity mechanism. The experiments in this thesis suggest that several aspects of this task do indeed seem to agree with present conceptions of unconscious knowledge but no evidence could be found that this knowledge is held at a conceptual level or that specific similarity plays any role in this task. Instead the experiments in this thesis suggest that this task may be better understood in terms of an abstraction mechanism which acquires perceptual information. Using the invariance detection paradigm, this thesis examines the effect of two types of task which measure performance above an 'objective threshold' of awareness. Performance on each task was not the same, suggesting that one cannot assume all direct tests measure the same knowledge despite being similar in nature. In addition, the finding that only the more sensitive of the two tasks could elicit information in the invariance detection paradigm suggests that the knowledge is extremely difficult to elicit. This also is a property of implicit learning and points to the digit invariance task being mediated by unconscious mechanisms. The finding of robust invariance detection in laboratory tasks suggests that one might expect to find similar learning for real world invariance. No evidence for this could be found, which suggests that either implicit learning is a laboratory artefact or that real world invariance learning does not operate in the same way that laboratory experiments suggest. These results suggest that laboratory experiments are required which replicate conditions under which real world learning might occur.