Category names and category learning

The thesis examines the role of verbal labels in category learning by adults. In an investigation of the effects of category learning and exemplar labelling on quantitative judgements about exemplars, it was found that only labels provided at the time of testing biassed subjects’ judgements, although this effect did rely on the labels having been previously learned Adults' default assumptions concerning the extension of novel names among newly learned categories were examined: subjects used an assumption of mutual exclusivity between names (as predicted by Markman, 1989), but did not adhere to the principle of linguistic contrast (Clark, 1987). In a series of experiments where exemplars were labelled with verbal (non-word) labels or complex visual patterns, category learning was superior with the verbal labels. This superiority spanned categories consisting of arbitrary collections of familiar or unfamiliar objects, and prototype-based polygon categories. Arbitrary collection learning was better when subjects reported inventing names for the non-verbal labels. When the verbal and non-verbal labels were compared on a speeded discrimination task, few errors were made but decision times were reliably shorter (c. 0.1 second) with the verbal labels. With other non-verbal labels which were faster to discriminate than the verbal labels, arbitrary collection learning was at a level intermediate between learning with the verbal arid original non-verbal labels. The role of category names as feedback was investigated in a prototype-based category learning task. Learning was no better with named exemplars or right-wrong feedback than when unaided, although learning with named exemplars plus right-wrong feedback was better than with right-wrong feedback only. No interaction between task difficulty and feedback conditions was found (cf. Homa and Cultice, 1984). Thus in these experiments verbal labels produced better category learning than non-verbal labels, even for schema-based categories which were learned equally well unaided. Four possible functions of category names in category learning are suggested as a framework for future investigations