Developing structural representations : their role in analogical reasoning
Recent research into the development of analogical reasoning has shown that young children are able to recognise and use relational similarity between situations, provided that they possess the necessary domain knowledge (Goswami, 1992). However, in most of the reported studies, the relational structure of the analogy has been made very salient. Circumstances where the relational structure of a problem has to be represented by the problem-solver themselves could result in differing performance. We do not know whether, or in what circumstances, children can correctly construct a representation of the relevant relational information. This thesis reports a series of experiments which investigate the role and development of structural representations for the purposes of analogical reasoning. The first two experiments tested whether primary aged children are able to construct an integrated external task representation by combining separate pieces of relational knowledge. Using series problems as a domain, they provided evidence that performance was not affected by the actual relation used, i.e. either spatial or non-spatial (abstract). However, it was observed that the order in which the task information was presented had an effect. The next four studies explored this by using spatial series problems. They showed that tasks which required a novel item to be placed to the left of (that is, at the front of) a partially ordered array inhibited performance. A further three experiments found that the reason for the inhibition was that unless the different pieces of relational information were highlighted as distinct items, they would be incorrectly integrated by using simple 'add-to-end' ordering rules. The final set of studies, using abstract evaluative relations in series problems, found that relational-highlighting effects generalised to these types of tasks. Also, the results showed that some evaluative relations were tied to either horizontal or vertical spatial representations and that performance was affected by how consistent the representation was with the child's experience of every-day life. The thesis showed that the ability to construct structural task representations is affected by features which are inherent in the presentation of specific tasks, and that incorrect structural representations in turn affect analogical mapping. These findings are discussed in terms of the 'generalised schemas' used during analogical mapping. It is suggested that these might be reconstructed using specific task information, rather than being retrieved intact from memory.