Covariation Analysis of Knowledge Representation: Some Developmental Studies.

Abstract In this paper it is argued that the most important contents of knowledge representations are sets of covariations abstracted from experience. Such covariations are frequently used collaterally, and at different representational levels, to constrain information among variables that would otherwise be too “fuzzy” or unreliable for predictability to be possible. These networks of covariations, it is argued, are the key contents of “what develops” in childhood and define the structures of the representations and the reasoning that are seen to emerge. The covariation structure of children's knowledge in one familiar and one unfamiliar domain, and across different age groups, was explored using functional measurement techniques and analyses of variance. Interaction effects were taken to reflect the usage of stored collateral covariations in reasoning and these were found for the most familiar domain even among 7 year olds. Interactions were fewer or absent in the less familiar domain. The results may suggest a new approach to the lawful description of children's knowledge and reasoning, and of developmental changes within them.