Toward a definition of creativity: construct validation of the cognitive components of creativity

The common way of defining creativity from a cognitive perspective involves divergent thinking production only (Guilford, 1957, Rose & Lin, 1984, Torrance, 1990). However, there is a growing body of evidence that the combination of convergent thinking production, divergent thinking production, and domain knowledge expertise are critical to creativity. Gardner's (1993) qualitative study of eminently creative individuals reveals domain knowledge, convergent thinking production, and divergent thinking production in the lives of these individuals. Three creativity process models (Amabile, 1996; Baer, 1988; Kirschenbaum, 1986) all require the use of domain knowledge, convergent and divergent production. Empirical evidence is needed to examine the opposing definitions in the same context. Using multiple-regression, this study compared the relationship of divergent production alone with a criterion for creativity, versus the relationship between a combination of divergent production, convergent production, and domain knowledge with the same criterion. The criterion was a creative product rating (CPR) made by qualified observers.

Two multiple regression analyses (a) compared the contributions to variance of divergent production, alone, to CPR and (b) convergent production, and domain knowledge with CPR for additional contribution to variance. Instruments that were used include predictors: (1) The Torrance Tests of Creative Thinking, Verbal Form (Torrance, 1990), for divergent production; (2) Structure-of-Intellect Test (NST) for convergent production; (3) Texas Assessment of Academic Skills, writing, for domain knowledge; (4) Harter's Intrinsic-Extrinsic Scale (Harter, 1981) a covariate for intrinsic motivation, and (5) the criterion, CPR, an interjudge rating for creative writing production (Amabile, 1996).

Frequency graphs of the predictors and CPR show linear relationships. Effect size for divergent production was R^ = .029 (n.s.). Effect size for the combined block was R2 = .056. Effect size changes for convergent production and domain knowledge were R^ = .025 (n.s.) and R^ = .002 (n.s.), respectively.

The effect size of all predictor variables is much smaller than we expect, if the combined model were valid. On the other hand, the relationship of the effect size for divergent production to the effect size for convergent production is about what we expect if the combined model were valid. The study served as a preliminary study. Problems and suggestions are discussed.