The Different Role of Working Memory in Open-Ended Versus Closed-Ended Creative Problem Solving: A Dual-Process Theory Account

Abstract

This study examined how working memory plays different roles in open-ended versus closed-ended creative problem-solving processes, as represented by divergent thinking tests and insight problem-solving tasks. With respect to the analysis of different task demands and the framework of dual-process theories, the hypothesis was that the idea generation in a divergent thinking test relies more on associative, effortless system 1 processing, but insight problem solving requires rule-based, resource-limited system 2 processing, in addition to system 1 processing. Since system 1 was suggested to be more active in resource deprivation conditions, Experiment 1 adopted the dual-task paradigm, which increased participants' working memory load. The results showed that divergent thinking performance was enhanced and insight problem-solving performance was hindered. Experiment 2 using the individual differences approach found that individuals' working memory capacity correlated with insight problem solving but not with divergent thinking performance, indicating a possible involvement of system 2 processing in insight problem solving. These findings suggested that open-ended and closed-ended creative problem solving involve different processes and helped to clarify some past inconsistencies when considering the relationship of factors with creativity.

Acknowledgments

This study was a part of first author's dissertation, supervised by the second author. This research was supported by grants to the first and the second authors from National Science Council of Taiwan (NSC 97-2410-H-431-014 and NSC 96-2413-H-002-008-MY2, respectively). Thanks to Mark A. Runco and other anonymous reviewers for their extremely helpful comments.

Notes

¹The fourth index, elaboration, is used especially for a figural question that refers to the degree of elaborating on a drawing by adding detailed decorations.

²Dominowski (1995) identified the features of insight problems as follows: (a) the solution of the problem is not beyond the knowledge of the problem solvers; (b) the initial attempt or hypothesis is usually strong but wrong, which creates obstacles to finding the solution; and (c) solving the problem requires change of problem representation or seeing the problem from a new perspective. The 2-4-6 problem is consistent with these properties.

³The fluency scores were simply the total number of responses generated by each participant. The flexibility scores represented the number of different categories of responses (e.g., the category of toys or decorations). The originality scores represented the sum of scores on each response, which was compared to the norm and scored as follows: responses appeared above 5% in the norm scored 0, responses appeared 2% to 4.99% in the norm scores 1, and responses appeared below 2% in the norm scored 2. The elaboration scores of the figural subtest were scored as the number of elaborated decorations in each response.

*p < .05. ^#05 <p < .1.

⁴Participants' scores on verbal and figural subtests of the CVCTT were highly correlated (Pearson's r ranged from .11 to .51) according to previous studies (Wu, 1998) and the results of Experiment 1. Therefore, Experiment 2 used only the verbal subtest for proper testing-time control.

Note. New-p = number of new-perspective hypotheses. Succ = final solution of 2-4-6 task (success or fail). V-flu = verbal fluency scores. V-flex = verbal flexibility scores. V-orig = verbal originality scores. WM = working memory capacity.

⁵According to the findings of Knoblich and his colleagues (1999), 93% of solutions were generated in the first 3 min (although 5 min was allowed).

Note. INS = insight problem. V-flu = verbal fluency scores. V-flex = verbal flexibility scores. V-orig = verbal originality scores. WM = working memory capacity.

⁶The hypothesis that individuals with low working memory capacity preferred system 1 processing did not imply that others who possessed higher working memory capacity preferred system 2 processing. Our data showed that individuals' working memory capacity did not correlate with their divergent thinking performance; it is possible that individuals with higher capacity could choose either system 1 or 2 processing mode.