Do individual differences in Iowa Gambling Task performance predict adaptive decision making for risky gains and losses?

Abstract

We relate performance on the Iowa Gambling Task (IGT), a widely used, but complex, neuropsychological task of executive function in which mixed outcomes (gains and losses) are experienced together, to performance on a relatively simpler descriptive task, the Cups task, which isolates adaptive decision making for achieving gains and avoiding losses. We found that poor IGT performance was associated with suboptimal decision making on Cups, especially for risky losses, suggesting that losses are weighted more than gains in the IGT. These findings were significant beyond several notable gender differences in which men outperformed women. Implications for the neuropsychological study of risk are discussed.

Keywords:

• Iowa Gambling Task
• Risk
• Adaptive decision making
• Reward sensitivity
• Punishment sensitivity
• Cups task
• Gender differences

This research was supported by National Science Foundation Grant No. SES 03–50984, and IIS 04–42586. We would like to thank Elaine Bossard, Melissa McElvain, Ilsa Weishar, and Madeleine Welsh for their help with data collection and manuscript preparation. We would also like to thank Paul Slovic and Lee Newman for their comments on this manuscript.

Notes

¹We use the word “domain” here to distinguish between gain- and loss-related choices. Thus, use of this term relates to the valence of the possible outcomes and should not be confused with substantive content domains, such as financial or health-related decision making.

²Notably, there were differences between groups in the number of initial advantageous cards selected. Since the first trial block is associated with early sampling of all decks and not necessarily advantageous decision making per se, we conducted an analysis of covariance (ANCOVA) to test whether an early amount of advantageous decks selected influenced subsequent performance on the remainder of the task. We found that this covariate did not influence subsequent performance between subjects, F(1, 356) = 0.6, ns, nor did we find a Block 1 × Trial Block interaction, F(3, 1068) = 2.15, ns. Thus, we do not consider this matter further.

³We conducted linear mixed-model analyses using IGT performance as a continuous variable predicting Cups task performance. The results of these analyses did not provide any different interpretations of the results than those obtained by categorizing IGT performance. Hence, for expository purposes, we present the latter.