Trait self-control is predicted by how reward associations modulate Stroop interference

Abstract

Although cognitive control is commonly identified as the basis of self-controlled behavior, correlations found between trait self-control and laboratory measures of cognitive control such as Stroop interference are typically low. Based on the notion that self-control requires the ability to refrain from rewarded behaviors, and inspired by the recent finding that Stroop interference is modulated by reward associations, we propose the idea that the modulation of interference by reward associations (MIRA) is a cognitive marker of trait self-control. Two independent samples of participants completed (1) a modified Stroop task designed to assess MIRA and (2) two common measures of trait self-control: the Brief Self-Control Scale (BSCS) and the Barratt Impulsiveness Scale (BIS-11). MIRA was strongly correlated with the BSCS and moderately correlated with two of the three subscales of the BIS-11. MIRA thus appears to reflect a cognitive endophenotype of individual differences in self-control, and perhaps of related mental disorders.

Appendix: Rationale for and consequences of not counterbalancing response–reward mappings

In the following, we explain the rationale for not counterbalancing response–reward mappings between subjects, and clarify the consequences for the central outcome measure of the present study: MIRA values. Individual MIRA values were calculated as RT differences between reward-related (RR) and reward-unrelated (RU) incongruent no-reward trials, and thus reflect the modulation of interference by reward associations. Importantly, reward associations are inevitably confounded with inhibited response buttons: for example, if rewards are given for correct responses on the second button (as was the case for all participants in the present study), then RR stimuli (‘2’ ‘222’ ‘2222’) will require to inhibit a response on the second button, whereas RU stimuli will require to inhibit a response on either the first (‘111’, ‘1111’), third (‘3’, ‘3333’), or fourth (‘4’, ‘444’) button. To control for this confound, earlier studies (Krebs et al., 2010, 2011, 2013) used between-subjects counterbalancing, randomly assigning participants to different response–reward mappings. This ensures the interpretability of across-subjects means of outcome measures: if positive/negative average MIRA values are observed, one can justifiably conclude that interference was increased/decreased by reward associations. The downside of such a design is that it reduces the interpretability of individual differences in outcome measures: differences in individual MIRA values are not only due to individual differences in the modulation of interference by reward associations, but also due to different response–reward mappings. Between-subjects counterbalancing was therefore not suitable for the present study. Instead, the same response–reward mapping was used for all participants to keep the confound of reward associations with inhibited response buttons constant. This design ensures the interpretability of individual differences in outcome measures: a relatively high MIRA value indicates that reward associations led to a relatively strong increase—or weak decrease—of interference. At the same time, it reduces the interpretability of across-subjects means: MIRA values do not only reflect the modulation of interference by reward associations, but also the (arbitrarily chosen) response–reward mapping. In line with this, the sign of MIRA values—which was negative for most participants in the present study—does not allow conclusions on whether interference was increased or decreased by reward associations. At this point, it must be said that for the present study, the interpretability of MIRA values should be considered as limited in the aforementioned sense only under the premise that RTs were actually affected by inhibited response buttons. The validity of this premise is illustrated in Fig. 3, which shows RTs for incongruent no-reward trials as a function of inhibited response buttons: although responses on the first, third, and fourth button were equally reward-unrelated (not associated with reward), RTs differed significantly depending on whether responses were inhibited on the first or third button, t(40) = 2.152, p = 0.038, first or fourth button, t(40) = 2.376, p = 0.022, and third or fourth button, t(40) = 4.708, p < 0.001.

Fig. 3

RTs for reward-related (RR) and reward-unrelated (RU) incongruent no-reward trials (data from both Experiments combined). Average MIRA values correspond to the vertical distance between the second bar (representing the mean RT for RR trials) and the horizontal dotted line (representing the mean RT for RU trials)