Working Memory in Children Predicts Performance on a Gambling Task

Abstract

ABSTRACT. The authors investigated whether working memory (WM) plays a significant role in the development of decision making in children, operationalized by the Children's Gambling Task (CGT). A total of 105 children aged 6–7, 8–9, and 10–11 years old carried out the CGT. Children aged 6–7 years old were found to have a lower performance than older children, which shows that the CGT is sensitive to participant's age. The hypothesis that WM plays a significant role in decision making was then tested following two approaches: (a) an experimental approach, comparing between groups the performance on the CGT in a control condition (the CGT only was administered) to that in a double task condition (participants had to carry out a recall task in addition to the CGT); (b) an interindividual approach, probing the relationship between CGT performance and performance on tasks measuring WM efficiency. The between-groups approach evidenced a better performance in the control group. Moreover, the interindividual approach showed that the higher the participants’ WM efficiency was, the higher their performance in the CGT was. Taken together, these two approaches yield converging results that support the hypothesis that WM plays a significant role in decision making in children.

Keywords:

• gambling task
• child development
• working memory
• decision making

Notes

¹At the end of the experiment each child was thanked for his or her participation by being given a small packet of M&M's (in addition to those M&M's he or she gained), irrespective of the performance in the experiment.

²To warrant a double task condition that resembles as much as possible the control condition we did not introduce a secondary task that would be orthogonal to the CGT (e.g., a tone counting task) but chose a secondary task that ensures keeping with an ecological situation very close to that of the standard CGT.

³Such modeling is appropriate since the choices are independent between children.

⁴Such a probability difference may not seem very important at first glance, but it actually represents more than 30% of the range of possible improvement in the probability between trial 20 and the largest probability value (1).

⁵Secondary task data were missing for one child.

⁶Note that this correlation remains significant when controlled for participant's age.

⁷We acknowledge that the difference in acquisition speed between the two groups must not hide a high within-group between-children variability in acquisition speed. Indeed, analyses using random coefficients models that allow for taking into account the between-children performance heterogeneity (for an introduction to random coefficients models in the framework of developmental psychology, see Wainwright, Leatherdale, & Dubin, 2007), that we do not present here, did not yield a significant group effect on the speed of acquisition. Such analyses show that, above and beyond a difference in group means between the two experimental conditions, there are interindividual differences that are left unexplained. Among the possible explanatory factors for these differences between individuals, emotional processes of the sort of those discussed in the introduction are ideal candidates.