Neural representation of expected value in the adolescent brain

Abstract

Previous work shows that the adolescent reward system is hyperactive, but this finding may be confounded by differences in how teens value money. To address this, we examined the neural ontogeny of objective value representation. Adolescent and adult participants performed a monetary gambling task in which they chose to accept or reject gambles of varying expected value. Increasing expected value had a stronger influence over gambling choices in adolescents relative to adults, an effect that was paralleled by greater activation in the ventral striatum in adolescents. This unique adolescent ventral striatum response remained even after matching groups on acceptance behavior. These behavioral and neural data suggest that the value of available options has a greater influence in adolescent versus adult choices, even when objective value and subjective choice are held constant. This research provides further evidence that hyperactivation of reward circuitry in adolescence may be a normative ontogenetic shift that is due to greater valuation in the adolescent brain.

Fig. 1.

Example of three trials from the mixed gambles task. Participants responded within 3,000 ms by pressing one of four keys. A jittered interstimulus interval followed before the subsequent trials. Gamble outcomes were not revealed during the scan.

Fig. 2.

The effect of increasing EV on behavioral responses for adolescents and adults. For both groups, increasing EV increased the likelihood of accepting a gamble. The influence of increasing EV on response was greater for adolescents versus adults.

Fig. 3.

Parametric analyses revealed neural activation that changes in proportion with increasing EV. (A) Regions showing increasing activation with increasing EV. (B) Regions showing decreasing activation with increasing EV. Data represent activation collapsed across all participants.

Fig. 4.

(A) Significant activation was observed for adolescent participants in left VS in response to parametrically increasing EV, z = 3.45, P = 0.012, cluster size = 28 voxels. No significant activation was observed for adult participants. (B) There was a significant developmental difference in left VS, z = 3.01, P = 0.0223.

Fig. 5.

(A) A performance-matched analysis only included data from adult and adolescent participants (indicated in gray box) who accepted >80% of gambles. (B) An HLM analysis confirms that the subset of participants highlighted in A do not differ in their behavior. (C) This analysis revealed greater activation in adolescents in right VS (12, 10, −6) (corrected).