The study examined neural substrates underlying the selection of monetary rewards in healthy adolescent volunteers. As predicted, choice of a reward was associated with greater activations in the prefrontal regions of dorsal ACC, ventrolateral OFC and mPFC, which have been associated with conflictual decision-making. The PFC activations (especially involving ACC) were most robust in response to the high- vs. low-risk choices followed by the combined high-and low-risk (risk/reward behavior) versus control condition, and lastly in comparison with equal-risk (50/50) options. This graded prefrontal activation across the three contrasts employed in this study supports the notion that high-risk selections posed greater cognitive challenge compared to purely reward-related behavior.
Despite greater PFC activation with high- than low-risk choice, an inverse relationship was observed between high-risk behavior and the magnitude of neuronal activation. These contradictory findings can be explained in the following manner. The adolescents, as a group, selected the low-risk options more frequently than the high-risk options. However, it appears that when adolescents made the high-risk selections, they were more conflicted because the probability of gaining a reward was less likely. The group contrast differences showing greater activation in response to high-risk choices might be driven by the subgroup of adolescents who selected the high-risk option less frequently. In contrast, the negative correlation between the proportion of high-risk choice and magnitude of PFC activation indicates that those adolescents who made the high-risk choice less frequently were more conflicted when they made this choice (thereby inducing greater PFC activation), whereas those who made the high-risk choice more frequently might have given greater consideration to the magnitude of reward than the probability of attaining this reward. These results are consistent with findings from other studies in adolescents (Bjork et al. 2007; Bjork et al. 2008
; Eshel et al. 2007
, van Leijenhorst et al. 2006
) and adults (Bjork et al. 2007; Carmichael and Price 1996
; Eshel et al. 2007
; Schoenbaum and Roesch 2005
; Taylor et al. 2006
). From a neurobiological standpoint, the negative correlation between high-risk behavior and PFC activations may reflect a propensity for impulsive decision-making, potentially mediated by a relative lack of OFC-mediated impulse control/response inhibition and ACC-mediated conflict monitoring/resolution.
The activation of dorsal ACC in response to high-risk choices suggests that adolescents in this study may have used their cognitive ability to resolve conflicts to take action rather than the emotional salience of the stimulus. The preference for low-risk option in this study may also reflect an aversive reaction to the conflictual decision-making resulting in a cognitively less challenging and affectively safe selection with the lower probability of a painful outcome. This view is supported by involvement of ACC during conscious processing of pain (Zaki et al. 2010). Similarly, activation of the ventrolateral OFC indicates some level of impulse control during reward-related behavior as proposed by Eshel et al. (2007)
. However, unlike findings from the study by Eshel et al. (2007)
, this study failed to observe the extension of OFC activation into the anterior insula. This could be due to a relatively lesser activation of the insula in adolescents compared to adults, and age-related differences in the function of harm-avoidance circuitry (Ernst et al. 2005
). However, involvement of more ventral ACC (pregenual ACC) in response to combined high- and low-risk choices as compared to equal-risk selections may reflect greater emotional salience of high- versus equal-risk stimuli (Hampton and O'Doherty 2007
). Nevertheless greater activation of these regions during the high-risk as compared to low-risk choices suggests that the participants were more conflicted when they contemplated the choice of high reward due to the low probability of attaining this reward. Taken together, these results suggest that probabilistic monetary rewards with varying levels of risk and magnitude activate distinct sub-regions within ACC and OFC in adolescents as well as adults.
Recruitment of bilateral mPFC during high-risk behavior, as observed in this study, has been documented earlier (Bush et al. 2002
). This is not unexpected as mPFC is part of an interconnected network organized around ACC, which encompasses a wide range of neural and neurochemical systems implicated in autonomic and visceral control (Carmichael and Price 1996
; Ongur and Price 2000
; Ongur et al. 2003
). In addition, mPFC has been implicated in the registration of reward versus punishment (Breiter et al. 2001
; Elliott et al. 1999
), and controls the reward-related behavior by tracking the magnitude of delivered reward (Breiter et al. 2001
; Knutson et al. 2003
). The later finding is supported by the observation of bilateral activation of mPFC in response to rewards of higher magnitude than lower magnitude in this study. Of note, higher SES was associated with lower activation in the right mPFC during the reward-related behavior. This could reflect greater mPFC-related “default” activity and less cognitive effort by adolescents from the lower SES compared to those from higher SES, despite more frequent high-risk selections. Based on new EEG data from UC Berkeley (accepted for publication in the Journal of Cognitive Neuroscience), children with low SES were found to have lower prefrontal response to unexpected novel stimuli than those from high SES. http://www.sciencedaily.com/releases/2008/12/081203092429.htm
With regard to reward magnitude, only one study in youngsters (Galvan et al. 2006
) and three studies in adults (Galvan et al. 2006
; Rogers et al. 2004
; Smith et al. 2009
) assessed the impact of the magnitude of reward on PFC activations. Galvan et al. (2006)
reported greater OFC activation with rewards of large magnitude in children and adults but not adolescents, while Rogers et al. (2004)
and Smith et al. (2009)
observed the activations of ACC and mPFC in response to rewards of larger magnitude than smaller magnitude. The adolescent sample size in the Galvan et al. (2006)
investigation was modest (n = 13), limiting the power to detect more modest effects. Overall, the findings from this study are consistent with results from these earlier studies and rewards of greater than smaller magnitude induced more significant activation of bilateral ACC, left OFC, and bilateral mPFC. It was interesting to observe a similar ventral extension of ACC activation in response to rewards of larger than smaller magnitude as observed with risky versus non-risky behavior in this study. These findings suggest that risky choices and rewards of larger magnitude may have a higher emotional salience than non-risky choices and rewards of smaller magnitude, respectively.
analysis revealed greater activation of the left ventral striatum in response to rewards of larger magnitude, when the probability of reward was maintained constant. This finding is consistent with results from an earlier study, which showed a direct correlation between reward magnitude and activation in the ventral striatum (Knutson et al. 2001
). In contrast to this, both ventral and dorsal portions of the striatum were activated while making a choice that involved higher -magnitude reward than lower magnitude reward, suggesting that both ventral and dorsal portions of the striatum are involved in the decision-making process involving monetary rewards (Belleine et al. 2007
; Ernst et al. 2005
; Galvan et al. 2007
). It has been suggested that the dorsal striatum is particularly involved during the selection and initiation process through the integration of sensorimotor, cognitive and emotional/motivational information (Balleine et al. 2007
; O'Doherty et al. 2004
), whereas the ventral striatum might be better able to critically ascertain the reward magnitude, particularly under uncertain conditions (O'Doherty et al. 2004
; Knutson et al. 2001
; Pagoni et al. 2002
). Consistent with a previous investigation, activation in the striatum was associated with the choice of a high-magnitude reward regardless of age but not specifically with high-risk/impulsive behavior (Galvan et al. 2007
In contrast to our hypothesis, chronological age did not have a significant influence on brain activations associated with the selection of monetary rewards. These results suggest that, within the adolescent population, individual differences appear to have a greater influence than chronological age per se. Previous investigations reported developmental differences among various age groups (Eshel et al., 2007
; Galvan et al., 2006
). Despite the developmental differences across age groups, there were significant individual differences within specific age groups and these individual differences contributed greater variance to the brain-behavior relationship in both adolescents and adults (Eshel et al. 2007
; Galvan et al. 2007
). Also, no relationship was observed between pubertal stage and PFC activations. However, this might be due to the limited variance in pubertal status within this sample, with majority of the adolescents in later pubertal developmental stage. Future investigations should recruit youth across different pubertal states and also include gonadal hormonal measures in order to better understand the association between pubertal status and the neurobiology of decision-making. Finally, adolescents from higher SES selected a high-risk reward more frequently than those from lower SES, suggesting that their economic advantage offers the choice to take greater monetary risks.
The findings from this study should be interpreted in the context of study limitations. The sample comprised of adolescent volunteers with stringent eligibility criteria, and the results might not be applicable to adolescents in the community. Also, the temporal differentiation of BOLD responses associated with the selection phase from the anticipation phase is limited, especially due to the lack of jitter in the WOF task used in this study. However, this is an issue that is not specific to this study because anticipation is probably present even before the selection since it contributes to the selection itself. Although activation of the ventral striatum observed in this study could be attributed to the anticipation phase of decision-making as reported earlier in healthy adolescents and adults (Bjork et al. 2004
; Knutson et al. 2001
), the striatum, particularly the dorsal portion, is also involved in the selection phase (Balleine et al. 2007
). Additionally, the selections could be impacted by the feedback phase as well. However, the regions involved in feedback (i.e., ventral striatum and mPFC) are different from the regions involved in selection (lateral PFC, OFC) (Ernst et al. 2004
). Thus, it is unlikely that there might be a systematic bleeding of feedback into selection in the regions examined. Finally, the task was not designed for even distribution of chance trials, which explains the significant inter-subject difference in the selection of 25% versus 75% probabilities within the 2575 wheel. This explains how the three subjects ended up choosing less than 10% of 25% (high-risk) option and were not analyzed in this study due to statistical reasons.
Despite these limitations, this study reports differential pattern of activation in response to varying levels of risk and reward probability without the confounding effects of reward magnitude and reaction times in a relatively large sample of adolescent subjects. The large sample size also provided the opportunity to control for important sociodemographic variables to examine the relationship between high-risk behavior and the magnitude of PFC activations. The unique contribution of high-risk behavior to PFC activations in ACC, OFC, and mPFC suggests that individual differences in risk-taking behavior might be important in recruiting the neural circuits associated with decision-making during adolescent development than chronological age or any other sociodemographic factors.