In contrast to the extensive literatures exploring the neural basis of mature incentive processing in non-human primates and human adults, fewer studies have specifically focused on the development of this system through adolescence in humans (May et al., 2004
; van Leijenhorst et al., 2006
; Bjork et al., 2007
; Bjork et al., 2004
; Ernst et al., 2005
; Eshel et al., 2007
; Galvan et al., 2006
). Collectively, studies indicate that adolescent incentive processing is supported by a similar neural circuitry as adults, including orbitofrontal cortex, basal ganglia (dorsal and ventral striatum, including nucleus accumbens), amygdala, and medial prefrontal cortex. However, as will be illustrated below, the manner in which these regions are recruited by adolescents differs during the course of incentive processing.
May et al. (2004)
found that children and adolescents recruit ventral striatum and orbital frontal cortex (similar to non-human primate reports) during the anticipation of reward or loss in a gambling task. This study was the first to apply event-related functional neuroimaging methods to child and adolescent incentive processing, but did not have an adult comparison group allowing for developmental comparisons to be made in terms of the recruitment of these primary regions. Studies which have investigated developmental differences between adolescents and adults in incentive processing have focused on different temporal aspects of incentive processing, leading to disparate conclusions. For example, Bjork et al. (2004)
compared blood oxygenation level dependent (BOLD) changes during an anticipatory period (i.e., before responding to receive incentive) in adolescents and adults using the monetary incentive delay (MID) task (Knutson et al., 2000
), a rewarded reaction time task. Briefly, in this task subjects first saw one of several geometric shapes, each of which was uniquely associated with a different magnitude of reward (money) available at trial end. Subjects then fixated a white crosshair for a variable delay period (i.e., the ‘anticipation’ period) after which they had to quickly respond via button press when a white square was flashed on the screen. If subjects responded while the square was still visible, they earned the promised reward. While adolescents performed similarly to adults on this task (by design), adolescents exhibited significantly less activation in the right ventral striatum (nucleus accumbens, NAcc) and extended-amygdala while anticipating responding for a reward (versus a condition where no reward was available). Ernst et al. (2005)
using fMRI examined changes in the BOLD response as subjects performed a rewarded decision-making task—the ‘wheel of fortune’ task. In this task, subjects had to choose via button press which half of a colored wheel they thought would be randomly picked by the computer (referred to as the ‘choice’ epoch). Each colored side was associated with a different magnitude of reward (win money) or punishment (lose money). Following a brief anticipation phase, subjects were presented with feedback about what color the computer selected (unbeknownst to the subjects, the color choice was selected at random but at a predetermined probability) and what incentive they received. During this feedback epoch (i.e., consummatory processing), adolescents demonstrated enhanced activity in the left nucleus accumbens, whereas adults exhibited more activity in the left amygdala, suggesting that adolescents are more sensitive to rewards (associated with NAcc) and adults are more sensitive to punishments (associated with amygdala) (Ernst et al., 2006
). Subsequent work manipulated the probability of receiving a reward by changing the relative size of the colored wheel slices in the Wheel of Fortune task (Eshel et al., 2007
). In this study, BOLD activity unique to the ‘choice’ epoch was investigated. Although behavioral performance did not differ across ages, adults activated OFC/VLPFC (BA 47, 10) and dorsal ACC (BA 32) significantly more than adolescents when making risky selections. These regions are known to contribute to aspects of cognitive control (Casey et al., 2001
) as well as the monitoring and resolution of conflicting decisions (Carter et al., 1998
). Results thus indicate that adolescents do not engage prefrontal regulatory mechanisms as much as adults when making risky choices. In a recent study, Bjork et al. (2007)
investigated the circuitry supporting rewarded decision-making using a novel monetary game of ‘chicken’ in which subjects had to choose when to bank accumulating rewards before the trial unpredictably terminated. Trials varied in terms of the penalty associated with losing (failing to bank winnings before trial stopped). Adolescents activated posterior mesofrontal cortex, a region reported to be recruited during pre-response conflict and during the monitoring and avoidance of errors (Ridderinkhof et al., 2004
), in a similar manner compared to adults in cases when a severe threat of loss was clear. However, under milder and more ambiguous conditions of risk, adolescents under-activated this region. Similarly, children (9 to 12 year-olds) compared to adults (18–26 year-olds) were found to recruit the anterior cingulate cortex more during high risk decision-making and engaged lateral orbitofrontal cortex more in response to negative compared to positive feedback (van Leijenhorst et al., 2006
). These results suggest that younger subjects have limitations in reward assessment that may underlie their apparent under-activity of rewards when valence is harder to assess.
Galvan et al. (2006)
using fMRI investigated BOLD differences in subjects performing a rewarded match-to-sample paradigm. Briefly, subjects saw one of three different visual cues (pictures of cartoon pirates) presented to the left or right of fixation, each of which was associated with a distinct reward value (different amounts of money). Following a brief delay, subjects saw two images of treasure chests to the left of right of fixation and were instructed to select (via button press and within 2 s) which chest appeared on the same side as the previous pirate picture. Subjects were then given feedback indicating if and how much they had won. Adolescents demonstrated an exaggerated response (higher magnitude of BOLD response) in NAcc relative to children or adults during the reward receipt epoch for large rewards. Furthermore, the extent (number of significantly active voxels) of NAcc activity in adolescents looked more like adults than children, overall. In OFC, adolescents looked more like children in terms of both extent and magnitude of activation. Results from this study were interpreted as reflecting a protracted development of OFC relative to NAcc and suggest that adolescents have limitations in the executive assessment of rewards and an overactive reward system.
Collectively, the studies suggest that the predictions of the hypo-and hyper-active models may not be mutually exclusive. For instance, Bjork et al. (2004)
found under-activity in ventral striatum during a period when adolescents anticipated responding
for rewards. This is a temporally distinct phase of incentive processing than that explored by Ernst et al. (2005)
and Galvan et al. (2006)
, studies which report adolescents had increased activity when receiving
reward. Thus, an important factor contributing to the hypo- versus hyper-active distinction may be the temporal stage of incentive processing under scrutiny—that is, distinct phases of incentive processing result in different patterns of activations.
Interestingly, Bjork et al. (2004)
did not observe significant differences in the ventral striatum between adolescents and adults performing the MID task during reward receipt, an epoch more directly comparable with Ernst et al. (2005)
. One factor that may underlie these contradictory results is a difference in the levels of cognitive load demanded by the different tasks. Bjork et al. (2004)
used a simple reaction time task where subjects simply responded to the appearance of a target, while the paradigms used by Galvan et al. (2006)
and Ernst et al. (2006)
required that subjects assess different responses and invoke working memory for instructions and past performance. More cognitively demanding tasks have been shown to recruit additional brain areas and/or increased activity within a single area (Rubia et al., 2000
) and may increase the likelihood of recruiting reward-related brain areas.
Finally, we note that conclusions based on comparison of BOLD responses across different age groups are a common concern. The challenge put forth by neuroscientists investigating the adult system is that it is not straightforward if BOLD activity changes in fMRI studies are due to actual differences in neuronal computations or an isolated artifact due to immaturities in the vasculature or gross head size differences. Counter to these arguments, however, we note that brain size is adult-like early childhood (see Brain Maturation during Adolescence, below) and that the feasibility of comparing BOLD responses across developmental age groups transformed into a common stereotaxic space has been well established (Brown et al., 2005
; Kang et al., 2003
; Wenger et el., 2004
). An additional concern is that performance differences in the scanner may lead to different levels or patterns of BOLD activity. We agree that this may be an effect in some studies. However, pediatric imaging studies frequently employ simple tasks easily performed by children (Luna et al., 2004a
; Galvan et al., 2006
) minimizing performance differences. Furthermore, when performance is equated across age groups (Bjork et al., 2004
; Schlaggar et al., 2002
), age-related functional differences are still observed.
Below, we next address why adolescents may demonstrate these particular patterns of functional brain activity—that is, what underlying brain mechanisms support these types of responses? From adolescence to adulthood, important brain structural and physiological changes occur with significant effects on brain function. Differences in brain maturational state, including thinning gray matter (e.g., synaptic pruning), increases in white matter (e.g., myelination), and neurotransmitter system differences, likely contribute to the particular functional patterns observed in adolescents and adults and are examined below.