The developmental psychopathology of motivation in adolescence

Abstract

Adolescence is a key period for the emergence of psychopathology, with many psychiatric disorders having their modal age-of-onset during this period. Relative to other periods of the lifespan, susceptibility to a number of psychiatric disorders is greatest during adolescence, particularly in females. In addition, disorders which emerge during adolescence appear to be more enduring and serious than those with a later onset. Although these psychiatric conditions may appear different from each other in terms of their associated behavioral signs or symptoms, this review will argue that they involve common alterations in motivational processes or disturbances in reward processing, although the direction of such changes (hypersensitivity vs. hyposensitivity to reward) and the stage of processing affected (reward anticipation vs. receipt) may differ across broader groupings of disorder. Recent behavioral, neuropsychological and neuroimaging research on reward processing in children, adolescents, and adults with these conditions will be described and evaluated. In addition, this article will consider what these studies tell us about their etiology and highlight gaps in our knowledge base. The review will also attempt to explain why adolescence is a period of elevated risk for the development of psychopathology, by discussing normative changes in reward processing in humans and animals.

1. Introduction

Recent epidemiological studies have documented dramatic increases in the prevalence of psychiatric disorders during the adolescent period. Although the causes of these increases are likely complex and multifactorial, it nevertheless appears that mid-adolescence is a key period in terms of vulnerability to psychiatric illness. As will be discussed below, these increases in prevalence appear greatest for mental disorders in which dysfunctional motivation or reward-related processes are implicated. It is also notable that normative developmental research has shown that there are alterations in motivational processes during the adolescent period, which may lead to heightened reward-seeking and risk-taking behavior at the population level. This article will begin by reviewing epidemiological data that illustrate the scale of the shifts in prevalence of mental illness during the teenage years and highlight the need to focus on the adolescent period. It will continue by describing the evidence for changes in the function of motivational circuitry during adolescence, by considering studies of motivation and reward processing in typically developing adolescent humans and animals. It will then relate this evidence to what is known about deficits or changes in motivational processes in common mental disorders, focusing particularly on disorders which show marked increases in prevalence during adolescence (depressive, conduct and substance use disorders). A greater understanding of these processes is critical, not just because it is important to help adolescents to successfully deal with this period of transition, but because although many of these mental disorders onset in adolescence, frequently these patterns of illness continue into adulthood. As such, enduring patterns of psychopathology often originate in adolescence, and those that do emerge during this period tend to be more severe and resistant to treatment. Consequently, this work has implications for understanding mental health across the lifespan.

2. The epidemiology of psychiatric illness in adolescence and age-related transitions in vulnerability

The results of the National Comorbidity Survey (NCS), a US population-based study of 9282 adults, showed that the median age of onset for any psychiatric disorder was 14 years (Kessler et al., 2005). This figure masked substantial variation between different classes of disorders: the median age of onset for both anxiety disorders and impulse control disorders was 11 years, whereas it was 20 years for substance use disorders and 30 years for mood disorders. While these figures are likely to have been influenced by inherent restrictions in the DSM-IV criteria (for example, in relation to the age-of-onset figure for impulse control disorders, it should be noted that attention deficit/hyperactivity disorder cannot be validly diagnosed if the symptoms emerge after age 7 years, and conduct disorder and oppositional defiant disorder are not normally diagnosed after age 18 years because they are assumed to convert into antisocial personality disorder in adulthood), it is nevertheless clear from these findings that many psychiatric disorders originate in childhood or adolescence. Further limitations of the NCS include the fact that it was cross-sectional and therefore relied on retrospective reports of age-at-onset information, and all participants were interviewed in adulthood. Consequently, there may have been a substantial time interval between the onset of the mental disorder and the diagnostic interview enquiring about its age-of-onset. Finally, it is well-established that resolved mental disorders reported in baseline interviews are less likely to be reported in follow-up interviews. All of these issues mean that the prevalence of mental disorders is likely to have been underestimated in the NCS and that, by studying adults alone, substantial error may have been introduced into the age-of-onset figures for different psychiatric disorders.

To address some of these methodological problems and gain more accurate information about the age-of-onset of common psychiatric disorders and the burden of psychiatric illness in adolescence, researchers at the National Institute of Mental Health recently conducted a large, nationally representative study investigating age-related transitions in the prevalence of mental disorders during adolescence, the National Comorbidity Survey-Adolescent Supplement (NCS-A). They interviewed 10,123 adolescents aged between 13 and 18 years, and found that almost half of the sample had at least one DSM-IV disorder, one-fifth of the sample had a mental disorder that had caused severe impairment, and 40% of those with any psychiatric disorder had at least one further comorbid disorder (Merikangas et al., 2010). The study also revealed striking changes in the prevalence of mental disorders across adolescence. The prevalence of major depressive disorder increased two-fold over the adolescent period, whereas rates of substance use disorders were six times higher in 17–18-year olds relative to 13–14-year olds. As such, this study convincingly demonstrated that adolescence is a key period for the emergence of psychopathology. It also provided information about the scale of the problem, underlining the need for improved understanding of the processes that lead to increased vulnerability to mental illness in adolescence.

In addition to this evidence that risk for psychopathology is at its highest in adolescence, a number of studies have suggested that disorders that emerge during adolescence are frequently more enduring and serious than those which first onset during adulthood. For example, individuals who begin using substances during adolescence are more likely to become addicted and display higher rates of substance use than those who start using substances during adulthood (Anthony and Petronis, 1995, Taioli and Wynder, 1991).

Finally, epidemiological studies highlight the importance of paying attention to sex differences in vulnerability to psychiatric illness, and the potential role of puberty in mediating some of these effects. In the Great Smoky Mountains Study, rates of major depression were similar in males and females up to age 12, but then diverged markedly such that depression became three times more common in females than males by age 16 (Angold and Costello, 2006). This increase in diagnosable depression was preceded by elevations in self-reported depression scores in girls from age 12 onwards (Angold et al., 2002). Early puberty is associated with increased risk for a number of negative outcomes, particularly in girls. Relative to those who undergo puberty on time or late, girls who experience early menarche are at higher risk for depression (Ge et al., 2003, Graber et al., 2004, Kaltiala-Heino et al., 2003), substance use disorders (Costello et al., 2007, Deardorff et al., 2005), and externalizing disorders (Caspi et al., 1993, Ge et al., 2002, Obeidallah et al., 2004). This is commonly viewed to be a consequence of appearing chronologically older and therefore becoming involved with older peers and romantic partners, but it is also possible that this is partly a consequence of earlier increases in gonadal steroids which alter the sensitivity of motivational circuitry in the brain.

Complementing these epidemiological studies using psychiatric diagnoses, a large number of criminological studies have revealed marked age-related transitions in engagement in criminal behavior. The prototypical pattern is a peak in the age-crime curve during adolescence – that is, individuals are more likely to engage in criminal behavior in mid- to late-adolescence than at any other period of the lifespan (Moffitt, 1993). This adolescent peak in the age-crime curve holds across countries and cultures, and is observed for multiple forms of criminal behavior including violent assault and property crimes (Hirschi and Gottfredson, 1983). Again, psychosocial processes such as social mimicry of deviant peers have been invoked to explain such changes (Moffitt, 1993), but factors within the individual such as alterations in the function of reward circuitry during this period may also contribute to these increases in delinquent behavior during adolescence. Another extremely robust finding from the criminological literature is that there are sex differences in criminal behavior, with males being substantially more likely than females to be convicted of a criminal offence (Moffitt et al., 2001). Sex differences in antisocial behavior are also observed in psychiatric epidemiological studies, with males being far more likely than females to meet diagnostic criteria for conduct disorder (Moffitt et al., 2001).

In summary, epidemiological and criminological studies have documented marked increases in vulnerability to psychopathology during the adolescent period. An improved understanding of the causes of these phenomena will be essential if we are to develop better treatments and preventive interventions. The next section will discuss relevant animal research investigating motivation and reward processing in adolescence.

3. Animal models of reward processing in adolescence

Evidence is accumulating that sensitivity to a range of social and non-social rewards is heightened during adolescence in animals. Relative to adults, adolescent rats engage more frequently in social interactions such as play-fighting, and display increased levels of novelty-seeking, risk-taking, and consumption of appetitive stimuli (Doremus-Fitzwater et al., 2010). For example, adolescent rats show an enhanced preference for spatial locations previously associated with a social partner compared with adults (Douglas et al., 2004). They also exhibit an increased preference for locations previously paired with novel objects (Douglas et al., 2003). Studies have demonstrated that the motivational value of palatable foods also peaks in adolescence. For example, rats consume significantly more sweetened milk and will work harder to receive it on post-natal day 50 (mid- to late-adolescence) than at any other point in the lifespan (Friemel et al., 2010). In a separate study that involved administering sucrose and quinine solutions to adolescent and adult rats, the adolescents were more sensitive than adults to the hedonic properties of sucrose and less sensitive to the aversive properties of quinine (Wilmouth and Spear, 2009).

Animal studies involving administration of psychotropic drugs during adolescence have also enhanced our understanding of alterations in reward processing, in addition to helping us characterize the underlying basis of the increased vulnerability to substance use disorders that occurs during this period (Schramm-Sapyta et al., 2009). This work has shown that, relative to adults, adolescent rats are generally more sensitive to the rewarding effects of substances (Belluzzi et al., 2004, Philpot et al., 2003, Torres et al., 2008) and less likely to experience negative side-effects when using substances (Schramm-Sapyta et al., 2006, Wilmouth and Spear, 2004). In addition, adolescent animals self-administer larger amounts of substances than adults (Doremus et al., 2005, Levin et al., 2003). There also appear to be qualitative differences between adolescent and adult rats in terms of the subjective effects of substances: for example, nicotine has anxiolytic effects in adolescent rats, but is anxiogenic in adults (Elliott et al., 2004). Finally, adolescent rats appear to experience less severe withdrawal effects than adults when the addictive substances they have been using are no longer available (O’Dell et al., 2004, O’Dell et al., 2006, Varlinskaya and Spear, 2004).

In summary, animal studies involving natural rewards and the administration of drugs of abuse demonstrate increased sensitivity to hedonic stimuli and enhanced motivation to work to receive appetitive stimuli in adolescence, relative to other periods of the lifespan. Adolescent rats also appear less sensitive than adults to aversive stimuli or the negative effects of psychotropic substances. These findings are important because they provide a potential mechanistic explanation for the observation that human adolescents show increased vulnerability to developing substance use disorders relative to adults. However, more work is needed to investigate whether these animal results translate to human adolescents, and to understand how these neurobiological influences interact with psychosocial and cultural factors in the genesis of human addiction.

The following section will consider behavioral and neuroimaging studies of reward processing across normative development to examine what they tell us about changes in vulnerability to psychopathology across the lifespan.

4. Normative studies of reward processing in human adults and adolescents

In order to place recent work on reward processing in neuropsychiatric disorders in context, it is first necessary to briefly review neuroimaging work on reward processing in healthy adults and adolescents. Early functional magnetic resonance imaging (fMRI) studies using healthy volunteers revealed activity in orbitofrontal cortex (OFC), ventral striatum, midbrain, and amygdala during the processing of intrinsically rewarding stimuli (e.g., attractive faces) and secondary rewards such as monetary gains (O’Doherty et al., 2001, O’Doherty, 2004). Subsequent work using event-related fMRI and experimental designs capable of discriminating between the neural correlates of reward anticipation and reward outcome suggested that different components of this circuit are activated during the anticipatory (e.g., ventral striatum) vs. the consummatory (medial OFC) components of reward processing (Knutson et al., 2001a, Knutson et al., 2001b, Knutson and Cooper, 2005).

The monetary incentive delay (MID) task (Knutson et al., 2000) has been widely used to probe reward-related brain function in both the normative and the psychiatric literature, and is worth briefly discussing here. It involves the presentation of visual cues signalling the opportunity to gain a monetary reward, avoid a monetary loss, or receive a control outcome (no gain or loss). The amounts available to be won or lost are generally systematically varied in magnitude and the subject's level of excitement prior to receiving the outcome is monitored in some versions of the task. The subject subsequently has to make a motor response to a target within a set interval. Finally, after a delay, the subject receives an outcome, which could either be rewarding, punishing, or neutral, and the value of his or her monetary endowment is updated. Strengths of the MID task include its ability to reliably elicit activation in reward-related circuitry, its event-related design which permits measurement of key motivational processes (reward anticipation), and its capacity to assess the relationship between subjective aspects of reward processing (e.g., excitement) and neural activity during the task. Furthermore, later versions of the MID task (e.g., Bjork et al., 2010b) permit the separation of different phases of reward processing by interposing a gap of several seconds between the presentation of cues predicting rewards and the receipt of rewarding outcomes. The MID task is, however, subject to a number of limitations: since the contingencies between the visual cues and the monetary amounts available to be gained or lost are pre-trained before the actual experiment, the MID task is not suitable for assessing reward learning (e.g., forming associations between neutral conditioned stimuli and unconditioned appetitive stimuli in an associative learning context). As subjects have to make a speeded button press to obtain rewards, the reward anticipation phase of the task is confounded by a motor preparatory component. The MID task has also been criticized for lacking ecological validity, although this is a general problem with most fMRI experiments. Lying in an fMRI scanner is an unpleasant and even frightening experience for some participants, and this may make it difficult to validly study reward-related brain processes, particularly in children or adolescents. Finally, because the MID task does not involve making choices (e.g., between options varying in risk or value) it may not accurately reflect the complexity of the contexts in which humans normally seek out rewards.

Given the increase in risk-taking and susceptibility to psychiatric illness and substance use disorders in adolescence, a number of recent studies have attempted to investigate whether there are differences between adolescents and adults in reward processing. Some of this work has been motivated by theories such as the triadic model of motivated behavior in adolescence (Ernst et al., 2006). This model suggests that there is an imbalance between neural systems mediating approach and avoidance in adolescence, such that approach- or reward-related brain circuitry is hypersensitive, whereas avoidance-related brain systems are relatively insensitive, in comparison with the level of sensitivity of these systems in adulthood. As well as this shift in the balance between approach and avoidance towards an enhanced tendency to approach rewarding stimuli, adolescents are argued to possess an immature regulatory system, potentially due to the slow development of the prefrontal cortex (particularly the anterior cingulate cortex) relative to other cortical regions. Although changes in anatomical connectivity over the adolescent period were not originally discussed in relation to the triadic model, we now know that the strength of frontal white-matter pathways increases linearly with age during adolescence, particularly in males (Perrin et al., 2008, Perrin et al., 2009). Consequently, immaturity of anatomical tracts connecting the prefrontal cortex and limbic structures involved in emotional behavior may contribute to the deficits in cognitive control or emotion regulation observed in adolescence. In addition to social and environmental factors, such as reductions in parental supervision and an increase in the availability of intoxicating substances with advancing age, this neural shift towards increased sensitivity to rewards and reduced sensitivity to aversive or punishing stimuli, coupled with poor regulatory capacity, may make the individual more vulnerable to impulsive or risky behavior.

Behavioral studies comparing the performances of adolescents and adults on emotional decision-making tasks, or evaluating the effects of increased motivation (for example, providing monetary rewards for good performance), have revealed that adolescents are more sensitive than adults to cues of reward or changes in motivational context. For example, a recent study that examined age effects on performance of a modified version of the Iowa Gambling Task, which is an index of decision-making under uncertainty, found that adolescents were more likely than adults to play from the advantageous decks (this was interpreted as evidence that sensitivity to rewards peaks in mid-adolescence), whereas adults were faster to learn to avoid playing from the disadvantageous decks (Cauffman et al., 2010). It should be noted that the Iowa Gambling Task has been criticized as a probe of decision-making due to its cognitive complexity; an apparent ‘deficit’ on the task could reflect a problem with short-term memory (failing to hold in mind which decks are associated with negative outcomes) or response reversal, or be underpinned by hypersensitivity to reward or hyposensitivity to punishment (Dunn et al., 2006). Individuals who enjoy taking risks may select the high-risk decks because they find the uncertainty attractive or rewarding. Finally, although the task is intended to simulate real-life decision making, most versions of the task do not involve playing for real money and thus apparent deficits in performance could reflect individual differences in intrinsic motivation or apathy.

Using an alternative gambling task in which subjects received immediate feedback about their decisions (the ‘hot’ or affective condition) or had to wait until the end of the trial to receive feedback (the ‘cold’ or deliberative condition), Figner et al. (2009) found that adolescents showed elevated risk-taking relative to adults, but only in the affective condition. Adolescents were also more sensitive to the motivational effects of potential monetary rewards on performance of an antisaccade task (Jazbec et al., 2006). While adolescents were generally less accurate than adults on the task, their performances normalized to adult levels when monetary incentives were available. This finding was recently replicated by Geier et al. who showed that adolescents made fewer inhibitory errors on rewarded relative to neutral antisaccade trials, whereas adults’ error rates were not significantly modulated by the presence of incentives (Geier et al., 2010).

Normative functional neuroimaging studies have revealed differences between adults and adolescents in reward processing, although the direction of such differences has not been consistent across studies. In one of the first developmental studies of this kind, Bjork et al. (2004) used the MID task to monitor brain activity during reward anticipation and outcome in healthy adolescents and adults. While adolescents and adults largely showed similar patterns of brain activity in both conditions, adolescents displayed less ventral striatal activity than adults during reward anticipation. This group subsequently replicated their finding of weaker ventral striatal activity during reward anticipation in adolescents compared with adults using a larger sample (Bjork et al., 2010b). In contrast with these reports, however, two studies found increased striatal responses to rewards in adolescents relative to adults (Ernst et al., 2005, Galvan et al., 2006). Most recently, Van Leijenhorst et al. (2010) assessed reward-related processing in a large group of children, adolescents, and young adults using a slot machine task in which rewards were received on a probabilistic basis, independent of the participants’ choice behavior. The adolescents showed larger ventral striatal responses than adults or children when receiving monetary rewards, whereas no group differences in this region were observed during the anticipation of rewards (Van Leijenhorst et al., 2010). These latter studies suggest that adolescent risk-taking may be a result of hypersensitive reward circuitry in the context of developmental delays in cognitive control, consistent with the triadic model.

In addition to these studies assessing reward processing using tasks that do not involve an explicit learning component, a recent study investigated reward processing in children, adolescents, and adults using a probabilistic learning task (Cohen et al., 2010). The authors fitted computational models to the resulting fMRI data to examine whether there were age-related changes in the magnitude or location of decision value or prediction error signals (phasic signals encoding a discrepancy between expected and actual outcomes, which are thought to play an important role in reinforcement learning and are particularly evident in dopaminergic projection regions such as the ventral striatum, midbrain, and orbitofrontal cortex). Relative to children and adults, adolescents displayed enhanced prediction error signals in the striatum and angular gyrus. Furthermore, adolescents and adults also differed in the location of the prediction error signal: in adolescents, this occurred in dorsal striatum (caudate nucleus), whereas in adults it was observed in the ventral putamen. Interestingly, children showed no reliable prediction error signal in the striatum. Thus, compared with adults, adolescents appear to show an enhanced neural response in reward-related circuitry when positive outcomes are experienced, particularly if those outcomes are better than expected (i.e., a larger positive prediction error). These apparent developmental differences in the neural correlates of reward learning may indicate that dopaminergic activity related to prediction error signals is greater in adolescents than in adults or children. Consequently, adolescents may learn more rapidly when they receive rewards and find rewarding cues or events highly salient.

This overview of normative research on developmental effects on reward processing raises a number of questions. First, what accounts for the inconsistency between studies showing deficient ventral striatal activity in adolescents relative to adults and studies reporting hypersensitive reward circuitry in adolescents? Possible explanations include differences between studies in terms of what constitutes ‘adolescence’ (pre-pubertal teenagers vs. late adolescents), and the use of different tasks which may have been optimized to measure anticipatory vs. consummatory components of reward processing, or may have confounded these components. This is problematic because reward anticipation and reward consumption may show distinct developmental trajectories across adolescence. Second, what are the implications of the fMRI results discussed above for understanding the heightened risk-taking and reward-seeking behavior observed in adolescence? Attenuated ventral striatal activity during reward anticipation could potentially lead to increased risk-taking and reward-seeking behavior in adolescents, to compensate for a reduction in sensitivity to rewards of small or moderate value. Alternatively, increased ventral striatal responses to rewarding outcomes or enhanced prediction error signals in adolescence could be interpreted as promoting heightened approach or reward-seeking behavior in this age group. Rather than assuming that a given sample of adolescents will show higher levels of risk taking behavior than adults or children, future studies should attempt to relate individual differences in risk-taking and reward-seeking behavior with neural activity during the anticipation and receipt of rewards. Furthermore, longitudinal studies of reward processing across adolescence are likely to be highly informative in resolving these questions, especially if they show that neural sensitivity to rewards is greatest at the same time as the individual shows a peak in their risk-taking or impulsive behavior.

5. Sex differences in reward processing and effects of the menstrual cycle

As noted in the epidemiology section (Section 1), there are marked sex differences in the prevalence of several psychiatric illnesses, with females being disproportionately more likely to develop internalizing disorders compared with males. However, this pattern appears to be linked with pubertal or maturational processes, as it only becomes evident from early adolescence onwards. In addition, early onset of puberty is a risk factor for multiple negative outcomes including depression. What is the source of these apparent puberty-related changes in vulnerability to depression?

There is evidence from neuroimaging studies that females show changes in the sensitivity of reward-related brain circuitry across the menstrual cycle. For example, healthy female volunteers show greater frontostriatal responses to monetary rewards in the follicular phase of the cycle, when the effects of estrogen are unopposed by progesterone, than in the luteal phase when progesterone levels are high (Dreher et al., 2007). Psychopharmacological studies have demonstrated that female rats show increased rates of cocaine self-administration (Jackson et al., 2006) and female humans display enhanced subjective responses to cocaine (Evans et al., 2002) or amphetamine (Justice and de Wit, 1999) in the follicular, relative to the luteal, phase of the menstrual cycle.

These shifts in the sensitivity of midbrain and PFC areas to rewards and hedonic properties of drugs across the menstrual cycle are observed in adult females with established cycles, but in female adolescents undergoing puberty, the different phases of menstruation may be shorter, extended in time, or only experienced irregularly (Golden and Carlson, 2008). In most cases, the interval between the onset of menarche and the establishment of regular menstrual cycles is approximately 5 years (Golden and Carlson, 2008). Assuming that fluctuations in concentrations of gonadal steroids, such as estrogen, exert similar effects on reward processing in juveniles, it is possible that irregular menstrual cycles may contribute to the dysregulation of reward processing in female adolescents.

This review has now considered relevant findings from animal work, normative research with humans, and the potential modulatory effects of gonadal steroids on reward processing. In the next section, I will discuss studies showing that psychiatric disorders that typically emerge during adolescence are associated with abnormal behavioral responses to reinforcement and dysfunction in brain reward systems.

6. Reward-related brain activity in psychiatric disorders emerging in adolescence

Research over the last decade has provided a number of insights into the neural bases of the internalizing and externalizing disorders that show age-related increases during adolescence. Relevant to this review, alterations in the sensitivity of reward-related circuitry have been reported in many of these conditions. Notably, common neural circuits are implicated, but the direction of the effects and the stages of reward processing affected appear to differentiate between different classes of mental disorder. Externalizing disorders and substance use disorders, which share common genetic risk factors (Kendler et al., 2003), appear to be associated with deficits in the activity of motivational circuitry during anticipation of rewards, whereas enhanced responses to rewarding outcomes are observed in some externalizing disorders. In contrast, internalizing disorders such as depression are associated with impaired striatal responses to rewarding outcomes. Table 1 summarizes the key findings from functional neuroimaging studies of several common externalizing and internalizing disorders that show increases in prevalence during adolescence or are first diagnosed in children and adolescents. Each of the disorders included in Table 1 will now be considered separately.

Table 1.

Summary of reward processing findings from fMRI studies of psychiatric disorders.

Group comparison	Reward anticipation	Reward outcome
Depressed adults vs. controls (Epstein et al., 2006, McCabe et al., 2009, Pizzagalli et al., 2009, Smoski et al., 2009, Steele et al., 2007, Surguladze et al., 2005; but see Knutson et al., 2008)	↓ Putamen	↓ Ventral striatum, ↓ Caudate
Depressed children vs. controls (Forbes et al., 2006, Forbes et al., 2009)	↓ Caudate	↓ Caudate, ↑ Medial OFC
Familial risk for depression vs. controls (Gotlib et al., 2010, Monk et al., 2008)	↓ Putamen	↓ Putamen, ↓ Lentiform nucleus (striatum)
Adults with antisocial personality disorder vs. controls (Vollm et al., 2010)	Not evaluated	↑ Subgenual cingulate, ↑ OFC
Adolescents with externalizing disorders vs. controls (Bjork et al., 2010a, Bjork et al., 2010b, Rubia et al., 2009)	↔ Ventral striatum, ↔ OFC	↑ Ventral striatum, ↑ Subgenual cingulate, ↓ OFC
ADHD adults vs. controls (Plichta et al., 2009, Strohle et al., 2008)	↓ Ventral striatum	↑ OFC
ADHD children vs. controls (Scheres et al., 2007)	↓ Ventral striatum	Not evaluated
SUD adults vs. controls (Buhler et al., 2010, David et al., 2005, Goldstein et al., 2007, Wrase et al., 2007)	↓ Ventral striatum	↑ Ventral striatum, ↑ Subgenual cingulate
SUD adolescents vs. controls (Peters et al., 2011)	↓ Ventral striatum	↔
Maltreated adults vs. controls (Dillon et al., 2009)	↓ Globus pallidus	↔
Maltreated adolescents vs. controls (Mehta et al., 2010)	↓ Ventral striatum	Not evaluated

Key: ↑ or ↓ activity in this region is increased or reduced, respectively, relative to the comparison group; ↔ indicates that there was no significant group difference for this condition; ADHD, attention-deficit/hyperactivity disorder; OFC, orbitofrontal cortex; SUD, substance use disorder.

6.1. Depression

Impairment in the ability to experience pleasure or reward (i.e., anhedonia) is a key symptom of major depressive disorder, and several studies have attempted to assess this phenomenon experimentally. In particular, recent work has investigated whether anhedonia can be demonstrated using laboratory-based behavioral tasks. There have also been concerted efforts to characterize the underlying neural and neurochemical basis of anhedonia.

A number of behavioral studies have revealed differences between depressed adults and control subjects in reward-related decision-making. Specifically, adults with depression appear less sensitive to the effects of potential reward on behavioral performance or when selecting between different options (Henriques and Davidson, 2000, Pizzagalli et al., 2005). Similar patterns of reduced sensitivity to high magnitude rewards during decision-making have been observed in depressed children and adolescents (Forbes et al., 2007), and appear to predict depressive symptoms later in adolescence (Forbes et al., 2007). These results could reflect an attenuation of normal reward seeking in depression or those at increased risk for developing depression.

There is convincing evidence for structural deficits in brain regions involved in processing reward-related information in major depressive disorder; such changes appear most pronounced in patients with early-onset depression (defined as onset before age 18). For example, in a recent large (n = 289) voxel-based morphometry (VBM) study, adults with depression (with or without comorbid anxiety disorders) were found to show gray matter volume reductions in subgenual and dorsal anterior cingulate cortex compared with controls (van Tol et al., 2010). Interestingly, this effect was primarily driven by volumetric reductions in subgenual anterior cingulate extending into medial orbitofrontal cortex in the subgroup of patients who reported an onset of depression prior to age 18. This latter finding is consistent with the results of an earlier study that reported reduced subgenual cingulate cortex volume in late adolescent and young adult females with depression relative to controls (Botteron et al., 2002). A recent meta-analysis of structural neuroimaging studies concluded that anterior cingulate and OFC volumes were robustly reduced in major depressive disorder (Koolschijn et al., 2009). This meta-analysis also found significant reductions in putamen and caudate nucleus volume in depressed individuals compared with controls, although these differences had only moderate or small effect sizes. Several VBM studies published since this meta-analysis have also reported reductions in striatal gray matter volume in depression (Kim et al., 2008, Wagner et al., 2011). Consequently, structural abnormalities in frontostriatal circuits are associated with depression, and may partly underpin the anhedonic symptoms observed in this condition.

Complementing these structural findings, functional neuroimaging studies have demonstrated reduced activity in reward-related regions in depression using a variety of experimental paradigms. For example, relative to control subjects, depressed adults were reported to show reduced responses in the ventral striatum and caudate nucleus when processing positive compared to neutral words (Epstein et al., 2006). There was also a negative correlation between the magnitude of ventral striatal responses to positive words and anhedonic symptoms in the depressed group. In a facial emotion processing study, Surguladze et al. (2005) found that whereas control subjects showed a positive relationship between striatal activation and increasing levels of happiness in facial expressions, depressed adults failed to display this pattern. Using a guessing game task that provided subjects with positive and negative feedback, it was reported that depressed adults show blunted responses to positive feedback in ventral striatum relative to controls (Steele et al., 2007). While the control subjects in that study showed faster reaction times on the next trial following positive feedback, the depressed patients did not alter their reaction times in subsequent trials when given positive feedback. Using an instrumental learning task and applying a computational model of reinforcement learning to the resulting fMRI data, the same group found that depressed individuals showed reduced prediction error signals in the striatum and midbrain (Gradin et al., 2011). Furthermore, there was a negative correlation between anhedonic symptoms in the depressed patients and the magnitude of the prediction error responses in these regions, providing evidence for a link between reduced dopaminergic activity and anhedonia. In a study involving the presentation of pleasant (chocolate) taste stimuli, adults in remission from depression showed lower ventral striatal and subgenual anterior cingulate responses to the pleasant stimuli (McCabe et al., 2009). Relative to controls, the subjects in remission from depression also showed reduced responses to the combined taste and sight of chocolate in the ventromedial prefrontal cortex. It should be noted that Knutson et al. (2008) found no differences between unmedicated depressed patients and control subjects in ventral striatal responses to reward anticipation using the MID task. In contrast, a study involving a decision-making task designed to probe brain activity during separate stages of reward selection, anticipation and feedback found reduced striatal activity during all stages of reward processing in patients with depression (Smoski et al., 2009). Lastly, in an important study, Pizzagalli et al. (2009) used the MID task to probe neural correlates of reward anticipation and outcome in adults with depression and healthy subjects. Relative to healthy controls, depressed subjects showed less activity in caudate nucleus and ventral striatum when experiencing rewarding outcomes (Pizzagalli et al., 2009). In the same study, structural MRI analyses revealed that caudate nucleus volumes were negatively correlated with both total scores and anhedonia subscale scores on the Beck Depression Inventory.

The majority of these studies indicate that the sensitivity of reward-related brain circuitry is attenuated in adult depression, and particularly forms of depression accompanied by clinically significant anhedonia. In addition, there is evidence that deficits in ventral striatal activity may be either a trait marker or an endophenotype for depression, since they are also observed in individuals who have recovered from depression. Consistent with this view, it has been argued that anhedonic symptoms, and by extension brain reward system abnormalities, represent promising endophenotypes for genetic and neuroimaging studies of depression (Hasler et al., 2004). Although less neuroimaging work has been performed with children or adolescents with depression, Forbes and colleagues have performed some highly informative studies in this area. In the first of these investigations, they observed reduced caudate nucleus and orbitofrontal cortex activity in depressed adolescents during reward-related decision making and the receipt of rewarding outcomes, relative to control subjects (Forbes et al., 2006). In an important extension of this work, Forbes et al. (2009) reported that depressed adolescents display reduced striatal activity during both the anticipation of rewards and the receipt of rewarding outcomes, relative to healthy controls. Furthermore, striatal activity was positively correlated with the participants’ ratings of positive affect in real-life situations, as measured using ecological momentary assessment. This work demonstrates that, although they can be criticized as lacking ecological validity, brain imaging measures of reward processing are capable of predicting the individual's ability to experience positive affect under naturalistic conditions. Interestingly, as well as adolescents with formal diagnoses of depression, girls at high familial risk for depression, who have never been depressed themselves, show disruptions in reward-related brain activity (Gotlib et al., 2010). Furthermore, a study by Monk et al. (2008) observed reduced nucleus accumbens responses to happy faces in adolescents at high familial risk for depression, relative to adolescents with no family history of depression. These latter results provide further evidence that impairments in the neural processing of reward may be a stable risk marker (or endophenotype) for depression that increases vulnerability in a probabilistic fashion.

Finally, a recent study that involved giving subjects positive and neutral social feedback, rather than monetary incentives, found increased amygdala responses to positive social feedback in depressed adolescents relative to controls (Davey et al., 2011). When considering both positive and control feedback conditions together, relative to a low-level baseline, the depressed adolescents showed increased responses in inferior prefrontal cortex, subgenual anterior cingulate cortex and insula compared with controls. This work suggests that depressed adolescents are more sensitive to social evaluation than controls.

6.2. Conduct disorder/oppositional defiant disorder

Neuropsychological studies have revealed that externalizing disorders such as conduct disorder or oppositional defiant disorder are associated with a heightened sensitivity to rewards and a reduced sensitivity to punishment (Fairchild et al., 2009, Matthys et al., 2004, van Goozen et al., 2004). Essentially, children and adolescents with these disorders appear to show an exaggerated form of the imbalance between approach and avoidance systems observed in healthy adolescents, as encapsulated in the triadic model (Section 4). Of interest, this behavioral alteration in sensitivity to positive and negative reinforcement may be partly driven by enhanced heart rate responses to rewards and blunted cardiac responses to losses in those with externalizing disorders (Luman et al., 2010a). Furthermore, a recent study using a variant of the Iowa Gambling Task found that differences in reward-related decision making were most pronounced in adolescents with conduct disorder and comorbid substance dependence, relative to controls and subjects with non-comorbid conduct disorder (Schutter et al., 2010). Since substance use disorders are often comorbid with externalizing disorders and may form part of an underlying externalizing spectrum (Kendler et al., 2003, Krueger et al., 2002, Krueger et al., 2005), this may reflect an effect of severity of externalizing behavior (those with both conduct disorder and substance dependence could be simply further along the externalizing spectrum), although it could also be interpreted as a consequence of chronic exposure to psychotropic substances on brain function.

Research investigating brain structure in externalizing disorders has demonstrated lower OFC gray matter volume in adolescents with conduct disorder and comorbid ADHD (Huebner et al., 2008). This is of interest because the orbitofrontal cortex is involved in representing the value of primary and secondary rewards (O’Doherty et al., 2001, O’Doherty, 2004). Studies have also shown reduced amygdala and anterior insula volumes in adolescents with conduct disorder (Fairchild et al., 2011, Huebner et al., 2008, Sterzer et al., 2007), which may contribute to their reduced sensitivity to aversive stimuli, as the amygdala is involved in mediating avoidance behavior whereas the insula in involved in the processing of risk and negatively valenced stimuli (Preuschoff et al., 2008). Recently, it has been reported that adult psychopaths show increased striatal volumes relative to control subjects (Glenn et al., 2010), while callous-unemotional personality traits were positively correlated with striatal gray matter volume in a group of adolescents with conduct disorder (Fairchild et al., 2011). These latter studies suggest that psychopathic or callous-unemotional traits, which are thought to designate a distinct subgroup of individuals with externalizing behavior, may be associated with structural abnormalities in reward systems which predispose to heightened reward-seeking behavior.

The fMRI literature on reward processing in externalizing disorders provides evidence for changes in reward-related activity, although the direction of the effects has not been wholly consistent across studies. Using a modified continuous performance task, adolescents with non-comorbid conduct disorder showed reduced OFC responses to infrequent monetary rewards relative to control subjects (Rubia et al., 2009), whereas a more recent study using the MID task found that a mixed group of adolescents with externalizing disorders showed enhanced subgenual cingulate cortex responses to monetary reward outcomes compared with controls (Bjork et al., 2010a). Interestingly, and in contrast to the situation observed in pure ADHD (see below), the externalizing and control adolescents showed similar neural responses during reward anticipation on the MID task. Consistent with this latter study, adults with antisocial personality disorder were found to show increased subgenual cingulate and OFC activation during receipt of rewards, relative to control subjects (Vollm et al., 2010). In a recent multi-modal imaging study of young adults, impulsive-antisocial psychopathic traits were positively correlated with ventral striatal activation during reward anticipation on the MID task (Buckholtz et al., 2010). The same report found a positive correlation between impulsive-antisocial traits and amphetamine-induced dopamine release in the ventral striatum in the same subjects, providing evidence that ventral striatal activity as assessed via fMRI is related to dopaminergic activity. Finally, a recent study found that adolescents with conduct disorder and comorbid substance use disorder displayed less activity in the striatum and multiple frontal and temporal regions during risky decision-making (Crowley et al., 2010). They also showed less activity in anterior cingulate and temporal cortex when experiencing gains, relative to control adolescents. While the use of different experimental paradigms and distinct clinical groups may explain some of these discrepancies between studies, further work with larger, better-characterized groups of subjects is merited. It is also notable that the results of the two studies using the MID task, which provided evidence for increased responsiveness to rewards in adolescents with externalizing disorders or adults with psychopathic traits, are consistent with the behavioral and psychophysiological findings described above. An interesting question for future research is whether enhanced sensitivity to rewards can be used to predict which children or adolescents will go on to develop conduct disorder or oppositional defiant disorder.

6.3. Attention-deficit/hyperactivity disorder

As noted in the epidemiology section (Section 1), although prevalence rates of ADHD do not increase over the adolescent period (Merikangas et al., 2010), this could reflect the constraints imposed by the DSM-IV criteria (American Psychiatric Association, 1994), which may prevent studies from detecting age-related changes in its prevalence. A number of etiological theories of ADHD, such as the dynamic developmental theory (Sagvolden et al., 2005) and the delay aversion model (Sonuga-Barke, 2005), hold that motivational deficits play a central role in the development of this disorder. In brief, the dynamic developmental theory proposes that symptoms of hyperactivity or impulsivity in ADHD are underpinned by altered reinforcement of novel behaviors and impaired extinction of behaviors that have previously been acquired, but are no longer adaptive or appropriate to the context. Reduced dopaminergic function, and particularly attenuated phasic dopaminergic signalling in mesolimbic circuits, is implicated in both reinforcement learning deficits. In contrast, the delay aversion model holds that individuals with ADHD find both delays and cues signalling delay aversive. Consequently, the imposition of delay elicits a strong negative emotional response and feelings of frustration in children with ADHD. The delay aversion model is an attempt to explain the robust finding that individuals with ADHD prefer small, immediate over larger, delayed rewards in choice tasks. This is argued to be driven by deficits in the function of frontostriatal reward circuits modulated by dopamine, which mean that delayed rewards are discounted more sharply in ADHD.

In line with these theoretical formulations of ADHD, there is now considerable evidence from behavioral and psychophysiological studies for alterations in sensitivity to reinforcement in ADHD (Luman et al., 2010b). As well as showing a stronger preference for small immediate over larger delayed rewards than typically developing controls (Sonuga-Barke et al., 1992, Sonuga-Barke et al., 2008), individuals with ADHD tend to select risky options that are potentially rewarding in the short-term, even if they may be unfavorable in the longer-term (Drechsler et al., 2008). The first study to investigate Iowa Gambling Task performance in ADHD found no differences between control and ADHD adolescents when playing the task for the first time; however, the control subjects learned to avoid selecting the unfavorable decks when playing the task for a second time, whereas the ADHD subjects’ performances failed to improve over time (Ernst et al., 2003a). Partly consistent with this finding, Toplak et al. (2005) found that adolescents with ADHD displayed deficits in Iowa Gambling Task performance relative to control subjects. Lastly, children with ADHD were found to be impaired in learning to select the advantageous option over the disadvantageous one (which provided rare large penalties), when playing a variant of the Iowa Gambling Task (Luman et al., 2008). This study also found enhanced heart rate responses to rewarding outcomes in ADHD relative to typically developing children, suggesting that ADHD is associated with increased psychophysiological sensitivity to rewards.

A number of structural MRI studies have provided data supporting the hypothesis that reward-related brain circuitry is altered in ADHD. Both children and adults with ADHD have been found to show striatal volume reductions. In particular, a landmark study which collected longitudinal MRI data across childhood and adolescence in a large group of ADHD and control subjects observed lower caudate nucleus gray matter volumes in the ADHD group (Castellanos et al., 2002). These structural changes were no longer present in the oldest age group studied (19 year olds), suggesting that they may have normalized as the ADHD group reached young adulthood. In addition, a meta-analysis of voxel-based morphometry studies of ADHD reported that reduced putamen volume was the most consistent neuroanatomical abnormality in this condition (Ellison-Wright et al., 2008). Finally, Seidman et al. (2011) recently observed reduced caudate nucleus volume in adults with ADHD relative to control subjects. These changes in striatal gray matter volume in ADHD are broadly consistent with animal work showing increased impulsive behavior and hyperactivity in rats following striatal lesions (Cardinal et al., 2001), and suggest that structural abnormalities in frontostriatal pathways may contribute to alterations in reward processing in ADHD.

Functional neuroimaging studies have also probed reward-related circuitry in ADHD. In an early PET study using the Iowa Gambling Task to study the neural substrates of decision-making, Ernst et al. (2003b) found that adults with ADHD showed reduced activity in insula and hippocampus relative to control subjects, whereas the ADHD group displayed increased activity in anterior cingulate cortex. Using the MID task in an event-related fMRI study, Scheres et al. (2007) found that adolescents with ADHD showed reduced ventral striatal activity during reward anticipation compared with healthy controls, whereas no group differences were observed for neural responses to reward outcomes. A subsequent study (Strohle et al., 2008) replicated this finding for reward anticipation in adults with ADHD, although interestingly the ADHD group showed enhanced OFC responses to reward outcomes, relative to controls. Supporting dimensional approaches to understanding ADHD, subclinical levels of ADHD symptoms were also found to be negatively correlated with ventral striatal activity during reward anticipation in healthy young adults (Stark et al., 2011). Lastly, an fMRI study using a paradigm which involved selecting between immediate and delayed rewards observed reduced ventral striatal activity in adults with ADHD compared with controls (Plichta et al., 2009). In the three studies that tested clinical groups with ADHD, the ventral striatal effects were found to be correlated primarily with the hyperactive/impulsive, rather than the inattentive, symptoms of ADHD. Considered together, these findings suggest that ADHD is associated with disrupted reward processing in the ventral striatum, which may be partly underpinned by structural changes in this region. Dysfunction in reward systems appears to be most strongly related to hyperactive/impulsive forms of ADHD, consistent with the dynamic developmental theory.

To investigate the potential role of dopaminergic dysfunction in the motivational deficits observed in ADHD, Volkow et al. (2009) used positron emission tomography to investigate dopamine transporter and dopamine receptor availability in adults with ADHD and healthy controls. They found that both transporter and receptor binding were reduced in the nucleus accumbens and midbrain of ADHD subjects relative to controls (Volkow et al., 2009). In subsidiary analyses, dopamine receptor availability was found to be negatively correlated with trait motivation (as measured using the Achievement Scale of the Multidimensional Personality Questionnaire) when considering the ADHD group alone (Volkow et al., 2010). As well as providing a possible mechanistic link between neurochemistry and motivational/reward processes, these observations may partly underlie the above fMRI results showing attenuated ventral striatal responses during reward anticipation, as animal work has shown that phasic dopaminergic activity plays a critical role in reward learning (Schultz, 1998).

6.4. Substance use disorders

A large number of behavioral studies have documented changes in decision-making in adults with substance use disorders, suggesting that these disorders are associated with alterations in reward processing. In particular, adults with substance dependence consistently show impaired performance on the Iowa Gambling Task (Bechara and Damasio, 2002, Dom et al., 2005). Although there are several potential explanations of such deficits, owing to the cognitive complexity of the Iowa Gambling Task (Dunn et al., 2006), one possible account is that individuals with substance dependence are hypersensitive to the large rewards conferred by the disadvantageous decks (even though these carry the risk of very large penalties). Evidence consistent with this view was provided by Bechara et al. (2002), who found that a subgroup of individuals with substance dependence showed increased psychophysiological responses when anticipating and receiving rewards. Furthermore, male substance abusers were reported to be more influenced by potential rewards available on a simulated gambling task than control subjects (Stout et al., 2005). Extending these results to adolescents, Leland and Paulus (2005) showed that teenage stimulant users were more likely to take risks in order to gain large rewards than control subjects, suggesting that hypersensitivity to reward may be a risk factor for experimentation with stimulant drugs. Using the Cambridge Gamble Task, which can be considered a index of decision-making under risk rather than uncertainty, Rogers et al. (1999) found that amphetamine-dependent individuals showed impaired decision-making and longer deliberation times compared with healthy controls. There was also a negative correlation between the quality of decision-making and years of amphetamine abuse. Finally, substance dependent individuals have been reported to discount delayed rewards more steeply than control subjects (Petry, 2002, Petry and Casarella, 1999), a finding which is also observed in relatively well-functioning late adolescents and young adults who use substances more heavily their peers (Kollins, 2003). Considered together, these findings provide support for the view that heightened sensitivity to potential rewards contributes to the suboptimal pattern of decision-making displayed by individuals with substance use disorders. The fact that hypersensitivity to rewards is observed in younger individuals who have only experimented with substances, suggests that this may be a vulnerability factor for, rather than a consequence of, substance abuse or dependence.

Structural MRI studies have not reported changes in striatal volumes in adults or adolescents with substance use disorders, but they have revealed reductions in prefrontal and insular cortical thickness (Makris et al., 2008) and lower gray matter volumes in ventromedial prefrontal cortex and anterior insula (Franklin et al., 2002) in cocaine-dependent adults. A recent study observed a negative correlation between orbitofrontal cortical thickness and the number of different drugs tried, in a group of adolescents who had been exposed to nicotine in utero (Lotfipour et al., 2009). As noted above, the OFC is critically involved in representing the reward value of stimuli (O’Doherty et al., 2001, O’Doherty, 2004), so these results suggest that structural alterations in brain reward circuitry may contribute to risk for substance experimentation in adolescence and problematic use in adulthood.

In a recent fMRI study using a variant of the MID task, adolescent smokers were reported to show less ventral striatal activity during reward anticipation than non-smokers matched on a number of important variables such as socioeconomic status and IQ (Peters et al., 2011). Intriguingly, this effect was still observed in a subgroup of adolescents who had only experimented with cigarettes, suggesting that reduced ventral striatal activity during reward anticipation might represent a vulnerability marker for substance use disorders, rather than being a consequence of exposure to nicotine or other neurotoxic constituents of cigarettes. This echoes previous findings showing attenuated striatal responses during the anticipation of monetary rewards in substance dependent adults (Buhler et al., 2010, Goldstein et al., 2007, Wrase et al., 2007), although see (Bjork et al., 2008). In contrast, individuals with substance dependence have been found to show enhanced responses to drug-related cues in reward-related circuitry, including the ventral striatum (Buhler et al., 2010, David et al., 2005).

6.5. Consequences of early adversity or childhood maltreatment on reward processing

There is now considerable evidence from epidemiological research that exposure to early adversity or childhood maltreatment represents a major risk factor for a range of psychiatric illnesses. While some studies have reported specific associations between certain forms of maltreatment and specific classes of psychiatric illness (Shanahan et al., 2008), most recent work has suggested that childhood adversity (and particularly forms of adversity related to family dysfunction) elevates risk for all classes of psychiatric disorder in a non-specific manner (Green et al., 2010, Kessler et al., 2010). In addition, studies investigating the effect of different environmental risk factors on specific disorders largely find that the overall number of risk factors is more influential than the presence of a specific risk factor (Biederman et al., 2002). As suggested by animal research demonstrating profound effects of maternal deprivation on brain reward systems (Pryce et al., 2004), one possible route whereby those who have been subjected to maltreatment are at increased risk for psychopathology is that early adverse experiences have disrupted brain circuitry involved in reward processing.

This hypothesis has been evaluated recently in humans using variants of the MID task in combination with functional neuroimaging methods. Relative to non-maltreated controls, young adults who had experienced emotional, physical or sexual abuse showed reduced striatal activity during reward anticipation and rated reward-predicting cues less positively (Dillon et al., 2009). However, the maltreated and control groups did not differ in neural responses to outcomes, such as monetary gains or losses. In a study of adolescents with documented histories of severe institutional deprivation, Mehta et al. (2010) evaluated the effects of early deprivation on reward processing. Compared with healthy controls, adolescents who had been deprived of maternal contact and social interaction in infancy were found to display abnormally low levels of ventral striatal activity during the anticipation of monetary rewards. Although these findings should be considered preliminary due to the relatively small sample sizes in each of the studies, this research has implications for our understanding of reward-related processes in psychiatric disorders, because psychiatric patients are more likely than comparison subjects to report histories of parental abuse or maltreatment. Consequently, it may be informative to systematically measure exposure to early adversity in future studies of reward processing in psychiatric populations.

7. Conclusions and future directions

The aim of this review was to highlight recent studies documenting age-related transitions in the prevalence of common psychiatric disorders and ask why adolescents appear to be more vulnerable to psychopathology than children or adults. Evidence from developmental research in animals and normative studies of risk-taking and reward processing across human development suggests that behavioral and neural sensitivity to rewarding or appetitive stimuli is heightened during adolescence. At the same time, adolescents appear to show reduced sensitivity to punishing or aversive stimuli compared with children or adults. These changes in sensitivity to reinforcement may be adaptive for many individuals and create opportunities as well as vulnerabilities (Dahl, 2004), but may nevertheless lead to increased levels of risky and impulsive behavior at the population level. Alterations in reinforcement sensitivity coupled with less efficient prefrontal regulatory systems, as hypothesized by the triadic model (Ernst et al., 2006), may also contribute to the peak in the age-crime curve during adolescence, and the dramatic increases in the prevalence of externalizing and substance use disorders during this developmental period. Behavioral and neuroimaging research has suggested that altered reward processing may be involved in the etiology of several psychiatric disorders. Impaired striatal activation during the consummatory phases of reward processing has been reported in internalizing disorders, and may be particularly linked to anhedonia. In contrast, externalizing disorders appear to be associated with deficits in striatal activation during reward anticipation, whereas several studies have reported enhanced striatal or medial OFC responses to receipt of monetary rewards in externalizing disorders. Importantly, there is substantial agreement across child, adolescent, and adult studies of reward processing in psychopathological conditions. Furthermore, individuals who are at increased familial risk for developing affective disorders appear to show deficits in reward-related brain activity. These results imply that alterations in reward processing are not simply a consequence of having a psychiatric disorder, but may precede the illness and increase risk for psychopathology in a probabilistic fashion. Consequently, they may be used as risk markers or endophenotypes for psychiatric disorders in future studies (Hasler et al., 2004). While this area of research is still relatively new, and many of the results described herein require replication in larger samples, it is clear that characterizing alterations in reward processing may help us to understand the etiology of many common psychiatric disorders. This work may also enable us to develop more effective treatments for these disabling conditions.

An important question for future research to address, building on data showing that externalizing disorders are more common in males and internalizing disorders more common in females, is whether the triadic model of motivated behavior (Ernst et al., 2006) applies equally well to both sexes. In my view, the triadic model explains the adolescent rise in externalizing disorders and risk-taking or reward-seeking behavior far better than it explains the increase in internalizing disorders observed during puberty in females. It therefore appears to apply better to male-typical than female-typical patterns of psychopathology. As such, it could be argued that deviations in the typical trajectory of brain development for females may lead to a different pattern – a hypersensitive avoidance system, possibly coupled with excessive activity in regulatory systems. In support of the view that the triadic model needs to be modified to account for sex differences, a number of structural MRI studies have reported that males show increased gray matter volumes in basal ganglia structures relative to females (Rijpkema et al., 2011), whereas gray matter volumes of prefrontal cortical (and particular, lateral OFC) structures are larger in females (Lenroot et al., 2007, Luders et al., 2009). The former volumetric difference may have consequences for dopaminergic function in the adolescent male brain, such that males show larger striatal responses to rewards and stronger positive prediction error signals than females. In contrast, as lateral prefrontal cortex regions are implicated in the processing of threatening or punishing stimuli (O’Doherty et al., 2001), females may be more sensitive to punishment or threat than males. Furthermore, females show a peak in white matter volumes at an earlier stage in their chronological development than males (Lenroot et al., 2007, Perrin et al., 2008), who appear to undergo an extended period of white matter maturation. This may have consequences for the function of regulatory systems in males, as white matter pathways involved in cognitive and emotional control may mature later in males than in females. This again supports the view that the original triadic model may apply better to males than females. These proposals extend the triadic model of motivated behavior to account for sex differences observed in prevalence of common psychiatric disorders, and provide testable predictions for future studies.

In addition to investigating sex differences in reward processing, there are a number of other issues that could be addressed in future work. First, more research is needed to characterize the relationship between laboratory (including fMRI) measures of reward processing on one hand and real-life experiences of positive affect and reward-seeking or risk-taking tendencies on the other. There have already been some important studies showing relationships between striatal activity and subjectively-experienced positive affect (Forbes et al., 2009) or motivation to work for rewards (Buhler et al., 2010), but further work on the ecological validity of reward processing measures is needed. An implicit assumption in much of the developmental neuroscience literature is that measures of reward processing predict risk-taking in real life situations, but this needs to be tested empirically. Second, a greater appreciation of the importance of individual differences (even within groups of healthy individuals) may help us to make sense of the inconsistent findings that have been a feature of the reward processing literature. Third, it may be informative to study reward learning across adolescent development and in psychiatric disorders using associative learning tasks (e.g., classical or instrumental conditioning with appetitive and aversive stimuli), as many of the fMRI reward tasks currently in use do not involve learning. Together with these changes in the complexity of experimental paradigms employed to assess reward processing, increased use of computational modelling techniques will be needed to disaggregate neural activation patterns during different phases of learning. Fourth, to enhance our understanding of the mechanisms of action of antidepressant or stimulant drugs used to treat psychiatric disorders, it would be of interest to characterize the effects of these medications on behavioral sensitivity to rewarding stimuli or neural responses during reward processing. For example, do effective antidepressant drugs normalize reward-related brain activity in depression, and does methylphenidate enhance ventral striatal responses in those with ADHD? Finally, although this review has described preliminary findings showing that reduced striatal responses to positive or rewarding stimuli may be a state-independent marker or endophenotype for depression, empirical research is needed to investigate whether changes in reward processing are causally related to the emergence of psychopathology using longitudinal designs.

Conflicts of interest

No conflicts of interest.