Reward Neurocircuitry in Adolescent Development and Decision Making
Meeting Summary: January 20, 2006
National Institute on Drug Abuse (NIDA)
National Institute of Mental Health (NIMH)
National Institute of Child Health and Human Development (NICHD)
National Institute of Neurological Disorders and Stroke (NINDS)
Adolescence is a time of increased risk taking, as well as increased decisional capacity. Importantly, it is a time of increased vulnerability to social influences and to the onset of psychopathology, such as mood and substance use disorders. An improved understanding of reward neurocircuitry in adolescent development and decision making may offer opportunities for prevention of illness and early intervention. To assess the state of our knowledge and research needs, an interdisciplinary workshop was held on January 20, 2006. Attendees included: preclinical and clinical researchers; neurobiological and behavioral scientists; pediatricians; and, developmental psychologists. Questions of interest included: What is known about the functional neuroanatomy and neurochemistry of reward neurocircuitry in humans? What is known about changes in reward/decision-making circuitry over the course of adolescence? How are developmental changes in reward circuitry impacted by puberty? What are the major methodological considerations in designing research to address these questions? The following is a summary of the major themes discussed at the meeting.
Although some work has been done in adult humans, most of what we know concerning reward neurocircuitry is based on the animal literature. Key brain structures include the nucleus accumbens, amygdala, and prefrontal cortex (PFC), though the exact function of PFC varies somewhat across species. Dopamine neurotransmission through this circuit has been thought to mediate hedonic pleasure or reward, particularly since dopamine antagonists block the rewarding effect of drugs.
Behavioral conditioning studies indicate a decrease in dopamine release over repeated reward trials. Dopamine release instead becomes a repsonse to the cue that signals the reward. The findings have led to the theory that dopamine signals a willingness to work for the reward or the salience ("wanting") of the reward. However, dopamine signaling is also activated by stress and various aversive stimuli. In addition, there are both tonic and phasic forms of dopamine release that seem to occur under different circumstances. Finally, the circuit itself has multiple interconnections between its various components, in addition to numerous connections with other brain structures (e.g., hypothalamus) that may modulate the ultimate output. It was proposed that multiple reward circuits may exist, each stimulated by different types of reward.
Definitions of reward vary across theories and studies, from those based on subjectively reported "hedonic" qualities to those based on behavioral observation. Early psychoanalytic theory posited multiple sources of "pleasure" that were weighed in making decisions on how to act. Subsequently, behavioral theory proposed a more quantifiable system where reward was equated with reinforcement, measured by an increase in an antecedent behavior.
More recently, neuroeconomics has proposed new ways to think about reward. For instance, computational theory hypothesizes that an organism keeps a running total of all future rewards expected and that the reward expected at any given time equals the difference between what it expects next and what it just expected. This "temporal difference learning" is analogous to the way the dopamine system actually responds (that is, phasically). Neurons respond to expectation of reward, and this response is greater when the animal receives a favored reward as opposed to a less favored one. Gradually, neuronal activity occurs in response to the reward cue, rather than the reward itself, until finally neuronal activity decreases altogether as reward becomes predictable.
Economic theory equates pleasure with utility. Decisions are made based on an expectation of reward (utility) in the future, and decisions are considered rational if they are consistent with expectations. There are multiple kinds of utility, such as experiential utility which is subjective and difficult to measure, and decision utility which is based on what an organism chooses. A number of corollaries may be relevant to understanding adolescent behavior. For example, prospect theory states that losses are more significant than gains; thus, subjects prefer smaller amounts of money that are guaranteed over larger amounts that can only be obtained by taking a risk. Adolescents may behave differently, however, since they seem to perceive risk differently than do adults. Temporal discounting refers to the idea that when a reward is delayed, its value to the subject is lessened or discounted: the greater the delay, the bigger the discount. Studies indicate that immediate rewards are always preferred over delayed rewards by younger subjects in both rats and humans. However, once a delay in receiving a reward is unavoidable, the relationship between size of the delay and size of the reward (or discount) may be consistent regardless of age.
In general, decision-making is based on drives, memories, and perceptual evaluations of the properties, valence, and contexts associated with a given stimulus. Ultimately, decision-making cannot be separated from reward, and when making choices, adolescents may assign different weights to rewards than adults do. State factors, such as arousal ("hot" vs. "cold" cognition) may also be particularly relevant to adolescent decision making. Most experiments occur under conditions of low arousal, but adolescence is not a low arousal developmental period. At this point, given the possibility of multiple reward circuits and the multiplicity of reward theories or frameworks from which to build experiments, it seems important to measure reward in as many ways as possible when addressing these issues in adults and especially in adolescents.
Changes during adolescence:
Although reward circuitry is largely conserved throughout development, increased risk-taking characterizes adolescence across species. A manifestation of increased risk-taking is an increased death rate during adolescence. Adolescent male rats show much stronger conditioning to novelty and social rewards than do adult rats, and alcohol seems to enhance social behavior in adolescent rats while decreasing it in adult rats. Although little data is available in humans, adolescents consistently report in surveys that they drink mostly to facilitate social interactions.
Recent neuroimaging studies have begun to compare adolescent vs. adult neural responses to reward. An important theme has been the balance between subcortical and cortical, especially prefrontal, systems that may modulate limbic structures. Both structural and functional neuroimaging studies have reported protracted developmental changes in prefrontal regions thought to subserve cognitive control during adolescence. For example, response inhibition tasks activate the anterior cingulate, ventral PFC (where activations correlate with performance) and dorsolateral prefrontal regions, where decreases in the focal extent of activations are sometimes seen with increasing age.
Few longitudinal studies are available, but one using a go/no-go task has shown enhanced activity in the inferior frontal gyrus, correlated with performance, across the age range of 7-14 years. These observations suggest a parallel between the activation of frontal control systems and the ability to maintain goal-oriented behaviors. Preliminary findings using diffusion tensor imaging suggest that myelination as a function of age (ages 7-31 years) parallels increasing efficiency on a go/no-go task. Frontostriatal, but not corticospinal, diffusivities predicted faster reaction times, independent of age and accuracy, particularly for trials expected to require greater cognitive control.
Imaging studies have also shown developmental changes in subcortical systems activated by reward manipulations. A study using a task modeled after one shown to stimulate dopamine firing in non-human primates found the activation of the nucleus accumbens reached adult levels by adolescence. In contrast, a decrease in the extent of activation seen in orbitofrontal cortex extended even beyond adolescence.
The neural response to reward magnitude appears to vary with age. Children show no significant differences in the activation of the nucleus accumbens in response to the relative magnitude of the reward, in contrast to adolescents and adults who show greater activation with larger rewards. Adolescents may be more sensitive to reward magnitude, as they have shown decreased nucleus accumbens activation to small (vs. medium or large) rewards, suggesting they experience this diminution as a negative, with more exaggerated increases in activation with large rewards. One study reported that accumbens activation correlated positively with subject ratings of how likely they were to engage in risky behavior within the next six months, but negatively with the perceived likelihood of negative consequences.
Some researchers have conceptualized the cognitive impulsivity, risk-taking and affective intensity seen in adolescent decision-making in terms of a functional equilibrium among reward/approach, avoidance and supervisory systems, and a set point within the system. Both a hyperactive and a hypoactive reward system have been hypothesized to characterize the adolescent period. Risk-taking might be viewed either as a result of a hyperactive reward system or, alternatively, as an attempt to tune up a hypoactive reward system.
Several neuroimaging paradigms have been used to interrogate these reward systems. A risk taking study employing a 'Wheel of Fortune' task separated out response selection, anticipation of reward, and feedback stages. Adolescents took significantly more risks than did adults and were happier when they won money, but were less upset than adults when they lost. In response to feedback, adolescents activated the nucleus accumbens more than did adults, while adults engaged the amygdala and PFC more than did adolescents. In contrast, another study reported significantly greater activation of the striatum in adults than adolescents, consistent with the notion of a hypoactive adolescent reward system. Many questions remain, including, importantly, the role of social influences on the reward and decision-making systems and hormonal influences on these systems.
Finally, other work has used food stimuli with fMRI to compare adult vs. adolescent responses to high vs. low fat foods. In adults, the high and low calorie foods showed activation in the amygdalar-hippocampal region and medial PFC. Comparing the two food conditions, the high calorie condition showed a large frontal activation (possibly reflecting an evaluative response) not seen in the low calorie condition, as well as activations in thalamus and hypothalamus. In adolescents, both high and low calorie foods activated the hippocampal and perihippocampal regions, as well as the fusiform gyrus. Adolescents failed to show the large frontal activation to high calorie foods seen in the adults, but did show some cingulate activation. Orbitofrontal activation during this task increased with age, while anterior cingulate activation decreased with age.
Overall, work in this area is sparse, with few studies focusing selectively on the adolescent period and no studies following subjects through different stages of adolescence. While the few studies available have focused on key structures of interest, the interactions within these systems, e.g., prefrontal modulation of limbic activity, have yet to be studied. Finally, it is important to recognize potential limitations in terms of ecological validity of the sorts of tasks used in this research.
Although little is known about which aspects of neural development are impacted by puberty, some have hypothesized that the development of prefrontal cortical systems is more closely linked to age, while changes in subcortical limbic systems may be influenced more by puberty. Themes of the discussion focused on the complex nature of puberty, measurement issues, and interactions with the social environment.
At the physiological and neuroendocrine levels, adolescence involves three different hormonal axes that tend to cascade together but whose onset and duration can vary independently. Gonadarchy, the cascade of hormones that activate the gonads, begins with an increase in the activity of neurons that cause cascades of change through the hypothalamus and the pituitary to release follicle stimulating hormone (FSH) and luteinizing hormone (LH), which activates the gonad. Activation of the gonad leads to releases of either estrogens or androgens, depending on the sex of the individual, which then feeds back and affects other systems. A second axis is growth, involving intense metabolic demands, and the growth hormone system. A third axis is adrenarchy, involving hormones such as dehydroepriandrosterone (DHEA). These hormones influence changes in skin, and axillary and pubic hair changes.
The cascading and interacting effects of the hormonal axes are complex, and there is much individual variation in the relationships among these different axes. The measurement of a peripheral level of a single hormone, either through blood or saliva, particularly if measured only one time, is unlikely to reveal significant effects on behavior. It is more likely the pattern of secretion or pulsing of the hormone and its interaction with other pubertal and neuronal processes that impacts behavior. Therefore, a single measurement of testosterone not correlating with aggressive behavior does not mean that testosterone is not a factor.
Other components of adolescence that interact with pubertal processes are also important to consider. For example, the rapid physical growth and onset of sexual maturation changes how others treat an adolescent and his or her social experience. Adolescence should be considered the result of a suite of changes with a relative synchrony involving neurobehavioral changes linked to pubertal changes, social interaction and social status changes, emotion regulation changes and motivational changes.
Puberty appears to be a period of heightened sensitivity to stress. Non-human primate research has shown that lower reproductive hormone levels are related to increased stress sensitivity and lower levels of serotonin gene expression. If differences in reproductive hormones vary with differences in stress sensitivity, then these hormones may play a role in guiding the behaviors seen during adolescence. Exposure of the brain to reproductive hormones in adolescence will vary considerably with these different patterns of pubertal development.
If researchers are rigorous in their conceptualization of puberty and in their measurements, hormones in urine and saliva can be used to investigate links to behavior. Questions should be focused and puberty should not be viewed as categorical or unidimensional.
Research needs, opportunities, strategies and approaches:
An increased understanding of the typical developmental trajectory of reward processing and decision-making will improve the identification of delays or deficits in individuals with disorders and promote a better understanding of how insults at different times in development are manifested. In addition, an increased understanding of individual differences apparent in behavioral measures that assess reward processing behaviors and decision-making, as well as individual variation in pubertal development is needed. Moreover, a rigorous conceptualization of puberty is needed to reflect evidence that it entails a suite of changes and not a categorical state.
Ecologically valid tasks, capable of identifying high-risk individuals and predicting development course, are needed to probe reward processing and decision-making during adolescence and may prove useful for developing effective interventions. Although using complex tasks to measure these behavioral domains may be useful, modeling to decompose such tasks may be challenging.
Improved understanding of the social neuroscience of adolescence is of critical importance as well. A systematic study of different rewards and better understanding of how to calibrate rewards so that they hold the same meaning for different age groups is needed. Designs that utilize the same task but different rewards would provide information as to whether the same circuitry is being activated.
Good behavioral assessments in animal models can provide valuable mechanistic information. In particular, studies of gene-environment interactions in risk-taking in animals and in humans are needed. Combining imaging with genetics may prove to be a fruitful endeavor.
Finally, an interdisciplinary approach and use of converging methodologies are likely to hold advantages for defining the neurobiological substrates that underlie reward processing and decision-making in adolescence. The techniques of behavioral therapy and applied analysis may inform the study of reward processing and decision-making from a quantitative perspective.