How does the brain evaluate whether the benefits of a decision outweigh the costs? A study now reveals that neurons in the anterior cingulate cortex encode costs and benefits, and altering brain activity here biases choices away from negative outcomes. These results link anterior cingulate cortex with the regulation of emotional states.
Humans and animals face a variety of decisions every day, many of which require experiencing some cost to obtain some benefit. Although you might fancy some fresh sushi for lunch, the downpour indicates your journey outside will be a wet one; perhaps grabbing a sandwich at the indoor canteen would be preferable. How does the brain evaluate whether the sushi will be rewarding enough to tolerate the cost of getting wet? Although we often make such choices with relative ease, decisions that have potentially negative consequences can evoke anxiety, and individuals who suffer from anxiety disorders and depression find cost-benefit decision-making particularly difficult1. In this issue of Nature Neuroscience, Amemori and Graybiel2 demonstrate not only that neurons in the primate pregenual anterior cingulate cortex (pACC) encode the costs and benefits of a decision, but that pACC controls whether to choose alternatives associated with potentially negative outcomes, implying a link between pACC's valuation functions and the regulation of emotional states such as anxiety.
The anterior cingulate cortex has long been considered to be a critical structure in emotion. Papez (1937) argued that the cingulate gyrus may be the “receptive region for the experience of emotion”3. Indeed, the anterior cingulate cortex has widespread interconnections with limbic, cognitive and motor areas, making it a potential interface at which cognitive, emotional and motivational states can influence behavior4. As investigators began to examine pACC function in decision-making tasks, it became clear that pACC neurons encode many of the variables that influence a decision and that damage to pACC impairs cost-benefit decision-making5,6,7. However, a critical issue is how to link neural correlates of decision-making in the healthy brain to clinical disorders associated with impaired decision-making in the unhealthy brain, particularly in pACC. Abnormal pACC activity is associated with anxiety-related psychiatric disorders, including depression, obsessive-compulsive disorder, schizophrenia and addiction, and disrupting the region in and around pACC is a common treatment for these disorders8,9. On the basis of this evidence, Amemori and Graybiel2 reasoned there might be a link between pACC's role in cost-benefit decision-making and the regulation of some forms of anxiety.
To explore this issue, the authors developed a cost-benefit decision task for macaque monkeys modeled after approach-avoidance (Ap-Av) protocols used in humans to quantify decision-related anxiety. On each trial, monkeys sat in front of a computer screen and made Ap-Av decisions between two options (Fig. 1a). The size of two colored stimuli indicated the negative (airpuff to the face) and positive (liquid reward) outcomes that would be experienced if the monkeys chose the approach option. Alternatively, the monkey could always choose the avoid option, which would provide a very small reward and had no negative outcome (airpuff). As a control task, monkeys were tested on an approach-approach (Ap-Ap) task that was identical to the Ap-Av task, except that both colored bars signaled potential reward and the monkey would not experience an airpuff. The authors reasoned that the anticipation of negative outcomes might evoke anxiety and that this should be specific to the Ap-Av task.
To test whether the Ap-Av task might evoke changes in anxiety levels, the authors administered the anti-anxiety drug diazepam to the monkeys and then tested their performance on the two different tasks. The results were clear: diazepam treatment significantly increased the number of approach choices in the Ap-Av task (Fig. 1b) in a dose-dependent manner, but it had no effect on the Ap-Ap task. In other words, anxiolytic drug treatment altered the decision boundary between the valuation of costs and benefits, causing subjects to be more willing to endure a negative outcome to obtain reward. This result indicates that the Ap-Av task provides an experimental context for exploring the neuronal representation of positive and negative emotional states.
To investigate the role of pACC in cost-benefit decision-making, Amemori and Graybiel2 recorded the electrical activity of over 1,000 neurons in a large region of the pACC, including both banks of the anterior cingulate sulcus and the underlying cingulate gyrus. This is particularly noteworthy because most single-neuron studies of pACC have primarily focused on the dorsal bank of the cingulate sulcus5. Consequently, we know relatively little about the functional properties of neurons in these more ventral regions of pACC (but see ref. 10), despite pACC's aforementioned association with psychiatric illness. By varying the amount of experienced reward and airpuff across a wide range of values and observing choice behavior, the authors were able to determine which task variables modulated pACC neuronal activity during cost-benefit decision-making, with a particular focus on whether neurons were sensitive to decision costs, benefits or the overall subjective value, or utility, of the choice.
Several noteworthy features emerged in the neuronal data. Most of the neurons fell into one of two clusters (Fig. 1c). P-type neurons were activated by task variables related to the expectation of large reward or high utility when the monkeys chose the approach option. In other words, these neurons appeared to be sensitive to positive motivational states. Conversely, N-type neurons were activated by task variables related to the expectation of small reward or low utility when the monkeys chose the avoidance option. Thus, N-type neurons seemed to be sensitive to negative motivational states. Furthermore, when the authors examined the spatial distribution of the P- and N-type neurons, they found that both types of neuron were present and intermixed across the region of pACC that was sampled, but, in a specific region of the ventral bank of the cingulate sulcus, the N-type neurons were significantly more prevalent than the P-type neurons.
Although previous studies have shown that neurons in frontal cortex, including pACC, often exhibit positive and negative coding relationships in decision-making tasks11,12, the current results imply that P-type and N-type neurons might make unique functional contributions to cost-benefit decision-making. To test whether there is a causal link between N-type neurons and negative motivational states that might promote avoidance decisions, the authors applied electrical microstimulation to different pACC sites at the time when monkeys evaluated the cost-benefit alternatives. If pACC has a critical role in decision-making, disrupting normal pACC activity with microstimulation should induce a change in choice behavior. Although stimulation had little effect in the Ap-Ap task when neither decision had an associated cost, microstimulation of pACC in the Ap-Av task produced a significant increase in avoidance decisions (Fig. 1b), but only when applied in the ventral bank of the cingulate sulcus, precisely where N-type neurons predominate. Subsequent administration of the anti-anxiety drug diazepam fully reversed the microstimulation effect in this subregion of pACC. Taken together, these findings provide convergent evidence from several techniques that suggests that pACC not only is critical for cost-benefit decision-making but also may be critical for regulating negative emotional states such as anxiety.
As impressive as these results are, they also raise several important questions for future research. First, other brain areas in the frontal cortex, including the orbitofrontal cortex13 and lateral prefrontal cortex14, contain neurons that are sensitive to positive and negative outcomes; thus, it remains an open question whether the P- and N- type neurons in pACC are performing functionally unique computations in comparison with positive and negative coding neurons in these other frontal areas. Second, although N-type neurons were preferentially located in the ventral bank of the cingulate sulcus, where microstimulation induced avoidance decisions, many of these N-type neurons encoded motivationally positive variables related to the reward, rather than motivationally negative variables related to the cost of the airpuff. This raises an important question about the mechanism of the microstimulation effect: does pACC microstimulation disrupt and decrease the influence of reward-related representations, or does pACC microstimulation activate and increase the influence of cost-related representations? Both of these mechanisms might produce an increase in avoidance decisions.
Finally, it would interesting to know whether pACC influences cost-benefit decision-making for other types of decision cost, such as physical effort. Several studies have shown that damage to anterior cingulate cortex biases choice toward actions that are associated with less effort even when a more rewarding option is available6,15. If pACC neurons also regulate whether particular rewarding outcomes are worth investing physical effort, this might provide new insight into why patients with clinical depression often exhibit a lack of physical activity; indeed, altered pACC activity is associated with clinical depression8.
The results of Amemori and Graybiel2 add to a growing literature linking pACC with cost-benefit decision-making5,6,7, but also suggest that these functions may be linked to the regulation of emotional states. A better understanding of the neuronal mechanisms that support cost-benefit decision-making may provide new insight into the pathophysiology of psychiatric disorders, including anxiety-related disorders, in addition to informing the potential efficacy and side effects of disrupting pACC function as a therapy for psychiatric disorders9.