This project has produced insights into the differential contributions of the amygdala and orbitofrontal cortex in learning stimulus-reward associations (Rudebeck and Murray, 2008). We studied, in much more detail than previously attempted, how subjects adapt when stimulus-reward contingencies change in the object-reversal task. After a change in stimulus-reward contingencies, the subject will nearly always choose the previously rewarded object, and the failure to obtain a reward causes a negative emotional reaction. Eventually, when the subject stops choosing that object and explores the alternative choice, it produces a reward that is not strongly expected, and this event triggers a positive emotional reaction. Subjects with orbitofrontal cortex lesions are slow to adapt their choices to these changing stimulus-reward contingencies. Based on trial-by-trial analyses of rewarded and unrewarded choices, we have found that subjects with orbitofrontal cortex lesions did not have a problem with using the negative feedback that resulted when the choice of a previously rewarded object led to no reward. Instead, these subjects benefited less than intact subjects from the positive feedback that followed each correctly performed trial after an error. Accordingly, their impairment resulted from an inefficiency in learning that a given choice would yield a positive outcome. Subjects with amygdala lesions showed the opposite pattern of results: they benefited more than intact subjects from each correctly performed trial that followed an error. These findings demonstrate that the orbitofrontal cortex and amygdala make different contributions to object-reversal learning and therefore to learning stimulus-reward associations. Subjects with orbitofrontal cortex lesions have a deficit in representing the updated value of a stimulus that has previously been of low value. They do not, as the traditional account has it, perform poorly on the object-reversal task because of perseveration, which is mild in these subjects. This realization is important because perseveration is of little relevance to diseases like major depressive disorder. An understanding of orbitofrontal cortex function in terms of an ability to use positive feedback to upgrade low valuations (of objects or self), on the other hand, has an obvious relevance to this disease. These patients have a low valuation of themselves and believe that only low-value events occur in their lives (or are likely to occur in the future). Our previous work has also shown that amygdala lesions, like lesions of the orbitofrontal cortex, cause a disruption of satiety-specific devaluation effects. But the previous experimental design left open whether the instrumental (response-reward) associations or Pavlovian (stimulus-reward) associations were guiding performance. To investigate the specificity of amygdala function along these lines, we designed a task in which responding relied on instrumental control of behavior and could not be explained via Pavlovian mechanisms. Subjects were trained to perform two different instrumental responses (tap and hold) on a touch-sensitive screen for two different food rewards. One of the foods was then devalued by selective satiation. As in our previous studies, control subjects showed a reduction in responses associated with a devalued outcome, but subjects with amygdala lesions failed to show this effect. This finding supports the idea that the amygdala is required for updating reward value and therefore plays a crucial role in guiding goal-directed behavior based on response-reward associations. This project has also made progress in understanding the role of distinct parts of orbitofrontal cortex in reward-guided behavior and emotion. Marked changes in reward-guided behavior and emotion are hallmarks of damage or dysfunction within the orbitofrontal cortex, but it is not a homogeneous structure. A lateral region broadly encompasses Walkers areas 11 and 13 and is heavily interconnected with sensory areas of cortex. A medial region, which includes Walkers area 14, is more densely interconnected with medial frontal cortex and autonomic structures. There has been speculation that these two regions play distinct roles in regulating emotion and reward-guided behavior, but direct evidence to this effect has been elusive. The current project used the selective satiation procedure to examine the ability of subjects to update their valuations of particular objects based on reward expectation. We found that lateral but not medial orbitofrontal cortex is essential for choosing objects based on their updated value. By contrast, medial orbitofrontal cortex but not lateral orbitofrontal cortex is important for the ability to stop responding to an object when it is no longer rewarded (i.e., in extinction). These results provide clear evidence for functional dissociations within the orbitofrontal cortex, and we are currently testing hypotheses about the contributions of medial vs. lateral orbitofrontal cortex in signaling expected outcomes. These results agree well with those described above for the object-reversal task: in both cases they point to a role for the lateral orbitofrontal cortex in updating reward expectations based on object-reward associations. In a further examination of extinction, which entails the abolition of a previously established object-reward association, we examined frontal areas medial to the orbitofrontal cortex. Evidence from one animal model had suggested that a part of the medial frontal cortex, the infralimbic cortex, plays an important role in mediating extinction. Lesions of this region in rats results in an increased expression of the older, original learning through the spontaneous recovery of responding. In the present project, however, we showed that this finding does not generalize to all animal models. In our animal model, lesions of infralimbic cortex performed no differently than controls;there was no spontaneous recovery of the previously extinguished behavior. (Lesions of another medial frontal area, the prelimbic cortex, also had no effect.) This finding shows that infralimbic cortex plays different roles in the two animal models studied to date.
Showing the most recent 10 out of 20 publications
