Previous work on this project has shown that value-based decision making requires both the OFC and the amygdala. We have used two kinds of tasks to investigate their mechanisms: (1) standard probabilistic and deterministic 2- and 3-choice visual discrimination learning tasks and (2) tasks that require subjects to make choices associated with rewards of different value (the devaluation task). We have found that both the amygdala and OFC are necessary when subjects need to make choices based on the current value of rewards, but not for choices based simply on the availability of rewards. Our earlier results indicated that when the value of one food was reduced by the subject consuming it to satiety, control subjects showed a reduction in choices of either stimuli or actions associated with the devalued food. In contrast, subjects with either bilateral damage to the amygdala or the OFC failed to show this devaluation effect. Based on the effects of permanent lesions of the OFC, it appears that the OFC is essential for either updating and registering the new value of the food outcome (value updating), or, alternatively, for linking that information to objects at the time of choice (goal selection). By examining the effects of temporary inactivations of the OFC we can disentangle value updating from goal selection. The key question for the study is when various brain areas make their contribution to the devaluation effect. To examine the contributions of OFCs components to goal selection, we reversibly inactivated either its anterior (area 11) or posterior (area 13) parts. We found that neurons in area 13 must be active during the selective satiation procedure to enable the updating of outcome valuations. After this updating has occurred, however, area 13 is not needed to select goals based on this knowledge. In contrast, neurons in area 11 do not need to be active during the value-updating process. Instead, inactivation of this area during choices causes an impairment. Taken together, the results from inactivating areas 11 and 13 demonstrate specialized functions for these two components of the macaque OFC. The posterior component, area 13, functions in conjunction with the basolateral amygdala to update the valuation of expected reward outcomes, based on an animal's current satiation state. The anterior component, area 11, plays a critical role in translating this knowledge into goals that produce an advantageous outcome. The inability to translate abstract valuation knowledge into advantageous choices resembles the goal neglect that occurs after damage to the frontal lobe in humans. An impairment in translating knowledge into behavioral goals could be fundamental to many addictive behaviors, compulsive disorders, and psychopathologies. Learning is not a unitary process and the neural circuitry that underlies learning likely depends on the type of association being learned. For example, action-based reinforcement learning may depend on different neural circuits than stimulus-based reinforcement learning. To probe ventral striatum (VS) contributions to different types of learning, we evaluated the performance of subjects with VS lesions on a probabilistic two-arm bandit reversal learning task that comprised two distinct types of learning blocks: stimulus-based (WHAT) and action-based (WHERE). Relative to controls, subjects with VS lesions had significant deficits in making stimulus-based reward associations in the WHAT blocks. By contrast, the groups did not differ in their ability to learn action-based reward associations in the WHERE blocks. We fit a reinforcement learning model to the subjects choice data to evaluate their choice consistency (inverse temperature), and positive and negative learning rates. We found that, relative to controls, subjects with VS lesions were more influenced by negative feedback (no reward) in the stimulus-based WHAT blocks. In addition, the subjects with VS lesions had an overall higher positive and negative learning rate in the action-based WHERE blocks. Overall, we find that the VS contributes to learning stimulus-based reward associations but not action-based reward associations. Thus, the VS is not relevant for motivating behavior independent of the learning modality. Rather, it is specifically important for learning the motivational value of stimuli. A considerable body of evidence has suggested a role for the amygdala in processing emotionally salient information in faces. For example, patients with amygdala damage fix their gaze on the eye region of faces to a lesser extent than controls. To evaluate amygdala contributions to social displays of emotion, we had subjects perform an attentional capture task in which they viewed pictures of a socially relevant portion of a subject face (eye, nose, or mouth) or nonsocial pictures. When the stimulus appeared at the subject's fixation point, they had to make a saccade away from it as quickly as possible. For comparison, we also permitted free viewing, during which subjects looked at faces voluntarily. In the free viewing condition, like previously published data, subjects with amygdala lesions spent less time than controls exploring faces, specifically the eye region. In the attentional capture task, we found that all subjects, lesioned and controls, made saccades more slowly when they saw social (vs. nonsocial) stimuli. Saccade latency was significantly faster in subjects with amygdala damage, compared to controls, especially on trials with a threatening expression formed by the mouth. These data indicate that subjects with amygdala damage are impaired in attending to certain social cues. Unlike in the free viewing condition, the effect of amygdala damage on the attentional capture task was driven by the mouth region, as opposed to the eyes. These data indicate that the amygdala plays a crucial role in processing and attending to social cues. Finally, selective, fiber-sparing excitotoxic lesions are a state-of-the-art tool for determining the causal contributions of different brain areas to behavior. We routinely use magnetic resonance imaging (MRI) to plan and to guide injections of excitotoxins into specific brain regions in order to produce selective cell loss (i.e., lesions) in those regions. In addition, we routinely assess the extent of excitotoxic lesions in vivo, using MRI-based methods, within a few weeks of the operation. Although the MRI-based lesion assessment is convenient and fast, we have not evaluated whether the results from this method provide an accurate picture of the extent of the lesion. To determine whether in vivo T2-weighted MRI accurately estimates damage following selective excitotoxic lesions of the amygdala, we compared lesion volume estimates for the amygdala obtained from MRI and traditional histology. Across 19 hemispheres from 13 subjects, MRI assessment consistently overestimated amygdala damage as assessed by the traditional method, namely, microscopic examination of Nissl-stained histological material. Two outliers suggested that near-complete MRI-estimated damage may predict actual damage of at least 45%, but more data from incomplete lesions are necessary to evaluate this hypothesis. Nevertheless, injection of excitotoxins routinely produces extensive damage (median = 82%) that correlates with total injection volume, validating the general success of the technique. The field will benefit from more research into in vivo assessment techniques, and additional evaluation of the accuracy of MRI assessment in different brain areas. For now, in vivo MRI assessment of excitotoxic lesions of the amygdala can be used to confirm successful injections, but MRI estimates of lesion extent should be interpreted with caution.
Showing the most recent 10 out of 20 publications
