It has been proposed that there is a fundamental dichotomy between the functions of PFo and PFm. PFo is thought to be essential for value-guided object choices whereas PFm is thought to be essential for value-guided action choices. To address this question, we have investigated the selective contributions of PFm to flexible object and action choices in two ways. One approach tests animals abilities to make flexible choices following changes in reward contingencies, with reward value held constant. A second approach tests animals abilities to adjust their choices based on the biological value of the outcome (i.e., food reward) revealed by that choice, with reward contingencies held constant. To evaluate flexible responses after changes in reward contingencies, we used both object-based and action-based reversal learning. The object-based task was carried out using two different objects;selection of one object was rewarded and selection of the other was not. The action-based task was carried out using two different actions performed on a joystick;performance of one action was rewarded and the other was not. After the subjects had learned the correct object or action, the reward contingencies were reversed and we measured the subjects ability to respond to this change by altering their choice of object or action. Animals with PFm lesions were impaired in using positive reinforcement to sustain selection of the correct object during the object-based reversal learning task. On the action-based reversal task, by contrast, animals with PFm lesions were impaired in using error feedback (i.e., nonreward) to guide their choice of action. The latter result differs from that of other investigators, who found that PFm damage led to an impairment in using reward feedback, as opposed to error feedback, in learning. In that study, however, the effects of PFm lesions were assessed only after learning had been well established. One possible explanation for the entire pattern of results is that PFm is sensitive to changes in reward information following performance of well learned reversals (whether object-based or action-based), whereas during initial learning of reversals, PFm is important for using error feedback to guide the choice of response. To evaluate animals abilities to adjust choices based on the value of a reward outcome, we used the reinforcer devaluation task. In this task, the value of rewards is varied by prior consumption of the food item that serves as a reward. Eating one kind of food leads to a selective satiation that decreases the value of that food. We found no effect of PFm lesions when animals were asked to link objects or actions with the current value of an outcome using a reinforcer devaluation paradigm. In summary, we found evidence implicating the PFm in linking both objects and actions with reward contingency but not in linking them with current value. Importantly, our data show that the role of PFm is not restricted to linking actions with reward outcomes, as previously suggested. Instead, the data indicate a more general role for PFm in using reward information to sustain effective choice behavior. Over the past few years we have developed a new method to investigate associations of actions with reward value. As before, we compared the effects on behavior of lesions of PFm and PFo. We also studied the role of the amygdala in this new task. Our new task was intended to improve on the shortcoming of our earlier study (described above) by using a test design in which we could evaluate choices between actions. Subjects were trained to perform two different actions (tap and hold) on a touch-sensitive screen. Repetitive tapping of the screen produced one food;constant holding of the screen produced a different food. Subjects were fed one of the foods to satiety, which lowered its value. This is the same procedure we have used in studying associations between objects and reward value. As expected, controls showed a reduction in the action associated with a devalued food, called a devaluation effect. In contrast, subjects with amygdala lesions failed to show this effect. This finding provides strong support for the idea that the amygdala is as important for the valuation of actions as it is for the valuation of stimuli (and objects). We also found that PFo lesions, like amygdala lesions, disrupted devaluation effects. PFm lesions produced equivocal results. Taken together, the foregoing results provide two pieces of evidence at odds with the idea that PFo and PFm are dedicated to object- and action-based processes, respectively. First, we found that PFm lesions disrupt object reversal learning as well as action reversal learning. Second, we found that PFo is essential for linking actions with the current value of foods. For the tap/hold task, PFo plays a necessary role in making choices between actions based on the current value of an outcome, much like it does for choices between objects. Thus, contrary to a current widely-held view, PFm contributes to object-value processing and PFo contributes to action-value processing. In keeping with the notion that learning about reward value is critical to understanding affect, we have also continued to study the role of the amygdala in stimulus-reward learning. We trained animals to perform a stimulus-choice task for fluid involving three levels of reward. Three stimuli, novel at the beginning of each session, were associated with different amounts of water: 0 ml, 0.2 ml or 0.4 ml. On each trial, two of the stimuli were presented and the subject was free to choose between them. All possible pairwise presentations were used in equal proportions. Through trial and error, subjects learned to choose the stimulus associated with the greater amount of water. Choice behavior was assessed both before and after bilateral excitotoxic lesions of the amygdala. On average, subjects learned the associations between stimuli and size of reward at a slower rate following amygdala lesions. Alterations in learning rate were mainly due to an inability to use nonreceipt of reward to guide subsequent choices. These results suggest that one role of the amygdala is to learn from nonreceipt of reward. Finally, we have continued to investigate the role of PFo in value-based decision making. An inability to reverse object-reward associations is a hallmark of injury or dysfunction within PFo in humans and animals. Several classic studies have reported severe impairments on object reversal learning after PFo damage. None of these studies used the modern lesion methods that spare axons passing nearby or through the site of the lesion. It is therefore possible that the impairments are due to inadvertent disruption of white matter caused by aspiration lesions of the PFo, which was the method used in the classic studies. If so, the deficit would reflect the behavioral effects of damage to fibers of passage either alone, or in addition to, PFo damage. To address this issue, animals with excitotoxic lesions of the PFo and controls were tested on an object reversal learning task, as well as on a reinforcer devaluation task. Both tasks are sensitive to aspiration lesions of the PFo. Excitotoxic lesions of the PFo were without effect on object reversal learning, but impaired monkeys abilities to choose adaptively in the reinforcer devaluation task. The magnitude of the impairment on the reinforcer devaluation task was comparable to that observed following aspiration lesions of the PFo. These findings suggest that the classic impairments in object reversal learning reported after aspiration lesions of the PFo are caused by damage to fibers passing near to the PFo, either alone, or in combination with damage to the PFo.
Showing the most recent 10 out of 20 publications
