Cynomolgus monkeys (Macaca fascicularis) were trained in a task that assessed their ability to associate visual stimuli with food reward. Acquisition of stimulus-reward associations was measured under 2 conditions, a 2-stimuli acquisition condition and a 1-stimulus acquisition condition. On each trial in the 2-stimuli condition, the positive (correct) and negative (incorrect) stimuli were presented side by side and the animal chose one by touching it; if the choice was correct, a food reward was dispensed. On each trial in the 1-stimulus condition, either the positive or the negative stimulus was presented alone; if the stimulus was the positive, it was followed by reward delivery, regardless of the animal's response to it, and if it was the negative, it was not followed by reward delivery. Thus, reward delivery was contingent upon the animal's response to the stimuli in the 2-stimuli condition but not in the 1-stimulus condition. The effect of acquisition trials under these 2 conditions was measured, in both conditions, by the animal's subsequent choice when presented with the 2 stimuli side by side. Following preoperative training in this task, the animals were first subjected to unilateral ablation of the inferotemporal cortex. This operation had little effect on the animals' learning ability. Then, the amygdala was ablated in the hemisphere contralateral to that in which the unilateral inferotemporal ablation had been carried out. This combination of crossed unilateral lesions of the amygdala and of the inferotemporal cortex, which disconnects the amygdala from the output of visual association cortex, produced a profound impairment in stimulus-reward-associative learning.(ABSTRACT TRUNCATED AT 250 WORDS)