It has been shown that damage to the human lateral frontal cortex results in a severe impairment on conditional associative tasks requiring learning of arbitrary associations between a set of stimuli and a set of responses (Petrides, M., Neuropsychologia, 1985, 23, 601-614; 1990, 28, 137-149). In these studies, which first demonstrated the impairment after frontal lesions, training was by a trial-and-error procedure, during which the subject performed the various responses when a given stimulus was presented and the experimenter provided feedback until the correct response was performed. In the present experiment, patients with unilateral frontal- or temporal-lobe excisions were tested on a visuo-motor conditional associative task with a modified procedure. The subjects had to learn arbitrary associations between a set of coloured stimuli and a set of hand postures. Training in the present experiment consisted of a series of demonstration trials followed by test trials. In the demonstration trials, the experimenter showed the subject the associations between the stimuli and the responses and, in the test trials that followed, the subject was tested on these associations. If an error was made on the test trials, the correct response was demonstrated by the experimenter. Despite these changes in the training procedure, namely the demonstration of the stimulus-response associations and the provision of the correct response immediately following an error, patients with left or right frontal-lobe excisions were severely impaired in learning this task. These findings, together with those of the earlier studies (Petrides, M., Neuropsychologia, 1985, 23, 601-614; 1990, 28, 137-149), demonstrate that the impairment in conditional learning after frontal lesions is not dependent on the type of the training procedure and therefore that it reflects a specific impairment in learning.