This preliminary study deals with the dynamics of the PR interval during exercise testing by using a Bruce protocol in 31 nonmedicated, normal patients with ages of 33 +/- 14 years. The behavior of the PR interval permitted its categorization into 2 groups. In Group I (27 of 31, 87.1% of patients) the PR interval showed the expected biphasic behavior with a gradual shortening with increasing exercise and a gradual lengthening during recovery. In contrast, the PR interval in Group II (4 of 31, 12.9% of patients) showed a triphasic behavior since, toward the end of recovery, there was a second decrease in duration. Because of this temporal relation to the phases of exercise, this paradoxical response, in analogy to what happened to the QT interval during exercise, could have been caused by a form of short-term memory or to varying, but normal, intergroup differences in autonomic function. However, further studies involving a greater number of patients are required to reach definite conclusions.