Preliminary studies have shown that a new caloric testing method described by Proctor and Dix offers several advantages over conventional methods. However, it had not been established that the Proctor-Dix method would prove reliable and practical when routinely applied in a clinical setting. Therefore, we administered the test to 60 patients referred for evaluation of vestibular function and compared the test scores with results from the conventional Fitzgerald-Hallpike test applied during the same test session. The analysis was based on the average slow phase eye speed of nystagmus during the most intense 20 seconds of individual responses. A close agreement was found between "right-left difference" and "directional preponderance" scores when results of the Proctor-Dix and Fitzgerald-Hallpike tests were compared in the 54 patients who completed all steps in the experiment. Larger differences between the results of the two tests were confined to subjects who were "normal" or only slightly abnormal. Order effects and additional analysis of correlations are discussed. The present study demonstrates that reliable results, comparable to those obtained from conventional methods, can be obtained when the new Proctor-Dix test is applied in a clinical setting.