This article examines the effects of irrelevant information on the multidimensional "same"--"different" task. Subjects were instructed to compare two geometric figures with respect to certain attributes but to ignore other attributes in making the "same"--"different" decision. The irrelevant attributes were chosen in such a way that they could not easily be ignored to see how the existence of irrelevant differences would affect the comparison process. As expected, the overall latencies were longer than is usually found in tasks with no irrelevant differences. However, the nature of the comparison process appeared unchanged. The usual finding of a "fast-same" phenomenon persisted even when irrelevant information was present. The similarity of the results in this task to results in the "same"--"different" task with no irrelevant features supports the idea that the same comparison mechanism is used whether or not irrelevant differences are present in the stimulus pairs. The results suggest a more general-purpose comparison mechanism than is usually included in models of this task. Two-process models of visual comparisons are thus ruled out entirely. A modified version of Krueger's noisy-operator theory does appear consistent with the results.