The psychophysics of numerical comparison: a reexamination of apparently incompatible data

Reaction-time studies of numerical comparison have used essentially two paradigms: classification, in which a target number must be labelled "larger" or "smaller" in comparison to a fixed standard, and selection, in which the larger (or smaller) number of a pair must be picked out. In previous studies, classification has yielded only a distance effect in RTs, whereas selection has also revealed magnitude (or minimum) and congruity effects. We used two experiments with two-digit number comparisons to find the reason for this discrepancy. In Experiment 1, we used a variant of the classification task with the standard changing on each trial. RTs increased along with the standard for "smaller" responses and decreased along with the standard for "larger" responses, in a manner reminiscent of magnitude and congruity effects. In Experiment 2, we again used classification, but the fixed standard 75 was not at the center of the range of target numbers (20, 21, ... 99). Close to the standard, RTs were faster for "larger" than for "smaller" responses, again a congruity effect. Our data show that magnitude and congruity effects can be obtained with two-digit numbers in classification as well as in selection tasks. A single equation, which implies that numbers are compared with respect to reference points at both ends of the continuum, describes the results from both tasks.