Following their qualitative review of the findings from 10 relevant studies, Graham and Ickes (1997) speculated that reliable gender-of-perceiver differences in empathic accuracy (a) were limited to studies in which the empathic inference form made empathic accuracy salient as the dimension of interest, and (b) therefore reflected the differential motivation, rather than the differential ability, of female versus male perceivers. These speculations were tested more rigorously in the present study, which examined a larger set of 15 empathic accuracy studies and applied the techniques of quantitative meta-analysis to test Graham and Ickes’(1997) moderating variable hypothesis. The hypothesis was strongly supported, consistent with a motivational interpretation previously proposed by Berman (1980) and by Eisenberg and Lemon (1983), which argues that reliable gender differences in empathy-related measures are found only in situations in which (a) subjects are aware that they are being evaluated on an empathy-relevant dimension, and/or (b) empathy-relevant gender-role expectations or obligations are made salient.