A modification of the visual evoked potential (VEP) technique, first employed by Braddick, Wattam-Bell and Atkinson [(1986) Nature, London, 320, 617-619] was used to estimate the orientation selectivity of 3-month-old infants. The orientation-selective VEP was recorded in response to various changes in the orientation of a square-wave grating. The magnitude (the square root of the power of the FFT) at the frequency of orientation change was assumed to represent the response to the change in grating orientation. Orientation sensitivity was then estimated by plotting the magnitude of the FFT at the frequency of orientation change as a function of the log of the orientation change in degrees. For each individual, the data were fit by nonlinear regression and threshold was defined as the largest orientation angle for which the magnitude of the FFT was zero. The results suggest that the orientation selectivity of 3-month-old infants (1.33 deg) is similar to that of adults (1.13 deg) tested with the same stimulus parameters (1 c/deg, 9 Hz). However, when adults are tested with stimulus parameters selected to optimize their VEP response (4 c/deg, 18 Hz) instead of those which optimize the infant's response, the orientation discrimination of adults improves by a factor of 2 (0.53 deg). The results obtained from adults under optimum stimulus conditions (4 c/deg, 18 Hz) approach the estimates reported in the literature for static stimuli.