Microiontophoresis of the inhibitory transmitter GABA was used to reversibly inactivate small sites of defined orientation specificity at a horizontal distance of some 600 microns from single cells recorded in area 18 of cat visual cortex, and the effects on orientation tuning were studied. The receptive fields of cells at the recording and inactivation sites overlapped extensively. During the inactivation of sites where the orientation preference differed by 45 degrees or more from that of a recorded cell (“cross-orientation” sites), 65% of 54 cells tested showed significant broadening of orientation tuning, with a mean increase in tuning width (measured at half the maximum response) of 93%, and an almost fourfold increase in the relative response to the orientation orthogonal to the optimum, compared with the response to the optimum; four cells essentially lost their orientation tuning. Broadening of tuning reflected an increase in response to nonoptimal orientations and was reversible upon termination of GABA application. The effects on orientation tuning typically peaked within 10–15 min of the onset of GABA iontophoresis with 50–100 nA ejecting currents, and could not be replicated by inactivating sites where the orientation preference was similar to that of a recorded cell; when the orientation preference at the inactivation sites was within 22.5 degrees of that of a recorded cell (“iso-orientation” sites), only 3 of 22 cells showed significant broadening of tuning, and in these cases, the effects were relatively weak (mean increase in tuning width of 39% and a negligible change in the relative response to the orientation orthogonal to the optimum). The effect of inactivating “iso-orientation” sites consisted primarily in an increase in response magnitude. The difference in the magnitude of the effects on orientation tuning elicited by inactivating “cross-orientation” and “iso-orientation” sites was highly statistically significant. Additionally, inactivation of “cross- orientation” or “iso-orientation” sites elicited differential effects on orientation tuning in 10 of the 13 cells in which direct comparisons were made. It is argued that the observed broadening of tuning was due to the loss of a “cross-orientation” inhibitory input, which normally sharpens orientation tuning by suppressing responses to nonoptimal orientations.