Previous studies of human contrast adaptation employing visually evoked potentials (VEP) have revealed contradictory results, namely, either a reduction or an enhancement in VEP amplitude. In a cross-adaptation experiment, we explored the possibility that differences in the temporal frequency of adapting and test patterns played a role. Phase-reversing checkerboard stimuli [1-deg check size, temporal frequency 8.5 or 17 reversals per second (rps)] served as adaptation and test pattern with contrasts of 0 or 97%. In 13 subjects, we recorded both retinal (PERG) and cortical (VEP) steady-state responses simultaneously. In a balanced block design, all four combinations of the temporal adaptation and test frequencies were employed. Contrast adaptation reduced the PERG amplitude by about 20% in every temporal condition (P < 0.001). The VEP amplitude was strongly affected by adaptation, but the effect differed in magnitude and sign depending on condition: With identical adaptation and test frequency, amplitude was reduced by 15% (P = 0.07) at 8.5 rps and by 38% at 17 rps (P < 0.05). Adapting at 8.5 rps and testing at 17 rps had a tiny (14%) insignificant effect, whereas adapting at 17 rps and testing at 8.5 rps revealed an amplitude enhancement of 27% (P < 0.05). These strong temporal cross-adaptation effects (in the VEP, but not in the PERG) suggest that the adaptable cortical mechanisms (gain control) can be narrowly tuned in their temporal properties. A sizable adaptation effect can even change its sign when varying the temporal frequency by a factor of two. This finding resolves contradictions between previous VEP adaptation studies and reconciles them with psychophysical findings.