(1) The purpose of this experiment was to characterize the responses of neurons in somatosensory cortex while the hand was actively moved (stroked) across a textured surface. Surfaces consisted of horizontal gratings that varied by spatial period or ridge-groove ratio (roughness). Surfaces were attached to rectangular blocks. TOP and BOTTOM halves of each block could contain surfaces of different roughness. (2) Velocity and force of the stroke were behaviorally constrained within certain limits and continuously measured and recorded during the stroke. (3) Response samples for each neuron were obtained for repeated presentations of each surface. Statistical analyses consisted of analysis of variance and t tests across surfaces on the data of each neuron, and summary statistics on groups of neurons with similar response characteristics. The interaction effects of behavioral variables (velocity and force) were examined and found not to be significant. (4) The sample mainly consisted of rapidly adapting neurons in area 3b of somatosensory area I (SI). Three main response types were found: (a) GRADED cells showed a monotonic increase in firing rate to increasingly rougher surfaces. This effect was seen in one-third of cells studied and is consistent with other reports. These cells seem to code roughness in the magnitude of their response, (b) In some cells, response to a BOTTOM surface depended on the roughness of the preceding TOP surface. This is analogous to contrast in the visual system. These CONTRAST cells are a novel finding in the somatosensory system, (c) Some cells only responded to surfaces that were completely smooth. These "OFF"response-type cells were seen in proximity to other cells that responded in a reciprocal fashion to surfaces with ridges, but not to smooth surfaces. SMOOTH cells did not respond to punctate or passively applied stimuli, and therefore could not be classified by adaptation of the responses. (5) An increase in firing rate as spatial period (roughness) increases (with a constant ratio of ridge to groove) seems contrary to vibratory models of texture perception. As spatial period increases, temporal frequency decreases, and thus "tuned" cells should show a decreased response rate. Yet GRADED cells showed an increased response. In addition, response varied on surfaces with different groove size, where spatial period, and thus temporal period, was constant. This suggests that in rapidly adapting neurons, at least for these simple surfaces, texture is coded by the magnitude of the firing rates rather than by its temporal fidelity. Reduced response to smoother surfaces does not exclude increased phase locking, however, so that GRADED cells may still be the same cells that respond to vibration.