Recordings were made from single mechanoreceptive afferents in the median or ulnar nerve of the anaesthetized monkey while the appropriate digital pad was stimulated by a textured surface moving at a constant velocity tangentially across the skin. The surface was swept across the afferent's receptive field many times, each time having been displaced sideways (laterally) by a small amount. The neural responses showed a temporal rhythm directly related to the period of the raised dots on the surface in the dimension parallel to the direction of movement. The responses also displayed direct dependence on the position of the dots and the period size in the lateral surface dimension. It was clear that only information from a large number of afferents could enable the discrimination of the textured surfaces examined. For large-dot (2.0 mm period) surfaces, increases of up to 8% in the period of the dots, in either surface dimension, produced a roughly linear increase in the mean response rate of every rapidly adapting (r.a.) afferent. There was virtually no change in the response rate of any slowly adapting (s.a.) afferent when the period of the dots was increased by similar small amounts, and the pacinian afferents (p.c.) displayed a wide range of behaviour to such period changes. In contrast, the mean response rate of p.c. afferents seemed most able of the three populations to transmit information about changes in the period in the case of small-dot (1.0 mm period) surfaces. The adequacy of a number of neural codes in accounting for all the psychophysical discrimination reported in the preceding paper (Lamb, 1983) was examined. A strong case could be made for a code involving the total or mean number of impulses evoked (a rate code), based on the r.a. afferent responses for the large-dot surfaces and, less certainly, on the p.c. afferent responses for the small-dot surfaces.