1. Psychophysical thresholds were determined at 162 points in the glabrous skin area of the human hand when slowly rising, triangular indentations of controlled amplitudes were delivered with a small probe. The method of constant stimuli was used with either the two alternative forced choice or the yes-no procedure. It was found that the distribution of the psychophysical thresholds varied with the skin region. Thresholds from the volar aspect of the fingers and the peripheral parts of the palm were low and their distribution was unimodal with a median of 11.2 micrometers. In contrast, there was an over-representation of high thresholds when observations from the centre of the palm, the lateral aspects of the fingers and the regions of the creases were pooled, and the distribution was slightly bimodal with a median of 36.0 micrometers. 2. Nerve impulses were recorded from single fibres in the median nerve of human subjects with percutaneously inserted tungsten needle electrodes. The thresholds of 128 mechanosensitive afferent units in the glabrous skin area of the hand were determined when stimuli were delivered to partly the same points as stimulated for the assessment of the psychophysical thresholds. Of the four types of units present in this area the Pacinian corpuscle (PC) and rapidly adapting (RA) units had the lowest thresholds with medians of 9.2 and 13.8 micrometers, followed by the slowly adapting type I and slowly adapting type II units with medians of 56.5 and 33.1 micrometers. There was no indication of a difference between thresholds of units located in different skin areas. 3. In the region of low psychophysical thresholds there was good agreement between the thresholds of the rapidly adapting and Pacinian corpuscle units and the psychophysical thresholds, particularly at the lower ends of the samples. In the skin regions of high thresholds, on the other hand, practically all psychophysical thresholds were higher than the thresholds of the most sensitive afferent units. Moreover, simultaneous recording of nerve impulses during a detection task indicated that subjects did not detect stimuli strong enough to elicit several impulses in afferent units in this region. 4. Circumstantial evidence led to the conclusion that detection was dependent on one impulse in one or a few rapidly adapting units under optimal conditons in the region of low psychophysical thresholds, whereas it seemed unlikely that activity in Pacinian corpuscle units was crucial. 5. The findings are consistent with the interpretation that human subjects are able to detect an input consisting of a single impulse in a single rapidly adapting unit.