The representation of joint position at rest and during movement was investigated in 44 muscle spindle primary afferents originating from the extensor carpi radialis brevis (ECRb) and extensor digitorum (ED) of normal human subjects. Position sensitivity was estimated for each afferent, and 43 of 44 were position sensitive. In each trial, six sequential ramp-and-hold movements (2-6 degrees, 2 degrees/s, total 24 degrees) flexed the relaxed wrist, beginning from the angle at which the afferent was just recruited. Joint position was represented by three specific features of afferent firing patterns: the steady-state firing rate during the 4-s hold period between ramps, the initial burst at the beginning of each ramp, and the ramp increase in firing rate later in the movement. The position sensitivity of the initial burst (1.27 +/- 0.90 pps/degree, mean +/- SD) was several times higher than that of the hold period (0.40 +/- 0.30 pps/degree) and not different from that of the ramp increase in firing rate (1.36 +/- 0.68 pps/degree). The wrist position sensitivities of ECRb and ED afferents were equivalent, as were their recruitment angles and angular ranges of position sensitivity. Muscle spindle afferents, both individually and as a population, were shown to represent static joint position via the hold rate and the initial burst. Afferents were recruited over the entire 110 degree range of wrist positions investigated; however, the angular range over which each feature represented joint position was extremely limited (approximately 15 degrees). The population response, based on the summed activity of the 43 afferents, was monotonically related to joint position, and it was strongly influenced by the afferent recruitment pattern, but less so by the position sensitivities of the individual afferents.