The present study compares how accurately two different but close velocities of movement are discriminated by populations of muscle spindle primary afferents whether or not one takes into account the direction of the movement and the preferred sensory directions of the units (i.e., the direction of movement to which the afferents are the most sensitive). The activities of 26 muscle spindle primary endings originating from the tibialis anterior, the extensor digitorum longus, the extensor hallucis longus, and the peroneus lateralis muscles were recorded in the lateral peroneal nerve. Their responses to movements imposed at two velocities (12.5 and 18 mm/s) were analyzed. These movements were straight-line movements imposed in eight directions and circular movements in both clockwise and anticlockwise directions. The encoding of the movement velocity was analyzed in two ways. First, the mean frequencies of discharge of the muscle spindle afferents were compared for the two velocities. Second, the data were analyzed using a "neuronal population vector model." This model is based on the idea that such neuronal coding can be analyzed in terms of a series of population vectors (i.e., mean contribution of all the muscle spindle afferents within one directionally tuned muscle) and by finally calculating a sum vector. The results showed no clear and consistent difference in the response frequency of the muscle spindle afferents for the two velocities of movement compared. Rather, the most consistently significant differences between the two velocities were in the lengths of the sum vectors. It is concluded that the encoding of two-dimensional movement velocity relies on populations of muscle spindle afferents coming from the whole set of muscles surrounding a particular joint, each muscle making an instantaneous, oriented, and weighted contribution to the sensory coding of the kinematics parameters.