1. We showed previously that neurons in the dorsal spinocerebellar tract (DSCT) may encode whole-limb parameters of movement and posture rather than localized proprioceptive information. Neurons were found to respond to hindlimb movements in the sagittal plane with maximum activity for foot placements in one direction and minimum activity for placements in the opposite direction. In contrast, movement direction is not specifically encoded by response activity when movements are restricted to a single joint. 2. We now describe the spatiotemporal characteristics of DSCT directional sensitivity for the responses of 267 neurons to small amplitude (0.5 cm) perturbations of the cat hindlimb. A small platform attached to the left hind foot was perturbed along four or eight directions in the sagittal plane, eliciting significant responses in 261 (98%) of the cells. The responses typically consisted of a sequence of peaks and troughs in poststimulus spike density lasting 150 ms or more following limb perturbation. 3. Peaks of activity in particular poststimulus intervals were broadly tuned for the direction of the perturbation, as determined by fitting the firing rates recorded in response to each perturbation direction to a cosine model. The parameters of the cosine model, namely the amplitude of modulation, the direction of maximum response, and the goodness of fit to the model, were computed for each 4 ms poststimulus interval. The parameters all showed the same tendency to wax and wane with respect to poststimulus time. For each period during which the cell activity was highly correlated with the tuning model, the tuning indicated a different best direction. Thus each cell's directional tuning could be characterized by a set of tuning maxima associated with specific poststimulus times, when the amplitude of the tuning reached a local maximum and the fit to the cosine model was highly significant (R(2) > 0.85). 4. Directions of the tuning maxima for the total population of cells were not uniformly distributed within particular poststimulus intervals. There was a statistically significant directional bias for upward directed perturbations in the poststimulus interval between 20 and 40 ms, followed by a period of downward bias from 45 to 55 ms. Between 60 and 85 ms, the distribution of tuning maxima was significantly skewed backward, whereas a very strong bias for the forward direction was present at about 100 ms. 5. Because the tuning was determined from responses to very small perturbations of the limb in a given posture, it was not clear whether the responses were related to specific joint angles or muscle lengths, or whether they somehow represented the kinematics of the whole Limb. To address this point, we examined the responses of 95 cells in two animals that were each tested in two different limb positions. One position was an approximation of the normal standing position. The other position consisted of a shortening of the limb axis (with major changes in all joint angles) in one animal, or a rotation of the limb axis backward (with little change in joint angles) in the other. 6. We compared each cell's responses to the same perturbations applied in the two limb positions and found they could be identical, scaled in time or magnitude, or completely different in the two positions. A greater percentage of cells with different responses was found in the experiment with the limb axis rotated. In the other experiment, in which there were major differences in joint angles in the two positions, the responses were mostly the same or scaled in time in the two positions. We also determined the population directional biases for the two positions in each experi ment, and found that phase differences between the vectors representing population biases for the two positions were minimized when they were measured relative to the orientation of the limb axis (limb coordinates) rather than to the extrinsic vertical (lab coordinates). 7. On the basis of this result, which was consistent with the data from seven other animals, we propose that DSCT activity may represent movement kinematics in a limb-centered reference frame. The directional sensitivity of the cells is defined by the orientation of the limb axis, which is the axis connecting the hip joint with the point of ground contact. The finding that preferred directions tend to cluster about directions that approximate limb movements along the limb axis (changes in limb length) or normal to it (changes in limb orientation) further suggests a two-dimensional coordinate system defined by limb length and orientation. Because the temporal components of the directional responses seem to occur independently, we also suggest that elements of the presynaptic circuitry may provide the basis for this two-dimensional representation and also be responsible for creating and shaping its temporal domain.
