1. In locally anesthetized cats, extracellular recordings were made from single neurons in the lateral cruciate gyrus of cerebral cortex. These neurons responded to natural activation of stretch receptors in single, contralateral, forelimb wrist muscles, typically with phasic excitation. Low-velocity stretches, which activate primary endings of muscle spindles, excited one set of neurons at a mean latency of 11 ms; high-velocity stretches, which principally activate Golgi tendon organs and/or secondary spindle endings, excited a second set at 18 ms. The cortical neurons showing threshold responses to low-velocity stretches were found exclusively within restricted columns, 0.5-2.0 mm in diameter, which were spatially separate for each muscle. Neurons exhibiting threshold responses to high-velocity stretches were present in high density within the same columns and were also distributed, although more sparsely, outside the columns. 2. These afferent columns were located in cytoarchitectonic area 4gamma, and were shown by intracortical microstimulation to coincide with the efferent columns for contraction of the same muscle from which in input rose. Discrete afferent columns were also found for single muscles in the peridimple region of sensory cortex (area 3a), spatially separate from the columns in motor cortex. The excitation of the columns in motor cortex by these inputs from muscle was independent of that in sensory cortex. 3. The role of the cerebellum in controlling these feedback systems to motor cortex was investigated by selective cooling of interpositus and dentate nucleus, respectively. Cooling of interpositus markedly reduced transmission in the high-threshold system; cooling of dentate had a similar effect on the low-threshold system. 4. The latency, threshold, and cooling data indicated that the low-threshold system to motor cortex utilizes extracerebellar pathways including medial lemniscus and is facilitated by dentate nucleus. The high-threshold system involves a transcerebellar pathway including interpositus nucleus. Both systems transmit velocity-related information, with each showing different and complementary sensitivity and dynamic range. 5. The results are discussed with reference to the cortical load-compensation mechanism postulated by Phillips (37-38).