The study of primate memory substrates, the CNS alterations which preserve conditioned responses, requires an experimental model that fulfills two criteria. First, the essential alterations must be in a technically accessible location. Second, they must persist without input from other CNS regions. The spinal cord is the most technically accessible and readily isolated portion of the primate CNS. Recent work has demonstrated that the spinal stretch reflex (SSR), the initial, wholly segmental response to muscle stretch, can be operantly conditioned and suggests that this conditioning may produce persistent spinal alteration. The present study attempted similar operant conditioning of the H-reflex, the electrical analog of the SSR. The primary goals were to demonstrate that spinal reflex conditioning can occur even if the muscle spindle is removed from the reflex arc and to demonstrate conditioning in the lumbosacral cord, which is far preferable to the cervical cord for future studies of neuronal and synaptic mechanisms. Nine monkeys prepared with chronic fine-wire triceps surae (gastrocnemius and soleus) electromyographic (EMG) electrodes were taught by computer to maintain a given level of background EMG activity. At random times, a voltage pulse just above M response (direct muscle response) threshold was delivered to the posterior tibial nerve via a chronically implanted silicon nerve cuff and elicited the triceps surae H-reflex. Under the control mode, reward always followed. Under the HR increases or HR decreases mode, reward followed only if the absolute value of triceps surae EMG from 12 to 22 ms after the pulse (the H-reflex interval) was above (HR increases) or below (HR decreases) a set value. Monkeys completed 3,000-6,000 trials/day over study periods of 2-3 mo. Background EMG and M response amplitude remained stable throughout data collection. H-reflex amplitude remained stable under the control mode. Under the HR increases mode (5 animals) or HR decreases mode (4 animals), H-reflex amplitude (EMG amplitude in the H-reflex interval minus background EMG amplitude) changed appropriately over at least 6 wk. Change appeared to occur in two phases: an abrupt change within the first day, followed by slower change, which continued indefinitely. Change occurred in all three triceps surae muscles (medial and lateral gastrocnemii and soleus). Under the HR increases mode, H-reflex amplitude rose to an average of 213% of control, whereas under the HR decreases mode it fell to an average of 68% of control. The results demonstrate that the H-reflex can be operantly conditioned.(ABSTRACT TRUNCATED AT 400 WORDS)