Significant problems regarding the measurement technique currently used to choose nerve roots for sectioning in the selective dorsal rhizotomy procedure have recently been reported. To better understand the source of these problems, a series of six experiments was performed in which the selective rhizotomy technique was applied to cats that were either intact, decerebrate, or spinalized. Measurements were made before and after partial rhizotomy. In decerebrate preparations, large, spontaneous changes in reflex threshold were observed over short periods of time, especially after partial rhizotomy was performed, and threshold changes greater than 1000% could be observed over 10-minute periods. Using constant-current stimulation of the dorsal root at threshold, the response of each ipsilateral leg muscle demonstrated frequent changes, and changes coincided with the variability in threshold estimation. In addition, very low thresholds were measured (0.1 to 0.25 mA) in half (3 of 6) of these experiments, these measurements being well below the currents customarily used for intraoperative decision making. Stimulation at twice threshold was found to consistently increase the pathological quality of the responses observed. Although intact animals never displayed contralateral responses when stimulated at threshold, such contralateral responses could readily be elicited at twice threshold. These laboratory results raise further doubts regarding the reliability of the measurement techniques now widely used for selective dorsal rhizotomy.