Five cats were trained to perform a forelimb ballistic flexion on a reaction time paradigm including an upper limit of about 400 ms for reinforcement (food pellets). They were implanted with a cyrogenic probe thermically insulated, except at the tip, by a vacuum jacket (outer diameter, 1.1 mm). Four cats had the probe inserted into the ventrolateral thalamic nucleus (VL), contralateral to the moving limb. During cooling they showed increased reaction times, which remained constant throughout daily sessions performed during many weeks, independent of the foreperiod but varying from 25 to 100 ms according to the subject. The temperatures used to upset the reaction times varied from +10 decrees C to -8 degrees C, depending on the localisation of the probe and on the insulation of the silver tip used to prevent nervous tissue reaction, but for each subject the reaction times always increased when the temperature was lowered. The fifth cat, with a probe inserted between VL and the Centre Median, showed a decrease of reaction times on cooling to 0 degrees C and an increase of the reaction times for a cooling at -10 degrees C. For one of the four cats with a probe properly inserted into the VL, strain-gauges were stuck on the lever to measure the latency of the decrease of the pressure exerted by the subject when the subject initiated the forelimb flexion in response to the CS. Reaction times and latencies of pressure changes were closely correlated with the movement onset, and they were equally delayed during cooling. This result demonstrates that it is not by slowing down movement velocity that reaction times are upset during VL cooling but by delaying the movement onset.