1. While the subject maintained a weak contraction in his finger flexor muscles, holding the metacarpophalangeal joints in 45 deg flexion, test torque pulses were applied which caused rapid finger extension movements and electromyographic (EMG) stretch reflex responses. Before each test pulse the fingers were passively flexed or extended ('post-short' and 'post-long' trials) for about 10 s. The EMG and joint deflection responses in the two types of trial were compared after averaging. 2. In the 'post-long' trials, the EMG reflex response showed a comparative increase in latency, with a reduction of the short-latency (M1) component and an enhancement of the medium-latency (M2) component. 3. The angular deflections were larger, and the turning points of the deflections, which indicated the start of the mechanical reflex responses, occurred later in the 'post-long' trials. These differences were not seen when the torque pulse was immediately preceded by a strong, brief isometric finger flexor contraction in the test position. 4. Immediately following the return to the test position the background finger flexor EMG activity was larger in the 'post-long' trials, a difference which gradually subsided over 15-20 s. A strong, brief contraction in the test position also eliminated this inter-trial difference. 5. The results are interpreted as manifestations of thixotropic after-effects in intra- and extrafusal muscle fibres. It is proposed that the M1 component of the stretch reflex is largely a response to the 'initial burst' of impulses in primary spindle afferents.(ABSTRACT TRUNCATED AT 250 WORDS)