Unloading of the forearm produces muscle activations in the antagonist triceps brachii as a result of stretch (Paulignan et. al. 1989). The long latency component of this stretch reflex increased in magnitude when the unloading was self-initiated and predictable in time as compared to when the unloading occurred by an external unpredictable source (McAllister & McDonagh, 2004). This study investigated whether the increase in reflex gain occurs with a time course that is related to the expected unloading onset. Eleven subjects (9 female) were seated upright with their right upper arm fixed to a support. An electromagnet attached underneath their right wrist supported a 1.8-kg load. They were instructed to maintain their loaded forearm in the horizontal at an elbow angle of 105^o. The subjects released the load from the electromagnet by pressing a switch with their left thumb. In the first (′fixed′) session the time delay between the switch press and the load release remained constant whilst the subject performed 60 consecutive trials at each of the following time delays: 10, 150, 300 and 450ms. In the second (′early′) session the subjects performed a total of 140 trials during which 128 trials at the 450ms delay were interspersed with 12 trials that had unexpectedly early delays; four at each of the 10, 150 and 300 ms time delays. The amplitude of the long latency reflex was calculated as the mean rectified EMG activity from the long head of the triceps brachii in the 35 to 75 ms following the load release. The data, presented as mean ± S.E.M, was analysed using a two-way Repeated Measures ANOVA (2 session (fixed, early) x 4 delay (10, 150, 300, 450 ms)). Differences between reflex amplitudes were assessed using Tukey HSD post-hoc tests. There was a significant interaction of session and delay (p<0.01) indicating that the effect of time delay on the reflex gain was dependent on the expected time of unloading. During the ′fixed′ session the reflex amplitudes were similar at all time delays (10 ms 11.3 ± 2.2 μV; 150 ms 12.1 ± 2.3 μV; 300 ms 15.6 ± 4.7 μV; 450 ms 13.2 ± 3.8 μV). However, during the ′early′ session the reflex amplitudes at both the 0 ms (1.3 ± 0.8 μV) and 150 ms (4.3 ± 1.2 μV) delays were considerably smaller than those at either the 300 ms (13.6 ± 4.5 μV) or 450 ms (18.4 ± 6.8 μV.) delays (p<0.05). Furthermore, the reflex amplitudes at both the 10 ms and 150 ms delays were significantly smaller when they occurred unexpectedly during the ′early′ session as opposed to when they occurred expectedly in the ′fixed′ session. (p<0.05). The results indicate that it is not the duration of the delay per se which modulates the reflex gain but rather the predictability of any particular delay. We conclude that the long latency reflex gain increased with a time course relative to and at least 150 ms in advance of the expected unloading onset.