Everyday experience suggests that individuals are adept at returning their limb to a remembered target position, for example, to pick up a glass of water. The remembered limb position is involved in establishing the motor programme for the next reach and does not modify the reach once it has started (Heath et al., 2004). This study has investigated the relative importance of vision, proprioception and efference copy of the motor command in laying down this memory. With local ethical approval, 20 healthy subjects (aged 21-23) were recruited and seated 27 cm from a target board printed on A2 paper; a start marker was placed 20 cm in front of the subject. The target board consisted of 5 randomly positioned numbered red circular targets 2 cm in diameter and with a black centre. Each trial consisted of 25 reach tasks carried out in random order. Each reach task was carried out in three stages cued by three tones 3 seconds apart. With the index finger of the dominant hand on the start marker the hand moved to the selected target on the first tone, on the second tone the experimenter passively moved the hand back to the start marker, on the third tone the subject actively moved their hand back to the target. Absolute error was measured as distance from the centre of the target. Subjects carried out the tasks both with and without blindfolds and with either active or passive initial placement of the hand on the target. In the blindfolded active trials the target position was established by the initial active reach. Blindfolded active trials were also carried out in the non-dominant hand. With vision, there was no difference (repeated measures ANOVA; P > 0.05) between the mean (±SEM) errors in active (1.4 ± 0.1 mm) and passive (1.3 ± 0.1 mm) trials although they were each very much more accurate (P < 0.05) than their corresponding blindfolded trial. When blindfolded, the active trials had a smaller (P < 0.05) error (26.9 ± 1.6 mm) than the passive trials (31.7 ± 1.6 mm). Furthermore, the blindfolded active trials performed using the dominant hand had a lower (P < 0.05) error (26.9 ± 1.6 mm) compared with the non-dominant hand (31.7 ± 2.1 mm). In this experimental scenario, limb position is clearly predominantly driven by vision. However, when blindfolded, actively positioning the hand produces a more accurate memory of target position than passively positioning the hand; this suggests that an efferent copy of the motor command may be held in memory. Memory for trials performed with the dominant hand produces a lower error than with the non-dominant hand; this could be due to a relative paucity of sensorimotor control of the non-dominant hand.