Uncertainty of the roles of proprioception and efference copy in visual spatial perception persists. Proprioception has won back some support recently mainly on the evidence gained from physiological experiments in man, and rather than being mutually exclusive, the two mechanisms have been presented as collaborating. Another view supported by human and animal experiments states that current visual spatial perception may be served by efference copy whereas proprioception is responsible for temporal adaptations of the system. Certain characteristics of visuomotor cells of the monkey cortex can be explained assuming an efference copy input. Anatomical data from different sources are not easily reconciled. Disagreement about the nerve pathway of muscle afferents weakens arguments based on the results of open loop experiments involving nerve lesions in monkeys. The assumed presence of Golgi tendon organs in human extraocular muscles is no longer tenable and instead, palisade endings similar to those of cats and monkeys and other, irregular nerve endings are described. But man differs in having a limited and patchy distribution of neurotendonous endings and moreover, they may develop only after infancy. The allocation of a sensory function to palisade endings in myotendinous cylinders appears secure, at least in cats. Detailed examination of muscle spindles in man reveals anomalies of structure sufficient to question their capacity to provide useful proprioceptive information. One of the anomalies is the atrophy of intrafusal muscle fibres, present in both infant and adult muscles, and it is proposed that the redundant sensory endings, which do not appear to degenerate, search for new targets and may account for the random presence of tendon nerve endings. These observations place the role of proprioception in human extraocular muscles in jeopardy; they are unsupportive of the recent physiological studies and favour efference copy.