Theories purporting to explain the cognitive processes underlying imitation and its taxonomic distribution have proliferated in recent years but a common assumption is that imitators must adopt a model's mental or visual perspective. Data on thirty-six adult subjects were used to test the hypothesis that imitative learning of knots would suffer a decrement proportional to the disparity of visual perspectives on the task that the subject held between observation and performance. There was no significant effect of visual perspective on performance, nor was there a trend in the predicted direction. This was in spite of significant preferences on the part of subjects for minimising the angle of disparity, assessed both by their behaviour and introspective report. The cognitive basis for human imitation is discussed in the light of these findings.