Several attempts to test the Gestalt hypothesis of autonomous change in the memory trace have shown trends which, though mutually consistent, do not fit the Gestalt hypothesis. Expt I aimed to reproduce these effects. Two figures were used (circles with gaps of 15° or 60°), one retention interval (15 sec) and two methods of testing: reproduction, in which the subject first copies the figure and then draws it from memory; and recognition, using the method of identical stimuli. In the recognition test, subjects tended to judge an identical 60° gap as ‘larger’ after 15 sec, implying closure of the gap in the memory trace, but the 15° gap showed no effect. No reliable changes in gap size occurred using the method of reproduction. Expt II tried to increase the sensitivity of this method by stressing the importance of reproducing the gap accurately. Again results were negative. Expt III tested the hypothesis that the recognition effect found was due to guessing habits. The initial stimulus was presented as before, but the recognition stimulus was replaced by the momentary illumination of a blank card purporting to be a brief presentation of the recognition figure. Subjects were thus forced to guess. The pattern of responses was almost identical with that found in the previous recognition study. Conclusions are: (1) the consistent trends previously found using the method of identical stimuli are probably due to guessing habits; (2) there is no evidence for autonomous change in the short-term retention of form.