This study is a partial replication, with variations, of the experiments of McGarrigle & Donaldson (1975) and Light et al. (1979). Sixty-four 5-year-old children were tested on traditional Piagetian tests of conservation of number under a standard control condition, and one of two experimental conditions: one (M) in which the task materials were apparently accidentally transformed by a mischievous monkey manipulated by a second experimenter, and another ( I ) in which the transformation was made to appear irrelevant to the main purpose of the task. The frequency of initial conserving responses was higher in the two experimental conditions than in the control, and this effect was strongest for the I condition. All groups of children but one appeared to exhibit a response set such that initial judgements were preserved in the second test condition. These results are discussed in relation to those of the other two studies, and their implications for Piagetian theory are considered.