A series of three experiments have been conducted which have examined the way that various forms of feedback influence children's progress in learning to tell the time as a result of computer-based training. The first study indicated that delays in presentation of feedback was not of assistance and this is different from some of the findings in the literature about adult learning. Somewhat surprisingly, the experiment also revealed that children in both a feedback condition and in a condition which did not provide feedback improved their post-test performance after working on a computer. The second experiment examined whether various features of the computer presentations could account for this finding, but no evidence was found to support the various hypotheses. The third experiment compared different forms of feedback. Once again, children in the no-feedback control group performed as well as those who received feedback. These counterintuitive findings are discussed in relation to Karmiloff-Smith's description of cognitive development. They also point to the need for evaluation to be carried out of 'common-sense' ideas which are used as a basis for software design.