This paper reports on the results of a study of English children's performance on a computer mathematics assessment compared with a pencil-and-paper assessment. Two matched samples of children were each assessed on one of two mathematics pencil-and-paper tests and assessed a month later on a cloned computer test. The performance scores were better on the computer tests than on the pencil-and-paper tests, although this was not statistically significant in every case. The paper goes on to discuss some of the differences between performance and approach to the question between the two media, and explores possible reasons for these differences. In conclusion, the computer tests were found to have an overall positive effect on children's performance, although not for every child and, in some instances, the computer assessment limited the way in which a question could be answered.