Gender differences in response to computers have been widely commented upon. This paper addresses the question of how far superficial aspects of the software in use can affect the performance of boys and girls on computer-based problem solving tasks. A controlled comparison was conducted between two versions of a route planning task which differed in terms of the scenario within which the task was framed (one involving 'pirates' and the other 'picnics'). Subjects were 52 eleven and twelve year-olds. There was a significant gender by software interaction, the girls' performance being markedly influenced by which version of the software they encountered. Following refinement of the software, a replication study was conducted with a further 48 children, and the same result was obtained. The findings are discussed in terms of the effects of confidence, familiarity and other factors upon children's responses to computers.