This paper describes an empirical study of fun, usability, and learning in educational software. Twenty five children aged 7 and 8 from an English primary school participated. The study involved three software products that were designed to prepare children for government initiated science tests. Pre and post tests were used to measure the learning effect, and observations and survey methods were used to assess usability and fun. The findings from the study demonstrate that in this instance learning was not correlated with fun or usability, that observed fun and observed usability were correlated, and that children of this age appeared to be able to differentiate between the constructs used to describe software quality. The Fun Sorter appears to be an effective tool for evaluating products with children. The authors discuss the implications of the results, offer some thoughts on designing experiments with children, and propose some ideas for future work.