This paper assesses the value of three learning style tests when used to examine the design of educational materials for teaching computer science at a distance. The paper presents three studies where three different learning styles were used to discriminate preference and performance in different contexts. The studies indicate that the Learning Style Questionnaire and the Group Embedded Figures Test are of little value. However the Cognitive Style Analysis proved useful in discriminating performance on imagery-rich materials in a simulated learning context. The paper argues that it may be necessary to match the theoretical basis of learning style with the context in which it is used in order to gain useful information. On the whole the studies showed that the value of learning style tests may be limited.