Learning (CAL) by the computing industry, there remained a shortage of suitable titles in some subject areas, including engineering. Investigation revealed that the most significant barrier to the exploitation of multimedia technology concerned justification and payback for the substantial amount of development effort required to produce software of this kind. It was found that the size of the potential audience for a programme was all too easily limited by the exorbitant computer system requirements and limited flexibility which tended to be built into the software by default. It was aimed to investigate whether the elements of a multimedia programme which contributed greatly to its computer system requirements, cost and inflexibility were so closely linked to its educational effectiveness. The research was experimental in nature. It involved the creation of several pieces of multimedia software, this being an experiment in itself since it allowed measurement of the amount of effort required to incorporate the various media into an educational programme. Two particularly significant pieces of software are described in detail in the thesis; an advisory system meant to promote design for testability among electronic engineers, and a CAL system offering an introduction to process planning. Both of these featured, in places, a highly interactive style, involving the dynamic generation of images and animations in response to users’ input. This represented a radical departure from the conventional approach to multimedia, which was normally based upon the sequential playback of prerecorded material. The process planning software was used with groups of students; their comments were invited and their performance was measured in a test which used a novel method to identify any students who had prior knowledge of the subject. (Correct answers from such people could not reasonably be claimed to indicate that learning had taken place, but the results of the remaining students provided a more accurate sample.) Knowing how well students had performed on each question, when taught in a variety of different styles, it was possible to compare the educational effectiveness of each approach. Since the amount of time spent adding each feature and medium to the software was known, it was then possible to identify which media had been the most efficient. It was found that interactivity is the most vital single ingredient in CAL software. Experimental results clearly showed that learning was most likely to occur when the subjects were required to play an active role. Attractive, informative media such as photographs and diagrams did generally help to facilitate learning, but the effect of these was comparatively minor. The author theorises that effective computer-based education does not necessarily involve extensive use of high quality digital video and the like; rather that the means to effective computer-based learning predate the multimedia era.