Research into the effectiveness of learner control in tutorial computer assisted instruction has produced equivocal results. In order to analyze these results, learner control is defined within a general model of decisions and information in instructional systems. When the research is categorized according to this model, the effectiveness of learner control is confirmed and a preliminary conclusion concerning the importance of information is reached. A case is made for an instructional design focus in learner control research based on model building rather than hypothesis testing.The notion of learner control has long held intuitive appeal for developers of computer assisted instruction (CAI) but its apparent potential for improving learning has never been experimentally established. Many studies comparing learner controlled materials with program controlled versions have contributed a large body of experimental evidence but no consensus has emerged. This article describes a meta-analysis of the body of research, the goal of which is to take a first step toward a consensus on the effective use of learner control.Examination of the research shows two major problems in drawing broad conclusions. First, the studies are conducted in conditions that are difficult to compare. Clearly the effectiveness of learner control in any study is partly dependent on the profile of the learner population, the topic of instruction and many other factors. Because the goal of this analysis is to establish the invariant aspects of effective learner control, these conditions are ignored. Any patterns which emerge through this experimental variation will be strong enough to merit further attention.The second problem in comparing experimental results concerns the definition of learner control. The control being investigated varies widely from study to study. The strategy in this article is to define a model of control which is then used 1 0 1994, Baywood Publishing Co., Inc.to categorize all the experimental systems. The patterns which emerge from this categorization lead to claims for the effectiveness of learner control.The primary claims are based on [U]' in which Kinzie and Sullivan speculate that the equivocal findings concerning learner control might result from differences in degree and type of control measured. Hence, the model is designed to identify the responsibility entrusted to the learner in each CAI environment: what kind of control and how much control. Applying the model to reported research leads to the conclusion that some amount of learner control is important but the type is not.Some secondary conclusions are suggested concerning a recent focus of research. Steinberg, in a review of learner control papers [l], notes increasing attention being paid to learner control with adaptive advice. The premise underlying this research is that learners taking control of their own learning should be provided with the information and strategies needed to make those choices. The analysis gives qualified support to the claim.In the ...