Some of the problems in research on aptitude treatment interactions (ATIs), are discussed. Among these problems are the inconsistency and lack of generality to many ATI findings, especially in classroom-based ATI investigations. It is suggested that such investigations may be of limited generality due to the nature of the phenomena. The importance of careful classification of aptitude and treatment variables is also discussed. Investigation of the interaction between individual differences in prior achievement and instructional method is proposed as one way to avoid some of the problems of other approaches, including examining interactions between cognitive processes and instructional method. Finally, it is suggested that ATI researchers precede formal investigations by an observational phase: instead of beginning research by generating a prediction as to what students are likely to do in instructional situations offering a variety of alternatives, primary observation of the students' selection of options is preferable. (Author/RL)