ABSTRACT
It may be of some relevance to relate the circumstances that led to the publication of the article. Over a number of years, my research interests had centered in various problems connected with the teaching and learning of foreign languages (Carroll, 1966), in particular, the prediction of success in learning foreign languages through measurements of "foreign language aptitude." (See Carroll, 1981.) Chiefly with the support of the Carnegie Corporation of New York, I had developed a battery of tests (Carroll & Sapon, 1959) that seemed under many circumstances to have high effectiveness in predicting the degree of such success. But they did not always have high validity; in fact, sometimes the validity coefficients were essentially zero. In preparation for a presentation that I had been invited to give at a symposium to be held at the University of Pittsburgh in February 1960 on Training Research and Education (Glaser, 1962), I resolved to survey all my results to try to see what factors, in the light of the varying settings in which my studies had been conducted, seemed to affect the predictiveness of the tests. It seemed that there were many varieties and conditions of foreign language instruction. One variable was the amount of intensiveness - from the highly intensive courses in which individuals were put under considerable pressure to keep up with a fast moving course in which they had as much as 5 to 8 hours of instruction per day, to the much less intensive, three-hours-per-week courses in liberal arts colleges and universities. Generally, aptitude measures were much more predictive in intensive courses than in non-intensive courses. But there were other variables: students might be uniformly highly motivated, or there might be wide individual differences in motivation. Instruction could be conducted in a rather "intellectual," formalistic way with much discussion of grammar and language features, or it could be conducted with much active practice, drill, and feedback. Thinking about the way in which all these variables might interact with one another and affect the predictiveness of aptitude measurements, I somehow got the idea of developing a general learning model that could explain my results. I remember that it was one Sunday morning in the summer of 1959 that in just a couple of hours, I made a rough draft of my model, listing classes of variables and postulating their relationship.
