Three strategies frequently employed in individualized instruction were tested in a three-way repeated measures design that involved eight computer exercises over a period of 6 weeks. Exercises compared individual with paired study, pretests with no pretests, and serial with parallel mastery. Posttests revealed an advantage for individual study and parallel mastery, while pretesting had no lasting effect. Questionnaire preferences for paired study improved under conditions of increased cognitive strain. A mechanism involving working memory is proposed and implications for both individualized and grouped instruction are discussed. (Author)