Hypotheses about the effects of anxiety on performance in computer-assisted instruction (CAI) were formulated and then tested. High school students and college students were used in the experiments. Learning materials used were in the subjects of science, mathematics, psychology, physics, and educational research. Two settings were compared: a laboratory setting and a computer-assisted instruction setting. An IBM 1500 CAI system was used in the experimentation. The findings supported the Trait-State Anxiety Theory in the prediction that HA (High Anxiety)-Trait subjects will show greater evaluations in A(Anxiety)-State than LA(Low Anxiety)-Trait subjects in learning situations in which the subject's personal adequacy is evaluated, or which pose a threat to self-esteem. The findings were also interpreted as generally consistent with the Spence-Taylor Drive Theory, although certain results suggested that the effects of anxiety on learning may not be the same for men and women. It was concluded that investigators should evaluate sex differences in research on anxiety and learning. References are given. (MF)