This study was designed to investigate how conditions of physical exercise affect human information processing. Sixteen subjects performed two information processing tasks (perception and decision) during two exercise conditions (endurance vs interval protocols) and during two control conditions (rest vs minimal load protocols). The control conditions required subjects either to perform the information processing tasks under resting conditions or while pedalling a bicycle ergometer at a minimal workload. Workload during the exercise protocols consisted of a fixed percentage of the subject's maximal workload. Each 40 min protocol consisted of five consecutive stages: practice, baseline, warming-up, exercise, and cooling-down, during which heart rate and ratings of perceived exertion were determined. In the perception task subjects had to identify a briefly presented row of three letters. In the decision task subjects had to indicate which of the outer numbers in a row of three digits was the larger. Results indicated that the two control protocols did not influence cognitive task performance; however, in the exercise protocols, increments in physical workload improved performance on the decision task and reduced performance on the perception task, while decrements in physical workload reduced performance on the decision task and improved performance on the perception task. Changes in mental task performance were not evident within protocol stages; only after stage transitions did changes in mental performance occur. We discussed possible theoretical approaches to explain these results and concluded that models advanced in the context of dual-task methodology seem most promising.