Information theory was used to qualify the difficulty of verbal discrimination (VD) learning tasks and to measure VD performance. Words for VD items were selected with high background frequency and equal a priori probabilities of being selected as a first response. Three VD lists containing only 2-, 3-, or 4-word items were created and equated for information content at approximately 20 bits. Lists were presented for 6 trials at 1.5 or 3.0 sec. per item with 20 Ss per list. Only correct responses were reinforced. The uncertainty (H) in the response distribution was calculated for each item on each trial and summed for each list. The H remaining after Trial 6 averaged over all 3-sec. conditions was 8.4 bits. For the 1.5-sec. conditions, it was 15.2 bits, indicating that information was being processed at a constant rate. The amount of information processed per trial was also constant for each list condition, although differences among lists reached borderline significance (p .05) at the 3-sec. rate. It appeared quite feasible to apply information theory to VD learning as an absolute measure of task difficulty and performance. (Author)