Effects of differences in mean stimulus information under two coding conditions and of sub-anaesthetic doses of nitrous oxide (15, 25 and 35% in oxygen) were investigated in two card-sorting experiments with student subjects In Expt. I, in which conventional playing cards were sorted into two, four or eight classes, the effect of the drug increased significantly with task complexity. Expt. II, in which cards bearing numerals were used, showed a drug effect which was independent of task complexity measured by mean information per stimulus. Neither result was to be explained in terms of a drug effect on the motor component of the tasks. Reasons for the difference between the two experiments are considered in relation to other evidence of effects of central nervous depressant drugs on input processes and short-term memory. The value of communication models for research on effects of drugs on human skills is discussed.