An experiment was made to determine the relation between performance in a linear information processing system, level of training and the input load on the system. Level of training was determined by number of problem experiences. Load was defined as the number of task units to be processed. The task was to detect, display, transform, re-display and correlate data concerning the temporal position and direction of sequences displayed on a rectangular grid. Two noise dots were included with each display presentation. Experimental variables were presented in a mixed factorial design. Experimental problems had a constant number of sequences to be processed after a short build-up period. The sequences in any one problem appeared at regular intervals and all bad the same time duration. Four experimental problems presented 2, 4, 8 or 12 sequences to the subjects. Analysis of total error indicated that performance is a linear function of load, and that an interaction between total error and system position resulted in no significant reduction in error as a function of training.