Reduction of task-relevant information in skill acquisition

Haider and Frensch (1996, 1999a, 1999b) argue that people learn through practice to separate task-relevant from task-irrelevant information, and that this is a major element in the acquisition of complex skills. However, in more complex tasks, there are multiple elements that are relevant to the solution. Here, we investigate a basic example of this kind of task. Where either of two information sources associated with a task may be used to determine a response, do individuals learn to ignore one source? If so, which attributes of the two information sources bias individuals towards selecting one source rather than the other? We report two experiments, each using a variant of the alphabetic arithmetic task (Haider & Frensch, 1996). We asked participants to verify strings, such as H[4] M N O P. In Experiment 1, participants completed either three or seven practice blocks in which an error occurring in the initial letter-digit-letter triplet always coincided with an error in the post-triplet string. All participants then completed a test block in which incorrect strings included only one type of error. The large string-length effect we obtained for correct strings suggests that participants learned to ignore one of the two sources of task-relevant information. There was a higher incidence of post-triplet errors over triplet errors in the test block. Experiment 2 followed a similar procedure to Experiment 1, but with two modifications. In two conditions, we asked participants to verify strings where we located the triplet at the end of the sequence. In two other conditions, we asked participants to verify strings used in Experiment 1, but presented in reverse order and by reading the strings from right to left. Results for the two triplet-last conditions are consistent with participants reducing to the pre-triplet string rather than to the triplet. Results for the two reverse-order conditions suggest that participants reduced to the triplet, as in Experiment 1. We propose that the hypothesised information reduction mechanism gives priority to information processed first, and acts to reduce processing time.