Delayed matching to sample: the effects of sample-set size on human performance

Abstract

The discriminative performance of students was assessed on a delayed-matching-to-sample task (DMST) using disks, presented by a computer, as stimuli. The size of the non-matching comparison stimuli was changed, for each participant, until each was 100% correct at 0.05 s delay. Then delays of 0.05, 2, 4, 8 and 16 s were paired with each of one, four, eight, 16, and 32 sample stimuli. Accuracy generally decreased over the one, four and eight samples, did not change consistently over the largest sample-set sizes and decreased as delay increased. Both delay and sample-set size had statistically significant main effects, their interaction was not significant. Fitted exponential functions gave a measure of discrimination at zero delay, a, and a measure of the rate of decrement in performance with increasing delay, b. As number of sample stimuli increased there was no systematic change in b, while a decreased most over one to four samples, decreased less with eight samples, and decreased least from 16 to 32 samples. These results suggest that the effects of varying sample-set size depends on the range of sizes studied, thus, they provide a possible explanation for some previous disparate findings. They also suggest that it might be proactive interference that leads to decreases in accuracy with increasing sample-set size.