Accuracy and speed strategies in scanning active memory

Abstract

The effect of speed and accuracy payoff on performance in a short-term memory scanning task was investigated under two conditions of familiarity of material and two conditions of payoff. Memory sets and probes were either drawn from a small well learned pool of words (thefamiliar set) or were sampled without replacement from a very large pool of words (theinfinite set). Under both payoffs the infinite memory sets were searched at a faster rate than the familiar sets. The speed payoff reduced the constant component of latencies and, contrary to previous findings, increased the rate of memory search for both sets. Errors increased from about 1% with accuracy payoff to 19% with speed payoff. Under speed payoff most of the errors were false positives for the familiar set and false negatives for the infinite set. Several models of the task were considered, but none proved entirely satisfactory.