LIST-LENGTH EFFECT - RECOGNITION ACCURACY AND VARIANCE OF UNDERLYING DISTRIBUTIONS

One of the central results of recognition memory is the list-length effect (Strong, 1912); recognition accuracy decreases as the number of events to be memorized increases. Global-matching memory models (MINERVA 2 and search of associative memory [SAM]) predict that recognition accuracy decreases because the underlying theoretical distributions representing unstudied and studied items increasingly overlap as the list length increases because each additional item adds variability. Consequently, the ratio of the variances of these distributions should approach 1.0 (equal variance). Two experiments were conducted to test this prediction. Experiment 1 used a multilist design and found that the variance ratios were equal and not different from 1.0 for short and long lists. This raises questions about the models' explanation of the list-length effect. Possible modifications explored within the SAM framework (adding a recall process to recognition and a continuous memory assumption) did not improve the model's ability to handle the data. In Experiment 2, subjects received only a single list (short or long). This eliminated difficulty in list discrimination or a variance contribution from previous lists a possible factors responsible for the equal variance ratios in Experiment 1. There was a small but significant increase in the variance ratio as list length increased, in line with model predictions. However, none of the models could fit the data with plausible parameter values.