Item

Abstract

In working memory tests, performance is normally assessed in terms of error rates. This is often a crude simplification, especially if testing memory for sequences. We thus developed a method to measure accuracy of performance in a spatial working memory task. Method: Objective: In working memory tests, performance is normally assessed in terms of error rates. This is often a crude simplification, especially if testing memory for sequences. We thus developed a method to measure accuracy of performance in a spatial working memory task. Method: Twelve subjects looked at a computer screen which was divided into four quadrants, 11.5 x 11.5 ° each. In a delayed-response design, subjects were presented with a sequence of one to six dots, each appearing in a random location within one of the four quadrants. After a delay of 5 s, subjects had to reproduce the sequence of the dots as accurately as possible using a joystick. We analysed both the correctness of the sequence (correct quadrant) and the accuracy of the remembered spatial location of each dot. For sequence analysis a modified Levenshtein algorithm was implemented. This algorithm evaluated deletions, insertions, substitutions in the sequences as well as swaps and calculated a correctness term (interval scale). Result: A pronounced load effect occurred for both correctness of sequences and spatial accuracy. Sequence performance declined with increasing load in an exponential fashion. At load 6, only 32 of the sequences were reproduced without error. The Levensthein correctness term dropped from 100 to 58 only. Spatial accuracy decreased linearly with increasing load from 0.75° (load 1) to 2.7° (load 6). Conclusion: The experimental design allows a quantitative approach to working memory performance on a high resolution scale. The Levenshtein algorithm is a promising tool for any sequence analysis problem in psychological tests.