Decomposing the Mean in the Problem Size Effect. Technical Report 2002-04

Response time (RT) distributions for large and small problem sizes were obtained under both speed- and accuracy-stressed instructions in a multiplication production task. Fitting the ex-Gaussian distributional model to the individual RT data allowed for the derivation of quantitative measures of distributional shape. Statistical results indicate that small problem size RT distributions differ from large problem size RT distributions with respect to both the mean of the normal component, m (larger for large problems) and the size of the tail, t, (larger for large problems). Accuracy instruction RT distributions also differ from speeded instruction RT distributions with respect to m (larger under accuracy instructions), and t (larger under accuracy instructions for large problem sizes only). Results support a strategy-choice explanation of the problem-size effect, and provide suggestions for the comparison of latencies obtained under speed- and accuracy-emphasized instructions