CONSTANT RATIO RULE FOR CONFUSION MATRICES FROM SHORT‐TERM MEMORY EXPERIMENTS

Abstract

Clarke (1957) has presented evidence that a constant ratio rule holds for auditory signal identification experiments. That is, the subset of a large confusion matrix may be used to predict a smaller confusion matrix based on data gathered under identical conditions. Evidence is presented that the constant ratio rule also holds for confusion matrices derived from short‐term memory experiments. The 10 times 10 confusion matrix presented by Conrad (1964) is ‘predicted’ from the appropriate submatrix of two 20 times 20 matrices, with moderate success. Besides confirming the applicability of the constant ratio rule to short‐term memory data, the experiment also provides indirect support for Conrad's (1965a) contention that order errors are a consequence of confusions.