On the concordance among empirical confusion matrices for visual and tactual letter recognition

Abstract

In this article, we examine the concordance among 19 empirical confusion matrices for visual and tactual recognition of capital letters of the alphabet. As a measure of concordance, we employed an index based on within-stimulus triads of letters. Unlike correlation measures of agreement that are based on a one-to-one matching of matrix elements, the selected index directly captures the internal structures of the confusion matrices prior to the comparison. Permutation tests revealed statistically significant concordance among 166 of 171 pairs of matrices in the study. Concordance of confusion structure among tactual matrices tended to be somewhat stronger than concordance among the visual matrices.