The statistical significance of the difference between two correlation coefficients can be calculated if the sample sizes are known. However, I wonder is there also a way to calculate the difference between two pair-wise related correlation matrices? The need for that arose when I administered the same questionnaire to two samples, and the one sample appeared to score lower and with less variation on the questionnaire than the other sample did. (I calculated the mean and the SD over the correlations per sample, and both are lower for the one sample than for the other.) I could apply a pair-wise t-test for dependent samples on the differences, in which case the number of correlations would be the sample size, but I expect objections from statisticians against that procedure.