The coefficient of concordance ( W) was computed by two methods, (1) by arithmetic averaging of the underlying rhos ( rs) and (2) by transforming the rss to an equal-interval scale before averaging. The first method is traditional, while the second method is, methodologically, more correct in that rs is not an equal-interval statistic. Method 2, for which a computer program had to be written (available from the authors), yielded Ws which were, on the average, 12% larger than those of Method 1. The magnitude of difference suggests that either method is satisfactory for indicating the gross degree of concordance. However, if a comparison between matched-groups is contemplated, then the more accurate method seems appropriate. Further, it was found that Method 2 was less affected by irregularities in the data, e.g., small number of atypical Ss, than was Method 1. Five sets of data, covering a wide range of Ws and Ns, were presented and the relationship between the corrected W, sample size, and the traditional W were also considered.