The expected behaviour of the "difference index" (Metzger, Shapiro, Mosimann & Vinton, 1968) has been examined under simple assumptions. Provided that it is only used to compare the compositions of proteins of equal or near-equal length, it can be interpreted to provide an estimate of the number of differences between the two sequences. Results with several pairs of proteins are in good agreement with estimates obtained by an alternative approach described previously (Cornish-Bowden, 1977).