The Cu,Zn superoxide dismutase (SOD) was examined earlier and found to behave in a very unclocklike manner despite (accepted point mutation, or PAM) corrections for multiple replacements per site. Depending upon the time span involved, rates could differ 5-fold. We have sought to determine whether the data might be clocklike if a covarion model were used. We first determined that the number of concomitantly variable codons (covarions) in SOD is 28. With that value fixed we found that the observations for SOD could fit reasonably well a molecular clock if, given 28 covarions, (i) there are approximately six replacements every 10 million years, (ii) the total number of codons is 162, (iii) the number of codons that are permanently invariable across the range of taxa from fungi to mammals is 44, and (iv) the persistence of variability is quite low (0.01). Thus, the inconsistent number of amino acid differences between various pairs of descendent sequences could well be the result of a fairly accurate molecular clock. The general conclusion has two sides: (i) the inference that a given gene is a bad clock may sometimes arise through a failure to take all the relevant biology into account and (ii) one should examine the possibility that different subsets of amino acids are evolving at different rates, because otherwise the assumption of a clock may yield erroneous estimates of divergence times on the basis of the observed number of amino acid differences.