Abstract
An apparent discrepancy among several evaluations of the sixth-order light-by-light contribution to the muon magnetic moment anomaly has been analyzed in detail. Possible causes examined include the register overflow due to the singular integrand, the malfunction of the adaptive mechanism in a Monte Carlo integration routine, the skewed nature of the error distribution for nonsquare-integrable integrals, and the underestimation of errors in an extrapolation procedure. It is shown that the disagreement arises mainly from the second and fourth causes in the calculation of Samuel and Chlouber, and can be resolved by increasing their error estimate by a reasonable amount.