Reliability estimates were calculated for the Mexican-American and non-Hispanic White standardization samples at eleven age levels on the Kaufman Assessment Battery for Children using computed variance components. These estimates were compared as ratios of error variances and evaluated using F distributions. Ratios for the two groups at all age levels proved to be nonsignificant, indicative of statistically equivalent reliabilities for the two groups. Coefficients were, with two exceptions, similar in magnitude. Although nota proof of nonbias, this demonstration supports the notion that error in this instance is a function that is constant across groups.