We test for the contribution of five strong candidate genes for obesity to quantitative variation for fatness in mice. The candidate loci are known through their major mutant phenotypes. We propose a randomization test for overall contribution of candidate genes, based on the empirical distribution of LOD scores from a quantitative trait locus (QTL) genome scan. The test is applied to data on body fat content and male gonadal fatpad weight from a QTL genome scan with an F2 population of C57BL/6J and DBA/2J inbred mice. The test is nonsignificant in this experiment for overall body fat content. QTLs detected at an experiment-wide significance level on chromosome 4, 6, 13 and 15 have effects on mean fatness of up to 19% between the homozygotes, but map to locations where there is no strong candidate gene. The test is significant for gonadal fat pad weight in males, and gives weak support for an association with the diabetes gene.