The influence of rootstock on average fruit weight was evaluated for a subset of data from a multilocation NC-140 apple [Malus sylvestris (L.) Mill. var. domestica (Borkh.) Mansf.] rootstock trial. Data for eight dwarf rootstocks were collected at four locations for 2 years. Analysis of covariance was used to evaluate the effect of rootstock on average fruit weight when crop density or number of fruit per tree was included in the linear model as a covariate. When number of fruit harvested per tree was used as a covariate, average fruit weight was not affected by rootstock in either year in Ontario. In Michigan and Virginia, rootstock and number of fruit per tree, but not the rootstock × number of fruit interaction, were significant, so common slopes models were used to estimate least squares means for average fruit weight. In general, trees on M.27 and P.1 produced the smallest fruit, and trees on B.9, M.9 EMLA, and Mac.39 produced the largest fruit. In New York the interaction of rootstock × number of fruit was significant, so least squares means were estimated at three levels of number of fruit per tree. Both years, at all levels of number of fruit, trees on M.26 EMLA produced the smallest fruit and trees on M.27 EMLA produced the largest fruit. Average fruit weight was most affected by number of fruit per tree when Mark was the rootstock. In general, results were similar when crop density was used as the covariate, except that trees on M.27 EMLA did not produce small fruit in Michigan and Ontario.