The paper reexamines the employer size-wage puzzle using NLSY79 data. The empirical results show that even for those who never receive any training from their employers, size-wage premium still exists and is quantitatively important. Wage increases associated with receiving on-the-job training are less in large establishments than in small ones. In addition, there is no evidence that starting wages in large establishments are lower than in small establishments. Theories that explain the size-wage puzzle using training and other endogenous productivity differences are not consistent with these new findings.