The increasing and concave age earnings profile is one of the most commonly accepted facts in economics. The human capital model attributes rising wage-tenure and wage-experience profiles to investments in worker productivity. Therefore, an implicit assumption in the human capital model is that training must be occurring throughout a worker's stay at an employer. However, the empirical training literature has thus far either implicitly or explicitly assumed that training is concentrated at the beginning of the employment relationship. Using data from the National Longitudinal Survey of Youth, we find that training in the later years of tenure is the norm rather than the exception. To eliminate the bias from belated information, we estimate a fixed effects wage equation with a two stage procedure that uses previous training and fixed individual and match-specific variables as instruments. While instrumental variables estimation is the standard method of correcting for measurement error, instrumental variables estimates are not consistent when the explanatory variable measured with error (in our case, training) can only take on a limited range of values. Nevertheless, the relative magnitudes of the OLS and the IV training coefficients provide information that enables us to correct for measurement error bias without any need for external information on misclassification rates. Our results provide strong support for the human capital model in that most wage growth beyond the first year of tenure is due to training.