Using German linked employer-employee data, this paper investigates the short-term impact of on-the-job training on wages. The applied estimation approach was first introduced by Leuven and Oosterbeek (2008). Wages of employees who intended to participate in training but did not do so because of a random event are compared to wages of training participants. The estimated wage returns are statistically insignificant. Furthermore, the decision to participate in training is associated with sizeable selection effects. On average, participants have a wage advantage of more than 4% compared to non-participants. (Contains 9 tables.)