This paper shows that marital status and gender crucially impact whether individuals receive certain types of on-the-job training. Using data from the British Household Panel Survey, we show robust evidence that when training is self-financed, married workers have significantly lower participation rates, whereas women have higher rates. The correlation between demographic characteristics and the likelihood of receiving employer-sponsored training is instead much weaker. We rationalize the relationship between training incidence and marital status with a simple two-period collective model of the household with limited commitment, where contemporaneous training decisions affect future bargaining power within the household. The core prediction of the model is confirmed empirically: the likelihood to participate in self-financed on-the-job training is negatively affected by higher levels of (a proxy for) intra-household bargaining power of the spouse of the individual. The results suggest there is scope for policy to increase workers' training participation rates by targeting individuals with weaker bargaining power within the household.