Theory presents two broad channels through which profit sharing can increase worker training. First, it directly increases training by alleviating hold-up problems and/or by encouraging co-workers to provide training. Second, it indirectly increases training by reducing worker separation and increasing training investments' amortization period. This article provides the first attempt at separately identifying these two channels. We confirm a strong direct effect, but also identify a weaker, more tenuous indirect effect. This suggests that profit sharing's influence on training is unlikely to operate primarily through its reduction on separations while simultaneously presenting the first evidence confirming the prediction of an indirect causation.