The gene effects for yield and yield component traits in faba bean genotypes of indigenous and exotic origins were estimated using generation mean analysis of five generations (P1 P2 F1 F2 and F3) in nine crosses for ten quantitative traits using two models. By five parameter model preponderance of additive and additive x additive interaction effects was observed with respect to  days to maturity, plant height, branches per plant, clusters per plant, pod length, pods per plant and 100 seed weight while preponderance of  dominance effect was observed in days to 50% flowering, seeds per pod and yield per plant. By joint scaling test preponderance of additive effects was observed for days to 50% flowering, days to maturity, plant height, clusters per plant, pods per plant and yield per plant and of dominance and dominance x dominance interaction effect with respect to branches per plant, pod length, seeds per pod and 100 seed weight. Duplicate epistasis was present in one cross each in plant height, pod length, seeds per pod and 100 seed weight, in two crosses each in days to 50% flowering, pods per plant and yield per plant and three crosses each in branches per plant and clusters per plant by five parameter model. However, duplicate epistasis was predominantly present in all the crosses for all the traits except a single cross each in days to 50% flowering, plant height, branches per plant and seeds per pod indicating that selection for improvement may be successful in later generations. It was concluded that the traits showing preponderance of fixable component of variance should be more responsive to selection than the traits with non-fixable component of variance and accordingly the preferable breeding strategy for faba bean improvement has been suggested.