We examined the effects of forest floor coverage on overland flow generation and soil erosion in mature Japanese cypress plantations with different coverage conditions: sparse understory and litter (uncovered plots), dense fern understory and litter (covered plots), and experimental removal of vegetative floor coverage on the covered plots (removal plots). We measured soil hydraulic properties and monitored overland flow and soil erosion in three replicated plots (approximately 1 x 2 m each) representing uncovered, covered, and removal conditions. Because of the strong water repellency of the surface soil, a substantial amount of overland flow occurred, even in covered plots. Nevertheless, the annual overland flow in covered plots was 37% of that in uncovered plots. Annual soil erosion in uncovered plots was 3.7 times greater than that in covered plots. Although overland flow in removal plots was similar to that in uncovered plots, soil erosion in the former was significantly greater than in the latter. These results suggest that differences in soil erodibility between the plots were essential determining factors of erosion and were no less important than floor coverage. We quantified the effects of floor coverage and soil erodibility independently and examined the relationship between coverage and erosion by applying an erosion model. In covered plots, floor coverage prevented 95% of soil detachment by raindrops, which was the dominant mechanism in reducing soil erosion, as compared with it inhibiting overland flow and resisting sediment transport. The soil erodibility of plots with ground cover was 4.5 times higher than that of uncovered plots. This implies that simply comparing plots with different coverage conditions is not sufficient for examining the effects of vegetation coverage on soil erosion.