1. Retention harvesting is advocated as an alternative to intensive timber harvesting, such as clear-cutting, to better maintain or facilitate recovery of biodiversity and other ecological values in managed forests. However, it is not clear how long the benefits of retention harvests persist.

2. We investigated responses of understory vascular plant cover, richness, diversity (inverse Simpson index), and composition to a gradient in dispersed retention (2% [clear-cut], 10%, 20%, 50% and 75% retention; unharvested reference [100% retention]) at 3, 6, 11, and 17 years after harvest, in four boreal mixed wood forest types (deciduous (broadleaf)-dominated, deciduous-dominated with conifer understory, mixed, and conifer-dominated) in western Canada.

3. Understory cover and richness tended to increase in the short-term (3 years), peaked at 6–11 years with differences following the gradient of harvesting intensity, then plateaued or declined in the second decade (17 years), by which time there were minimal or no differences among harvesting levels, including the reference. Responses for diversity were minimal. In contrast, composition varied along the gradient of harvesting intensity and showed little recovery towards the unharvested condition over the 17-year period. Generally, for plant community composition, clear-cut and lower retention treatments (10%, 20%) were similar to one another but differed from the higher retention and unharvested reference treatments.

4. Synthesis and applications. Retention harvests can moderate the negative impacts of harvesting and facilitate the recovery of biodiversity. Our results suggest that for the cover, richness and diversity of understory vascular plants, this moderating influence is weak and short-lived. However, higher levels of retention can temper changes in understory composition relative to the unharvested forest, but full recovery is likely to be slow and will be complicated by post-harvest regeneration dynamics.