Retention harvesting (also called tree retention or structural retention), in which live mature trees are selectively retained within harvested stands at different retention levels and in different patterns (aggregated to dispersed), is increasingly being used to mitigate the negative impacts of forest harvesting on biodiversity. However, the effectiveness of combining different patterns of retention harvesting for conservation and recovery of understory vascular plants in the long-term is largely unknown. To address this gap, we compared understory vascular plant diversity, abundance, and composition between aggregated retention and five levels of surrounding dispersed retention (0% = clearcut, 10%, 20%, 50%, 75%) 15 years post-harvest. We also investigated the influence of dispersed retention on the ability of embedded retention patches to support plant communities characteristic of unharvested forests, and whether it varies by patch size of aggregated retention (0.20 ha or 0.46 ha) and position within patches (edge or interior). Species richness, diversity, and cover were higher in the dispersed retention than in the patch retention as the harvested areas favored early-seral plant species. Graminoid cover was greater at the edges than in the interior of large patches. Retention patches as small as 0.2 ha more effectively supported shade-tolerant (forest interior) plant communities when they were surrounded by higher levels of dispersed retention (as compared to patches retained within clearcuts). Overall, the combined use of both aggregated and dispersed retention within a given cutblock benefits both late- and early-seral plant species and thus could effectively conserve understory plant assemblages in harvested landscapes. Sustainable forest management should therefore consider using a range of retention patch sizes combined with varying levels of surrounding dispersed retention in harvest designs to achieve objectives for plant conservation.