Accurate estimation of the responses of understory plants to natural and anthropogenic disturbance is essential for understanding efficacy and non-target effects of management and restoration activities. However, ability to assess changes in abundance of understory plants that result from disturbance may be hampered by inappropriate sampling methodologies. Conventional methods for sampling understory plants may be robust for common, well distributed species, but may fail to adequately characterize the abundance of less-common species, which are often the taxa of management concern. I tested conventional and novel approaches to sampling understory plants to determine their efficacy (in terms of number of replicates and time required) for quantifying abundance of plants of varying frequency and spatial heterogeneity on three control and three thinned-and-burned treatment units located within the western Montana block of the Fire and Fire Surrogates Project (FFS) — a largescale investigation of the effects of fuel-hazard reduction treatments on a variety ofecosystem components. In each treatment unit, I used four sampling methods (modified Whittaker plots, Daubenmire transects, point line intercept transects, and strip adaptive cluster sampling) to estimate the cover of 24 understory species that vary in abundance. Compared to Daubenmire and point line intercept transects, modified Whittaker plots estimated cover with the lowest variances and, consequently, for the majority (67%) of species required the smallest sample sizes to accurately measure cover. However, this greater sampling efficiency was offset by increased time required to sample. For species grouped by growth-form and for common species, all three conventional sampling designs (i.e. Daubenmire transects, modified Whittaker plots, and point line intercept transects) were capable of estimating cover with a 50% relative margin of error with reasonable sample sizes (3-36 plots or transects for growth form groups; 8-14 for common species); however, increasing the precision to 25% relative margin of error required sampling sizes that may be logistically infeasible (11-143 plots or transects for growth-form groups; 28-54 for common species). In addition, all three designs required enormous sample sizes to estimate cover of non-native species as a group (29-60 plots or transects) and of individual less-common species (62-118 plots or transects), even with 50% relative margin of error. Strip adaptive cluster sampling was the only method tested that efficiently sampled less-common species: for Cirsium arvense, an invasive non-native plant, adaptive sampling required five times fewer replicates than needed for modified Whittaker plots and 20 times less than for Daubenmire or point line intercept transects. My findings suggest that conventional designs may not be effective for accurately estimating the abundance of newly establishing, non-native plants as a group or of the majority of forest understory plants, which are characterized by low abundance and spatial aggregation. Novel methods such as strip adaptive cluster sampling should be considered in investigations for which cover of these species is a primary response variable.