Seed supply often limits the size and scope of restoration projects that require active revegetation. To meet demand from more and larger tallgrass prairie restoration projects in the Great Plains, U.S.A., seed is wild-harvested-collected from remnant habitats-using agricultural combine harvesters. We investigated the potential impacts of wild-harvest by comparing prairie remnants of northwestern Minnesota that varied in their histories of harvest frequency but were otherwise similar. We asked: (1) Do wild-harvested prairies differ in species composition from unharvested prairies? (2) If so, can life history traits be used to predict the response of prairie communities to wild-harvest? We conducted a retrospective study of 17 prairies harvested for seed frequently (annually/biennially), infrequently (2-3 times), or not at all. We sampled vegetation at 45 points within each site, recording all species present within 0.25 m2 quadrats. To address the first question, we used non-metric multidimensional scaling and Mantel tests, followed by analysis of variance contrasts to identify any species less likely to occur on frequently harvested sites ('harvest-negative'). For the second question, we used logistic regression to test whether lifespan, clonality, and seed production predicted harvest-negative species. Plant community composition in frequently harvested prairies differed from that of infrequently or unharvested prairies. Fourteen species, generally short-lived and non-clonal, were classified as harvest-negative. Our results suggest that frequent wild-harvest disrupts reproduction of species relying on seed, and that life history traits may provide a basis for predicting a species' response to wild-harvest. [ABSTRACT FROM AUTHOR]