Interactions among seeds, mediated by granivorous rodents, are likely to play a strong role in shrubsteppe ecosystem restoration. Past studies typically consider only pairwise interactions between preferred and less preferred seed species, whereas rangeland seedings are likely to contain more than 2 seed species, potentially leading to complex interactions. We examined how the relative proportion of seeds in a 3-species polyculture changes rodent seed selectivity (i.e., removal) and indirect interactions among seeds. We presented 2 rodent species, Peromyscus maniculatus (deer mice) and Perognathus parvus (pocket mice), in arenas with 3-species seed mixtures that varied in the proportion of a highly preferred, moderately preferred, and least preferred seed species, based on preferences determined in this study. We then conducted a field experiment in a pocket mouse—dominated ecosystem with the same 3-species seed mixtures in both “treated” (reduced shrub and increased forb cover) and “untreated” shrubsteppe. In the arena experiment, we found that rodents removed more of the highly preferred seed when the proportions of all 3 seeds were equal. Moderately preferred seeds experienced increased removal when the least preferred seed was in highest proportion. Removal of the least preferred seed increased when the highly preferred seed was in highest proportion. In the field experiment, results were similar to those from the arena experiment and did not differ between treated and untreated shrubsteppe areas. Though our results suggest that 3-species mixtures induce complex interactions among seeds, managers applying these results to restoration efforts should carefully consider the rodent community present and the potential fate of removed seeds.