Escape from specialist natural enemies is frequently invoked to explain exotic plant invasions, but little attention has been paid to how generalist consumers in the recipient range may influence invasion. We examined how seed preferences of the widespread generalist granivore Peromyscus maniculatus related to recruitment of the strongly invasive exotic Centaurea stoebe and several weakly invasive exotics and natives by conducting laboratory feeding trials and seed addition experiments in the field. Laboratory feeding trials showed that P. maniculatus avoided consuming seeds of C. stoebe relative to the 12 other species tested, even when seeds of alternative species were 53–94% smaller than those of C. stoebe. Seed addition experiments conducted in and out of rodent exclosures revealed that weakly invasive exotics experienced relatively greater release from seed predation than C. stoebe, although this was not the case for natives. Seed mass explained 81% of the variation in recruitment associated with rodent exclusion for natives and weak invaders, with larger‐seeded species benefiting most from protection from granivores. However, recruitment of C. stoebe was unaffected by rodent exclusion, even though the regression model predicted seeds of correspondingly large mass should experience substantial predation. These combined laboratory and field results suggest that generalist granivores can be an important biological filter in plant communities and that species‐specific seed attributes that determine seed predation may help to explain variation in native plant recruitment and the success of exotic species invasions.

Indirect interactions are important for structuring ecological systems. However, research on indirect effects has been heavily biased toward top-down trophic interactions, and less is known about other indirect-interaction pathways. As autogenic ecosystem engineers, plants can serve as initiators of nontrophic indirect interactions that, like top-down pathways, can involve both trait-mediated indirect interactions (TMIIs) and density-mediated indirect interactions (DMIIs). Using microcosms, I examined a plant --> predator --> consumer interaction pathway involving the exotic autogenic ecosystem engineer Centaurea maculosa; native Dictyna spiders (which exhibit density and trait [web-building] responses to C. maculosa); Dictyna's insect prey, Urophora affinis; and Urophora's host plant (a secondary receiver species) to quantify DMIIs and TMIIs in an autogenic engineered pathway. Both DMIIs and TMIIs were strong enough to reduce Urophora populations, but only DMIIs, which were 4.3 times stronger than TMIIs, were strong enough to also reduce Urophora's fecundity and increase the fecundity of its host plant. Prior field studies support these results, suggesting that the differences between DMIIs and TMIIs are even stronger in nature. This study illustrates that autogenic ecosystem engineers can initiate powerful indirect interactions that generally parallel predator-initiated interactions but also differ in important functional ways.