Community assembly theory asserts that the contemporary composition of ecological communities may depend critically on events that occur during the formation of the community; a phenomenon termed "historical contingence." We tested key aspects of this theory using plant communities in over 200 experimentally created vernal pools at a field site in central California, USA. The experiment was initiated in 1999 with construction of vernal pool basins into which different seeding treatments were imposed to evaluate the effects of dispersal limitation, order of colonization ("priority effects"), and frequency of colonization on plant community composition. We tracked the abundance and distribution of five focal species for seven years following seeding and observed strong but transient effects of seeding, as well as order and frequency of colonization. All five species occurred with higher frequency in seeded pools vs. unseeded control pools, demonstrating dispersal limitation. Three of four species exerted strong priority effects, with much higher abundance in pools in which they were seeded first compared to pools in which they were seeded in the second year of the study, one year after other species were seeded. We tested for effects of frequency of colonization using one species, the endangered Lasthenia conjugens, and observed much higher abundance in frequently vs. infrequently seeded pools for the first four years following seeding. Finally, we observed that the strength of priority effects varied significantly with water depth for one of the species groups, which demonstrates that abiotic context can strongly influence species interactions. We conclude that several aspects of historical contingence play key roles in the early formation of vernal pool plant communities. But we also observed reversals in community trajectories, suggesting that in this system historical effects may be lost within a decade.