Cheatgrass (Bromus tectorum L.) is a winter annual weed that presents a serious obstacle to rangeland restoration in the Intermountain West. The objective of this study was to evaluate factors regulating the size and persistence of cheatgrass carryover seed banks on semiarid sites in western Utah. We prevented current-year seed production in each of four habitats, then tallied emerging seedlings over the next 4 yr. Two iterations of the study were conducted during consecutive years. One year before initiation of each iteration, we estimated seed rain at each site. Above-average precipitation in 1998–1999 resulted in relatively high seed rain (13942 seeds·m⁻²) for the first iteration, whereas seed rain for the second iteration averaged only 3567·m⁻² because of drought conditions in 1999–2000. Mean total number of seedlings emerging from carryover seeds for the first and second iterations were 1304 and 270 seedlings·m⁻². Seedling emergence from carryover seed was positively correlated with production-year seed rain (R² = 0.69). The fraction of seed rain that carried over tended to be lower when precipitation the year following production favored fall emergence of the transient seed bank. First-year emergence of carryover seeds averaged 96% of total emergence, whereas third-year emergence averaged < 1% and was zero for six of eight cases. Carryover seeds persisted somewhat longer at the xeric black greasewood site than at more upland sites. Our study shows that cheatgrass seeds rarely persist beyond the second carryover year even on semiarid sites. Emergence from the carryover seed bank can be predicted from site attributes and precipitation patterns in previous years.