Miocene-Pliocene strata in the Pacific Northwest preserve a rich record of landscape evolution and coincident faunal shifts. Considerable research efforts in the past century have been aimed at understanding major drainage reorganization and its relation to tectonics, volcanism, climate change, and aquatic biota. Many studies have focused on fish fossils, which show that Miocene fish diversity, particularly salmonoids, displays great adaptive plasticity. However, the details and mechanisms for river reorganization are still debated. Here we present new and recent detrital zircon provenance results from modern and ancestral river sands collected throughout Oregon, Washington, and Idaho. We synthesize our new results and interpretations with existing paleontological evidence for basin isolation and drainage capture. Detrital zircons from the Columbia Basin (CB) consistently show populations derived from the Snake River Plain (SRP) throughout late Miocene-Pliocene time. However, comparisons of Miocene-Pliocene detrital zircons from the CB to modern major rivers and tributaries in the CB and SPR show that the upstream eastern SRP is a major contributor. CB strata do not require zircons sourced from the western SRP, where Pliocene Lake Idaho existed in a large, deep, and occasionally internally drained basin. Based on the age and provenance results, we suggest that the transiting Yellowstone Hotspot divided the modern SRP into two basins: the western basin was isolated and possibly closed, while the eastern basin drained northward into the modern Clark Fork and Columbia Rivers. This scenario is consistent with fish, mollusk, and rodent fossil evidence from the SRP and CB. In addition, the detrital zircon data indicate a Miocene confluence of the Columbia and Clearwater rivers south of the Saddle Mountains anticline, but north of the current Columbia-Snake River confluence. We also find that the Salmon River may have been captured by the Clearwater River sometime between 4.6 and 8.5 Ma. Prior to this time, the Salmon River likely drained into the SRP. Lastly, we find that faunal localities in southern Oregon suggested to contain evidence for fluvial connection between the western SRP and California are ~2.5 Ma, younger than the incision of Hells Canyon.