Due to changing cropping practices in perennial grass seed crops in western Oregon, USA, alternative rotation systems are being considered to reduce weed infestations. Information is generally lacking regarding the effects of alternative agronomic operations and herbicide inputs on soil weed seed bank composition during this transition. Six crop rotation systems were imposed in 1992 on a field that had historically produced monoculture perennial ryegrass (Lolium perenne L.) seeds. Each system plot was 20 x 30 m, arranged in a randomized complete block design, replicated four times. Twenty to thirty soil cores were sampled in June 1997 from each plot. The weed species composition of the cores was determined by successive greenhouse grow-out assays. In addition to seed density, heterogeneity indices for species evenness, richness, and diversity were determined. The most abundant species were Juncus bufonius L. and Poa annua L. Changes in seed bank composition were due to the different herbicides used for the rotation crop components. Compared to the other rotation systems, no-tillage, spring-planted wheat (Triticum aestivum L.) and oat (Avena sativa L.) reduced overall weed seed density and richness, but did not affect weed species evenness or diversity. When meadowfoam (Limnanthes alba Hartweg ex Benth.) succeeded wheat in rotation, weed species richness was unaffected, but evenness and diversity were reduced, compared to the other rotation systems. For meadowfoam in sequence after white clover (Trifolium repens L.), crop establishment method (no-tillage and conventional tillage) had no effect on weed seed species density, evenness, or diversity.