The effects of the planting method (transplanting vs. direct seeding), rice cultivar (Sen Pidao vs. Phka Rumduol), and herbicide application (admixture of bentazone and cyhalofop-butyl) on weeds and weed seed banks were quantified in unflooded, shallowly flooded, and deeply flooded paddy fields in Cambodia in 2005 and 2006. Broad-leaved weeds infested more toward maturity in 2006 than in 2005, particularly in directly seeded plots. Weed dry weights at pre-heading and/or maturity were consistently reduced by herbicide application and Phka Rumduol cultivar, while weed numbers increased under unflooded condition. The proportion of sedges was consistently larger in directly seeded and non-herbicide plots at pre-heading. Larger numbers and dry weights of sedges and total weeds in 2005 caused larger seed bank sizes of sedges and total weeds in 2006, which further caused their infestation in 2006. A greater weed dry weight at 62 days after sowing in 2005 resulted in larger seed banks of Cyperus iria and Fimbristylis miliacea, which were most severe under non-herbicide, direct-seeded treatment, while that at rice maturity resulted in larger seed banks of Cyperus difformis, Scirpus juncoides, and Lindernia antipoda. Overall, sedges shared the majority of the total weed seed bank, followed by broad-leaved weeds and then grasses. A lower yield in 2005 led to significantly larger seed bank sizes of sedges, but not of grasses or broad-leaved weeds. Rice yield reduction was consistently related to larger numbers of sedges by heading and those of grasses at maturity.