Biological nitrogen fixation (BNF) is an important but often elusive component of N budgets. We are estimating BNF by two regionally important winter cover crops, woollypod vetch (Vicia dasycarpa) and bell beans (Vicia faba), using the difference method and the natural abundance method. To determine the percent of N derived from the atmosphere (%Ndfa) by the difference method, we grew vetch and bell beans in single-species plots adjacent to plots of reference species, oats (Avena sativa) and mustard (Brassica japonica). BNF estimates from the difference method varied with choice of reference species and were often negative because of greater reference species biomass production. The fundamental assumption of this method, that legume and reference species access N from similar soil pools and assimilate it at similar rates, was violated. The natural abundance method, however, uses reference plants to represent the d ¹⁵N signature of soil N available for plant growth instead of the amount of available N. Based on a trial in 2002-03, we suspected that a long history of legume cover crops lowered the soil d ¹⁵N, making it similar to the legumes' d ¹⁵N. In 2003-04 we planted legume and reference species in fields with a range of 0-20 years of annual legume cover cropping. The results did not show a clear relationship between d ¹⁵N and cover crop history. BNF estimates were generally high (for vetch with oats as reference, 42-95 %Ndfa; for bell beans with oats as reference, 87-103 %Ndfa). Such high values are questionable, particularly in fields with high levels of available soil N. The choice of reference species had a lower effect on natural abudance %Ndfa estimates than by the difference method estimates. Neither method provides a reasonably small range of fixation estimates for use in budgets.