The proportion of N in soybean derived from biological N 2 fixation ( P atm) was estimated in a pot experiment using soil labelled with ammonium chloride (4.67 mg N pot − at 22.85, 41.40, 60.00 and 78.50 atom % 15N). A model of total N and 15N accumulation by soybean and the 15N enrichment of plant-available soil N was used to estimate P atm independent of a non-fixing reference plant, and compared to estimates of P atm derived from the use of non-fixing soybean and ryegrass as reference plants. The decline in 15N enrichment of the plant-available soil N (a soil) was described by first-order kinetics of the form, a soil = b exp(− Dt). Decline constants ( D) for the atom% 15N excess of KCl-extractable NO 3 −-N were 0.0571, 0.0225 and 0.0211 d − for soybean, non-fixing soybean and ryegrass, respectively, and were independent of fertilizer 15N abundance. By fitting the exponential function to the enrichment of the mineral-N and the NO 3 −-N pools, the 15N enrichment of soil N being assimilated by the plant at any time was predicted. Total N and 15N uptake by soybean showed a strong relationship with time of the form, TN leg = c exp(NU t or T 15N leg = c'exp(NU' t), respectively. Using these equations and the 15N enrichment of the KCl-extractable mineral-N or NO 3 −-N pools, the uptake of soil N was estimated and was then used to calculate Paim- This proportion ( P' atm) increased with time after sowing to reach a maximum of 0.6, 90 days after sowing, and was independent of the soil N pool used to estimate the uptake of soil N. Estimates of P atm calculated by the isotope dilution method varied from 0.33 to 0.84 depending on the reference plant used. Ryegrass consistently produced higher estimates of P atm than non-fixing soybean. Under conditions of rapidly decreasing soil 15N enrichment, modelled estimates of P atm were similar to estimates obtained using non-fixing soybean as the reference plant. There was good agreement between P atm calculated from the model (0.46) and by isotope dilution (0.49) at 103 days after sowing using non-fixing soybean as the reference. Failure to correct for the contribution from seed-N, when total N accumulation by the plants was small caused an overestimation of N 2 fixation because the unlabelled seed N was included in the estimation of P atm.