NH3/ was the principal product from soybean bacteroids, prepared by various procedures, when assayed in solution in a flow chamber under N2 fixation conditions. In addition, small quantities of alanine were produced (reaching 20% of NH3/ under some conditions). Some 15N was assimilated by bacteroids purified from soybean root nodules on Percoll density gradients and shaken with 15N2 and 0.008 atm O2. Under these conditions, accounted for 93% of the (15)N fixed into the soluble fraction. This fraction contained no measurable [15N]alanine. Neither these bacteroids nor those prepared by the previously used differential centrifugation method, when incubated with exogenous alanine under non-N2-fixing conditions, gave rise to NH3 from alanine. Therefore, contamination of bacteroid preparations with enzymes of plant cytosolic origin and capable of producing NH3 from alanine cannot explain the failure to detect [15N]alanine [as reported elsewhere: Waters, J. K., Hughes, B. L., II, Purcell, L. C., Gerhardt, K. O., Mawhinney, T. P. & Emerich, D. W. (1998). Proc Natl Acad Sci USA 95, 12038-12042]. Cell-free extracts of the bacteroids as used in the 15N experiments contained alanine dehydrogenase and were able to produce alanine from pyruvate and. Other experiments with alanine dehydrogenase in extracts of cultured rhizobia and bacteroids are reported and discussed in relation to the 15N experiments. Possible reasons for the differences between laboratories regarding the role of alanine are discussed. It is concluded that NH3 is the principal soluble product of N2 fixation by suspensions of soybean bacteroids ex planta and that should continue to be considered the principal product of N2 fixation which is assimilated in vivo in soybean nodules.