Predicting N availability from legumes to a subsequent crop has been problematic. We tested the hypothesis that corn (Zea maysL.) grain yield and whole plant N accumulation could be predicted from N mineralization indexes of soil samples containing representative amounts of incorporated residues from the previous crop. Soil samples were taken from a crop rotation study conducted at four locations in Minnesota, in which corn was grown following eight crop treatments, including fallow, alfalfa (Medicago sativaL.), soybeans [Glycine maxL. (Merr.)], corn, and wheat (Triticum aestivumL.). Corn received from 0 to 224 kg of fertilizer N/ha. Soil was procured from the plow layer during the 2 weeks before planting and to 1.5 m (for inorganic N) within 1 week after planting. Subsamples were subjected to acid permanganate, autoclave, and glucose extractions, inorganic N determination, and aerobic and anaerobic incubations. With stepwise multiple regression, 1 week of aerobic incubation contributed as much as did incubation times up to 12 weeks to models of grain yield and total N uptake at physiological maturity. Results of acid permanganate, autoclave, and glucose extractions, and of anaerobic incubation did not consistently contribute to the models. Over all locations, topsoil inorganic N and 1 week of aerobic incubation explained between 65 and 81|X% of the variability in grain yield and total N accumulation of nonfertilized corn. For fertilized corn, N application rate alone accounted for the majority of variability in grain yield and total N uptake. Two independent crop rotation experiments provided data used to validate the predictive capability of the regression models. Despite promising relationships derived from the initial experiment, results from validation experiments were not reliably predicted by these equations. Although analyses of soil samples containing crop residues for inorganic soil N and a particular N mineralization index may relate well to yield and N uptake by corn in a given year, variability among years may preclude general use of these models for predictive purposes.