The automated cleaning, grading, and monitoring of grain throughout the grain handling system would maintain, if not improve, Canada's ability to be successful in the global grain market. A machine vision system is currently being developed for use with such systems in the Department of Biosystems Engineering, University of Manitoba. One measurement characteristic that is relatively easy to use is the reflectance characteristic of grains. Reflectance characteristics of 8 cereals, 3 oilseeds, 8 pulse seeds, and 27 specialty seeds were measured using a spectrophotometer (Model: Cary 5, Varian Canada Inc., Mississauga, ON). Using Canada Western Red Spring (CWRS) wheat samples, the effects on reflectance characteristics of growing region, moisture content, grade, and amount of foreign material were quantified. To assess the capability of reflectance features for grain classification, thirteen features were extracted from the reflectance data based on slope-ratio, ratio, and normalized area. Discriminant analysis using the hold-out method was used to determine the classification accuracies. Procedure STEPDISC was used to determine the contribution of each feature to the model. Reflectance characteristics successfully classified (100% accuracy) the oilseeds, seven of the eight classes of cereals, five of the eight classes of pulses, and twenty of the twenty-seven classes of specialty see s. Ratio features contributed more to the classification accuracies than did the slope-ratios or the area under the reflectance curve features. Based on the intuitive selection of features, the wavelengths that best classified the bulk grain samples were 800, 1050, and 1250 nm. Classification accuracies for cereals and pulses were higher when normal estimation was used. Reflectance characteristics did not successfully classify the grading characteristics of CWRS wheat.