The use of spectral information for the discrimination of crops and weeds is reviewed, discussing research results covering most of this century. This includes discussions about the modelling of reflection from vegetation using optical and chemical information, indicating various problem areas (e.g. cell size, directional reflection, chemical composition). Models covered are those based on both the Kubelka-Munk theory and descriptive methods. This leads to a selection of spectral ranges, indices and ratios which might give useful information for vegetation discrimination ranging from individual plants to whole fields of crops. A range of experiments covering different spectral regions are reviewed and the usefulness of the resulting spectral information is discussed.