This paper discusses the identification of various trash types in cotton (non-lint material/foreign matter) using soft computing techniques, such as, Fuzzy Logic and Neural Network based approaches. Trash identification provides the basis for computing the trash content in ginned cotton. The effectiveness of a hybrid neuro-fuzzy structure, namely the Adaptive Network-Based Fuzzy Inference System, to classify trash types is compared with other techniques. A correlation between trash content computed by Agricultural Marketing Service and those computed by the Southwestern Cotton Ginning Research Laboratory is presented.