Cotton is currently graded on its color, leaf, and preparation. Individual component measurements for color and trash are currently reported along with the cotton grades. The cotton industry now requires a method to identify all types of trash in a sample. This article presents an image processing system, test of feature invariance, and two preliminary pattern classification methods used to identify the types of trash in cotton samples. In one approach, classical grouping was performed using divisive hierarchical clustering based on a normalized Euclidian distance metric. Clustering was able to classify 568 trash objects in the training data set with 92-percent accuracy into bark, stick, and leaf/pepper categories. Separation between leaf and pepper could be handled by defining an area cutoff in the pepper-leaf continuum. In the second approach, neural networks were used to minimize identification error in a training data set. Using 75% of a 562 object data set for training, neural networks were able to classify the remaining 134 trash objects into bark versus non-bark categories with a 99.3 percent accuracy (one piece of bark misidentified). The non-bark objects were further classified by another network into either stick or a combined leaf/pepper category with 96.4 percent accuracy. Combining both networks gave an accuracy of 96.3 percent. A paper 'Neural networks and clustering to categorize cotton trash" will be available from the laboratory.