Beginning with a lower half matrix of similarities or dissimilarities between all distinct pairs of objects of stimuli, MAPCLUS yields a discrete representation of those stimuli in the form of a solution in which the clusters are not constrained to the nested form characteristic of hierarchical clustering. There are at least two and usually three separate groupings of information which MAPCLUS must be given: the program parameters, the input data, and optionally, an initial configuration. The paper explains how to run the program, noting how to adjust program parameters, how to select among competing solutions, and how to choose an appropriate number of clusters. It concludes with examples of interpreting the output. Eleven references and technical data are appended.