The aim of this paper was to build an empirical and interpretable classification system of challenging behaviors in people with intellectual disability. We analyzed 19 challenging behavior labels using non metrical multidimensional scaling. Bi- and tri-dimensional MDS solutions were supported both by a multiple regression analysis of 10 bipolar scales about different characteristics of challenging behaviors and a joining and k-means cluster analyses. For multiple regression analysis, predictor variables were the values of MDS dimensions and criterion variables were the average ratings attained by bipolar scales. The bi-dimensional solution distinguishes between biological and social etiology (dimension 1) and overt and covert behaviors (dimension 2).