Correspondence analysis is a variant of principal component analysis for use with categoric data. It has been applied to information describing the frequencies of structural certain molecular features and the skin sensitization potential of a database of organic compounds. Correspondence analysis allows the graphical study of the interrelationship between the relative frequency of structural features of non, weak, moderate and strong sensitizers and the individual class membership. Structural features resulting in protein reactivity are shown to be strongly associated with strong sensitizers; other features, not related to reactivity, are associated with non, or weak, sensitizers. The analysis also indicates structural features whose description is too vague and needs more refinement.