Latent class analysis (LCA) has emerged as the best suitable statistical tool to identify separate dimensions (latent classes) when analyzing dichotomous data; its objective is to categorize people into classes using the observed items and to identify those items that best distinguish between classes. LCA was applied to the Peters et al. delusions inventory, an inventory in a dichotomous format (Yes/No) aimed at investigating proneness to delusion in the general population. The study involved 82 patients diagnosed with a psychotic disorder and 210 well-matched healthy controls from the community. Four classes were identified in the sample: a normative one, and 3 classes traceable to the 3 major dimensions of psychosis, i.e., paranoia, grandiosity/hypomania, and the schizophrenia-like profile. The coherent multidimensional structure of the model emerging from LCA of Peters et al. delusions inventory suggests that single clusters of symptoms may be indicative of specific diagnostic categories within the spectrum of psychoses, allowing a more subtle determination of their boundaries and correlates.