The relatively obscure etiology and pathophysiology of schizophrenia cannot be completely ascribed to a paucity of ``®ndings''. In fact, the several hundred thousand publications on schizophrenia in the past century have ``reported'' thousands of distinct ®ndings. While a majority of such ®ndings have subsequently not been corroborated, many other ®ndings have been replicated and are currently considered to be ``established''. Several contributors to this issue of the Journal participated in a two-day workshop to clarify constructs that attempt to explain the nature of schizophrenic illness. The complete list of participants included Drs. Francine Benes, William T. Carpenter, Jr., Lynn DeLisi, Peter Falkai, Robert Freedman, Patricia Goldman-Rakic, Anthony Grace, John F. Greden, John M. Kane, Matcheri Keshavan, Peter Liddle, Robin Murray, John Olney, Rajiv Tandon, John Waddington and Daniel Weinberger. Since models are designed to explain a body of facts, the set of established ®ndings in schizophrenia that such models should explain needs to be explicitly stated. In conjunction with Drs. Matcheri Keshavan and Richard J. Wyatt (and expanded from Wyatt et al., 1988), a list of 72 ®ndings was developed and all participants were asked to rate the reproducibility of each ®nding and whether it was intrinsic or primary to schizophrenia on a 0±4 scale (Table 1). The average of these ratings is presented in Table 1. As data about various aspects of schizophrenia have burgeoned, models or constructs around which such data can be organized become increasingly important. Models are best conceptualized as working hypotheses that have both an explanatory and a predictive function; i.e., they must explain a certain body of ``established'' facts and also generate a set of testable predictions. In the absence of ``unifying'' hypothesized constructs, the ®eld might become innundated with large mounds of undigested data that collectively do not make sense (this is probably already the case!). We currently cannot see the forest for the trees. Unfortunately, since a vast majority of scienti®c papers in schizophrenia (>90%) do not specify the particular construct or model that they are based upon or are attempting to clarify, it is unclear how the new information they provide ®ts into the existing body of information about schizophrenia. There are currently several dierent types of models or constructs in schizophrenia that attempt to explain dierent aspects of the disease: etiological, pathophysiologic, pathologic, neurochemical, neuropsychologic, treatment response, etc. While a signi®cant number of models currently utilized in schizophrenia are modular, some attempt to explain several dierent facets of the illness. Having unifying constructs, however, is not an unmixed blessing. One can ``force'' data arti®cially to ®t the construct. Dierent models might endeavor to explain dierent pieces of data and ignore other relevant data, leaving one with the task of integrating multiple constructs (instead of attempting to integrate Journal of Psychiatric Research 33 (1999) 467±471