The lack of theories and methodological weakness have been pointed out as two distinct but related problems in empirical management information systems (MIS) research. Reinforcing the existing belief that too much attention has been devoted to ''what'' as opposed to ''why'' or ''when'' relationships exist, this paper focuses on a subset of model building and methodology issues involving the systematic discovery and representation of causal relationships. Our analysis of the existing empirical MIS literature reveals the need to build richer causal models, to increase the flexibility of model representation, to integrate the isolated worlds of pure latent and pure manifested variables, and to provide a fighter linkage between the exploratory and confirmatory research phases. Based on philosophy of science and advances in the fields of experimental economics and sociology, we propose a foundation for developing richer models by explicitly considering the exogeneity and endogeneity of constructs and a manipulative account of causality, and by recognizing the role of incentives, agent, and organizational characteristics in MIS models. Since richer models require more flexible tools and techniques, the paper describes the representational shortcomings and statistical pitfalls of factor-analytic methods commonly deployed in empirical research. We suggest that weak exploratory phase tools and approaches may allow violations of causal assumptions to pass undetected to the confirmatory phase. Since confirmatory tools like LISREL also make factor-analytic assumptions, these violations are not likely to be detected at the confirmatory phase either. We propose using TETRAD, a non-parametric tool, at the exploratory phase for its ability to accommodate a wide variety of causal models. The findings are summarized within an integrated framework, which enhances the likelihood of discovering relationships through richer theoretical support and powerful exploratory analysis.