A pathway is a set of mechanistic or causal relationships between components such that the result of altering a component is the production of variation in other components of the pathway. A consequence of pathway structure is characteristic patterns of variation and co-variation among phenotypes in a population. For example, the pattern of gene expression for a set of genes can be explained in terms of underlying pathway structure. In theory, the structure of gene expression or related variation patterns can be used to infer the structure of the underlying pathway and a large number of approaches have been suggested for this purpose. We have developed a statistically rigorous inference approach that discovers pathway structure as a by-product of the model based inference. The approach makes use of a broad class of structural linear statistical models and a computationally efficient Bayesian inference approach. We have found that our method can identify complex pathway structures including feedback loops and reciprocal regulation motifs with relatively small sample sizes. We are currently applying our method to discover pathway structure for a number of model organisms including yeast and humans.