A considerable body of evidence supports the notion that the neurofunctional substrate of working memory is not only related to the integrity of the prefrontal cortex, but also to the concerted interplay of widespread interacting networks including the parietal cortex, subcortical regions and cerebellar areas. Modern functional brain imaging techniques such as functional magnetic resonance imaging (fMRI) have provided a detailed picture of functional neuroanatomy subserving working memory functions. Most of the earlier functional studies were directed toward the identification of brain areas subserving specific cognitive domains in terms of a functional segregation. More recently, different multivariate techniques were employed to specifically address measures of functional or effective connectivity. Structural equation modeling (SEM) or path analysis is one of the most often used methods to model interactions among covarying brain areas in an explicitly model-based approach. The present review will focus on basic methodological issues of SEM for the analysis of fMRI datasets in studies of working memory. Aside from a discussion of previous studies and their essential findings, advanced methodological issues and caveats as well as future perspectives of the method will be addressed.