Legal doctrinal scholarship engages with the problems of legal practice: it systematizes, comments on, evaluates and debates what goes on in law. These activities do not occur in a vacuum: they are embedded in scholarly traditions and theories. This paper discusses the role of the theoretical frameworks used in legal research and has two related aims. First, it aims to provide some practical conceptualizations and guidelines regarding theoretical and normative frameworks that are useful to understand and conduct legal research. Second, it aims to investigate the relationships between different kinds of normative frameworks and their relationship to empirical work. In the second part, an argument is made for a pragmatist understanding of the interplay between normative theorizing and empirical study. How do these work together in judgments about the state of the law?

Quantitative empirical research into legal decisions must be conducted using statistical tools that are appropriate for the data involved. Court decisions are one example of a domain where the data is intrinsically hierarchical (i.e., multilevel), since decisions are made on individual cases by decision-makers in courts located across geographical (or jurisdictional) areas. Past research into court decisions has often either neglected higher level variables or incorrectly used single-level statistical models to analyze multilevel data. The lack of a clear understanding about when and why multilevel statistical models are required may have contributed to this situation. In this paper, we identify the problems of estimating single-level models on hierarchically structured data, and consider the advantages of conducting multilevel analyses under these circumstances. We use the example of criminal sentencing research to illustrate the arguments for the use of multilevel models and against a single-level approach. We also highlight some issues to be addressed in future sentencing studies.