Mixture modeling: a person-centered approach to studying communication and learning

Instructional communication scholars have traditionally adopted a process-product paradigm to estimate how teacher communication behaviors associate with student learning outcomes (Cortez et al., 2006). This traditional paradigm has generated much foundational research on effective teaching. At the same time, this approach might be appropriately described as narrow because it deemphasizes the fact that students are unique learners with their own roles, responsibilities, motivations, and abilities (and so on) that they bring into their learning environments. Substantively speaking, this process-product approach is limited because it overemphasizes the importance of how effective teaching, both principally and generally, fosters the same learning outcomes for all students in the same way (effective teaching is assumed to result in learning for all students despite their uniqueness in who they are). Statistically speaking, process-product scholarship typically examines communication and student learning relationships using the general linear model (e.g., correlation, t-test, analysis of variance, ordinary least-squares regression). This paradigm takes a variablecentered approach when scholars associate communication variables with learning variables. Taking a variable-centered approach has been foundational to the discipline, but it assumes that students from a sample belong to a single population. Assuming that students come from a homogeneous population yields a single parameter estimate for a communication and/or learning association; that is, one statistical estimate will suffice for all students in a study. For instance, if an estimated correlation is r = .30, it is implied that this is the correlation for all students in the population. Similarly, in confirmatory factor analysis, if a factor loading is λ = .88, this is the estimated factor loading for everyone. A variable-centered approach places the emphasis on variables rather than people by providing single estimates that describe relationships between variables under study. Alternatively, the analytical focus can be shifted from variables to people through the application of finite mixture modeling which offers a person-centered approach to studying communication and learning. Unlike a variable-centered approach, a person-centered approach allows for population heterogeneity to the extent that the sample embodies an unknown mixture of homogeneous subpopulations. In the truest application of mixture modeling (a direct application), the goal is to uncover latent