From Ordinal Representations to Representational Profiles: A Primer for Describing and Modelling Social Representations of History

Abstract

Social representations theory is rich in explanatory power and broad in scope. This very complexity often leads to cases where predictions derived from the theory are difficult to operationalize and test. We argue that in many cases this is because social representations theory requires statistical models and analytic techniques that are uncommon in other social science traditions. In this chapter we outline a series of analytic methods and describe examples for their use in both improving description and testing predictions relating to social representations of history. We offer this overview as a methods primer for four complementary analytic methods for the study of social representations. These four methods are: (1) ordinal models assessing naming prevalence, (2) dimensional models assessing relational representations, (3) factorial representations focusing on unitary concepts, and the most recent addition to our toolbox: (4) representational profiles: latent class analysis allowing the assessment of representational profiles. We focus much of our primer on this method, and argue that latent class models, and factor mixture modelling in general has immense potential for the empirical assessment of social representations. This is because such models allow the assessment of categorical models of different types of representations, where those representations can represent different emerging factor structures derived inductively from the data. We finish by formally outlining a series of six premises for the theory and measurement of representational profiles using this novel approach. Syntax documenting a worked example of the Latent Class Model tested in one of our earlier papers using Mplus is appendicized, and additional supplementary material posted online.