In psychology, measurement of mental attributes relies on statistical models: so-called ‘measurement models’. Existing measurement models are developed within a particular framework in which the attributes are assumed to be latent common causes of observed behaviors. However, more recently, ‘network theories’ of psychological attributes have gained traction in psychology, in which the assumption that the attribute acts as a common cause of the observed behaviors is explicitly rejected. Instead, behaviors, cognitions and feelings interact directly with each other, forming a complex system that can be represented as a network.

Although the development of psychological network theory has led to many new analytic techniques, it has not yet resulted in a psychometric framework that can guide measurement. For example, claims such as that some people are more severely depressed than others, or that a particular school intervention has increased a person’s ability, require that levels of these psychological attributes can be compared across people or time. The traditional framework of latent variable models accommodates such comparisons by estimating people’s positions on these attributes, but these models make assumptions that are inconsistent with a network theory of these constructs. At the same time, it is unclear how to best make such comparisons from network models of these constructs. I will discuss some possible steps in developing a measurement model for psychological attributes that are studied within a network framework, and discuss challenges involved. The shift from latent variable theory to network theory in psychology also provides an interesting context to consider Chang’s (2004) notion of ‘epistemic iteration’, which describes how theory about a concept and measurement practices for that concept iteratively revise each other.

Programme