Trait-State Models

Introduction

Many psychological attributes have state as well as trait versions. This is most obvious for mood states. A classical example is Spielberger's state and trait anxiety. State anxiety is assessed via items in which the actual mood states have to be rated (‘right now I feel …’) whereas trait anxiety is assessed by items asking for the general mood state (‘in general I feel …’). However, it is interesting to note that the concept of states was introduced to personality research only during the 1960s and 1970s. By contrast, the concept of traits has guided personality research from its very beginning, even though the distinction between states and traits appears to be as old as thinking about human mind and behaviour.

Traditionally, some psychological properties are classified as states, such as well-being and feeling anxious, while others are said to be traits such as extraversion, neuroticism, etc. However, the more general approach is certainly to assume that each psychological property is statelike to some degree and therefore also traitlike to some degree. Observed mood states are not only due to (bio-psycho-social) situations but also depend to some degree on permanent characteristics of the person. Similarly, observed trait measures not only depend on permanent characteristics of the person, but are also due to situations to some [Page 1042]degree. This is the substantive background for latent state-trait (LST) theory and its associated latent state-trait models to be described in the following sections.

Latent State-Trait Theory

In the 1990s, LST theory has been introduced defining states (of whatever variable) as a property of a person-in-a-situation and traits as a property of a person. Furthermore, specific structural equation models for longitudinal data have been developed that can be used to disentangle state and trait components of whatever variable.

Originally, LST theory was developed as a generalization of classical test theory (CTT), designed to take into account that psychological assessment does not take place in a situational vacuum. Hence, from a substantive perspective, LST theory may also be viewed as a methodological development coping with arguments raised in the person-situation debate. Whereas in CTT, aside from measurement error, there is only one single factor (persons) determining the variance of an observable variable, LST theory explicitly assumes two factors instead: persons and situations. Even though the standard models of LST theory are not designed to disentangle situation effects and the effects of interaction between persons and situations, interaction effects are an integral part of the theory.

The core of LST theory consists of two decompositions: (a) the decomposition of any observable variable into latent state and measurement error variables, and (b) the decomposition of any latent state into latent trait and latent state-residual variables, the latter representing situational and/or interaction effects. Latent state and latent trait variables are defined as special conditional expectations. A score of a latent state variable is defined to be the conditional expectation of an observable variable given a person-in-a-situation, whereas a score of a latent trait variable is the conditional expectation of this observable variable given a person.

...