Examining Conditional Relevance in Situational Action Theory Using Generalised Additive Modelling

Abstract Tests of Situational Action Theory’s (SAT) conditional relevance of controls have typically relied on linear parametric interaction terms or comparisons across categorised moral configuration groups. While these approaches have contributed to our understanding of SAT, they may also constrain how the effects of controls (e.g. self-control and deterrence) vary across the moral configurations implied by the theory. In contrast, the present study reanalyses survey data on self-reported speeding collected from 919 Australian drivers to examine the use of a generalised additive modelling approach. Here, the effects of self-control, perceived certainty of detection, and perceived severity of punishment were permitted to vary nonlinearly across the morality–moral context space. Results indicated substantial nonlinear variation in the relevance of self-control across moral configurations, whereas certainty displayed a largely linear effect and severity showed little systematic influence. However, although this more flexible approach was better able to capture the underlying morality–moral context space than traditional linear approaches, the emerging patterns did not align with theoretical expectations regarding the conditional relevance of controls, nor did they more closely approximate those observed in the prior analysis of the same data. These findings suggest that the challenge in examining the conditional relevance of controls likely reflects not only modelling complexities, but also broader measurement challenges, both of which require further investigation.