Médiateurs et modérateurs : implications théoriques et méthodologiques dans le domaine du stress et de la psychologie de la santé

English

Mediators and moderators : theoretical and methodological implications in stress and health psychology research The concepts of mediators and moderators are often confused in psychological research. Most empirical studies use these two terms interchangeably and analyse their effects using the same statistical approaches. Nevertheless, it is very important to distinguish between their properties as it permits the conceptualization of complex theoretical models and different levels of analysis. In the domain of stress research, the distinction between these two types of variable is also essential because it underscores the historical evolution of different models. The earliest behavioural models of stress (stimulus-response) were simple interactional models, where by negative consequences of stress could be explained jointly by environmental and personal variables, and above all by their incompatibility. In this context, moderators play an important role. Numerous personal variables (personality traits) and environmental variables (social support) serve as of buffers between stressors (work, life events) and outcomes (well-being, illness). Although these interactional models are important for emphasizing individual vulnerability to stress, they do not explain the underlying psychological processes at work. For this reason, the evolution of models requires the identification of psychological and/or biological mediators that explain how predictive variables (environment and personality) have an impact on individual health. The transactional model of stress is an example of such neo-behavioural models. The identification of distinct models and variables with specific properties also requires the selection of appropriate statistical analyses. Testing the effects of moderators does not use the same kind of analysis as for testing the effects of mediators. ANOVA and hierarchical regression analyses are the most appropriate methods for testing moderator effects, while path analyses with Lisrel allow for the examination of mediator effects of a given variable relative to two other variables. Finally, the clarification of these different issues has value beyond the domain of stress theory.

English

RÉSUMÉ Afin d’atténuer les confusions relatives à l’utilisation par les chercheurs en psychologie des notions de médiateur et modérateur, nous nous proposons d’exposer leurs propriétés respectives. Nous montrons comment ces variables s’intègrent dans les modèles de la psychologie de la santé et plus précisément dans l’explication du stress. Nous distinguons dans ce sens les modèles déterministes, des modèles interactionnistes simples et complexes. Par des exemples issus de recherches empiriques dans le champ du stress professionnel notamment, nous exposons le rôle de chacune de ces variables psychologiques sur la santé des individus. Enfin, nous expliquons les méthodes d’analyse quantitative susceptibles de tester leurs effets respectifs.