Are the sources of interest the same for everyone? Using multilevel mixture models to explore individual differences in appraisal structures

Abstract

How does personality influence the relationship between appraisals and emotions? Recent research suggests individual differences in appraisal structures: people may differ in an emotion's appraisal pattern. We explored individual differences in interest's appraisal structure, assessed as the within-person covariance of appraisals with interest. People viewed images of abstract visual art and provided ratings of interest and of interest's appraisals (novelty–complexity and coping potential) for each picture. A multilevel mixture model found two between-person classes that reflected distinct within-person appraisal styles. For people in the larger class (68%), the novelty–complexity appraisal had a stronger effect on interest; for people in the smaller class (32%), the coping potential appraisal had a stronger effect. People in the larger class were significantly higher in appetitive traits related to novelty seeking (e.g., sensation seeking, openness to experience, and trait curiosity), suggesting that the appraisal classes have substantive meaning. We conclude by discussing the value of within-person mixture models for the study of personality and appraisal.

Keywords:

• Interest
• Appraisal
• Personality
• Multilevel modelling
• Latent-class analysis
• Finite mixture models

Acknowledgements

This research was presented at the University of Geneva, the University of Erlangen, and the 2007 meeting of the Society for Personality and Social Psychology.

We thank Jim Villano and Anna Waters for assistance with data collection.

Notes

When exploratory methods are used, one is naturally concerned about the stability of the final model, particularly when many aspects of a model could be varied. The model's central effects—the coefficients, number of classes, and class proportions—don't appreciably differ due to changing the centring, to freely estimating the random effects’ within-class variances, or to allowing interest's Level 2 and residual Level 1 variances to vary across classes. Similarly, running an alternate specification of the model (fixed within-person paths that vary across classes, instead of between-person random effects that vary across classes) yields essentially identical findings, as one would expect.