Emotion and appraisal: A study using ecological momentary assessment

Abstract

This study employed Ecological Momentary Assessment to test predictions from appraisal theories of emotion about the relationships between emotions and appraisals, using a sample of police officers from Singapore. Strong support was obtained for the predictions, thus demonstrating ecological validity of appraisal theories while circumventing shortcomings of previously used methods in appraisal studies. The results also indicate that the emotions were accounted for by specific configurations of appraisals over and above those accounted for by individual constituent appraisals.

Acknowledgements

This research was funded by grant no. R-107–000–007–012 from the National University of Singapore with supplemental funds from the Singapore Ministry of Home Affairs, Singapore Police Force.

We would like to express our appreciation to Phoebe C. Ellsworth for her comments on earlier drafts of this article, and Richard Gonzales, Richard Lewis, and Susan A. Murphy for their statistical advice.

Notes

¹Note that there are other ways of manipulating appraisals, such as by anagrams (Pecchinenda & Smith, 1996) and video-games (Kappas & Pecchinenda, 1999).

²The data for negative emotions were also analysed with two other methods. In the first method, 0 was assigned to all observations where no emotion was indicated (i.e., rated as 1) and 1 was assigned to all observations where the emotion was indicated (i.e., rated as 2 or above). We applied HLM Bernoulli regression, a multi-level modelling technique suitable for binary outcome variables, to these data (Bryk & Raudenbush, 1992). The results were highly similar to those produced by Poisson regressions. The second method involved randomly selecting an observation out of a pool where an emotion was indicated from participants who indicated at least one emotional episode and randomly selecting one observation from participants who did not indicate an emotional episode. When selecting an episode from participants with at least one emotional episode, priority was given to episodes with higher emotion ratings. In other words, for a hypothetical participant with seven observations where Anger was rated as 2 and four observations where Anger was rated as 3, one observation was randomly selected only from the pool of the four observations where Anger was rated as 3. This made it more likely that episodes of higher emotional intensities were selected since these observations are rare. Thus, in this procedure, each participant contributed one data point, which together formed a more normal distribution. Applying linear regression to this between-participant data produced results similar to those produced by HLM Poisson regression (although there are fewer significant results due a loss in power). Nevertheless, the fact that these two different techniques produced findings consistent with those from HLM Poisson regression suggests that the results from Poisson regression are reliable. Finally, all three analyses produced comparable results for testing the configuration hypothesis.