On the relative effectiveness of affect regulation strategies: A meta-analysis

Abstract

To examine the effectiveness of various affect regulation strategies and categories of affect regulation strategies, a meta-analysis was conducted. Results generally indicate that reappraisal (d=0.65) and distraction (d=0.46 for all studies; d=0.95 for studies with a negative or no affect induction) are the most effective regulation/repair strategies, producing the largest hedonic shift in affect. The effectiveness of different categories of regulation/repair strategies depended on the valence of the preceding affect induction. Results also indicate that stronger affect inductions and the use of bivariate affect measures will provide a richer understanding of affect regulation. Additionally, not all specific strategies or categories of strategies have been researched and the impact of individual differences on affect regulation has received relatively little attention. Finally, results indicate that control conditions in affect regulation research may not provide a valid point for comparison, as they facilitate effective affect repair.

Keywords:

• Emotion
• Mood
• Affect regulation
• Affect repair
• Meta-analysis

Acknowledgements

Portions of this paper were presented at the 8th annual meeting of the Society for Personality and Social Psychology (January, 2007).

We would like to thank Randy Larsen, Don Saucier, and Tirza Shulman for their comments regarding this work. Special thanks also go to Eva Gilboa-Schechtman for her many helpful comments throughout the editorial process. Additionally, we would like to thank the students that aided in coding the studies included in this meta-analysis: Kate Duangdao, Jobeth Kincaid, and Megan Tripp.

Notes

¹ Larsen's taxonomy is not directly presented in this meta-analysis because the coders for this study indicated perfect agreement between this classification scheme and the superordinate classification scheme of Parkinson and Totterdell (1999). Both taxonomies utilise a behavioural vs. cognitive distinction. Although there are obvious theoretical differences between the two, in practice for these coders Larsen's (2000) change situation vs. change emotion distinction was equal to the engagement vs. avoidance distinction made by Parkinson and Totterdell (1999). Additionally, although an analysis of the taxonomy developed by Thayer, Newman, and McClain (1994) was initially conducted (and is available on request), the categories in this taxonomy contain strategies of vastly differing effectiveness (i.e., the social support, venting, and gratification category contains two effective and one ineffective strategy) and/or are ill-defined (i.e., direct tension reduction). As such, there was little informative value gained from presenting these analyses.

² It should be noted that, although some of the strategies presented by authors could be considered exemplars of another specific strategy (i.e., a cognitive reasoning task could be considered distraction, listening to music could be considered a pleasant and rewarding activity, etc.), specific strategies utilised for each case are recorded as stated by the author. This allows for the analysis of strategy effectiveness to be carried out at the lowest and most specified level.

³ The use of this method for calculation of effect sizes creates some difficulties. First, the extant data clearly demonstrate that positive and negative affect can exist largely independent of one another (see Hemenover & Schimmack, 2007; Larsen, McGraw, & Cacioppo, 2001; Larsen, McGraw, Mellers, & Cacioppo, 2004) and thus, the use of a univariate measure of affect change is contrary to the actual structure of emotion. However, 16 cases (effect sizes) utilise univariate measures of affect and, to represent the entire literature on the subject, must be included. Second, some repair strategies are better suited for NA change while others are better suited for PA change. Although this method of effect size calculation can be problematic for the 16 cases (effect sizes) reporting both PA and NA change values, it is the only way to include all relevant data (and make meaningful comparisons between strategies) in this meta-analysis.

⁴ The use of multiple effect sizes from a single study violates the assumption of independence in meta-analytical techniques. However, because of the moderation of effect sizes by the affect regulation strategy employed, in this meta-analysis, it was necessary to use multiple effect sizes from a single study.

⁵ There are obvious assumption violations due to the differences in cell sizes in many of the comparative analyses. As such, the implications of these analyses should be viewed with some caution. Additionally, there is a great deal of overlap between category analyses (i.e., the inhibition specific strategy and act happy category contain the same two effect sizes). This again requires caution when interpreting these analyses.

⁶ As it was hypothesised that significant differences would exist between different specific strategies and different categories of strategies, a priori significance values are utilised in all difference tests.

⁷ Sources included in the meta-analysis are marked with an asterisk.