In search of durable positive psychology interventions: Predictors and consequences of long-term positive behavior change

Abstract

A number of positive psychology interventions have successfully helped people learn skills for improving mood and building personal resources (e.g., psychological resilience and social support). However, little is known about whether intervention activities remain effective in the long term, or whether new resources are maintained after the intervention ends. We address these issues in a 15-month follow-up survey of participants from a loving-kindness meditation intervention. Many participants continued to practice meditation, and they reported more positive emotions (PEs) than those who had stopped meditating or had never meditated. All participants maintained gains in resources made during the initial intervention, whether or not they continued meditating. Continuing meditators did not differ on resources at baseline, but they did show more PE and a more rapid PE response to the intervention. Overall, our results suggest that positive psychology interventions are not just efficacious but of significant value in participants’ real lives.

Keywords:

• positive emotions
• interventions
• broaden-and-build
• meditation
• loving-kindness
• adherence
• long-term effects
• positive emotion reactivity

Acknowledgments

This study was supported by the National Institute of Mental Health grant MH59615. The authors thank their colleagues John Kraemer and John Eargle for statistical advice and Elizabeth Bartmess and Kathryn Irish for comments on the manuscript.

Notes

1. Participants who did not take part in the meditation workshop showed results that are very similar to those who did participate, and did not continue to meditate. These two groups are combined under ‘non-continuers.’ One participant, who did not participate but who did have a meditation practice at follow-up, is included under ‘continuers.’ Excluding this participant has negligible effects on the results presented in this article.

2. In order to present a model that can be expressed more intuitively, we reduced the two predictors used in the previous paragraph to one: a raw difference score between week 5 and baseline. This does not change the results (OR = 1.54–21.24, p = 0.01), though it reduces the effect size (R ²= 0.27).