Gratitude Practice and Mental Health: What 145 Studies Say About Well-Being

Why did gratitude attract scientific attention?

Positive psychology has studied since the 1990s what makes people psychologically resilient, not just what breaks them down. Gratitude became one of the most studied constructs — partly because it is easy to measure and easy to test: asking someone to write down a few things they are grateful for is simpler than delivering psychotherapy.

Prior to 2025, accumulated meta-analyses presented an ambiguous picture: effect sizes ranged from non-significant to moderate depending on the outcome and type of intervention. Critics pointed to possible publication bias, heterogeneity of protocols, and cultural specificity.

In July 2025 the most large-scale meta-analysis to date was published (Choi et al., Proceedings of the National Academy of Sciences), which attempted to address these objections directly.

What did the PNAS 2025 meta-analysis show?

Choi, Cha, McCullough and co-authors analyzed 145 papers, 163 samples, 727 effect sizes, and 24,804 participants from 28 countries. This is the first cross-cultural meta-analysis of gratitude interventions with pre-registered moderator variables.

Key results by outcome:

• Overall well-being: Hedges' g = 0.19 [95% CI: 0.15–0.22], p < 0.001
• Positive affect: g = 0.27 [0.22–0.33]
• Composite well-being score: g = 0.28 [0.14–0.42]
• Life satisfaction: g = 0.18 [0.12–0.24]
• Happiness: g = 0.17 [0.12–0.22]
• Depressive symptoms: g = 0.15 [0.09–0.22]
• Negative affect: g = 0.12 [0.07–0.17]

In the RCT-only subgroup — g = 0.20 [0.17–0.24]. Sensitivity analysis with publication-bias correction did not change the conclusion.

How do protocols differ — and what works best?

The meta-analysis distinguished several types of interventions by effect size:

• "Counting blessings" (gratitude journaling): g = 0.17 [0.12–0.22]
• "Recalling past gratitude": g = 0.25 [0.16–0.33]
• "Counterfactual thinking" (imagining that something good had not happened): g = 0.36 [0.19–0.53]
• Mixed and combined approaches: g = 0.26 [0.19–0.34]

An important caveat: effect size did not depend significantly on number of sessions or duration of the intervention. This means a longer practice does not necessarily yield additional effect gains — at least within the time horizons studied.

What do clinical scales say: anxiety and depression?

A Brazilian systematic review and meta-analysis (Diniz et al., Einstein / Sociedade Beneficente Israelita Brasileira, 2023; 64 RCTs in the review, 33 in the quantitative synthesis) measured effects using standard clinical scales:

• Anxiety (GAD-7): −1.63 points (95% CI: −2.37 to −0.89, p < 0.0001)
• Depression (PHQ-9): −1.86 points (95% CI: −2.89 to −0.83, p = 0.0004)
• Life satisfaction (SWLS): +0.48 points (95% CI: 0.21–0.75, p = 0.005)
• Psychological well-being (MHC-SF): +0.29 points (95% CI: 0.17–0.41, p < 0.00001)

A reduction of 1.63 points on GAD-7 and 1.86 points on PHQ-9 is a sub-threshold improvement at high baseline scores, but represents a meaningful shift in healthy populations or with mild symptomatology.

Are there physiological mechanisms?

A systematic review of 19 RCTs (Wang and Song, Frontiers in Psychology, 2023) assembled physiological data. In patients with chronic heart failure, an 8-week gratitude journal was associated with increased parasympathetic activity (heart rate variability). Among pregnant women, a three-week practice compared with standard care reduced cortisol levels during waking and sleeping periods. Several trials recorded reductions in inflammatory markers — CRP, TNF-alpha, IL-6, sTNFr1. In young women, diastolic blood pressure decreased.

An important caveat: the review did not calculate a pooled effect size for cardiovascular parameters — the quality and design of included trials were too heterogeneous. Physiological data are interpreted as preliminary.

Sense of purpose and mortality: a related data angle

Gratitude is conceptually close to a sense of meaning and purpose in life. A meta-analysis by Sutin and colleagues (Psychological Medicine, 2026) combined 25 samples, including 8 cohorts with individual-level data — 488,765 participants followed for up to 32 years, 48,928 deaths. A higher sense of purpose in life was associated with HR = 0.76 (95% CI: 0.70–0.83) for all-cause mortality — approximately 24% lower risk per standard deviation on the scale. After adjustment for depression the association persisted (HR = 0.91). The association replicated across cohorts in the US, Europe, and Asia.

These are observational data; the direction of causation has not been established. Nevertheless, the scale of the sample and the length of follow-up make this association one of the most robust in the literature on psychological well-being and health.

What are the cultural limitations?

One of the key findings of PNAS 2025: significant effects were found in the US, Germany, South Korea, China, and Canada — but not in France, Japan, the Netherlands, or the UK. Hong Kong showed the largest effect (g = 0.60), Japan the smallest (g = 0.04). None of the seven pre-registered cultural predictors — individualism, GDP, religiosity, and others — explained this difference at the Bonferroni correction level.

Practical takeaway: a null effect in some cultures is not a methodological error but a real result. The mechanism behind this cultural variability is not yet understood.