The role of positive emotion in harmful health behavior: Implications for theory and public health campaigns

The study investigates whether specific positive emotions, particularly gratitude, can attenuate appetitive risk behaviors (ARBs) such as cigarette smoking. Although prior meta-analyses suggest that inducing broadly positive affect does not reliably reduce ARBs, the authors argue this conclusion may be premature because most research has focused on undifferentiated positive mood rather than discrete positive emotions. Grounded in the Appraisal Tendency Framework and social-functional theories of emotion, the paper hypothesizes that gratitude—characterized by appraisals of gain and other-focus—promotes prosocial goals and long-term cooperation, potentially enabling individuals to forgo immediate gratification and thereby reduce craving and smoking. The research aims to test correlational and causal links between gratitude and smoking across multiple methods and populations, and to evaluate whether existing public health campaigns leverage gratitude effectively.

Prior work in emotion and decision-making demonstrates robust effects of discrete emotions on judgment and choice across domains including risk-taking and intertemporal choice. Despite evidence linking positive affect to better health, meta-analyses have found no reliable causal effect of positive (vs. neutral) affect on reducing ARBs or related cognitions, and in some cases amusement increased ARB indicators. The Appraisal Tendency Framework posits that each emotion carries distinct appraisal patterns and implicit goals that shape downstream decisions beyond the eliciting context. Sadness, for example, involves loss appraisals and self-focus, increasing reward-seeking and addictive use. Gratitude, in contrast, reflects appraisals of gain and other-focus, binding social relationships, fostering prosociality, and reducing impatience—mechanisms central to ARBs. Evidence shows trait and induced gratitude reduce economic impatience, suggesting potential protective effects on behaviors driven by craving. Positive emotions are not always well differentiated, but theories of discrete positive emotions (e.g., gratitude, awe) motivate testing emotion-specific effects on ARBs rather than broad positive mood.

The research comprises five studies using secondary and primary data, including nationally representative US samples, an international sample across 87 countries, and preregistered experiments with adult smokers. Open science practices were followed, with preregistrations, data, and code posted to OSF, and power targets set for small to medium effects. Study 1a: Secondary analysis of the National Study of Youth and Religion (NSYR) wave 3 (N = 2,485; age 17–24; 2007–2008) and wave 4 (N = 2,003; age 20–32; 2012). Trait gratitude was measured by three items (McCullough et al. 2002; α = 0.50 wave 3; α = 0.62 wave 4). Sadness measured by single items. Smoking status/frequency operationalized differently across waves; logistic and linear regressions assessed associations, controlling for age, gender, and sadness. Study 1b: Secondary analyses of MIDUS datasets: MIDUS 2 (N = 4,963, 2004–2006), MIDUS 2 Biomarker (N = 1,255, 2004–2009), MIDUS 3 (N = 3,294, 2013–2014), MIDUS Refresher (N = 3,577, 2011–2014), MIDUS Refresher Biomarker (N = 863, 2012–2016). Gratitude measured via single item or two items (α = 0.71 in biomarker waves); compassion measured by four items (α ≈ 0.44–0.50). Smoking status binary (current regular smoker). Logistic regressions tested gratitude vs. compassion as predictors, with robustness checks controlling demographics (age, gender, SES), general positive affect (10 or 14-item scales; α ≈ 0.93–0.94), and general negative affect (14 items or composite depressive/anxious symptoms; α ≈ 0.81–0.93). Study 2: International cross-sectional sample from the Psychological Science Accelerator (N = 21,644, 87 countries/regions; baseline emotions measured). Positive emotions (gratitude, love, hope, inspiration, serenity) measured by single items (modified DES). Negative affect was a five-item composite (α = 0.83). Individual-level outcome: intention in the next week to use too much tobacco (e.g., smoke/vape) or other recreational drugs (7-point scale; highly skewed; analyzed as continuous and dichotomous). Multilevel models with random country intercepts controlled for age, gender, education, negative affect, other positive emotions. Country-level analyses aggregated emotions for countries with ≥30 participants, and linked WHO current tobacco use prevalence (N = 51), Euromonitor retail cigarette volume per capita (N = 52), and World Bank GDP per capita. Linear regressions assessed relations; multicollinearity checked via VIF (<10). Study 3: Preregistered experiment with adult smokers recruited via MTurk CloudResearch high-quality pool (initial N = 600; final N = 546 after preregistered exclusions; age 18–71). Baseline emotions (20 items) and baseline craving (QSU-brief; three items; α = 0.96) were recorded. Participants were randomly assigned to gratitude, compassion, sadness, or neutral conditions via a two-part induction (validated video + writing task). Post-induction craving was remeasured. Pre–post change in craving was the dependent variable. Pairwise t-tests contrasted conditions; exploratory mediation tested whether change in self-reported emotions mediated effects. A preregistered replication (gratitude vs neutral; N = 194 after exclusion) repeated the procedure; combined-sample analyses provided more precise estimates. Study 4: Preregistered experiment with adult smokers intending at least minimally to quit (screened via Contemplation Ladder ≥1). Recruited via MTurk CloudResearch (initial N = 200; final N = 169 after exclusions). Random assignment to gratitude vs neutral inductions (same as Study 3). Outcomes: helping behavior (words of advice to two smokers), incentivized self-reported willingness to enroll in a free online cessation program (BecomeAnEx) quantified via lottery ticket titration (0–8), and actual enrollment verified by screenshot within one week. Logistic and t-test analyses assessed effects; a preregistered replication (N = 450 after exclusion) repeated the design; combined-sample analyses estimated overall causal effect. Exploratory mediation tested change in self-reported emotions mediating enrollment. Study 5: Preregistered assessment of emotions evoked by the CDC Tips from Former Smokers campaign videos. Adult smokers recruited via MTurk (final N = 194 after exclusions). Each participant rated five randomly selected videos from 81 total. After each video, participants rated intensity (0–8) of 44 emotions (compiled from three taxonomies) and frequency (percentage >0). Analyses summarized intensity and frequency, and contrasted gratitude vs sadness using t-tests and chi-square.

Study 1a: Trait gratitude significantly associated with lower likelihood of smoking (logistic regressions; standardized b = −0.17, P < 0.001 in wave 3; b = −0.34, P < 0.001 in wave 4; combined b = −0.23, odds ratio = 0.80, P < 0.001). Gratitude inversely correlated with frequency of smoking (linear regressions; b = −0.08, P < 0.001 in wave 3; b = −0.15, P < 0.001 in wave 4; combined b = −0.11, P < 0.001). Effects held after controlling age, gender, sadness. Study 1b: Across five MIDUS datasets, trait gratitude predicted lower likelihood of smoking (logistic regressions; b range −0.40 to −0.19, Ps < 0.001; combined b = −0.26, odds ratio = 0.77, P < 0.001). General positive affect sometimes predicted lower smoking, consistent with partial overlap among positive emotions. Trait compassion did not predict smoking status in any wave (b range −0.12 to 0.04, Ps > 0.121; combined b = −0.01, odds ratio = 0.99, P = 0.743). Results robust controlling demographics, general positive and negative affect. Study 2 (individual level): Current feelings of gratitude associated with significantly lower intentions to use tobacco/other recreational drugs across control specifications (multilevel models; b range −0.08 to −0.04, Ps < 0.001; N = 87 countries). Among five positive emotions (gratitude, love, hope, inspiration, serenity), only gratitude consistently showed significant relations and often the largest effect size. Study 2 (country level): Higher average gratitude predicted lower prevalence of tobacco use (linear regressions; b range −0.57 to −0.31, P ≤ 0.010; N = 51) and lower retail volume of cigarette sales per capita (b range −0.76 to −0.53, Ps < 0.001; N = 52), robust to controls for other positive emotions, negative affect, and log GDP per capita. Gratitude was the only positive emotion consistently significant across specifications. Study 3: Gratitude reduced craving to smoke relative to neutral (b = −0.31, SE = 0.12, t(542) = −2.48, P = 0.013, d = 0.37), compassion (b = −0.33, SE = 0.12, t(542) = 2.72, P = 0.007, d = 0.33), and sadness (b = −0.67, SE = 0.12, t(542) = −5.46, P < 0.001, d = 0.61). Compassion did not differ from neutral (b = 0.02, SE = 0.12, P = 0.844). Sadness increased craving vs neutral (b = 0.36, SE = 0.12, t(542) = 2.94, P = 0.003, d = 0.40). Mediation: change in self-reported gratitude mediated gratitude vs neutral effect on craving (b = −0.15, 95% CI [−0.29, −0.006], P = 0.041; proportion 0.43); happiness and compassion changes also showed significant indirect effects. Replication (gratitude vs neutral): b = −0.40, SE = 0.15, t(192) = −2.79, P = 0.006, d = −0.40; indirect effects via gratitude (b = −0.20, P = 0.022) and happiness (b = −0.08, P = 0.048). Combined samples: main effect b = −0.37, SE = 0.09, t(457) = −3.98, P < 0.001, d = −0.39; mediation via gratitude (b = −0.17, P = 0.002; proportion 0.47) and happiness (b = −0.07, P = 0.004; proportion 0.20). Study 4: Gratitude increased actual enrollment in a smoking cessation program vs neutral in the initial study (M_gratitude = 40%, M_neutral = 24%; odds ratio = 2.16, z = 2.28, P = 0.023, d = 0.43). No significant effects on number of words of advice (P = 0.483) or self-reported intent (lottery titration; P = 0.795). Replication: directional but nonsignificant increase (28% vs 22%; odds ratio = 1.36, z = 1.41, P = 0.158). Combined samples: significant increase in enrollment (31% vs 23%; odds ratio = 1.55, z = 2.39, P = 0.017, d = 0.24), implying a 35% relative increase. Mediation: change in self-reported gratitude significantly mediated enrollment (b = 0.11, 95% CI [0.01, 0.21], P = 0.025; proportion 0.43); other emotions did not. Study 5: Among 44 emotions evoked by CDC Tips videos, gratitude ranked 21st in intensity and 20th in frequency; sympathy, compassion, and sadness were most intense/frequent. Gratitude was less intense than sadness (t(193) = −18.39, P < 0.001, d = −1.32) and less frequent (χ² = 39.06, P < 0.001; odds ratio = 0.25).

The findings address the central question of whether discrete positive emotions can reduce appetitive risk behaviors by demonstrating that gratitude specifically, and not broadly positive affect nor compassion, is linked to lower smoking and reduced craving, and can causally increase engagement in cessation behavior. Triangulating across correlational evidence (US national and global samples), experiments on craving (with emotion specificity and replications), and real-world behavior (program enrollment), the results support the appraisal-based mechanism wherein gratitude’s gain appraisals and other-focus foster prosocial goals, patience, and preferences for long-term health over short-term rewards. Mediation by self-reported gratitude across experimental studies strengthens the conclusion that the emotion itself, not just incidental features of inductions, drives outcomes. Practically, current antismoking campaigns predominantly evoke sadness and compassion, missing opportunities to harness gratitude’s protective effects; integrating gratitude-inducing content could reduce craving and enhance cessation engagement. The work refines theories of emotion and decision-making by incorporating discrete positive emotions into models of addictive behavior, and challenges meta-analytic conclusions that positive emotions lack causal effects on ARBs.

Our research unveils the pivotal role of positive emotions, particularly gratitude, in shaping addictive risk behaviors. This significant departure from the conventional focus on negative emotions enriches our understanding of the emotional dynamics involved in addiction. Our research offers a more nuanced and empirically grounded foundation for designing effective intervention strategies at the population level.

The studies focus primarily on smoking, though theory extends to other appetitive risk behaviors; further research should test generalizability. Gratitude inductions arose outside the smoking context; inducing gratitude within antismoking campaigns may require tailored approaches and could function differently over time given affective adaptation. Study 4’s enrollment outcome, while promising, does not establish long-term cessation; longitudinal follow-up is needed. Future work should disentangle the necessary vs sufficient roles of the key appraisal dimensions (sense of gain and other-focus). Gratitude is not the only candidate positive emotion; others such as awe, which reduces impatience and benefits health, warrant investigation.