The effect of induced happiness versus sadness on three types of self-control tasks

ABSTRACT

How can people succeed in situations that require self-control? One mechanism might be through positive emotions such as happiness, although previous studies are limited and have yielded inconsistent results. Three studies (total N = 424) therefore examined the effect of induced happiness versus sadness on three different self-control tasks that required delaying gratification (Study 1), persistence on an aversive task (Study 2), and breaking a habitual response (Study 3). Results showed that happiness improved performance on these tasks compared to sadness, while controlling for baseline emotions, practice performance, and trait self-control. This research advances our causal knowledge of the facilitating effect of happiness versus sadness on tasks that measure different self-control components, and has important practical implications that we discuss.

KEYWORDS:

• Emotions
• happiness
• sadness
• self-control
• self-control performance

Self-control refers to the ability to “override impulses to act as well as the ability to make oneself initiate or persist in [a] boring, difficult, or disliked activity” (Carver, 2010, p. 766). To date, there is little doubt that the capacity for self-control is associated with a wide range of positive outcomes, such as good physical and mental health, harmonious interpersonal relations, and improved work and academic performance (Crescioni et al., 2011; Finkel & Campbell, 2001; Tangney et al., 2004), whereas the lack of self-control is associated with numerous negative outcomes such as overeating, addictive and criminal behaviors, as well as financial and health problems (Moffitt et al., 2011; Vohs & Heatherton, 2000). In light of this, advancing our understanding of how people succeed at exerting self-control has the potential to help people achieve important positive life outcomes (Gillebaart & de Ridder, 2015).

Building on insights from the broaden-and-build theory (Fredrickson, 2001), one potential mechanism for improving self-control might be through increased positive emotions. Indeed, research suggests that emotions can be used to achieve self-control success, with positive emotions generally enhancing self-control relative to negative emotions (e.g., Aspinwall, 1998; DeSteno, 2018). However, despite these promising findings, other research indicates that positive emotions may not always improve self-control (e.g., Crossfield & Damian, 2021; Winterich & Haws, 2011). It has therefore been proposed that more research is needed to better understand the causal effects of emotions on self-control, especially since previous discrepant results are likely due to a number of situational, individual, and methodological factors such as variations in the emotion induction technique, the specific emotions examined, and the behavior or task used to measure self-control (e.g., Schmeichel & Inzlicht, 2013; Tornquist & Miles, 2019, 2022). Thus, the current research investigates how positive vs. negative emotions influence performance on various tasks that require self-control, holding the emotion induction method and the type of emotions examined constant. Specifically, Study 1 investigates how happiness vs. sadness influence performance on a task that requires delaying gratification. Study 2 investigates how happiness vs. sadness influence performance on a task that requires persistence on an aversive task. Finally, Study 3 examines how happiness vs. sadness influence performance on a task that requires breaking a habitual response. In addition, in all three studies, we account for the potentially confounding effects of baseline emotional states, practice performance, and trait self-control, allowing us to draw more convincing conclusions regarding the effect of happiness vs. sadness on self-control (Kahlert et al., 2017).

Self-control: an overview

The terms self-control, self-regulation, willpower, and self-discipline are often used interchangeably to refer to the same process. Self-control is considered to be closely related to other constructs such as impulsivity and Big Five conscientiousness (Friese & Hofmann, 2009; Tangney et al., 2004), although these constructs are not entirely the same. For example, although Friese and Hofmann (2009) found a negative correlation between self-control and impulsivity across studies, the correlations ranged from −.58 to −60, suggesting that self-control and impulsivity are not completely redundant concepts. Furthermore, while self-control predominantly operates in situations where there is a conflict between impulses and long-term goals, conscientiousness can operate with or without impulses (de Boer et al., 2011). Moreover, even though lumping self-control with related constructs such as conscientiousness has been successful in predicting various life outcomes, splitting self-control from these related constructs has been particularly successful in predicting outcomes such as achievement (Duckworth & Seligman, 2017). Thus, self-control is typically studied as a separate construct.

Moreover, while self-control theories often conceptualize self-control as involving the inhibition of impulses (e.g., Tangney et al., 2004), there are additional ways in which self-control has been conceptualized in the literature. For example, the classic delay of gratification concept emphasizes that self-control involves choosing delayed but larger rewards over immediate but smaller rewards (Mischel, 1974), whereas more recent concepts emphasize that self-control involves enacting and persisting in aversive activities (Carver, 2010). Thus, these models highlight that self-control involves several important, yet somewhat distinct, components or abilities (i.e., inhibition, delaying gratification, and persistence), although these models have in common that self-control concerns decisions in which people sacrifice short-term desires in favor of greater long-term benefits (de Ridder et al., 2011; Fujita, 2011).

Theoretical models have also distinguished between self-control as a stable trait that differs between individuals, and self-control as a state that can change across time and situations (de Ridder et al., 2012; Tangney et al., 2004). For instance, the elaborated process model explains failures in state self-control in terms of the goals that are active during task performance. Specifically, shifts in attention, motivation, and emotion due to changing goals can account for self-control failures during task performance (Inzlicht et al., 2014). According to this model, people have “have-to” goals and “want-to” goals, and if they have been pursuing the former in a task with high self-control demands, people become motivated to pursue the latter, enjoyable goals, instead of continuing exerting self-control. This causes a shift in emotion and attention toward the “want-to” goals, at the expense of continuing to pursue the “have-to” goals. Although this model can help us understand why people fail at self-control, little is known about what makes people succeed at self-control (Gillebaart & de Ridder, 2015).

Emotions and self-control

One potential mechanism for improving self-control might be through emotions. Emotions are “experiential, physiological, and behavioral responses to personally meaningful stimuli” (Mauss & Robinson, 2009, p. 209), which can be organized according to various perspectives. Dimensional perspectives organize emotions according to valence, contrasting pleasant with unpleasant emotions (e.g., happy vs. sad), and/or according to arousal, contrasting high with low arousal emotions (e.g., surprised vs. quiet). In contrast, discrete perspectives assume that each emotion has its own unique profile and therefore distinguish distinct emotions from each other (e.g., anger vs. contempt; Ekman, 1992; Mauss & Robinson, 2009; Russell, 1980). While debates between these perspectives remain, researchers agree that both perspectives make important contributions in understanding emotions (Ekman, 2016; Harmon-Jones et al., 2017).

Given the strong impact of emotional states on wellbeing, motivation, and behavior (Hartmann et al., 2023), researchers have been interested in examining how emotions influence self-control. These studies have proposed that positive emotions can enhance self-control relative to negative emotions (Aspinwall, 1998; DeSteno, 2018; Tornquist & Miles, 2019), both when these emotions are induced in the laboratory (Raghunathan & Trope, 2002; Winterich & Haws, 2011), and when they are naturally experienced in real-life (Niermann et al., 2016; Vinci et al., 2017). For example, Winterich and Haws (2011) found that inducing a positive (vs. negative) mood in participants reduced their unhealthy snack preferences, and Raghunathan and Trope (2002) found that inducing a positive mood in caffeine consumers decreased their intentions toward caffeine consumption. In a more naturalistic setting, Niermann et al. (2016) found that increased positive affect predicted more time spent exercising that day, whereas the reverse was true for negative affect. Vinci et al. (2017) similarly found that positive emotions were associated with a lower risk of relapse in smokers. This research provides evidence that positive emotions can promote behaviors that require self-control such as food consumption and physical exercise.

Although studies examining how emotions impact performance on lab-administered self-control tasks are somewhat scarce, some studies have found that presenting participants with negative vs. neutral words or images in a Stroop or Stop-signal task impairs their performance on these tasks (Frings et al., 2010; Kalanthroff et al., 2013). The assumption in these studies is that the emotional stimuli capture attention and thus receive prioritized processing, thereby disrupting performance (i.e., emotional interference), although it remains debated whether this effect is due to the threat, negativity, or arousal of the emotional stimuli (Song et al., 2017). Moreover, research has found that presenting participants with positive vs. neutral words in a Stroop task improves their performance (Liu et al., 2018), providing initial evidence that positive emotions can facilitate performance on lab-administered self-control tasks.

It is important to note, however, that these effects tend to be attributed to cognitive mechanisms, and that these studies have typically not measured emotional processes, making it difficult to draw conclusions regarding the impact of emotions on performance on lab-administered self-control tasks. One exception is research conducted by Tornquist and Miles (2022) who induced and measured emotions prior to a task requiring self-control. They found both that people believed positive emotions would facilitate their self-control performance more than negative emotions, and that positive emotions actually enhanced performance on the lab-administered self-control task. However, this research was designed to test the links between personality, beliefs about emotions, preferences to complete an emotion induction, emotions, and self-control. Therefore, more research is needed to directly test the causal effect of emotions on lab-administered self-control tasks.

One possible explanation for why positive emotions facilitate self-control comes from the broaden-and-build theory (Fredrickson, 2001). This theory proposes that, in contrast to negative emotions, positive emotions enhance peoples’ intellectual, physical, and social resources, and motivate them toward future achievements, which should ultimately make challenges such as self-control easier to manage (Wilcox et al., 2011; Winterich & Haws, 2011). The mood-maintenance theory similarly suggests that people who experience positive emotions are motivated to maintain these emotional states and therefore refrain from engaging in actions that may reduce them (such as immediate indulgence), which consequently promotes more virtuous choices (Andrade, 2005; Clark & Isen, 1982). Thus, positive emotions seem to increase motivation and prompt people to exert self-control and, consequently, achieve long-term goals.

Contrary to the above findings, however, research indicates that positive emotions may not always improve self-control. For example, Crossfield and Damian (2021) conducted Bayesian analyses and found evidence that presenting participants with negative vs. neutral vs. positive words in a Stroop task did not affect their performance. Moreover, Evers et al. (2013) found that people who experienced positive emotions consumed more unhealthy food compared to controls, and that they consumed similar amounts as those who experienced negative emotions. Similarly, Tornquist et al. (2023) conducted Bayesian analyses and found no effects of interventions aimed at increasing positive emotions on food consumption and physical exercise during COVID-19. Such discrepancies are likely due to a number of individual, contextual, and methodological factors such as variations in the type of emotions examined, the tasks used to induce emotions, and the behavior or task used to measure self-control (DeSteno, 2018; Patrick et al., 2009; Schmeichel & Inzlicht, 2013; Tornquist & Miles, 2019; Winterich & Haws, 2011).

For example, although Winterich and Haws (2011) held the self-control task constant (i.e., unhealthy snack preferences), they induced different emotions across studies (Study 1: happiness and hope; Study 2: hope, happiness, pride, and neutral; Study 3: hope and pride; Study 4: positive and negative mood) via various emotion induction tasks (i.e., writing and reading tasks), and found significant main effects of emotions on self-control in only some of the studies. Likewise, although Albarracin and Hart (2011) held the emotion induction task constant by asking participants to write an emotional letter to a friend, they induced different emotions through this writing task across studies (Study 1: happiness, anger, and neutral; Study 2 and 3: happiness and anger; Study 4: happiness and sadness) and varied the tasks used to measure performance (Study 1 and 4: ability to solve intellectual problems; Study 2: willingness to read articles; Study 3: memory for celebrity pictures). They found main effects of emotions on the outcomes in most, but not all, of their studies, although it is important to note that this study was designed to examine the interaction between priming type (i.e., viewing action words such as “go” vs. inaction words such as “pause”) and emotion induction (positive vs. negative) on generic performance rather than the main effect of emotions on self-control specifically.

Together, this research suggests that different ways of manipulating emotions and measuring self-control can lead to different conclusions regarding the impact of emotions on self-control, which is consistent with other lines of evidence suggesting that using different comparison groups, emotion induction methods, and measurement techniques can yield different effects (e.g., Karlsson & Bergmark, 2015; Lin et al., 2021). Thus, the picture of how positive and negative emotions influence self-control appears to be more complicated than initially envisioned, and more research is needed to examine the causal impact of positive emotions on self-control (Schmeichel & Inzlicht, 2013; Tornquist & Miles, 2019).

Specifically, given previous inconsistent findings, it seems particularly important to test the robustness of the effect of positive vs. negative emotions on self-control. For example, if positive, compared to negative, emotions help people to overcome current self-control challenges by increasing their personal resources and motivation toward future achievements (broaden-and-build theory: Fredrickson, 2001; Winterich & Haws, 2011) and if the effect of positive vs. negative emotions on self-control is indeed robust, we might expect to find that the facilitating effect of positive emotions on self-control will replicate across different self-control tasks that measure somewhat different aspects of self-control, which has, to our knowledge, never previously been tested.

Present studies

Prior research has found that positive emotions can facilitate self-control behaviors such as physical exercise and food consumption (Niermann et al., 2016; Winterich & Haws, 2011), although others have failed to link positive emotions with these behaviors (Evers et al., 2013; Tornquist et al., 2023). Likewise, although research indicates that positive emotions may enhance performance on lab-administered self-control tasks such as the Stroop task (Liu et al., 2018), others have failed to replicate these findings (Crossfield & Damian, 2021). Such discrepancies could partly be due to differences in methodology between these studies, such as the different emotion induction methods used, the type of emotions examined, and the behavior or task used to measure self-control. Therefore, in the current research we hold the emotion induction and emotion ratings constant while varying the self-control tasks as a better way to investigate how positive and negative emotions influence self-control.

Specifically, participants in all three studies completed the same emotion inductions and emotion ratings to ensure that their emotional experiences were similar across studies. Participants also practiced their respective self-control task prior to completing the emotion inductions to ensure that they understood that the upcoming task would require self-control (Tornquist & Miles, 2022). We particularly focused on contrasting the effects of states of happiness with states of sadness on performance on tasks that require self-control. We did so because happiness and sadness are basic emotions commonly experienced by the general population, they elicit consistent bodily sensations across people, and are often used as comparison emotions in emotion research (Ekman, 2016; Hartmann et al., 2023; Kowalska & Wróbel, 2017; Lench et al., 2011). Happiness, and conceptually overlapping emotions (e.g., joy), reflects approach behaviors, and is often associated with increased motivation, effort, and activation, and allows people to progress toward their goals (Hartmann et al., 2023; Johnson, 2020). Sadness, and conceptually overlapping emotions (e.g., grief), reflects avoidance behaviors and is often associated with increased withdrawal, less reward-seeking, and goal disengagement, and allows people to cope with their losses (Arias et al., 2020; Wrosch & Miller, 2009), although the impact of sadness on goal disengagement can sometimes be adaptive when goals are unattainable (Barlow et al., 2022).

In addition, given that self-control involves several important components or abilities, and that studies testing the impact of manipulated emotions on lab-administered self-control tasks are lacking and/or have yielded inconsistent findings, the current research aims to add to this body of research by measuring performance on three different lab-administered self-control tasks that measure somewhat different aspects of self-control. Study 1 examined whether assigning participants to a happiness or sadness induction leads to enhanced performance on a task that asked them to choose between small-immediate and large-delayed monetary awards, which is commonly used to measure the ability to delay gratification in the lab (Duckworth & Kern, 2011; Hagger et al., 2010, Li, 2008; Tuk et al., 2011). Study 2 tested whether assigning participants to a happiness or sadness induction facilitates performance on an aversive anagram task (the latter often used to measure persistence: e.g., Imhoff et al., 2014, Muraven et al., 1998; Tornquist & Miles, 2022). Finally, Study 3 investigated whether assigning participants to a happiness or sadness induction improves performance on a sentence completion task that requires breaking a habitual response (i.e., response inhibition). Thus, even though these tasks have in common that they involve voluntary control in the service of greater goals (i.e., self-control), they measure different components of self-control (i.e., the ability to delay gratification, persistence, and response inhibition).

Finally, given that participants’ baseline emotional experiences and practice performance are likely to strongly influence subsequent engagement with the emotion inductions and performance on the self-control tasks (see Negrão et al., 2022; Sorjonen et al., 2019), we controlled for these variables in our studies. Likewise, because trait self-control predicts emotional experiences in performance contexts and is considered a key ingredient for goal attainment (Milyavskaya & Inzlicht, 2017; Tornquist & Miles, 2022), this trait is also likely to influence engagement with the emotion inductions and self-control tasks, and we therefore accounted for trait self-control in our studies. Thus, the methods in our three studies were identical except the self-control task performed by the participants, allowing us to draw more convincing conclusions regarding the effects of happiness vs. sadness on three different self-control tasks, while accounting for the potentially confounding effects of baseline emotions, practice performance, and trait self-control. This research therefore provides a crucial step for advancing our causal knowledge of how happiness vs. sadness influence different aspects of self-control, which may provide a foundation for designing interventions to improve these emotional states and self-control in individuals.

All three studies were approved by the local Ethics Committee and all data and materials have been made available via the Open Science Framework, and can be accessed online (https://osf.io/6bqax/?view_only=a728d0b50e3a4659b88268ecd174819a). Herein, we report a sensitivity power analysis conducted in G*Power (Faul et al., 2009) for our key hypothesis test in each study, which is thought to be the most informative power analysis as it allows researchers to report the minimum effect size their experiment had 80% power to detect (Perugini et al., 2018).

Study 1

To test how happiness vs. sadness influence performance on a task that requires delaying gratification, participants in Study 1 completed measures of current emotions and were assigned to either a happiness or sadness induction. Specifically, they were asked to recall a personal event that made them feel happy or sad, respectively. The ability to delay gratification was measured by having participants complete a task that asked them to choose between small-immediate and large-delayed monetary rewards.

Method

Participants

Psychology undergraduates (N = 157) participated in exchange for course credit (M_age = 21.85 years, SD_age = 6.59; 138 females; 59% White, 19% Hispanic/Latino, 9% Asian American/British, 5% Black American/British, 8% Mixed/Multiple/Other, and 1% Native American/Pacific Islander). A sensitivity power analysis (N = 157; power criterion = .80; alpha = .05) for the effect of condition on self-control with baseline emotions, practice performance, and trait self-control as covariates showed that our study was powerful enough to detect small-to-medium effects (f = .23).

Materials

Trait self-control

The Brief Self-Control Scale (Tangney et al., 2004) was used to measure trait self-control. Participants rated 13 items (e.g., “Sometimes I can’t stop myself from doing something, even if I know it is wrong”) on a scale from 1 (not at all like me) to 5 (very much like me). Final scores are the mean of the 13 items (9 items are reverse scored; Cronbach’s α = .81), and higher scores indicate greater trait self-control.

Current emotions

Emotions were chosen from the Modified Differential Emotions Scale (Fredrickson et al., 2003). Each emotion was defined by three adjectives (e.g., joy: joyful, glad, or happy) to ensure the same understanding of these constructs among the participants. Participants rated how much they currently experienced four positive emotions (proud, serene, hopeful, joyful), four negative emotions (angry, guilty, anxious, sad), and two filler items (i.e., tired, concentrated) that were embedded with the other emotions (Tamir & Ford, 2012) on a scale of 1 (not at all) to 7 (extremely), although the target emotions herein are joy (i.e., happiness) and sadness. Higher scores indicate greater experiences of that emotion.

Emotion induction

Emotions were induced by randomly assigning participants to either a happiness or sadness induction where they were asked to recall a personal event that made them feel very happy and very sad, respectively. Participants were asked to spend at least five minutes writing about their experiences. The writing task was chosen because it has been effective in inducing emotions in several prior studies (e.g., Forgas, 1999, 2011; Lench & Levine, 2005; Tornquist & Miles, 2022). Average time spent on the writing task was 8.41 minutes (SD = 3.56).

Self-control performance

To measure self-control performance, we used the Monetary Choice Questionnaire (MCQ; Kirby et al., 1999), which has been used to measure self-control or self-discipline in prior research (e.g., Duckworth & Seligman, 2005; Tuk et al., 2015). As such, participants made 27 choices, with each asking them to choose between receiving a smaller hypothetical amount of money (e.g., $55) today or a larger hypothetical amount of money (e.g., $75) later in time (e.g., 61 days). Participants who choose smaller immediate rewards in this task exhibit lower self-control because they fail to inhibit their desire for an immediate payoff in favor of a long-term more beneficial one. In line with prior work (e.g., Tuk et al., 2015), the number of times participants chose the delayed reward over the immediate reward was our measure of self-control. A score of zero indicates that the participant always chose the immediate reward, whereas a score of 27 indicates that the participant always chose the delayed reward. Thus, we measured participants’ hypothetical rather than actual monetary choices as an index of self-control performance, with a higher score indicating enhanced self-control.

Procedure

Participants completed questions about demographics (age, gender, ethnicity), trait self-control, and current emotions. Participants were then told that they would practice a choice task, and that they would complete the actual choice task later. Thus, participants were presented with three practice choices, each asking them to choose between a small-immediate reward and a larger-delayed reward (e.g., whether they would prefer to get $34 today or $50 in 30 days). Participants were told that they were going to make very similar choices later in the study and were asked to take the choices seriously. After making their three choices, participants were asked to complete a writing task, which involved recalling a past event. For this task, participants were randomly assigned to either the happiness or sadness inductions, in which they were instructed to write about a happy or sad personal event, respectively, for at least five minutes. Participants then re-rated their current emotions (i.e., manipulation check) and completed the MCQ self-control task. Each pair of choices in the MCQ was presented on a separate page, and the small-immediate choice was presented on the left and the large-delayed choice on the right. Participants made each choice by clicking on it, before proceeding to the next choice. After making their choices, participants were thanked and dismissed. The study took 15 minutes to complete.

Results and discussion

Manipulation check: How did the emotion induction influence emotions?

We conducted a MANCOVA to test whether participants in the happiness induction experienced more happiness than participants in the sadness induction, and whether participants in the sadness induction experienced more sadness than participants in the happiness induction. Emotion induction (happiness, sadness) was a between-subject factor, baseline happiness and sadness were covariates, and experiences of happiness and sadness after the emotion induction were dependent variables.

The analysis revealed a significant multivariate effect of emotion induction, F(2, 152) = 63.37, Wilks’ λ= .55, p < .001, ${\eta }_p^2$= .46; there was a univariate effect of emotion induction on happiness, F(1, 153) = 103.25, p < .001, ${\eta }_p^2$= .40, and a univariate effect of emotion induction on sadness, F(1, 153) = 72.68, p < .001, ${\eta }_p^2$= .32. Participants assigned to the happiness induction experienced more happiness (M = 4.78, SE = .15) than those assigned to the sadness (M = 2.59, SE = .15; p < .001) emotion induction. Participants assigned to the sadness induction experienced more sadness (M = 3.69, SE = .16) than those assigned to the happiness (M = 1.69, SE = .17; p < .001) emotion induction. Thus, the emotion induction was successful.

Key analysis: how did the emotion inductions influence self-control on the delay gratification task?

We conducted an ANCOVA to test whether participants in the happiness induction demonstrated better self-control than participants in the sadness induction. Emotion induction (happy, sad) was a between-subject factor, baseline happiness and sadness, practice performance, and trait self-control were covariates, and performance on the delay gratification task was the dependent variable. The ANCOVA showed that participants assigned to the happiness induction (M = 12.86, SE = .43) performed significantly better on the delayed gratification task than participants assigned to the sadness induction (M = 11.37, SE = .43), F(1, 151) = 5.85, p = .02, ${\eta }_p^2$= .04, while controlling for baseline emotions, practice performance, and trait self-control (see Table 1). This finding suggests that happiness, compared to sadness, enhance performance on a self-control task that requires delaying gratification.

Table 1. Analysis of covariance (ANCOVA) summary table for the effect of emotion induction on self-control, with trait self-control (TSC), practice performance, and baseline (T1) happiness and sadness as covariates.

	F	p	${\eta }_p^2$
Study 1
TSC	.47	.50	.003
Practice Performance	40.14	<.001	.21
T1 happiness	.28	.60	.002
T1 sadness	2.24	.14	.02
Emotion induction	5.85	.02	.04
Study 2
TSC	.12	.73	.001
Practice Performance	81.73	<.001	.40
T1 happiness	1.12	.29	.009
T1 sadness	.36	.55	.003
Emotion induction	4.13	.04	.03
Study 3
TSC	.68	.41	.005
Practice Performance	376.67	<.001	.75
T1 happiness	.14	.71	.001
T1 sadness	1.16	.28	.009
Emotion induction	9.74	.002	.07

Study 1: N = 157; Study 2: N = 131; Study 3: N = 133. Significant at the p < .05 level.

Study 2

To test how happiness vs. sadness influence performance on a task that requires persistence on an aversive task, participants in Study 2 completed measures of current emotions and were assigned to either a happiness or sadness induction where they were asked to recall a personal event that made them feel happy or sad, respectively. To measure self-control, participants completed a self-control task that asked them to generate as many words as possible from 15 anagrams.

Method

Participants

Participants (N = 134) were recruited through Amazon Mechanical Turk (MTurk) in exchange for $1.50 (M_age = 35.87 years, SD_age = 10.20; 64 females; 79% White, 8% Hispanic/Latino, 7% Asian American, 6% African American, and 1% Other). Only adults living in the U.S.A. participated; all participants were fluent in English. A sensitivity power analysis (N = 134; power criterion = .80; alpha = .05) for the effect of condition on self-control with baseline emotions, practice performance, and trait self-control as covariates showed that our study was powerful enough to detect small-to-medium effects (f = .24).

Materials

Trait self-control

Participants completed the same self-control measure as in Study 1 (Cronbach’s α = .88).

Current emotions

Participants completed the same emotion measure as in Study 1.

Emotion induction

Participants completed the same emotion inductions as in Study 1. Average time spent on the writing task in Study 2 was 6.83 minutes (SD = 1.77).

Self-control performance

Based on prior research suggesting that attempting to generate words from anagrams requires self-control (Imhoff et al., 2014; Muraven et al., 1998; Tornquist & Miles, 2022), we used an anagram task to measure this construct. Specifically, participants were asked to generate words from 15 anagrams (e.g., LSEPE and FFRAEVO: Calef et al., 1992). The generated words could be of any length. To make the task more challenging, however, participants were encouraged to use all the letters (three anagrams were unsolvable for this type solution e.g., SICLAPI). The words also needed to be real English words and no letters could be used twice. Participants could stop the task and continue to the next part of the study whenever they wanted to. We used this procedure because self-control is the process of persisting in behaviors that may be boring or unpleasant but that must be completed to achieve a goal (Carver, 2010; Davisson, 2014). Thus, participants who persisted longer on the anagram task despite being able to discontinue whenever they wished demonstrated enhanced self-control. Time spent on the task was therefore our measure of self-control performance (Määttänen et al., 2021).

Procedure

The procedure was almost identical to the procedure in Study 1. That is, participants in Study 2 completed questions about demographics, trait self-control, and current emotions. They were then asked to practice an anagram task, and were told that they would complete the actual anagram task later. Specifically, participants were presented with three practice anagrams (e.g., YOMDAN) and were asked to generate words from each one, while following the rules described above. After practicing the anagram task, participants completed the mood-induction writing task that involved recalling a past event. Participants were randomly assigned to either the happiness or sadness inductions and were instructed to write about a happy or sad personal event, respectively, for at least five minutes. Participants then indicated their current emotions again (i.e., manipulation check) and completed the self-control task, which asked them to generate words from 15 anagrams. The anagrams were presented on the same page and participants indicated their solutions using an empty textbox next to each anagram. When they could no longer (or did not want to) generate more words, participants clicked the “Next Page” button to advance to the next screen where they were thanked and dismissed. The study took 15 minutes to complete.

Results and discussion

Manipulation check: How did the emotion inductions influence emotions?

We conducted the same MANCOVA as described in Study 1. The analysis revealed a significant multivariate effect of emotion induction, F(2, 129) = 35.60, Wilks’ λ= .64, p < .001, ${\eta }_p^2$= .36; there was a univariate effect of emotion induction on happiness, F(1, 130) = 45.60, p < .001, ${\eta }_p^2$= .26, and a univariate effect of emotion induction on sadness, F(1, 130) = 58.98, p < .001, ${\eta }_p^2$= .31. Participants assigned to the happiness induction experienced more happiness (M = 5.13, SE = .18) than participants assigned to the sadness induction (M = 3.35, SE = .19; p < .001). Moreover, participants assigned to the sadness induction experienced more sadness (M = 4.03, SE = .19) than participants assigned to the happiness induction (M = 2.00, SE = .19; p < .001). These findings suggest that the emotion induction was successful.

Key analysis: How did the emotion inductions influence self-control on the anagram task?

We conducted the same ANCOVA as described in Study 1, but with performance on the anagram task as the dependent variable. Because three people had extreme scores on the anagram practice task (i.e., they practiced the task for more than 20 minutes), they were excluded from the analysis. Results showed that participants assigned to the happiness induction (M = 5.09, SE = .29) performed better on the anagram task than those assigned to the sadness induction (M = 4.23, SE = .30), F(1, 124) = 4.13, p = .04, ${\eta }_p^2$= .03, while controlling for baseline emotions, practice performance, and trait self-control (see Table 1). As in Study 1, this finding suggests that happiness, compared to sadness, improves performance on a self-control task that requires persistence.

Study 3

To test how happiness vs. sadness influence performance on a task that requires breaking a habitual response (i.e., response inhibition), participants in Study 3 completed measures of current emotions and recalled a personal event that made them feel either happy or sad. They then performed a sentence completion task which first asked them to complete the stem of various sentences with as many fitting/related words as possible, and then to break this habitual response by completing various sentences with as many unfitting/unrelated words as possible.

Method

Participants

Psychology undergraduates (N = 133) participated in exchange for course credit (M_age = 24.32 years, SD_age = 6.81; 105 females; 32% White, 39% Hispanic/Latino, 16% Asian American/British, 8% Mixed/Other, 3% Black American/British, and 1% Native American/Pacific Islander). A sensitivity power analysis (N = 133; power criterion = .80; alpha = .05) for the effect of condition on self-control with baseline emotions, practice performance, and trait self-control as covariates showed that our study was powerful enough to detect small-to-medium effects (f = .24).

Materials

Trait self-control

Participants completed the same self-control measure as in Study 1 and 2 (Cronbach’s α = .84).

Current emotions

Participants completed the same emotion measure as in Study 1 and 2.¹

Emotion induction

We used the same emotion induction as in Study 1 and 2. Average time spent on the writing task in Study 3 was 9.11 minutes (SD = 3.43).

Self-control performance

To measure self-control performance, we used a sentence completion task that has been used in several prior studies to measure response inhibition, which is crucial for successful self-control and goal attainment (Burgess & Shallice, 1997; de Ridder et al., 2011; Martin et al., 2019; Tangney et al., 2004). Participants were first presented with 13 sentences with the last word of each sentence missing (e.g., “The crime rate has gone up this … ”) and were asked to complete the sentence with words that fit at the end of the sentence. In our example, words such as “month” and “summer” would correctly complete the sentence. Participants were then presented with 13 additional sentences with the last word of each sentence missing (e.g., “He mailed the letter without a … ”) and were asked to complete the sentence with words that were unrelated to the sentence in every way, which predominately requires response inhibition as participants are required to inhibit the natural impulse to complete the sentence with a fitting word. In our example, unrelated words such as “dog” and “rain” would count as correct responses, whereas related words such as “stamp” and “address” would be incorrect. Participants were encouraged to generate as many words as possible and were told that they could continue to the next task when they could no longer (or did not want to) generate new words. Time spent carrying out this task was strongly correlated with the number of unrelated words generated (r = .76, p < .001), so we used the former to objectively measure self-control, with more time indicating better self-control.

Procedure

The procedure in Study 3 was similar to the previous two studies. Participants completed questions about demographics, trait self-control, and current emotions. They then practiced a sentence task, and were told that they would complete the real sentence task later. Specifically, participants were presented with four practice sentences; two asked them to complete the sentence stem with fitting words and two asked them to complete the sentence stem with unrelated words. After completing the practice sentences, participants were asked to complete the mood induction task, where they were randomly assigned to either the happiness or sadness induction by writing about a happy or sad personal event, respectively, for at least five minutes. Participants then indicated their current emotions (i.e., manipulation check) and completed the self-control task, which asked them to first complete 13 sentences with fitting words and then complete 13 sentences with unrelated words. Each sentence was presented on a separate page and participants wrote their responses in a textbox next to each sentence stem. When they could no longer think of new words, participants clicked the “Next Page” button to advance to the next screen where they received the next sentence stem to complete. After completing this task, participants were thanked and dismissed. The study took 20 minutes to complete.

Results and discussion

Manipulation check: How did the emotion inductions influence emotions?

As in Study 1 and 2, we conducted a MANCOVA to test this. The analysis revealed a significant multivariate effect of emotion induction, F(2, 128) = 66.04, Wilks’ λ = .49, p < .001, ${\eta }_p^2$= .51; there was a univariate effect of emotion induction on happiness, F(1, 129) = 96.33, p < .001, ${\eta }_p^2$= .43, and a univariate effect of emotion induction on sadness, F(1, 129) = 82.98, p < .001, ${\eta }_p^2$= .39. Participants assigned to the happiness induction experienced more happiness (M = 4.82, SE = .15) than participants assigned to the sadness induction (M = 2.58, SE = .17; p < .001), and participants assigned to the sadness induction experienced more sadness (M = 3.82, SE = .17) than participants assigned to the happiness induction (M = 1.67, SE = .16; p < .001). Thus, the emotion induction was successful.²

Key analysis: How did the emotion inductions influence self-control on the sentence completion task?

We conducted the same ANCOVA as previously but with performance on the sentence completion task as the dependent variable. The ANCOVA showed that participants assigned to the happiness induction (M = 15.75, SE = .49) performed better on the sentence completion task than participants assigned to the sadness induction (M = 13.50, SE = .52), F(1, 127) = 9.74, p = .002, ${\eta }_p^2$= .07, while controlling for baseline emotions, practice performance, and trait self-control (see Table 1). In line with Study 1 and Study 2, this finding suggests that happiness, compared to sadness, facilitates performance on a self-control task that requires breaking a habitual response.³

General discussion

Building on insights from the broaden-and-build theory (Fredrickson, 2001), in three studies we examined whether experimentally inducing happiness vs. sadness in participants influenced their performance on three different lab-administered self-control tasks that measured the ability to delay gratification, persistence on an aversive task, and response inhibition. Our results showed that participants in a happy state performed better on all three self-control tasks compared to those in a sad state, while controlling for the potentially confounding effects of baseline emotions, practice performance, and trait self-control. Thus, our findings provide strong causal evidence for the facilitating effect of happiness on self-control, and suggest that this effect extends to tasks that measure different components of self-control, which has never previously been tested.⁴

While some prior studies have shown positive emotions can enhance self-control relative to negative emotions, other studies have failed to corroborate this claim (Schmeichel & Inzlicht, 2013; Tornquist & Miles, 2019). Indeed, these inconsistent results could be due in part to methodological differences between studies. Specifically, prior studies have varied the type of positive and negative emotions examined, the tasks used to induce emotions, and the task used to measure self-control (e.g., Winterich & Haws, 2011).

In contrast, the methods employed in our three studies were identical (including the method of emotion induction, the emotions examined, and emotion ratings) except for the self-control tasks performed by participants, and we controlled for baseline emotions, practice performance, and trait self-control. This allows us to draw more convincing conclusions regarding the effects of emotion induction on self-control. When implementing these procedures, we found that participants assigned to the happiness induction demonstrated better performance on all three self-control tasks, as compared to participants in the sadness induction. These findings support the broaden-and-build theory (Fredrickson, 2001) which proposes that positive emotions such as happiness enhance peoples’ personal intellectual resources and motivate them toward future achievements, making current self-control dilemmas easier to overcome by broadening the thought-action repertoire (e.g., Wilcox et al., 2011). In contrast, negative emotions such as sadness tend to narrow the thought-action repertoire toward threat-related stimuli, thus having a negative impact on self-control. These data also extend this theory by showing that the effect of happiness vs. sadness generalize to challenging tasks that require different forms of self-control.

Our findings are also consistent with prior studies showing that positive emotions enhance self-control relative to negative emotions (e.g., Tornquist & Miles, 2022), but inconsistent with both research reporting no significant effects of emotion valence on self-control (e.g., Crossfield & Damian, 2021), and research reporting significant effects of emotions on self-control in some, but not all, of their studies (e.g., Winterich & Haws, 2011). Importantly, our findings extend prior work by providing compelling evidence that, when employing a widely-used writing task to induce two basic, commonly experienced emotions (i.e., happiness and sadness: Hartmann et al., 2023; Lench et al., 2011), happiness improves self-control compared to sadness across different lab-administered tasks measuring different elements of self-control. Specifically, these data further show the facilitating effect of happiness on self-control tasks that measure delaying gratification, persistence on an aversive task, and response inhibition. According to various models of self-control, these abilities are considered three important, yet somewhat, distinct aspects of self-control (e.g., Carver, 2010; Mischel, 1974; Tangney et al., 2004). Thus, our findings suggest that the effect of happiness vs. sadness is robust, which has, to our knowledge, never been tested.

In fact, studies examining how emotions influence performance on self-control tasks have often used single self-control tasks that yield inconsistent results. For example, some studies have presented participants with emotional stimuli in a Stroop task and measured their performance on this task. Using this technique, Liu et al. (2018) showed that positive vs. neutral words enhanced Stroop performance, whereas Crossfield and Damian (2021) failed to replicate this effect. However, other lines of research suggest that Stroop tasks may be less valid measures of self-control compared to other lab-administered self-control tasks (e.g., delay gratification tasks: Duckworth & Kern, 2011). Thus, our results show that, when using validated self-control tasks, induced happiness improves performance on these tasks relative to sadness.

Furthermore, while our results replicate previous studies such as Tornquist and Miles’s (2022) and Pyone and Isen’s (2011) findings that positive emotions improved performance on lab-administered self-control tasks, these studies were not specifically designed to test the causal effect of induced positive vs. negative emotions on self-control, and the former only controlled for baseline emotions whereas the latter did not account for any confounding variables. Thus, we add to this prior work by carrying out a direct examination of the impact of induced happiness vs. sadness on various self-control tasks while ruling out the confounding effects of baseline emotions, practice performance, and trait self-control, which is crucial for securing a better understanding of the effects of positive and negative emotions such as happiness and sadness on self-control.

Interestingly, we found that the effect sizes were similar in magnitude and ranged from small-to-medium in our three studies (${\eta }_p^2$= .03, .04, and .07), suggesting that emotion inductions eliciting happiness may have a modest effect on self-control tasks that require delaying gratification, persistence on an aversive task, and breaking a habitual response when compared to sadness. However, it can also be argued that because the effect size was slightly higher in Study 3, happiness might be particularly beneficial for self-control tasks that involve breaking a habitual response, perhaps because breaking a habit is particularly challenging and requires a lot of effort (Lally & Gardner, 2013). The latter interpretation is consistent with research arguing that the effect of emotions on self-control may depend on the specific type of self-control required by the task (Tornquist & Miles, 2019). Nonetheless, this interpretation is speculative and future studies should examine whether the facilitating effect of happiness on self-control is moderated by the type of self-control required.

Moreover, it has been argued that positive emotions may influence self-control differently depending on whether the task requires the initiation or inhibition of behaviors (initiatory vs. inhibitory self-control: Tornquist & Miles, 2019; see also de Boer et al., 2011), and that delay-discounting and anagram tasks may involve inhibitory and initiatory self-control, respectively (Imhoff et al., 2014; Tuk et al., 2015). The present studies, however, suggest that happiness may improve performance on self-control tasks that require both the initiation and inhibition of behaviors, given that the obtained effect sizes were almost identical in Study 1 and Study 2. But, our research was not designed to test this, and future studies should examine how happiness vs. sadness influence performance on tasks that were specifically designed to measure initiatory vs. inhibitory self-control. Finally, we note that, while these effect sizes are comparable with the effects obtained in previous similar studies and in psychological research in general (Tornquist & Miles, 2022; Schäfer & Schwarz, 2019), it is crucial for future studies to examine factors that may more strongly predict self-control success (e.g., positive thinking: Hennecke et al., 2019).

While our findings contribute to understanding the causal influence of happiness and sadness on three self-control tasks that measure different components of self-control, future research should investigate whether these findings replicate outside the lab, and should also extend the types of self-control tasks employed. For example, Study 1 measured participants’ hypothetical rather than actual monetary choices, and it is not clear whether our findings would replicate in more realistic scenarios. In terms of extending to other self-control tasks, studies could also examine how happiness vs. sadness influence performance on other common self-control tasks that involve pain tolerance, crossing out letters, squeezing a handgrip, solving complex math problems, and retyping text while avoiding certain keys (Duckworth & Kern, 2011; Hagger et al., 2010; Muraven et al., 1998).

Of central concern is that although these tasks, as well as the tasks used in the present research, are thought to require self-control, it has been proposed that self-control tasks administered in the lab may actually require very little self-control, especially in comparison to self-control behaviors carried out in real-life (Shenhav, 2017). One key challenge for future research on the effect of emotions on self-control therefore involves developing tasks that are even higher in the demands they place on self-control. One possibility would also be to conduct a study that employs the experience sampling method to assess the emotions that people experience in conjunction with engaging in various self-control behaviors in real-life, and test which emotions that predict success in these situations. It would, of course, be challenging to experimentally induce emotions in studies using this method, which is the most rigorous means of testing the causal impact of emotions on outcomes (Siedlecka & Denson, 2019).

Furthermore, given that the current study aimed to test how happiness vs. sadness influenced self-control, future studies could contrast distinct positive emotions with each other and examine how these emotions influence self-control. While pride and hope generally seems to have a facilitating effect on self-control behaviors such as food consumption (Patrick et al., 2009; Winterich & Haws, 2011), much less is known about how these and other distinct positive emotions influence performance on common lab-administered self-control tasks.³ Future research could take a more nuanced perspective on the exact positive emotions that might be most helpful when performing tasks that require self-control. For example, given that joy creates the urge to play and pride creates the urge to envision achievements, the former might be particularly useful when initiating physical self-control behaviors while the latter might be more useful when performing intellectual self-control tasks (see Fredrickson, 2001, for a review of distinct positive emotions). Relatedly, given that it has been recommended that the effect of an emotion on an outcome should be compared with other emotions and neutral groups (Lench et al., 2011), future studies should test whether assigning participants to a happiness-eliciting condition improves self-control compared to a neutral control condition.

Moreover, researchers should explore how emotional arousal (high vs. low) influences performance on lab-administered self-control tasks. Given that happiness typically involves moderately high arousal whereas sadness typically involves moderately low arousal (Posner et al., 2005), an alternative interpretation of our findings could be that moderately high arousal emotions (such as happiness) facilitate self-control relative to moderately low arousal emotions (such as sadness). We believe this is unlikely given that previous research has shown that elevated emotional arousal actually impairs self-control tasks such as resisting M&M candies (Fedorikhin & Patrick, 2010; Schmeichel & Inzlicht, 2013),⁴ which is not what we found in our studies. Future studies could perhaps assess how emotions that capture different combinations of positive and negative valence and high and low arousal (i.e., sadness, happiness, fear and calmness) influence self-control.

Although one of the most common and valid measures of current emotional experiences is self-reported questionnaires, one possible drawback with this measure is that people may be unwilling to report on their emotional states, particularly when these emotional states are negative in valence (Mauss & Robinson, 2009). To assess people’s emotional experiences and their impact on self-control, researchers could consider using additional emotion assessments such as physiological measures of skin conductance levels or bodily sensation mapping, which is a self-report tool where people draw their feelings in the format of bodily sensations in a body silhouette rather than reporting their feelings on a scale (Hartmann et al., 2023; Nummenmaa et al., 2014).

Moreover, even though we accounted for some methodological differences between previous studies by holding the emotion induction, emotions examined, and emotion ratings constant, there are other methodological differences to consider. For example, we found that the internal consistency for baseline negative emotions was higher in Study 2 (α = .93) than in Study 1 (α =.77) and Study 3 (α =.71), perhaps due to differences in sample characteristics between our studies. That is, compared to our younger student sample in Study 1 (M_age = 21.85) and Study 3 (M_age = 24.32), our older Amazon MTurk sample in Study 2 (M_age = 35.87) might have been less likely to discriminate between anger, guilt, anxiety, and sadness, given that they tend to be more stable in their experiences of different negative emotions than younger adults who may experience a great deal of anger compared to other negative emotions (Löckenhoff et al., 2008), resulting in higher correlations between our negative emotions in Study 2.

Finally, in all of our three studies, we had sufficient power to test the effects of the emotion inductions on self-control while controlling for baseline emotions, practice performance, and trait self-control, but future research should conduct well-powered studies that would allow them to control for additional potentially confounding variables (e.g., demographic factors such as socioeconomic status and other personality traits: Larsen & Ketelaar, 1991; Najdzionek et al., 2023; Tornquist & Miles, 2022), and to examine mediation and moderation effects to enhance our understanding of how and why happiness improves self-control (Schmeichel & Inzlicht, 2013). Given that theories (e.g., broaden-and-build and mood-maintenance theories) and research studies indicate that happiness promotes motivation, and that motivation facilitates goal attainment (Andrade, 2005; Benita et al., 2023; Clark & Isen, 1982; Fredrickson, 2001; Hartmann et al., 2023), future research could examine whether motivation mediates the relationship between happiness and self-control. Moreover, given that negative emotions such as sadness impair self-control, people who are able to effectively regulate these emotions should be more likely to succeed at self-control than those who are unable to regulate these emotions. Indeed, research has shown that people who accept and understand their negative emotions are more likely to progress toward their goals than people who suppress their negative emotions (Benita et al., 2021). Future studies could test whether the effect of sadness on self-control is moderated by habitual use of different emotion regulation strategies.

Implications

Given that self-control failures are common and can have devastating consequences for individuals and society, it is essential to understand the factors that influence self-control (Gillebaart & de Ridder, 2015). In an attempt to address this issue, the findings of the current research provide strong causal evidence that one of these factors is positive emotions, particularly states of happiness. Interventions to increase positive emotions might therefore have the potential to improve self-control and, consequently, have enormous social impact. Indeed, one way to improve self-control might be to teach people strategies to increase their positive emotions. For example, given that expressing gratitude and visualizing one’s best possible self may lead to short- and long-term increases in positive emotions (Carrillo et al., 2019; Sin & Lyubomirsky, 2009; but see; Tornquist et al., 2023), it is possible that regularly performing these activities (for example, by writing a gratitude diary) might lead to emotions that prompt people to exert self-control behaviors and, consequently, achieve desirable goals. Another promising, but less researched, intervention in relation to states of happiness might be Happiness through Goal-Setting Training (HTGST), which appears effective in improving people’s reasons for goal-pursuit through life satisfaction, positive affect, negative affect and work engagement (Ehrlich, 2023), which might subsequently promote self-control and goal attainment. The current research may also provide a foundation for developing new interventions to improve states of happiness and self-control in individuals. This might have great potential given that many existing interventions to improve self-control often fail (e.g., Miles et al., 2016).

Conclusion

The current research extends prior work by demonstrating that induced happiness can help people succeed on tasks that measure different aspects of self-control (i.e., delaying gratification, persistence, and response inhibition) relative to induced sadness, while accounting for the effects of baseline emotions, practice performance, and trait self-control. This research advances our causal knowledge of the facilitating effect of happiness on self-control, and may therefore have significant practical implications. Specifically, it has the potential to inform the design of interventions to increase states of happiness and improve self-control and, consequently, help people attain desirable life outcomes. Future research should examine how and why happiness improves performance on different self-control tasks and behaviors, as well as the role of other related factors (e.g., other emotions and positive thinking) in self-control success.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data and materials are available in the Open Science Framework repository (https://osf.io/6bqax/?view_only=a728d0b50e3a4659b88268ecd174819a).

Notes

1. Scores were averaged to form one composite score for positive emotions and one for negative emotions. The Cronbach’s alpha was acceptable in Study 1 (positive emotions α = .85; negative emotions α = .77), Study 2 (positive emotions α = .88; negative emotions α = .71), and Study 3 (positive emotions α = .88; negative emotions α = .71).

2. We also conducted the manipulation checks with overall experiences of positive and negative emotions after the emotion induction as dependent variables. The results were identical as when using happiness and sadness as dependent variables. Results are available at the OSF website.

3. We conducted supplementary ANOVAs to test whether participants in the happiness induction demonstrated better self-control than participants in the sadness induction, but without baseline emotions, practice performance, and trait self-control as covariates. Results showed that there was no significant main effect of the emotion induction on self-control in Study 1, F(1, 155) = 2.20, p = .14, = .01, or in Study 2, F(1, 132) = 2.83, p = .10, = .02. However, there was a significant, albeit small, effect of the emotion induction on self-control, F(1, 131) = 4.60, p = .03, = .03 in Study 3; participants assigned to the happiness induction performed better on the sentence completion task than participants assigned to the sadness induction. This shows that failing to eliminate important confounding variables identified in the literature can lead to misleading conclusions regarding the effect of emotions on self-control, which could explain previous inconsistent findings.

4. We conducted exploratory correlations to examine the associations between specific positive and negative emotions experienced after the emotion inductions (i.e., proud, serene, hopeful, joyful, angry, guilty, anxious, sad) and performance on all three self-control tasks, while controlling for baseline emotions and practice performance. Given the number of tests performed, Bonferroni correction was applied by adjusting the significance level: statistical significance was accepted at p < .003. Emotions were not significantly correlated with performance on any of the three self-control tasks (all p’s > .05), suggesting that distinct emotions and emotional arousal did not influence self-control in the present studies.