Do Character Strengths-Based Interventions Change Character Strengths? Two Randomized Controlled Intervention Studies

The VIA classification (Peterson & Seligman, 2004) describes “good character” through 24 positively valued traits known as character strengths. The 24 traits range from creativity and curiosity to kindness and humility but also include gratitude and spirituality. The science of character strengths has culminated in a multi-dimensional definition of character strengths as positive personality traits that reflect core identity, lead to positive outcomes for oneself and others, and contribute to the greater good (Niemiec, 2017). Character strengths have in common that they are all expected to contribute to the well-being of oneself and others; empirical findings (e.g., Hausler, Strecker, et al., 2017; Wagner et al., 2020) have widely confirmed this assumption. Notably, the relationships between character strengths and positive outcomes are also evident when analyzing the associations at the within-person level (Gander, Hofmann, et al., 2020).

Further, Wagner and Gander (2024) argued, in line with ideas of state-trait isomorphism (Fleeson, 2001; Fleeson et al., 2002; see also van Allen & Zelenski, 2018), that in addition to character traits, which describe the general tendency to act in accordance with a character strength across situations and time, we might also consider character strengths states, that is, specific enactments or displays of character strengths at the moment. Wagner and Gander (2024) showed that people report higher positive affect when displaying higher levels of character strengths states than usual, regardless of their character strengths trait level. Thus, these findings further corroborate the positive association between character strengths and positive outcomes.

The observation that displaying character strengths enhances well-being regardless of one’s disposition provides a foundation for developing generic, “one-fits-all” positive psychology interventions (Ruch et al., 2020). These interventions focus on specific character strengths, such as gratitude (e.g., writing a letter of gratitude, Seligman et al., 2005), kindness (e.g., counting self-performed acts of kindness, Otake et al., 2006), or humor (e.g., writing down three funny things, Gander et al., 2013), with the goal of increasing well-being.

The VIA classification does not only emphasize the overall role of character strengths for well-being but also introduces the notion that the individual configuration of character strengths is relevant. Drawing on Allport’s (1961) idea that each individual has some traits that are particularly descriptive of them, Peterson and Seligman (2004) suggested that each person has a set of three to seven character strengths, called signature strengths. These signature strengths are typical of this person, and their enactment is considered especially fulfilling. However, empirical research on the relevance of distinguishing between signature and non-signature strengths has been limited, with few studies (e.g., Blanchard et al., 2020) exploring this concept. Moreover, intervention studies have not consistently supported the notion that targeting an individual’s most prominent character strengths yields greater effectiveness than other character strengths (Proyer et al., 2015; Rust et al., 2009). Despite this, the concept of signature strengths has proven valuable for practical applications (Niemiec, 2017) and serves as a starting point for developing personalized interventions.

In an influential study using randomized, placebo-controlled online interventions, Seligman et al. (2005) evaluated two signature strengths exercises. Participants received feedback on their five highest-ranking character strengths and were instructed to (i) use them more often or (ii) use them in a new way. While the first exercise only led to transient increases in well-being, the latter exercise went along with higher well-being for up to six months after the intervention. Several replication studies have supported the positive effects of the latter exercise, sometimes demonstrating stronger effects compared to other well-established positive psychology interventions like the “three good things” exercise (Gander et al., 2013; Khanna & Singh, 2019; Mitchell et al., 2009; Mongrain & Anselmo-Matthews, 2012). A meta-analysis encompassing 14 studies on interventions based on signature strengths reported average effect sizes (Hedge’s g) ranging from g = .21 to .42 across several well-being and ill-being measures (Schutte & Malouff, 2019).

Studies on the effectiveness of various signature strengths-based interventions have yielded mixed results. While the “using signature strengths a new way” intervention has demonstrated consistent effectiveness, other variants have yielded less consistent findings. Some studies reported that learning about and observing one’s own (Dolev-Amit et al., 2021) or one’s partner’s (Habenicht & Schutte, 2023) character strengths could already have beneficial effects. However, other studies have not observed any effects in interventions that only ask participants to identify and use their character strengths (e.g., Seligman et al., 2005). As a result, it remains unclear what specific aspects of signature strengths-based interventions contribute to their success. In other words, what are the active ingredients (see Pawelski, 2020) of personalized CSIs? Is knowledge about one’s strengths sufficient to yield increases in well-being? Is the enactment of character strengths relevant, or is it necessary to enact them in novel ways, which could help broaden participants’ thought-action repertoire by displaying new behaviors or old behaviors in new contexts (Fredrickson, 2001)? Given substantial variability in samples, outcomes, and designs across studies, it is difficult to make sound conclusions when comparing their results. Previous studies have already systematically compared variants of popular positive psychology interventions such as “three good things” (Bahník et al., 2015), “acts of kindness” (Ko et al., 2021), or “gratitude letter” (Regan et al., 2023). In the present study, we aim to narrow this gap by systematically comparing different variants of personalized CSIs in one study, which allows for meaningful comparisons between the different variants.

Previous research has also been limited in two ways. First, some studies have solely focused on short interventions that typically last no more than one week. Second, in other studies, longer programs have been tested, but the effectiveness of individual components within those programs has yet to be thoroughly evaluated. Considering that substantial changes may require more time to unfold and that intervention intensity has been identified as a relevant moderator of intervention effectiveness (e.g., van Agteren et al., 2021), it would be desirable to implement longer interventions with multiple steps. This approach could enhance effectiveness by allowing for a more comprehensive intervention process. Thus, after comparing different variants of CSIs in Study 1, we will test a multi-component program based on signature strengths in Study 2.

In the literature, it is commonly assumed that one of the central working mechanisms of character strengths-based interventions is that they lead to a more frequent enactment of character strengths. The more frequent enactment can, in a bottom-up process, lead to changes in character strengths traits, which, in turn, lead to higher well-being (e.g., Quinlan et al., 2012). Two studies on interventions focusing on specific character strengths (i.e., three funny things or gratitude letter) have found immediate changes in states (i.e., momentary amusement or gratitude) after the intervention (Gander, Proyer, et al., 2020; Heekerens et al., 2022).

However, to our knowledge, no study has directly assessed whether character strengths were enacted more frequently following signature-strengths interventions¹ or whether the interventions yielded changes in the character strengths traits. Notably, a recent review of behavioral outcomes of strengths-based interventions (Bates-Krakoff et al., 2022) did not include any study that measured character strengths-related behavior (i.e., character strengths states) as an outcome.

The present research can also be situated within the broader field of personality change interventions (for overviews, see Hudson, 2021; Jackson et al., 2021). Building on principles identified in therapeutic change processes (Allemand & Flückiger, 2017), interventions to promote volitional personality change have also been designed and tested in non-clinical populations in recent years (e.g., Allan et al., 2018; Hudson et al., 2019; Stieger et al., 2021). These studies have yielded promising results, highlighting that it is not motivation to change that is the primary driver of personality change but rather the enactment of trait-related behaviors (Hudson et al., 2019). This finding underscores the role of personality states in facilitating longer-term changes in personality traits, which is also emphasized in theoretical accounts of volitional personality trait change (e.g., Magidson et al., 2014).

In this set of two randomized controlled intervention studies, our primary goal is to expand the understanding of the effects of signature strengths-based interventions in three ways. First, Study 1 systematically compares different variants of signature-strengths interventions to determine which variants are most effective in increasing well-being (and reducing ill-being) and which components are necessary for the interventions to be effective. Second, Study 2 tests the effectiveness of a multi-component program composed of several variants of strengths-based interventions. Third, both studies aim to address a crucial gap in the extant literature on character strengths-based interventions, that is, whether they result in actual changes in character strengths or broader personality traits (see Ruch et al., 2020). To achieve this, we use character strengths states, that is, the enactment of character in the past week, and character strengths traits as well as personality traits as outcomes in the present studies.

In summary, through these two studies, we seek to gain insights into the effects of signature strengths-based interventions, their most effective variants, and their potential to bring about changes in an individual’s character strengths.

Study 1 tested an expansion of the interventions outlined in the Strengths Builder program found in Niemiec and McGrath (2019), comparing seven variants of CSIs, namely, spotting strengths (IC1): learning about the VIA classification and recognizing strengths in other people without learning about one’s strengths; observing strengths (IC2): observing the use of one’s signature strengths in everyday life; using signature strengths in a new way (IC3): enacting one’s signature strengths in a novel manner; using signature strengths with minor challenge (IC4): using one’s signature strengths to overcome an everyday challenge; using signature strengths with larger challenge (IC5): using one’s signature strengths to overcome a major challenge; using signature strengths for others (IC6): using one’s signature strengths for the benefit of other people; forming a habit (IC7): trying to form a habit in using one’s signature strengths. These intervention conditions were compared with a placebo control condition (PCC) that received instructions to write about early memories. A more detailed description of the instructions is given in Table 1 (see supplementary materials for full instructions).

Using signature strengths in a new way (IC3) represented the original exercise suggested by Seligman et al. (2005). The idea of IC4, IC5, IC6, and IC7 was to test variants that might go along with increased effectiveness; they should provide participants with a focused area of application (IC4, IC5; see Niemiec, 2017) or aim at instructing people to use their strengths for prosocial purposes (IC6) since earlier studies suggested that interventions aimed at benefiting others (vs. oneself) go along with larger increases in well-being (Nelson et al., 2016). IC7 aimed at facilitating long-term changes by building habits for regular strengths practice, in line with ideas on personality change (Bleidorn et al., 2021; Hennecke et al., 2014). IC1 and IC2 represented variants assumed to be less effective than the original version (IC3) since they did not include behavioral instructions. IC2 aimed at examining whether a strengths-based intervention is also effective when the use of one’s strengths is only observed and documented (similar to the “identifying signature strengths” intervention reported in Seligman et al., 2005), while IC1 examined whether just learning about character strengths and applying this knowledge to other people goes along with positive effects (see Habenicht & Schutte, 2023, for the life domain of romantic relationships).

In line with earlier research, we examined the effects of these interventions on well-being and ill-being (i.e., perceived stress). We also considered the effects of the interventions on social well-being, character strengths traits, and character strengths states, that is, the frequency of displaying character strengths-relevant behaviors for one week.

We hypothesized that IC3, IC4, IC5, IC6, and IC7 would show larger increases in well-being and character strengths states, and larger decreases in stress compared to the placebo control condition. Further, we expected larger increases in social well-being in IC6 than in the placebo control condition. Effects on character strengths states and on character strengths traits were examined on an exploratory basis without formulating specific hypotheses.

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. This study was not preregistered. The R-code underlying all analyses, the intervention materials, and supplementary Tables and Figures are available from the project’s OSF page (https://osf.io/7n9fd/). The raw data can unfortunately not be made publicly available because the consent form used in this study excluded the disclosure of data to third parties.

We determined our sample size based on power calculations using an ANCOVA approach (i.e., predicting the posttest scores by the pretest scores and the condition) and used effect sizes from earlier studies (Gander et al., 2018; partial $η2$ = .02) when comparing an intervention condition with a placebo control condition. The power analyses (power = 80%, correlation between pre- and posttest = .80, one-tailed tests) suggested that 150 participants are required per condition. We aimed to collect a total sample of 1,200 participants who completed the intervention. Due to time constraints, we had to terminate recruitment shortly before reaching this threshold. Inclusion criteria were being at least 18 years old, currently not being in psychological or psychopharmacological treatment, and not using illegal drugs. Further, we excluded participants who did not complete the assigned exercise (in all conditions, including the placebo control condition).

A total of 2,791 participants registered, of whom 2,496 provided basic demographic information. Minors, participants currently in psychotherapeutic or psychopharmacological treatment, users of illegal drugs, and those who failed to complete the pretest were excluded (n = 600). A total of 1,896 participants were randomly assigned to the seven conditions. At posttest, N = 1,163 participants completed the measures and indicated that they conducted the assigned exercise. We used this sample for all our analyses (see Figure 1). Sensitivity analyses based on power simulations suggested that with this sample, we could detect differences among conditions of β ≥ .16 with a power of at least .80.

Participants were primarily women (82.3%; 17.2% men, 0.5% other), aged 18 to 78 years (median = 45, SD = 12.15). Most participants were German (70.0%), Swiss (23.0%), or Austrian (4.2%). About half of the sample (52.2%) held a degree from a university, 17.3% held a degree from a university of applied sciences, and 15.7% held a diploma allowing them to attend universities or universities of applied sciences; 13.7% completed vocational training, and 1.1% completed mandatory education.

The local ethics committee approved the study, and participants gave informed consent. The study was conducted online and advertised through university mailing lists and press releases. Data collection took place from May 2018 to April 2020. The study was advertised as a training program for character strengths based on scientific findings. Participants were unaware that there were various conditions. After providing informed consent, participants registered online, provided basic demographic information, completed the pretest, and were assigned to one of eight conditions (i.e., seven character strengths-based interventions and one placebo control condition; see Table 1) using an automated pseudo-random number algorithm. They then received detailed instructions and worksheets for their assigned exercises for seven days in a single file. At the immediate posttest and at the 2-, 4-, and 12-week follow-ups, participants were prompted via e-mail to revisit the website to complete the follow-up assessments. In addition, at the immediate posttest, participants indicated whether they had completed the assigned exercise. Those who did not were excluded from subsequent analyses. After the last follow-up, they received automatic feedback on their character strengths and well-being scores. There was no financial compensation for participation.

The Authentic Happiness Inventory (AHI; Seligman et al., 2005) measures the subjective assessment of overall well-being and covers both subjective and psychological well-being. It uses 24 sets of five statements (e.g., 1 = “I have sorrow in my life” to 5 = “My life is filled with joy”). Proyer et al. (2017) report good psychometric properties. The AHI was administered at pre-test, post-test, 2-, 4-, and 12-week follow-up. Internal consistencies were high at all time points (ω ≥ .95).

The Perceived Stress Scale (PSS; Cohen et al., 1983), used in the German adaptation by Büssing et al. (2013), assesses the subjective stress level during the last month using ten items ranging from 1 = “never” to 5 = “very often.” A sample item is “In the last week, how often have you felt nervous and ‘stressed’?”. We used an adapted version of the scale, asking for the frequency of symptoms during the last week (instead of the last month). The PSS was administered at all time points (i.e., at pre-test, post-test, and 2-, 4-, and 12-week follow-up). Internal consistencies were high at all time points (ω ≥ .90).

The Relationships Scale from the Comprehensive Inventory of Thriving (CIT; Su et al., 2014), used in the German adaptation by Hausler, Huber, et al. (2017), is an 18-item measure for the assessment of social well-being and encompasses aspects of perceived support, community, trust, respect, loneliness, and belongingness. All items use a 5-point scale ranging from 1 = “strongly disagree” to 5 = “strongly agree.” A sample item is “There are people I can depend on to help me.” The CIT was administered at all time points. Internal consistencies were high at all time points (ω ≥ .96).

The Character Strengths State Rating Form (CSSRF; Gander, Wagner, et al., 2022) is a 24-item measure for assessing the 24 character strengths states from the VIA classification using 1 item per strength. Participants indicate the frequency with which they have displayed strengths-related behaviors in a specific period. In the present study, participants were asked to refer to the past week. A sample item is “To be creative, to produce something useful, whether with unusual solutions or by designing something” (creativity). All items are answered on a 7-point scale ranging from 1 = “not at all” to 7 = “all the time.” Wagner and Gander (2024) report good convergence of averaged character strengths states with traits. The CSSRF was administered at all time points.

The VIA Inventory of Strengths (VIA-IS; Peterson et al., 2005), used in the German adaptation by Ruch et al. (2010), is a 240-item trait measure for the assessment of the 24 character strengths from the VIA classification using ten items per strength. All items are answered on a 5-point Likert-style scale ranging from 5 = “very much like me” to 1 = “very much unlike me.” A sample item is “I find the world a very interesting place” (curiosity). The VIA-IS was completed at pretest and the first follow-up. Internal consistencies were ω ≥ .74 across all scales and time points. Participants received feedback on their highest five character strengths (“signature strengths”) based on their results in the VIA-IS at pretest. Before computing the intraindividual rank order of character strengths, character strengths scores were scaled in relation to data from a large population sample (Ruch et al., 2010) in order to account for differences in (item-) difficulty among the character strengths (see Wagner & Ruch, 2023).

Participants completed the AHI, PSS, CIT, and CSSRF at all measurement time points: pretest, posttest, and follow-ups after 2, 4, and 12 weeks. Additionally, they completed the VIA-IS at pretest and the 2-week follow-up. Additionally, participants completed further measures not relevant to the purpose of the present study, including measures on personality (MRS-25; Schallberger & Venetz, 1999), humor (BENCOR; Ruch, 2012), and further positive traits (LTS; Kaufman et al., 2019).

Since each participant in IC3 to IC7 was instructed to train a different subset of character strengths, namely those five strengths for which they reported the highest pretest scores in the VIA-IS, we computed different indices for character strengths states and traits: One index averaging all the trained strengths traits (called “signature strengths”) and states (called “signature strengths states”) and one index for the remaining strengths (called “non-signature strengths” and “non-signature strengths states,” respectively). These indices were computed for all conditions, including those not instructed to train strengths (i.e., IC1, IC2, and PCC).

For the conditions, dummy variables were computed, contrasting the seven intervention conditions with the placebo control condition. All outcome variables (i.e., well-being, stress, social well-being, signature and non-signature strengths states and traits) were z-standardized based on the sample mean and standard deviations at pretest. Thus, the regression weights indicate the degree to which the intervention conditions differed from the placebo control condition after the intervention in the metric of the baseline standard deviation, analogous to the interpretation of Cohen’s d.

For the main analyses, we used longitudinal multilevel models. For model selection, we compared different models with increasing complexity (see supplementary Table S1). We estimated the models using maximum likelihood, and decisions on model comparisons were based on $χ2$- difference tests and Akaike information criteria. We started with an intercept-only model and then controlled for the baseline of the dependent variable and added a linear effect of time (posttest = 0, follow-up 1 = 2, follow-up 2 = 4, follow-up 3 = 12). A comparison with a model using a logarithmic effect of time (post-test = 0, follow-up 1 = 1.10, follow-up 2 = 1.61, follow-up 3 = 2.57; more substantial rates of change at the beginning than at the later time points) suggested a better fit to the data, and we, therefore, proceeded with this coding of time. We then added random slopes for time and tested whether an additional quadratic effect of time would yield a better fit, which it did not. Correction for autocorrelation did yield a better fit for some dependent variables. Therefore, we implemented this correction for all models. Finally, we introduced the dummy-coded condition (intervention conditions vs. placebo control condition) as a predictor. We also tested whether a differential slope among the conditions better fits the data, which was not the case. Thus, our final models for the main analyses predicted the dependent variable by the pretest scores, logarithmic time, and the conditions while allowing for random intercepts and slopes and correcting for autocorrelation.

We computed ordinary linear regressions to study changes in character strengths traits that were only assessed at two time points, predicting the scores at the 2-week follow-up by the pretest scores and the dummy coded conditions.

For all analyses, we do not report p-values but instead report 95% confidence intervals for the estimated parameters. We used the guidelines by Funder and Ozer (2019) for interpreting effect sizes; that is, Cohen’s d ≥ .10 (corresponds to Cohen’s ω ≥ .05; η² ≥ .003) denotes a very small effect, Cohen’s d ≥ .20 (Cohen’s ω ≥ .10; η² ≥ .01) a small, Cohen’s d ≥ .41 (Cohen’s ω ≥ .20; η² ≥ .04) a medium, and Cohen’s d ≥ .63 (Cohen’s ω ≥ .30; η² ≥ .09) a large effect.

We used R (Version 4.3.0; R Core Team, 2023) and the R-packages tidyverse (Version 2.0.0; Wickham et al., 2019), effectsize (Version 0.8.3; Ben-Shachar et al., 2020), nlme (Version 3.1.162; Pinheiro et al., 2023), papaja (Version 0.1.1; Aust & Barth, 2022), parameters (Version 0.21.0; Lüdecke et al., 2023), paramtest (Version 0.1.0; Hughes, 2017), psych (Version 2.3.3; Revelle, 2023), pipeR (Version 0.6.1.3; Ren, 2016), lavaan (Version 0.6.15; Rosseel et al., 2023), and MBESS (Version 4.9.2; Kelley, 2022), for all our analyses.

First, we compared those who completed the posttest and indicated that they completed the assigned exercises (N = 1,163) with those who did not (n = 733). These analyses suggested that those who dropped out were, on average, four years younger and more likely to be men. Further, dropouts reported lower well-being and social well-being and higher levels of stress at baseline but did not differ from completers regarding character strengths states (see supplementary Table S2).

Next, we analyzed whether there were differences among the conditions in those who completed the exercise. The conditions did not differ regarding age, gender, or dependent variables at baseline (see supplementary Table S2).

Means and standard deviations of all conditions at all time points are given in Figure 2 and supplementary Table S3. Multilevel model results are given in Tables 2 and 3. Descriptive results (Figure 2) suggested that all conditions reported an increase (or decrease in stress, respectively) after the intervention until the last measurement point 12 weeks later. Over time, however, the increases became less pronounced.

Analyses suggested that well-being increased in all conditions from the posttest to the last follow-up (see Table 2). The conditions IC3 and IC7 reported higher levels of well-being after the intervention than the control condition (PCC). Stress decreased after the intervention but remained unchanged from the posttest to the last follow-up. There were no differences after the intervention between any of the intervention conditions and the PCC. Social well-being increased from the posttest to the last follow-up. The condition IC3 reported higher levels of social well-being after the intervention than the PCC.

Both signature and non-signature character strengths states increased from the posttest to the last follow-up (see Table 3). The conditions IC3 reported higher scores in signature strengths states, and IC3 and IC7 reported higher levels in non-signature strengths states after the intervention than the PCC. For character strengths traits, IC3 and IC4 reported higher scores in signature strengths traits after the intervention, while no differences in non-signature strengths were observed among the conditions.

Study 1 showed that the “using signature strengths in a new way”-intervention (IC3) is effective for increasing well-being and social well-being. Further, the newly developed intervention focusing on “forming a habit” (IC7) also led to increases in well-being. Both interventions also led to increases in signature (IC3) and non-signature (IC3 and IC7) character strengths states. At the same time, IC3 (in addition to IC4: using signature strengths with a minor challenge) also led to increases in signature strengths traits. Effect sizes ranged from very small to small and were on the lower end or below the effect sizes we could detect reliably, given our sample sizes.

Thus, the study confirmed earlier findings on the effectiveness of the “using signature strengths in a new way”-intervention for increasing well-being (e.g., Schutte & Malouff, 2019) and extended its effects on social well-being. Although not directly targeted in this intervention, aspects such as loneliness, belonging, and perceived social support and respect were also affected after using their signature strengths in a new way. We are unaware of other studies that examined the effects of this intervention on specific domains of well-being. Future studies might assess potential effects on multiple dimensions of well-being (or flourishing, e.g., Su et al., 2014) to examine whether such interventions lead to general, global increases in well-being that affect all domains or whether only specific domains of well-being are affected. It might also be relevant to investigate the impact of strengths-based interventions on well-being in different life domains, as recently demonstrated for personality change interventions (Olaru et al., 2023), given that character strengths have been found to be relevant across different life domains (Wagner et al., 2021).

Further, the results confirmed the effectiveness (in terms of well-being increases) of a newly developed intervention aimed at forming signature-strengths based habits. We find this intervention particularly interesting because it is the only intervention tested that explicitly aims to implement longer-term changes. From an applied view, positive psychology interventions have been criticized for focusing on short-term, “one-off” interventions (van Woerkom, 2021); exercises aiming to build habits might increase the chances of attaining sustainable, long-term changes. Indeed, we did not observe a decline in well-being at the later time points; however, the same applied to all other tested interventions, including the placebo control condition. Thus, further research using more and larger-spaced follow-ups is required to examine whether indeed more sustainable effects can be achieved with this intervention.

No effects on ill-being (i.e., stress) were observed for all interventions. This finding is somewhat in line with earlier research showing that the effects of positive psychology interventions on well-being are more robust than on ill-being (e.g., White et al., 2019). Also, unexpectedly, all other tested new variants of character strengths-based interventions did not show any positive effects on the (well-being and ill-being) outcomes. Thus, to foster one’s well-being, it is not enough to learn about character strengths and think of the positive qualities of other people in everyday life (IC1). Neither does merely observing one’s display of character strengths (IC2) have beneficial effects. Also, using one’s signature strengths for dealing with minor (IC4) or major challenges (IC5) or for the benefit of other people (IC6) did not go along with any changes in well-being. It is possible that trying to use one’s strengths for dealing with minor or major challenges goes along with an increased focus on these potentially distressing challenges, which might have prevented any positive effects on well- or ill-being. Also, we did not assess whether participants considered the exercise helpful for coping with these challenges — it is also possible that the exercise helped in dealing with these challenges without affecting one’s global well- or ill-being. However, it remains unclear why the intervention aimed at using one’s strengths to benefit other people did not show any positive effects in this study. This finding disagrees with similar studies that compared the effects of performing kind acts for oneself versus for others and reported stronger increases for those who engaged in prosocial behaviors (Nelson et al., 2016).

In Study 2, we built on the findings obtained in Study 1 and tested a multi-component strengths intervention program (Niemiec & McGrath, 2019) that lasted four weeks. The program included components that help participants recognize and apply strengths and a component that helps them build a habit of doing so. In Study 1, IC3 and IC7 were the most effective interventions overall, so they were presented in Weeks 3 and 4, preceded by the components IC1 and IC2. We considered IC1 and IC2 to be helpful initial steps as they could help increase the awareness of character strengths and the reflection on how they are enacted in daily life. Reflective processes are thought to be key mechanisms that drive enduring changes in personality traits (e.g., Baumert et al., 2017), so encouraging participants’ self-reflection before instructing them to change their behavior might be beneficial. Table 4 shows the contents of the four intervention components.

The four-week program was developed in an effort to adapt existing interventions to potentially increase the effect in a way that it might be more likely “to result in downstream improvements in psychological functioning and health” (Kubzansky et al., 2023, p. 174), as was recently recommended by a group of researchers. Findings regarding the effects of intervention duration on the outcomes have been mixed to date. On the one hand, a recent meta-analysis (Carr et al., 2021) found no differences between shorter and longer (≥ eight sessions) positive interventions for most outcomes (except depressive symptoms, which improved more in longer interventions), and an experimental study even reported that extending the duration of an intervention to two weeks produced smaller effects than the one-week version (Gander et al., 2013). On the other hand, Roberts et al. (2017) reported a nonlinear relationship between the duration of a therapeutic intervention and its effect on personality traits. They found the effectiveness to increase with intervention length, but only up to eight weeks.

In the present study, we were also interested in whether the intervention yielded changes not only in character strengths states and traits but also in broader personality traits as described by the five-factor model of personality. Given the overlap between character strengths and the dimensions of the five-factor model of personality (McGrath et al., 2020; Ruch et al., 2023), it seems plausible that an intervention program based on signature strengths would also result in changes in broader personality traits. We did not, however, formulate any specific expectations about the directions in which the intervention might lead to changes in personality traits.

Further, we built on previous studies that showed that beliefs about the malleability of well-being might moderate the effects of positive psychology interventions (Gander, Proyer, et al., 2022). In the present study, we also studied the role of beliefs about the malleability of personality and aimed to explore whether the intervention would yield changes in those beliefs caused by the engagement with one’s personality traits and well-being.

The study’s overall goal was to compare the effects of a four-component signature strengths-based positive psychology intervention on well-being and stress with those of a waitlist condition. We preregistered (https://aspredicted.org/w48ks.pdf) the following hypotheses: We expected that the program would be well-received by the participants; i.e., they would report positive subjective ratings of acceptance and subjective benefits from the intervention. In comparison to the waiting list control condition, we expected that participants in the intervention condition would report (a) larger increases in well-being, (b) larger decreases in stress, (c) larger increases in character strengths traits, (d) larger increases in character strengths states, (e) larger increases in beliefs about the malleability of well-being and personality, and (f) larger changes in the five-factor model dimensions of personality. We also expected that the improvements of the intervention group after the intervention would be stable until the follow-up².

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. Before data collection, we preregistered the hypotheses and the analysis plan³ (https://aspredicted.org/w48ks.pdf). The R code underlying all analyses, the intervention materials, correlations among all study variables, and supplementary Tables and Figures are available from the project’s OSF page (https://osf.io/7n9fd/). The raw data can unfortunately not be made publicly available because the consent form used in this study excluded the disclosure of data to third parties.

Based on pre-registered a priori power calculations, we initially determined that sample sizes of at least 100 participants who completed the entire program, including follow-up, per condition would be required to detect small effects with a power of 0.80 or greater. However, considering the likelihood of substantial dropout rates, our objective was to have sample sizes of at least 200 for both conditions. Ultimately, we could not reach the targeted sample sizes and conducted the analysis using the actual sample size we obtained. Inclusion criteria were being at least 18 years old, currently not being in psychological or psychopharmacological treatment, and not using illegal drugs.

A total of 1,162 participants registered, of whom 795 provided basic demographic information. Participants in psychotherapeutic or psychopharmacological treatment, users of illegal drugs, and those who failed to complete the pretest were excluded (n = 278). We randomly assigned 523 participants to the two conditions. A total of N = 254 participants completed the measures at week 4 (posttest for the intervention condition) or week 8 (follow-up for the intervention condition). We analyzed all participants who completed the measures at week 4 or week 8 (see Figure 3). Sensitivity analyses based on power simulations suggested that given this sample size, we could detect differences between the conditions at week 4 or 8 of β ≥ .22 and differences from week 4 to week 8 of β ≥ .16 with a power of at least .80. However, moderation effects (i.e., the interaction between condition and the moderator) would have required effect sizes of at least β ≥ .24 to be detected with a power of ≥ .80. Because we did not expect to obtain moderation effects of this magnitude, we decided to deviate from preregistration and not analyze moderation effects.

Participants were mostly women (77.8%; 20.5% men, 1.5% other) and were aged 19 to 87 (median = 42, SD = 13.37). The participants lived all over the world but were predominantly from North America (37.5%), Western Europe (35.6%), Pacific Asia (10.9%), Eastern Europe (5.0%), or Southern Asia (4.6%). Approximately one-third of the sample (32.9%) possessed a Bachelor’s degree, while 39.2% held a Master’s degree, and 10.5% had obtained a Doctorate. Additionally, 17.4% of the participants reported having other educational levels.

According to the local ethics committee guidelines, this study did not require ethical review, and participants gave informed consent. The study was conducted online and advertised through university mailing lists, social media, and the VIA Institute on Character website. Data collection took place from February 2019 to May 2023. The study was advertised as a training program for character strengths based on scientific findings. Participants were informed that they would be assigned to the intervention condition or the waiting list control condition. After providing informed consent, participants registered online, provided basic demographic information, completed the baseline measures at week 0, and were randomly assigned to the two conditions using an automated pseudo-random number algorithm. The intervention condition received detailed instructions and worksheets for the 4-week intervention in a single file. The waiting list condition received the intervention materials after waiting for four weeks. After four weeks (posttest for the intervention condition) and eight weeks (follow-up for the intervention condition, posttest for the waiting list condition), participants were prompted via e-mail to revisit the website to complete further assessments (see Figure 4). After the assessments at week 8, they received automated feedback on their character strengths and well-being scores. There was no financial compensation for participation.

As in Study 1, we used the Authentic Happiness Inventory (AHI; Seligman et al., 2005) to assess overall well-being. The AHI was completed at all time points (i.e., week 0, 4, and 8). Internal consistencies were high at all time points (ω >= .97).

The Perceived Stress Scale (PSS; Cohen et al., 1983) assesses subjective stress and was administered at all time points. Internal consistencies were high at all time points (ω ≥ .91).

The Character Strengths State Rating Form (CSSRF; Gander, Wagner, et al., 2022) assesses the 24 character strengths states from the VIA classification and was administered at all time points. Internal consistencies were high at all time points (ω ≥ .94).

The VIA Inventory of Strengths (VIA-IS; Peterson et al., 2005) assesses the 24 character strengths traits from the VIA classification and was administered at all time points. Internal consistencies were ω ≥ .72 across all scales and time points.

Additionally, the short form of the Big Five Inventory–2 (BFI-2-S; Soto & John, 2017) assesses the dimensions of the five-factor model of personality with 30 items (6 items per dimension). The items use a 5-point scale ranging from 1 = “strongly disagree” to 5 = “strongly agree.” A sample item is “I am someone who tends to be quiet” (Extraversion). The BFI-2-S was administered at all time points. Internal consistencies were ω ≥ .73 across all scales and time points.

Further, we assessed two kinds of beliefs about the malleability of well-being and personality: Implicit beliefs about personality (IBP) and implicit beliefs about well-being (IBWB), assessing the degree to which people believe that personality, or well-being, is fixed, as opposed to being malleable. For the assessment of IBP, we used three items suggested by Dweck et al. (1995). A sample item is “Everyone is a certain kind of person, and there is not much that can be done to really change that.” For the assessment of IBWB, we adapted three items suggested by Dweck et al. (1995) for the assessment of implicit beliefs about morality to well-being: “A person’s satisfaction with life is something very basic about them, and it cannot be changed much,” “Whether a person is happy or not is deeply ingrained in their personality. It cannot be changed very much.”, “There is not much that can be done to change a person’s level of being satisfied with life.”. All items use a 6-point scale ranging from 1 = “strongly agree” to 6 = “strongly disagree.” Both scales were administered at all time points. Internal consistencies were ω ≥ .84 for both scales across all time points.

Finally, we assessed acceptance of the intervention (“Which answer best describes how much you liked the training?”) on a 7-point scale (1 = “disliked extremely,” 2 = “disliked,” 3 = “disliked a little,” 4 = “neutral,” 5 = “liked a little,” 6 = “liked,” 7 = “liked extremely”) and subjective benefit of the intervention (“Did you benefit from the training?”) on a 5-point scale (1 = “not at all,” 2 = “slightly,” 3 = “moderately,” 4 = “very,” 5 = “extremely”).

We again computed indices of signature and non-signature strengths traits and states based on the scores at week 0 in the VIA-IS. The condition was dummy coded (0 = control condition, 1 = intervention condition). We also created a dummy for the time points after the intervention (0 = week 4, 1 = week 8). All outcome variables (i.e., well-being, stress, signature and non-signature strengths traits and states, five-factor-model personality traits, and beliefs about the malleability of well-being and personality) were z-standardized based on the pretest sample means and standard deviations.

We used multilevel models for the main analyses, predicting the scores of the outcomes by the pretest scores, the condition, time, and the condition × time interaction. Thereby, the variable for condition indicates whether the intervention differed from the control condition at week 4, the time variable indicates whether the control condition increased from week 4 to week 8. The condition × time interaction indicates whether the changes in the intervention condition differed from the changes in the control condition from week 4 to the week 8.

As in Study 1, we report 95% confidence intervals instead of p-values and interpret effect sizes according to the guidelines suggested by Funder and Ozer (2019). We used R (Version 4.3.0; R Core Team, 2023) and the same packages as in Study 1 for the analyses.

Analysis of Dropouts. Those who completed the measures at weeks 4 or 8 (n = 254) and those who failed to do so (n = 269) did not differ in terms of demographics or baseline scores of dependent variables. However, dropout was considerably higher in the intervention condition than in the waitlist condition (see supplementary Table S4).

Analysis of Comparability of Conditions at Baseline. When comparing the characteristics of the conditions at baseline, no relevant differences were observed regarding demographics or dependent variables (see supplementary Table S4).

Acceptance and Subjective Benefit. Participants in the intervention condition indicated liking the intervention program overall: 72% indicated a response higher than “neutral,” and only 8% indicated not liking the program. The average of the 7-point scale was M = 5.27 (SD = 1.21). Regarding subjective benefits, 8% said they did not benefit from the intervention, 43% described their benefit as “moderate,” while 32% even felt their benefit was either “very” or “extremely” strong. The average of the 5-point scale was M = 3.08 (SD = 1.00).

Means and standard deviations of all conditions at all time points are given in Figure 5 and Supplementary Table S5. The results of multilevel models are given in Table 5. Results suggested that the intervention condition (IC) reported higher levels of well-being at week 4 than the waitlist control condition (CC). Further, well-being increased in the CC from week 4 to week 8 (when they received the intervention), and this increase was higher than the changes in the IC (who already completed the intervention at week 4) from week 4 to week 8. Nonetheless, the IC maintained their well-being levels throughout week 8. The same pattern was obtained for stress: At week 4, the IC reported lower scores than the CC, and the reduced stress levels were maintained at week 8. In the CC, stress decreased when they received the intervention from week 4 to week 8, and these changes were stronger than the changes in the IC during this period. Regarding beliefs about the malleability of well-being and personality, no changes were observed in any of the conditions.

Both signature and non-signature character strength states increased when the conditions received the interventions: In the IC from week 0 to week 4 and in the CC from week 4 to week 8. For non-signature strengths states, the changes in the CC from week 4 to week 8 exceeded those in the IC. For character strengths traits, no effects of condition were observed, but the CC reported an increase in signature and non-signature strengths when they received the intervention. Participants in the IC reported lower levels of neuroticism than the CC at week 4, while the scores of the CC decreased from week 4 to week 8. For extraversion, the CC reported an increase from week 4 to week 8, which was larger than the changes in the IC during this period. For openness, an increase in the CC from week 4 to week 8 was observed, but no differences between the conditions. No changes or differences were reported in agreeableness or conscientiousness.

Study 2 showed that a four-week multi-component strengths intervention effectively increases well-being and reduces stress. These effects could be observed during the periods in which the conditions received the intervention: from week 0 to week 4 for the intervention condition, and from week 4 to week 8 for the waitlist control condition. Further, in the intervention condition, the initial increases were maintained at week 8. Effect sizes of the changes due to the intervention were of small to medium size.

Further, the intervention also led to increases in both signature and non-signature–character strengths states of small to medium size, and these effects also persisted up to week 8. For character strengths traits, the pattern was less clear: While participants reported increases of small size in signature and non-signature strengths traits from week 4 to week 8, there were no differences between conditions, and the changes, therefore, cannot be attributed to the intervention.

However, the results indicated that the intervention also affected five-factor model traits, predominantly neuroticism: Both conditions reported decreases after receiving the intervention, and the intervention condition retained these lower scores at week 8. Further, we noted some increases in extraversion and openness, which occurred in the control condition after participants had completed the intervention. Also, all changes in the five-factor model personality traits were of very small to small size and should, therefore, be interpreted with caution.

Beliefs regarding the malleability of personality and well-being were widely unaffected by the intervention, and we, therefore, conclude that our strengths-based intervention was ineffective in fostering the belief that well-being or personality could be changed. One possible explanation is ceiling effects: Participants reported already relatively strong beliefs in the malleability of well-being (M = 5.0 on a scale ranging from 1 to 6) and personality (M = 4.3 on a scale ranging from 1 to 6) before the intervention. Additionally, in the present study, we assessed general change beliefs, that is, whether people believe that well-being and personality are modifiable in general. We did not assess instrumental change beliefs, namely whether participants think their well-being can be influenced and whether they know how to change their well-being. In a previous study, the latter was mainly relevant for intervention effectiveness (Gander, Proyer, et al., 2022). Future studies might also assess participants’ beliefs about the malleability of the specific trait targeted in an intervention since this might differ from general beliefs about the malleability of personality. Finally, while the intervention mentioned that character strengths are expected to be modifiable, it did not emphasize changing participants’ beliefs. Thus, future interventions should be informed more strongly by earlier successful studies focusing on changing mindsets (e.g., Yeager et al., 2016).

The present set of two studies examined the effects of seven variants of character strengths-based interventions versus a placebo control condition in Study 1 and a 4-week multi-component character strengths-based intervention versus a waitlist control condition in Study 2. A broad array of outcomes was considered, including (subjective, psychological, and social) well-being, ill-being (stress), (character strengths) states, and (character strengths and five-factor model) traits. The overarching goals of these studies were examining (I) what aspects of character strengths-based interventions are effective for fostering well-being, (II) whether these interventions also affect character and personality states and traits, and (III) whether a longer multi-component intervention consisting of several variants of strengths-based interventions is more effective than single interventions.

Regarding (I), the findings of Study 1 confirmed the assumption that in signature strengths-based interventions, the novelty aspect is crucial: As in the original study by Seligman et al. (2005), the interventions that merely instructed participants to display more of an already established behavior (i.e., IC4, IC5, and IC6) did not yield any positive effects unless the intervention also aimed at fostering a habit (IC7). Thus, the positive effect of strengths-based interventions can neither be explained by an increase in the focus on positive aspects (as in IC1 and IC2) nor by just displaying one’s signature strengths more often – only when strengths are used in a novel way (IC3) or are deliberately practiced (IC7), a positive change is achieved. At the same time, this finding that merely displaying one’s signature strengths more often does not lead to higher well-being disagrees with earlier observational studies that reported higher well-being when someone displayed a character strength (regardless of whether it is typical of the person or not) more often than usual (Wagner & Gander, 2024). One possible explanation is that it might make a difference whether the increased display of a strength is instructed (as in the present study) or whether it naturally occurs (as in observational studies), in which case the behavior may also be demanded by situational characteristics (e.g., Fleeson et al., 2002). Also, while this study considered character strengths displays during the past week, Wagner and Gander (2024) assessed daily displays of character strengths. Thus, it is possible that displaying strengths more frequently than usual is only beneficial in the short term, while in the longer run, the additional efforts take a toll and equalize the initial benefits, as has been suggested for other personality traits (Gallagher et al., 2011). Overall, more research on situational displays of character strengths is urgently needed.

The present studies had a more applied focus and aimed to evaluate the effectiveness of different components that may be used in a multi-component intervention; thus, they were not designed to test the working mechanisms of the interventions directly. However, in future studies, a more fine-grained analysis of the active ingredients of the “using signature strengths in a new way” intervention (IC3) would also be of interest. Is novelty not only a crucial aspect of the intervention but even more important than displaying character strengths? A comparison to a purely novelty-based intervention not related to character strengths (“doing something new”) would be necessary to answer this question. This approach was previously used to test the effect of novelty in a kindness intervention (Buchanan & Bardi, 2010). The study, with 86 participants across three conditions (two interventions and an inactive control), found no significant differences between “acts of kindness” and “acts of novelty” interventions. However, the small sample size may have limited the study’s power to detect meaningful differences. Notably, “acts of kindness” showed descriptively larger gains in life satisfaction compared to “novelty” (d = .21). Thus, it seems likely that at least some – but not all – of the effectiveness of the “using signature strengths in a new way” stems from the effects of novelty (or novelty and variety), which has been suggested as a candidate basic psychological need (Bagheri & Milyavskaya, 2020; González-Cutre et al., 2016) and is also used in interventions in other domains as a catalyst for change (e.g., Fissler et al., 2013).

Regarding (II), the presented studies further confirmed the assumption that character strengths-based interventions indeed foster character strengths, which has rarely been tested directly (except by Proyer et al., 2013). Both studies suggested that character strengths states and traits increased following the intervention, although the results were more robust for states than for traits, which aligns with expectations. Character strengths states, that is, the display of strengths-related behaviors, should be affected immediately and strongly by the intervention, while character strengths traits, which also include aspects of self-identity (e.g., seeing oneself as a creative person) in addition to strengths-related behaviors, should require more time and effort to be affected by interventions. Interestingly, signature character strengths traits and states (i.e., those strengths targeted in the intervention conditions) and non-signature states and traits (i.e., those not targeted) were more or less similarly affected by the intervention. This finding might suggest that strengths-based interventions do not necessarily target only specific strengths but generally improve psychological functioning and self-efficacy (see van Woerkom et al., 2016), which results in higher scores for states and traits, but also well-being.

Further, this finding is also not surprising given the robust intercorrelations among character strengths: In the present study, the highest correlation observed for each character strength (trait) with any of the other character strengths was, on average, r = .58. Thus, it is very likely that targeting one character strength (e.g., love of learning) also leads to changes in other character strengths (e.g., curiosity) due to their close association. Another possible explanation is that people often seek to improve the character strengths they perceive as lacking (Gander & Wagner, 2023). Therefore, they might be more motivated to change non-signature strengths than signature strengths. Consequently, participants might not strictly adhere to the intervention’s focus on signature strengths but also work on improving non-signature strengths. Additionally, the possibility of methodological artifacts, such as ceiling effects or an overrepresentation of harder-to-train strengths among the signature strengths, also has to be considered. However, this seems unlikely in the present study as very few participants reached the scale maxima for both character strengths states and traits (approximately 2% and 0%, respectively). Additionally, although some strengths (e.g., love of learning, prudence, appreciation of beauty) were more frequently identified as signature strengths compared to others (e.g., hope, kindness, humor), the distribution across different character strengths and conditions was relatively even (see supplementary Table 6). Hence, it is unlikely that these methodological artifacts strongly biased the results.

Finally, the results from the multi-component intervention in Study 2 suggested that strengths-based interventions might influence not only well-being and character strengths but also five-factor model dimensions. The most consistent findings were observed for neuroticism, which decreased in both conditions at the expected time points. However, caution is advised in interpreting this preliminary finding due to the absence of an active control condition in Study 2. Furthermore, our study design may not fully capture the nuanced effects of the interventions. Since each participant had a unique set of signature strengths, the impact on five-factor model dimensions like extraversion could vary significantly. For instance, strengths linked to high extraversion (like zest or humor) may produce different results compared to strengths less or negatively related to extraversion (e.g., fairness, humility). This variability suggests that the overall effects on traits like extraversion could be diluted. Therefore, while these preliminary findings are promising, they underscore the need for further research into the effects of positive psychology interventions on personality dimensions. It is conceivable that engaging in a different way with traits that are already regularly displayed might offer a more accessible approach to reducing neuroticism compared to adopting behaviors that are not yet frequently demonstrated.

Regarding (III), the 4-week multi-component intervention in Study 2 showed more substantial increases than the single interventions in Study 1, suggesting that a combination of several interventions in a comprehensive program could improve the effects of the interventions. Given that single interventions usually go along with small effect sizes, this is an important contribution to practical applications. While past findings regarding the effects of intervention duration on the outcomes have been mixed (e.g., Carr et al., 2021; Gander et al., 2013), our findings suggest that a meaningful combination of intervention components might indeed be more effective than single interventions. However, we hasten to emphasize that direct comparisons between Study 1 and Study 2 are limited by different study designs: Study 1 used an active placebo control condition, while Study 2 used a waitlist control condition. It has repeatedly been reported (e.g., Carr et al., 2021) that using active control conditions represents a much more stringent design, which also accounts, among others, for expectancy effects but usually goes along with smaller effect sizes than inactive controls, such as waitlist conditions.

Our dropout analyses revealed that in Study 1, those who completed the assigned exercise (and were analyzed) were happier, less stressed, and reported higher levels of social well-being, signature strengths states and traits, as well as non-signature strengths traits. Also, they were better educated, older, and more often women than those who failed to do so. In Study 2, while the dropouts did not differ from the completers regarding demographic or dependent variables at baseline, there was a differential dropout between conditions: Those who dropped out were more frequently in the intervention condition. Presumably, this is because those in the control condition were yet to receive the intervention manual and, therefore, were more inclined to adhere to the protocol than those in the intervention condition, who already received the manual and were less motivated to complete the posttest. Besides these differences, there were no differences among the conditions for the completers, and the results regarding the effectiveness of the interventions are therefore not biased. At the same time, the differences between completers and dropouts limit the generalizability of the findings. Therefore, our results only apply to those who completed an intervention.

While the present study presents initial evidence that strengths-based interventions might not only affect well-being but also be effective strategies for fostering character and personality traits, we want to add that we do not consider changes in self-reports following an intervention as a confirmation of “true” personality change. Since participants were instructed to practice character strengths, increases in character strengths states and traits may be partly attributed to demand characteristics or to halo effects due to increased well-being after the intervention. Other data sources, such as informant reports, would thus be required to settle this question. Nonetheless, we tentatively conclude that strengths-based exercises that include using signature strengths in a new way or forming strengths-based habits are promising candidates for deliberate character and personality change.

When examining the effects of the interventions on character strengths states and traits, we computed indices for signature and non-signature strengths by averaging the trained- and un-trained strengths. Of course, it is possible that not all character strengths are equally malleable (e.g., Gander, Hofmann, et al., 2020) and that changes in some strengths drive the effects. In addition, character strengths differ in their relationships with well-being (e.g., Wagner et al., 2020) as well as with personality traits (e.g., Ruch et al., 2023). Thus, comparing the effects on specific strengths would also be highly interesting. However, the design of our study and the limited sample size preclude detailed scrutiny of individual (trained) character strengths. Future research, ideally employing designs tailored to compare the malleability of specific strengths (e.g., focusing on one particular strength instead of a set of five, as in our study), could shed more light on this important question. Another related question for future research looking at the effects of interventions targeted at specific character strengths would be studying correlated change between character strengths and well-being (and ill-being) in the context of interventions. In the context of volitional personality change in the five-factor model traits, Olaru et al. (2023) reported correlated changes in personality traits and satisfaction in different life domains. Given the differential associations of character strengths with well-being dimensions (Wagner et al., 2020) and life domains (Wagner et al., 2021), we would also expect such correlated change in character strengths-based interventions. An important consideration for future studies on personalized character strengths-based interventions is allowing participants to select the character strength(s) they want to focus on during the intervention. Given that a number of theoretical approaches highlight the role of motivation for change (e.g., Hennecke et al., 2014), targeting those traits that individuals desire to change might be a fruitful approach to increase motivation. We know that people’s desire to change character strengths tends to be driven by what they perceive as lacking and the character strength’s contribution to well-being (Gander & Wagner, 2023). Therefore, instructing them to focus on signature strengths when they do not want to increase them might harm some people’s motivation. Future studies could also address this question by assessing motivation for change directly.

Our findings (1) further corroborate the effectiveness of the “using signature strengths in a new way” - intervention for increasing (subjective, psychological, and social) well-being, (2) confirm the effectiveness of a newly developed intervention focused on forming a strengths-based habit, and (3) provide first evidence that character strengths-based interventions might not only affect well-being, but also foster character strengths states and traits, and might also be effective for changing broader personality traits, such as neuroticism. While further research is necessary, primarily forming a strengths-based habit might be a promising approach for the deliberate training of character strengths.

Fabian Gander: Conceptualization, Investigation, Data Curation, Formal analysis, Writing - Original Draft, Writing – Review and Editing, Visualization.

Lisa Wagner: Conceptualization, Writing - Original Draft, Writing – Review and Editing.

Ryan M. Niemiec: Conceptualization, Investigation, Writing – Review and Editing.

We are grateful to Damaris Burri McColgan for her contribution to the development, translation, and implementation of the interventions and to Mara Stewart for checking the translations of the website and the interventions.

The preparation of this manuscript was supported by the VIA Institute on Character and a research grant from the Swiss National Science Foundation (100014_172723).

This study was supported by the VIA Institute on Character, which holds the copyright to the VIA Inventory of Strengths and employs one of the authors (RMN).

Supplemental Tables are available from https://osf.io/7n9fd/

The data supporting this study’s findings are available on request from the corresponding author. Unfortunately, the data cannot be made publicly available as the consent form excludes data sharing with third parties. The R code for the statistical analyses can be viewed at https://osf.io/7n9fd/.