1 Introduction

The World Health Organization (2021) defines well-being as a positive state and resource for everyday life, encompassing quality of life and the ability to contribute to society with a sense of purpose and meaning. Interventions designed to increase well-being, also called positive psychology interventions (PPI, Seligman & Csikszentmihalyi, 2000), have emerged to complement psychological interventions designed to reduce psychopathology (for reviews, see Bolier et al., 2013; Carr et al., 2021; Sin & Lyubomirsky, 2009). Setting valued goals (Locke & Latham, 1990, 2002), developing optimism (Malouff & Schutte, 2017), and imagining one’s best future self (Sheldon & Lyubomirsky, 2006) are examples for effective future-oriented PPIs that address well-being by either increasing approach motivation or by reducing avoidance behavior or motivation. Not striving towards meaningful goals, on the other hand, can be a detriment to well-being (Steel, 2007).

According to self-determination theory (SDT, Deci & Ryan, 2008), social contexts that promote a sense of competence, autonomy, and belonging lead to a high degree of intrinsic motivation, which in turn relates to higher performance, learning, and well-being. In the context of goals, research indicates that people who focus on intrinsic goals have higher well-being than people who emphasize extrinsic goals such as accumulating wealth, gaining social status, or experiencing prestige (Kasser & Ryan, 1996; Sheldon et al., 2004). Addressing cognitions related to perceived obstacles while striving towards meaningful goals, may therefore be a valuable strategy for improving well-being.

We investigate a brief writing intervention called fear-setting. Fear-setting focuses on the visualization of perceived or imagined negative consequences (fears) related to making progress in meaningful goals or initiating personal changes (for a description of an early Stoic exercise based on this idea, see Irvine, 2009). The intervention attempts to change specific cognitions related to goal attainment motivation through a deliberate assessment and reappraisal of perceived threats (Beck & Haigh, 2014). Defining and reappraising fears should help reduce key cognitions and self-defeating beliefs (e.g., “I can’t do it” or “I am a failure”), which promote cognitive-affective states that in turn prevent individuals from taking action. Self-defeating beliefs are often linked to impaired motivation (Powers et al., 2007; Steel & König, 2006), lowered self-efficacy (Schwarzer & Jerusalem, 1995), or optimism (Schwarzer & Knoll, 2003) to strive towards personally meaningful goals. Other reasons for inaction may be fear of failure (Conroy et al., 2002) including fear of shame and embarrassment or losing self-esteem (Burka & Yuen, 2007). The intervention also involves risk management and planning through pre-empting what to do in the face of challenges and therefore changing the debilitating cognitions (“I’m afraid x might happen, but if I do y, it becomes less likely”, or “If this happens, I will do x or ask y for support”; e.g., Schwarzer & Knoll, 2003). Furthermore, through imagining the benefits of partial success and through reflecting on the costs of waiting, fear-setting reframes the situation and makes rewards of trying and costs of waiting more salient (O’Donoghue & Rabin, 1999). Finally, fear-setting likely improves affective states because reappraising fears addresses both the perceived probability of success or a good future (related to positive affect, Schubert et al., 2020) as well as the cost of striving towards a goal such as fear of failure (an indicator of negative affect, Sagar & Stoeber, 2009).

1.1 Fear-Setting Writing Intervention

Table 1 presents the adaption of fear-setting in the current study (Irvine, 2009). Through six steps, participants choose a goal, define and rate the likelihood of their fears related to that goal, imagine strategies to prevent fears from becoming true and imagine ways to repair the damage if fears do come true. In the last steps, fear-setting involves reflecting on the benefits of trying or a partial success and on the costs of maintaining the status quo or passivity. Fear-setting goes beyond the Stoic version of negative visualization as it is a combination of exercises which draw upon various evidence-based mechanisms and interventions such as goal setting (Locke & Latham, 2002), optimism (Glaesmer et al., 2008), cognitive restructuring (Clark, 2013), problem- and emotional-oriented coping (Folkman & Lazarus, 1985), and defensive pessimism (Norem, 2008).

Table 1 Fear-setting instructions and suggested mechanisms
Full size table

1.2 The Current Study

In this randomized controlled trial with an intervention condition and a wait-list control group, we investigated the effects of fear-setting on well-being related outcomes immediately after the intervention (posttest) and one week after the intervention (follow-up) as a novel, accessible, and multi-mechanism writing intervention for increasing well-being. Fear-setting is a popular intervention among positive psychology practitioners, popularized by a TED talk (see Ferriss, 2017) and discussed by various news outlets (e.g., Business Insider, CNBC, USA Today, INC.com). Therefore, a secondary goal of the study was to empirically test the effectiveness of the intervention. We hypothesized that (1) participants in the intervention condition would report a stronger increase in motivation to act, stronger decrease in fear of failure, stronger increase in self-efficacy, and stronger increase in optimism from baseline to posttest than participants in the control condition (cognitive outcomes); and (2) participants in the intervention condition would report stronger increases in positive affect and stronger decreases in negative affect from baseline to posttest than participants in the control condition (affective outcomes).

2 Method

2.1 Participants

For this study, we aimed to recruit at least 40 participants in both conditions who finish the follow-up survey (min N = 80, see pre-registration for the power-analysis, https://osf.io/bkmwn). To account for potentially higher dropout in the intervention condition, we used a randomization ratio of fear-setting condition to waitlist control group of 1.5 to 1 (Dumville et al., 2006). The study was advertised as an online workshop to overcome procrastination and decision paralysis (named “Reach Your Goals” study). We recruited participants via social media and in undergraduate seminars and lectures in the psychology department at two German universities. We only included participants who were at least 18 years old.

The baseline survey was filled out 160 times (see CONSORT Flow diagram). Six participants filled out the baseline survey twice and we kept their first responses. Five participants were excluded: One person failed two out of two attention checks (“please indicate ‘somewhat agree’ on this item”), one person gave nonsensical answers (indicating the same response on every item), and three persons could not be matched to their answers on the posttest and follow-up survey. The final sample included 149 participants, 122 women and 27 men, with an average age of 29.5 years (SD = 10.3). Most participants had A-levels/high school diploma (45.6%), followed by 40.3% participants with academic degrees, and the remainder with non-academic degrees such as a traineeship or secondary school diplomas (14.1%). All participants provided informed consent prior to the study. Psychology students were eligible for course credit for their participation. The ethics committee at Goethe University Frankfurt approved the study (No 2019-43) (Fig. 1).

Fig. 1
figure 1

CONSORT flow diagram

Full size image

2.2 Procedure and Study Design

Data were collected via online questionnaires at study registration (baseline), immediately after the intervention (posttest), and one week later (follow-up). The 75-minute intervention was conducted within a standardized online workshop by a psychologist using a live video interaction platform with group sizes varying between 8 and 23. We did not expect effects on the group level because the interventions were implemented individually, and the grouping was for practical purposes only. Congruently, intraclass correlations for our cognitive and affective outcomes at posttest and follow-up ranged from 0.00 to 0.06 and would be considered small by conventional standards (Hox, 2002). The workshop included a short input on Stoicism and positive psychology, instructions and subsequent delivery of the fear-setting writing exercise, and a buddy and plenary exchange during which participants defined the first step towards their personal goal (see supplement M for detailed workshop and intervention materials, https://osf.io/jgqwz). Directly after the workshops, participants in the fear-setting condition filled out the posttest survey.

Participants in the control group received no instructions during the intervention period and were not informed about their group membership. Instead, they received scheduled emails to fill out the posttest and the follow-up survey at the same time as the intervention group. After filling out the follow-up survey, participants in the control group received an automatic email informing them that they had been randomized to the control group and inviting them to participate in the fear-setting workshop scheduled one week later.

2.3 Intervention

In the first step of fear-setting (see Table 1), participants defined a goal or desired personal change. In the second step, participants were asked to write down ten or more feared negative consequences related to their goal or desired personal change. They then rated the likelihood of these fears becoming true on a scale from 1 (lowest likelihood) to 10 (highest likelihood). Participants were asked to name at least ten fears, encouraging them to articulate even improbable fears in order to help them realize that some fears are irrational. Step three and four are about risk management. In the third step, participants reflected on strategies to prevent feared outcomes from becoming true or at least to reduce their likelihood. In the fourth step, participants wrote about strategies to use if feared outcomes do come true. Steps five and six are about reframing the problem and making rewards of trying and costs of waiting more salient. In the fifth step, participants listed ways that an attempt could benefit them, such as from increased self-esteem or newly learnt skills. In the sixth step, participants reflected about ways that their passivity and maintaining the status quo has caused them problems.

2.4 Measures

2.4.1 Motivation to Act

To measure motivation to act, we used the Questionnaire on Current Motivation (QCM; Rheinberg et al., 2001) with adapted 18 items of the subscales probability of success (McDonald’s omega [ω] = 0.84 for t0, ωt1 = 0.87, ωt2 = 0.89) and fear of no success (ωt0 = 0.81, ωt1 = 0.82, ωt2 = 0.83). We adapted items so they would capture the state regarding personal goals (e.g., ‘At the moment, I think I can cope with the difficulty of my goal or plan’). We adapted items for fear of failure, self-efficacy, and optimism in the same way (see below). Participants answered using nine items on a 7-point scale ranging from 1 (strongly disagree) and 7 (strongly agree).

2.4.2 Fear of Failure

We measured fear of failure regarding the personal goal using the Performance Failure Appraisal Inventory (PFAI; Conroy et al., 2002) with 11 adapted items of the subscales fear of experiencing shame and embarrassment (ωt0 = 0.88, ωt1 = 0.88, ωt2 = 0.91) and fear of devaluing one’s self-estimate (ωt0 = 0.82, ωt1 = 0.82, ωt2 = 0.83) on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree) (e.g., ‘Regarding my goal or plan, I will find solutions if I put my mind to it’).

2.4.3 Self-Efficacy

We used and adapted the German version of the Generalized Self-Efficacy Scale (GSES) by Schwarzer and Jerusalem (1995) to capture the subjective expectation of self-efficacy regarding the personal goal with set of ten items on a 4-point scale ranging from 1 (not at all true) to 4 (exactly true) (ωt0 = 0.89, ωt1 = 0.87, ωt2 = 0.93). An example item is ‘Regarding my goal or plan, I am confident that I could deal efficiently with unexpected events’.

2.4.4 Optimism

Optimism regarding the personal goal was assessed using three adapted items from the German version of the 10-item Life Orientation Test-Revised (LOT-R; Glaesmer et al., 2008) on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree) (ωt0 = 0.82, ωt1 = 0.90, ωt2 = 0.90). We used the items ‘At the moment, I’m optimistic about my future regarding my goal or plan’, ‘At the moment, I expect the best regarding my goal or plan’, and ‘At the moment, I expect more good things to happen to me than bad regarding my goal or plan’ (as used in Heekerens et al., 2022).

2.4.5 Affect

Current affect was measured using the German version of the Positive and Negative Affect Schedule (PANAS; Breyer & Bluemke, 2016), consisting of ten items each for positive (e.g., ‘active’; ωt0 = 0.86, ωt1 = 0.91, ωt2 = 0.91) and negative affect (e.g., ‘irritable’; ωt0 = 0.89, ωt1 = 0.77, ωt2 = 0.92), on a 5-point scale from 1 (not at all) to 5 (extremely).

2.5 Statistical Software

Hypotheses tests, exploratory analyses, additional and sensitivity analyses, and measurement equivalence testing were done using Mplus version 8.7 (Muthén & Muthén, 2017).

2.6 Statistical Analysis

We analyzed our data using structural equation models that allow us to test underlying assumptions of longitudinal models (e.g., measurement equivalence; Millsapp, 2011) [1]. In the present study, we were interested in the effect of an intervention compared to a waitlist control group across three measurement occasions (baseline, posttest, follow-up). On each measurement occasion, outcomes were assessed using multiple variables (indicators) that were expected to load on common factors. Because meaningful across-time comparisons (e.g., of latent means) require strong scale invariance (equal indicator loadings and intercepts) across time, we tested this prerequisite in a first step of our analysis (for an introduction, see Geiser, 2020; Millsapp, 2011). For this purpose, we defined models assuming that a single factor is present at each measurement occasion and that all indicators load onto the same factors across time (configural scale invariance). These models were compared against more constrained models assuming equal loadings (weak scale invariance), and the weak scale invariance models were then compared against even more constrained models additionally assuming equal intercepts (strong scale invariance). All models include indicator-specific factors (Eid et al., 1999) to address the issue of indicator specificity in our longitudinal data set. For scales with ten or more items, three parcels were formed by aggregating three or four items (Matsunaga, 2008). We compared the models using chi-squared difference tests and assumed that the more constrained model fits worse for p-values below 0.01. Except for models including the momentary optimism scale, which only fulfilled criteria for weak scale invariance, strong measurement invariance over time can be assumed in all our models.

To investigate group differences in change over time, we used latent change score (LSC) models that analyze true state score changes across time, which has the advantage that measurement error is accounted for (Steyer et al., 1997). As in our measurement equivalence tests, all models include indicator-specific factors. Change score variables are formulated to reflect changes between neighboring measurement occasions (i.e., baseline and posttest, or posttest and follow-up). We discuss results for models with change score variables reflecting changes relative to the baseline measurement in the Additional Analysis section. Model fits were appropriate for all models (as indicated by CFI < 0.97, RMSEA > 0.05, and SRMR 0.08; Hooper et al., 2008; Hu & Bentler, 1999). To obtain effect sizes, we calculated the standardized difference in mean changes from pretest to posttest \(d\) (Becker, 1988; Viechtbauer, 2010). We interpreted these according to the guidelines suggested in Cohen (1988) with d = 0.2 as a small effect, d = 0.5 as a medium effect, and d = 0.8 as a large effect.

3 Results

3.1 Preliminary Analysis

Prior to the main analysis, we conducted a MANOVA to test whether participants in the intervention and waitlist control conditions differed in their baseline scores on cognitive and affective outcomes. Results indicated no difference between conditions, F(11, 148) = 0.46, p = .92. Table S1 displays descriptive statistics per condition and time point (means, standard deviations, and correlation coefficients).

Table S2 describes the wide range of different goals participants who completed the fear-setting writing exercise (n = 59). The categories of goals, in descending frequency, were professional and education development (e.g., write a doctoral thesis), personal and emotional development (e.g., become more confident), personal life style and well-being (e.g., move in with a friend), health and fitness (e.g., lose 50 kg), creative and artistic pursuits (e.g., write a fantasy novel), and financial and material goals (e.g., buy a home).

3.2 Test of Hypotheses

3.2.1 Motivation to Act

According to our hypothesis 1, we expected a stronger increase in motivation to act a personal goal from baseline to posttest in the intervention compared to the control condition. Average perceived probability of goal success, a subscale of motivation to act with a range from 1 to 7, was 4.9 at baseline and 5.4 at posttest in the intervention condition compared to 5.0 at baseline and 5.2 at posttest in the control condition. The corresponding small-to-medium effect size from the standardized difference in mean change from baseline to posttest was d = 0.27, 95% CI [0.01, 0.53]. Consistently, Table 2 shows that condition (0 = control, 1 = intervention) predicted an increase in latent baseline to posttest change scores of perceived success probability, b = 0.49, 95% CI [0.10, 0.88], and a subsequent insignificant decrease at follow-up, b = -0.19, 95% CI [-0.56, 0.19]. Average perceived fear of no goal success, another subscale of motivation to act on a scale from 1 to 7, was 4.5 at baseline and 4.1 at posttest in the intervention condition compared to 4.7 at baseline and 4.4 at posttest in the control condition. The corresponding effect size was d = -0.07, 95% CI [-0.33, 0.20]. Concurringly, Table 2 reveals a trend but no significant group difference in latent baseline to posttest change scores, b = -0.49, 95% CI [-1.02, 0.04] and no change in latent posttest to follow-up changes, b = -0.02, 95% CI [-0.47, 0.43].

Table 2 Parameter estimates for latent change score models on differences between the fear-setting condition (n = 86) and the waitlist control condition (n = 63) on primary outcomes in changes from baseline to posttest, posttest to follow-up, and baseline to follow-up
Full size table

3.2.2 Fear of Failure

According to hypothesis 1, we expected a stronger decrease in fear of failure from baseline to posttest in the intervention than in the control condition. Average fear of experiencing shame and embarrassment, a subscale of fear of failure on a scale from 1 to 5, was 3.1 at baseline and 2.7 in the intervention condition compared to 3.2 at baseline and 2.8 at posttest in the control condition. The standardized difference in mean changes from baseline to posttest was d = -0.08, 95% CI [-0.35, 0.19]. Table 2 displays a trend in favor of the fear-setting condition but no significant group difference in latent baseline to posttest change scores of experiencing shame and embarrassment, b = -0.30, 95% CI [-0.63, 0.03] and no significant group differences in latent change scores from posttest to follow-up, b = -0.04, 95% CI [-0.28, 0.21]. Average fear of devaluing one’s self-estimate, another subscale of fear of failure on a scale from 1 to 5, was 3.0 at baseline and 2.6 at posttest in the intervention condition compared to 3.0 at baseline at 2.7 at posttest in the control condition. The corresponding effect size was d = -0.06, 95% CI [-0.32, 0.21]. For fear of devaluing one’s self-estimate, no significant group difference in latent baseline to posttest change scores were found, b = -0.24, 95% CI [-0.60, 0.12] and no significant differences from posttest to follow-up, b = 0.11, 95% CI [-0.15, 0.36].

3.2.3 Self-Efficacy

According to our first hypothesis, we expected a stronger increase in self-efficacy from baseline to posttest in the intervention compared to the control condition. Average self-efficacy (sum scores ranged from 10 to 40) was 26.4 at baseline and 27.9 at posttest in the intervention condition compared to 25.4 and 27.4 in the control condition. The related effect size was d = -0.13, 95% CI [-0.40, 0.14]. According to Table 2, there were no significant group differences in latent baseline to posttest change scores for self-efficacy, b = 0.07, 95% CI [-0.05, 0.19] and for differences from posttest to follow-up, b = -0.02, 95% CI [-0.16, 0.13].

3.2.4 Optimism

According to hypothesis 1, we also expected a stronger increase in optimism from baseline to posttest in the intervention compared to the control condition. Average optimism with sum scores ranging from 3 to 15 was 10.4 at baseline and 11.4 at posttest in the intervention condition compared to 10.3 at baseline and 11.3 at posttest in the control condition (see Table 3). The associated effect size was d = 0.01, 95% CI [-0.31, 0.32]. There was no significant group difference in latent baseline to posttest change scores of optimism, b = 0.19, 95% CI [-0.11, 0.49], and from posttest to follow-up, b = 0.00, 95% CI [-0.28, 0.28]. [2]

Table 3 Parameter estimates for latent change score models on differences between the fear-setting condition (n = 86) and the waitlist control condition (n = 63) on secondary outcomes in changes from baseline to posttest, posttest to follow-up, and baseline to follow-up
Full size table

3.2.5 Positive Affect

According to hypothesis 2, we expected a stronger increase in positive affect from baseline to posttest in the intervention compared to the control condition. Average positive affect on a scale from 1 to 5 was 2.9 at baseline and 3.2 at posttest in the intervention condition compared to 2.8 at baseline and 2.8 at posttest in the control condition suggesting a greater increase in the intervention condition. The corresponding medium effect size was d = 0.40, 95% CI [0.05, 0.74]. Consequently, Table 3 shows that condition (0 = control, 1 = intervention) predicted an increase in latent baseline to posttest change scores of positive affect, b = 0.45, 95% CI [0.18, 0.73] and a subsequent (insignificant) decrease from posttest to follow-up, b = -0.30, 95% CI [-0.63, 0.02].

3.2.6 Negative Affect

According to hypothesis 2, we expected a stronger decrease from baseline to posttest in the intervention compared to the control condition. The average score for negative affect on a scale from 1 to 5 was 1.8 at baseline and 1.5 at posttest in the intervention condition compared to 1.8 at baseline and 1.6 at posttest in the control condition. The standardized difference in mean changes from baseline to posttest was d = -0.14, 95% CI [-0.45, 0.17]. Table 3 shows the insignificant group differences in latent baseline to posttest change scores of negative affect, b = -0.08, 95% CI [-0.21, 0.05] and the insignificant differences in change scores from posttest to follow-up, b = 0.23, 95% CI [-0.04, 0.50].

3.3 Correction for Multiple Tests

To account for alpha-error accumulation and to correct for the influence of multiple hypothesis tests, we calculated Benjamini-Hochberg corrected p-values (False Discovery Rates; Benjamini & Hochberg, 1995). All p-values refer to the intervention effect on the change of the respective variable from baseline to posttest (see Tables 2 and 3 for details). As per our pre-registration, we conducted one-sided tests. The corrected p-values are 0.024 (perceived success probability), 0.066 (fear of no goal success), 0.066 (fear of experiencing shame and embarrassment), 0.125 (fear of devaluing one’s self-estimate), 0.126 (self efficacy), < 0.001 (positive affect), 0.126 (negative affect), and 0.110 (momentary optimism).

3.4 Dropout Analysis

To ensure that dropout did not introduce bias into our results, we conducted a detailed dropout analysis on the final sample (n = 149) comparing study completers (n = 101) to dropouts after baseline (n = 39) and dropouts after the posttest (n = 9) on age, gender, and baseline scores on primary and secondary outcomes of the study. There were no differences in age or gender between completers or dropouts. While there were some baseline differences in positive affect between dropouts after baseline (M = 2.8, SD = 0.6) and study completers (M = 3.1, SD = 0.8), this difference was equally distributed between the intervention and the control group, t(37), p = .214. In addition, there were some baseline differences on fear of failure (self-estimate) between dropouts after baseline (M = 3.1, SD = 1.1) and dropouts after the posttest (M = 3.9, SD = 0.5) and a trend for self-efficacy. Because of the small group size of dropouts after the posttest (n = 9), we conducted two-sample Wilcoxon rank-sum tests to compare the intervention to the control condition. There were no significant differences between the intervention and control condition on primary or secondary outcomes (all p > .05). Overall, the dropout patterns in this study appear to not overall skew the overall results.

3.5 Additional Analysis

To provide additional information regarding intervention effects, we calculated latent change score models with change score variables reflecting changes relative to the baseline measurement, specifically to investigate direct changes from baseline to follow-up. Results in Tables 2 and 3 indicate significantly higher latent baseline to follow-up change scores for perceived success probability, b = 0.49, 95% CI [0.05, 0.93], and positive affect, b = 0.45, 95% CI [0.01, 0.89], as well as significantly lower scores for shame and embarrassment, b = -0.57, 95% CI [-1.10, -0.04] in the intervention compared to the control condition. [3]

4 Discussion

In this study, we investigated the effects of a brief writing intervention called fear-setting, which is designed to increase well-being in the context of striving towards a personally meaningful goal (Irvine, 2009). Fear-setting is already used in practice but there has been no evidence about its effectiveness so far.

4.1 Intervention Effects on Key Cognitions and Affect

In support of our first hypothesis we found that the fear-setting intervention increased the perceived probability of successfully acting towards a personally meaningful goal with a small-to-medium effect size, which is a factor in the overall construct motivation to act (e.g., ‘At the moment, I think I can handle the difficulty of my goal or project’, Rheinberg et al., 2001). For motivation to act (subscale fear of no success), we found no significant benefit of the intervention condition above the waitlist control condition, but the direction of the effect was as expected. According to temporal motivation theory, increasing the perceived chances of success at a given task increases the utility of taking action (\(\text{u}\text{t}\text{i}\text{l}\text{i}\text{t}\text{y}=\text{c}\text{h}\text{a}\text{n}\text{c}\text{e} \text{o}\text{f} \text{s}\text{u}\text{c}\text{c}\text{e}\text{s}\text{s}\times \text{v}\text{a}\text{l}\text{u}\text{e} \text{o}\text{f} \text{s}\text{u}\text{c}\text{c}\text{e}\text{s}\text{s})\), thus increasing the likelihood of action and reducing procrastination (Steel, 2007; Steel & König, 2006). Our study provides tentative evidence that fear-setting can be used to increase motivation to act towards meaningful goals or desired personal changes by changing the perceived likelihood of success. Since the effect for motivation to act (subscale fear of no success) was in the expected direction (p = .066), future research with high sample sizes should reconsider this potential effect. Together, these results suggest that the fear-setting intervention may support individuals in perceiving success as more likely.

For fear of failure, the results were mixed. We found that, after applying corrections for multiple testing (Benjamini & Hochberg, 1995), the fear-setting intervention did not significantly reduce fear of failure from baseline to posttest (subscales fear of experiencing shame and embarrassment and fear of devaluing one’s self-estimate), but the direction of the effect was as expected. However, examining the results from baseline to follow-up, our data indicates significantly higher decreases in fear of failure (shame and embarrassment) in the intervention compared to the control condition. Striving towards meaningful goals is often associated with fear of failure because failure could mean embarrassment in front of others or having to devalue one’s self-estimate (Conroy et al., 2002). When reasonable beliefs become biases such as ‘I can’t do it’ or ‘I will fail’, these biases can become debilitating (Beck & Haigh, 2014). Fear-setting may not reduce fears but instead prevent them from becoming debilitating biases, which could explain why participants saw their success as more likely and were therefore more motivated to act. More research on the fear-setting intervention is needed in order to examine its potential effect on fear of failure.

As expected, participants in the fear-setting intervention reported increased positive affect immediately after the intervention with a medium effect size. In line with self-determination theory (Deci & Ryan, 2012) and goal setting theory (Locke & Latham, 2002), our study provides evidence that deliberately thinking about personally meaningful goals and discovering ways how to attain these goals increases positive affect, especially feeling excited, inspired, and lively (high arousal/positive valence affect, see Posner et al., 2005). Thinking about partial success and pre-empting solutions to challenges are two parts of fear-setting that directly address the future. Both thinking about a favorable future event (e.g., success) and changing one’s expectation toward a more favorable future outcome has been linked to increased positive affect in many well-controlled studies (for a review, see Quoidbach et al., 2015). For negative affect, the study suggests no favorable effect of fear-setting compared to the waitlist control condition. Since the fear-setting intervention instructs individuals to constructively engage with their fears related to goals or personal changes, thinking about their fears may inherently evoke unpleasant emotions, even if that relates to higher motivation to act and positive affect.

From the results on positive affect, it follows that fear-setting should also increase optimism, which was not the case in this study. Optimism was measured in relation to the personal goal or desired change (e.g., ‘At the moment, I expect more good things than bad things to happen to me in relation to my goal or intention’). However, the items used to measure optimism were still more general in nature and implied more than success or failure of the endeavor (e.g., experiences while working toward a goal), which could be negative.

Other than expected, self-efficacy was unaffected by the intervention. Bandura (1977) suggests to increase self-efficacy individuals need to have mastery experiences at a given task, observe similar others do well at the task, receive verbal encouragement, and have favorable emotional states, none of which, with the exception of favorable emotional states, were direct targets in fear-setting. Self-efficacy interventions are usually domain-specific (e.g., dietary behavior, physical exercise) and fear-setting is general. In addition, several meta-analyses have found no or weak effects of self-efficacy interventions (e.g., Ashford et al., 2010). To increase self-efficacy, longer, multi-session and specific interventions may be needed.

4.2 Limitations and Directions for Future Research

The study has several limitations. First, we told participants that the goal of the intervention was to overcome procrastination and we used a waitlist control condition. Our design does not allow us to rule out the influence of expectancy effects (e.g., our participants may have felt compelled to respond to expectations unintentionally created by our team; Rosenthal, 1994). Furthermore, our intervention was advertised as a study on overcoming procrastination and decision paralysis, which may have drawn in particularly motivated or struggling participants (for a review on selection bias, see Infante-Rivard & Cusson, 2018). This was also reflected in the data, as we observed changes in the expected direction in the control condition on most response variables (see also Table S1). Possibly, it could be that individuals with low baseline motivation to change or approach goals may benefit more from fear-setting. In addition, it could be that factors other than the proposed intervention mechanisms were driving our effects (e.g., interaction and perceived support in the workshop groups). To clarify if the fear-setting writing intervention compared with a psychoeducation workshop still has effects beyond the social component, future research should compare fear-setting to a control group with a social activity (e.g., group discussions on well-being, socializing without a focus on well-being, or recreational activities). Future research could also compare fear-setting to active control groups (e.g., goal-setting, best possible self) to distinguish which interventions are most effective at motivating individuals to pursue different kinds of goals. Second, defensive pessimism involves setting low expectations for goal pursuits and thinking through obstacles or setbacks and how to avoid or then cope with them (reflected in steps three and four of fear-setting, i.e., prevent and repair, see Norem & Illingworth, 1993). However, the study did not measure if participants were defensive pessimists (Norem, 2001), which could be an important moderator for the fear-setting intervention. Third, fear-setting is a broad intervention that targets multiple mechanisms of change at once. Future research could elucidate which mechanisms drive effect sizes, which could then help optimize the intervention. Finally, participants indicated a large range of goals and personal changes they had been putting off (e.g., ‘finish my thesis’, ‘break off contact with former romantic partner’, and ‘lose 50kg’). Attaining some of these goals might be more difficult than others. We decided to allow all goals, decisions, and personal changes to test fear-setting in its most general form and to also show the flexibility of the writing exercise. Future research could limit the options to focus on specific domains (e.g., health habits, academic achievement, and so on).

We make several suggestions for future research. First, future research may wish to make changes to the fear-setting intervention and to how it is delivered. For example, other writing interventions produce stronger effects when the dosage is higher (i.e., when they are repeated more than once, see Travagin et al., 2015). Another change could be to offer fear-setting as a self-administered intervention through an app or a website (without a workshop) to make the intervention more accessible and decrease its low-threshold. The effects of self-administered fear-setting may be somewhat lower than in a group-setting as meta-analyses on PPI have repeatedly shown that PPI are more effective in group settings than in self-administered settings (Bolier et al., 2013; Carr et al., 2021; Sin & Lyubomirsky, 2009). Furthermore, we used fear-setting as an intervention to increase well-being. Since fear-setting is also an intervention that premeditates challenges and helps discover ways to deal with such challenges (similar to defensive pessimism; Norem, 2008), fear-setting could be studied as an intervention to improve problem-oriented coping. Next, future studies may wish to investigate fear-setting in populations that are especially prone to procrastination, including persons high on neuroticism or low on conscientiousness (Eerde, 2004; Steel, 2007) or university students (Ferrari et al., 1995; Steel & Ferrari, 2013). In addition, we framed fear-setting in the context of self-determination theory, which emphasizes the role of autonomy, competence, and belonging on intrinsic goals, motivation, and well-being, but we did not analyze the impact of social contexts on the effects of fear-setting. Future research could explore moderating factors within the social context (e.g., norms and culture, social support).

5 Conclusion

This was the first study to evaluate the effectiveness of fear-setting as a writing intervention to increase well-being. The study provides initial evidence that fear-setting may increase positive affect and increase motivation for individuals who want to strive towards an important goal or make a personal change or difficult decision. Fear-setting may also reduce fears of failure related to anticipating shame and embarrassment. We found no effects of fear-setting for self-efficacy, optimism, and negative affect. We suggest more research examining fear-setting as an intervention designed to increase well-being and to increase motivation to reach for goals or make difficult decisions or personal changes that individuals procrastinate on.

Footnotes.

  1. 1.

    In the pre-registration, we wrote that we would use linear mixed models to evaluate the effects of the fear-setting intervention over time. Instead, we used latent change score models to accurately model measurement errors across time, which improved statistical power. The main results (significant effects for motivation to act - subscale success probability - and positive affect) do not change when compared to using linear mixed models.

  2. 2.

    Although the strong measurement invariance solution was a significantly worse fit than the weak measurement invariance solution for momentary optimism, fit indices indicate an appropriate fit for the corresponding LCS model (χ² (33, N = 149) = 727.63, p < .001, CFI = 0.99, RMSEA = 0.05, 95% CI [0.00, 0.08], SRMR = 0.05).

  3. 3.

    We report additional results for exploratory outcomes (life satisfaction, depression, and self-esteem) in supplementary material, Table S3.