The Effect of Counterfactual Potency on Behavioral Intentions

Abstract

This research examines how counterfactual potency (CP), the multiplicative effect of the likelihoods of the “if” and “then” clauses of counterfactuals, determines the effects of counterfactuals on behavioral intentions. In Study 1, we found that participants who read highly (vs. minimally) mutable vignettes perceived the counterfactuals as more likely and endorsed relevant intentions more. However, CP did not mediate the effect of mutability on intentions. In Studies 2 and 3, we found that CP directly affected intentions and also mediated the effects of mutability on intentions when mutability was specifically manipulated via controllability (Study 2) or norm violation (Study 3). Finally, Study 4 used archival reaction time data to show that more concrete counterfactuals were perceived as more likely and subsequently facilitated intentions. Taken together, the current research provides evidence that more likely counterfactuals facilitate behavioral intentions.

Keywords

counterfactual thinking counterfactual potency likelihood perception intention endorsement accessibility

Imagining what might have been, known as counterfactual thinking, is a pervasive form of mental simulation (Summerville & Roese, 2008). When engaging in counterfactual thinking, people imagine conditional propositions with an “if” clause describing an alternative behavior that did not occur, and a “then” clause depicting a consequent, imagined outcome (Roese & Epstude, 2017). For example, after receiving an unsatisfactory grade, students may think that “If only I had studied more, then the exam would have gone better.” Counterfactuals can help people prevent similar negative outcomes or procure a better outcome in the future (Roese & Epstude, 2017). Upward counterfactuals, which simulate a favorable alternative to reality, can facilitate intentions, regulate behaviors, and improve performance (e.g., Smallman & Roese, 2009). However, not all counterfactual thoughts are equally functional. Some counterfactuals can result in fast and strong intention formation, whereas some other counterfactuals can lead to relatively slow and weak intention formation (e.g., Smallman, 2013; Walker et al., 2016). Why does the strength of counterfactual effects on intention formation vary? Past research has focused on counterfactual content (Smallman, 2013), group membership (Walker et al., 2016), and temporal distance (Smallman & McCulloch, 2012) to answer this question. Investigating factors that alter counterfactual effects on intentions is crucial to understand the role of counterfactuals in forming future behavioral intentions. A key role of functional counterfactual thinking is to provide information that helps improve relevant future behavior and avoid mistakes in the future. One important unexplored factor is likelihood perceptions of counterfactuals.

Likelihood perceptions affect various mental simulations, including both counterfactual thoughts about an imagined alternative past as well as imagined possible futures. Counterfactual thinking always entails an (often implicit) evaluation of the likelihood of the simulated past that might have happened (Petrocelli et al., 2011). For instance, how much a driver endorses the thought “If I’d taken another route, I wouldn’t have been late” may depend on whether they took a detour that day (in which case following the usual route seems plausible) versus followed their established commute (in which case an alternative route may seem unlikely), as well as whether the alternative route was likely to have been faster than the actual route. For future events, increased likelihood perceptions lead people to judge that the event will happen and form behavioral intentions accordingly (Carroll, 1978). That is, the thought “I’ll want a jacket if it’s going to be snowy” seems more likely, and packing a jacket more crucial, if one is traveling to Copenhagen in January than to Cairo in August. Given the importance of likelihood in both counterfactual thoughts and plans about the future, it may be the case that likelihood perception of counterfactual thinking influences future judgments just as the likelihood of a future event affects future judgments. The current research thus examines the relationship between likelihood perception of counterfactual thinking, known as counterfactual potency,¹ and intention formation.

Likelihood and Regulatory Functions of Counterfactual Thinking

Counterfactual thinking often serves a functional role in regulating behavior by reflecting a desired goal (Roese & Epstude, 2017). For instance, one may think that “If I had studied, I would have passed the exam,” and this counterfactual thought embodies a desirable goal state (e.g., passing the exam) and an alternative action (e.g., studying) needed to achieve the goal. Upward counterfactual thinking can regulate behavior by identifying better outcomes that could have been obtained. This regulatory function can occur by broadly increasing motivation or by influencing specific behavioral intentions (Roese & Epstude, 2017; Smallman & Summerville, 2018; see Figure 1).

Figure 1.

Two distinct pathways from counterfactual thinking to behavior and the (potential) roles of counterfactual potency.

Counterfactuals can have diffuse motivational benefits through their influences on negative affect and control perception (Smallman & Summerville, 2018). That is, increased negative affect (Myers et al., 2014) and perceived control over future events (Nasco & Marsh, 1999) induced by counterfactual thinking can maintain and enhance motivation (Markman et al., 2008).

Judgments of counterfactual likelihood activate diffuse motivational effects both through its effects on affect and perceived control for future events. Higher likelihood induces higher responsibility perception and more intense negative emotions (Petrocelli et al., 2011). In addition, higher responsibility perception for a past event intensifies feelings of regret (Zeelenberg et al., 2000). Such increased negative affect enhances motivation (Carver & Scheier, 1990). Furthermore, more likely counterfactuals entail an alternative action (e.g., usual behavior) that is more controllable than an alternative action (e.g., unusual behavior) in less likely counterfactuals (Petrocelli et al., 2011). Thus, likelihood may increase a sense of control over future events (see also Nasco & Marsh, 1999).

In addition to these diffuse motivational effects, counterfactuals can affect behavioral intentions specifically related to the contents of a counterfactual by identifying a cause of the past outcome, facilitating relevant behavioral intentions (Smallman & Roese, 2009). In this respect, a counterfactual provides a lesson that helps avoid a similar mistake in the future. However, in contrast to the body of work showing a relationship between likelihood and diffuse motivation, so far there has been no investigation of the relationship between likelihood and intention formation. The current research thus focuses on how the perceived likelihood of counterfactuals influences behavioral intentions.

How Might Likelihood (CP) Influence Behavioral Intentions?

A counterfactual is triggered when people perceive a discrepancy between a desired goal and actual state. To reduce such discrepancy and obtain the goal, individuals search for possible alternatives (Roese & Epstude, 2017). If the identified alternatives are not likely to have happened or helped realize the goal, however, these alternatives would not effectively reduce the discrepancy. Indeed, counterfactuals focused on unlikely antecedents can serve as a self-justifying excuse and reduce motivation (McCrea, 2008).

The perceived likelihood of a counterfactual is related to two necessary conditions for a past experience to trigger an intention: the memory of that experience must be both cognitively available and relevant to the future intention. If an event is forgotten, it cannot be used to generate an intention. Likewise, a remembered but apparently irrelevant experience will also not be used to generate an intention. Even if a memory can be easily recalled, individuals do not use this knowledge unless it is appropriate or causally relevant to the desired outcome (Ajzen, 1977; Higgins, 1996). Counterfactual thoughts that are subjectively higher in likelihood are more likely to meet both of these criteria than thoughts with lower perceived likelihood.

First, higher likelihood increases memory encoding, availability, and accessibility. One fundamental function of memory is to establish references from past experiences and use them in the future (Anderson & Milson, 1989). When a simulated event is perceived to be likely or plausible, the memory of that event tends to be thoroughly encoded, readily available, and in turn more accessible (Kahneman & Tversky, 1982; Pezdek et al., 1997; Sherman et al., 1983). Likewise, if simulated counterfactual thoughts are regarded as likely or plausible, the likelihood of the activation of stored knowledge (e.g., a counterfactual per se and a lesson from it) would increase. That is, a similar situation in the future activates accessible stored knowledge; counterfactuals more likely to have happened would more easily come to mind.

Second, higher likelihood could increase the relevance of the counterfactual and thus use of specific knowledge. In particular, “then likelihood” (TL) involves the perceived likelihood of an alternative outcome occurring (i.e., the likelihood of a “then” clause) given that an alternative action (i.e., an “if” clause) had taken place. TL therefore implies how likely an alternative action (identified in the “if” clause) would have helped achieve the desired goal (identified in the “then” clause). For example, “If I hadn’t been texting, the accident might not have happened” is likely to be high in TL and thus relevant to the goal of safer future driving if the accident occurred because of distraction. However, if distraction was not a contributing factor (e.g., the accident occurred when another car rear-ended the driver at a red light), then TL will be low and the thought not relevant to future intention generation. Thus, thoughts high in TL indicate that the alternative is highly relevant to the goal. Among available and accessible counterfactuals in memory, individuals would use the ones that are likely to lead to achieve the goal. TL identifies functional behaviors for the future.

Taken together, likelihood appears to be crucial to the association of counterfactuals and behavioral intentions by meeting two necessary criteria: likely counterfactuals should be both more accessible in memory and more relevant to desired goal states. Higher likelihood counterfactuals should therefore facilitate intentions.

Evidence From Past Research

Although no work has directly examined the role of counterfactual likelihood perceptions on behavioral intentions, counterfactual content related to the accessibility or relevance of the counterfactual affects intentions. In particular, factors increasing the vividness or concreteness of counterfactuals enhance the effect of counterfactuals on intentions, consistent with the proposed facilitating effect of greater likelihood. For instance, describing negative events as having occurred in the near (vs. distant) past more easily facilitated relevant behavioral intentions (Smallman & McCulloch, 2012). Likewise, participants formed relevant intentions more quickly when considering concrete actions (e.g., wearing sunscreen) but not abstract alternative actions (e.g., taking precautions; Smallman, 2013).

These papers provide indirect evidence that likelihood changes behavioral intentions. Prior research demonstrates that temporal distance is associated with mental representations and likelihood. Specifically, simulating temporally closer events (e.g., cleaning the house tomorrow) induces concrete representation (e.g., vacuuming the floor), whereas imagining temporally distant events (e.g., cleaning the house next year) evokes abstract representation (e.g., showing one’s cleanliness). Furthermore, concrete (vs. abstract) representation and vivid (vs. pallid) events are judged to be more likely due to the ease of retrieval or imagining (e.g., Sherman et al., 1985; Wakslak & Trope, 2009). Kahneman and Tversky (1982) and Teigen (1998) also maintain that when a hypothetical event is easily generated, that event is perceived to be more likely. Thus, the facilitating effects of near past events and concrete actions in past studies of the effect of counterfactuals on behavioral intentions may have resulted from heightened likelihood perception.

Overview of Current Work

Across four studies, this research examines the relationship between likelihood and intention formation. We predict that likelihood perception of counterfactuals will drive intentions. In Study 1, we provide preliminary evidence for the effect of likelihood on intention endorsement using vignettes varying in the ease of imagining. In Study 2, we specifically examine controllability while accounting for event repeatability and consistency. Study 3 considers an alternative explanation of the results of Study 1 by testing the differences in likelihood and intention endorsement according to different types of events. We predict that unusual (vs. usual) behaviors in the past will induce higher likelihood, and in turn increased likelihood will predict intention endorsement. However, both likelihood and intention endorsement are not expected to differ based on how nearly an event might have been avoided (e.g., missing a flight by 5 vs. 50 minutes). In Study 4, we measure likelihood of counterfactual statements used in Smallman (2013) and combine these newly collected likelihood ratings with archival data to examine if likelihood is higher for concrete versus abstract counterfactuals, and if likelihood predicts faster archival reaction times for intention judgment from Smallman (2013). Together, these data on facilitating effects of likelihood on intention formation illustrate an important mechanism by which likelihood perception of counterfactuals facilitates behavioral intentions.

Study 1

Study 1 was designed to provide initial evidence that likelihood would increase intention endorsement. To manipulate likelihood, we altered details of vignettes to make an alternative more (high likelihood) or less (low likelihood) easily imagined, which should in turn affect likelihood (e.g., Kahneman & Tversky, 1982). In particular, we manipulated whether a causal situation was usual or unusual, or whether the counterfactual outcome was numerically close versus distant to reality. Negative events occurring under unusual (vs. usual) circumstances make counterfactuals more salient and easier to simulate (Kahneman & Miller, 1986; Teigen, 1998). For example, reading a story about an accident while driving an unusual versus usual route to home generates a counterfactual thought more readily (Walker et al., 2016). Likewise, if a negative outcome is close (vs. distant) to the desired outcome, individuals easily imagine an alternative outcome. Missing a flight by 5 minutes more easily brings about a counterfactual thought than missing a flight by 30 minutes does (Kahneman & Tversky, 1982). The ease with which the counterfactual is imagined determines the subjective plausibility of the event (Kahneman & Tversky, 1982; Stanley et al., 2017). Therefore, we predict that CP ratings will be greater when individuals simulate high mutability situations relative to low ones, and higher CP ratings would in turn induce stronger intention endorsement.

Method

Preregistration information for Study 1 can be found at https://aspredicted.org/5u8p9.pdf. Materials and data for all studies can be accessed at https://osf.io/pdcv9/?view_only=836ab139f2964faeb6f64a8ab3503cbb.

Participants

Participants were recruited at a mid-sized Midwestern university and compensated with course credit. The preregistered minimum sample size was 175 (based on the recommendation of Fritz & MacKinnon, 2007, to detect a moderate mediated effect size at 80% power) and we collected 192 responses, M_age = 19.14, SD_age = 1.14; 53.4% female, 45% male; 71.2% White, 21.5% Asian, 5.8% Black, 1% Native Hawaiian/other Pacific Islander, .5% American Indian or Alaskan Native; 2.6% Hispanic, Latinx, or Spanish origin. One response was eliminated because all intention ratings were missing.

Material and procedure

After providing informed consent, participants were seated at a computer in an individual cubicle. They were randomly assigned to either the high or low mutability condition and instructed to imagine the situations vignettes described. Each participant read six vignettes one at a time. Past research about counterfactuals has commonly employed a between-participants design (e.g., Walker et al., 2016) to avoid possible carryover effects of counterfactual thinking. For instance, if a highly mutable vignette precedes a minimally mutable one, counterfactuals generated while reading the highly mutable vignette can lead participants to consider alternatives on the low mutability event (Galinsky et al., 2000). For this reason, we chose to use a between-participants design.

Each vignette describes a negative event that could have been avoided. For the high mutability condition, participants imagined situations involving unusual actions or close outcomes. For the low mutability condition, participants imagined situations with routine actions or distant outcomes. After reading each vignette, participants in both conditions read the same counterfactual statement (e.g., “If only I had taken the most direct route, then I would have avoided the accident.”).

For each counterfactual, participants completed IL and TL items anchored at very unlikely (1) and very likely (7). For IL, participants were asked to consider the “If only” part of the statement and rate their likelihood perception (e.g., “How likely is it that you would have taken the most direct route?”). For TL, they were instructed to focus on the “then” part of the counterfactual, assuming that the “If only” part had taken place (e.g., “Given that you had taken the direct route, how likely is it that you could have avoided the accident?”). Then they rated their intentions to engage in the relevant behavior (e.g., “In the future, how likely are you to take the most direct route?”) on a slider set to the midpoint of very unlikely (0) to very likely (100).

Results

Preliminary analysis

CP was calculated by multiplying IL and TL (Petrocelli et al., 2011). CP ratings were submitted to a 2 (high and low mutability conditions) × 6 (vignettes) mixed-model analysis of variance (ANOVA) with repeated measures on the second factor. Excluding one story (“airport” in Figure 2), which showed an inconsistent pattern with other stories, resulted in a non-significant interaction effect, F(3.79, 713.26) = .67, p = .61, $η_{p}^{2}$ = .004.² We assumed that high and low mutability would induce high and low CP ratings, respectively. The main effect of condition was significant, F(1, 188) = 63.02, p < .001, $η_{p}^{2}$ = .25, indicating that the remaining five story kernels (10 vignettes) produced different levels of CP ratings.

Figure 2.

Differences of CP ratings according to mutability conditions and vignette in Study 1.

Main analysis

We expected that CP would positively predict behavioral intentions. Because of the repeated measures, the data are interdependent in nature. To account for repeated CP and intention ratings across multiple vignettes, a multilevel regression approach was employed regressing intentions on CP nesting trials within participants (see Supplemental Materials for analysis of aggregate measures for all studies). Consistent with the hypothesis, CP ratings had a significant effect on intentions at both within-participant, γ = .90, SE = .10, p < .001, 95% confidence interval (CI) = [.70, 1.09], and between-participant, γ = .66, SE = .18, p < .001, 95% CI = [.32, 1.01], levels. More likely counterfactuals led to stronger intention endorsement.

We also examined whether CP mediated the relationship between mutability condition and intention. We used multilevel structural equation modeling (MSEM) testing multilevel mediation following the procedures of Preacher et al. (2010). The mediation model (see Figure 5) contained the Level 2 (between-participant level) predictor (mutability condition), with the mediator (CP) and outcome (intention) assessed at Level 1 (within-participant level), which had both between- and within-participant variation. Because the predictor could exert its impact only at a between-person level, we considered only the between-participant level of mediation effect.

As shown in Figure 5, on the between level, the results indicated a positive effect of mutability condition on CP, γ = .8.94, SE = 1.14, 95% CI = [6.71, 11.18]; the easier it was to generate alternatives, the more plausible counterfactuals were. Mutability condition also significantly altered intention endorsement, γ = 15.48, SE = 2.98, 95% CI = [9.63, 21.33] when CP was controlled for. Moreover, CP had a positive effect on intentions, γ = .87, SE = .19, 95% CI = [.49, 1.25]; however, when mutability was controlled for, CP did not account for variance in intention endorsement, γ = .07, SE = .23, 95% CI = [−.39, .53]. The estimate for the indirect effect revealed that CP did not mediate the effect of mutability on intention endorsement, γ = .64, SE = 2.09, 95% CI = [−3.46, 4.74].

Discussion

The primary purpose of Study 1 was to provide preliminary evidence for the association between likelihood (CP) and intention. The multilevel regression analysis indicated that likelihood was a strong predictor of intention endorsement. That is, participants who indicated that counterfactuals were more likely to have occurred formed stronger intentions compared with those who perceived the counterfactuals as less likely. However, CP did not mediate the impact of mutability on intention endorsement.

There are several alternative explanations for the findings. First, perceiving an opportunity to alter the outcomes of future events is crucial for counterfactuals to be functional (Markman et al., 1993). However, we asked participants to rate their intentions in general (e.g., “In the future, how likely are you to take the most direct route?”), which might have resulted in different levels of repeatability perceptions. Thus, in the case of those who perceived an event as not repeatable, even potent counterfactuals may have been irrelevant to intentions. Second, intention ratings could be affected by consistency between usual behavior (in the low mutability condition) and behavior in the future. We addressed these two issues in Study 2.

Third, different mutability factors might have confounded the outcomes. The vignettes used in Study 1 varied in norm violation or closeness to trigger different levels of likelihood. However, closeness and norm violation were not fully crossed in Study 1. One of the stories manipulated closeness and the others manipulated norm violation. Moreover, some stories that contained either usual or unusual behaviors also included details that might inform a closeness perception (e.g., “. . . you got to the appointment 30 minutes late . . . ”). Therefore, in Study 3, we examine closeness and norm violation seperately. We discuss this relationship further in Study 3.

Fourth, the current results may have been influenced by a consistency bias or demand characteristic, as participants were asked to rate their intentions after completing the CP measures. That is, rating a counterfactual as “very likely” versus “very unlikely” may have caused participants to choose a corresponding answer (“very likely”) to the intention question. Also, asking the IL and TL questions right before the intention measure could increase a chance that participants recognize the purpose of the study. Therefore, in Study 4, we measure participants’ CP ratings for counterfactual prime statements that were used in Smallman (2013) and compare these new ratings with archival reaction time data for intention judgment. This method allows us to examine the facilitating effect of likelihood and avoid demand characteristics and consistency bias.

Study 2

To address concerns about repeatability and consistency from Study 1, in Study 2 we manipulated mutability levels via controllability of the antecedent action and by modifying the intention measure to specifically instruct participants to imagine the decision happening again in the near future to ensure that the counterfactual addressed a repeatable behavior. Consistent with Study 1, we expected to find greater CP ratings when participants simulated higher mutability situations, and stronger intention endorsement following greater CP ratings.

Method

Preregistration information for Study 2 can be found at https://aspredicted.org/sw4e2.pdf.

Participants

College students³ fluent in English were recruited from the Prolific online recruitment platform and compensated with US$1.40. We preregistered a minimum of 175 participants (Fritz & MacKinnon, 2007) to detect a moderate mediated effect size at 80% power and collected 183 responses. After eliminating four incomplete responses, 179 responses were analyzed, M_age = 24.79, SD_age = 6.54; 48% female, 44.7% male, 5% non-binary, 1.2% multigender, .6% gender-queer/non-conforming; 58.1% White, 14.5% Asian, 13.5% multiracial, 7.8% Hispanic, Latinx, or Spanish origin, 5% Black, .6% American Indian or Alaskan Native, .6% Middle Eastern or North African.

Materials and procedure

After providing informed consent, participants were randomly assigned to either the high or low mutability condition. Each participant read six vignettes one at a time. Individuals in the high mutability condition read events where they had control over the focal action (e.g., “After a lot of thought, you decided to go to The Smile instead of Daydream.”), whereas those in the low mutability condition imagined situations that were not controllable (e.g., “You had to go to The Smile restaurant because Daydream was closed for maintenance.”).

After reading each vignette, participants read the counterfactual corresponding to the event and completed the same CP measures as in Study 1. Next, they were asked to imagine a similar situation in the near future and rate their intention to engage in the relevant behavior on a slider set to the midpoint of very unlikely (0) to very likely (100). (e.g., “Think about the next time you decide where to eat [within the next few weeks]. How likely are you to go to Daydream?”) Following each intention measure, participants rated two questions about realism (e.g., “I think that there are situations like the story I just read in real life.”) on a 7-point Likert-type scale. (Including realism as a covariate did not change direction or significance of any reported effects; see analyses in Supplemental Materials.)

Results

Preliminary analysis

We submitted calculated CP (IL × TL, Petrocelli et al., 2011) to a 2 (high and low mutability conditions) × 6 (vignettes) mixed-model ANOVA with repeated measures on the second factor. We expected to see higher CP ratings in the high (vs. low) mutability condition, which would be supported by a main effect of condition. As expected, we found the main effect of condition, F(1, 177) = 17.07, p < .001, $η_{p}^{2}$ = .09, indicating significant differences between high and low conditions in CP ratings. However, one vignette regarding “sunscreen” showed an inconsistent pattern with the other vignettes (see Figure 3). Thus, we used the remaining five vignettes for further analyses. The main effect of condition remained significant after excluding the “sunscreen” vignette, F(1, 177) = 20.07, p < .001, $η_{p}^{2}$ = .10. There was a significant main effect of vignette as well, F(3.64, 643.39) = 2.76, p = .03, $η_{p}^{2}$ = .02, which demonstrates different effects of the stories of vignettes on CP levels and suggests that a multilevel approach is appropriate to analyze CP. The interaction effect of condition and vignette was not significant, F(3.64, 643.39) = .67, p = .60, $η_{p}^{2}$ = .004.

Figure 3.

Differences of CP ratings according to mutability conditions and vignette in Study 2.

Main analysis

First, we predicted that CP would predict stronger behavioral intentions. As in Study 1, we used a multilevel regression approach nesting trials within participants to account for repeated CP and intention ratings across multiple vignettes. Consistent with the prediction, CP significantly predicted intentions at both within-participant, γ = 1.15, SE = .08, p < .001, 95% CI = [.99, 1.31], and between-participant levels, γ = .85, SE = .18, p < .001, 95% CI = [.49, 1.21], such that the more likely counterfactuals were, the stronger participants endorsed relevant behavioral intentions.

We also examined the mediation effect of CP on the link between mutability condition and intention endorsement using MSEM (Preacher et al., 2010). On the between level, mutability condition had a significant effect on CP, γ = 6.09, SE = 1.35, p < .001, 95% CI = [3.44, 8.74], and CP subsequently increased intentions, γ = .96, SE = .21, p < .001, 95% CI = [.55, 1.36] when condition was accounted for. Thus, as shown in Figure 5, CP mediated the impact of mutability condition on intention endorsement, γ = 5.82, SE = 1.89, p = .002, 95% CI = [2.12, 9.52]. In other words, participants appeared to endorse behavioral intentions relevant to mutable counterfactuals because these counterfactuals were perceived as likely to have occurred.

Discussion

We conducted Study 2 to account for event repeatability and consistency between usual behavior and behavior in the future that could be responsible for null mediation effects in Study 1. We tested our hypothesis that CP would drive intention endorsement using a different manipulation varying controllability to eliminate the consistency effect and different intention measures to address event repeatability. The results showed that CP positively predicted intention endorsement, which is consistent with the results of Study 1. Furthermore, as we initially predicted, the findings demonstrated CP as a mediator in the relationship between mutability condition and intention endorsement. In other words, more controllable (i.e., more mutable) past events led participants to perceive the relevant counterfactuals as more likely to have happened, and more likely counterfactuals subsequently enhanced endorsement for the relevant behaviors. Thus, the null mediation results in Study 1 could be partly due to event repeatability and the consistency between usual behavior and behavior in the future, and Study 2 provided evidence supporting the hypothesis that likelihood would affect intention endorsement.

Study 3

In Study 1, we examined the effects of likelihood on intention endorsement with the assumption that the ease of imagining counterfactuals would increase likelihood, and subsequently intention endorsement. The results revealed no mediation effect of likelihood on the relationship between mutability and intention. However, different factors of mutability may have distinct effects. In Study 1, norm violation and closeness were not cross-balanced. Thus, if norm violation and closeness have distinct effects on likelihood, an effect specific to one of these might have been masked.

According to Roese (1997), norm violation and closeness play different roles in counterfactual generation. He identifies two distinct components of counterfactual generation: counterfactual activation and counterfactual content. Counterfactual activation refers to whether counterfactual processing is initially triggered, and counterfactual content refers to mutated (or altered) antecedents that undo the outcome. When individuals consider only factual information about the past, counterfactual processing is “switched off.” If they start pondering an alternative outcome (e.g., “How could I have avoided the accident?”), counterfactual thinking is “switched on” or activated. Then people generate counterfactuals with specific content (e.g., “if I had been more careful or driven slowly”) that could have resulted in a different outcome (e.g., “I would have been able to avoid the accident”).

Roese (1997) argues that closeness determines counterfactual activation, whereas norm violation affects content of counterfactuals. When missed opportunities are perceived as close, only slight effort is needed to obtain the goal. Thus, perceived closeness motivates individuals to activate counterfactual thinking by simulating an alternative outcome. In contrast, violation of an a priori norm causes individuals to consider a scenario in which the norm had occurred. That way, they could return to the normal state in a counterfactual. Thus, normality determines what antecedent is mutated (Kahneman & Miller, 1986; Roese, 1997), but does not itself turn on counterfactual processing.⁴

This distinction is critical to the current research because likelihood perceptions are closely linked with counterfactual content rather than mere activation. Activation is a necessary precondition for likelihood judgments but should not have a direct impact of the magnitude of estimation. In contrast, likelihood perceptions distinguish between different counterfactual content. Of CP components, the IL (if likelihood) component estimates the antecedent likelihood, whereas the TL (then likelihood) component represents the association between the alternative antecedent and outcome. Therefore, different counterfactual content would change each likelihood.

In Study 3, we therefore manipulate norm violation and closeness and explore their influences on CP and its impact on intention endorsement. We predict that unusual behaviors (i.e., violating the norm) will increase CP because unusual behaviors determine counterfactual content, and CP will be rated based on the likelihood of this content. However, near misses (i.e., close outcomes) will not change CP because near misses merely switch on counterfactual processing, and CP will not differ by how easy it is to activate counterfactuals, since the counterfactual is explicitly presented to participants and no activation is required. More important, we expect that CP will mediate the influence of norm violation but not mediate the impact of closeness on intention endorsement.

Method

Preregistration information for Study 3 can be found at https://aspredicted.org/cg9p2.pdf.

Participants

A power analysis indicated that 171 to 265 participants are necessary for a 2 (norm violation: unusual and usual behavior) × 2 (closeness: near and far miss) between-participants design with an effect size (.20 < f < .25) at 90% power. Therefore, we preregistered a goal of a minimum of 171 participants and collected 249 responses for this study, M_age = 30.86, SD_age = 10.36; 51.4% male, 46.2% female, 1.2% non-binary, .4% agender; 71.5% White, 17.7% Asian, 6.4% Black, 1.2% Native Hawaiian/other Pacific Islander, 1.2% American Indian or Alaskan Native; 9.2% Hispanic, Latinx, or Spanish origin. Participants fluent in English were recruited from the Prolific online recruitment platform and compensated with US$1.

Materials and procedure

After providing informed consent, participants were randomly assigned to one of the 2 (norm violation) × 2 (closeness) fully between-participants conditions. The norm violation manipulation involved usual (e.g., “You drove to campus like you normally do”) and unusual (e.g., “You decided to drive to campus instead of taking the bus like you normally do”) behaviors. The closeness manipulation consisted of near (e.g., “. . .you ended up getting 79 out of 100, which means you got a C”) and far (e.g., “. . .you ended up getting 75 out of 100, . . .”) misses. After reading each of three vignettes, participants read a corresponding counterfactual and completed the same CP and intention measures as in Study 1.

Results

Preliminary analysis

IL and TL values were multiplied to create CP ratings, which were submitted to a 2 (norm violation: usual and unusual condition) X×2 (closeness: far and close condition) × 3 (vignette) mixed-model ANOVA with the third factor measured within-subjects (see Figure 4). The three-way interaction was not significant, F(1.92, 469.24) = 2.06, p = .13, $η_{p}^{2}$ = .01. Although there was a significant interaction of norm violation and vignette, F(1.92, 469.24) = 3.49, p = .03, $η_{p}^{2}$ = .01, CP ratings were consistently higher in all vignettes containing unusual versus usual behaviors. Thus, this interaction effect merely means that some of the vignettes more sensitively changed CP than the other(s). Vignette did not interact with closeness, F(1.92, 469.24) = .40, p = .66, $η_{p}^{2}$ = .002. There was a main effect for vignette, F(1.92, 469.24) = 8.50, p < .001, $η_{p}^{2}$ = .03, which implies that the target counterfactual was more potent for some vignettes than others.

Figure 4.

Norm violation × closeness × vignette on CP ratings.

The first prediction in Study 3 was that norm violation would result in higher CP. Consistent with this prediction, we found a main effect for norm violation, F(1, 245) = 34.34, p < .001, $η_{p}^{2}$ = .12, d = .75; Bonferroni-adjusted comparisons: p < .001, 95% CI = [4.66, 9.37], indicating that events involving unusual behaviors led to greater likelihood relative to those with usual behaviors. The second prediction was that closeness would not affect CP. As expected, a null main effect for closeness, F(1, 245) = .02, p = .88, $η_{p}^{2}$ < .001, d = .03; Bonferroni-adjusted comparisons: p < .001, 95% CI = [−2.17, 2.54], showed that participants perceived similar levels of likelihood no matter how close or distant the desired outcome was from the actual outcome. A two one-sided tests (TOST) equivalence test (Lakens et al., 2018) confirmed that the observed effect size, d = .03, fell inside the boundary values of the medium effect size (−.50 < d < .50), t(246.54) = 3.73, p < .001, which indicates the difference of likelihood between the closeness conditions is statistically equivalent. The interaction of norm violation and closeness was not significant, F(1, 245) = .003, p = .96, $η_{p}^{2}$ < .001. In short, the two different mutability determinants have distinct associations with likelihood such that norm violation predicts likelihood whereas closeness does not.

We also used multilevel regression nesting trials within participants to examine the predictions of the positive impacts of norm violation on CP as well as the null relationships of closeness with CP. CP ratings were regressed on norm violation (0 = usual, 1 = unusual behavior) and closeness (0 = far, 1 = near miss). As expected, there was a significant effect of norm violation on CP ratings, γ = 7.02, SE = 1.19, p < .001, 95% CI = [4.69, 9.35]. Moreover, there was no effect of closeness, γ = .27, SE = 1.27, p = .83, 95% CI = [−2.22, 2.76]. These findings reaffirm the positive influence of norm violation on CP and the null relationship between closeness and CP.

The main hypothesis of the current research was that likelihood would predict intention endorsement. To test this, we regressed intention on CP ratings using multilevel regression and found the predicted significant effect of CP on intention at both within level, γ = .90, SE = .12, p < .001, 95% CI = [.66, 1.13], and between level, γ = .1.06, SE = .22, p < .001, 95% CI = [.62, 1.50].

As in Studies 1 and 2, we next conducted mediation analyses to test the mediation effect of likelihood on the relationships between two mutability factors and intentions. Given that there were no interaction effects of norm violation and closeness on either CP or intention ratings, it was reasonable to investigate two separate mediation models in which the initial predictor is either norm violation or closeness. Thus, we tested the indirect effect of each of the mutability factors separately on intentions via CP. We employed a MSEM paradigm to analyze multilevel mediation (Preacher et al., 2010). Similar to the data structure in Studies 1 and 2, the design of Study 3 corresponded to a 2-1-1 design with the Level 2 between predictor (either norm violation or closeness) and Level 1 mediator (CP) and outcome (intention). Indirect effects were identified at the between level because in a 2-1-1 design the indirect effects existed only at the between level.

For norm violation (see Figure 5), the estimates for the a path (the impact of norm violation on CP) and b path (the impact of CP on intention) were significant. The multiplicative product of these estimates showed a significant mediation effect of CP on the relationship between norm violation and intention, γ = 4.70, SE = 2.15, p = .03, 95% CI = [.49, 8.91]. For closeness (see Figure 5), only the b path (the effect of CP on intention) was significant. There was no significant indirect effect for closeness, γ = .29, SE = 1.34, p = .83, 95% CI = [−2.34, 2.91]. That is, CP mediated the influence of norm violation but not closeness on participants’ intention endorsement in the predicted pattern.

Figure 5.

A 2-1-1 multilevel mediation model in Studies 1, 2, and 3.

Discussion

Study 3 addressed a possible confounding effect of mutability factors in Study 1 by examining whether different determinants of counterfactual mutability, norm violation and outcome closeness, serve different roles in likelihood and its effect on intention. We predicted that violating a norm would lead to higher likelihood ratings relative to following the norm, as norm violations change the content and thus likelihood associated with a counterfactual. Furthermore, we expected that outcome closeness (near or far misses) would not change likelihood because it is associated with counterfactual activation and outcomes rather than counterfactual content and antecedents (Roese, 1997).

Study 3 supported the main hypothesis that more likely counterfactuals would induce stronger intention endorsement. Consistent with our predictions about specific factors playing different roles in likelihood, only norm violation, but not closeness, resulted in differences in likelihood and intention endorsement. Unusual versus routine behaviors caused participants to perceive their counterfactuals as more likely. This was not the case when closeness perception was considered. Likewise, likelihood mediated the effect of norm violation but not closeness on intention endorsement. This implicates that likelihood could be a proximal factor determining the strength of intention endorsement only when counterfactuals involve norm violation. Taken together, Study 3 identified that violating an a priori norm induces high likelihood of simulated counterfactuals, and this heightened likelihood increases intention endorsement. Also, perceived closeness exerts no effect on likelihood nor behavioral intentions. Thus, this study explains a possible reason for the null mediation effect of likelihood on the link between mutability and intention endorsement in Study 1, demonstrates the utility of likelihood in a context of norm violation, and shows distinct roles of mutability factors.

Study 4

Although Studies 2 and 3 found that more likely counterfactuals lead to available intentions being endorsed more strongly, it does not speak to the facilitating effect of likelihood on intention accessibility. One of the two necessary conditions for past experiences to trigger behavioral intentions is for these thoughts to be accessible as well as relevant. Although Studies 1 to 3 demonstrate the role of likelihood in relevance, they did not examine accessibility. Study 4 addresses this gap by examining accessibility using archival data from a reaction time paradigm. In addition, having participants rate CP and intention questionnaires in sequence creates potential demand or consistency concerns. In Study 4, we therefore collected CP responses and combined them with archival reaction time data. That is, we measured CP of the counterfactual stimuli from Smallman (2013) and examined the relationships between CP ratings generated in this study and archival reaction time data from Smallman (2013).

Smallman (2013) investigated whether the concreteness of a counterfactual affected its ability to facilitate an intention judgment in a reaction time paradigm. Given that the procedure varied the antecedents of a counterfactual and measured the reaction time for an associated counterfactual, it offered an ideal archival data set for the goals of the current research. In Smallman’s studies, she used 300 action phrases that described either a specific behavior, a category of behavior, or a trait. Participants made relevant intention judgments after reading those phrases paired with a counterfactual or factual cue, and their reaction times for the judgments were measured.

In the current study, we measured CP ratings of the same counterfactual stimuli as described above, consisting of a counterfactual cue and an action phrase. We predict that stimuli involving specific behaviors will bring about higher CP ratings compared with those with categories of behavior or traits because specific behaviors induce more concrete, vivid mental representation. More centrally, we also predict that counterfactual primes rated as more likely will be associated with faster reaction times for intention judgments.

Method

Preregistration information for Study 4 can be found at https://aspredicted.org/g2jc5.pdf. A total of 184 participants were recruited at a mid-sized public university in the Midwest United States to generate CP ratings on counterfactual prime statements taken from Smallman (2013). Participants were assigned to one of four counter-balancing conditions. Each of the four mutations (two specific behaviors, a category of behavior, and a trait) was rated for each of the 100 negative events (100 event descriptions and 400 prime statements in total) across all participants. However, one specific behavior per item was dropped later in analysis because it was not simultaneously examined with categories of behavior and traits in the original study (Smallman, 2013).

At the beginning of each trial, a negative event (e.g., “spilled food on shirt”) was presented. Participants then read one of three types of phrases paired with a counterfactual cue (“could have”): either a specific behavior (e.g., “used napkins”), a category of behavior (e.g., “eaten neatly”), or a trait (e.g., “been less sloppy”). Then they rated the same CP measures used in the Studies 1 and 2. After five practice trials, participants completed the main trials. In the main trials, 100 event descriptions were randomly presented, and each participant read 50 specific behaviors, 25 categories of behavior, and 25 traits presented in randomized order.

In Smallman (2013), each participant completed 100 trials of a sequential priming paradigm including 50 counterfactual and 50 control trials. In each trial, they read a negative event and a prime statement, which appeared below the event description after 2 seconds. This prime statement consisted of a cue and an action phrase. In the counterfactual trials, a counterfactual cue was paired with a statement involving a specific behavior, a category of behavior, or a trait. In the control trials, a control cue (e.g., “In the last week have”) was paired with one of the types of action phrases. In both conditions, participants chose either “yes” or “no” to decide either whether taking the action/trait could have changed the outcome of the negative event (the counterfactual condition) or whether they had performed the action/had the trait in the past week (the control condition). After this prime judgment, participants completed the target (intention) task, asking them to make a judgment about relevant future behaviors. An intention cue (“In the future I will”) appeared on the screen. After 2 seconds, an action corresponding to the action phrase shown earlier (e.g., “use napkins”) appeared below the intention cue. Participants then selected either “yes” or “no” to indicate whether they would perform the action in the future. The reaction times (RTs) for “yes” responses were measured and log-transformed for analysis (see Smallman, 2013, for other details of data processing). In the current research, we used the average reaction time for “yes” responses to the intention judgment for each prime statement.

Results

We calculated CP ratings for counterfactual prime statements by multiplying IL and TL ratings. Therefore, each prime statement (e.g., “using napkins”) has CP ratings from newly collected responses and intention RTs from Smallman (2013). We averaged CP ratings and archival log-transformed RTs for each prime statement and considered prime statement as a unit of analysis.

We expected that prime statements describing specific behaviors would lead to higher likelihood relative to those including categories of behavior and traits. We first regressed CP ratings on prime statement, coding specific behavior as 1 and the others as 0. As predicted, CP ratings on specific behaviors, M = 31.58, SD = 3.77, were significantly higher than the others, M = 30.31, SD = 4.04, b = 1.27, t(298)= 2.62, p = .01, 95% CI = [.32, 2.22]. Furthermore, we conducted a one-way ANOVA to test the distinctions of primes (specific behaviors, categories of behavior, and traits) in terms of CP. The primes differed in CP, F(2, 297) = 6.26, p = .002, η² = .04, such that specific behaviors, p = .001, 95% CI for difference = [.83, 3.02], and categories of behavior, M = 30.96, SD = 3.68, p = .02, 95% CI for difference = [.22, 2.40], were perceived as more likely than traits, M = 29.66, SD = 4.30. However, CP of specific behaviors did not differ from CP of categories of behavior, p = .27, 95% CI for difference = [−.48. 1.71].

We then examined the relationships between likelihood and intention judgment. We regressed RTs for intention judgments that had been preceded by counterfactual primes on CP. As predicted, more likely counterfactuals predicted faster intention formation, b = −.02, t(296) = −3.38, p < .001, 95% CI = [−.03, −.01]. We also performed this analysis for trials where the intention judgment had been preceded by a control rather than counterfactual prime. Notably, there was no significant relationship between CP and intention RTs preceded by control primes, b = .01, t(295) = 1.06, p = .29, 95% CI = [−.004, .01], indicating that the relationship is between CP and intention facilitation, not some other lexical property of the intention judgment that would have been the same following a control prime.

We also tested if likelihood mediated the effects of different prime statements on intention judgments (see Figure 6), using the same dummy variables with specific behavior as the comparison group. Significant indirect effects demonstrate to what extent the indirect effects of prime statements describing specific behaviors differ from the indirect effects of primes that contained categories of behavior and traits, respectively, to predict intention judgments via CP. There was a significant indirect impact of traits on intention RTs through CP, b = .03, SE = .01, 95% CI = [.01, .06]. However, the indirect effect of categorical behavior targets on intention RTs via CP was not significant, b = .01, SE = .01, 95% CI = [−.003, .03]. The findings suggest that, as predicted, counterfactuals involving traits are seen as less likely and subsequently make the relevant intentions less accessible relative to counterfactuals containing specific behaviors. Although categories of behavior (vs. specific behaviors) in counterfactuals did not yield a significant indirect effect on intention, the trend implies that counterfactuals containing categories of behavior tend to reduce likelihood compared with specific behaviors and thus decrease the accessibility of corresponding intentions.

Figure 6.

Mediation model in Study 4.

We next compared the indirect effect of specific behavior with the others that included both categories of behavior and traits. Compared with primes with categories of behavior and traits, prime statements that described specific behavior were perceived as more likely, b = 1.40, SE = .48, 95% CI = [.45, 2.35], and subsequently led to faster intention formation indirectly through CP, b = −.02, SE = .01, 95% CI = [−.04, −.004]. As before, we repeated these analyses for intention RTs for trials preceded by a control prime; the indirect effects did not differ by type of prime statements (specific behaviors vs. behavioral categories: b = −.003, SE = .004, 95% CI = [−.01, .005]; specific behaviors vs. traits: b = −.01, SE = .01, 95% CI = [−.03, .01]; specific behavior vs. the others: b = .01, SE = .01, 95% CI = [−.01, .02]). These outcomes indicate that likelihood increases when counterfactual content is more concrete, and increased likelihood facilitates intention formation. Moreover, this relationship is contingent on counterfactual thought, rather than lexical content, as control primes did not show this effect.

Discussion

Study 4 examined whether the perceived likelihood of a counterfactual influences the accessibility of relevant behavioral intentions. To eliminate potential demand characteristics or consistency bias that could have contributed to the results of Studies 1, 2, and 3, in Study 4 we analyzed data from two different sample groups using the same stimuli. We asked current participants to provide CP ratings for stimuli, and Smallman (2013) measured RTs of intention judgments after reading each stimulus. Therefore, each stimulus has an averaged CP rating generated in a novel participant sample, which cannot affect its (archival) intention RT.

Regarding the facilitating effect of likelihood on intention, we predicted that simulating more concrete counterfactuals would increase likelihood, and increased likelihood would lead to faster intention judgment. Consistent with this prediction, specific behavior primes predicted greater CP ratings compared with more abstract primes (categories of behavior and traits), and greater CP in turn produced faster intention judgments.

This study indicates that higher likelihood leads counterfactuals to better facilitate intention judgments. That is, individuals can more easily access relevant intentions when simulating more likely counterfactuals. As a result, they can readily connect past negative events to plans for taking better actions in the future. In addition, this study offers a new account of the mechanism of the influence of counterfactual content on intention judgment. Smallman (2013) suggested several possible mechanisms, such as increased sensitivity to relevant concrete information after negative events. However, this study provides evidence that likelihood perception of counterfactuals plays an important role in responding to behavioral intentions.

General Discussion

People often reflect on mistakes, mentally alter their behaviors, and imagine a better or worse outcome. Previous research has shown that imagining an alternative behavior that may have undone the past problem (upward counterfactual) can be functional by facilitating relevant intentions (e.g., Smallman & Roese, 2009). However, when a similar situation to the counterfactual actually happens in the future, some counterfactual thoughts are more salient than others, resulting in different functional effects of counterfactuals. In addition, little research has explored the effects of likelihood perception on counterfactuals’ functionality in spite of the fact that likelihood perception carries a great deal of weight in later judgment (Reyes et al., 1980; Sherman et al., 1985). The current research addresses that gap and explores a reason why the functionality of counterfactuals differs by showing that likelihood perception of counterfactuals, or counterfactual potency, mediates the functional effects of counterfactuals.

The results of the current research indicate that a counterfactual affects the relevant behavioral intention via likelihood perception of the counterfactual. Although previous research established that likelihood could have a diffuse functional effect via its effects on emotional states (Petrocelli et al., 2011), it was unclear how likelihood might influence intention judgments directly. The current results from the four studies suggest that likelihood is an important determinant for counterfactuals to be impactful on behavioral intentions. That is, merely generating a counterfactual does not lead this counterfactual to be functional. Instead, the more counterfactuals are perceived as likely, the more readily and strongly individuals form relevant behavioral intentions.

Furthermore, the present results provide the possibility that likelihood may serve as the most proximal determinant of counterfactuals’ effects on relevant behavioral intentions. Past research has identified various factors (e.g., counterfactual content, temporal distance, group membership) that affect the strength of counterfactual and relevant intention formation (Smallman, 2013; Smallman & McCulloch, 2012; Walker et al., 2016). In Studies 2 and 3, the results showed the mediating effects of CP in the relationship between mutability (varying in terms of norm violation and controllability) and intentions. Also, in Study 4, we found that likelihood mediated the different impacts of counterfactual content on intentions, which indicates that more concrete counterfactuals facilitate behavioral intentions because these counterfactuals are more likely to have occurred. Theoretically, the use of knowledge depends on its judged usability (appropriateness or relevance of using knowledge in a certain circumstance; Higgins, 1996), and CP contains the judged usability between a forgone cause and a desired outcome. Thus, close distance (Smallman & McCulloch, 2012) and ingroup membership (Walker et al., 2016) may have led to more functional counterfactuals because of higher likelihood perceptions of the counterfactuals. It should be noted, however, that evidence from the present research does not eliminate the possibility of CP having an indirect effect on intentions. Future research should further investigate how CP forms relevant intentions. For instance, some counterfactuals equally high in CP can bring about different levels of affective responses. If corresponding intentions were not altered by the different levels of affect following the counterfactuals, it would strengthen the direct role of CP in facilitating and endorsing relevant behavioral intentions.

This research also provides insights into reasons why counterfactuals are sometimes not functional. Some dysfunctional outcomes of counterfactuals could be due to low likelihood. For example, a counterfactual statement, such as “If I had studied more, I would have received a better grade,” can be functional. According to functional accounts of counterfactual thinking (Roese & Epstude, 2017), this counterfactual is likely to induce intentions to study more in the future. However, McCrea (2008) argues that a counterfactual can serve an excuse to self-handicap and enhance self-esteem at the expense of changing behavior (see also Petrocelli et al., 2012). If one simulates a counterfactual as an excuse, that counterfactual should entail an unlikely behavior or a situation that was difficult to control. That is, in the case of self-handicapping, the counterfactual might be rephrased as “I might have studied more but I was too busy,” which implies a low likelihood of the alternative behavior to have occurred. In other words, a counterfactual that is an excuse for the past outcome would be associated with low likelihood, and subsequently low likelihood would decrease both motivation and behavioral intentions for future performance. Future research should examine whether self-handicapping involves less likely counterfactuals.

One caveat is that high likelihood will not always bring about positive consequences. If a detrimental behavior (e.g., taking an unprescribed ADHD medication to study or gambling behavior) is expected to result in the desired outcome or positive consequences (e.g., a better grade or earning money), a highly likely counterfactual would increase the likelihood that an individual engages in this detrimental action. For instance, participants who generated counterfactuals about a vignette in which they imagined taking unprescribed ADHD medication to study late but ended up having side effects and receiving a terrible grade actually formed more favorable attitudes toward taking unprescribed ADHD medication (Ramos et al., 2016). This dysfunctional consequence may be because a counterfactual, such as “If only I had not taken the ADHD medication, my health and grade would have been better,” was perceived as unlikely. Instead, even after reading the negative scenario, participants may have believed that it is very likely that the medication would have actually resulted in a better grade without any side effects. Therefore, it is possible that high likelihood can promote dysfunctional behaviors.

There are a number of constraints on generality to consider in the current research. First, the participant sample was a predominately White, American undergraduate sample. Recent research has identified important cross-cultural differences in frequency of counterfactual thought due to cultural differences in fatalism and equifinality beliefs (Maitner & Summerville, 2022). Thus, the mediating processes for counterfactuals to affect behavior may also be culturally bounded in ways that the current sample cannot address. In addition, the current research involved predetermined counterfactuals provided by the experimenters. In natural settings, individuals can simulate multiple counterfactuals. Although past research often finds similar patterns for ratings of CP in experimenter-provided counterfactuals and in the content of spontaneous counterfactuals, the current results cannot directly speak to the processes involved in spontaneous counterfactual generation.

Simulating hypothetical worlds always involves likelihood perception. When an alternative action that in fact did not happen is likely to have occurred and be associated with a desired outcome, a similar circumstance in the future triggers this action. A counterfactual that contains an alternative action either unlikely to have happened or unlikely to have changed the outcome is disregarded as this action would not be beneficial. This research demonstrates that counterfactual thinking exerts its influences on behavioral intentions to varying degree according to likelihood. The perceived likelihoods of alternative worlds ensure that we do not repeat the same mistakes again and instead fulfill our unobtained goals in the future.

Supplemental Material

sj-docx-1-psp-10.1177_01461672221105958 – Supplemental material for The Effect of Counterfactual Potency on Behavioral Intentions

Supplemental material, sj-docx-1-psp-10.1177_01461672221105958 for The Effect of Counterfactual Potency on Behavioral Intentions by Woo J. Kim and Amy Summerville in Personality and Social Psychology Bulletin

Footnotes

Acknowledgements

We thank Brielle Johnson, Morgan Bodenstedt, Jess Boemker, Joey Canter, Emma Clark, Nick Contreras, Morgan Hapner, Abby Hart, Reagan Heck, Rachel Salerno, Will Sampson, Trevor Sass, Saidi Wadesisi, and Cindy Wang for their help with data collection and Rachel Smallman for sharing the archival data used in Study 4.

Authors’ Note

Studies 1 and 3 were part of the first author’s master’s thesis.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Study 3 was supported by Miami University via Thesis Research Grant awarded to the first author.

ORCID iDs

Woo J. Kim

Amy Summerville

Supplemental Material

Supplemental material is available online with this article.

Notes

References

Ajzen

(1977). Intuitive theories of events and the effects of base-rate information on prediction. Journal of Personality and Social Psychology, 35(5), 303–314. https://doi.org/10.1037/0022-3514.35.5.303

Anderson

J. R.

Milson

(1989). Human memory: An adaptive perspective. Psychological Review, 96(4), 703–719. https://doi.org/10.1037/0033-295X.96.4.703

Carroll

J. S.

(1978). The effect of imagining an event on expectations for the event: An interpretation in terms of the availability heuristic. Journal of Experimental Social Psychology, 14(1), 88–96. https://doi.org/10.1016/0022-1031(78)90062-8

Carver

C. S.

Scheier

M. F.

(1990). Origins and functions of positive and negative affect: A control-process view. Psychological Review, 97(1), 19–35. https://doi.org/10.1037/0033-295X.97.1.19

Fritz

M. S.

MacKinnon

D. P.

(2007). Required sample size to detect the mediated effect. Psychological Science, 18(3), 233–239. https://doi.org/10.1111/j.1467-9280.2007.01882.x

Galinsky

A. D.

Moskowitz

G. B.

Skurnik

(2000). Counterfactuals as self-generated primes: The effect of prior counterfactual activation on person perception judgments. Social Cognition, 18(3), 252–280. https://doi.org/10.1521/soco.2000.18.3.252

Higgins

E. T.

(1996). Knowledge activation: Accessibility, applicability, and salience. In Higgins

E. T.

Kruglanski

A. W.

(Eds.), Social psychology: Handbook of basic principles (pp. 133–168). Guilford Press.

Kahneman

Miller

D. T.

(1986). Norm theory: Comparing reality to its alternatives. Psychological Review, 93(2), 136–153. https://doi.org/10.1037/0033-295X.93.2.136

Kahneman

Tversky

(1982). The simulation heuristic. In Kahneman

Slovic

Tversky

(Eds.), Judgment under uncertainty: Heuristics and biases (pp. 201–208). Cambridge University Press.

10.

Lakens

Scheel

A. M.

Isager

P. M.

(2018). Equivalence testing for psychological research: A tutorial. Advances in Methods and Practices in Psychological Science, 1(2), 259–269. https://doi.org/10.1177/2515245918770963

11.

Maitner

A. T.

Summerville

(2022). “What was meant to be” versus “what might have been”: Effects of culture and control on counterfactual thinking. Journal of Personality and Social Psychology, 123(1), 1–27. https://doi.org/10.1037/pspa0000295.

12.

Markman

K. D.

Gavanski

Sherman

S. J.

McMullen

M. N.

(1993). The mental simulation of better and worse possible worlds. Journal of Experimental Social Psychology, 29(1), 87–109. https://doi.org/10.1006/jesp.1993.1005

13.

Markman

K. D.

McMullen

M. N.

Elizaga

R. A.

(2008). Counterfactual thinking, persistence, and performance: A test of the reflection and evaluation model. Journal of Experimental Social Psychology, 44(2), 421–428. https://doi.org/10.1016/j.jesp.2007.01.001

14.

McCrea

S. M.

(2008). Self-handicapping, excuse making, and counterfactual thinking: Consequences for self-esteem and future motivation. Journal of Personality and Social Psychology, 95(2), 274–292. https://doi.org/10.1037/0022-3514.95.2.274

15.

Myers

A. L.

McCrea

S. M.

Tyser

M. P.

(2014). The role of thought-content and mood in the preparative benefits of upward counterfactual thinking. Motivation and Emotion, 38, 166–182. https://doi.org/10.1007/s11031-013-9362-5

16.

Nasco

S. A.

Marsh

K. L.

(1999). Gaining control through counterfactual thinking. Personality and Social Psychology Bulletin, 25(5), 557–569. https://doi.org/10.1177/0146167299025005002

17.

Petrocelli

J. V.

Percy

E. J.

Sherman

S. J.

Tormala

Z. L.

(2011). Counterfactual potency. Journal of Personality and Social Psychology, 100(1), 30–46. https://doi.org/10.1037/a0021523

18.

Petrocelli

J. V.

Seta

C. E.

Seta

J. J.

Prince

L. B.

(2012). “If only I could stop generating counterfactual thoughts”: When counterfactual thinking interferes with academic performance. Journal of Experimental Social Psychology, 48(5), 1117–1123. https://doi.org/10.1016/j.jesp.2012.03.017

19.

Pezdek

Finger

Hodge

(1997). Planting false childhood memories: The role of event plausibility. Psychological Science, 8(6), 437–441. https://doi.org/10.1111/j.1467-9280.1997.tb00457.x

20.

Preacher

K. J.

Zyphur

M. J.

Zhang

(2010). A general multilevel SEM framework for assessing multilevel mediation. Psychological Methods, 15(3), 209–233. https://doi.org/10.1037/a0020141

21.

Ramos

A. M.

Becker

Biemer

J. A.

Clark

Fields

Smallman

(2016). The role of counterfactual thinking on attitudes toward ADHD medication use. Substance Use & Misuse, 15(4), 508–516. https://doi.org/10.3109/10826084.2015.1126739

22.

Reyes

R. M.

Thompson

W. C.

Bower

G. H.

(1980). Judgmental biases resulting from differing availabilities of arguments. Journal of Personality and Social Psychology, 39(1), 2–12. https://doi.org/10.1037/0022-3514.39.1.2

23.

Roese

N. J.

(1997). Counterfactual thinking. Psychological Bulletin, 121(1), 133–148. https://doi.org/10.1037/0033-2909.121.1.133

24.

Roese

N. J.

Epstude

(2017). The functional theory of counterfactual thinking: New evidence, new challenges, new insights. Advances in Experimental Social Psychology, 56, 1–79. https://doi.org/10.1016/bs.aesp.2017.02.001

25.

Sherman

S. J.

Cialdini

R. B.

Schwartzman

D. F.

Reynolds

K. D.

(1985). Imagining can heighten or lower the perceived likelihood of contracting a disease: The mediating effect of ease of imagery. Personality and Social Psychology Bulletin, 11(1), 118–127. https://doi.org/10.1177/0146167285111011

26.

Sherman

S. J.

Zehner

K. S.

Johnson

Hirt

E. R.

(1983). Social explanation: The role of timing, set, and recall on subjective likelihood estimates. Journal of Personality and Social Psychology, 44(6), 1127–1143. https://doi.org/10.1037/0022-3514.44.6.1127

27.

Smallman

(2013). It’s what’s inside that counts: The role of counterfactual content in intention formation. Journal of Experimental Social Psychology, 49(5), 842–851. https://doi.org/10.1016/j.jesp.2013.05.004

28.

Smallman

McCulloch

K. C.

(2012). Learning from yesterday’s mistakes to fix tomorrow’s problems: When functional counterfactual thinking and psychological distance collide. European Journal of Social Psychology, 42(3), 383–390. https://doi.org/10.1002/ejsp.1858

29.

Smallman

Roese

N. J.

(2009). Counterfactual thinking facilitates behavioral intentions. Journal of Experimental Social Psychology, 45(4), 845–851. https://doi.org/10.1016/j.jesp.2009.03.002

30.

Smallman

Summerville

(2018). Counterfactual thought in reasoning and performance. Social and Personality Psychology Compass, 12(4), Article e12376. https://doi.org/10.1111/spc3.12376

31.

Stanley

M. L.

Stewart

G. W.

De Brigard

(2017). Counterfactual plausibility and comparative similarity. Cognitive Science, 41(S5), 1216–1228. https://doi.org/10.1111/cogs.12451

32.

Summerville

Roese

N. J.

(2008). Dare to compare: Fact-based versus simulation-based comparison in daily life. Journal of Experimental Social Psychology, 44(3), 664–671. https://doi.org/10.1016/j.jesp.2007.04.002

33.

Teigen

K. H.

(1998). When the unreal is more likely than the real: Post hoc probability judgements and counterfactual closeness. Thinking and Reasoning, 4(2), 147–177. https://doi.org/10.1080/135467898394193

34.

Wakslak

Trope

(2009). The effect of construal level on subjective probability estimates. Psychological Science, 20(1), 52–58. https://doi.org/10.1111/j.1467-9280.2008.02250.x

35.

Walker

R. J.

Smallman

Summerville

Deska

J. C.

(2016). Motivated by us but not by them: Group membership influences the impact of counterfactual thinking on behavioral intentions. Social Cognition, 34(4), 286–305. https://doi.org/10.1521/soco.2016.34.4.3

36.

Zeelenberg

van der Pligt

de Vries

N. K.

(2000). Attributions of responsibility and affective reactions to decision outcomes. Acta Psychologica, 104(3), 303–315. https://doi.org/10.1016/S0001-6918(00)00034-2

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB