Abstract
Causes of behavior are often classified as either dispositional (e.g., personality) or situational (e.g., circumstances). However, the disposition–situation dichotomy confounds locus (internal vs. external) and stability (unstable vs. stable) of attribution, rendering it unclear whether locus or stability drives changes in dispositionality. In the present research, we examine the dispositional shift—that is, psychologically distant (vs. near) events are attributed to dispositional (vs. situational) causes. Using construal level theory, we hypothesize that the dispositional shift is caused by a change in stability (but not necessarily locus) of attribution. Two experiments support this hypothesis. In Experiment 1, distant (vs. near) future events were attributed to more stable causes. In Experiment 2, actions by a socially distant person (vs. oneself) were also attributed to more stable (but also more internal) causes. Thus, important psychological manipulations, here psychological distance, can influence causal dimensions selectively, supporting the independence of stability and locus of attribution.
Keywords
How people respond to an event often depends on what they think has caused the event. For example, a person who thinks poverty is caused by poor people’s habitual laziness will be less likely to fight poverty compared to a person who thinks poverty is caused by unjust wages (Weiner, Osborne, & Rudolph, 2011). Psychologists therefore have spent great effort to understand the processes underlying causal attribution (Reeder, 2013). Crucially, research on causal attribution has been conducted with two different foci (Kelley & Michela, 1980). On the one hand, antecedent-focused research aims at understanding how people cognitively construe explanations of events. In other words, such attribution research explains the cognitive underpinnings of the path from events to its attributed cause. On the other hand, consequence-focused research aims at understanding the various consequences of attributions once they are made. Hence, these theories aim at explaining the path from attributions to emotions, motivations, and behavior. The present research concentrates on antecedent-focused attribution research.
Research on the antecedents of attribution, at least from the 1970s onward, has often been based on a dichotomy that conceptualizes attribution as being either dispositional (e.g., personality of the actor) or situational (e.g., social pressure). For example, Heider (1958) postulated that humans prefer to identify dispositional causes of behavior. Dual-process theories of attribution contrast dispositional attributions, which are postulated to be automatic, with effortful corrections toward situational attributions (e.g., Gilbert, 1989). Moreover, in naive theories, positive outcomes are attributed predominantly to situations, while negative outcomes are attributed predominantly to dispositions (Ybarra, 2002). The disposition–situation dichotomy is usually defined as a difference in locus of attribution with the actor’s skin as boundary; that is, any cause internal to the actor is considered dispositional and any external cause is considered situational (e.g., Malle, 2006; Storms, 1973). For example, the fundamental attribution error has been defined as “people’s behavior is caused externally (by situations) rather than internally (by dispositions)” (Sabini, Siepmann, & Stein, 2001, p. 1). Similarly, Uleman, Saribay, and Gonazales (2008) use the term internal–external dichotomy as a synonym for disposition–situation dichotomy.
At the same time, dispositional causes are assumed to be stable while situational causes are assumed to be unstable (e.g., Jones & Nisbett, 1972; Pronin & Ross, 2006); more specifically, “[d]ispositional properties are the invariances that make possible a more or less stable, predictable, and controllable world. They refer to the relatively unchanging structures and processes” (Heider, 1958, p. 80). Put another way, “chronic dispositional judgments refer to those explanations that attribute behavior to chronic, invariant causes which remain with the person indefinitely” (Costabile, 2011, p. 447). The fact that dispositions are typically assumed to be both internal and stable also emerges in definitions of the actor–observer effect. As an example, Liberman and Trope (2008) suggest: “people’s explanation of their own behavior emphasizes concrete situational factors that operate at the moment of action, whereas their explanation of others’ behavior emphasizes stable and personal dispositions” (p. 1203; emphasis added). In sum, the disposition–situation dichotomy confounds locus and stability of attribution (for discussions of this problem, see also F. D. Miller, Smith, & Uleman, 1981; Robins, Spranca, & Medelsohn, 1996).
Dissecting Dispositionality
Although antecedent-focused attribution research has often regarded internal causes as necessarily more stable than external causes, there are at least two reasons why stability and locus of attribution are, in principle, orthogonal. First, several well-established psychological constructs can be considered potential inner causes of behavior but are at the same time unstable, for example, brief emotional bursts and transient states of fatigue. Likewise, several situations are considered powerful determinants of behavior but at the same time can be highly stable, for example, enduring work pressure, long-term unemployment, and lifetime imprisonment.
Second, consequence-focused attributional research has long recognized that causes of actions vary along several dimensions. The most widely used dimensional model in this area of research distinguishes locus (internal to external), stability (stable to unstable), and controllability (controllable to uncontrollable; Weiner, 1985; for a different dimensional model, see Abramson, Seligman, & Teasdale, 1978). 1 Crucially for our argument, evidence suggests that even with the cause’s locus held constant, emotional, motivational, and behavioral consequences of attribution can differ vastly depending on a cause’s stability and controllability. For example, in the case of poverty, both lack of effort and a physical handicap are internal causes but—differing in their stability and controllability—still elicit very different reactions (Weiner et al., 2011).
Based on these arguments, we suggest that stability and locus may serve independent functions—not only when influencing consequences of attributions, but also for person perception, examining antecedents of attributions. Accordingly, locus and stability might be independently influenced by psychologically meaningful manipulations. Surprisingly, however, research in person perception has hardly ever examined attribution dimensions independently. In the present research, we use psychological distance as a factor that has the potential to selectively influence attribution dimensions. More specifically, we test the hypothesis that psychological distance will influence stability, but not necessarily locus, of attribution.
The Dispositional Shift
Psychological distance has been found to induce a dispositional shift: More (vs. less) psychological distance from an event increases the rate of dispositional (vs. situational) attributions—no matter whether the distance is temporal, spatial, or social. Temporally distant (vs. near) events have led to more dispositional attributions (see also Burger & Rodman, 1983; Funder & van Ness, 1983; Moore, Sherrod, Liu, & Underwood, 1979; Nussbaum, Trope, & Liberman, 2003; Peterson, 1980; Pronin & Ross, 2006; but see D. T. Miller & Porter, 1980; Sanna & Swim, 1992). Similarly, events occurring at a distant (vs. near) location elicited more dispositional attributions (Henderson, Fujita, Trope, & Liberman, 2006). And events happening to another person (vs. oneself), and thus at greater social distance, also increased dispositional attributions—called the actor–observer effect (e.g., Green & McClearn, 2010; Jones & Nisbett, 1972; Nisbett, Caputo, Legant, & Marecek, 1973; Semin & Fiedler, 1989; Storms, 1973; cf. Malle, 2006).
A prominent explanation for the dispositional shift (see Henderson et al., 2006; Nussbaum et al., 2003) is based on construal level theory (CLT; Trope & Liberman, 2010). CLT posits that different kinds of distance (time, space, social distance, and hypotheticality) have similar effects on psychological processes (Trope & Liberman, 2010). More specifically, psychological distance is postulated to increase abstract, high-level construals. “[T]he […] function of high-level construals is to enable people to mentally transcend the here and now by forming a representation consisting of the invariant features of the available information” (Trope & Liberman, 2010, p. 450). Thus, distance is postulated to increase abstractness of information processing, characterized by a focus on central and temporally enduring features, as “in practice, details are too variable and unpredictable, and we, therefore, resort to abstraction, sacrificing detail to achieve stability and predictability” (Liberman, Trope, & Rim, 2011, p. 144). During causal attribution, therefore, high distance should lead to a focus on temporally stable features of an action. Moreover, as dispositions are assumed to be more stable than situations, CLT predicts that high psychological distance will increase dispositional compared to situational attributions (Trope & Liberman, 2010).
Up to now, research on the dispositional shift has relied on the coarse dispositional–situational distinction. Thus, it remains unclear whether the dispositional shift is primarily driven by a change in locus, stability, or both. To gain insight into the processes driving the dispositional shift, the present research examined the influence of psychological distance on causal dimensions of attribution separately. Specifically, causal attribution was measured using the Revised Causal Dimension Scale (McAuley, Duncan, & Russell, 1992), which assesses locus, stability, and internal and external controllability of causes. 2 CLT postulates that psychological distance, via abstraction, leads to a focus on enduring aspects of an action. Thus, the dispositional shift should result from a change in stability of attribution. After all, CLT claims that abstract representations are more stable—not that abstract representations are more internal. Therefore, we hypothesize that high (vs. low) psychological distance will lead to attributions to more stable (vs. unstable) causes. For locus and controllability, we do not see that CLT would predict different attributions for different degrees of psychological distance, as neither internal compared to external nor controllable compared to uncontrollable causes typically differ in their abstractness.
Experiment 1
Experiment 1 tested the CLT-based hypothesis that temporally distant events are attributed to more stable causes than temporally near events. For this, we asked participants to imagine events either in the near or distant future and to provide explanations for these events.
We report all measures and manipulations in both Experiments 1 and 2. There were no data exclusions. Sample size was determined before any data analysis, and all experiments were preregistered. Materials and data as well as links to the preregistrations can be found at https://osf.io/g 5 xhm/
Method
Power Analysis
We conducted an a priori power analysis using G*Power (Faul, Erdfelder, Buchner, & Lang, 2009) using a preliminary study (of which Experiment 1 is an exact replication) to estimate the effect size. For a two-tailed t test between two independent means with d = .63, α = .05, and 1 − β = .90, the power analysis yields a required sample size of N = 110 (preregistration 3 : https://osf.io/ksyzu).
Participants and Design
In exchange for monetary compensation, 120 people (80 female and 40 male; aged 18–65 years, M age = 27 years, SD age = 10 years) participated. Participants were randomly assigned to either low or high temporal distance in a 2 (temporal distance: near vs. distant; between) × 4 (causal dimension: stability, locus, personal controllability, external controllability; within) mixed-model design.
Procedure
Participants were asked to imagine themselves experiencing eight different situations (taken from Peterson et al., 1982) consecutively. Each situation was to happen either in the near future (between in about 1 day and about 10 days) or in the distant future (in about 1 to about 4 years).
More specifically, participants first read a one-sentence description of a situation. Depending on the experimental condition, participants were instructed to imagine this situation to take place either in a few days (e.g., In 2 days, you will meet a friend who will compliment you on your appearance) or in a few years (e.g., In 2 years, you will meet a friend who will compliment you on your appearance). Participants were then asked to indicate the main cause for the outcome of the situation (e.g., What do you think will be the main cause for your friend's complimenting you on your appearance?). After stating it, participants rated this main cause using the Revised Causal Dimension Scale (McAuley et al., 1992). More specifically, using 9-point rating scales, participants rated the cause on four dimensions—temporal stability (temporary to permanent, variable over time to stable over time, changeable to unchangeable); locus (e.g., inside of you to outside of you); personal controllability (e.g., not manageable by you to manageable by you); and external controllability (e.g., not under the power of other people to under the power of other people) with 3 items for each dimension. For each situation, 12 evaluations were performed on two consecutive screens, locus and stability items on the first screen and personal and external control items on the second screen. Participants could switch freely between the two screens and alter any value until they moved on to the next situation. This should avoid one of the dimensions dominating responses (by preventing one dimension being evaluated first without considering the other dimensions).
After evaluating the causal dimensions of the first situation’s cause, the next situation description ensued. Participants completed the sequence of reading the description, naming the main cause and evaluating the main cause for eight different situations (half of them positive, the other half negative) in random order (situation descriptions are taken from Peterson et al., 1982, selected by Nussbaum et al., 2003).
After participants completed all eight situations, as a manipulation check for the temporal distance, they were asked to recall when the situations were to have taken place. Specifically, participants were asked Please indicate within which time frame you were supposed to imagine the situations happening by choosing between the two options within the next few days and in one year at the earliest. Lastly, participants stated their age and gender.
Results and Discussion
The manipulation check indicated that participants were generally aware of the temporal distance within which they were to imagine the situations. Eighty-seven percentage of the participants chose the distant option (in 1 year at the earliest) in the temporally distant condition, while 14% chose this option in the temporally near condition, χ2(1) = 64.5, p < .001. 4
For each causal dimension, mean attributions were calculated across all situations. Supporting our hypothesis, stability of attribution was higher for participants in the temporally distant condition than for participants in the temporally near condition, t(118) = 2.81, p = .006, d = .51, 95% CI [0.15, 0.88]; . Thus, high compared to low distance led to a medium-sized increase in stability of attribution.
An exploration whether any of the other causal dimensions were also affected by temporal distance yielded no significant effects, locus: t(118) = 1.52, p = .13, d = .28, 95% CI [−0.08, 0.64]; personal control: t(118) = 0.38, p = .70, d = .07, 95% CI [−0.29, 0.43]; external control: t(118) = 1.12, p = .27, d = .20, 95% CI [−0.16, 0.56]; . Moreover, a 2 (distance: near, distant) × 4 (causal dimension: locus, stability, personal control, external control) analysis of variance (ANOVA) yielded a significant interaction between distance and causal dimension, F(3, 116) = 2.96, p = .035,
Means (and Standard Deviations) for Attribution Dimensions in Experiment 1, Separate for High and Low Temporal Distance.
The correlations between the different attribution dimensions were mostly small to medium-sized, see Table 2. 5 Specifically, the two target dimensions, stability and locus, were weakly positively correlated, sharing about 6% common variance.
Correlation Coefficients (and p values) Between the Attribution Dimensions in Experiment 1 Separately for High and Low Temporal Distance.
As mentioned when explaining the power analysis, Experiment 1 was the direct replication of an initial test of the present hypothesis. These two experiments constitute all experiments varying temporal distance that we performed. Even though we did not preregister a combined analysis of both experiments, this combined analysis, by its increased power, leads to more precise estimates. A 2 (distance: near, distant) × 4 (causal dimension: locus, stability, personal control, external control) ANOVA including participants from both Experiment 1 and the pretest yielded a significant main effect of causal dimension, F(3, 225) = 118.64, p < .001,
Means (and Standard Deviations) for Attribution Dimensions in the Combined Analysis of Experiment 1 and the Pretest, Separate for High and Low Temporal Distance.
Thus, for stability, we observed a medium-sized effect, whereas for the other causal dimensions, the effect size estimates are very small, and even the upper bounds of the 95% CI for these effect sizes are clearly below the medium-sized effect we observe for stability of attribution. In sum, the present results suggest that high (vs. low) distance increases attribution to stable causes, while there is no evidence that distance affects the other causal dimensions. Thus, we find first evidence that stability and locus of attribution can vary independently as they are differentially influenced by the present distance manipulation.
Experiment 2
Experiment 2 replicated Experiment 1 using a different distance dimension, namely social distance. Participants made causal attributions imagining events either happening to themselves (low social distance) or to an acquaintance (high social distance). As CLT predicts analogous psychological changes for all distance dimensions, applying CLT to the present context results in the hypothesis that participants will attribute events to more stable causes for high compared to low distance.
Method
Power Analysis
An a priori power analysis using G*Power (Faul et al., 2009) for a two-tailed t test between two independent means with α = .05, 1 − β = .90, and d = .65 6 yields a required sample size of N = 102 (preregistration: https://osf.io/xntfy).
Participants and Design
In exchange for monetary compensation, 102 people (85 female and 17 male; aged 18–47 years, M age = 23 years, SD age = 5 years) participated. Participants were randomly assigned to a social distance condition in a 2 (social distance: low vs. high; between) × 4 (causal dimension: stability, locus, personal controllability, external controllability; within) mixed-model design.
Procedure
Social distance was manipulated by participants’ doing the attribution task either for themselves (low social distance) or for an acquaintance (high social distance). Specifically, for the high social distance condition, participants were first asked to provide the name of a person in their wider social circle. They were asked to choose a person of their own gender whom they did not know well and did not often interact with. This name was included in the situation descriptions (e.g., high social distance: Claudia meets a friend who compliments her on her appearance; low social distance: You meet a friend who compliments you on your appearance). Additionally, items from the Revised Causal Dimension Scale were adapted so that the internal locus anchor and personal control anchor always referred to the target person (in the preceding example, the person who receives the compliment; the anchors of the sample locus item would be, for low distance, inside of you and outside of you, and for high distance, inside of Claudia and outside of Claudia). In contrast to Experiment 1, no point in time was mentioned for the events to occur. Apart from these differences, the procedure of the attribution task was identical to Experiment 1.
After completing all eight attributions, participants in the high social distance condition were asked to name their relation to the target person and to rate their closeness on a 9-point scale ranging from 1 (not at all close) to 9 (very close). Finally, they provided demographic information, were thanked, and payed.
Results and Discussion
Participants in the high social distance condition adhered to the instructions and chose target persons they were not especially close to (e.g., their fellow students, distant friends, colleagues). The mean score of the closeness rating was M = 3.63 (SD = 2.03). 7
As in Experiment 1, mean attributions were calculated for each causal dimension across all situations. Supporting our hypothesis, stability of attribution was higher for high social distance than for low social distance, t(100) = 4.33, p < .001, d = .86, 95% CI [0.45, 1.26]. Thus, high compared to low distance led to a large increase in stability of attribution.
An exploration whether any of the other causal dimensions were also affected by temporal distance yielded a significant effect for locus of attribution, with more internal attributions for high social distance compared to low social distance, t(100) = 2.39, p = .019, d = .47, 95% CI [0.08, 0.87]. The differences between high and low social distance for controllability did not quite reach significance; personal control: t(100) = 1.87, p = .064, d = .37, 95% CI [−0.02, 0.76]; external control: t(100) = 1.63, p = .11, d = .32, 95% CI [−0.07, 0.71]. A 2 (distance: near, distant) × 4 (causal dimension: locus, stability, personal control, external control) ANOVA yielded a significant interaction between distance and causal dimension, F(3, 98) = 5.55, p = .001,
Means (and Standard Deviations) for Attribution Dimensions in Experiment 2, Separate for High and Low Social Distance.
For correlations between the different attribution dimensions, see Table 5. The shared variance between the two target dimensions, locus and stability, was 16% in the low distance condition and 49% in the high distance condition, which is considerably more than in Experiment 1.
Correlation Coefficients (and p values) Among the Attribution Dimensions in Experiment 2 Separated for High and Low Social Distance.
In both Experiments 1 and 2, distance increased stability of attribution. However, as a result of the social distance manipulation, locus of attribution also varied, becoming more internal with high compared to low distance, while we found no difference in locus of attribution for the temporal distance manipulation. One possible explanation for this difference is that the operationalization in Experiment 2 entails, in addition to a change in distance, a change in perspective (from first to third person), constituting an actor–observer manipulation. Accordingly, a change in person could plausibly lead to a change in locus of attribution, for example, because of differing naive theories about the causation of own versus other people’s behaviors (e.g., Chiu, Hong, & Dweck, 1997). Another possible reason is that social distance might, independently of a change in person, influence locus of attribution whereas temporal distance does not influence locus of attribution. On the other hand, this unpredicted effect might be spurious and therefore needs to be replicated before drawing conclusions on whether or not our manipulations for temporal and social distance indeed influence locus of attribution differently.
General Discussion
The cognitive underpinnings of the path from behaviors to causal attribution are usually measured and theorized about using the dichotomy between dispositional and situational causes. However, this dichotomy confounds locus and stability of attribution. Questioning the validity of this confound, the present experimental observations suggest locus and stability of attribution to be at least partially dissociable. While the two dimensions were correlated (their shared variance varied between 6% and 49%), they responded differentially to experimental manipulations. More specifically, we observed a medium-sized effect of temporal distance on stability of attribution (more stability at higher distance), but no significant effect on locus. In contrast to temporal distance, social distance significantly affected both stability and locus of attribution, with more internal attributions for observers (high distance) compared to actors (low distance). However, the effect on locus was considerably smaller than the effect on stability.
Taken together, we observed locus and stability to be differentially sensitive to manipulations of psychological distance, which indicates that the two attribution dimensions are at least partially independent. Therefore, we suggest, when examining antecedents of causal attribution, locus and stability should generally be measured independently. The present results show that the distinction between stability and locus of attribution is not only conceptually possible but also psychologically meaningful.
The dissociation between stability and locus is also important for testing mechanisms and theories of attribution. More specifically, the present results clarify previous research on the dispositional shift, the observation that high (vs. low) distance leads to more dispositional (vs. situational) attributions. The dispositional shift has previously been interpreted as resulting from a change in locus of attribution (e.g., Pronin & Ross, 2006). However, when measuring dimensions of causal attribution separately, we find that locus of attribution was only influenced by social distance but not by temporal distance, whereas stability of attribution was influenced by both social and temporal distance. With this, the present work provides the first evidence for a common mechanism of the dispositional shift for different forms of distance. However, contrary to the assumption that the dispositional shift is caused by a change in locus of attribution, we found that the dispositional shift was caused by a change in stability of attribution—more distant events were attributed to more stable causes.
The present results support CLT. CLT predicts that increasing distance focuses attention on central and temporarily invariant features of the event or object (Trope & Liberman, 2010). Thus, CLT specifically predicts attributions to more stable causes with high compared to low distance. Therefore, by supporting the specific predictions derived from CLT, the present results lend additional support to CLT. Another attribution dimension that is interesting from a CLT perspective is globality—a reason can be relatively global (influencing many outcomes) or specific (confined to few outcomes). Globality is similar to stability, with the latter referring to invariance (vs. variability) over time in contrast to invariance across multiple situations at the same time. CLT postulates that abstract representations are both more global and more stable (Trope & Liberman, 2010). Accordingly, while we did not measure globality, because it is neither represented in the causal dimension model (Weiner, 1985) nor in the Causal Dimension Scale (McAuley et al., 1992), we would nevertheless expect distance to affect both stability and globality.
The present results are also relevant for research on mental state inferences, a more recent trend in attribution research. Research on mental state inferences examines attributions to specific internal causes, for example, to goals in contrast to traits (Korman & Malle, 2016; Malle, 2011; Olcaysoy Okten & Moskowitz, 2018; Reeder, 2009; van Overwalle, van Duynslaeger, Coomans, & Timmermans, 2012; for a review see Moskowitz & Olcaysoy Okten, 2016) From a causal dimension perspective, goals and other mental states are internal and unstable, while traits are internal and stable. Thus, the question whether a behavior is attributed to goals or traits can be seen as the question whether an internal attribution is unstable or stable.
For mental state inferences, the present results suggest that increasing psychological distance should increase trait compared to goal attributions. This has indeed been observed (Rim, Uleman, & Trope, 2009; see also Idson & Mischel, 2001). However, the two approaches—mental state attributions and the present approach using causal dimensions—differ in the scope of causes they examine. While mental state inference paradigms are confined to specific internal causes, the dimensional model of attribution examines both internal and external causes. Thus, the present approach has the broader scope. Returning to the causes of poverty, for example, we would expect high distance (compared to low distance) to lead to poverty being attributed to more stable external causes (e.g., the labor market) and less to unstable external causes (e.g., bad luck).
Footnotes
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) received no financial support for the research, authorship, and/or publication of this article.