This is an open access article distributed under the terms of the Creative Commons
Attribution License (Power asymmetry exists in a wide range of interpersonal relationships, and people often experience situations in which they have power over others or vice versa (Overall et al., 2016; Smith & Hofmann, 2016). Since the seminal work by Kipnis (1972) that revealed the negative consequences of power asymmetry, especially on the powerless, scholars of diverse disciplines have investigated how power asymmetry results in negative social behaviours such as aggression (Overall et al., 2016), exploitation (Nikiforakis et al., 2014), and non-cooperation (Gordon & Puurtinen, 2021). In this research, we hypothesised that reputational concern would explain why power shapes such negative behaviours and attempted to experimentally establish the relationship between power and reputational concern, using two commonly used power priming methods.
Since Kipnis (1972), numerous studies have demonstrated that power leads individuals to display a wide range of disinhibited social behaviours that negatively influence the powerless and/or groups, including aggression, dehumanization, nonconformity to group norms, and exploitation (e.g., Gordon & Puurtinen, 2021; Keltner et al., 2003; Lammers & Stapel, 2011). Previous studies on reputation, however, have consistently demonstrated that such negative social behaviours often lead to establishing a negative reputation, which further results in negative reputational consequences such as ostracism and punishment (Feinberg et al., 2014). In other words, individuals are generally not motivated to display such behaviour since it would severely damage their reputation and their social lives (Wu et al., 2016).
Power has been operationally defined as having asymmetric control over a valued resource (Magee & Galinsky, 2008), the capacity to resist others’ influence (Lammers et al., 2016), the ability to control others’ outcomes or behaviours (Thibaut & Kelley, 1959), and the ability to control, modify, or influence others via rewards and punishments (Sivanathan et al., 2008). The powerful, by definition, can asymmetrically influence others, and even if they establish a negative reputation, it is unlikely that they will receive negative reputational consequences such as punishment; they are immune from negative reputational consequences. Thus, it can be hypothesised that power would liberate individuals from reputational concern (H1).
On the other hand, the powerless are susceptible to negative reputational consequences. Given that previous studies have found that the powerful often take advantage of and exploit the powerless (Gordon & Puurtinen, 2021; Hilbe et al., 2016), the powerless would be likely to be subject to negative behaviour by the powerful. Thus, it would be of vital importance for the powerless to actively seek a way to avoid receiving negative reputational consequences (i.e., establishing a positive reputation). Accordingly, we hypothesised that power would amplify reputational concern among the powerless, encouraging them to seek a positive reputation (H2).
Recent studies have provided relevant underpinnings to the predictions about the relationship between power and reputational concern; Petkanopoulou et al. (2019), for instance, found that the powerless anticipated negative consequences (e.g., establishing a negative reputation and potential punishments) for expressing anger. Consequently, they found that the powerless were less likely to directly express anger towards others compared to the powerful. More relatedly, Cai and Wu (2017) employed power manipulation and found that individuals in the high power priming condition felt less fear of receiving negative evaluation from others than those in the low power priming condition. Thus, these previous studies point to the presence of the relationship between reputational concern and power.
However, since these studies focused on the powerful vs. the powerless dichotomy, their findings cannot distinguish between the effect of being powerful (H1) and being powerless (H2). In other words, it remains uncertain to what extent the effect of being powerful (i.e., liberation from reputational concern) and being powerless (i.e., amplification of reputational concern) together explain the observed association between reputational concern and power. Thus, this research aimed to test the two hypotheses using two commonly used power priming methods with a control condition, which can separately address the hypotheses.
Study 1: Episodic Priming
In this study, we employed episodic power priming (Galinsky et al., 2003) to test the hypotheses. In this priming paradigm, participants are asked to recall and describe a situation in which they had power over others or someone else had power over themselves. This manipulation has been used in various studies and has successfully impacted a wide range of dependent variables (Galinsky et al., 2015; Sturm & Antonakis, 2015). The method has been favoured and employed in numerous studies for several reasons (Sturm & Antonakis, 2015); firstly, it is easy to implement and does not involve ethical concerns. Secondly, it can affect the sense of power without conscious awareness. Thirdly, these methods are thought to be superior to role induction-based manipulation (e.g., leader vs. follower/boss vs. subordinate role plays), as these methods are free from potential influences of participants’ expectations about how they are supposed to behave with an assigned role. For the apparent robustness and advantages of the technique, we used episodic priming for the first study.
We preregistered a brief description of the experimental procedure, a target sample size and its justification, data exclusion criteria, and the following operationalised hypotheses; H1: powerful priming would decrease reputational concern; H2: powerless priming would increase reputational concern. We preregistered them (see Supplementary Materials), and data, study material, analysis code, codebook, and supplementary results are available (see Supplementary Materials).
Method
We employed a single factor between-subject design with three levels (priming: high power vs. low power vs. control). An a priori power analysis revealed that 285 participants would be sufficient to detect an effect size of d = .37 (Fast et al., 2012) with a statistical power of .80 at alpha = .05. To account for any data exclusion, 300 university students were recruited in exchange for partial course credit, and 288 participants (Mage = 19.63, SD = 4.06, 236 females, 52 males) fully completed the study.
As a cover story, participants were informed that the study was designed to investigate the relationship between personality and interpersonal relationships. After giving consent, participants were randomly assigned to one of the three priming conditions: the high power, the low power, and the control conditions. Following Fast et al. (2012), participants in the high power condition were asked to recall and describe an event in which they had power over another individual or individuals. Those in the low power condition were asked to recall and describe an event in which someone else had power over themselves. In the control condition, participants were asked to recall and describe a social interaction during the previous day. In the high and low power conditions, we provided the following description of what power meant; “By power, we mean a situation in which you controlled the ability of another person or persons to get something they wanted or were in a position to evaluate those individuals.” To ensure that participants would seriously complete the recall task, we asked them to spend at least three sentences describing the event. Survey completion time and how long they spent on the recall task were recorded.
After the power manipulation, participants answered four questions measuring reputational concern (Wu et al., 2015; e.g., "I do not consider what others say about me," α = 77) and eight items measuring the sense of power (Anderson & Galinsky, 2006); e.g., "If I want to, I get to make the decisions," α = 80). We then introduced six items measuring participants’ perceived status (Yu et al., 2019) for exploratory purposes (see supplementary results). These items were all measured with a 7-point scale ranging from 1 = strongly disagree to 7 = strongly agree. Lastly, participants provided some demographic information (sex and age), and they were debriefed.
Results and Discussion
Following the preregistration, we excluded five participants based on their survey completion time (two standard deviations away from a mean completion time). A 1 x 3 (priming: high power vs. low power vs. control) between-subject ANOVA on reputational concern revealed that the effect of power priming was not significant, F(2, 280) = 1.01, p = .36, = .01. In addition, we created hypothesis-relevant contrasts: Contrast 1: high power vs. control; Contrast 2: low power vs. control. Planned contrasts yielded non-significant contrasts, ts < 1.42, ps > .16. Participants in the high power condition (N = 88, M = 5.27, SD = 1.11) did not experience significantly more reputational concern than those in the control condition (N = 105, M = 5.16, SD = 1.07). Similarly, those in the low power condition (N = 90, M = 5.37, SD = 0.93) did not experience significantly less reputational concern than those in the control condition. Thus, the hypotheses were not supported.
As a manipulation check, a 1 x 3 (priming: high power vs. low power vs. control) between-subjects ANOVA on the sense of power was conducted. Strikingly, the effect of power priming was not significant, F(2, 280) = 0.82, p = .44, = .01. Planned contrasts did not yield significant effects, ts < 1.06, ps > .29. Participants in the high power condition (N = 88, M = 4.50, SD = 0.89) did not have a significantly higher sense of power than those in the control condition (N = 105, M = 4.52, SD = 0.80). Similarly, those in the low power condition (N = 90, M = 4.64, SD = 0.69) did not have a significantly lower sense of power than those in the control condition. Thus, these results indicated that the power manipulation was not successful.
As a non-preregistered exploratory analysis, we took an opportunity to code text responses to the recall task to further examine whether the manipulation was, in fact, unsuccessful. Two independent coders inspected each text response, and we excluded five participants in the control condition whose recalled event involved power differences and/or did not involve social interactions, two participants, in the high power condition whose recalled event did not involve an experience of holding power over others, and four participants in the low power condition whose recalled event did not involve an experience of being powerless. In addition, they coded the valence of the recalled event (see supplementary results for more detail) for an exploratory purpose. With the data exclusion, we re-examined whether the sense of power and reputational concern significantly varied depending on the priming condition. Consistent with the analyses without the data exclusion, the effect of the manipulation was not significant for both (see supplementary results for more detail).
Finally, as a non-registered exploratory analysis, we computed a correlation between reputational concern and the measured sense of power, r = -.06, p = .30. Thus, despite the manipulation not being successful, the nonsignificant correlation suggests that a situationally induced sense of power is not significantly associated with reputational concern.
Episodic power manipulation has been widely used in previous studies (for reviews, see Galinsky et al., 2015; Sturm & Antonakis, 2015) in various languages (Hashimoto & Karasawa, 2022; Schmid & Mast, 2013; Scholl & Sassenberg, 2014). Nevertheless, we did not find evidence that episodic power priming significantly affected participants’ sense of power. Accordingly, we could not address whether power would liberate the powerful from reputational concern (H1) and whether it would amplify reputational concern among the powerless (H2). While some previous studies found that power manipulation indeed affected the sense of power (e.g., Fast et al., 2012), a recent preregistered study with sufficient statistical power reported that episodic power manipulation was unsuccessful (Heller & Ullrich, 2017). Thus, consistent with Heller and Ullrich (2017), our study casts doubt on the robustness of the manipulation.
That being said, it should be noted that, because it was an online survey, the intensity and depth of recollection might be insufficient. The median completion time of the recall task, which includes time for reading instructions, recalling an event, and describing the event, was 138.84 seconds. The instructions had 78, 74, and 41 words in the high power, low power, and control conditions, respectively. Given that Brysbaert (2019) found that English silent reading speed is 238 words per minute, it can be inferred that participants, especially those in the high and low power conditions, spent far shorter than two minutes recalling and describing the event. In addition, the study was advertised as a 6-min long survey, and this might have further pushed participants to quickly complete the task. Overall, the online nature of the survey might not sufficiently encourage individuals to engage in the recall task.
In addition, we would like to point out that our measurement of the sense of power is often used as a measurement of trait, rather than state, sense of power, while some previous studies showed that the power manipulation was successful using the same scale as a manipulation check (Fast et al., 2012). Thus, participants might have thought that the items were about their trait power, and this might have resulted in the nonsignificant effect of the manipulation, which was designed to influence situational power. Therefore, while our study, together with Heller and Ullrich (2017), casts doubt on the effectiveness of (online) episodic power priming, it would be desirable to first conduct an additional experimental study to overcome the methodological limitations in the present study (an issue we return to in Study 3)
Study 2: Semantic Priming
In Study 1, we employed one of the most used power priming methods, episodic priming. However, the manipulation was unsuccessful, presumably due to the online nature of the study, and we failed to address the hypotheses. Thus, in the present study, we chose another commonly used power priming method that would be less susceptible to the potential methodological limitation of the online episodic power priming discussed in the previous study: semantic power priming (Anderson & Galinsky, 2006; Galinsky et al., 2008; Mast et al., 2009). We preregistered the same hypotheses, a brief explanation of the study procedure, data exclusion criteria, and the target sample size (see Supplementary Materials). Data, study material, analysis code, codebook, and supplementary results are available in the Supplementary Materials.
Method
We employed a single factor between-subject design with three levels (priming: high power vs. low power vs. control). Following the rationale employed in Study 1, we recruited 300 British participants via Prolific Academic (Mage = 34.83, SD = 12.37, 208 females, 92 males).
After giving consent, participants were presented with 20 fragmented words (e.g., CONT_OLL_NG, _GALITARI_N, and OB_DIE_T) and asked to fill out the blanks and answer the completed words. Twenty words were relevant to dominance and subordination in the high and low power conditions, respectively. In the control condition, twenty words were power-neutral. These word fragments were used in Study 2 of Mast et al. (2009), and valence, word length, and word frequency were matched across three conditions (see Mast et al., 2009 for more detail).
After finishing the word completion task, participants answered the questions measuring reputational concern (α = .85) and the sense of power (α = .89). Participants then provided demographic information (age and sex) and were debriefed.
Results and Discussion
A 1 x 3 (priming: high power vs. low power vs. control) between-subject ANOVA on reputational concern revealed that the effect of power priming was not significant, F(2, 285) = 0.07, p = .94, < .001. Following Study 1, hypothesis-relevant contrasts were created: Contrast 1: high power vs. control; Contrast 2: low power vs. control. Planned contrasts indicated that these contrasts were not significant, ts < 27, ps > .78. Participants in the high power (N = 96, M = 4.99, SD = 1.07) and the low power (N = 96, M = 5.05, SD = 1.20) conditions did not experience significantly different levels of reputational concern compared to those in the control condition (N = 96, M = 5.01, SD = 1.08). Thus, the two hypotheses were not supported.
We then checked whether semantic power priming influenced the sense of power. A 1 x 3 (priming: high power vs. low power vs. control) between-subject ANOVA on the sense of power yielded a nonsignificant effect of semantic power priming, F(2, 285) = 0.48, p = .62, = .003. In addition, planned contrasts found that those in the high power (N = 96, M = 4.52, SD = 0.91) and the low power (N = 96, M = 4.43, SD = 1.00) conditions did not feel significantly different levels of sense of power than those in the control condition (N = 96, M = 4.38, SD = 0.95), ts < 0.96, ps > .34. Thus, semantic power priming was not successful.
Finally, as a non-preregistered exploratory analysis, we computed a correlation between the measured sense of power and reputational concern. Experimentally induced sense of power was not significantly associated with reputational concern, r = -.01, p = .85. In line with Study 1, this does not support our expectation that the sense of power is related to reputational concern.
Previous studies using semantic power priming reported that the manipulation successfully influenced dependent variables of different kinds, such as risk-taking behaviour (Anderson & Galinsky, 2006, Study 3), creativity (Galinsky et al., 2008, Experiment 1), interpersonal sensitivity (Mast et al., 2009, Study 2), and moral thinking (Lammers & Stapel, 2009, Study 1). In addition, semantic power priming in Study 2 should be less susceptible to the methodological limitations associated with online studies than episodic priming. Thus, it is striking that the present preregistered study did not influence reputational concern or the sense of power, even though we used one of the most elaborate sets of word fragments carefully piloted and developed by Mast et al. (2009). Given that the manipulation was unsuccessful in the present study, we again failed to address the hypotheses. We would like to note, however, that as in Study 1, the sense of power scale items were not worded in a way that explicitly referred to state rather than trait power, and this might have reduced the effect of the power manipulation on the sense of power.
Study 3
In Study 1, we employed episodic power priming (Galinsky et al., 2003), but the manipulation was not successful, presumably due to methodological limitations. Firstly, engagement in the recall task in Study 1 might not be sufficient, judging from the participants’ study completion time. Secondly, the items we used to measure the sense of power were arguably worded to refer to a trait rather than a situational sense of power. Thus, in the present study, we sought to modify the study to overcome the limitations and re-test our hypotheses with episodic power priming.
We first modified the experimental instruction for the recall task. In Study 1, we told participants to recall a past event, and participants were simply asked to describe the recalled event in at least three sentences. In order to make sure that participants’ engagement in the task would be sufficient, we asked them multiple questions regarding the recalled event (e.g., whom they were interacting with, where the event took place, emotions they experienced during the event, etc.) and asked them to provide as much detail as possible. In addition, Study 1 was advertised to take approximately 6 minutes in total, and this might make participants rush, leading to insufficient engagement in the task. In the present study, we thus recruited participants from Prolific, advertising it as a 10–15-minute-long survey.
Secondly, we modified the items to measure the sense of power. Körner et al. (2022) modified the instructions for the original scale so that they could measure a state rather than a trait sense of power. They did not change the wording of the original scale items, but they instructed participants, “Please tick the option that applied most to you at the moment.” We employed the same instruction. Additionally, we added “currently” to each scale item to make it clear that participants are asked about their current state (e.g., “Currently, I feel that I can get others to listen to what I say”).
As an extension, we introduced four items designed to measure how difficult/easy participants found it to recall the event (Lammers et al., 2017). In response to discussions on the effectiveness of priming that called for careful considerations of potential moderators (Cesario, 2014; Stroebe & Strack, 2014), Lammers et al. (2017) reported that measured ease of retrieval moderated the relationship between episodic power manipulation (high power vs. low power manipulation) and several dependent variables (e.g., confidence, disobedience, and unethical behaviour) such that the effect of episodic priming on those variables was reduced or nonsignificant among participants who found it difficult to recall the event. Thus, we measured ease of retrieval to exploratorily investigate if it moderates the effectiveness of the manipulation and its effect on reputational concern.
Prior to the data collection, we preregistered the hypotheses, study material, and full analysis codes, including data exclusion criteria and analytic strategies (see Supplementary Materials). Data, study material, analysis codes, codebook, and supplementary results are available in the Supplementary Materials.
Method
We employed a single factor between-subject design with three levels (priming: high power vs. low power vs. control) between-subject design. Following Study 1, we collected 300 participants from Prolific Academic, and we had 301 completed responses (Mage = 40.14, SD = 14.42, 147 females, 150 males). The number of male and female participants does not add up to the total sample size as there were participants who preferred not to indicate their sex or did not identify as male or female. Except for the above-mentioned changes, the present study was identical to Study 1. After giving consent, participants first completed the recall task and then they answered questions measuring their state sense of power (α = .91), reputational concern (α = .87), and ease of retrieval (α = .91). Ease of retrieval was measured with four items using a 7-point scale ranging from 1 = extremely difficult to 7 = extremely easy (Lammers et al., 2017). Lastly, they provided demographic information (sex and age) and were debriefed.
Results and Discussion
Following preregistration, we excluded 43 participants based on two criteria. Firstly, participants whose completion time was far from a median completion time by three median absolute deviations. Secondly, two independent coders examined text responses to the recall task and excluded those who did not recall an event involving social interactions. We also excluded those in the control condition who recalled an event involving power asymmetries (e.g., manager-subordinate interactions) and those in the high and low power conditions who recalled an event that did not involve holding power over others and lacking power, respectively. This left 87, 83, and 88 participants in the control, high power, and low power conditions, respectively (Mage = 39.97, SD = 13.86, 121 females, 133 males). We conducted a sensitivity power analysis, and with the final sample size, we should be able to detect a small-to-medium effect of = .04.
First, we conducted a 1 x 3 (priming: control vs. high power vs. low power) between-subject ANOVA on the situational sense of power and found a significant effect, F(2, 255) = 6.26, p = .002, = .05. As preregistered, we conducted planned comparisons using two dummy-coded variables: Contrast 1: high power vs. control; Contrast 2: low power vs. control. We found that while Contrast 1 was significant (t = 2.44, p = .02), Contrast 2 was not (t = -1.01, p = .31). Thus, participants in the high power priming condition (M = 4.77, SD = 0.94) felt significantly more powerful than those in the control condition (M = 4.37, SD = 1.08), and the high power manipulation was successful. Yet, those in the low power priming condition (M = 4.21, SD = 1.14) did not feel significantly less powerful than those in the control condition, and the low power priming was again unsuccessful.
Next, we ran a 1 x 3 (priming: control vs. high power vs. low power) between-subject ANOVA on reputational concern. The effect of priming was not significant, F(2, 255) = 0.22, p = .81, = .002. Planned contrasts also revealed that Contrast 1 and Contrast 2 were not significant, |ts| < .06, ps > .57. Participants in the high power condition (M = 5.10, SD = 1.17) and in the low power condition (M = 4.99, SD = 1.14) did not experience significantly different levels of reputational concern compared to those in the control condition (M = 5.00, SD = 1.25). Overall, given that the high power manipulation was successful, the results did not support H1. By contrast, the low power priming failed, and we could not sufficiently address H2.
We then computed a correlation between the measured sense of power and reputational concern, and we did not find a significant relationship, r = .05, p = .39. Consistently with Studies 1 and 2, the results did not support our expectation that the sense of power would be related to reputational concern. Ease of retrieval was not significantly correlated with reputational concern or sense of power, rs < .10, ps > .11.
Finally, we included the main effect of ease of retrieval and its interaction with priming in the previous ANOVA models and examined a moderating role of ease of retrieval reported by Lammers et al. (2017). The interaction effect on the sense of power was not significant, F(2, 252) = 0.37, p = .70, = .003. We then conducted the planned contrast analyses with the main effect of ease of retrieval and its interaction with the two contrasts included. Ease of retrieval did not significantly interact with Contrast 1 and Contrast 2 to predict the sense of power, |ts| < .09, ps > .52.
We found a significant interaction between ease of retrieval and priming on reputational concern, F(2, 252) = 6.21, p = .002, = .05. To follow up the interaction, as a non-preregistered analysis, we compared estimated marginal means of reputational concern in the control condition with those in the high and low power priming conditions at three levels of ease of retrieval: M = 5.66, and 1 SD (1.25) above and below the mean. Despite the significant interaction, we did not find significant comparisons, ps > 14. In other words, our hypotheses were not supported regardless of how easy it was for participants to recall an event. While Lammers et al. (2017) found that the recall task would be ineffective when ease of retrieval was relatively low, we did not find such results on reputational concern. It could be that the effect of ease of retrieval observed in Lammers et al. (2017) was specific to certain dependent variables (e.g., confidence, disobedience, and unethical behaviour). We report other exploratory analyses on ease of retrieval in online supplementary results.
In summary, in Study 3, we sought to overcome the methodological issues in Study 1, which arguably contributed to the failed manipulation. Namely, we encouraged participants to engage in the recall task in more depth and worded the sense of power scale items so that they explicitly reflected situational rather than trait power. In this study, the high power priming was successful. Yet, the low power manipulation was not. However, we would like to note that post-hoc pairwise comparisons revealed that those in the high power priming condition felt significantly more sense of power than those in the control condition and the low power priming condition, |ts| > 2.44, |ps| < .04. Thus, one could argue that the power manipulation was overall successful, and the source of the effect is those in the high power priming condition feeling more powerful rather than those in the low power priming condition less powerful. A majority of previous studies have predominantly focused on the high power vs. low power dichotomy, and the effect of lacking power is understudied (Schaerer et al., 2018). Correspondingly, previous studies that included both the control and low power condition are rare (Schaerer et al., 2018). Thus, given the paucity of research on the influence of lacking power, our data would be a valuable addition to the existing literature and calls for further research on the issue.
We failed to find support for the first hypothesis that the high power priming would reduce reputational concern. Yet, the low power manipulation was still unsuccessful, and we could not test the second hypothesis about the effect of the low power priming on reputational concern. Together with the absence of a significant correlation between reputational concern and a situationally induced sense of power, the results suggest that one’s own power would not be related to reputational concern.
General Discussion
In three studies with two commonly used power priming methods, we aimed to test the hypothesis that high and low power manipulation would reduce and increase reputational concern. Despite the widespread use, semantic priming in Study 2 failed to manipulate the sense of power, and it did not allow us to test the relationship between the sense of power and reputational concern. By contrast, while episodic manipulation in Study 1 was not successful, the modified version of high power episodic priming in Study 3 was partly successful, and we found that high power priming did not reduce reputational concern, inconsistently with our prediction (H1). Overall, these three studies yielded valuable insights not only into our understanding of the relationship between reputational concern and power but also into the effectiveness of power priming methods.
Power and Reputational Concern
Across three studies, we have obtained evidence against the meaningful relationship between one’s own power and reputational concern. However, we argue that it would be premature to conclude that power does not influence reputational concern.
Firstly, previous studies have pointed out that power can be construed as either an opportunity to control others or responsibility for others (Sassenberg et al., 2012; Scholl et al., 2018; Tost & Johnson, 2019). Our rationale behind the hypothesised relationship between holding power and reputational concern was that powerholders are not susceptible to negative reputational consequences and, therefore, they would not be as concerned about their reputation as their powerless counterparts. If participants in the high power condition construed power as responsibility, the fact that they are relatively free from negative reputational consequences would not be so salient that their level of reputational concern decreased.
Secondly, we focused on one’s own power (i.e., sense of power) and found no evidence for its link to reputational concern. We would like to point out that the effect of power can be broken down into two sources: the effect of holding and lacking power (i.e., one’s own power) and the effect of facing others holding and lacking power over oneself (target power). Given that we hypothesised that powerholders would be less concerned about their reputation because others could not punish them for their negative behaviour and reputation, it would be low target power rather than one’s own high power that is more closely related to reputational concern. Likewise, we predicted that the powerless would be highly concerned about their reputation because they are subject to reputational consequences from powerholders, and it is likely that high target power rather than one’s own low power would be associated with increased reputational concern. Strikingly, a recent systematic review by Feenstra et al. (2022) revealed that only 3.8% of the reviewed studies manipulated target power, and the potential role of target power is relatively understudied in the existing literature, calling for further studies.
Overall, it would be too early to draw any conclusions about the relationship between power and reputational concern. Given the ubiquity and the crucial role of reputation in social lives, the potential relationship between them deserves further scholarly investigation. More specifically, we argue that it is of vital importance to incorporate power construal and type of power (i.e., one’s own power vs. target power) into future studies.
Recently, scholars have started applying economic game paradigms to manipulate power (Gordon & Puurtinen, 2021; Harrell & Simpson, 2016; Sivanathan et al., 2008; van Dijk et al., 2020); in these studies, participants in the powerful condition are typically given the actual ability to control resources and punish others, whereas participants in the powerless condition are subject to asymmetric influence by the powerful. In other words, instead of manipulating participants' sense of power, this approach allows scholars to manipulate the actual power. In addition, such experimental settings would allow scholars to manipulate how power should be construed by participants. Thus, the hypothesised relationship between power and reputation would be better studied with these methods. As suggested by Sturm and Antonakis (2015), the manipulation of actual power with such methods also provides ecologically more valid evidence as to how power shapes human cognition and behaviour. Therefore, departing from the commonly used power priming methods would be a future avenue for research on the relationship between power and reputation.
Power Manipulation
Episodic and semantic priming methods have been dominant research methods in the field of research on power for several reasons, and a number of studies have, in fact, reported that these methods successfully influenced a wide range of cognition, affects, and behaviours (Sturm & Antonakis, 2015). In Study 1, episodic power priming was unsuccessful, which is consistent with Heller and Ullrich (2017). Yet, it was successful when we overcame methodological limitations that arguably contributed to the failure of the manipulation (i.e., insufficient engagement in the recall task and wording of the scale items) in Study 3. Thus, our results, together with Heller and Ullrich (2017), place caveats on the use of the paradigm, suggesting that researchers should carefully design the task, especially when it is executed online.
To our knowledge, Heller and Ullrich (2017) were the first to publish a null effect of episodic power priming with a well-powered study. Our studies were also sufficiently powered and preregistered. These recent findings suggest that the most commonly used methods to induce the sense of power do not always work. Given the ample discussions on the effectiveness of priming methods in general (Cesario, 2014; Hoogeveen et al., 2018; Stroebe & Strack, 2014; van Elk & Lodder, 2018), our research calls for a comprehensive systematic review of the effect of the power priming manipulations. This would help us identify what conditions should be met so that power priming effectively influences the sense of power and, it would also reveal how much publication bias contributes to the seemingly robust effect of power priming (Ferguson & Brannick, 2012; Rosenthal, 1979) in the power priming literature.
Conclusions
In summary, we sought to establish the relationship between one’s own power and reputational concern, but, overall, we did not find supporting evidence across three preregistered studies. Having said that, our research has yielded valuable insights into the relationship between power and reputational concern and left promising future directions to further elucidate the relationship. Since power asymmetries are ubiquitous and reputation plays a pivotal role in social lives, we believe that the current research will invite further studies and discussion on the issue. In addition, our preregistered studies added valuable data to the recent scholarly debate on the effectiveness of power priming methods and will contribute to cumulative science on the topic.