How other people see us—as warm or cold, as competent or incompetent—very much determines how they evaluate and treat us across all domains in life. Hence, self-presentation, the process by which we try to control our social images, constrains our behaviour in virtually all social encounters and is an essential feature of human life (Goffman, 1959; Leary, 1995). In the present research, we examine how people use the dynamics between two fundamental dimensions of social perception, warmth and competence, when self-presenting to reach an important goal.
Research has long noted the presence of two key dimensions in our perceptions and judgment of others (e.g., Rosenberg, Nelson & Vivekanathan, 1968; Zanna & Hamilton, 1972). The dimensions have been postulated under different names, including agency and communion (Wojciszke, 2005) and intellectual versus social goodness (Rosenberg et al., 1968; for an overview, see Abele & Wojciszke, 2014), but with essentially identical content. The current research builds on findings from the social perception literature (e.g., Fiske, Cuddy, Glick & Xu, et al., 2002), which typically uses the distinction between warmth, having to do with perceived intent, comprising friendliness and trustworthiness, and competence, which is related to perceived ability, including intelligence and skill (Abele, Cuddy, Judd & Yzerbyt, 2008). The two dimensions are used universally when people differentiate others, both at the individual and at the group level (Abele & Wojciszke, 2007, 2014; Fiske, Cuddy & Glick, 2007; Judd, James-Hawkins, Yzerbyt & Kashima, 2005; for recent reviews, see Fiske, 2015; Yzerbyt 2016). Whereas warmth and competence were long conceived of as positively related (Rosenberg, et al., 1968), recent efforts suggest that they often manifest a negative relationship. Specifically, a group or a person described as high in competence is frequently assumed to be less warm than another person or group described as low in competence; whereas, a target described as low in warmth is often seen as more competent than a target high in warmth. Yzerbyt, Provost, and Corneille (2005) called this tendency to differentiate social targets by contrasting them on the two dimensions the compensation effect (for reviews, see Kervyn, Yzerbyt & Judd, 2010; Yzerbyt, 2018).
Although research has demonstrated compensation when we form judgments of others, less is known about its use in self-presentation. Intuitively, one might assume that we would try to maximize both how warm and how competent we appear. However, given the trade-off between warmth and competence in social perception, it seems reasonable that this compensation could be utilized when we cultivate impressions of ourselves. Thus, in some situations, people may attenuate their warmth to appear competent, and in other settings, they may downplay their competence to appear warm.
Indeed, in a set of recent studies, Fiske and colleagues provided evidence suggesting compensation effects in self-presentation. Holoien and Fiske (2013) asked participants either to make a warm or a competent impression of themselves on others. Participants instructed to appear warm displayed lower competence than a control group; whereas, those encouraged to make a competent impression downplayed their warmth. Further, Swencionis and Fiske (2016) investigated participants’ self-presentation when imagining an interaction with a lower- or higher-status target. Participants imagining interacting with the low-status target downplayed their competence relative to their warmth; whereas, those in interaction with a higher-status target decreased their warmth relative to competence.
While these studies provide evidence that people may sometimes engage in self-presentational compensation, they lack a critical feature of many everyday self-presentations: a strategic self-presentational goal. Specifically, in these studies, participants were merely assigned the goal to convey a warm or competent self-image (Holoien & Fiske, 2013) or were given no specific goal at all for the interaction (Swencionis & Fiske, 2016). However, in naturalistic situations, self-presentation is typically tailored to attain a specific goal, such as getting a new job or being believed by another person, for which the target’s social image may be important. It seems conceivable that while people may promote one dimension at the expense of the other when the impressions they make are not instrumental to attain a specific goal, they may well try to maximize their chances by coming across as both warm and competent when a favourable impression determines whether they reach a desired outcome or not. Hence, to understand people’s use of social compensation in everyday life, we need to study how people spontaneously choose to self-present when a desired outcome depend on them making a favourable impression on others. The current studies extend existing research by investigating how people use the compensatory dynamics between warmth and competence to reach a potentially important goal. Experiment 1 examines participants’ strategies during a job interview. In Experiments 2 and 3, we confront people with information about a crime and ask them, in the role of witness or suspect, to appear as credible as possible.
Next to the advancement of one or the other dimensions or the promotion of both of them, recent work on the two fundamental dimensions also points to other patterns. As a case in point, evidence suggests that competence has primacy over warmth when it comes to the self (Abele & Wojciszke, 2007, 2014; Wojciszke, 2005). It is possible that this primacy would bias compensatory dynamics in self-presentational strategies. Specifically, people may be more willing to downplay their warmth when competence is critical to reach a valued goal than to give up their competence when the situation suggests the importance of conveying a warm impression. In the current efforts, we explore this issue by examining self-presentation across situations in which the given goal suggests an emphasis on either competence or warmth. Because existing gender stereotypes are strongly related to warmth and competence (Eagly, 1987), we decided to control for gender in all three studies. Given the prescriptive nature of gender stereotypes, it seems possible that downplaying warmth in situations requiring high competence may be more hazardous for women than men (Rudman, 1998; Rudman & Fairchild, 2004); whereas, attenuating competence in a situation suggesting the importance of warmth may more often backfire for men than for women (Bosak, et al., 2018). This could mean that women more often than men choose not to attenuate their warmth relative to competence and that men more often than women choose not to downplay their competence.
In the contexts examined in the current research, we expect that participants would be motivated to convey an overall positive self-image, hence, they would not try to appear directly negative in either warmth or competence. Rather, we hypothesize that they may sacrifice being perceived as somewhat lower on one dimension to appear more favourably on the other dimension.
In sum, the current research replicates and extends previous findings by investigating the compensatory dynamics between warmth and competence when people spontaneously choose a self-presentational strategy to reach a potentially important goal.
Experiment 1: Downplaying warmth to get a desired job
We presented participants with an advertisement for a highly qualified manager position or for a less qualified non-manager position and asked them to indicate how they wanted to appear during an employment interview to get the job. We hypothesized that, overall, participants would strive to appear competent given the prevalence of this dimension in a professional context (e.g., Louvet, Rohmer & Dubois, 2009). We further expected that the motivation to appear competent would be higher for a highly qualified compared to less qualified position. Importantly, we also expected that, to appear highly competent, interviewees for the highly qualified position would downplay their warmth relative to their competence and relative to those interviewed for a less qualified job. In addition, we hypothesized that this tendency would be less pronounced among women than among men.
Stockholm university students (N = 120, 60 females, mean age = 25.33 years, SD = 6.00) participated in exchange for a $4 lottery ticket. For a medium-sized effect, this sample size would yield approximately 80% power.
We used a 2 (position: manager vs. non-manager) × 2 (participants’ gender: male vs. female) × 2 (dimension: warmth vs. competence) mixed-model design with the last factor measured within participants.
Participants were randomly assigned to read one of two job advertisements (downloaded from the Swedish public employment office website, www.arbetsformedlingen.se). The highly qualified manager advert concerned a position as a high-level business control manager, stressing such traits as being fearless and analytical. The less qualified non-manager advert concerned a mental health rehabilitator position, emphasizing such traits as being noncompetitive and respectful (see ESM for the adverts). Participants were then asked to imagine that they had applied for the position and had been called to an employment interview. They rated on 12 self-presentation traits how they wanted to appear during the interview to make the employer select them. The traits, derived from Kervyn, Yzerbyt, Judd, and Nunes (2009), measured the extent to which participants wanted to appear competent (positive valence: skilled, determined, competent; negative valence: disorganized, negligent, messy) and warm (positive valence: caring, tolerant, warm; negative valence: cold, selfish, insensitive). Scales ranged from 1 (not at all) to 7 (very). All data from the three experiments can be retrieved online on the Open Science Framework (OSF) repository at the following link: https://osf.io/pbfh4.
Self-presentational ratings were averaged for each dimension and valence, combined into indices of competence and warmth after reversal of the negative traits, (Cronbach’s α’s = .63 and .83, respectively), and analyzed with a 2 (position: manager vs. non-manager) × 2 (participants’ gender: male vs. female) × 2 (dimension: warmth vs. competence) mixed-model ANOVA with the last factor varying within participants.
A main effect of position (F(1,116) = 18.86, p < .0001, ηp2 = .14) revealed higher ratings for participants applying for a non-manager (M = 6.08, SD = 0.57) than those applying for a manager position (M = 5.67, SD = 0.51). There was also a main effect of gender (F(1,116) = 7.35, p < .008, ηp2 = .06), with lower ratings given by men (M = 5.74, SD = 0.52) than by women (M = 6.00, SD = 0.61). These effects were qualified by a significant gender by position interaction (F(1,116) = 4.16, p < .05, ηp2 = .03). Whereas the gender difference emerged for participants applying for a non-manager position (M = 5.85, SD = 0.52 and M = 6.31, SD = 0.53), for men and women respectively (F(1,58) = 11.12, p < .001, ηp2 = .16), no difference emerged for the manager position (M = 5.63, SD = 0.49 and M = 5.70, SD = 0.54), for men and women respectively (F(1,58) = .23, ns).
More importantly, while ratings across conditions and dimensions were above the scale midpoints, a dimension main effect showed that, overall, participants wanted to appear more competent (M = 6.38, SD = 0.59) than warm (M = 5.36, SD = 1.04) (F(1,116) = 120.90, p < .0001, ηp2 = .51). This effect was qualified by a position by dimension interaction (F(1,116) = 53.44, p < .0001, ηp2 = .32). As predicted, participants’ self-presentational strategies differed depending on the position they were trying to get (see Figure 1).
One way to probe this interaction is to look at the relative importance of both dimensions as a function of position. Whereas those interviewed for a less qualified position slightly preferred competence (M = 6.25, SD = 0.68) over warmth (M = 5.91, SD = 0.87) (F(1,58) = 6.29, p < .05, ηp2 = .10), those interviewed for a highly qualified position showed a strong preference for competence (M = 6.51, SD = 0.45) over warmth (M = 4.82, SD = 0.91) (F(1,58) = 181.93, p < .0001, ηp2 = .76). Alternatively, checking how position affected the importance given to each dimension, participants in the manager condition stressed competence slightly more than their non-manager counterparts (F(1,116) = 6.34, p < .05, ηp2 = .05). In sharp contrast, and supporting our compensation hypothesis, participants applying for the highly qualified position downplayed their warmth substantially relative to participants in the less qualified condition (F(1,116) = 48.87, p < .0001, ηp2 = .34).
Our first study sends an encouraging message with respect to our hypotheses. Extending previous work (Holoien & Fiske, 2013; Swencionis & Fiske, 2016), the results show that people may spontaneously alter the relative balance of the two dimensions in how they present themselves when trying to reach an important goal. Thus, when competence was critical to reach a valued goal, as when one is applying for a highly qualified job, participants not only stressed their competence, but also they did so at the expense of warmth, in contrast to applicants for a less-demanding job. In other words, and although there is no logical reason for participants to downplay how warm they are when applying for a qualified job, they preferred to appear significantly lower on these traits compared to participants trying to get a non-manager position. Interestingly, this self-presentational strategy meant that those applying for the highly qualified job tried to come across less favourably overall than those trying to get a less qualified job. While this at first might seem irrational, it makes perfect sense based on the social compensation effect, according to which, people should infer higher competence in someone who is a little less warm.
We found no evidence of an interaction between participant gender and dimension in self-presentational strategies. The fact that women downplayed their warmth relative to competence as much as the men did suggests that the experimental situation did not strongly cue gender-congruent behaviour (e.g., Rudman, 1998). Possibly, this could be due to our measures of competence and warmth, which did not include the most strongly gender-typed traits of the two dimensions (e.g., dominating, assertive/emotional, dependent), but rather less gender-typed ones (Rudman, Moss-Racusin, Phelan & Nauts, 2012). The higher overall scores for women could signal a strategy to overcompensate potential negative preconceptions about women’s job suitability (Eagly & Karau, 2002; Heilman, Martell & Simon, 1988).
In this experiment, we used two authentic job positions with a high or low demand for applicant qualifications. Although this material gives the study high external validity, it makes it possible that the positions differed in aspects other than qualifications (e.g., status, domain, gender-typing), constituting a potential threat to the study’s internal validity. However, whichever of these aspects was driving the obtained effects, they can all be assumed to consistently cue demands on applicants’ competence. Moreover, we note that previous research has demonstrated that the job domains most prototypical of agency and communion, two dimensions overlapping substantially with competence and warmth (Abele & Wojciszke, 2014), were jobs in the same two domains used in the present study (business and care provider, respectively; Blasberg, Rogers & Paulhus, 2014). This research also showed that participants asked to fake ideal candidates for the business job scored higher on a scale measuring agency and lower on communality measures than those posing as ideal candidates for the care provider position.
In any case, in our next set of studies we exclude potential biases in the stimulus material by using scenarios in which only participants’ roles differ.
Experiment 2: Downplaying warmth to convince the law
In Experiment 1, we compared people’s strategies in an employment situation, where successful self-presentation could make a difference between getting and not getting a desired position. In Experiment 2, we wanted to test the generality of these findings in an entirely different context, where the real-life stakes are far from trivial. We also decided to turn to a setting where even less information about the optimal self-image is provided. Admittedly, job adverts often provide cues as to what characteristics are being expected from applicants to meet the position requirements. What would happen in a situation where information regarding the requested traits is more limited and people have to fill in more on their own?
We studied how people who are either the witness or the suspect of a hypothetical crime choose to self-present to appear as credible as possible during a police interrogation. The value of eyewitness testimony largely depends on the witness’ ability to remember the target event correctly (e.g., Buck & Warren, 2009). Presumably, when reporting a crime as a witness, the primary aim should therefore be to maximize one’s apparent competence. In line with compensation, we propose that this could be served by witnesses downplaying their warmth. When interviewed as a suspect, however, the balance between competence and warmth could become more delicate. Evidence suggests that crime suspects are perceived as less guilty and are given more lenient sentences if they appear sensitive and emotional, rather than cold and unemotional (Heath, Grannemann & Peacock, 2004), possibly due to a presumed relation between innocence and warmth. As a suspect, there could be a motivation to convince the police that one is a warm person who could not possibly commit a crime. Thus, we conjectured that people in this role would choose to not stress their competence relative to their warmth and, possibly, to show a reversal in preferences, downplaying competence to appear relatively warmer. Furthermore, we expected that men would be more reluctant than women to downplay their competence in this situation.
Swedish participants (N = 160) representing different vocational backgrounds were recruited to an online survey distributed with the Qualtrics software. Of these, 37 failed to complete the whole survey, leaving 123 participants (66 females, mean age = 37.1 years, SD = 10.56).
We utilized a 2 (role: witness vs. suspect) × 2 (participants’ gender: male vs. female) × 2 (self-presentational dimension: warmth vs. competence) mixed design with the last factor measured within participants.
Participants were randomly assigned to one of two versions of a theft scenario. The witness version was written from the perspective of a customer who witnessed a clerk in a grocery store steal money from the register and who reported the theft to the manager. The suspect version was written from the innocent clerk’s perspective, who had been accused of theft after a customer had claimed having seen the clerk steal money from the register (see ESM for scenarios). Participants were then asked to imagine that they had been called to a police interrogation regarding the theft. Witness participants were told that they had a good memory of the event and wanted the police to understand that. Suspect participants learned that they were innocent and wanted to convince the police of this.
After reading their assigned scenario, participants indicated how they wanted to appear during the interrogation. The self-presentational traits were 16 scales from Kervyn, et al., (2009) measuring participants’ preferred appearance (competence positive valence: capable, skilled, determined, competent; negative valence: disorganized, lazy, unreliable, incompetent; warmth positive valence: caring, tolerant, sociable, warm; negative valence: cold, hostile, insensitive, unpleasant). We chose to extend and slightly change the traits used in Experiment 1 to test the generality of the compensation effect. Scales ranged from 1 (not at all) to 7 (very). Participants were then shown a list with 24 words, pretested to convey the user’s competence and warmth (six words high and six words low on each dimension, see Holoien & Fiske, 2013). They were instructed to select 12 words that they would prefer to use during the interrogation.
The ratings were combined into indices of competence and warmth as in Experiment 1 (Cronbach’s α’s = .70 and .80, respectively) and submitted to a 2 (role: witness vs. suspect) × 2 (participants’ gender: male vs. female) × 2 (trait dimension: warmth vs. competence) mixed-model ANOVA with the last factor varying within participants.
A gender main effect (F(1,119) = 19.23, p < .0001, ηp2 = .14) revealed that men (M = 5.67, SD = 0.70) gave lower ratings than women (M = 6.12, SD = 0.43). No higher-order effect involving gender approached significance (all ps>.13). There was again a significant effect of dimension (F(1,119) = 35.0, p < .0001, ηp2 = .23). Participants gave higher ratings for competence (M = 6.12, SD = 0.66) than for warmth traits (M = 5.70, SD = 0.80). More importantly, the dimension by role interaction was significant (F(1,119) = 8.19, p < .005, ηp2 = .06). As predicted, the specific role participants endorsed had them self-present in different ways (see Figure 2).
We probed the interaction by first looking at the weight given to the two dimensions as a function of participants’ role. Whereas suspects slightly preferred competence (M = 6.06, SD = 0.69) over warmth (M = 5.86, SD = 0.87) (F(1,59) = 4.26, p < .05, ηp2 = .07), witnesses strongly favored competence (M = 6.18, SD = 0.63) over warmth (M = 5.55, SD = 0,70) (F(1,60) = 42.48, p < .0001, ηp2 = .41). Looking at how role affected the importance given to each dimension, role had no impact on the preference for competence traits (F(1,119) = 1.13, ns); whereas, witnesses chose to appear less warm than the suspects (F(1,119) = 4.43, p < .05 ηp2 = .04).
Each word that participants selected was multiplied by the word’s competence and warmth ratings from pretests (see Holoien & Fiske, 2013, Table 1). Ratings of the 12 words selected by each participant were then averaged and submitted to a 2 (role: witness vs. suspect) × 2 (participants’ gender: male vs. female) × 2 (trait dimension: warmth vs. competence) mixed-model ANOVA. The analysis again showed a significant effect of dimension (F(1,119) = 16.24, p < .0001, ηp2 = .12). The words participants selected were overall higher in competence (M = 2.27, SD = 0.11) than in warmth (M = 2.16, SD = 0.29). Importantly, the dimension by role interaction was also significant (F(1,119) = 3.96, p < .05, ηp2 = .03). In line with the pattern on the self-presentational ratings, follow-up analyses showed that while suspects did not prefer competence- (M = 2.25, SD = 0.11) over warmth-related words (M = 2.20, SD = 0.26) (F(1,59) = 2.28, ns), witnesses clearly preferred to use competence- (M = 2.28, SD = 0.12) rather than warmth-related words (M = 2.13, SD = 0.31) (F(1,60) = 16.69, p < .0001, ηp2 = .22). There were no effects of role in participants’ preferences for words suggesting warmth or for those suggesting competence.
|Competence-related (long words, power, work)|
|Men||8.28 (3.75)||7.88 (3.88)||8.38 (5.06)||8.07 (3.59)|
|Women||9.59 (3.98)||9.39 (3.95)||10.36 (4.59)||7.91 (4.07)|
|All||8.96 (3.90)||8.73 (3.96)||9.35 (4.91)||7.98 (3.84)|
|Warmth-related (positive and negative emotions, social processes, family, friend)|
|Men||3.46 (1.36)||3.67 (1.70)||3.13 (2.01)||3.62 (2.12)|
|Women||3.14 (1.57)||3.40 (1.87)||2.89 (2.03)||4.11 (2.06)|
|All||3.29 (1.47)||3.52 (1.79)||3.02 (2.01)||3.89 (2.08)|
Replicating and extending the results from our first experiment, Experiment 2 showed that people strategically use the dynamics between competence and warmth when self-presenting to reach a valued goal. We hypothesized that participants in the role of crime witnesses would want to stress their competence when interrogated by the police. Indeed, accuracy is a critical facet in the evaluation of eyewitness memory (e.g., Buck & Warren, 2009), and it makes sense that witnesses would want to convince others that they could be useful information sources. In line with this hypothesis, participants in the role of witness did their best to appear competent, while at the same time downplaying their warmth. This pattern was evident in participants’ self-presentational ratings, as well as in the words they preferred to use during the interrogation.
We also expected that when in the role of the suspect, appearing more competent than warm might not be seen as the optimal strategy. Supporting this hypothesis, competence in self-ratings and word choice was not significantly preferred over warmth in this condition. At the same time, participants in the role of suspect downplayed their self-presentational warmth much less than the witnesses. Interestingly, endorsing the role of the suspect did not lead participants to reverse their preference for competence and warmth traits.
As in Experiment 1, we obtained a gender main effect, with women scoring higher than men on both warmth and competence across conditions. Because this effect generalized across two experiments with different self-presentational contexts, it seems to suggest a default strategy for women to overcompensate in their self-presentations, possibly as an attempt to defeat negative stereotypes about women’s abilities and, in particular, with respect to abilities (Eagly & Karau, 2002).
Experiment 3: The suspect’s dilemma: When your competence could bring you down
The fact that participants in our first two studies were globally tempted to present themselves as being competent nicely dovetails with the recurrent finding that people very much express concern for competence when it comes to themselves, in contrast to the preferential focus on warmth when it comes to others (Abele & Wojcizske, 2007, 2014; Abele, Bruckmüller & Wojcizske, 2014). However, we think there could be circumstances when people would rather be seen as more warm than competent. One such situation may arise when people are suspected of having committed a particularly dreadful crime, such as manslaughter. In this setting, their fate depends on convincing people that their character would make them unable of such aggressive and hostile behavior. Thus, an increase in crime severity could lead participants in the role of the suspect to a point where they would want others to see them as warm as possible, even warmer than competent. Experiment 3 tested this idea. In this experiment, we examined participants’ self-presentational strategies by looking both at their trait ratings, and at the semantic content of their self-descriptions.
US participants (N = 303, 161 females, mean age = 32.7 years, SD = 10.22) were recruited through Amazon Mechanical Turk and paid USD 0.50 for their participation.
We used a 2 (role: witness vs. suspect) × 2 (crime: theft vs. manslaughter) × 2 (participants’ gender: male vs. female) × 2 (self-presentational dimension: warmth vs. competence) mixed design with the last factor measured within participants.
Participants were randomly presented with one of two crime scenarios in the role of either witness or suspect. Half of the participants read the same theft scenarios as in Experiment 2. The others read one of two versions of a manslaughter scenario. The witness version described from the customer’s perspective how the clerk in a grocery store attacked a man who had demanded money and thrown the man into a shelf, where he hit his head so badly that he died. The suspect version described from the clerk’s perspective how a man, after demanding money, attacked the clerk but missed and fell, hitting his head so badly that he died (see ESM for scenarios). Participants were then asked to imagine they had been called to a police interrogation regarding the crime. Instructions and ratings of preferred appearance followed those of Experiment 2. In addition, ostensibly to help the police gauge their credibility, participants were asked to write a few sentences describing themselves. We used the Linguistic Inquiry and Word Count (Pennebaker, Booth, Boyd, & Francis, 2015) text analysis software to determine the semantic content related to competence and warmth in participants’ self-descriptions. We used default categories to estimate competence (words longer than six letters, power, and work) and warmth (positive and negative emotion, social processes, family, and friend).
Ratings were combined into indices of competence and warmth as before (Cronbach’s α’s = .86 and .88, respectively) and analyzed with a 2 (crime: theft vs. manslaughter) × 2 (role: witness vs. suspect) × 2 (participants’ gender: male vs. female) × 2 (trait dimension: warmth vs. competence) mixed-model ANOVA, with the last factor varying within participants.
A main effect of gender (F(1,295) = 14.17, p < .0002, ηp2 = .05) again showed that men (M = 5.89, SD = 0.86) gave lower ratings than women (M = 6.21, SD = 0.61). There was also a significant effect of dimension (F(1,295) = 61.25, p < .0001, ηp2 = .17), with ratings higher for competence (M = 6.24, SD = 0.80) than for warmth (M = 5.89, SD = 0.93). Dimension interacted significantly with role (F(1,295) = 38.50, p < .0001, ηp2 = .12) and with crime (F(1,295) = 9.25, p < .003, ηp2 = .03).
More importantly, the predicted three-way interaction involving dimension, role, and crime proved significant (F(1,295) = 5.61, p < .02, ηp2 = .02) (see Figure 3).
Probing this three-way interaction, we first examined the dimension by role interaction as a function of crime. The pattern for theft was significant (F(1,145) = 7.56, p < .007, ηp2 = .05) and fully replicated the one obtained in Experiment 2 (see Figure 3). Whereas suspects globally preferred competence (M = 6.36, SD = 0.77) over warmth (M = 6.04, SD = 1.04) (F(1,74) = 11.72, p < .001, ηp2 = .14), this was significantly more the case for witnesses (competence M = 6.33, SD = 0.75, warmth M = 5.66, SD = 0.87) (F(1,71) = 61.06, p < .001, ηp2 = .46). Looking at the interaction from the complementary viewpoint, witnesses chose to appear less warm than the suspects (F(1,145) = 5.29, p < .05, ηp2 = .04); whereas, role had no impact on the use of competence traits (F(1,145) = 0.0, ns).
Confirming our hypotheses, the dimension by role interaction for manslaughter was significantly stronger (F(1,150) = 35.90, p < .001, ηp2 = .19) and revealed the predicted compensation pattern (see Figure 3). Again, witnesses favored competence (M = 6.39, SD = 0.65) over warmth (M = 5.78, SD = 0.85) (F(1,76) = 56.98, p < .001, ηp2 = .43), but this time, suspects tended to prefer warmth (M = 6.07, SD = 0.90) over competence (M = 5.88, SD = 0.92) (F(1,74) = 2.81, p < .10, ηp2 = .04). Looking at the impact of roles on the prevalence of the two dimensions, witnesses again presented themselves as less warm than the suspects (F(1,150) = 3.66, p < .06, ηp2 = .02). However, and as predicted, suspects in this condition relied on competence much less than witnesses (F(1,150) = 16.36, p < .001, ηp2 = .10).
Another analytic strategy is to look at the role by crime interaction as a function of dimension. Considering warmth first, next to a gender effect (F(1,295) = 13.56, p < .001, ηp2 = .04), there was only a main effect of role (F(1,295) = 8.95, p < .003, ηp2 = .03), suggesting that crime severity failed to moderate the fact that witnesses self-presented as significantly less warm than suspects. Turning to competence and, again, to a gender effect (F(1,295) = 8.34 p < .005, ηp2 = .03), data revealed the presence of both a role effect (F(1,295) = 8.76, p < .004, ηp2 = .03) and a crime effect (F(1,295) = 4.57, p < .04, ηp2 = .02). Interestingly, and as predicted, the role by crime interaction was also significant (F(1,295) = 8.59, p < .004, ηp2 = .03). Confirming the previous analyses from a different angle, participants’ role had no impact on the emphasis of competence in the case of theft (F(1,145) = 0.0, ns); whereas, suspects relied on competence much less than witnesses in the case of manslaughter (F(1,150) = 16.36, p < .001, ηp2 = .10).
The proportion of words related to the different LIWC categories were averaged for each dimension (see Table 1) and subjected to the same mixed-model ANOVA as the self-presentational ratings. The analysis revealed a strong main effect of dimension (F(1,295) = 370.97, p < .001, ηp2 = .56), showing that competence-related words were used more often overall than words related to warmth. A gender main effect indicated more target words for women than for men (F(1,295) = 4.65, p < .05, ηp2 = .02). These effects were qualified by a gender by dimension interaction (F(1,295) = 5.14, p < .05, ηp2 = .02). While women used more competence-related words (M = 9.28, SD = 4.20) than men (M = 8.16, SD = 4.12) (F(1,295) = 5.88, p < .05, ηp2 = .02), there were no gender differences for words related to warmth (women M = 3.40, SD = 1.93; men M = 3.46, SD = 1.82) (F(1,295) = 0.17, ns).
In line with the results of the self-presentational ratings, the analysis also showed a dimension by role interaction (F(1,295) = 6.34, p < .05, ηp2 = .02). We probed this interaction by first looking at the weight given to the two dimensions as a function of participants’ role. Participants in the role of suspect used more competence- (M = 8.36, SD = 3.90) than warmth-related words (M = 3.70, SD = 1.94) (F(1,148) = 154.29, p < .001, ηp2 = .51), but this preference was even stronger among the witness participants (competence M = 9.16, SD = 4.44; warmth M = 3.15, SD = 1.77) (F(1,147) = 217.22, p < .001, ηp2 = .60).
When looking at how role affected the type of words used, participants in the role of suspect tended to choose fewer competence-related words than the witness participants (F(1,295) = 3.10, p < .1, ηp2 = .01). Conversely, participants interrogated as suspects used more warmth-related words than those in the role of witness (F(1,295) = 6.41, p < .05, ηp2 = .02). For the open-ended descriptions, there were no main or interaction effects involving crime.
Consistent with our earlier findings, Experiment 3 confirmed that when participants self-present to reach a desired goal, they choose to downplay impressions on one dimension and to bring the other dimension to the fore. Specifically, we fully replicated the pattern from self-presentational ratings in Experiment 2 in that both suspects and witnesses to a theft preferred to display more competence than warmth during a police interrogation; whereas, witnesses downplayed their warmth much more than the suspects. Importantly, and extending these findings, our results further corroborated the notion that people sometimes trade their competence to appear really warm. Thus, when interrogated about a more severe crime, namely manslaughter, witnesses again favored competence over warmth and displayed less warmth than the suspects. In contrast, suspects tended to favor appearing warm rather than competent and, as predicted, relied less on competence than the witnesses. Thus, even though competence seems to enjoy predominance over warmth when it comes to thinking about and presenting ourselves, our results suggest that this balance may shift when one’s outcomes depend on convincing others that they are dealing with a person whose good intentions should not be questioned. When looking at the semantic content in participants’ open-ended descriptions of themselves, results mirrored the overall pattern in self-presentational ratings. Thus, participants in the role of witnesses showed a stronger preference for competence- over warmth-related words than the suspects. Moreover, participants in the role of suspects described themselves with more warmth-related words than the witnesses, while they at the same time tended to use less words related to competence.
As in Experiments 1 and 2, women scored higher than men on both dimensions. Hence, in accord with the Swedish women participating in our first two experiments, American women taking part in this third experiment showed a pattern suggesting that they actively counteracted stereotypes that women may be less competent than men in general.
Although intuition would have us assume that in most situations people choose to maximize both how competent and how warm they appear, the present findings suggest that this is an overly simplistic view on people’s self-presentational strategies. Across three experiments, participants spontaneously used a trade-off between warmth and competence in their preferred appearance, demonstrating that they were most sensitive to how perceivers would balance the dimensions given the situational demands. Thus, when competence was critical to achieve a goal, as when applying for a highly qualified job or when being interrogated as a crime witness, participants not only strongly emphasized their competence, but also they consistently downplayed their warmth. In contrast, self-presentational emphasis shifted from competence to warmth when favourable outcomes appeared to depend on being perceived as a well-intended person, as was the case when participants endorsed the role of the suspect of a violent crime.
These results are entirely consistent with the compensation effect found in intergroup relations and impression formation (Judd, et al., 2005; Kervyn, et al., 2009, 2010; Yzerbyt, et al., 2005, 2008). Moreover, our data nicely enrich previous demonstrations of self-presentational compensation (Holoien & Fiske, 2013: Swencionis & Fiske, 2016). Clearly, compensation occurs not only when people are assigned the goal to convey a warm or competent self-image or when they interact with others in a hierarchical setting, but also when they spontaneously manage their impression to reach a strategic goal for which the target’s social image may prove important (e.g., getting a job, being believed by others).
The specific mechanisms that make people compensate in their self-presentations were not explored in the current research. Presumably, the phenomenon depends upon prior knowledge, implicit or explicit, regarding social perceivers’ expectations of a balance between warmth and competence in social targets. In fact, our results suggest that people are well tuned to these expectations and spontaneously adjust the balance to promote the optimal self-image given their current goals. Research in the intergroup domain suggests that compensation in social perception may result from distributive justice concerns (Yzerbyt, et al., 2008), as well as people’s desire to appear non-prejudiced (Yzerbyt & Cambon, 2017; Yzerbyt, 2018). The exact nature of the beliefs that relate to compensation in self-presentation deserves future attention.
Although self-presentational compensation emerged consistently across the three studies, the effects were stronger in Experiment 1, investigating self-presentation in an employment interview, than in the studies focusing on a legal context. A possible explanation for this pattern is that participants in Experiment 1 could collect some information as to what was expected of them from the advert that they read. In contrast, participants in Experiment 2 and 3 had no detailed information as to what the most desirable image would be. Hence, in the latter studies, variability in participants’ preconceptions of what would be the optimal self-image in the given situation could influence strategies moreso than in Experiment 1.
Another notable aspect of our findings was the tendency for participants to be more willing to give up being seen as nice than to give up conveying an impression of competence. As a matter of fact, whereas warmth ratings fell below competence ratings across virtually all conditions, indicating the importance of competence, it took a strong situational cue—being accused of a violent crime—to make participants downplay their competence and bring their warmth on the forefront. This pattern aligns with evidence showing that competence has primacy over warmth when it comes to the self (Abele & Wojcizske, 2007, 2014; Abele, et al., 2014). Although Holoien and Fiske (2013) did not discuss this aspect of their findings, we note that their data also reveal stronger effects sizes when participants downplay warmth to appear competent than when downplaying competence to appear warm.
An interesting message coming from our data concerns the lack of interactions between participants’ gender and self-presentational dimension. All three experiments could count on a sufficient number of participants from each gender, and there is no reason to assume that power was particularly low for testing such interaction effects. Hence, the experimental situations used in the current study seemingly did not cue gender-congruent behaviour (e.g., Rudman, 1998). This could be due to our measures of competence and warmth, which included the most strongly gender-typed traits of the two dimensions and less gender-typed ones (Rudman, Moss-Racusin, Phelan & Nauts, 2012). It could also be that the particular position (high or low qualified job) or the particular role (witness or suspect) exerted more of an impact than gender. This pattern is reminiscent of the work by Eagly and Steffen (1986) about the impact of social roles when attributing characteristics to men and women. These authors found that people inferred more masculine versus feminine traits among men and women, respectively, but only when the target’s profession was left unspecified. This seems to suggest that features generally associated with the specific position/role may have shaped the self-presentation strategies observed here independently of self-presenters’ gender. Whether other contexts are open to more sophisticated strategies involving gender is definitely a topic for future research.
While we found no gender by dimension interactions, the results across our three experiments consistently showed that women preferred to appear both warmer and more competent than men did. Hence, women seemed to promote themselves in their self-presentations, and one may speculate that this move comes as an attempt to defeat negative stereotypes about women’s abilities (Eagly & Karau, 2002). Clearly, the generality of this pattern in other contexts, and potential mechanisms behind it, deserves further investigation.
Our three studies all focused on situations in which one dimension was more relevant for self-presentation than the other. Even if in such situations people may strategically use compensation to increase observers’ perception that they possess the relevant characteristic, it could be asked whether self-presentational compensation is a general strategy used in every context or only when there are clear cues in the situation suggesting that one dimension is more important than the other. Based on the current findings, along with evidence showing that people value competence more than warmth when it comes to the self (Wojciszke, 2005), it could be argued that a default strategy for most people would be to downplay their warmth relative to competence. This is speculative and goes beyond the realm of the current data. Future studies should address this question and specifically test the conditions during which people do and do not use compensation in self-presentation.
The current studies suffer from some limitations. First, the stimulus materials used in Experiment 1 differed in aspects other than qualifications. Although we argue that these differences should consistently have acted as cues to the jobs’ competence-related demands, strong conclusions about which factor was driving the obtained effects cannot be drawn.
Admittedly, some of the effects obtained in the current research were small and revealed only weak effect sizes. We thus emphasize the fact that some specific results, such as the downplaying of warmth relative to competence for suspects in Experiment 3, should be interpreted with caution. However, the fact that the overall pattern, with participants downplaying the less critical dimension relative to the more important one when self-presenting in order to reach a goal, replicated across different contexts in our three studies, lends support to the strength of the current message.
To conclude, the present studies extend previous research by demonstrating compensatory dynamics between warmth and competence when people spontaneously choose a self-presentational strategy to reach an important goal. Our findings further suggest that these compensational strategies are used in such varied contexts as employment interviews and police interrogations. Moreover, corroborating evidence on peoples’ preference for competence over warmth when it comes to describing the self, our results suggest that people more willingly give up warmth to convey a competent impression than they do competence to convey a warm impression. Future studies should examine the specific mechanisms, and indeed people’s motivations, that drive the effect. Even more fascinating, further work should dig into the ramifications of compensatory self-presentation on observers’ perceptions and behaviors toward the actor.