Abstract
Two experiments examined how low and high self-esteem people regulate attention in the face of uncertainty about their partner’s caring. We primed risk regulation processes by leading experimental participants to believe their partner’s caring and responsiveness was in question. We then assessed directed attention to the partner’s positive and negative qualities using a dot-probe paradigm. High, but not low, automatically directed attention away from their partner’s negative traits in response to uncertainty.
Love to faults is always blind, always is to joy inclined.
Puppet lovers in their bliss, turn away from all of this.
But reality does intrude on such perceptions. No matter how ideal partners seem to be, they inevitably engage in negative or rejecting behavior (Kelley, 1979). Uncaring partner behavior typically motivates people to derogate their partner to distance themselves from such hurts (Murray, Holmes, & Collins, 2006). But not everyone takes this bait. High self-esteem people are not only more likely to idealize their partner than low self-esteem people (Murray et al., 1996a) but also strengthen such idealized perceptions in the face of uncertainty about their caring (Cavallo, Murray, & Holmes, 2013; Gomillion & Murray, 2014).This article examines whether high self-esteem people are so practiced in idealization maintenance that they automatically direct attention away from their partner’s faults in the face of acute uncertainty.
Self-Esteem and Relationship Goal Priorities
To connect, people need to make themselves vulnerable—by overlooking a partner’s faults and transgressions, seeking support, and soliciting sacrifices. But connecting in these ways invites the possibility of a partner’s rejection, motivating people to be cautious and hesitant to depend on a partner. According to the risk regulation model, trust in a partner’s caring functions as a cue to safety in navigating these goal conflicts (Murray et al., 2006).
Being more trusting signals the likelihood of the partner’s responsiveness in risky situations, allowing people to seek greater closeness. In contrast, being less trusting signals the likelihood of the partner’s rejection, motivating people to distance from the partner and avoid potential hurts (Campbell, Simpson, Boldry, & Rubin, 2010; Derrick, Leonard, & Homish, 2012; Murray, Holmes, & Griffin, 2000; Murray, Rose, Bellavia, Holmes, & Kusche, 2002; Overall & Sibley, 2009). Unfortunately, low self-esteem people have greater trouble establishing a sense of trust in their partner’s caring than high self-esteem people (Murray et al., 2000; Murray, Holmes, Griffin, Bellavia, & Rose, 2001). Consequently, high and low self-esteem people usually regulate risk by prioritizing opposite interpersonal goals. In situations that provoke acute uncertainty about a partner’s caring, low self-esteem people typically pursue self-protection goals (Cameron, Stinson, Gaetz, & Balchen, 2010; Ford & Collins, 2010), whereas high self-esteem people typically pursue connection goals (Cavallo, Fitzsimons, & Holmes, 2009).
Such chronically divergent goal priorities have dramatic effects on partner idealization. When low self-esteem people are led to believe their partner will discover their secrets and become disaffected, they self-protectively derogate their partner’s qualities on explicit measures (Murray et al., 2002). But when high self-esteem people are led to believe that their partner has serious complaints about them, they idealize their partner’s qualities all the more (Gomillion & Murray, 2014; Murray et al., 2002). In fact, high self-esteem people idealize their partner more than lows in dating and marital relationships because they chronically trust that their partner loves and values them (Murray et al., 2000, 2001).
Selectively Myopic? Directed Attention and Risk Regulation
Because prior research examined idealization maintenance as a controlled output of risk regulation, it might have obscured a vulnerability that high self-esteem people actually share with lows. Priming partner rejection automatically activates the state goal to self-protect for lows and highs alike (Cavallo et al., 2009). Therefore, high self-esteem people might experience the automatic inclination to derogate their partner in risky situations but resist this inclination because explicit measures afforded highs the self-regulatory control needed to idealize their partner. Indeed, highs report feeling closer to their partner in risky situations when they have self-regulatory resources available to prioritize their preferred goal to connect over the state goal to self-protect (Cavallo, Holmes, Fitzsimons, Murray, & Wood, 2012).
However, high self-esteem people may not be entirely defenseless without self-regulatory control. Motivational states automatically direct attention toward features of social objects that could satiate such motivations and away from those features that could frustrate such motivations (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn, 2007; Moskowitz, 2002). For instance, desiring interpersonal acceptance heightens attention to the smiling faces in a visual array of faces (DeWall, Maner, & Rouby, 2009). Fearing rejection locks attention on social threat cues, such as angry faces (Fox, Russo, & Dutton, 2002; Mogg & Bradley, 2002;Mogg, Philippot, & Bradley, 2004). Desiring romantic attachment directs the attention of singles toward attractive potential suitors (DeWall, Maner, Deckman, & Rouby, 2011; Maner, Gailliot, Rouby, & Miller,2007), whereas fearing a romantic partner’s infidelity directs attention toward romantic rivals who could tempt a partner to stray (Maner, Miller, Rouby, & Gailliot, 2009). Conversely, people committed to a dating partner automatically direct attention away from desirable alternative partners (Maner, Gailliot, & Miller, 2009; Maner, Rouby, & Gonzaga, 2008); people who prefer to maintain psychological distance in relationships also direct their attention away from distressing relationship information (Edelstein & Gillath, 2008).
For high self-esteem people, chronic goals to connect to the partner could selectively direct attention toward positive qualities that reinforce the partner’s value in risky situations or direct attention away from negative qualities that undermine it. In making judgments, people usually weigh negativity more heavily than positivity (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Reeder & Brewer, 1979). In romantic relationships, a partner’s misdeeds also have greater power to undermine trust in the partner’s caring than good deeds have to bolster it (Holmes & Rempel, 1989) and negative events detract more from satisfaction than positive events add to it (Rusbult, Johnson, & Morrow, 1986; Weiss, Hops, & Patterson, 1973).
Because bad has a stronger pull than good (Baumeister et al., 2001), deflecting attention from a partner’s negative qualities may do more to quell uncertainty than focusing attention on positive qualities can do to squelch it. Given this asymmetry, we reasoned that a partner’s negative qualities might be a stronger target of directed attention than the partner’s positive qualities. We expected high self-esteem people to direct their attention away from their partner’s negative qualities when they had reason to question their caring. In so doing, highs could shield chronic goals to connect against state goals to self-protect by diverting their attention away from partner features that could deflect them from their preferred pursuit of connection (Dreisbach & Haider, 2008; Goschke & Dreisbach, 2008; Vogt, De Houwer, & Moors, 2011).
The chronic priority low self-esteem people put on self-protection should only reinforce the state goal to self-protect in risky situations. Therefore, we did not expect low self-esteem people to direct their attention away from their partner’s negative traits in response to risk. Instead, we expected lows to be at least as attentive—if not more attentive—to their partner’s negative qualities when they had reason to question their partner’s caring. By keeping an eye on these potential cues to rejection, low self-esteem people could effectively shield themselves against future hurts, satisfying the goal to self-protect (Pietrzak, Downey, & Ayduk, 2005).
Study 1
Study 1 activated risk regulation processes by leading participants in the relationship threat condition to believe their partner would uncover their unsavory secrets, sowing the seeds for conflict (Murray et al., 2002). We then measured attention to the partner’s positive and negative qualities using a dot-probe task (MacLeod, Mathews, & Tata, 1986). We expected threatened highs, but not lows, to direct attention away from their partner’s negative qualities, evidencing a connectedness-promoting blind spot to such cues.
Method
Participants
One hundred and fourteen individuals involved in exclusive relationships participated in exchange for course credit. Two participants were lost due to a computer malfunction, and 10 participants were excluded (see Results), leaving 102 participants. Participants (40 men) averaged 19.67 years of age (standard deviation [SD] = 1.97); relationships averaged 15.86 months (SD = 12.90).
Procedure
Participants first completed demographic questions and the Rosenberg (1965) Self-Esteem Scale. Participants were then randomly assigned to relationship threat (threat, affirmation, or control) and target (partner or people in general) conditions. All participants first listed three secret aspects of the self that they wish to keep hidden from their partner. In the relationship threat condition, participants were told that partners eventually discover these “secret selves,” leading to increased conflict. In the affirmation control condition, participants were told that the discovery of secret selves leads to increased closeness. Participants in the baseline control condition were not given any feedback. Next, participants in the partner target condition generated four of the traits they liked the most in their partner and four of the traits they disliked the most in their partner (Appendix A). In the general target condition, participants generated four traits that they liked the most in people in general and four traits that they disliked the most in people in general. This latter condition was included to demonstrate that biases in attention are specific to the partner. Participants then completed the dot-probe task, fillers, and a manipulation check that asked them to describe how their secret selves might affect conflict (1 = might decrease, 9 = might increase). Then they were debriefed.
Dot-Probe Task
We adapted the dot-probe task developed by MacLeod, Mathews, and Tata (1986) to assess attentional bias. On each trial, participants first stared at a fixation cross at the center of the screen. After 500 ms, stimuli appeared to the left and right of the fixation cross. The stimuli were positive-neutral and negative-neutral word pairings. These word pairings consisted of the positive and negative traits idiographically generated by the participants and 16 preprogrammed neutral stimuli (e.g., “drinkable,” “italic,” and “underlined”). After 500 ms, the stimuli disappeared and a target probe appeared in place of one of the stimuli. Participants then pressed a key to indicate whether the probe appeared on the left versus right side of the screen. The probe remained until the participant responded. Participants first completed 20 practice trials, during which they received feedback on their performance in the form of red “Xs” that appeared on every miss. Then they completed 128 randomized test trials. Reaction times (RTs) were recorded in milliseconds.
Results and Discussion
We removed incorrect trials (4.02% of data), trials with RTs shorter than 200 ms or greater than 850 ms (3.27% of data), 2 participants who listed no negative traits, 7 participants who erred on more than 15% of trials (Mogg & Bradley, 2002; Mogg et al., 2000; Koster, Crombez, Verschuere, & De Houwer, 2004), and 1 participant who had no valid trials for neutral probes. Exclusion and error rates did not differ by condition.
Next, we calculated attentional bias scores for both positive and negative traits. For attentional bias to positive traits, we subtracted the mean RT for neutral stimuli presented opposite positive traits from the mean RT for positive traits. For attentional bias to negative traits, we subtracted the mean RT for neutral stimuli presented opposite negative traits from the mean RT for negative traits. More positive bias scores indicate attention directed away from the traits (i.e., slower response times for probes presented in a trait quadrant vs. neutral quadrant). More negative bias scores indicate attention directed toward the traits (i.e., faster response times for probes presented in a trait quadrant vs. neutral quadrant). Attentional bias scores for positive and negative traits were normally distributed in both studies.
Directed Attention
We conducted hierarchical regression analyses predicting attention to positive and negative traits from (1) the main effects of relationship threat (i.e., dummy-coded vectors compared the threat and baseline control and the affirmation and baseline control conditions), the main effect of target (1 = partner traits, −1 = general traits), and the centered main effect of self-esteem, (2) all possible two-way interactions, and (3) the three-way interactions. Table 1 reveals the expected Self-Esteem × Target × Threat versus Control Condition interaction predicting attentional bias to negative, β = −.30, t(91) = 1.98, p = .051, f 2 = .05, but not positive traits. No main or interactive effects involving the comparison between the affirmation and the baseline control conditions emerged. 1
Model Coefficients for Attentional Bias.
Note. *p < .05. **p < .01. ***p < .001.
Figure 1 presents the predicted score for attention to negative traits. We decomposed this interaction to reveal (1) the effects of self-esteem on attention to the partner’s negative traits in the relationship threat versus control condition and (2) the effects of relationship threat on attention to the partner’s negative traits among low and high self-esteem people.
Self-esteem and Threat condition predicting an attentional bias for negative traits in partner and other target conditions. More positive bias scores indicate attention directed away from the traits.
The simple effects of self-esteem
The predicted two-way Self-Esteem ×Threat interaction emerged for participants focused on partner traits, β = .37, t(91) = 1.90, p = .06 (Figure 1, left panel) but not for those focused on people in general, β = −.17, t(91) = −.89, p = .38 (Figure 1, right panel). In the relationship threat condition, high self-esteem participants were significantly slower to respond to probes presented behind negative partner traits than lows, β = .52, t(91) = 2.04, p = .04, suggesting their attention was directed away from these traits. Self-esteem did not predict attention to negative partner traits in the baseline control condition, as expected β = −.16, t(91) = −.63, p = .53.
The simple effects of relationship-threat
The predicted two-way Threat ×Target interaction was significant for high, β = .42, t(91) = 1.98, p = .051, but not for low self-esteem participants, β = −.14, t(91) = −.74, p = .46. High self-esteem participants were significantly slower to respond to probes presented behind negative partner traits in the threat than control condition, β = .44, t(91) = 1.74, p = .09 (Figure 1, left panel), again suggesting their attention was directed away from these traits. This bias did not emerge when highs were focused on negative traits in people in general, β = −.25, t(91) = −1.05, p = .30 (Figure 1, right panel). The simple effect of threat was not significant for low self-esteem people focused on negative partner traits, β = −.20, t(91) = −.89, p = .37, or for lows focused on negative traits in people in general, β = .04, t(91) = .16, p = .88.
Manipulation Check
Although experimental participants (M = 6.67) expected their secret selves to generate more conflict in their relationship than controls (M = 6.62), this predicted main effect was not significant. This manipulation of relationship threat has induced significant uncertainties on manipulation checks in prior research (Cavallo et al., 2009; Murray et al., 2002).
Study 2
Study 2 activated risk regulation processes by leading participants in the relationship threat or uncertainty condition to believe their partner was likely to take advantage of them. We then measured attentional bias to the partner’s positive and negative qualities using the dot-probe task. We expected high, but not low, self-esteem people to direct their attention away from their partner’s negative traits in the face of uncertainty about their partner’s caring.
Method
Participants
One hundred and nineteen individuals involved in exclusive relationships participated for course credit. Three participants were lost due to a computer malfunction and 16 further participants were excluded (see Results section), leaving 100. Participants (31 men) averaged 19.13 years of age (SD = 2.76); relationships averaged 22.15 months (SD = 25.02).
Procedure
Participants completed demographic questions and the Rosenberg (1965) Self-Esteem Scale. Participants were then randomly assigned to relationship threat (threat or control) and target (partner or people in general) conditions. All participants then read a series of vignettes based on a manipulation used by Cavallo, Fitzsimons, and Holmes (2010). These vignettes depicted conflict of interest situations in relationships, such as a fictional couple disagreeing about which movie to see on a date, with one partner ultimately accommodating the other partner’s needs. Participants rated how likely they would be to compromise in each of these situations. In the threat condition, participants then received feedback indicating they were overestimating the quality of their relationship and leaving themselves vulnerable to being taken for granted by their partner. In the control condition, participants were told that they had an accurate and positive view of relationships, promoting responsiveness on the part of their partners. As in Study 1, participants then generated traits in their own partner (Appendix B) or people in general, completed the dot-probe task, fillers, and a manipulation check that asked whether they were underestimating or overestimating the quality of their relationship.
Results and Discussion
We removed incorrect test trials (2.44% of data), trials with RTs shorter than 200 ms or greater than 850 ms (0.98% of data), and all trials for 10 participants who did not list negative partner traits and 6 participants who made more than 15% of errors. We then conducted hierarchical regression analyses predicting attentional bias to positive and negative traits from (1) the main effects of threat (1 = threat, −1 = control), target type (1 = partner traits, −1 = general traits), and self-esteem, centered, (2) the two-way interactions, and (3) the predicted three-way interaction. Table 2 reveals the three-way Threat × Target × Self-Esteem interaction was significant for attentional bias to negative, β = −.29, t(93) = 2.80, p = .006, f 2 = .08, but not positive, traits. Figure 2 presents the predicted scores.
Self-esteem and Threat condition predicting an attentional bias for negative traits in partner and other target conditions. More positive bias scores indicate attention directed away from the traits.
Model Coefficients for Attentional Bias.
Note. *p < .05. **p < .01. ***p < .001.
The simple effects of self-esteem
The predicted Self-Esteem × Threat interaction emerged for participants focused on partner traits, β = .33, t(93) = 2.22, p = .03 (Figure 2, left panel). In the threat condition, high self-esteem participants were significantly slower to respond to probes presented behind negative partner traits than low self-esteem participants, β = .36, t(93) = 1.99, p = .05, again suggesting their attention was directed away from these traits. The simple effect of self-esteem predicting attention to negative partner traits was not significant for participants in the control condition, β = −0.32, t(93) = −1.33, p = .19.The Self-Esteem × Threat interaction was marginally significant for participants focused on negative traits in people in general, β = −.23, t(93) = −1.76, p = .08 (Figure 2, right panel). In the threat condition, self-esteem did not predict RTs to negative traits in people in general, β = −.15, t(93) = −.74, p = .46. But, in the control condition, high self-esteem participants tended to be slower to identify probes presented behind negative traits for people in general compared to lows, β = .32, t(93) = 1.85, p = .07.
The simple effects of relationship threat
The Threat × Target interaction was significant for high self-esteem, β = .29, t(93) = 2.08, p = .04, and low self-esteem participants, β = −.29, t(93) = −1.99, p = .05. High self-esteem people were significantly slower to respond to probes presented behind negative partner traits in the threat than control condition, β = .40, t(93) = 1.99, p = .05 (Figure 2, left panel). However, the simple effect of threat was not significant when highs were focused on negative traits for people in general, β = −.18, t(93) = −.94, p = .35 (Figure 2, right panel). The simple effect of threat was not significant for low self-esteem people focused on negative partner traits, β = −.27, t(93) = −1.31, p = .19, or negative traits in people in general, β = .30, t(93) = 1.53, p = .13.
Manipulation Check
Participants in the threat condition (M = 1.96) were more likely to report that they were overestimating the quality of their relationship than control participants (M = 1.48).
Meta-Analytic Summary
To examine the consistency of the lower order effects, we conducted a meta-analysis using Winer’s (1971) method of combined t-tests. Table 3 summarizes the results. The Self-Esteem × Threat interaction was significant for participants responding to negative partner traits and marginal for participants responding to negative traits in people in general. The Threat × Target interaction was significant for people high in self-esteem and marginal for participants low in self-esteem. In the relationship threat, but not the control condition, high self-esteem participants were significantly more likely than lows to direct attention away from their partner’s negative traits. Highs in the threat condition were also significantly more likely to direct attention away from negative partner traits than controls. Low self-esteem participants were more likely to turn their attention toward their partner’s negative traits in the relationship threat than control condition, albeit not significantly. In the control, but not the relationship threat condition, highs tended to be slower to respond to negative traits in people in general than lows. 2
Meta-Analytic Z Scores for Negative Attentional Bias.
Note. *p < .10. **p < .05.
Alternative Explanations
There are at least three possible alternative explanations for the selective myopia evidenced by high self-esteem people. Because we introduced the relationship threat before participants generated the traits, threatened highs might have listed less serious traits that were easier to turn a blind eye toward. Ten undergraduate research assistants blind to experimental condition and the trait’s specified valence rated all traits from 1 = very positive to 7 = very negative). Although participants listed less serious negative traits in their own partner, the observed three-way interaction was still significant when we controlled for the main and interactive effects of trait ratings in Study 1, β = .68, t(82) = 3.31, p < .001, and in Study 2, β = .33, t(86) = 2.18, p < .05. The second and third possibilities are that high self-esteem people more readily turned their attention away from their partner’s negative traits because they were involved in more satisfying or longer relationships. But, in separate analyses that replaced self-esteem with satisfaction and self-esteem with relationship length, the three-way Satisfaction × Threat × Target and Relationship Length × Threat × Target interactions was not significant (all ps > .18).
General Discussion
High self-esteem people are even more adept at defusing threats than prior research suggests. Although self-regulatory control can help high self-esteem people prioritize connection goals (Cavallo et al., 2012), such control is not an absolute necessity. High self-esteem people automatically diverted attention away from negative partner qualities when they actually had reason to be vigilant and attend to precisely those qualities. The selective myopia of threatened high self-esteem people emerged relative to both threatened low self-esteem people and high self-esteem people in the control condition. Also, their selective myopia could not be attributed to high self-esteem people generating less negative partner qualities, to highs being more satisfied, or to highs being involved in longer relationships.
The present findings have limitations. First, the results were not letter-perfect. In each study, threatened high self-esteem people appeared selectively myopia relative to threatened low self-esteem people. However, in the control condition, highs also tended to turn their attention away from negative traits in people in general (relative to controls). We hesitate to interpret this unexpected and marginal effect, but it echoes other evidence suggesting that highs are more approach oriented than low self-esteem people (McGregor, Nash, & Inzlicht, 2009). In Study 1, a tried and true relationship threat manipulation utilized also revealed no effect on the manipulation check (Cavallo et al., 2009; Murray et al., 2002). Fortunately, Study 2 did reveal such an effect and the selective myopia evidenced by highs in Study 1 replicated.
Second, people came to the lab with already high versus low self-esteem and already prematched to their partner. Therefore, we cannot rule out the possibility that high self-esteem people possess partners who differed in some objective way from the partners of low self-esteem people. Raters did judge the negative partner qualities provided by high self-esteem people to be just as negative as the qualities low self-esteem people provided. High self-esteem people are also just as likely to possess low self-esteem as high self-esteem partners (Murray et al., 2000). However, the automatic ease that highs evidenced turning away from their partner’s negative qualities could still have a basis in relationship reality. For instance, their willingness to draw closer to their partner in the face of risk could motivate partners to behave well (Cavallo et al., 2013), making their negative traits less perceptually salient and easier to ignore.
Third, the present studies raise more questions about low self-esteem people than they answer. Low self-esteem people often devalue their partner when they have reason to doubt their caring (Gomillion & Murray, 2014; Murray et al., 1998; 2000). Yet, lows were not significantly more vigilant to their partner’s negative qualities in the threat than control condition. This equivocal evidence for vigilance could reflect the goal conflict low self-esteem people face in relationships. Although lows typically self-protect, they still strongly desire connection and actually forego self-protection when it is safe (Cameron et al., 2010; Marigold, Holmes, & Ross, 2007, 2010; Murray, Derrick, Leder, & Holmes, 2008; Murray et al., 2005). Had we provided lows in the control condition with a stronger affirmation experience, it might have provided a better point of contrast for detecting vigilance to their partner’s faults in the face of threat.
Finally, the present studies left a major issue unaddressed—the effect that automatically turning attention away from a partner’s faults has on relationships. Future research could examine whether training people to turn their attention away from their partner’s faults makes it easier to sustain idealized perceptions. It might also explore whether being selectively inattentive to partner faults has any downside. For instance, forgiving an unrepentant partner undermines self-concept clarity (Luchies, Finkel, McNulty, & Kumashiro, 2010). Consistently turning an automatically blind eye to a partner’s faults could also allow offending partners to reoffend (McNulty, 2010). But when people are feeling most vulnerable, turning a selective blind eye to a partner’s negative qualities could also afford people the sense of optimism needed to attend directly to problem areas (Overall, Fletcher, Simpson, & Sibley, 2009).
Conclusion
The present experiments reveal a novel trick in the relationship toolbox of high self-esteem people. In the face of acute uncertainty about their partner’s caring, high self-esteem people automatically diverted their attention away from their partner’s faults. The present findings thus point to an important role for low-level cognitive processes in risk regulation.
Footnotes
Appendix A
Appendix B
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: National Science Foundation Grant #1101589.