Attachment and the Management of Empathic Accuracy in Relationship-Threatening Situations

Abstract

“To understand all is to forgive all.”

French proverb

“To understand all is to forgive nothing.”

English epigram

As these contrasting views suggest, insight into what a partner is thinking or feeling can cut both ways. In some contexts, knowing more may strengthen the ties that bind partners together. In others, knowing more may threaten or even destroy a relationship. For example, in relationship-threatening situations that cannot be averted, some people may be motivated to ignore, disregard, or misinterpret the damaging thoughts or feelings their partners could be harboring, not wanting to become entangled in the issues implied by those negative thoughts and feelings. Other people, however, may want to “get inside the head” of their partners, even if the knowledge gained might hurt or destabilize the relationship. Who are these people, and how do they “manage” empathic accuracy enroute to protecting—or failing to protect—themselves or their relationships?

Though it is a topic of considerable theorizing and speculation (Ickes & Simpson, 1997; 2001), little empirical research has investigated how people manage empathic accuracy, especially during relationship-threatening interactions. Melding principles of attachment theory (Bowlby, 1969, 1973, 1980) and the revised Empathic Accuracy Model (Ickes & Simpson, 2001), we derived and tested a set of predictions that address how individuals who have different attachment orientations differentially “manage” empathic accuracy in relationship-threatening contexts.

Individuals begin developing working models of the self and others based on interactions with significant others early in life. Across time, these models influence the way in which people think, feel, and behave in close relationships (Bowlby, 1969, 1973, 1980). Beliefs and expectations are central components of working models, which involve “if/then” propositions about what attachment figures are likely to do in certain contexts (e.g., if I feel vulnerable, then I can count on my partner for support). Working models also contain rules distilled from experiences with past attachment figures that guide behavior with respect to current attachment figures. For example, working models affect information processing by influencing which aspects of a romantic partner’s behavior are attend to or ignored, the inferences and judgments that are made about a partner’s behavior, and which partner actions are or are not remembered (Collins, Guichard, Ford, & Feeney, 2004). Variations in working models give rise to individual differences in attachment orientations (styles).

Two dimensions underlie individual differences in adult romantic attachment orientations (Brennan, Clark, & Shaver, 1998; Simpson, Rholes, & Phillips, 1996). Avoidance reflects the degree to which people are uncomfortable with closeness and emotional intimacy. Highly avoidant individuals tend to be less invested in their relationships, claim to value their relationships less, and strive to maintain psychological and emotional independence from their partners (Bowlby, 1973; Hazan & Shaver, 1994). Anxiety reflects the degree to which individuals worry and ruminate about being rejected or abandoned by their partners. Highly anxious individuals are chronically concerned that their partners might leave them, do not love them, or are unwilling to help them cope with distressing situations.

One of the central functions served by attachment orientations is the regulation of negative affect (Kobak & Sceery, 1988; Simpson, 1990). According to Kobak and Sceery (1988), secure individuals directly acknowledge distress when they experience it and turn toward attachment figures for comfort and support. Highly avoidant individuals divert attention away from the source of distress, do not acknowledge being upset, and use self-reliant tactics to control and mitigate negative affect. Highly anxious people direct their attention toward the source of distress and focus on it, particularly when they believe that attachment figures might be unresponsive to their needs. As a result, highly anxious people have difficulty reducing and containing negative affect. Given these tendencies, highly anxious and highly avoidant individuals cope with stressful events less effectively than do less avoidant and less anxious (i.e., more secure) persons (Mikulincer & Florian, 1998).

Mikulincer and Shaver (2003) developed a process model that specifies the conditions under which the attachment system should be activated in adults who have different attachment orientations. Because highly anxious individuals want to avoid losing their partners/relationships, they use tactics associated with a hyperactivating strategy (e.g., ruminating about worst-case scenarios, exaggerating potentially threatening cues, remaining vigilant to signs that their partners might leave them), especially when a potential relationship threat is detected. One such tactic may be motivated empathic accuracy with respect to what the partner is thinking and feeling during relationship-threatening interactions. In other words, one manifestation of vigilance should be greater empathic accuracy in relationship-threatening situations relative to less threatening ones. Preliminary evidence for this phenomenon was first reported by Simpson, Ickes, and Grich (1999), who found that highly anxious women were more empathically accurate than less anxious women when each woman tried to infer her romantic partner’s thoughts and feelings as he rated and discussed attractive women with her. This effect, however, has never been documented in men, and it has not been investigated in discussions that center on significant relationship issues (e.g., intimacy, jealousy, conflict), many of which involve problems internal to (rather than external to) the relationship.

When highly avoidant individuals detect potential threats to their independence in relationships, they should try to keep their attachment systems deactivated. Unlike anxious individuals, highly avoidant individuals work to inhibit and control their emotions using deactivating tactics such as ignoring, dismissing, or withdrawing from threats, and/or suppressing threat-related thoughts. One of the most efficient ways to limit, control, and curtail information that might activate the attachment system is to simply “stay out of the partner’s head” (i.e., to be less empathically accuracy when partners might be harboring threatening/distressing thoughts and feelings). Thus, highly avoidant individuals should use “frontline” strategies designed to ward off activation of their attachment systems from the very outset. No research to date has examined whether highly avoidant people are less empathically accurate during actual relationship discussions.

Study 1

In Study 1, married couples participated in a videotaped problem-resolution task. Each couple was asked to identify and try to resolve either a major or a minor relationship problem that centered on an intimacy or a jealousy issue. Following each discussion, each partner completed an empathic accuracy task followed by post-discussion measures. We predicted that highly avoidant individuals would exhibit lower levels of empathic accuracy than less avoidant people, regardless of the severity (major vs. minor) of the discussed topic. In contrast, highly anxious individuals should display more vigilance (greater empathic accuracy) than less anxious individuals when discussing a major relationship issue.

The discussion topics (intimacy vs. jealousy) were chosen because they are primary sources of concern and represent potential danger zones (Ickes & Simpson, 2001), particularly for insecurely attached people. We anticipated that highly anxious individuals would display higher levels of empathic accuracy when discussing a severe (major) intimacy topic given that such issues may be more common in established marriages, whereas jealousy issues might vary in terms of their relevance to certain marriages. However, intimacy and jealousy should both have the potential to activate the attachment systems of highly avoidant people, leading them to display uniformly lower levels of empathic accuracy.

Results

Descriptive Statistics

Descriptive statistics are reported in Table 1. Means for the ease-of-inference index are normed to the 7-point rating scale used to assess this dimension. The empathic accuracy index could range from 0 (total inaccuracy in inferring the content of the partner’s thoughts/feelings) to 100 (perfect accuracy). According to the raters, it was easier to infer the thoughts/feelings from the behavior of men than it was from women. No other gender differences emerged. Correlations among these variables are shown in Table 2.

Table 1

(Study 1) Descriptive Statistics

	Men:	Women:
	Mean	Mean	Mean Difference	Matched–Pairs t-tests
Empathic Accuracy	26.01 (15.59)	26.03 (14.98)	.02 (18.00)	t = .21, ns
Attachment Avoidance	27.89 (8.82)	24.11 (8.43)	3.78 (12.18)	t = 3.03, p < .01
Attachment Anxiety	25.14 (9.06)	25.95 (10.44)	.81 (11.33)	t = −.70, ns
Ease-of-Inference (observer-rated)	3.94 (1.00)	3.57 (.93)	.37 (1.19)	t = 3.02, p < .01

Open in a separate window

Note: N = 95 couples. Standard deviations appear in parentheses below each mean.

Table 2

(Study 1) Correlations Among the Variables

	A	B	C	D	E	F	G	H
A	--	−.14	−.09	−.08	.21	−.08	−.10	−.11
B		--	−.07	−.03	−.15	.00	.16	−.09
C			--	−.01	.14	.12	.33**	.08
D				--	−.14	−.00	.12	.24*
E					--	−.12	.06	−.28**
F						--	.15	.03
G							--	−.03
H								--

Open in a separate window

Note: N = 95 couples. All correlations are two-tailed. Higher scores indicate higher values on each variable.

A = Male empathic accuracy; B = Male avoidance; C = Male anxiety; D = Male ease-of-inference (observer-rated); E = Female empathic accuracy; F = Female avoidance; G = Female anxiety; H = Female ease-of-inference (observer-rated).

^*p < .05

^**p < .01

Manipulation check

To test the effectiveness of our primary manipulated variable (being assigned to discuss a major vs. a minor relationship problem), we examined the amount of stress reported by each participant during the discussion on the 3-item stress measure. As expected, individuals assigned to discuss more severe problems reported more stress (M = 15.43) than did those in the less severe condition (M = 12.34), t(188) = 2.81, p < .01.

The APIM Analyses

Because husbands’ and wives’ scores were significantly correlated (indicating some degree of statistical interdependence), we analyzed the data using the Actor-Partner Interdependence Model (APIM; Kashy & Kenny, 2000; Kenny, 1996). The APIM allows one to estimate the degree to which dyad members’ responses or behaviors are associated with factors attributable to the actor (the individual providing the response/behavior) and to the actor’s partner. The APIM, therefore, estimates both actor effects (the effect that an individual’s predictor variable score has on his/her own outcome score) and partner effects (the effect that an individual’s partner’s predictor variable score has on the actor’s outcome score). Because the APIM models the statistical interdependence that exists between partners in relationships, it provides separate, statistically independent tests of actor and partner paths. Specifically, the effects of the actor’s independent variable score on the actor’s dependent measure control for the partner’s independent variable score, and vice versa. Using this approach, the dyad is treated as the unit of analysis, and actor and partner effects are tested with the proper degrees of freedom.

Tests of Predictions

The APIM analyses were conducted using the PROC MIXED program in SAS 9.0. Actor effects are reported as regression coefficients, all of the independent variables are standardized, and the primary dependent variable (empathic accuracy) is unstandardized. All predictor variables were centered on the grand sample mean (Aiken & West, 1991). The dependent variable in each analysis was each actor’s empathic accuracy score. The predictor variables were: actor and partner scores on attachment anxiety and avoidance, problem topic (jealousy vs. intimacy), and problem severity (major vs. minor). We also entered the two-way interactions between each attachment score (for both actor and partner) and problem topic and problem severity, the two-way interaction between problem topic and problem severity, and all theoretically-relevant three-way interactions. Actor gender and actor ease of partner readability were entered as covariates. Gender did not significantly interact with any of the predictor variables, meaning that none of the effects reported below were qualified by gender differences. Moreover, there were no significant attachment anxiety by attachment avoidance effects for actors or partners. All statistically significant findings that emerged are reported below.

First, as predicted, a significant main effect for attachment avoidance indicated that highly avoidant individuals were less empathically accurate than their less avoidant (more secure) counterparts, b = −.33, t(141) = −2.26, p < .03. This finding, which is consistent with Rholes et al. (2007), indicates that highly avoidant people are less likely to “get into the heads” of their romantic partners during relationship-threatening interactions. There was no interaction between attachment avoidance and problem severity, b = .02, t(140) = 0.12, ns.

Second, a three-way interaction between problem topic, actor attachment anxiety, and problem severity emerged, b = −.31, t(147) = −2.35, p < .02 (see Figure 1). The two-way interaction between actor anxiety and condition severity predicting actor empathic accuracy in the intimacy condition was also significant (b = −.66, t(146) = 2.36, p < .02).^3,4 Among couples who discussed intimacy issues, highly anxious individuals (actors) displayed higher empathic accuracy when the problem was relatively more severe compared to less anxious individuals, who showed the opposite pattern. Simple slope analyses, however, revealed that neither the high anxiety nor the low anxiety regression lines differed significantly from zero (both ts < 1.50, ns). Among couples who discussed jealousy topics, no interaction between anxiety and problem severity was found.

Open in a separate window

Figure 1

The interaction of actors’ attachment anxiety, type of problem, and severity of problem predicting actors’ empathic accuracy. The top figure is for couples who discussed intimacy issues; the bottom figure is for those who discussed jealousy issues. Values are plotted for individuals scoring 1 SD above and 1 SD below the mean for each predictor variable. Lo AAnx = low actor anxiety; Hi AAnx = high actor anxiety; Lo sev = low problem severity; Hi sev = high problem severity.

All of these effects remained statistically significant (all ps < .05) when the partner’s readability ratings were partialed out. This outcome is important because it indicates that these findings are not due to the degree to which participants expressed what they were thinking/feeling during their discussions.

Study 2

Study 2 was designed to replicate, clarify, and extend the findings of Study 1. If the effects found in Study 1 are robust, they should also emerge when a different type of romantic relationship (long-term dating relationships), attachment-relevant interaction (conflict resolution), and method of measuring distress (observer-ratings of how distressed each partner appeared during the discussion) are examined. In Study 1, we manipulated topic severity by asking couples to discuss an intimacy or a jealousy topic that posed a major or a minor relationship problem. However, we could not ensure that every couple maintained their assigned level of problem severity throughout the entire discussion. Accordingly, we used a more direct and sensitive measure of discussion severity in Study 2; we had trained observers code each participant’s level of distress during the videotaped discussion with his/her dating partner. Moreover, to further generalize the findings beyond the methods used in Study 1, we had partners discuss an unresolved conflict in their relationship.

Specifically, we asked long-term dating couples in Study 2 to identify and try to resolve a current conflict in their relationship. Immediately following each discussion, each partner completed the empathic accuracy task privately. Similar to Study 1, we predicted that highly avoidant individuals would display lower levels of empathic accuracy in general during the conflict resolution task. We also predicted that highly anxious individuals would display comparatively greater empathic accuracy if they were rated as more distressed during their conflict discussions, but lower empathic accuracy if they appeared less distressed.

If our predicted effects are robust, they should also remain significant after statistically controlling for several alternative constructs. Thus, we also tested for the potential influences of each participant’s neuroticism, relationship satisfaction, and relationship length. Highly neurotic individuals might be more empathically accurate, especially if they feel threatened (Karney & Bradbury, 1995). Because attachment anxiety correlates moderately with neuroticism (Brennan & Shaver, 1995), this possibility must be ruled out. Insecurely attached people also have relationships that tend to be shorter-lived (J. Feeney, 2008; Kirkpatrick & Davis, 1994) and less satisfying (J. Feeney, 2008; Simpson, 1990). It is also conceivable that people who are more satisfied with their relationships or have dated their partners for longer periods of time might display heightened empathic accuracy when distressed, given that such people might have more to lose if their relationships ended (e.g., Rusbult, 1980). Thus, the effects of relationship satisfaction and stability must also be controlled.

Results

Descriptive Statistics

Table 3 reports the descriptive statistics for Study 2. Women and men displayed low-to-moderate levels of empathic accuracy, similar to previous conflict resolution studies of empathic accuracy (e.g., Fletcher & Thomas, 2000). Participants were rated as displaying low-to-moderate levels of stress/anxiety, and participants’ thoughts/feelings were rated as moderately difficult to infer based on what they said/did during their discussions. Matched-pairs t-tests revealed one marginally significant gender difference. Similar to Study 1, men’s thoughts/feelings were rated as slightly more “readable” than women’s. Zero-order correlations among the Study 2 variables are shown in Table 4.

Table 3

(Study 2) Descriptive Statistics

	Men:	Women:
	Mean	Mean	Mean Difference	Matched–Pairs t-tests
Empathic Accuracy	20.22 (13.58)	22.24 (13.64)	−1.95 (16.72)	t = −1.11, ns
Attachment Avoidance	25.69 (8.27)	25.97 (7.79)	−.28 (11.44)	t = −.24, ns
Attachment Anxiety	27.53 (9.38)	29.02 (10.00)	−1.49 (11.61)	t = −1.26, ns
Stress/Anxiety (observer-rated)	24.77 (1.83)	25.02 (1.73)	−.25 (1.49)	t = −1.68, ns
Ease-of-Inference (observer-rated)	4.06 (1.23)	4.35 (1.12)	−.29 (1.46)	t = −1.85, p = .068

Open in a separate window

Note: N = 96 couples. Standard deviations appear in parentheses below each mean.

Table 4

(Study 2) Correlations Among the Variables

	A	B	C	D	E	F	G	H	I	J
A	--	−.16	−.19	−.05	.03	.25*	−.09	.03	.03	.29**
B		--	.16	.15	−.07	.08	−.01	.23*	−.05	.07
C			--	.06	−.02	−.02	.14	.28**	−.15	−.08
D				--	.01	−.06	.08	−.12	.65**	−.08
E					--	.09	−.22*	−.11	−.07	.22*
F						--	−.27*	.02	.00	.02
G							--	.16	.11	−.02
H								--	−.08	−.02
I									--	−.02
J										--

Open in a separate window

Note: N = 96 couples. All correlations are two-tailed. Higher scores indicate higher values on each variable.

A = Male empathic accuracy; B = Male avoidance; C = Male anxiety; D = Male stress/anxiety (observer-rated); E = Male ease-of-inference (observer-rated); F = Female empathic accuracy; G = Female avoidance; H = Female anxiety; I = Female stress/anxiety (observer-rated); J = Female ease-of-inference (observer-rated).

^*p < .05

^**p < .01

APIM Analyses and Tests of Predictions

Because partners’ scores were significantly correlated, we analyzed the data using the APIM (Kashy & Kenny, 2000; Kenny, 1996). Actor effects are reported as regression coefficients, the independent variables are standardized, and the primary dependent variable (empathic accuracy) is unstandardized. All predictor variables were centered on the grand sample mean (Aiken & West, 1991).

The APIM analyses were conducted using the PROC MIXED program in SAS 9.0. The dependent measure was each actor’s empathic accuracy score. The ease-of-inference index was entered as a covariate to control for variability in how easy it was to infer each partner’s thoughts/feelings based on what s/he said or did during the conflict discussion. The predictor variables were actor and partner scores on attachment anxiety and avoidance, the actor’s observer-rated stress/anxiety index, actor gender, and the condition to which each couple was assigned (major vs. minor conflict). All two-way and theoretically relevant three-way interactions were also entered. No significant three-way interactions emerged. None of the interactions involving gender were significant, so none of the effects reported below are qualified by gender differences.

Both of the effects we predicted were found. First, a main effect for avoidance indicated that highly avoidant individuals were less empathically accurate in general than were their less avoidant counterparts, b = −.37, t(142) = −2.65, p < .01. This replicates what we found in Study 1, and it is consistent with other research showing that highly avoidant people use deactivating strategies in relationship-threatening situations (Fraley & Shaver, 2000; Mikulincer & Florian, 1998). It also provides further evidence that highly avoidant people do, in fact, limit the monitoring of their partners’ thoughts and feelings in attachment-relevant situations.

Second, as depicted in Figure 2, the interaction between actor’s attachment anxiety and observer-ratings of actor’s stress/anxiety was found, b = .13, t(150) = 1.92, p = .056.⁵ Compared to less anxious people, highly anxious individuals displayed greater empathic accuracy if they were rated as more stressed/anxious during their discussions, but less empathic accuracy if their stress/anxiety was rated as lower. Simple slope analyses revealed that neither slope differed significantly from 0 (both ts < 1.50, ns). Thus, similar to Study 1, highly anxious people displayed heightened empathic accuracy when they were more distressed than when they were less distressed relative to less anxious people. These results suggest that highly anxious individuals tend to become more vigilant to what their partners are thinking and feeling when their partners might be harboring relationship-threatening thoughts/feelings.

Open in a separate window

Figure 2

The interaction of actors’ attachment anxiety and observer-rated stress/anxiety predicting actors’ empathic accuracy. Values are plotted for individuals scoring 1 SD above and 1 SD below the mean for each predictor variable. Lo Anx = low actor anxiety; Hi Anx = high actor anxiety; Lo Stress = low observer-rated stress; Hi Stress = high observer-rated stress.

Both of the effects reported above also held when the “readability” of partners’ thoughts/feelings (the ease-of-inference index) was partialed (both ps < .05). This discounts the possibility that highly anxious individuals become more empathically accurate when distressed because their partners’ thoughts/feelings are more directly conveyed in their discussions, or that highly avoidant individuals are generally less empathically accurate because their partners’ thoughts/feelings are more difficult to infer from their statements or actions.

Discriminant Validity Analyses

To provide further evidence for the discriminant validity of these effects, we controlled for three potential confounds: relationship length, relationship satisfaction, and neuroticism. When each potential confound was partialed, both of the effects reported above remained statistically significant or marginally significant.

We also re-conducted the Study 2 analyses to determine whether the effects remained significant when the proportion (the percentage) of negative-to-total thoughts and feelings reported by each actor’s partner was controlled. Insecurely attached people may, for example, have partners who report a higher percentage of negative thoughts/feelings relative to their total thoughts/feelings during the conflict discussion, which might have impacted the empathic accuracy results. Both of the effects reported above, however, were significant when the proportion of negative-to-total partner thoughts/feelings was statistically controlled (both ps < .05). These predicted effects, therefore, are robust.

General Discussion

The goal of this research was to investigate how individuals who have different attachment orientations “manage” empathic accuracy in relationship-threatening situations with their romantic partners. In two social interaction studies, we examined patterns of empathic accuracy associated with attachment anxiety and avoidance in different attachment-relevant interactions (intimacy vs. jealousy problem discussions, and conflict resolution discussions) that entailed different types of relationships (marital and dating) and different methods of measuring distress (experimentally manipulated and observer-rated). Study 1 tested empathic accuracy patterns in married couples who discussed a major or minor relationship problem that focused on intimacy or jealousy issues. Study 2 tested empathic accuracy patterns in long-term dating couples who discussed an unresolved conflict in their relationship. For both studies, we hypothesized that highly avoidant people should display lower levels of empathic accuracy in general than less avoidant (more secure) people. This hypothesis was supported in both studies. In fact, in many cases, the level of empathic accuracy exhibited by highly avoidant persons was slightly above zero, representing almost total inaccuracy.

We also hypothesized that highly anxious people would display greater empathic accuracy when discussing relationship issues that were more versus less relationship-threatening. Although highly anxious individuals did not experience significant increases in empathic accuracy in higher relative to lower threat conditions, they did display significantly greater empathic accuracy when discussing relationship issues that were more versus less relationship-threatening compared to less anxious (i.e., more secure) people. Perhaps to protect themselves or their relationships from harmful information, less anxious people experienced slight decreases in empathic accuracy when more relationship-threatening problems were discussed. This pattern emerged for high versus low anxious spouses who discussed more severe relationship-relevant intimacy problems (in Study 1) and for high versus low anxious dating partners who discussed relationship conflicts that generated more observer-rated distress (in Study 2).

Our confidence that these effects are likely to be the result of motivational tendencies of highly anxious and highly avoidant individuals is bolstered by the fact that the results of Study 2 remain even after both partners’ scores on neuroticism, relationship satisfaction, and relationship length were partialed out. In other words, the relation between attachment anxiety and greater empathic accuracy when individuals are more distressed is not attributable to the variance that attachment anxiety shares with these potential confounds.

These findings clarify the conditions that generate empathic accuracy and inaccuracy in romantic relationships when individuals who have different attachment orientations discuss difficult, potentially relationship-threatening topics. When interpreting the results, a few key points must be kept in mind. First, in Study 1, the experimentally manipulated variable of relationship problem severity generated increases in empathic accuracy only in those people who, according to attachment theory, should experience such increases—highly anxious people. Second, the experimental instructions in both studies were designed to decrease the likelihood that empathic inaccuracy could have resulted from participants’ censoring or failing to report their own actual thoughts and feelings or the thoughts/feelings inferred from what their partners were thinking/feeling during the discussions. In both studies, each partner was explicitly instructed to be as accurate as possible, and all participants reported during debriefing that they followed these instructions. This suggests that the low levels of empathic accuracy exhibited by highly avoidant persons—in some cases approaching 0—were not attributable to a failure to perform the empathic accuracy task properly. Moreover, our patterns of empathic accuracy are not the result of anxious participants’ partners conveying their thoughts/feelings more clearly, or the result of avoidant participants’ partners’ thoughts/feelings being more difficult to infer. In both studies, the effects remained significant after controlling for the partner’s readability.

The current findings contribute to our understanding of attachment and empathic accuracy in two significant ways. First, both studies provide the first evidence that highly anxious people may be more motivated to “get into their partner’s head” during relationship-threatening discussions. Second, both studies provide the first evidence that highly avoidant people display lower levels of empathic accuracy in general, even if the topics and issues being discussed are not likely to be highly threatening in nature. No prior studies, including the one by Simpson et al. (1999), have documented that highly avoidant people display empathic inaccuracy as a “default” strategy during attachment-relevant interactions with their romantic partners.

Footnotes

Contributor Information

Jeffry A. Simpson, University of Minnesota, Twin Cities Campus.

John S. Kim, University of Minnesota, Twin Cities Campus.

Jennifer Fillo, University of Minnesota, Twin Cities Campus.

William Ickes, University of Texas at Arlington.

W. Steven Rholes, Texas A&M University.

M. Minda Oriña, Saint Olaf College.

Heike A. Winterheld, California State University at Eastbay.

References