Adult Development and Aging

On the Importance of Relationship Quality: The Impact of Ambivalence in Friendships on Cardiovascular Functioning

Julianne Holt-Lunstad, Ph.D. Brigham Young University

Bert N. Uchino, Ph.D., Timothy W. Smith, Ph.D., and Angela Hicks, Ph.D. University of Utah

ABSTRACT

Background: Social relationships are reliably related to rates of morbidity and mortality. One pathway by which social relationships may influence health is via the impact of relationship quality on cardiovascular reactivity during social interactions. Purpose and Method: This study exam- ined the effects of the quality of a friendship on cardiovascu- lar reactivity when speaking about positive or negative life events with an ambivalent or supportive friend. To examine this, 107 healthy male and female adults (and their same- sex friend) were recruited. Results: Results revealed that participants exhibited the greatest levels of systolic blood pressure reactivity when discussing a negative event with an ambivalent friend compared to a supportive friend. We also found higher resting levels of heart rate and lower respiratory sinus arrhythmia among those who brought in ambivalent friends than those who brought in supportive friends. Conclusion: Individuals may not be able to fully relax in the presence of ambivalent friends and may not benefit from support during stress. This research may help clarify the health-related consequences of differing types of social relationships.

(Ann Behav Med 2007, 33(3):278–290)

INTRODUCTION

Social relationships play a central role in our everyday lives, and evidence suggests that these relationships influ- ence physical health (1,2). Social support is inversely related to the prevalence and incidence of several diseases; the most common being coronary heart disease (CHD) (3–5). These findings are often taken as evidence for the health benefits of social relationships, however, even close relationships are not entirely positive (6). Although social relationships can be sources of warmth, caring, nurturance, and understanding, they can also be sources of conflict, criticism, jealousy, and rejection (7). Research from the Terman Life Cycle study suggests that past negativity in social relationships predicts earlier mortality (8,9).

Therefore, a more complete understanding of the health- related consequences of social relationships requires con- sideration of both their positive and negative aspects. Of importance, positivity and negativity are separable dimen- sions (10–13), and many individuals construe their relation- ships as having a mix of positive and negative feelings (e.g., an overbearing mother, competitive friend, volatile romance [14]). As noted by others, the implications of this ‘‘ambivalence’’ within relationships have not been adequately considered in social support theory or research (7,14,15).

Given that ambivalent relationships are characterized by a mix of both positive and negative feelings, it is unclear what the implications of these relationships may be for our health. Can individuals benefit from the positivity associa- ted with ambivalent relationships or is negativity more sali- ent, leading to harmful effects? Although more research is needed, our prior findings suggest that ambivalent relation- ships are distinct from supportive (i.e., primarily positive), indifferent (i.e., not particularly positive or negative), and aversive (i.e., primarily upsetting) relationships and may be related to detrimental outcomes (16). For example, older adults with higher numbers of ambivalent relation- ships within their network had higher levels of depressive symptoms and greater cardiovascular reactivity (CVR) during stress (14). In addition, ambulatory systolic blood pressure (SBP) was highest when participants were inter- acting with individuals they rated as normally feeling ambivalent toward compared to those toward whom they felt primarily positive, indifferent, and primarily negative (17). Thus, individuals may not be able to benefit from the positivity associated with ambivalent relationships and such ambivalence may be more detrimental than sim- ple negativity.

Prior research indicates that ambivalent relationships may be health relevant; however, they have failed to eluci- date the context in which ambivalent ties may be harmful. For instance, although the ambulatory blood pressure study (17) did assess the quality of the interaction, infor- mation on the exact nature of the interaction (e.g., whether they were actually seeking support or simply having a cas- ual conversation) was not obtained. In our prior work we have found that individuals did not benefit from support received from an ambivalent friend compared to a support- ive friend (18). However, this study did not include other interaction contexts (e.g., discussion of positive events) as

Reprint Address: J. Holt-Lunstad, Ph.D., Department of Psy- chology, Brigham Young University, 1024 SWKT, Provo, UT 84602. E-mail: [email protected]

# 2007 by The Society of Behavioral Medicine.

278

comparisons. Participants in this study were also not ran- domly assigned to relationship conditions, hence making this study open to alternative explanations (e.g., due to existing personality differences). Thus, we need to better understand the context that may explain how such relation- ships impact health.

Two potential broad mechanisms could underlie the associations between ambivalent relationships and detri- mental cardiovascular outcomes. First, interactions with ambivalent ties may entail significant interpersonal stress (i.e., stress enhancing hypothesis). If a social network mem- ber is primarily a source of negativity, one may habituate to the aversive relationship by using specific coping strategies (e.g., discounting or simply avoiding these people). How- ever, an ambivalent network member (someone you feel particularly positive and negative toward) may be less pre- dictable, less easily discounted or avoided, and thus may be associated with heightened interpersonal stress (14,19). Thus, individuals may not be able to relax when in the pres- ence of or while interacting with ambivalent network ties. A second pathway involves interference with social support (i.e., support interference hypothesis). Social relationships that are supportive are thought to have an effect on health by buffering the negative effects of stress. However, despite the positivity that exists in ambivalent relationships indivi- duals may not be able to benefit from support from such ties, because the coexisting negativity may lead them to either avoid disclosure or question the intent, sincerity, or possible subsequent consequences of support (20). Although the stress-enhancing and support interference hypotheses rep- resent distinct pathways by which ambivalence may influ- ence health, they are not necessarily competing hypotheses as both may potentially be operating (e.g., it is possible that stress may be operating to interfere with support).

We tested these hypotheses by examining the impact of ambivalent relationships on CVR in a controlled experimental design. Exaggerated CVR to stressful situations may influence the development and=or expression of CHD (for reviews, see 21,22). We randomly assigned participants to discuss positive, negative, and neutral life events with an ambivalent or supportive friend. If ambivalent relationships are general sources of stress (i.e., stress-enhancing hypoth- esis), we would expect increased reactivity when individuals are discussing both positive and negative events with an ambivalent friend as compared to a supportive friend. How- ever, if individuals simply do not benefit from support during times of stress (i.e., support interference hypothesis), then we should see increased reactivity primarily when participants disclose negative events to an ambivalent friend.

METHOD

Design

Three between-participants factors were included: relationship quality (supportive, ambivalent), event (positive,

negative), and participant gender (male, female). The depen- dent variables were residualized change scores for physiologi- cal measurements and several self-report measures. Participants were randomly assigned to the conditions of relationship quality and event discussion. Relationship type and partner gender was held constant, with all participants required to bring a same-sex friend.

Participants

Participants were recruited from introductory psy- chology courses at the University of Utah, received extra course credit, and included 56 women and 51 men (N ¼ 107). The participant’s friend received $15 compen- sation. Consistent with prior research (e.g., 23), the follow- ing self-reported inclusion criteria were used to select healthy participants: no hypertension, no cardiovascular prescription medication, no history of chronic disease with a cardiovascular component (e.g., diabetes), no recent his- tory of psychological disorder (e.g., major depressive dis- order), no tobacco use, and no consumption of more than 10 alcoholic beverages per week.

Procedure

Testing was divided into two sessions. In the first, part- icipants were asked to complete the Social Relationships Index SRI (detailed next) in which they rated their social network. We then asked the participants to contact and bring in a specific friend from their list of rated relation- ships. This individual was selected by the experimenter based on each participant’s random assignment to relation- ship (supportive or ambivalent) condition.1 Participants were unaware of the relationship condition. Approximately 2 weeks later, participants brought their friend to the lab- oratory (see Table 1 for summary of procedures and mea- sures). Upon arrival both individuals completed an informed consent document and a demographic question- naire. Participants’ self-reports of health were checked for reliability against the inclusion criteria. Participants and their friend were told that because cardiovascular function- ing can differ according to whether one is speaking or not, we needed to be certain that participant was speaking dur- ing certain portions of the experiment. The friends were told that part of their responsibility during the experiment was to keep their friend talking. The friend was told, ‘‘Sometimes the best way to do this is to be supportive and sometimes it is to stir things up a bit; however, all we ask is that you say what would come natural to you in the situation.’’

Participants were then escorted by themselves to a separate sound-attenuated room where four mylar bands

1 Two individuals did not have both ambivalent and support- ive relationships listed on their SRI. When dropping these indivi- duals from our analyses the results remain consistent; therefore, results reported include these participants’ data.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 279

were placed in the tetrapolar configuration for impedance cardiograph recordings according to published guidelines (24). An occluding cuff was placed on the left arm. They were seated in a comfortable chair and asked to complete questionnaires (see measures). Following this adaptation period of approximately 20 min, the friend was escorted into the room and seated next to the participant. The par- ticipant and friend were instructed to relax for the next 12 min while resting measures of cardiovascular function were obtained. Both were told not to talk during the rest period, and a curtain was drawn between them. As a neu- tral comparison to the event discussion, the participant and friend were instructed to discuss with each other for 4 min what they do during a normal weekday, alternating speak- ing and listening for 1-min epochs to hold constant speech effects on CVR (e.g., 25).

Before the event task began, a second baseline was also obtained. The procedures were identical to the first with the exception that it was 7 min in length. Participants were then asked to list up to five past experiences that were particularly negative=positive (depending on random assignment) that they felt comfortable discussing with their friend as part of our experiment. These were rated on a scale of 1 to 5 on degree of importance, positivity, and negativity, in comparison to all possible negative=positive events in their life. The experimenter selected a situation= event that was rated as moderately high such that all dis- cussions, regardless of condition, would be relatively equivalent on these aspects to control for intensity. Exam- ples of negative events discussed included almost getting in a car accident, making a commitment but then didn’t fol- low through, meeting girlfriend’s family, boyfriend still maintains friendship with ex-girlfriend, getting fired from part-time job, and finals. Examples of positive events discussed included boyfriend flying in from out of town for surprise visit, spending the day with Mom, getting

a promotion at work, and making the dance team. The participant and friend discussed the event for a total of 6 min. Participants were asked to discuss three aspects of the selected situation=event: first to describe in detail the experience=event, second to talk about his or her thoughts and feelings regarding this situation=event, and finally to discuss how they handled the situation=event and=or how they might have changed anything. After each aspect, the friend was given 1 min to respond. Thus, again they alter- nated speaking and listening for 1-min epochs to hold constant speech effects on CVR.

Cardiovascular assessments were taken during each dis- cussion task and the last 5 min of each baseline. Heart rate (HR), impedance-derived measures, and respiratory sinus arrhythmia (RSA) were obtained continuously, while systolic blood pressure (SBP) and diastolic blood pressure (DBP) were obtained once every 90 sec during baseline and at the begin- ning of each minute during the discussion tasks. Self-reported state anxiety was assessed before and after each discussion task. Perceptions of the friend during each of the discussions were rated using the Impact Message Inventory (IMI; see description next) and completed immediately following each discussion. At the completion of the valenced discussion part- icipants completed a rating of their thoughts and feelings regarding the discussion as a manipulation check.

Measures

Self-Report Measures

Social relationships index (SRI). The SRI was deve- loped as a self-report version of the social support inter- view (26,27). Participants rated up to 10 friends in terms of how helpful and upsetting they were from 1 to 6 (extremely), and on length of time known, amount of con- tact per week, importance, predictability, and how likely they are to go to this person for support. In our prior work,

TABLE 1

Physiological and Psychological Dependent Measures Taken During Each Experimental Epoch

Experimental Epoch Length of Epoch Measurement(s)

Prescreening SRI: rating up to 10 friends Prebaseline 1 Consent forms, background and health questionnaires, ratings of friend (e.g.,

SRI, QRI) Baseline 1 12 min SBP, DBP, HR, RSA, PEP, TPR, CO Postbaseline1 State anxiety, Neutral discussion 4 min SBP, DBP, HR, RSA, PEP, TPR, CO Postneutral Discussion State anxiety, IMI Baseline 2 7 min SBP, DBP, HR, RSA, PEP, TPR, CO Postbaseline 2 State anxiety Prep for Discussion 2 Experience rating Discussion 2 6 min SBP, DBP, HR, RSA, PEP, TPR, CO Postdiscussion 2 State anxiety, IMI, Postdiscussion Rating, and thought listing

Note. SRI ¼ social relationships index; QRI ¼ quality of relationships index; SBP ¼ systolic blood pressure; DBP ¼ diastolic blood pressure; HR ¼ heart rate; RSA ¼ respiratory sinus arrhythmia; PEP ¼ pre-ejection period; TPR ¼ total peripheral resistance; CO ¼ cardiac output; IMI ¼ impact message inventory.

280 Holt-Lunstad et al. Annals of Behavioral Medicine

these measures of positivity and negativity were temporally stable with significant 2-week test–retest correlations (r ¼ .81, p < .001 for positivity; r ¼ .83, p < .001 for negativity) (14). Scores were used to determine the relation- ship qualities (i.e., supportive, ambivalent, aversive, indif- ferent) in the participant’s social network. We selected a friend who was rated as either supportive or ambivalent (depending on random assignment) for them to bring to the laboratory. Although our model also includes relation- ships that would be classified as aversive (high negativity, low positivity) and indifferent (low positivity, low nega- tivity), it would have been unfeasible and unlikely that we could get participants to bring such relationships in with them to the lab, so we focused primarily on friendships. Friends were rated in terms of three different contexts: (a) when they need support such as advice, understanding, or a favor; (b) when they are excited, happy, or proud of some- thing; and (c) during routine daily interactions, conver- sations, or activities. The individual selected to come in with the participant was the individual listed who most highly characterizes the assigned condition (i.e., supportive or ambivalent) across all three contexts. Consistent with our prior work, the criterion used for the classification of supportive was a rating of a 3 or greater on measures of positivity and a rating of 1 (not at all) on negativity. For the classification of ambivalent, a rating of a 3 or greater on the measures of positivity and a rating of greater than 1 on negativity were required.

Quality of relationships index (QRI). The QRI (28) assesses supportive and conflictual aspects of relationships and was used as an independent check on the SRI-based relationship classification. This 29-item inventory includes

three scales that assess perceptions of social support, conflict, and depth in a specific relationship. The internal reliability was found to be high (.85, .91, and .84, respect- ively) for ratings of friends (28).

IMI, form II. The IMI (29) is a circumplex-based inventory designed to assess perceptions of another’s interpersonal behavior, along the dimensions of friendliness versus hostility and dominance versus submissiveness. It contains 32-items with 4 items per octant. Kiesler and colleagues (30) provided evidence supporting the circumplex structure of the IMI and demonstrated its adequate psychometric properties. It has been found to be sensitive to similar interpersonal manipulations in prior research (31).

State anxiety scale. A short form of the Spielberger State-Trait Anxiety Scale was administered to participants prior to and following each discussion (32). This measure was used as a manipulation check of the stressfulness of discussing a negative versus a positive event=situation. Prior work has found the internal consistency of the scale to be high (Cronbach’s a ¼ .78–.80).

Additional manipulation checks. To ensure that manipulations were having their intended effect, a short questionnaire was devised that we refer to as the Post- Discussion rating (see Table 2 for list of items). Participants were asked to rate their current feelings on a 6-point scale from 1 (not at all) to 6 (extremely). For

TABLE 2

Mean Post Discussion Ratings as a Function of Relationship Quality

Postdiscussion Ratings Supportive Ambivalent

1. How open were you to disclosing this event with your friend? 5.39 (0.86) 5.33 (0.89) 2. How comfortable did you feel discussing this event with your friend?�� 5.44 (0.72) 4.96 (1.13) 3. How helpful was your friend during the discussion?�� 4.59 (1.14) 3.98 (1.41) 4. How challenging was the event discussion task? 2.81 (1.47) 2.83 (1.40) 5. How upsetting was your friend during the discussion?�� 1.19 (0.39) 1.54 (0.88) 6. How mixed and conflicted were your thoughts and feelings toward your friend during the discussion?�� 1.30 (0.60) 1.65 (0.78) 7. How effortful was it to do the discussion task? 2.70 (1.34) 2.63 (1.26) 8. How threatening was the event discussion task? 1.43 (0.86) 1.65 (0.91) 9. How difficult did you find the discussion task? 2.11 (1.19) 2.24 (1.33)

10. How natural was this discussion compared to normal discussions with this friend?� 3.52 (1.38) 2.91 (1.42) 11. To what extent did your friend respond as normally as he or she would outside this experiment? 4.24 (1.32) 3.85 (1.41) 12. How familiar was your friend with the event that you discussed?y 3.56 (1.89) 4.15 (1.64) 13. When discussing this event with this person in the past, how helpful has your friend been?� 4.63 (1.31) 4.06 (1.50) 14. When discussing this event with this person in the past, how upsetting has your friend been?�� 1.20 (0.53) 1.62 (0.93) 15. To what extent have you discussed this event with this friend before? 2.41 (1.11) 2.61 (1.12) 16. To what extent have you discussed this event with anyone before? 2.69 (1.10) 2.78 (0.90)

Note. Values are mean (standard deviation). yp < .10. �p < .05. ��p < .01. ���p < .001.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 281

Items 15 and 16, the scale was from 1 (never) to 6 (at great lengths).

Cardiovascular Measures

Blood pressure. A Dinamap Model 8100 monitor (Critikon Corporation, Tampa, FL) was used to measure SBP and DBP. The Dinamap uses the oscillometric method to estimate blood pressure. Blood pressure assessments were obtained via a properly sized occluding cuff positioned on the upper left arm of the participant according to the manufacturer’s specifications. Mean SBP and DBP for each epoch were averaged across minutes to increase the reliability of these assessments.

Impedance-derived measures. A Minnesota Impedance Cardiograph Model 304B was used to measure the electrocardiogram, basal thoracic impedance, and the first derivative of the impedance signal (dZ=dt). Cardiac output (CO), total peripheral resistance (TPR), and pre- ejection period (PEP) were derived according to published guidelines (33). For a more detailed description of these measures and methodology see Uchino et al. (34).

RSA. RSA provides a noninvasive measure of parasympathetic control of the heart and can vary according to age and posture2 (see 35). RSA was calculated based on the digitized interbeat intervals that were checked and edited for artifacts using the detection algorithm of Berntson and colleagues (35). After linear detrending, the heart period time series was band pass filtered from .12 to .40 Hz using an interpolated finite impulse response filter (36). RSA was then calculated as the natural log of the area under the heart period spectrum (calculated by a Fast Fourier Transform and scaled to msec2=Hz). RSA was calculated on a minute- by-minute basis and aggregated across minutes within each epoch to increase reliability.

RESULTS

Manipulation Checks and Self-Report Measures

Relationship Classification and Correlates

To verify the classification of relationship types, we examined its stability across sessions. The ratings of posi- tivity and negativity taken together indicate that our classi- fication criteria were met in most cases, and the overall

classification remained relatively stable.3 The test–retest correlations for ratings of positivity and negativity are r ¼ .60 (p < .0001) and r ¼ .72 (p < .0001), respectively.

In a second set of analyses we examined the association between our initial social relationship classification with the independent measure of social relationships (i.e., QRI). Results revealed a significant main effect for relationship quality (as rated on the SRI) on the Support, Conflict, and Depth scales of the QRI (see Table 3). Friends classified as ‘‘Supportive’’ on the SRI were also rated higher on support and depth, and lower in conflict, than were friends classified as ‘‘Ambivalent.’’ Finally, we examined if friendships characterized as ‘‘ambivalent’’ or ‘‘supportive’’ differed on other relationship dimensions.4

There was no significant difference on length of time they have known these friends and amount of contact with these friends. There was also no significant difference in predict- ability of supportive and ambivalent friends. However, friends classified as supportive had higher ratings of impor- tance than friends classified as ambivalent, although both types of friends were perceived as highly important. For items assessing how likely the participant was to go to their friend during the three (negative, positive, and neutral) contexts, we found a relationship categorization main effect for the positive, negative, and neutral contexts. Part- icipants’ were more likely to seek out their supportive friend than ambivalent friends in positive situations (when ‘‘really excited, happy or proud of something’’), during stressful times (‘‘When you need support such as advice, understanding, or a favor. . .’’), and during ‘‘routine daily interactions, conversations, or activities.’’ Despite the relative difference, ratings of likelihood to seek out the friend were high across both supportive and ambivalent friends.

2 Based on the average age of our sample and RSA being

assessed in the seated position, the relative mean population RSA value would be 6.6.

3 There were 19 participants originally classified as supportive

that increased in negativity at Time 2 such that they no longer met the initial classification criteria. Likewise, there were 8 parti- cipants that were originally classified as ambivalent that decreased in negativity at Time 2 such that they no longer met the initial cri- teria. Subsequent analyses were performed dropping these parti- cipants, and all major findings were consistent with what has been reported.

4Despite randomization to the relationship conditions we found that our groups were not equivalent on numbers of sup- portive and ambivalent friends in their network. That is, indivi- duals randomly assigned to bring in a supportive friend had significantly more supportive network friends (p < .05), whereas individuals randomly assigned to bring in an ambivalent friend had significantly more ambivalent network friends (p < .05). Because of this, we reran our main analyses controlling for the number of supportive and ambivalent friends, and our physiologi- cal findings were unchanged. Therefore, it is unlikely that these broader social network differences were responsible for our findings.

282 Holt-Lunstad et al. Annals of Behavioral Medicine

Event Ratings and Postdiscussion Ratings

The second factor that we manipulated was the valence of the discussion topic. Ratings of topic intensity indicated no significant group differences (M ¼ 2.9; 3 ¼ moderate), suggesting that the topic selection procedure was effective and differences between conditions could not be simply explained in terms of the psychological intensity of the positive and negative events.

A number of other items intended to be manipulations checks were also included in our Postdiscussion Rating Scale. There were several main effects for relationship classification (see Table 2). When asked about discussing the same event with their friend in the past, we found a Relationship�Condition interaction on ratings of upset, F(1, 97) ¼ 7.12, p < .01; x2 ¼ 0.30. In follow-up compari- sons participants perceived their ambivalent friends as more upsetting in the past than supportive friends when discussing the current negative event (p < .05) but not when discussing the current positive event.

The IMI

The IMI was completed twice, once after each dis- cussion. Analysis of variance (ANOVA) results revealed a main effect for relationship categorization for ratings on

the Friendliness and Dominance dimension, taken after both the neutral discussion and the event discussion (see Table 3). Supportive friends were perceived as greater in friendliness and less dominant than ambivalent friends dur- ing both discussions.

Anxiety

A mixed ANOVA was used to examine levels of anxi- ety across our four epochs (first baseline, neutral dis- cussion, second baseline, event discussion). There was a main effect for relationship classification during the first resting baseline, neutral discussion, and the second base- line; however, the final rating taken at the completion of the event discussion was not significant (see Table 3). Over- all, the Period�Relationship Classification interaction was not significant. Thus, ratings of anxiety were consist- ently higher among those who had brought in an ambiva- lent friend than among those who brought in a supportive friend. We also examined change in anxiety while control- ling for baseline. Results revealed only a main effect for the event discussion, F(1, 102) ¼ 5.02, p < .05, x2 ¼ 0.16, such that discussing negative events was associated with larger increases in ratings of anxiety than discussing posi- tive events.

TABLE 3

Psychological Variables According to Relationship Quality

Supportive M (SD) Ambivalent M (SD) F x2

SRI Positive Time 1 5.40 (0.63) 4.34 (0.92) 46.59���� 0.30 Positive Time 2 5.31 (0.63) 4.66 (0.68) 25.59���� 0.19 Negative Time 1 1.03 (0.17) 2.74 (1.03) 140.1���� 0.57 Negative Time 2 1.18 (0.30) 2.09 (0.70) 77.11���� 0.42 Time known (years) 6.42 (6.28) 6.75 (7.66) 0.06 �0.88 Contact (days per week) 4.21 (2.30) 4.31 (2.14) 0.05 �0.91 Predictability 4.35 (1.38) 4.33 (1.46) 0.00 �0.99 Importance 5.36 (0.99) 4.88 (1.17) 4.81� 0.54 Likelihood to seek out friend

Stressful situations 4.97 (1.02) 4.40 (1.23) 6.3� 0.69 Positive situations 5.12 (0.99) 4.55 (1.18) 7.24�� 0.67 Neutral situations 5.10 (1.07) 4.61 (1.16) 4.6� 0.60

QRI Support 25.64 (2.70) 22.87 (3.27) 21.78���� 0.37 Conflict 16.06 (3.08) 22.36 (6.21) 43.2���� 0.54 Depth 19.75 (2.76) 17.83 (3.89) 8.5�� 0.18

IMI Friendliness (N) 4.59 (1.13) 3.22 (1.37) 31.01���� 0.46 Friendliness (E) 4.52 (1.29) 3.26 (1.72) 17.92���� 0.32 Dominance (N) �0.83 (0.63) �0.35 (0.98) 8.63�� 0.18 Dominance (E) �0.98 (0.69) �0.62 (1.06) 4.37� 0.09

Anxiety Baseline 1 1.35 (0.40) 1.51 (0.46) 4.19� 0.11 Discussion 1 (N) 1.38 (0.40) 1.71 (0.46) 14.71��� 0.33 Baseline 2 1.30 (0. 36) 1.55 (0.46) 8.60�� 0.22 Discussion 2 (E) 1.69 (0.55) 1.88 (0.58) 2.37 0.05

Note. The scale for SRI ranged from 1 (not at all) to 6 (extremely). SRI ¼ social relationships index; QRI ¼ quality of relationships index; IMI- ¼ impact message inventory; N ¼ neutral discussion; E ¼ event (positive or negative) discussion. y p < .10. �p < .05. ��p < .01. ���p < .001. ����p < .0001.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 283

Primary Analyses

The major physiological dependent measures were SBP, DBP, and HR given these measures have been linked to significant health outcomes (37,38). In addition, because blood pressure and HR are multiply determined cardio- vascular endpoints, cardiac impedance derived measures of PEP, RSA, CO, and TPR were also used as dependent measures in an effort to better assess the underlying deter- minants responsible for the changes in CVR across con- ditions.

An average baseline value was calculated for each of our physiological variables to increase the reliability of these assessments (39). During the neutral and event dis- cussions, the participant and friend alternated speaking for 1 min each so that we had an index of when the partici- pant was speaking versus listening. Because there were no differential effects of our manipulations as a function of speaking and listening on any of our physiological vari- ables, an average for the neutral and an average for the event discussion were calculated to increase the reliability of these assessments (39). Change scores were then com- puted as an index of reactivity (40). Because basal level can affect reactivity scores (41), baseline measures of physiological functioning were entered into all of the reac- tivity analyses as a covariate. Finally, effect sizes for signifi- cant findings were calculated using omega-squared (42).

Baseline and Reactivity During the Neutral Task

Because participants were randomly assigned to the conditions of relationship quality and event discussion, we did not expect group differences on any of our cardio- vascular assessments during initial baseline. However, con- sistent with the state anxiety assessment during baseline, we found a relationship categorization main effect for baseline HR, F(1, 98) ¼ 5.28, p ¼ .02, x2 ¼ 0.11, with higher HR among those that brought in ambivalent friends than those who brought in supportive friends (see Table 4). Consistent with prior research, there was also a gender main effect for SBP, F(1, 100) ¼ 38.63, p < .0001, x2 ¼ 0.52, with higher resting SBP among men (M ¼ 118.84 mmHg) than women (M ¼ 109.95 mmHg).5

We also examined the impact of the relationship cate- gorization (i.e., supportive, ambivalent) and gender of par- ticipant on CVR during the neutral discussion while controlling for the first baseline. There were no significant main effects or interactions for SBP, DBP, or HR reactivity.

5Upon inspection, we found a participant who demonstrated extreme baseline data, which was six standard deviations above the mean; therefore, this participant’s data were excluded for all analyses of physiological measures. When this participant was identified as an outlier, we reanalyzed our physiological data. It is important to note that although degree of freedom and means changed slightly, none of our findings changed significance.

T A

B L E

4

M e a n

C a rd

io v a s c u la

r R

e s p o n s e

D u ri n g

R e s t

a n d

D is

c u s s io

n T

a s k s

a s

a F

u n c ti o n

o f

R e la

ti o n s h ip

Q u a lit

y

B a

se li

n e

1 N

eu tr

a l

D is

cu ss

io n

R ea

ct iv

it y

B a

se li

n e

2 P

o si

ti ve

D is

cu ss

io n

R ea

ct iv

it y

N eg

a ti

ve D

is cu

ss io

n R

ea ct

iv it

y

S u

p p

o rt

iv e

A m

b iv

a le

n t

S u

p p

o rt

iv e

A m

b iv

a le

n t

S u

p p

o rt

iv e

A m

b iv

a le

n t

S u

p p

o rt

iv e

A m

b iv

a le

n t

S u

p p

o rt

iv e

A m

b iv

a le

n t

S B

P (m

m = H

g )

1 1

4 .8

9 (1

1 .7

3 )

1 1

4 .4

9 (8

.7 7

) 5

.1 4

(1 0

.2 7

) 7

.8 8

(7 .0

0 )

1 1

7 .1

2 (1

1 .6

5 )

1 1

6 .2

3 (7

.6 6

) 6

.2 0

(4 .6

1 )

3 .3

8 (4

.8 8

) 5

.4 (5

.5 3

) 9

.6 2

(4 .8

4 )

D B

P (m

m = H

g )

6 3

.2 5

(6 .9

8 )

6 4

.9 8

(7 .2

9 )

8 .7

2 (8

.5 7

) 5

.9 1

(5 .3

8 )

6 4

.5 2

(7 .6

1 )

6 5

.3 1

(7 .0

6 )

7 .9

6 (5

.7 3

) 5

.7 7

(5 .5

9 )

6 .9

4 (5

.8 4

) 7

.7 5

(6 .2

2 )

H R

(b p

m )

6 6

.0 2

(1 0

.5 3

) 7

0 .2

5 (1

0 .1

5 )

6 .9

2 (3

.8 7

) 6

.6 8

(5 .1

0 )

6 6

.3 6

(1 0

.2 0

) 7

0 .7

7 (1

0 .2

3 )

5 .8

7 (2

.7 9

) 4

.1 5

(4 .5

6 )

6 .0

7 (4

.2 4

) 7

.0 9

(4 .4

0 )

R S

A (n

a tu

ra l

lo g

) 6

.9 3

(0 .9

2 )

6 .0

5 (1

.2 8

) �

0 .3

3 (0

.6 9

) �

0 .3

0 (0

.9 1

) 6

.7 4

(0 .8

6 )

5 .9

2 (1

.2 8

) �

0 .1

2 (0

.6 9

) �

0 .0

8 (0

.5 9

) �

0 .3

0 (0

.6 4

) �

0 .3

3 (0

.8 2

) P

E P

(m s)

1 0

6 .0

5 (1

2 .5

2 )

1 0

4 .0

5 (1

1 .9

2 ) �

3 .2

5 (5

.4 0

) �

4 .1

1 (6

.3 2

) 1

0 6

.2 6

(1 2

.1 5

) 1

0 4

.8 2

(1 0

.0 0

) �

3 .9

6 (5

.8 9

) �

2 .9

0 (4

.3 7

) �

3 .7

1 (5

.8 6

) �

5 .3

1 (7

.9 1

) T

P R

(D y

n es

-s �c

m �

5 )

7 7

7 .4

3 (2

0 9

.2 7

) 8

2 2

.9 8

(3 4

4 .7

2 )

1 0

6 .5

6 (1

2 3

.0 6

) 8

5 .2

8 (1

2 6

.4 1

) 8

0 3

.1 2

(2 2

3 .3

3 )

8 6

7 .3

6 (3

7 4

.1 2

) 1

2 2

.9 9

(1 6

2 .8

4 )

9 7

.1 2

(1 8

8 .5

6 )

8 7

.9 9

(1 2

6 .8

5 )

1 0

5 .9

0 (1

2 6

.0 3

) C

O (l = m

in )

9 .1

7 (2

.1 7

) 9

.2 6

(2 .8

3 ) �

0 .2

2 (0

.7 7

) �

0 .2

5 (1

.0 6

) 9

.0 6

(2 .2

0 )

8 .9

9 (2

.8 3

) �

.2 6

(1 .0

8 ) �

.3 1

(0 .8

5 ) �

.1 0

(0 .7

9 ) �

.0 2

(1 .2

4 )

N o

te .

S ta

n d

a rd

d ev

ia ti

o n

s a

re in

p a

re n

th es

es .

S B

P ¼

sy st

o li

c b

lo o

d p

re ss

u re

; D

B P ¼

d ia

st o

li c

b lo

o d

p re

ss u

re ;

H R ¼

h ea

rt ra

te ;

b p

m ¼

b ea

ts p

er m

in u

te ;

R S

A ¼

re sp

ir a

to ry

si n

u s

a rr

h y

th m

ia ;

P E

P ¼

p re

-e je

ct io

n p

er io

d ;

T P

R ¼

to ta

l p

er ip

h er

a l

re si

st a

n ce

; C

O ¼

ca rd

ia c

o u

tp u

t.

284 Holt-Lunstad et al. Annals of Behavioral Medicine

Baseline and Reactivity to Discussing Positive or

Negative Events with Supportive or Ambivalent Friends

Preliminary analyses of the second baseline replicated the gender main effects for SBP, F(1, 102) ¼ 33.86, p < .0001, x2 ¼ 0.48, and replicated the relationship main effect for HR, F(1, 98) ¼ 6.98, p < .01, x2 ¼ 0.15. Due to the lack of relationship categorization differences in reac- tivity during the neutral task, simple change scores (event discussion � second baseline) were used while controlling for gender and second baseline for the remainder of our analyses.

Even though preliminary analyses showed that parti- cipants perceived the positive and negative events to be equally intense, a significant main effect was found for SBP reactivity, F(1, 100) ¼ 7.74, p < .01, x2 ¼ 0.24. The negative event discussion evoked greater SBP reactivity than the positive event discussion. These data are consist- ent with research suggesting a negativity bias on physio- logical function (43).

We next evaluated our primary hypotheses about the context in which ambivalent relationships may be detri- mental. Results revealed a statistical interaction between relationship quality and the event topic on reactivity for SBP, F(1, 100) ¼ 13.05, p ¼ .0005, x2 ¼ 0.36; see Figure 1. Consistent with the support interference hypothesis, con- trasts reveal participants exhibited greater SBP reactivity when discussing a negative event with an ambivalent friend (M ¼ 9.6 mmHg D) than with a supportive friend (M ¼ 5.4 mmHg D), F(1, 48) ¼ 8.72, p < .01, x2 ¼ 0.70. Further, among participants interacting with an ambiva- lent friend, those discussing a negative event displayed greater SBP reactivity (M ¼ 9.6 mHg D) than did those dis- cussing a positive event (M ¼ 3.4 mHg D), F(1, 49) ¼ 20.99, p < .0001, x2 ¼ .90. There was no significant difference between discussing positive and negative events with a supportive friend. These results, combined with our results from the neutral interaction, suggest that ambivalent relationships are not just general sources of ten- sion during interactions but may be distressing primarily in

the context of seeking support during stress. However, when discussing positive events there was a significant dif- ference between discussions with supportive and ambiva- lent friends, F(1, 50) ¼ 5.24, p < .05, x2 ¼ 0.19. Unexpectedly, discussing a positive event with an ambiva- lent friend was associated with lower reactivity.

Follow-Up Analyses

Clarification of SBP Finding During

Positive Discussions

Given the unexpected finding that individuals discussing a positive event with an ambivalent friend showed the lowest levels of reactivity we performed several focused follow-up analyses. Two possible explanations for this finding are that participants were able to get more excited telling their sup- portive friend about something fantastic that occurred recently or that participants interacting with ambivalent friends disengaged from the positive event discussion task. To test these explanations, a repeated measures ANOVA of the neutral reactivity and the event reactivity was explored directly contrasting neutral and positive discussions with ambivalent friends and supportive friends. Results revealed a significant Relationship�Period interaction, F(1, 49) ¼ 4.39, p < .05, x2 ¼ 0.16. Upon further inspection, there was no significant difference in reactivity during the neutral task and the positive discussion for those talking to their supportive friend. In contrast, our data show that there was a significant decrease in reactivity between the neutral task (M ¼ 8.16 mmHg) and the positive event condition (M ¼ 3.37 mmHg) for those with ambivalent friends, F(1, 25) ¼ 7.78, p ¼ .01, x2 ¼ 0.33. Overall, these data do not support the possibility of exaggerated reactivity among those discussing positive events with a supportive friend; rather, the data suggest a possible disengagement from the task for those interacting with ambivalent friend.

Analyses of Underlying Determinants of Heart Rate and

Blood Pressure

We followed up our significant results by analyzing the impedance-derived measures associated with the underly- ing determinants of HR (i.e., PEP and RSA) and BP (i.e., TPR, CO). There was also a significant difference dur- ing the first baseline period for relationship categorization on RSA, F(1, 92) ¼ 17.91, p < .0001, x2 ¼ 0.35, and second baseline period, F(1, 92) ¼ 15.92, p < .0001, x2 ¼ 0.32, with greater RSA among those who brought in a supportive friend than those who brought in an ambivalent friend. To examine the potential mediating effect of RSA, we reanalyzed the baseline HR data using RSA as a covariate and the effect became nonsignificant. This was consistent across the first, F(1, 91) ¼ 0.04, p ¼ .84, and the second, F(1, 91) ¼ 0.03, p ¼ .86, baseline. These data suggest that the baseline differences in HR are due to parasympathetic influences on the heart. There were

FIGURE 1 Systolic blood pressure reactivity according to relationship classification and discussion.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 285

no other significant main effects or interactions for RSA, PEP, TPR, or CO (see Table 4 for details). As a result, no clear evidence of the underlying determinants for the Relationship�Event interaction for SBP was found.6

Analysis for Potential Mediation

Because we found a significant interaction on the Post- discussion Rating of past upset between relationship quality and event discussion valence that was in the same direction as our interaction effect for SBP reactivity, we did explora- tory analyses to examine potential mediation. Thus, we were interested in the PDR item that asked ‘‘When discuss- ing this event with this person in the past, how upsetting has your friend been?’’ Results revealed that past upset was sig- nificantly correlated to SBP reactivity (p < .05). Because the primary stress-enhancing effect was found in the nega- tive event discussion, we repeated this analysis looking at ambivalent-supportive friend differences only in the nega- tive event discussion. The previously significant simple main effect was rendered nonsignificant when controlling for perceptions of how upsetting they had been in the past. Thus, past upsetting feelings toward the friend when they had discussed the current negative event mediated the relationship quality differences on SBP reactivity.

DISCUSSION

The purpose of this study was to address two questions regarding the specificity of the association between social relationships and health. First, are ambivalent relationships associated with health relevant processes? Second, under what conditions are ambivalent ties detrimental? Our data suggest that interactions with ambivalent ties were generally viewed more negatively and that individuals interacting with an ambivalent friends showed greater SBP reactivity primarily when disclosing a negative personal event. More- over, this effect was mediated by how upsetting the friend had been in the past when discussion the negative event. Thus, to the extent that SBP reflects a detrimental health- relevant process, ambivalent relationships are a potentially negative influence on health, especially during discussion of prior stressful personal events.

One main goal of this study was to evaluate two broad mechanisms potentially linking ambivalent ties to detri- mental outcomes. First, it is possible that individuals are unable to receive or perceive effective social support from ambivalent relationships during times of need. Second, social interactions with ambivalent network ties may gener- ally be associated with significant levels of interpersonal stress (44). Our main finding for the physiological data, highest SBP reactivity associated with discussing negative

events with the ambivalent friend, is consistent with the support interference hypothesis. Given the joint positivity and negativity associated with ambivalent relationships it may be that when participants seek out support from an ambivalent friend that the support received is not trusted or they may feel blamed instead of understood. It is also possible that ambivalent friends may be perceived as more critical given that they were rated as lower in warmth and higher in control than relationships characterized as supportive. Together, our data suggest that support from such ambivalent ties in times of stress may be less effective, with such effects being mediated by how upsetting the friend had been in past discussions of the negative event. These data may also suggest a more context-specific interpretation of the stress-enhancing hypothesis such that ambivalent friends may be associated with increased inter- personal stress only when discussing negative events.

Although our reactivity data appear consistent with the support interference hypothesis, other data indicate that we cannot discount the stress-enhancing hypothesis as our data also demonstrate that ambivalent relationships may have more subtle effects on physiological processes. For example, self-reports of anxiety were consistently higher among those who brought in an ambivalent friend than a supportive friend across every portion of the experi- ment. We also found consistently higher levels of HR and lower RSA among those who brought in ambivalent friends than those who brought in supportive friends. Because baseline measures were assessed with only a cur- tain separating them from their friend, it suggests that the mere presence or the anticipation of interacting with an ambivalent friend may be taxing for these individuals. Therefore, those paired with ambivalent friends may have been unable to fully relax and hence regulate their cardio- vascular system to the extent participants with supportive friends were able to do. It is important to mention that although RSA is a marker of vagal (parasympathetic) con- trol of the heart (45), respiratory rate and amplitude can influence RSA independent of vagal control (46). Some studies suggest that ‘‘corrected’’ and ‘‘uncorrected’’ assess- ments produce similar conclusions under stress paradigms (47–49). However, one limitation of our findings is that we do not have respiratory data to directly examine this issue. Therefore, it is possible that participants who brought in an ambivalent friend may have had lower RSA than those who brought in a supportive friend simply because they were breathing at a faster rate during the baseline period. Future research will be needed to examine the plausibility of this alternative explanation.

One relatively unexpected but interesting finding indi- cated the lowest levels of reactivity occurred among those interacting with ambivalent friends in positive contexts. Our ancillary analyses suggest the possibility that indivi- duals may have disengaged from the discussion as evidenced by their lower reactivity compared to the neutral discussion condition. Relevant research by Gable and

6 The lack of findings for impedance-derived measures poten-

tially mediating blood pressure changes may be due to the greater measurement error inherent in deriving CO and TPR.

286 Holt-Lunstad et al. Annals of Behavioral Medicine

colleagues (50) suggests that this disengagement may have important intra-and interpersonal consequences. Social capitalization, telling others about positive events in one’s life, is thought to be associated with greater positive affect over and above the positivity associated with the event. There are multiple potential mechanisms that influence self and relationship processes including: the process of reliving the event can boost positive affect, sharing the event strengthens the relationship, and=or it can serve to boost self-esteem by being seen favorably in the eyes of others. Thus, if individuals disengage when discussing positive events with ambivalent friends, they may not experience the upward spiral of positive affect and well-being (51). However, such explanations are speculative and future stu- dies will be necessary to examine these possibilities.

Overall, our data suggest that ambivalent relationships do not afford the same emotional and physical benefits that supportive relationships do and may be potentially detrimen- tal. One possible reason for these deleterious influences is that general perceptions of the relationship may be driving participants’ interpretation of the situation. A person’s his- tory of experiences (positive and=or negative) with a friend provides a lens or schemata by which current and future events are interpreted. For example, an ambivalent friend who is behaving supportively in our study may be assumed to be ‘‘trying to look good for the experimenters.’’ Thus, it may not be anything in particular the friend is doing in the moment but rather the individual’s perception of that friend that may be driving the response. Our mediational analyses are consistent with this interpretation as how upsetting the friend had been in the past discussions of the negative event appeared responsible for the differences in SBP reactivity. Taken together, these findings suggest that one’s history and overall perception of the friend may be more important than the behavior during the interaction.

There are several limitations of this study that warrant discussion. The first concerns the generalizability of the sample. This study only included college-age participants who were healthy. In addition, our sample primarily com- prised White participants. The extent to which these find- ings can be expanded to other ages, socioeconomic status, health status, and ethnicities cannot be determined. In addition, all participants were required to bring in a same-sex friend. It is possible that participants may have reacted differently if they were interacting with an opposite sex individual and=or someone of a different relationship type (i.e., romantic partner, sibling, coworker). It is also possible that the effect of relationship type on CVR may be underestimated in this study due to some degree of imprecision in the classifications (see footnote 3).

Another limitation of this study is the ecological val- idity to social support or disclosure processes. It is possible that the structured (alternating minutes) conversation reduced the naturalness and flow of the discussions between friends. Although it is difficult to determine pre- cisely the extent to which the naturalness was reduced,

we did behavioral code the videotaped interactions,7 and these data suggest that any effect this may have had was equivalent when discussing events with supportive and ambivalent friends. Alternatively it is possible that people may not actually choose to seek out their ambivalent friend when they need support. Although we found that the like- lihood to seek out supportive friends was relatively higher than ambivalent friends, ratings of likelihood were still (between moderate and very) high.

It is also unclear whether having ambivalent friends would be better or worse than having no friends at all (social isolation, e.g., 52). Although both may be bad, con- flict and support are separate dimensions (53). There are some data to suggest that negativity in social relationships has a stronger effect than support (26,54,55). Alternatively, because we find that most of our participants have both supportive and ambivalent individuals in their network, it is also possible that their supportive relationships can buf- fer the negative effects of having ambivalent relationships (see 56). It is also possible that individuals may benefit from noninterpersonal disclosure such as diary=journal entries (57,58) and prayer (59). Future research will be needed to directly contrast these possibilities.

We should also note that the actual health relevance of this study is less clear, because it was a cross-sectional study using young participants. Many of the health outcomes predicted by social relationships, including cardiovascular disease, develop slowly over time. Despite our young sample, the average length of time they had known their friend was 6.6 years—approxi- mately one third of their lifetime. Our data attest to the stability of relationships even among young college students who are typically in a transitional stage socially. Due to the stability of many other types of social relationships (e.g., familial ties, spouse; see 60), the health effects of these ambivalent social relationships may be more important. It is also possible that any potential negative impact of ambivalent relationships may be limited if individuals can avoid contact with them. Our data, however, revealed no difference in the amount of contact participants have with ambivalent and supportive friends. Future studies examining ambivalent feelings among married couples may be especially promising given the strong links of marriage to health (61,62). Likewise, studies examining ambivalent relationships among older adults or clinical popu- lations may also provide insight on the relative impact of social relationships in more health relevant populations.

Despite the limitations, this study has a number of strengths. First, the increased ecological validity is associa- ted with the use of an existing friendship. Many prior lab- oratory studies examining social support have been limited

7 We had two raters, blind to condition, rate the videotaped

interactions. Reliability between the coders for each (between .41 and .69) was significant (p < .05). Based on these ratings, it appears that the structured nature of the conversation did not dif- ferentially impact the flow of conversation. Interested readers can get a copy of these analyses by contacting the primary author.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 287

(63) by the use of strangers, confederates, or the exper- imenter to operationalize support, in that they fail to cap- ture nuances associated with existing relationships (i.e., one’s history with that person). A further strength of this study is that we randomly assigned participants to bring in a particular type of relationship (supportive or 4ambivalent) from their larger network. By randomly assigning the type of friend (supportive=ambivalent) for individuals to bring in as part of the experiment, we are able to prevent individuals from self-selecting their friend and potentially biasing our results. Last, these data were strengthened by independent confirmation of relationship type and the inclusion of standardized measures of basic interpersonal dimensions (i.e., IMI).

Overall, these results may have important implications for the conceptualization andassessment of social relation- ships in the health domain. Simply assessing levels of social support (e.g., positivity) may obscure the natural hetero- geneity that exists in people’s relationships by ignoring co- occurring levels of negativity. This point may be especially important to consider in developing social support interven- tions (10). Our results suggest that implementing a support intervention (without regard to extent of positivity and negativity within that relationship) could potentially lead to detrimental outcomes. More generally, it is our hope that such a multidimensional approach to the study of social relationships may be useful in understanding the underlying associations between social relationships and health.

REFERENCES

(1) Berkman LF: The role of social relations in health pro- motion. Psychosomatic Medicine. 1995, 57:245–254.

(2) Cohen S: Psychosocial models of the role of social support in the etiology of physical disease. Health Psychology. 1988, 7:269–297.

(3) Berkman LF, Glass T, Brissette I, Seeman TE: From social integration to health: Durkheim in the new millennium. Social Science and Medicine. 2000, 51:843–857.

(4) House JS, Landis KR, Umberson D: Social relationships and health. Science. 1988, 241:540–545.

(5) Uchino BN: Social Support and Physical Health: Under- standing the Health Consequences of Our Relationships. New Haven, CT: Yale University Press, 2004.

(6) Braiker HB, Kelly HH: Conflict in the development of close relationships. In Burgess RL, Huston TL (eds), Social Exchange in Developing Relationships. New York: Aca- demic, 1979, 135–168.

(7) Major B, Zubek JM, Cooper ML, Richards C: Mixed mes- sages: Implications of social conflict and social support within close relationships for adjustment to a stressful life event. Journal of Personality and Social Psychology. 1997, 72:1349–1363.

(8) Friedman HS, Tucker JS, Schwartz JE, et al.: Psychosocial and behavioral predictors of longevity: The aging and death of the ‘‘Termites.’’ American Psychologist. 1995, 50:69–78.

(9) Tucker JS, Friedman HS, Wingard DL, Schwartz JE: Mari- tal history at midlife as a predictor of longevity: Alternative

explanations to the protective effects of marriage. Health Psychology. 1996, 15:94 –101.

(10) Finch JF, Okun MA, Barrera M, Zautra AJ, Reich JW: Positive and negative social ties among older adults: Measurement models and the prediction of psychological distress and well-being. American Journal of Community Psychology. 1989, 17:585–605.

(11) Fiore J, Becker J, Coppel DB: Social network interactions: A buffer or a stress? American Journal of Community Psychology. 1983, 11:423–439.

(12) Kiecolt-Glaser JK, Dyer CS, Shuttleworth EC: Upsetting social interactions and distress among Alzheimer’s Disease family caregivers: A replication and extension. American Journal of Community Psychology. 1988, 116:825–837.

(13) Ruehlman LS, Karoly P: With a little Flak from my friends: Development and preliminary validation of the Test of Negative Social Exchange (TENSE). Psychological Assess- ment. 1991, 3:97–104.

(14) Uchino BN, Holt-Lunstad J, Uno D, Flinders JB: Heterogen- eity in the social networks of young and older adults: Predic- tion of mental health and cardiovascular reactivity during acute stress. Journal of Behavioral Medicine. 2001, 24:361–382.

(15) Coyne JC, DeLongis A: Going beyond social support: The role of social relationships in adaptation. Journal of Consulting and Clinical Psychology. 1986, 54:454–460.

(16) Uchino BN, Holt-Lunstad J, Smith TW, Bloor L: Hetero- geneity in social networks: A comparison of different mod- els linking relationships to psychological outcomes. Journal of Social and Clinical Psychology. 2004, 23:123–139.

(17) Holt-Lunstad J, Uchino BN, Smith TW: Relationship qual- ity predicts ambulatory blood pressure during social inter- actions. Health Psychology. 2003, 22:388–397.

(18) Uno DF, Uchino BN, Smith TW: Relationship quality moderates the effect of social support given by close friends on cardiovascular reactivity in women. International Jour- nal of Behavioral Medicine. 2002, 9:243–262.

(19) Smith TW, Ruiz JM, Uchino BN: Vigilance, active coping, and cardiovascular reactivity during social interaction in young men. Health Psychology. 2000, 19:382–392.

(20) Lepore SJ: A social-cognitive processing model of emotion- al adjustment to cancer. In Baum A, Andersen BL (eds), Psychosocial Interventions for Cancer. Washington, DC: American Psychological Association, 2001, 99–116.

(21) Smith TW, Ruiz JM: Psychosocial influences on the devel- opment and course of coronary heart disease: Current sta- tus and implications for research and practice. Journal of Consulting and Clinical Psychology. 2002, 70:548–568.

(22) Treiber FA, Kamarck T, Schneiderman N: Cardiovascular reactivity and development of preclinical and clinical dis- ease states. Psychosomatic Medicine. 2003, 65:46–62.

(23) Cacioppo JT, Malarkey W, Kiecolt-Glaser JK, et al.: Het- erogeneity in neuroendocrine and immune responses to brief psychological stressors as a function of autonomic cardiac activation. Psychosomatic Medicine. 1995, 57: 154–164.

(24) Sherwood A, Allen M, Fahrenberg J, et al.: Methodological guidelines for impedance cardiography. Psychophysiology. 1990, 27:1–23.

(25) Friedman E: The effects of normal and rapid speech on blood pressure. Psychosomatic Medicine. 1982, 44: 545–553.

288 Holt-Lunstad et al. Annals of Behavioral Medicine

(26) Pagal MD, Erdly WW, Becker J: Social networks: We get by with (and in spite of) a little help from our friends. Journal of Personality and Social Psychology. 1987, 53: 793–804.

(27) Uchino BN, Kiecolt-Glaser JK, Cacioppo JT: Age-related changes in cardiovascular response as a function of a chronic stressor and social support. Journal of Personality and Social Psychology. 1992, 63:839–846.

(28) Pierce GR, Sarason IG, Sarason BR: General and relation- ships-specific perceptions of social support: Are two con- structs better than one? Journal of Personality and Social Psychology. 1991, 61:1028–1039.

(29) Kiesler DJ, Anchin JC, Perkins MJ, et al.: The Impact Message Inventory: Form II. Palo Alto, CA: Consulting Psychologists Press, 1985.

(30) Kiesler DJ, Schmidt JA, Wagner CC: A circumplex inventory of impact messages: An operational bridge between emotion and interpersonal behavior. In Plutchik R, Conte HR (eds), Circum- plex Models of Personality and Emotions. Washington, DC: American Psychological Association, 1997, 221–244.

(31) Nealy JB, Smith TW, Uchino BN: Cardiovascular responses to agency and communion stressors in young women. Jour- nal of Research in Personality. 2002, 36:395–418.

(32) Marteau TM, Bekker H: The development of a six-item short form of the state scale of the Spielberger state-trait anxiety inventory (STAI). British Journal of Clinical Psychology. 1992, 31:301–306.

(33) Sherwood A, Allen M, Fahrenberg J, et al.: Methodological guidelines for impedance cardiography. Psychophysiology. 1990, 27:1–23.

(34) Uchino BN, Uno D, Holt-Lunstad J, Flinders JB: Age- related differences in cardiovascular reactivity during acute psychological stress in men and women. Journal of Geron- tology. 1999, 54:339–346.

(35) Berntson G, Quigley K, Jang J, Boysen S: An approach to artifact identification: Application to heart period data. Psychophysiology. 1990, 27:586–598.

(36) Neuvo Y, Cheng-Yu D, Mitra S: Interpolated finite impulse response filters. IEEE Transactions on Acoustics, Speech, and Signal Processing, ASSP—32. 1984, 563–570.

(37) Manuck SB, Kaplan JR, Clarkson TB: Behaviorally induced heart rate reactivity and atherosclerosis in Cynomolgus monkeys. Psychosomatic Medicine. 1983, 45:95–108.

(38) Light KC, Dolan CA, Davis MR, Sherwood A: Cardio- vascular responses to an active coping challenge as predic- tors of blood pressure patterns 10 to 15 years later. Psychosomatic Medicine. 1992, 54:217–230.

(39) Kamarck TW, Jennings JR, Debski TT, et al.: Reli- able measures of behaviorally-evoked cardiovascular reactivity from a PC-based test battery: Results from stu- dent and community samples. Psychophysiology. 1992, 29:17–28.

(40) Llabre MM, Spitzer SB, Saab PG, et al.: The reliability and specificity of delta versus residualized change as measures of cardiovascular reactivity to behavioral challenges. Psycho- physiology. 1992, 28:701–711.

(41) Benjamin LS: Facts and artifacts in using analysis of covari- ance to ‘‘undo’’ the law of initial values. Psychophysiology. 1967, 4:187–206.

(42) Hays W: Statistics. New York: Holt, Rinehart, and Winston, 1988.

(43) Taylor S: Asymmetrical effects of positive and negative events: The mobilization minimization hypothesis. Psychology Bulletin. 1991, 110:67–85.

(44) Coyne JC, DeLongis A: Going beyond social support: The role of social relationships in adaptation. Journal of Consulting and Clinical Psychology. 1986, 54:454–460.

(45) Berntson GG, Bigger JT, Eckberg DL: Heart rate varia- bility: Origins, methods, and interpretive caveats. Psycho- physiology. 1997, 34:623–648.

(46) Grossman P, Kollai M: Respiratory sinus arrhythmia, cardiac vagal tone, and respiration: Within- and between-individual relations. Psychophysiology. 1993, 30:486–495.

(47) Berntson GG, Cacioppo JT, Binkley PF, et al.: Autonomic cardiac control. III. Psychological stress and cardiac response in autonomic space as revealed by pharmacologi- cal blockades. Psychophysiology. 1994, 31:599–608.

(48) Cacioppo JT, Berntson GG, Binkley PF, et al.: Autonomic cardiac control II: Basal response, noninvasive indices, and autonomic space as revealed by autonomic blockade. Psychophysiology. 1994, 31:586–598.

(49) Grossman P, Van Beek J, Wientjes C: A comparison of three quantification methods for estimation of respiratory sinus arrhythmia. Psychophysiology. 1990, 27:702–714.

(50) Gable SL, Reis HT, Impett EA, Asher ER: What do you do when things go right? The intrapersonal and interpersonal benefits of sharing positive events. Journal of Personality and Social Psychology. 2004, 87:228–245.

(51) Fredrickson BL: The role of positive emotions in positive psychology: The broaden-and-build theory of positive emo- tions. American Psychologist. 2001, 56:218–226.

(52) Cacioppo JT, Hawkley LC, Crawford E: Loneliness and health: Potential mechanisms. Psychosomatic Medicine. 2002, 64:407–417.

(53) Cacioppo JT, Berntson GG: Relationship between attitudes and evaluative space: A critical review, with emphasis on the separability of positive and negative substrates. Psycho- logical Bulletin. 1994, 115:401–423.

(54) Rook KS: The negative side of social interaction: Impact on psychological well being. Journal of Personality and Social Psychology. 1984, 46:1097–1108.

(55) Schuster TL, Kessler RC, Aseltine RH: Supportive inter- actions, negative interactions, and depressed mood. American Journal of Community Psychology. 1990, 18: 423–438.

(56) Lepore SJ: Social conflict, social support, and psychologi- cal distress: Evidence of cross-domain buffering effects. Journal of Personality and Social Psychology. 1992, 63:857–867.

(57) Smyth JM: Written emotional expression: Effect sizes, out- come types, and moderating variables. Journal of Consulting and Clinical Psychology. 1998, 66:174–184.

(58) Frisina PG, Borod JC, Lepore SJ: A meta-analysis of the effects of written emotional disclosure on health outcomes of clinical populations. Journal of Nervous and Mental Disease. 2004, 192:629–634.

(59) Pargament KI: Religious methods of coping: Resources for the conservation and transformation of significance. In Shafranske EP (ed), Religion and the Clinical Practice of Psychology. Washington, DC: American Psychological Association, 1996, 215–239.

Volume 33, Number 3, 2007 Relationship Quality and Cardiovascular Function 289

(60) Fingerman KL, Hay EL, Birditt KS: The best of ties, the worst of ties: Close problematic, and ambivalent social relationships. Journal of Marriage and Family. 2004, 66: 792–808.

(61) Farr W: Influence of marriage on the mortality of the French people. In Humphreys N (ed), Vital Statistics: A Memorial Volume of Selections from Reports and Writings

of William Farr. Methuen, NY: The Scarecrow Press (Orig- inal work published 1858), 1994, 438–441.

(62) Kiecolt-Glaser JK, Newton TL: Marriage and health: His and hers. Psychology Bulletin. 2001, 127:472–503.

(63) Christenfield N, Gerin W, Linden W, et al.: Social support effects on cardiovascular reactivity: Is a stranger as effective as a friend? Psychosomatic Medicine. 1997, 59:388–398.

290 Holt-Lunstad et al. Annals of Behavioral Medicine