RESEARCH ARTICLE
Comparing Visible and Invisible Social Support: Nonevaluative Support Buffers Cardiovascular Responses to Stress Julie A. Kirsch & Barbara J. Lehman*† Department of Psychology, Western Washington University, Bellingham, WA, USA
Abstract Previous research suggests that in contrast to invisible social support, visible social support produces exaggerated negative emotional responses. Drawing on work by Bolger and colleagues, this study disentangled social support visibility from negative social evaluation in an examination of the effects of social support on negative emotions and cardiovascular responses. As part of an anticipatory speech task, 73 female participants were randomly assigned to receive no social support, invisible social support, non-confounded visible social support or visible social support as delivered in a 2007 study by Bolger and Amarel. Twelve readings, each for systolic blood pressure, diastolic blood pressure and heart rate were taken at 5-min intervals throughout the periods of baseline, reactivity and recovery. Cardiovascular outcomes were tested by incorporating a series of theoretically driven planned contrasts into tests of stress reactivity conducted through piecewise growth curve modelling. Linear and quadratic trends established cardiovascular reactivity to the task. Further, in comparison to the control and replication conditions, the non-confounded visible and invisible social support conditions attenuated cardiovascular reactivity over time. Pre- and post-speech negative emotional responses were not affected by the social support manipulations. These results suggest that appropriately delivered visible social support may be as beneficial as invisible social support. Copyright © 2014 John Wiley & Sons, Ltd. Received 9 April 2013; Revised 15 November 2013; Accepted 20 November 2013 Keywords invisible social support; anticipatory stress; social evaluation; negative emotions; cardiovascular activity *Correspondence Barbara J. Lehman, Department of Psychology, Western Washington University, 516 High Street, Bellingham, WA 98225, USA. † Email: [email protected] Published online 22 January 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smi.2558
Although social support is clearly beneficial to physical health (House, Landis, & Umberson, 1988), its effectiveness varies according to situational fit and by support provider characteristics (Uchino, 2009). Previous research has conceptualized social support as either one’s potential access to social support (perceived available social support) or the actual exchange of resources such as advice or emotional comfort (enacted social support; Uchino, 2009). Evidence linking perceived social support to better physical health has consistently shown a positive and causal relationship, whereas evidence linking enacted social support to positive health outcomes has sometimes demonstrated weak or negative relationships (e.g. Cohen & Wills, 1985; Kaplan et al., 1994). Investigations of the mechanisms by which social support exerts positive effects on health may help elucidate inconsistencies in the effectiveness of social support. Although social support may directly promote physical and emotional outcomes, the stress-buffering model
Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
suggests that social support is particularly important for alleviating negative emotions and reducing physiological responses at times of stress (Cohen & Wills, 1985). Because excessive and prolonged cardiovascular activity has been implicated in the etiology of hypertension and coronary heart disease (Brosschot, Gerin, & Thayer, 2006; Trieber et al., 2003), it is important to determine how social support can alleviate such responses. Research has demonstrated that enacted social support does not always effectively buffer stress responses, especially for women (e.g. Allen, Blascovich, Tomaka, & Kelsey, 1991; Taylor et al., 2010). Allen et al. (1991) found that socially supported women experienced more heart rate (HR) and blood pressure elevations compared with unsupported men and women. Other researchers have found no differences between socially supported or unsupported individuals (Anthony & O’Brien, 1999). These inconsistencies suggest that the effect of enacted social support on stress responses merits closer scrutiny. 351
Effective Social Support
The current study tested a recent conceptualization of social support that may explain inconsistencies in the effects of enacted social support. Enacted social support is often manipulated visibly, meaning that recipients are aware of receiving such support. Both correlational research (Bolger, Zuckerman, & Kessler, 2000) and experimental (Bolger & Amarel, 2007) research demonstrate that unsolicited visible support exacerbates negative emotional responses to stress. Visibly enacted support is often confounded with negative social evaluation (Thorsteinsson & James, 1999) because recognizing one’s own need for support may elicit evaluative concern. It is therefore unclear whether visibly enacted support remains ineffective when the potential for negative social evaluation from the support provider is removed. The present study tested the extent to which social support visibility affects cardiovascular and emotional responses to stress in contexts both high and low in negative social evaluation. Social support visibility as a moderator of responses to stress According to Bolger and Amarel’s (2007) social support model, support can be provided either before an individual solicits support or after an individual solicits it. Bolger et al. (2000) proposed that individuals who are aware of receiving unsolicited social support during experiences of stress tend to have lower self-esteem and reduced self-efficacy. Additionally, recipients may feel incapable of coping with the stressor and negatively evaluated by the support provider. Because negative social evaluation and lack of control lead to stress appraisals (Dickerson & Kemeny, 2004), unsolicited visible support may exacerbate stress responses. In a meta-analysis of laboratory social support research, Thorsteinsson and James (1999) attributed null and adverse effects of social support to negative social evaluation by the support provider. Support conditions in which the support provider could observe and therefore potentially evaluate the recipient’s performance were least effective in attenuating responses to stress. Therefore, social support may be more effective when there is no potential for provider negative social evaluation (Kors, Linden, & Gerin, 1997). In summary, visible social support may be less effective when the support provider has or could evaluate the recipient’s ability to cope with the stressor. Reducing the recipient’s awareness of support may minimize the costs associated with visible support. The benefits of such invisible support were first tested in a diary study of intimate partners under stress. Over the course of 1 month, stressed individuals who reported low frequency of received support but whose partner ranked his or her own actions as highly supportive rated themselves low on anxiety and depression (Bolger et al., 2000) as measured at the end of each day. In everyday life, invisible social support might take the form of providing indirect advice or guidance or 352
J. A. Kirsch and B. J. Lehman
completing household chores that would be appreciated by the recipient but not recognized as socially supportive acts. Since social support was not manipulated, variables such as relationship quality and support content may have influenced negative emotional responses to stress. In a laboratory study that served as a model for the current research, Bolger and Amarel (2007) manipulated social support visibility by using an anticipatory speech paradigm. To avoid cross-gender interactions, the authors used all female experimenters, confederates and participants. A confederate partner provided invisible, visible or no social support after she listened to a participant’s practice speech. Recipients of invisible support showed the smallest increase in negative emotions, while recipients of visible support reported the highest increase. Thus, participants fared much worse when they received visible social support compared with when they received no support at all. These findings indicate that visible social support is costly to recipients experiencing a stressful event. Visibly enacted support may be detrimental because it directs attention towards the stressor and increases feelings of uncontrollability or negative social evaluation. Invisible support may instead enhance feelings of support availability without directing attention to the stressor or invoking negative social evaluation (Bolger & Amarel, 2007). For example, in the context of their laboratory task, visible social support took the form of a confederate directly offering advice on how the participant should organize a required speech (e.g. ‘You should… end with a strong conclusion’), while in the invisible support condition, the confederate offered the advice in the form of a question posed to the experimenter (e.g. ‘Isn’t it best to… end with a strong conclusion?’). In a later study, Howland and Simpson (2010) demonstrated that for intimate couples, invisible support was more likely to deflect attention from the recipient’s problems and enhance perceptions of coping resources, thereby decreasing negative emotional responses. A reduction in negative emotional responses may help to attenuate cardiovascular stress responses and influence the onset and course of cardiovascular diseases and hypertension (Feldman, Cohen, Hamrick, & Lepore, 2004). Therefore, social support visibility may be an important mechanism by which social support exerts its effects on stress and cardiovascular health. The present study The current study was designed to extend and clarify social support research and theory. Health psychologists have broadly discussed the applicability of invisible social support for physical health outcomes (e.g. Taylor et al., 2010), yet physiological responses have not been reported in the literature. This study forms an empirical test of the importance of invisible social support for both emotional and physiological stress reactivity. Further, although previous research suggests that visible support exerts an emotional cost, the reasons for Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
this cost remain unclear. Because Bolger and Amarel (2007) used the same confederate to provide support and to listen to the participant’s practice speech, the study may have confounded social support visibility with negative social evaluation. Indeed, visibly supported participants reported that the supportive confederate felt the participant had difficulties with her speech task. In a follow-up study, Bolger and Amarel removed the communication of negative evaluation by having the confederate first assure the participants that she didn’t think the participant needed help. Despite this assurance, participants in the visible support condition fared no better than participants who did not receive any support. Furthermore, this type of instrumental social support may be especially likely to be misunderstood (Trobst, 2000) and may provoke negative social evaluation. Because negative social evaluation has a potent influence on physiological and emotional responses to stress, it seems likely that feelings of negative social evaluation strengthen the association between visible social support and negative stress responses. The current study implemented a visible social support condition that minimized the potential for negative social evaluation by the support provider. This non-confounded visible social support condition was compared with invisible social support, with visible social support as manipulated by Bolger and Amarel (2007) and with a control group. Invisible and nonconfounded visible support conditions were manipulated before participants practised their speeches. The social support provider in these two conditions did not observe the participants’ performance and thereby could not evaluate it. Although a potentially evaluative confederate partner remained present in all conditions, only in the replication condition did this confederate provide social support after the practice speech, directly importing Bolger and Amarel’s (2007) visible social support condition. In this condition, the participant gave a practice speech in front of the confederate, who then provided social support in the form of advice to the participant. Because the participant had just given the practice speech, this support could likely be viewed as negative feedback about her performance, thereby raising concerns for social evaluation and potentially minimizing the benefits of the support. The primary objectives of the current study were to determine whether social support affected emotional and cardiovascular reactivity to an anticipatory speech task and to test whether the effects of support visibility were separate from negative social evaluation. We hypothesized that the forms of social support not confounded with possible negative social evaluation from the support provider would be the most effective at alleviating stress. Therefore, participants in the invisible and non-confounded visible conditions were predicted to show attenuated cardiovascular and negative emotional responses to stress. Follow-up tests then determined whether invisible social support was more beneficial than Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
Effective Social Support
visible social support. A difference between the invisible and non-confounded visible support conditions would suggest that invisible support is uniquely beneficial, while a non-significant comparison would indicate that negative social evaluation may have inflated estimates of invisible support in previous laboratory investigations.
Method Participants and design Participants were 73 female undergraduate students. The sample was 68% Caucasian, 10% Asian American, 5% African American and 15% identified as multi-ethnic/ other; 2% did not identify their ethnicity. Participants were randomly assigned to one of the four conditions: invisible support (N = 20), non-confounded visible support (N = 17), control (N = 18) and replication visible support (N = 18). Of the 92 women who signed up to participate, 19 did not produce usable data. Omitted participants were equally distributed across conditions. As with Bolger and Amarel’s (2007) study, nine participants (10%) were excluded for suspicion. Four participants discontinued participation because they found the speech task too distressing, three participants were excluded because they recognized a confederate and two had incomplete data. Lastly, in accordance with the recommendations of Tabachnick and Fidell (2007) and further described in the results section, two participants were excluded because their cardiovascular readings were outliers. Consistent with previous research, none of the participants had a history of a cardio-respiratory disease or were taking prescribed medication that interfered with cardiovascular responses (Goyal, Shimbo, Mostofsky, & Gerin, 2008). Participants also refrained from smoking and caffeine consumption for at least 12 h before the study. Procedure Participants signed up for a 1.5-h session titled ‘Better grades: Psychological and physiological processes affecting student and teacher interactions’. Participants reported individually to a waiting area, where another participant (Confederate 1) was also waiting. A female experimenter led participants to the main experimental room and explained the cover story, which held that the main purpose was to understand the psychological and physiological processes related to grades and grading. The participant and Confederate 1 then provided informed consents and were fitted with blood pressure monitors. Two practice readings were taken so that the participant was familiar with the sensation of the cuff. Two additional minutes of rest were allowed before baseline measures were taken. See Figure 1 for a timeline of the design and measures. Baseline cardiovascular and negative emotion measures were taken. The experimenter then explained that the participant and Confederate 1 were partners 353
Effective Social Support
J. A. Kirsch and B. J. Lehman
Figure 1. Sampling schedule of the 12 cardiovascular readings obtained at 5-min intervals showing hypothesized magnitude change for reactivity
and would be randomly assigned to prepare either a written or oral speech to be graded by two graduate teaching assistants. The participant was always assigned the speech task and provided 3 min for task preparation and 3 min to practice in front of Confederate 1. The participant and Confederate 1 were led to separate rooms for preparation. Participants assigned to the invisible, non-confounded visible or control conditions interacted with a second confederate (Confederate 2) before speech preparation. Confederate 2 delivered the social support manipulation described in the succeeding text. After task preparation, the participant and Confederate 1 were brought together for the practice talk, which was to be video recorded for future analysis. Confederate 1 was told to listen in a neutral and unresponsive manner to emulate the graders. Participants practised their speeches for up to 3 min. In the replication condition only, Confederate 1 delivered support after the participant completed her practice speech. While anticipating her ‘final’ speech, the participant provided cardiovascular readings, manipulation checks, negative emotions and social evaluation concern. The participant was then told that because of time constraints, she would not complete the final speech task; rather, the previously recorded practice talk would be graded. Cardiovascular recovery was assessed while a final questionnaire was completed. Participants were debriefed following Mills (1976). Social support manipulations Table I shows the scripts and timing of each support condition. The invisible and non-confounded visible support manipulations were modifications of support found in Bolger and Amarel (2007). Confederates offered the practical advice that participants should ‘summarize what you are going to say and end with a strong conclusion’, (p. 467). Two female confederates, 354
both in their early 20s, were required for the control, the non-confounded visible and the invisible social support conditions. Confederate 1 acted as the partner who did not provide any social support. Confederate 2 delivered the social support manipulation, ostensibly as a participant who just finished delivering her final speech. Early in the study, the experimenter set the context for the invisible, non-confounded visible and control manipulations by mentioning that simultaneous study sessions were being conducted. Participants later came in contact with an ‘other’ participant (Confederate 2) who appeared flushed and was reported to have just finished delivering her final speech. In each condition, Confederate 2 asked, ‘Are you going next?’, as an attempt to engage and gain the attention of the participant. The script then diverged by support condition. As shown in Table I, in the non-confounded visible condition, Confederate 2 spoke directly to the participant after the experimenter momentarily left the room. For the invisible and control conditions, the experimenter remained in the room with the confederate and participant and asked Confederate 2, ‘Do you have any questions before you fill out these questionnaires?’ The confederate responded according to the script. The fourth experimental condition was a direct replication of Bolger and Amarel’s (2007) visible social support manipulation. After the participants’ practice speech, the experimenter asked Confederate 1, ‘Do you have any questions before we move on?’ As shown in Table I, the confederate provided support using the exact words from Bolger and Amarel. Measures and materials The following measures, adapted from Bolger and Amarel (2007), were embedded in a series of questionnaires designed to maintain the cover story. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
Effective Social Support
Table I. Social support timing and confederate scripts Condition
Social support provider
Timing of manipulation
Social support script, which follows the experimenter prompt, ‘Do you have any questions?’
Support conditions formed by the Confederate’s response to the experimenter’s question. Control Invisible
Confederate 2 Confederate 2
Before speech preparation Before speech preparation
‘No, I don’t.’ ‘Yes, I have got a question about what I did during the talk. For this kind of thing isn’t it most important to summarize what you are going to say at the beginning and make a strong conclusion at the end? That’s what they told me in my Comm. class.’
Visible support conditions where Confederate directly speaks to the participant. Non-confounded visible
Confederate 2
Before speech preparation
Replication
Confederate 1
After practice speech
Negative emotional responses Negative emotions were assessed at pre- and poststress by using the shortened nine-item Profile of Mood States (Lorr & McNair, 1971) developed by Bolger and Amarel (2007). Participants rated their current feelings and experiences on a five-point scale from 0 (not at all) to 4 (extremely). Coefficient alpha for baseline and post-stress negative emotions were 0.79 and 0.88, respectively.
Social evaluation This two-item measure assessed the degree that participants felt negatively evaluated by their support provider. Participants in the invisible, visible and control conditions rated both Confederate 2 (who was referenced as the ‘other participant’) and Confederate 1 (who was referenced as the ‘student partner’). Participants in the replication condition rated only Confederate 1 because they did not interact with Confederate 2. Social evaluation items, ‘The other participant/student partner seemed to think I would do fine’ and ‘The other participant/student partner seemed to think I would have a hard time’, were rated on a fivepoint scale from 1 (strongly disagree) to 5 (strongly agree); higher values indicate more social evaluative concern. Cronbach’s alpha was 0.63 for the student partner and 0.68 for the other participant. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
‘No, I don’t’ Experimenter then leaves the room. Confederate says to participant, ‘I just used what I learned from my Comm. class. They said it’s most important that you summarize what you say at the beginning and end with a strong conclusion.’ ‘I would like to say something to [Participant] if that’s all right. You know, to give a good talk it’s probably most important to summarize what you’re going to say at the beginning, and also to make a strong conclusion at the end.’
Suspicion check After the support manipulation, participants rated their feelings of suspicion about the study on a scale of 0 (not at all suspicious) to 4 (very suspicious). Participants were excluded if their suspicion related to the social support manipulation. Cardiovascular responses Systolic blood pressure (SBP), diastolic blood pressure (DBP) and HR were measured using a Spacelabs 90217 blood pressure monitor (Spacelabs Healthcare, Snoqualmie, WA, USA). The Spacelabs monitor has been validated by the British Hypertension Society and by the Association for the Advancement of Medical Instrumentation (Baumgart & Kamp, 1998). Seated readings were taken on the left arm. To increase reliability, 12 cardiovascular readings were spaced approximately 5 min apart (Goyal et al., 2008). Figure 1 represents the sampling schedule. For tests of task stressfulness, two measures formed the baseline, six constituted reactivity and the last four indicated recovery. Data were analysed somewhat differently for tests of social support. For these tests, prior to the social support manipulation, cardiovascular readings would not be anticipated to differ by condition; therefore, the first four readings (two baseline and two reactivity measures taken from before the manipulation but after the participant was aware of the impending speech) 355
Effective Social Support
served as the pre-support measures, the next four were post-support measures (two taken prior to the practice speech and two taken while awaiting the final speech) and the final four readings again measured recovery. Statistical strategy for cardiovascular responses Hierarchical growth curve modelling tested the social support hypotheses and the overall stressfulness of the speech task (see Llabre, Spitzer, Saab, & Schneiderman, 2001; Waugh, Panage, Mendes, & Gotlib, 2010). In contrast to repeated measures analysis of variance (ANOVA), one traditional method of analysing laboratory stress tasks, this paper draws on the power available through all 12 cardiovascular readings and uses growth curve models to test hypothesized patterns of change (e.g. linear and quadratic) at each task period (baseline, reactivity and recovery). A two-level hierarchical linear model (HLM) was specified for each cardiovascular parameter (SBP, DBP and HR). Level 1 consisted of the 12 data points obtained during the experimental session. Level 2 modelled slopes and intercepts for each participant. Random effects were tested but were kept in the model only if they predicted the outcome at p < 0.10. Degrees of freedom were approximately 70 for tests of random effects and were more than 634 for tests of fixed effects. Several steps were used to build each model. Piecewise growth curve modelling at level 1 tested cardiovascular response patterns at different task periods (Llabre et al., 2001). Cardiovascular responses were modelled through sets of contrast-coded variables that corresponded to the three theorized patterns shown in Figure 2A (magnitude change, linear trend and quadratic curve; following Waugh et al., 2010). For modelling magnitude change, values were assigned by giving the task period under consideration values of 1, with other periods coded as 0. For linear slope, values were contrast codes created to represent linear relationships centred on the middle of the task period. For example linear slope for the six reactivity measures were coded as 5, 3, 1, 1, 3 and 5. For quadratic curve, the contrast codes represented a quadratic relationship centred with positive values in the middle of the task period, creating an inverted U shape. For example, when modelling the quadratic pattern for recovery, the measures were coded as 1, 1, 1 and 1. These contrast coded variables therefore represented three different regression equations that were fit to the cardiovascular responses at each of the three task periods. Consistent with the piecewise growth-curve modelling approach, the task periods of reactivity and recovery were analysed as separate models. Social support hypotheses were tested at level 2. Following Nezlek’s (2011) guidelines for testing categorical predictors in HLM, a contrast-coded variable was constructed to test the prediction that the non-evaluative social support conditions would attenuate cardiovascular responses. For this comparison, participants in the invisible and visible conditions were coded as 1, and the 356
J. A. Kirsch and B. J. Lehman
control and replication conditions were coded as +1. This level 2 variable was tested as a moderator of the patterns of level 1 cardiovascular responses for presupport manipulation, post-support manipulation and recovery task periods. A significant positive coefficient would confirm the predicted relationship (Nezlek, 2011). As noted previously, the first two reactivity measures occurred before the social support manipulation and were therefore examined as part of the pre-support manipulation time period. The task periods of presupport manipulation, post-support manipulation and recovery were analysed separately in HLM.
Results Data screening and tests of manipulation checks Following standard screening procedures, cardiovascular responses were examined for potential errors or biologically improbable readings (Marler, Jacobs, Lehoczky, & Shapiro, 1988) and graphed. Though all three parameters were approximately normally distributed and no biologically improbable readings were found, SBP, DBP and HR readings for two participants were identified as outliers because they were set apart from the rest of the distribution and were more than three standard deviations (SDs) above the mean (Tabachnick & Fidell, 2007). Cardiovascular responses were therefore analysed both with and without the outliers. Although the analyses reported in the succeeding text exclude these outliers, the results including the outliers were generally consistent with those presented here. A separate analysis of social support visibility manipulation checks and assessments of task importance suggested that the manipulations were successful and that participants viewed their topic as somewhere between ‘moderately’ and ‘quite a bit’ important. Additional details on these analyses are available from the authors. Cardiovascular responses Test of task stressfulness The three hypothesized patterns shown in Figure 2A (magnitude change, linear slope and quadratic curve) were entered simultaneously as level-1 predictors of cardiovascular responses. Figure 2B illustrates the mean cardiovascular readings obtained throughout the stress task. Table IIA and B shows both the piecewise growth curve results for SBP, DBP and HR at reactivity and the descriptive statistics for cardiovascular responses for the 12 readings. The statistically significant and positive trends of magnitude change and quadratic curve at reactivity confirm that the stress task elicited cardiovascular responses. Test of social support The contrast-coded variable corresponding to the prediction that non-evaluative social support would Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
Effective Social Support
Figure 2. Graph of change models tested (A) and observed cardiovascular response patterns for diastolic blood pressure (DBP), systolic blood pressure (SBP) and heart rate (HR) during the speech task (B). See Table II for the cardiovascular means and for the tests of significance of the models shown in Figure 2A.
alleviate cardiovascular responses to the speech task was tested as a moderator of the cardiovascular patterns of magnitude change, linear slope and quadratic curve. Table III illustrates the results of the hierarchical growth curve model for SBP, DBP and HR. Consistent with expectations of random assignment, social support did not moderate magnitude change, linear slope or quadratic curve for the pre-support manipulation readings. Note that for social support analyses, the two cardiovascular reactivity readings that were taken prior to the social support manipulation were coded as pre-support manipulation readings, since social support differences would not be anticipated for those readings. As a result, the patterns of change for the post-support readings more closely follow a negative linear slope than a quadratic curve, and the patterns in Table III differ from those in Table II, which only tested the effects of task stressfulness. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
During the post-support manipulation task period, the social support model significantly moderated linear slope for SBP, t(704) = 2.29, p = 0.023; DBP, t(634) = 2.42, p = 0.016; and HR, t(70) = 2.73, p = 0.008.1 Figure 3 illustrates the negative linear patterns for SBP, DBP and HR. This figure indicates that participants supported prior to the practice speech (the invisible and non-confounded visible conditions) had steeper negative linear slopes than participants supported after (replication visible condition) or who did not receive any support. Follow-up tests separately contrasted theoretically the important differences between the post-support readings for the visible and invisible conditions and between the control and replication visible conditions. However, the 1 With the two outliers included in the analysis, social support moderated linear slope for SBP at p = 0.056, DBP (p = 0.031) and HR (p = 0.007).
357
358
Magnitude change Linear slope Quadratic curve Magnitude change Linear slope Quadratic curve Magnitude change Linear slope Quadratic curve
118.24 121.63 121.68 75.82 77.41 77.43 73.90 75.88 75.91
Reactivity
Intercept 6.89 (0.73) 0.07 (0.10) 0.78 (0.17) 3.24 (0.55) 0.01 (0.08) 0.21 (0.12) 4.04 (0.55) 0.01 (0.08) 0.49 (0.15)
Coefficient (SE)
Recovery
Comparing Visible and Invisible Social Support: Nonevaluative Support Buffers Cardiovascular Responses to Stress Julie A. Kirsch & Barbara J. Lehman*† Department of Psychology, Western Washington University, Bellingham, WA, USA
Abstract Previous research suggests that in contrast to invisible social support, visible social support produces exaggerated negative emotional responses. Drawing on work by Bolger and colleagues, this study disentangled social support visibility from negative social evaluation in an examination of the effects of social support on negative emotions and cardiovascular responses. As part of an anticipatory speech task, 73 female participants were randomly assigned to receive no social support, invisible social support, non-confounded visible social support or visible social support as delivered in a 2007 study by Bolger and Amarel. Twelve readings, each for systolic blood pressure, diastolic blood pressure and heart rate were taken at 5-min intervals throughout the periods of baseline, reactivity and recovery. Cardiovascular outcomes were tested by incorporating a series of theoretically driven planned contrasts into tests of stress reactivity conducted through piecewise growth curve modelling. Linear and quadratic trends established cardiovascular reactivity to the task. Further, in comparison to the control and replication conditions, the non-confounded visible and invisible social support conditions attenuated cardiovascular reactivity over time. Pre- and post-speech negative emotional responses were not affected by the social support manipulations. These results suggest that appropriately delivered visible social support may be as beneficial as invisible social support. Copyright © 2014 John Wiley & Sons, Ltd. Received 9 April 2013; Revised 15 November 2013; Accepted 20 November 2013 Keywords invisible social support; anticipatory stress; social evaluation; negative emotions; cardiovascular activity *Correspondence Barbara J. Lehman, Department of Psychology, Western Washington University, 516 High Street, Bellingham, WA 98225, USA. † Email: [email protected] Published online 22 January 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smi.2558
Although social support is clearly beneficial to physical health (House, Landis, & Umberson, 1988), its effectiveness varies according to situational fit and by support provider characteristics (Uchino, 2009). Previous research has conceptualized social support as either one’s potential access to social support (perceived available social support) or the actual exchange of resources such as advice or emotional comfort (enacted social support; Uchino, 2009). Evidence linking perceived social support to better physical health has consistently shown a positive and causal relationship, whereas evidence linking enacted social support to positive health outcomes has sometimes demonstrated weak or negative relationships (e.g. Cohen & Wills, 1985; Kaplan et al., 1994). Investigations of the mechanisms by which social support exerts positive effects on health may help elucidate inconsistencies in the effectiveness of social support. Although social support may directly promote physical and emotional outcomes, the stress-buffering model
Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
suggests that social support is particularly important for alleviating negative emotions and reducing physiological responses at times of stress (Cohen & Wills, 1985). Because excessive and prolonged cardiovascular activity has been implicated in the etiology of hypertension and coronary heart disease (Brosschot, Gerin, & Thayer, 2006; Trieber et al., 2003), it is important to determine how social support can alleviate such responses. Research has demonstrated that enacted social support does not always effectively buffer stress responses, especially for women (e.g. Allen, Blascovich, Tomaka, & Kelsey, 1991; Taylor et al., 2010). Allen et al. (1991) found that socially supported women experienced more heart rate (HR) and blood pressure elevations compared with unsupported men and women. Other researchers have found no differences between socially supported or unsupported individuals (Anthony & O’Brien, 1999). These inconsistencies suggest that the effect of enacted social support on stress responses merits closer scrutiny. 351
Effective Social Support
The current study tested a recent conceptualization of social support that may explain inconsistencies in the effects of enacted social support. Enacted social support is often manipulated visibly, meaning that recipients are aware of receiving such support. Both correlational research (Bolger, Zuckerman, & Kessler, 2000) and experimental (Bolger & Amarel, 2007) research demonstrate that unsolicited visible support exacerbates negative emotional responses to stress. Visibly enacted support is often confounded with negative social evaluation (Thorsteinsson & James, 1999) because recognizing one’s own need for support may elicit evaluative concern. It is therefore unclear whether visibly enacted support remains ineffective when the potential for negative social evaluation from the support provider is removed. The present study tested the extent to which social support visibility affects cardiovascular and emotional responses to stress in contexts both high and low in negative social evaluation. Social support visibility as a moderator of responses to stress According to Bolger and Amarel’s (2007) social support model, support can be provided either before an individual solicits support or after an individual solicits it. Bolger et al. (2000) proposed that individuals who are aware of receiving unsolicited social support during experiences of stress tend to have lower self-esteem and reduced self-efficacy. Additionally, recipients may feel incapable of coping with the stressor and negatively evaluated by the support provider. Because negative social evaluation and lack of control lead to stress appraisals (Dickerson & Kemeny, 2004), unsolicited visible support may exacerbate stress responses. In a meta-analysis of laboratory social support research, Thorsteinsson and James (1999) attributed null and adverse effects of social support to negative social evaluation by the support provider. Support conditions in which the support provider could observe and therefore potentially evaluate the recipient’s performance were least effective in attenuating responses to stress. Therefore, social support may be more effective when there is no potential for provider negative social evaluation (Kors, Linden, & Gerin, 1997). In summary, visible social support may be less effective when the support provider has or could evaluate the recipient’s ability to cope with the stressor. Reducing the recipient’s awareness of support may minimize the costs associated with visible support. The benefits of such invisible support were first tested in a diary study of intimate partners under stress. Over the course of 1 month, stressed individuals who reported low frequency of received support but whose partner ranked his or her own actions as highly supportive rated themselves low on anxiety and depression (Bolger et al., 2000) as measured at the end of each day. In everyday life, invisible social support might take the form of providing indirect advice or guidance or 352
J. A. Kirsch and B. J. Lehman
completing household chores that would be appreciated by the recipient but not recognized as socially supportive acts. Since social support was not manipulated, variables such as relationship quality and support content may have influenced negative emotional responses to stress. In a laboratory study that served as a model for the current research, Bolger and Amarel (2007) manipulated social support visibility by using an anticipatory speech paradigm. To avoid cross-gender interactions, the authors used all female experimenters, confederates and participants. A confederate partner provided invisible, visible or no social support after she listened to a participant’s practice speech. Recipients of invisible support showed the smallest increase in negative emotions, while recipients of visible support reported the highest increase. Thus, participants fared much worse when they received visible social support compared with when they received no support at all. These findings indicate that visible social support is costly to recipients experiencing a stressful event. Visibly enacted support may be detrimental because it directs attention towards the stressor and increases feelings of uncontrollability or negative social evaluation. Invisible support may instead enhance feelings of support availability without directing attention to the stressor or invoking negative social evaluation (Bolger & Amarel, 2007). For example, in the context of their laboratory task, visible social support took the form of a confederate directly offering advice on how the participant should organize a required speech (e.g. ‘You should… end with a strong conclusion’), while in the invisible support condition, the confederate offered the advice in the form of a question posed to the experimenter (e.g. ‘Isn’t it best to… end with a strong conclusion?’). In a later study, Howland and Simpson (2010) demonstrated that for intimate couples, invisible support was more likely to deflect attention from the recipient’s problems and enhance perceptions of coping resources, thereby decreasing negative emotional responses. A reduction in negative emotional responses may help to attenuate cardiovascular stress responses and influence the onset and course of cardiovascular diseases and hypertension (Feldman, Cohen, Hamrick, & Lepore, 2004). Therefore, social support visibility may be an important mechanism by which social support exerts its effects on stress and cardiovascular health. The present study The current study was designed to extend and clarify social support research and theory. Health psychologists have broadly discussed the applicability of invisible social support for physical health outcomes (e.g. Taylor et al., 2010), yet physiological responses have not been reported in the literature. This study forms an empirical test of the importance of invisible social support for both emotional and physiological stress reactivity. Further, although previous research suggests that visible support exerts an emotional cost, the reasons for Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
this cost remain unclear. Because Bolger and Amarel (2007) used the same confederate to provide support and to listen to the participant’s practice speech, the study may have confounded social support visibility with negative social evaluation. Indeed, visibly supported participants reported that the supportive confederate felt the participant had difficulties with her speech task. In a follow-up study, Bolger and Amarel removed the communication of negative evaluation by having the confederate first assure the participants that she didn’t think the participant needed help. Despite this assurance, participants in the visible support condition fared no better than participants who did not receive any support. Furthermore, this type of instrumental social support may be especially likely to be misunderstood (Trobst, 2000) and may provoke negative social evaluation. Because negative social evaluation has a potent influence on physiological and emotional responses to stress, it seems likely that feelings of negative social evaluation strengthen the association between visible social support and negative stress responses. The current study implemented a visible social support condition that minimized the potential for negative social evaluation by the support provider. This non-confounded visible social support condition was compared with invisible social support, with visible social support as manipulated by Bolger and Amarel (2007) and with a control group. Invisible and nonconfounded visible support conditions were manipulated before participants practised their speeches. The social support provider in these two conditions did not observe the participants’ performance and thereby could not evaluate it. Although a potentially evaluative confederate partner remained present in all conditions, only in the replication condition did this confederate provide social support after the practice speech, directly importing Bolger and Amarel’s (2007) visible social support condition. In this condition, the participant gave a practice speech in front of the confederate, who then provided social support in the form of advice to the participant. Because the participant had just given the practice speech, this support could likely be viewed as negative feedback about her performance, thereby raising concerns for social evaluation and potentially minimizing the benefits of the support. The primary objectives of the current study were to determine whether social support affected emotional and cardiovascular reactivity to an anticipatory speech task and to test whether the effects of support visibility were separate from negative social evaluation. We hypothesized that the forms of social support not confounded with possible negative social evaluation from the support provider would be the most effective at alleviating stress. Therefore, participants in the invisible and non-confounded visible conditions were predicted to show attenuated cardiovascular and negative emotional responses to stress. Follow-up tests then determined whether invisible social support was more beneficial than Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
Effective Social Support
visible social support. A difference between the invisible and non-confounded visible support conditions would suggest that invisible support is uniquely beneficial, while a non-significant comparison would indicate that negative social evaluation may have inflated estimates of invisible support in previous laboratory investigations.
Method Participants and design Participants were 73 female undergraduate students. The sample was 68% Caucasian, 10% Asian American, 5% African American and 15% identified as multi-ethnic/ other; 2% did not identify their ethnicity. Participants were randomly assigned to one of the four conditions: invisible support (N = 20), non-confounded visible support (N = 17), control (N = 18) and replication visible support (N = 18). Of the 92 women who signed up to participate, 19 did not produce usable data. Omitted participants were equally distributed across conditions. As with Bolger and Amarel’s (2007) study, nine participants (10%) were excluded for suspicion. Four participants discontinued participation because they found the speech task too distressing, three participants were excluded because they recognized a confederate and two had incomplete data. Lastly, in accordance with the recommendations of Tabachnick and Fidell (2007) and further described in the results section, two participants were excluded because their cardiovascular readings were outliers. Consistent with previous research, none of the participants had a history of a cardio-respiratory disease or were taking prescribed medication that interfered with cardiovascular responses (Goyal, Shimbo, Mostofsky, & Gerin, 2008). Participants also refrained from smoking and caffeine consumption for at least 12 h before the study. Procedure Participants signed up for a 1.5-h session titled ‘Better grades: Psychological and physiological processes affecting student and teacher interactions’. Participants reported individually to a waiting area, where another participant (Confederate 1) was also waiting. A female experimenter led participants to the main experimental room and explained the cover story, which held that the main purpose was to understand the psychological and physiological processes related to grades and grading. The participant and Confederate 1 then provided informed consents and were fitted with blood pressure monitors. Two practice readings were taken so that the participant was familiar with the sensation of the cuff. Two additional minutes of rest were allowed before baseline measures were taken. See Figure 1 for a timeline of the design and measures. Baseline cardiovascular and negative emotion measures were taken. The experimenter then explained that the participant and Confederate 1 were partners 353
Effective Social Support
J. A. Kirsch and B. J. Lehman
Figure 1. Sampling schedule of the 12 cardiovascular readings obtained at 5-min intervals showing hypothesized magnitude change for reactivity
and would be randomly assigned to prepare either a written or oral speech to be graded by two graduate teaching assistants. The participant was always assigned the speech task and provided 3 min for task preparation and 3 min to practice in front of Confederate 1. The participant and Confederate 1 were led to separate rooms for preparation. Participants assigned to the invisible, non-confounded visible or control conditions interacted with a second confederate (Confederate 2) before speech preparation. Confederate 2 delivered the social support manipulation described in the succeeding text. After task preparation, the participant and Confederate 1 were brought together for the practice talk, which was to be video recorded for future analysis. Confederate 1 was told to listen in a neutral and unresponsive manner to emulate the graders. Participants practised their speeches for up to 3 min. In the replication condition only, Confederate 1 delivered support after the participant completed her practice speech. While anticipating her ‘final’ speech, the participant provided cardiovascular readings, manipulation checks, negative emotions and social evaluation concern. The participant was then told that because of time constraints, she would not complete the final speech task; rather, the previously recorded practice talk would be graded. Cardiovascular recovery was assessed while a final questionnaire was completed. Participants were debriefed following Mills (1976). Social support manipulations Table I shows the scripts and timing of each support condition. The invisible and non-confounded visible support manipulations were modifications of support found in Bolger and Amarel (2007). Confederates offered the practical advice that participants should ‘summarize what you are going to say and end with a strong conclusion’, (p. 467). Two female confederates, 354
both in their early 20s, were required for the control, the non-confounded visible and the invisible social support conditions. Confederate 1 acted as the partner who did not provide any social support. Confederate 2 delivered the social support manipulation, ostensibly as a participant who just finished delivering her final speech. Early in the study, the experimenter set the context for the invisible, non-confounded visible and control manipulations by mentioning that simultaneous study sessions were being conducted. Participants later came in contact with an ‘other’ participant (Confederate 2) who appeared flushed and was reported to have just finished delivering her final speech. In each condition, Confederate 2 asked, ‘Are you going next?’, as an attempt to engage and gain the attention of the participant. The script then diverged by support condition. As shown in Table I, in the non-confounded visible condition, Confederate 2 spoke directly to the participant after the experimenter momentarily left the room. For the invisible and control conditions, the experimenter remained in the room with the confederate and participant and asked Confederate 2, ‘Do you have any questions before you fill out these questionnaires?’ The confederate responded according to the script. The fourth experimental condition was a direct replication of Bolger and Amarel’s (2007) visible social support manipulation. After the participants’ practice speech, the experimenter asked Confederate 1, ‘Do you have any questions before we move on?’ As shown in Table I, the confederate provided support using the exact words from Bolger and Amarel. Measures and materials The following measures, adapted from Bolger and Amarel (2007), were embedded in a series of questionnaires designed to maintain the cover story. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
Effective Social Support
Table I. Social support timing and confederate scripts Condition
Social support provider
Timing of manipulation
Social support script, which follows the experimenter prompt, ‘Do you have any questions?’
Support conditions formed by the Confederate’s response to the experimenter’s question. Control Invisible
Confederate 2 Confederate 2
Before speech preparation Before speech preparation
‘No, I don’t.’ ‘Yes, I have got a question about what I did during the talk. For this kind of thing isn’t it most important to summarize what you are going to say at the beginning and make a strong conclusion at the end? That’s what they told me in my Comm. class.’
Visible support conditions where Confederate directly speaks to the participant. Non-confounded visible
Confederate 2
Before speech preparation
Replication
Confederate 1
After practice speech
Negative emotional responses Negative emotions were assessed at pre- and poststress by using the shortened nine-item Profile of Mood States (Lorr & McNair, 1971) developed by Bolger and Amarel (2007). Participants rated their current feelings and experiences on a five-point scale from 0 (not at all) to 4 (extremely). Coefficient alpha for baseline and post-stress negative emotions were 0.79 and 0.88, respectively.
Social evaluation This two-item measure assessed the degree that participants felt negatively evaluated by their support provider. Participants in the invisible, visible and control conditions rated both Confederate 2 (who was referenced as the ‘other participant’) and Confederate 1 (who was referenced as the ‘student partner’). Participants in the replication condition rated only Confederate 1 because they did not interact with Confederate 2. Social evaluation items, ‘The other participant/student partner seemed to think I would do fine’ and ‘The other participant/student partner seemed to think I would have a hard time’, were rated on a fivepoint scale from 1 (strongly disagree) to 5 (strongly agree); higher values indicate more social evaluative concern. Cronbach’s alpha was 0.63 for the student partner and 0.68 for the other participant. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
‘No, I don’t’ Experimenter then leaves the room. Confederate says to participant, ‘I just used what I learned from my Comm. class. They said it’s most important that you summarize what you say at the beginning and end with a strong conclusion.’ ‘I would like to say something to [Participant] if that’s all right. You know, to give a good talk it’s probably most important to summarize what you’re going to say at the beginning, and also to make a strong conclusion at the end.’
Suspicion check After the support manipulation, participants rated their feelings of suspicion about the study on a scale of 0 (not at all suspicious) to 4 (very suspicious). Participants were excluded if their suspicion related to the social support manipulation. Cardiovascular responses Systolic blood pressure (SBP), diastolic blood pressure (DBP) and HR were measured using a Spacelabs 90217 blood pressure monitor (Spacelabs Healthcare, Snoqualmie, WA, USA). The Spacelabs monitor has been validated by the British Hypertension Society and by the Association for the Advancement of Medical Instrumentation (Baumgart & Kamp, 1998). Seated readings were taken on the left arm. To increase reliability, 12 cardiovascular readings were spaced approximately 5 min apart (Goyal et al., 2008). Figure 1 represents the sampling schedule. For tests of task stressfulness, two measures formed the baseline, six constituted reactivity and the last four indicated recovery. Data were analysed somewhat differently for tests of social support. For these tests, prior to the social support manipulation, cardiovascular readings would not be anticipated to differ by condition; therefore, the first four readings (two baseline and two reactivity measures taken from before the manipulation but after the participant was aware of the impending speech) 355
Effective Social Support
served as the pre-support measures, the next four were post-support measures (two taken prior to the practice speech and two taken while awaiting the final speech) and the final four readings again measured recovery. Statistical strategy for cardiovascular responses Hierarchical growth curve modelling tested the social support hypotheses and the overall stressfulness of the speech task (see Llabre, Spitzer, Saab, & Schneiderman, 2001; Waugh, Panage, Mendes, & Gotlib, 2010). In contrast to repeated measures analysis of variance (ANOVA), one traditional method of analysing laboratory stress tasks, this paper draws on the power available through all 12 cardiovascular readings and uses growth curve models to test hypothesized patterns of change (e.g. linear and quadratic) at each task period (baseline, reactivity and recovery). A two-level hierarchical linear model (HLM) was specified for each cardiovascular parameter (SBP, DBP and HR). Level 1 consisted of the 12 data points obtained during the experimental session. Level 2 modelled slopes and intercepts for each participant. Random effects were tested but were kept in the model only if they predicted the outcome at p < 0.10. Degrees of freedom were approximately 70 for tests of random effects and were more than 634 for tests of fixed effects. Several steps were used to build each model. Piecewise growth curve modelling at level 1 tested cardiovascular response patterns at different task periods (Llabre et al., 2001). Cardiovascular responses were modelled through sets of contrast-coded variables that corresponded to the three theorized patterns shown in Figure 2A (magnitude change, linear trend and quadratic curve; following Waugh et al., 2010). For modelling magnitude change, values were assigned by giving the task period under consideration values of 1, with other periods coded as 0. For linear slope, values were contrast codes created to represent linear relationships centred on the middle of the task period. For example linear slope for the six reactivity measures were coded as 5, 3, 1, 1, 3 and 5. For quadratic curve, the contrast codes represented a quadratic relationship centred with positive values in the middle of the task period, creating an inverted U shape. For example, when modelling the quadratic pattern for recovery, the measures were coded as 1, 1, 1 and 1. These contrast coded variables therefore represented three different regression equations that were fit to the cardiovascular responses at each of the three task periods. Consistent with the piecewise growth-curve modelling approach, the task periods of reactivity and recovery were analysed as separate models. Social support hypotheses were tested at level 2. Following Nezlek’s (2011) guidelines for testing categorical predictors in HLM, a contrast-coded variable was constructed to test the prediction that the non-evaluative social support conditions would attenuate cardiovascular responses. For this comparison, participants in the invisible and visible conditions were coded as 1, and the 356
J. A. Kirsch and B. J. Lehman
control and replication conditions were coded as +1. This level 2 variable was tested as a moderator of the patterns of level 1 cardiovascular responses for presupport manipulation, post-support manipulation and recovery task periods. A significant positive coefficient would confirm the predicted relationship (Nezlek, 2011). As noted previously, the first two reactivity measures occurred before the social support manipulation and were therefore examined as part of the pre-support manipulation time period. The task periods of presupport manipulation, post-support manipulation and recovery were analysed separately in HLM.
Results Data screening and tests of manipulation checks Following standard screening procedures, cardiovascular responses were examined for potential errors or biologically improbable readings (Marler, Jacobs, Lehoczky, & Shapiro, 1988) and graphed. Though all three parameters were approximately normally distributed and no biologically improbable readings were found, SBP, DBP and HR readings for two participants were identified as outliers because they were set apart from the rest of the distribution and were more than three standard deviations (SDs) above the mean (Tabachnick & Fidell, 2007). Cardiovascular responses were therefore analysed both with and without the outliers. Although the analyses reported in the succeeding text exclude these outliers, the results including the outliers were generally consistent with those presented here. A separate analysis of social support visibility manipulation checks and assessments of task importance suggested that the manipulations were successful and that participants viewed their topic as somewhere between ‘moderately’ and ‘quite a bit’ important. Additional details on these analyses are available from the authors. Cardiovascular responses Test of task stressfulness The three hypothesized patterns shown in Figure 2A (magnitude change, linear slope and quadratic curve) were entered simultaneously as level-1 predictors of cardiovascular responses. Figure 2B illustrates the mean cardiovascular readings obtained throughout the stress task. Table IIA and B shows both the piecewise growth curve results for SBP, DBP and HR at reactivity and the descriptive statistics for cardiovascular responses for the 12 readings. The statistically significant and positive trends of magnitude change and quadratic curve at reactivity confirm that the stress task elicited cardiovascular responses. Test of social support The contrast-coded variable corresponding to the prediction that non-evaluative social support would Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
J. A. Kirsch and B. J. Lehman
Effective Social Support
Figure 2. Graph of change models tested (A) and observed cardiovascular response patterns for diastolic blood pressure (DBP), systolic blood pressure (SBP) and heart rate (HR) during the speech task (B). See Table II for the cardiovascular means and for the tests of significance of the models shown in Figure 2A.
alleviate cardiovascular responses to the speech task was tested as a moderator of the cardiovascular patterns of magnitude change, linear slope and quadratic curve. Table III illustrates the results of the hierarchical growth curve model for SBP, DBP and HR. Consistent with expectations of random assignment, social support did not moderate magnitude change, linear slope or quadratic curve for the pre-support manipulation readings. Note that for social support analyses, the two cardiovascular reactivity readings that were taken prior to the social support manipulation were coded as pre-support manipulation readings, since social support differences would not be anticipated for those readings. As a result, the patterns of change for the post-support readings more closely follow a negative linear slope than a quadratic curve, and the patterns in Table III differ from those in Table II, which only tested the effects of task stressfulness. Stress Health 31: 351–364 (2015) © 2014 John Wiley & Sons, Ltd.
During the post-support manipulation task period, the social support model significantly moderated linear slope for SBP, t(704) = 2.29, p = 0.023; DBP, t(634) = 2.42, p = 0.016; and HR, t(70) = 2.73, p = 0.008.1 Figure 3 illustrates the negative linear patterns for SBP, DBP and HR. This figure indicates that participants supported prior to the practice speech (the invisible and non-confounded visible conditions) had steeper negative linear slopes than participants supported after (replication visible condition) or who did not receive any support. Follow-up tests separately contrasted theoretically the important differences between the post-support readings for the visible and invisible conditions and between the control and replication visible conditions. However, the 1 With the two outliers included in the analysis, social support moderated linear slope for SBP at p = 0.056, DBP (p = 0.031) and HR (p = 0.007).
357
358
Magnitude change Linear slope Quadratic curve Magnitude change Linear slope Quadratic curve Magnitude change Linear slope Quadratic curve
118.24 121.63 121.68 75.82 77.41 77.43 73.90 75.88 75.91
Reactivity
Intercept 6.89 (0.73) 0.07 (0.10) 0.78 (0.17) 3.24 (0.55) 0.01 (0.08) 0.21 (0.12) 4.04 (0.55) 0.01 (0.08) 0.49 (0.15)
Coefficient (SE)
Recovery
