Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Network Connections and Salivary Testosterone Among Older U.S. Women: Social Modulation or Hormonal Causation?

Network Connections and Salivary Testosterone Among Older U.S. Women: Social Modulation or... Abstract Objectives This study examined potentially bidirectional connections of older U.S. women’s salivary testosterone with their social network connections. Methods Data were from the 2005–2006 and 2010–2011 waves of the National Social Life, Health and Aging Project (NSHAP), a national probability sample of older U.S. adults. Autoregressive cross-lagged panel models tested linkages of women’s testosterone with their social networks. Results Consistent with recent biological theory suggesting social modulation of hormones, a higher kin proportion in one’s egocentric (person-centered) network, arguably a stable compositional feature, negatively predicted women’s testosterone levels. In contrast, findings for tie strength were consistent with hormonal regulation of women’s sociality—with both perceived support from friends and family, and closeness to network members, negatively influenced by testosterone. Discussion Rather than being a static and exogenous biological factor, older women’s testosterone levels seem partly an outcome of their social context. Implications for sexual health and hormone therapy are discussed. However, this androgen also influences dimensions of their intimate networks critical to successful aging. Findings suggest the need for social scientists to engage with the neuroendocrine literature, which offers suggestions on linkages of hormones with specific network patterns. Challenge hypothesis, Hormones, Kin proportion, Social evolution, Tie strength A range of studies links higher testosterone to specialized “life-history strategies” centered around prolific short-term and nonmonogamous partnering (Archer, 2006; Wingfield, Hegner, Dufty, & Ball, 1990). Traits facilitating these effects include risk taking and impulsivity, and are accompanied by a range of “antisocial” behaviors, reactive aggression, as well as lower empathy. Consistent with this diminished prosocial capacity, recent cross-sectional evidence from the nationally representative NHANES (National Health and Nutrition Examination Survey) indicates negative linkages of men’s testosterone with cooperative, supportive relationships beyond the intimate partnership (Gettler, 2016). In other words, the well-documented underinvestment in monogamous partners or in parenting care, among men with elevated testosterone, may be paralleled by a lack of investment in strong social relationships in general (Archer, 2006). While theory predicts similar patterns among women, their testosterone-sociality linkages remain unexplored. Moreover, causal direction in associations of testosterone with social relationships remains unclear. Both clinical and an older sociological literature appear to assume hormones are exogenous to social processes—such that any connection between the two is in the form of hormonal causation (Archer, 2006; Halpern, Udry, & Suchindran, 1998; Udry, 2000). In contrast, recent studies indicate behavioral and relational modulation of testosterone levels, suggesting that close ties may prospectively lower testosterone (Archer, 2006; Pollet, van der Meij, Cobey, & Buunk, 2011; van Anders, Steiger, & Goldey, 2015). In turn, a large and growing literature indicates women’s network ties have consistent associations with a range of better outcomes in late life, including physical and mental health (Smith & Christakis, 2008). Thus, predictors of these crucial social resources become important to explore. While studies have yielded a large number of “risk factors,” none have examined hormones. Moreover, testosterone also has deep physiological linkages with multiple health dimensions in late life (Bouman, Jan Heineman, & Faas, 2005; Cauley, Gutai, Kuller, Ledonne, & Powell, 1989). If this sex steroid does indeed influence network outcomes, therefore, it may also confound previously published associations of social ties with behaviors and health conditions (Smith & Christakis, 2008). Conversely, as discussed below, any influence of a woman’s egocentric (person-centered) web of relationships on her testosterone levels would have serious implications for the vast literature on testosterone’s apparent effects on sexual health (Bachmann et al., 2002; Pluchino et al., 2013). The current study filled these gaps by examining linkages of older women’s salivary testosterone with their social ties, as well as potential causal direction in such associations. Specifically, it used data from the first two waves of the National Social Life, Health and Aging Project (NSHAP)—a national probability sample of older U.S. adults—along with autoregressive cross-lagged (ARCL) panel models (Curran, 2000) to simultaneously address reciprocal influences of women’s testosterone on their social relationships and vice versa. The latter included one structural indicator of a nurturant relational context—the proportion of kin in one’s egocentric network—as well as two indicators of tie strength: perceived support from friends and family, and closeness to network members. It was conjectured that kin proportion, as a stable compositional feature, would be driven by a range of factors other than hormones. Thus, any linkage with testosterone would more likely follow a social modulation than hormonal causation pattern. In contrast, closeness to and perceived support from one’s social alters (i.e., network members) could take either form. Two research questions were examined: (a) do older women’s elevated testosterone levels predict their network kin proportion and tie strength? And (b) are these effects bidirectional? Testosterone and Social Relationships Explanatory models of testosterone-prosociality linkages derive from the “challenge hypothesis.” The original version of this conceptual framework concentrated on species differences in mating and social propensities. Among monogamous species, it also addressed temporal changes in them, catalyzed by testosterone (Wingfield et al., 1990). Briefly, the model proposes that this sex hormone may facilitate a trade-off between mating and parenting effort. Thus, rather than stable relationships and parental care, those with higher testosterone may invest in short-term and promiscuous mating strategies (Archer, 2006). These patterns may begin early in life, such that an individual is gradually channeled into a sexual career marked by more risky and frequent partnering. Behaviors associated with such an orientation may include a range of antisocial activities, all of which can be viewed as putting shorter-term goals before longer-term ones—as also aggression, dominance, status striving, and a general lowering of social empathy. Rather than having a “localized” association with sexuality, in other words, high testosterone might be linked more generally to life trajectories marked by less norm conformity and higher interpersonal conflict. As noted, clinical as well as an older sociological literature appears to assume hormonal causation with these associations, such that testosterone lowers prosocial propensities and heightens promiscuity (Archer, 2006; Halpern et al., 1998; Udry, 2000). In contrast, a newer “social modulation” model suggests the inverse—that is, that “exposure” to a stable and high-quality partnership, and constrained sexual opportunities, may prospectively lower testosterone (Archer, 2006; Das & Sawin, 2016; Pollet et al., 2011). Similarly, behaviors such as wielding power may also raise testosterone (van Anders et al., 2015), while participating in emotionally supportive relationships may lower it (Gettler & Oka, 2016). Per evolutionary logic, continuously high testosterone levels and their associated behaviors are suboptimal, since they elevate survival costs (from increased sexual competition and aggression) as well as reproductive ones (through less involvement in offspring, which are thus less likely to survive and pass on ancestral genes). In monogamous species, such costs are controlled by maintaining testosterone at lower baseline levels, and making this hormone responsive to environmental cues (Wingfield et al., 1990). Far from being a static and exogenous biological factor, in other words, testosterone levels are in large part a dynamic outcome of a changing social context. Thus, for instance, testosterone rises when competition (especially over potential mates) or interaction with sexually desirable partners is imminent. Human mating systems have been described as involving mild polygamy, together with high parental care, such that testosterone shows substantial ecological responsiveness (Archer, 2006; Geary, 2000). More constrained exposure to arousal-inducing social signals, therefore, may prospectively lower testosterone. In other words, while testosterone’s linkages with sexualized life trajectories and associated behaviors seem clear, causal direction at a given stage of the life course does not. Human literature on the challenge hypothesis has largely focused on men, with far fewer studies examining women’s patterns (Archer, 2006). A new generation of scholarship, rooted in feminist biological theory, is beginning to fill these gaps (van Anders, 2012). Importantly, it finds that while there is similarity across genders in linkages of testosterone to behaviors, there is also considerable variation. These differences may apply not simply to the relative strength of associations, but also to the specific behaviors or ecological factors tied to this sex steroid among women or men. Evolved functions, social roles, and/or sociocultural experiences related to femininity and masculinity are all speculated to play a role. For instance, certain activities (e.g., handshake grip strength) may be testosterone-enhancing for men but not women, because only the former have been taught to value them as competitive (van Anders, Goldey, & Kuo, 2011). Beyond the challenge hypothesis, a new Steroid/Peptide Theory of Social Bonds (S/P Theory) addresses these same testosterone-prosociality-partnership interconnections—uniting testosterone and oxytocin in a single model, and explicitly focusing on women’s patterns (van Anders et al., 2011, 2015). The framework divides intimacy into sexual (sexual contact between individuals that may be oriented around pleasure, reproduction, power, etc.) and nurturant (warm loving contact between individuals that enhances social bonds) subtypes. Both types increase oxytocin. However, only sexual intimacy is linked to elevated testosterone. Nurturant intimacy, in contrast, may have negative associations with it. Importantly, such an orientation of testosterone along a nurturance–competition gradient, in this theoretical model, applies to social relationships in general, beyond the pair bond. Current biological literature also assumes any social or relational modulation is due to spontaneous psychobiological downregulation of testosterone in nurturant, noncompetitive contexts where empathy and dampening of aggression are more needed (Gettler & Oka, 2016). In turn, evolutionary kin selection theory (also knows as the “inclusive fitness” model) strongly indicates networks rich in relatives are more likely than any other to provide precisely such an environment (Foster, Wenseleers, & Ratnieks, 2006). The model, one of the main evolutionary explanations for altruism, predicts behavior to maximize the fitness of one’s relatives, even at a cost to one’s own survival and reproduction. However, a longstanding network literature indicates a second potential mechanism through which such kin-centered contexts can influence testosterone—social control of antisocial (and hence testosterone-enhancing) behaviors. Embeddedness in a dense network of “stakeholders,” such as concerned family members, can serve as a crucial source of control over one’s counter-normative activities (Coleman, 1988; Portes, 1998; Umberson, 1987). Indeed, such relational inhibition may be gender-specific—such as enforcement of norms that proscribe older women’s but not men’s sexuality (Das, Laumann, & Waite, 2012). If so, such network composition may also dampen the former’s testosterone—although, as a structural feature of one’s web of relationships, it is less likely than tie strength to be influenced by testosterone levels. Overall, then, rather than a gender-neutral framework, building theory on hormone-sociality linkages requires careful attention to factors that may be specifically important to one gender or the other. Despite these clear and potentially conflicting theoretical predictions, no empirical analyses yet exist on linkages of women’s testosterone with their social ties. Accordingly, the current study examined three conjectures. To test social modulation, it was hypothesized that: Hypothesis 1.Women’s higher network kin proportions will predict their lower testosterone. Hypothesis 2. Women’s stronger social ties will predict their lower testosterone. Similarly, the inverse hormonal causation conjecture led to the following hypothesis: Hypothesis 3. Women’s elevated testosterone will predict their weaker social ties. Method Participants Data were from the first two waves of the U.S. National Social Life, Health, and Aging Project (NSHAP), fielded in 2005–2006 and 2010–2011, respectively. NSHAP is a national area probability sample of community residing older adults with an oversampling of Blacks and Hispanics, designed specifically to query bio-social patterns in late life. The first wave of NSHAP comprised 3,005 participants born between 1920 and 1947 (aged 57–85 years at the time of interview), with a response rate of 75.5%. In the second wave (which had an overall response rate of 76.9%), 2,261 of these individuals (including 1,184 women) were reinterviewed. The total analytic sample for the current study was comprised of 1,162 of these women. As described below, IPA weights were used to adjust for across-wave attrition. Procedure An in-person questionnaire was administered in either English or Spanish by a field interviewer in the participant’s home. Most interviewers were experienced personnel given further training by NORC at the University of Chicago, and remained with the project throughout the survey period. In addition to self-report, data included assessments of physical and sensory function, height and weight, and salivary, blood, and vaginal mucosal samples—all collected at the time of interview by nonmedically trained interviewers. Participant consent was obtained prior to interview. Institutional Review Boards at the Division of the Social Sciences and NORC at the University of Chicago approved data collection procedures (O’Muircheartaigh, English, Pedlow, & Kwok, 2014; Smith et al., 2009). Reanalysis of NSHAP secondary data was approved by a McGill University Research Ethics Board. Table 1 shows the summary statistics for all variables used in the analyses. For all self-reports, question wording was identical across waves. Table 1. Descriptive Statistics for Variables Used in Analyses Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Note: All estimates are weighted to account for differential probabilities of selection and differential nonresponse. Design-based standard errors are given in parentheses. SES = Socioeconomic status. aContinuous variable. bDichotomous variable. cOrdinal variable. dIn hourly units, on 24-hr scale. eSame as in Wave 1. fTime invariant covariate. Only Wave 1 values relevant. Open in new tab Table 1. Descriptive Statistics for Variables Used in Analyses Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Note: All estimates are weighted to account for differential probabilities of selection and differential nonresponse. Design-based standard errors are given in parentheses. SES = Socioeconomic status. aContinuous variable. bDichotomous variable. cOrdinal variable. dIn hourly units, on 24-hr scale. eSame as in Wave 1. fTime invariant covariate. Only Wave 1 values relevant. Open in new tab Measures: Social Ties Three wave-specific indicators were included in this group. First, kin proportion in network was derived from NSHAP’s egocentric (respondent-centered) social network data, and ranged continuously from 0 to 1 (Cornwell, Schumm, Laumann, & Graber, 2009). Respondents were asked to enumerate individuals with whom they discussed important matters—a standard name generator in network studies (Waite & Das, 2010). This “network roster” allowed up to five people to be named. They were then probed to characterize their relationship with each nominated network member, as one of eighteen distinct types. The kin proportion variable was simply the ratio of family relationships to all ties. Second, perceived support from friends and family combined four Likert scales—of being able to open up to and rely on friends and on family. The questions were included in the main part of the interview, and were drawn from the 2002 Health and Retirement Study (Cornwell & Waite, 2009). Since these items only contained three categories each—ranging from 0 (hardly ever or never) to 2 (often)—summary indexes based on Cronbach’s α were considered suboptimal. Instead, principal component analysis based on a polychoric correlation matrix was used (Kolenikov & Angeles, 2009). The first principal component had a Wave 1 eigenvalue of 2.28 and explained 57% of the total variance in the items, more than twice as much as the next component. The corresponding Wave 2 figures were 2.22% and 56%, respectively. Accordingly, this component was adopted as the measure of perceived support. Finally, the network roster data yielded another measure of tie strength: closeness to network members. Participants rated each such social alter on a scale of 1 (“not very close”) to 4 (“extremely close”). The summary measure was an average of these alter-specific scores. Measures: Salivary Testosterone Testosterone was derived from saliva samples taken during the biomeasure collection portion of the in-home interview (Gavrilova & Lindau, 2009; Kozloski, Schumm, & McClintock, 2014). Participants were encouraged not to eat during the interview, even though eating and dental care are reported to have no effect on the replicability of steroid levels in saliva (Gröschl, Wagner, Rauh, & Dörr, 2001). Following previously validated protocols (Granger et al., 2007), passive drool was used to collect whole unstimulated saliva, which was then frozen until assay. Salivary enzyme immunoassays were conducted using commercially available kits. Wave 1 assays were by Salimetrics Laboratories (State College, PA); in Wave 2, specimens were sent to Dresden LabService GmbH (Dresden, Germany) for duplicate assays using identical assay kits from Wave 1. Due to extreme right skew, testosterone values—originally in pg/mL—were log transformed for analysis. Measures: Control Variables This block of self-reports was comprised of both time varying and time invariant covariates, and was controlled in all models. A first such measure was a participant’s age (in years), entered linearly as a continuous variable. This variable was conceptualized as time invariant (and included only the Wave 1 indicator) since it was perfectly correlated with time, with Wave 2 age a perfect linear function of the preceding value. Ethnicity was indexed through a set of dummy variables for Black, Hispanic, and Other, with non-Hispanic White as the reference. Time-varying covariates (measured at each wave) included, first, a dichotomous indicator of marital status. In addition, network size (total number of alters enumerated in the network roster) was included to avert confounding of testosterone’s linkages with tie strength or kin composition by the sheer size of one’s egocentric network. A next time varying covariate was a standardized summary index for a participant’s socioeconomic status—comprised of a participant’s wave-specific self-reported household wealth, household income, educational attainment, perceived household income relative to acquaintances and to American families in general. Cronbach’s alpha at Wave 1 was 0.75 and at Wave 2 was 0.76. A dichotomous measure indicated current consumption of any sex hormone supplements—included to net out any pharmacological influences on endogenous testosterone. NSHAP collected a complete log of currently used medications during the in-home interview, by direct observation using a computer-based log. The Multum® drug database, based on the hierarchical classifications of the American Hospital Formulary Service, was used for coding drug names (Qato, Schumm, Johnson, Mihai, & Lindau, 2009). A continuous indicator for saliva collection time (in hourly units, on a 24-hr scale) was also controlled, given long-established circadian rhythms in older women’s testosterone (Vermeulen, 1976). Similarly, waist size (in inches) was included to net out potential linkages of this sex steroid with visceral adiposity (Kozloski et al., 2014). Physical health indicators included one’s total number of diagnosed conditions. Participants were asked about any lifetime diagnosis of a range of medical conditions, of which nine—heart attack, arthritis, ulcers, asthma, stroke, hypertension, diabetes, cancer, and enlarged prostate—were combined into a single score (Williams, Pham-Kanter, & Leitsch, 2009). Poor functional health was indexed by a continuous additive score for impaired Activities of Daily Living (Katz, 1983). Specifically, difficulties with routine everyday activities such as dressing, bathing, eating, getting in or out of bed, and using the toilet were queried through Likert scales ranging from 0 (no difficulty) to 3 (unable to do). These were then summed up to create the additive score. Statistical Analysis Three models were run, respectively examining linkages of women’s testosterone with their network kin proportion (Table 2), with perceived support from family and friends (Table 3), and with closeness to network members (Table 4). Analyses used an ARCL approach (Curran, 2000). Thus, cross-lagged paths were specified to estimate reciprocal influences for all time varying factors—the four substantively important variables, as well as all time varying covariates. In addition, for each such variable, autoregressive paths examined dependency of Wave 2 on Wave 1 values. Finally, all Wave 2 measures were regressed on time invariant covariates. Supplementary Figure 1 illustrates the model. Table 2. Autoregressive Cross-Lagged Model for Kin Proportion in Network, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 2. Autoregressive Cross-Lagged Model for Kin Proportion in Network, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 3. Autoregressive Cross-Lagged Model for Perceived Support from Friends and Family, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 3. Autoregressive Cross-Lagged Model for Perceived Support from Friends and Family, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 4. Autoregressive Cross-Lagged Model for Closeness to Network Members and Salivary Testosterone among Older U.S. Women: Coefficients (standard errors) Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 4. Autoregressive Cross-Lagged Model for Closeness to Network Members and Salivary Testosterone among Older U.S. Women: Coefficients (standard errors) Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab It is noted that ARCL models do not disaggregate between- and within-subjects effects. A newer autoregressive latent trajectory specification with structured residuals (ALT-SR) allows extraction of within-person cross lagged effects (Berry & Willoughby, 2016). Per reviewer recommendation, the latter was tried. Convergence could not be achieved—arguably due to the number and diversity of time varying covariates. Hence, inferences were based on the more conventional and widely used ARCL method. Another analytic issue was selective attrition—that is, nonreinterview of the Wave 1 participant, whether through mortality or nonresponse, due to factors also influencing covariates. To take this selection process into account, pooled logistic regression was used to fit a predictive model for attrition across waves. Covariates included a participant’s demographic and health attributes (age, education, gender, race, self-rated physical and mental health, and number of health conditions diagnosed over the lifetime). Based on predicted probabilities from this model, stabilized inverse-probability-of-attrition (IPA) weights were created (Weuve et al., 2012, 2015). Thus, participants with characteristics associated with a lower probability of continuation were assigned larger weights, “compensating” for their underrepresentation in the second wave. Finally, all analyses were weighted by the product of these IPA weights with Wave 1 population weights that adjusted for the intentional oversampling of Blacks and Hispanics and incorporated a non-response adjustment based on age and urbanicity (O’Muircheartaigh et al., 2014). IPA weights were generated with the Stata 14.1 statistical package (Stata Corp, 2014). All other analyses were conducted in Mplus Version 7.1 (Muthen & Muthen, 1998–2012). Standard errors were adjusted for sample stratification (sampling strata independently) and clustering (sampling individuals within each of 100 primary sampling units). Estimation was through a robust weighted least squares estimator using a diagonal weight matrix. Results Tables 2–4 respectively show results from cross-lagged analyses for linkages of testosterone with the kin proportion in one’s close personal network, with perceived support from friends and family, and with closeness to network members. Given the range of estimates, the focus below is on those conceptually important for pattern inference. Since each model had a mix of variable types, unstandardized coefficients and standard errors are presented. Autoregressive effects remained substantively identical across models; hence, each is stated once. Table 2: Testosterone and Kin Proportion in Network Both autoregressive effects in this model were significant—that of Wave 1 on Wave 2 kin proportion (Coeff. = 0.46, p < .01), and that of testosterone across waves (Coeff. = 0.41, p < .01). In other words, there seems to be considerable temporal constancy in both of these factors. However, even net of this apparent stability, and consistent with social modulation of testosterone, the cross-lagged effect of Wave 1 kin proportion on Wave 2 testosterone also reached significance (Coeff. = −0.21, p < .01). Contrary to hormonal causation arguments, however, the same was not true of the inverse effect, of Wave 1 testosterone on Wave 2 kin proportion. As with other biological processes, then, testosterone may have stochastic as well as stable components—with at least the former open to social influence (Lykken & Telegen, 1996). From a longer temporal perspective, the stable component may also be subject to such influence. Early life factors may imprint behavioral patterns that produce and maintain differential testosterone levels (Archer, 2006)—a conjecture that could not be tested with the NSHAP sample of older adults. Table 3: Testosterone and Perceived Support from Friends and Family In contrast to objective kin proportion, subjective perception of support from friends and family at Wave 1 did not have a cross-lagged effect on women’s Wave 2 testosterone. Rather, consistent with hormonal causation, it was Wave 1 testosterone that negatively affected Wave 2 perceived support (Coeff. = −0.18, p < .01). Older women with higher levels of this sex steroid, in other words, do seem to experience more subjective isolation even in their most proximal social ties. As with testosterone and kin proportion, Wave 1 perception of support also had a positive autoregressive effect on Wave 2 values (Coeff. = 0.35, p < .01). Table 4: Testosterone and Closeness to Network Members As with all of the other substantively important variables, the autoregressive effect of Wave 1 on Wave 2 closeness to one’s social network members was positive (Coeff. = 0.37, p < .01). Cross lagged effects supported a hormonal causation inference, consistent with Table 3 but contrary to Table 2 patterns. Specifically, Wave 1 testosterone had a negative effect on Wave 2 closeness to network (Coeff. = −0.08, p < .01), while the inverse linkage failed to reach significance. Discussion The inclusion of biological indicators in population-representative and longitudinal social surveys offers an unprecedented opportunity to explore proximal connections of life course patterns with hormonal or other biological processes. The present study, the first to use a national probability sample to examine potential causal direction in linkages of testosterone with social networks, extended this growing literature. To recall, theory deriving from the challenge hypothesis indicates that testosterone may be negatively associated with having strong, nurturant and kin-centered social ties. While men’s linkages have been explored, women’s have not—a gender discrepancy that holds across the human literature (van Anders et al., 2015). Causal direction also remains unclear, with rival conceptual models suggesting inverse effects. Moreover, patterns may vary across network dimensions. While a high kin proportion, as a stable compositional feature, would more likely influence women’s testosterone than vice versa (Hypothesis 1), the latter’s associations with tie strength could take either a similar social modulation (Hypothesis 2) or a hormonal causation (Hypothesis 3) form. Per conjecture, cross-lagged effects for kin connections indicated a sociality-to-hormones direction: while high Wave 1 proportions of relatives in one’s egocentric network had a negative effect on Wave 2 testosterone, the inverse was not true. In contrast to this clear social modulation finding, those for tie strength were fully consistent with hormonal causation (i.e., Hypothesis 3 rather than 2). For both perceived support from friends and family, and closeness to one’s network members, it was Wave 1 testosterone that negatively influenced Wave 2 relational outcomes rather than vice versa. Findings add to the emerging evidence that far from being linked solely to men’s competitive propensities and aggression—the predominant focus in the first generation of studies based on the challenge hypothesis—testosterone has distinct and nuanced implications for women’s sociality. In turn, these social factors may be critical for healthy aging. If testosterone negatively influences women’s tie strength, for instance, it also arguably lowers availability of resources channeled through such relationships. Network theorists have long argued that deep social connections rich in mutual trust can enhance a person’s perceived agency, lower discomfort with uncertainty, and reduce resistance to change (Krackhardt, 1992). Late life can be a time of high uncertainty, when individuals experience fundamental changes in the structure of both their families and their broader social network. Children leave home; retirement uproots individuals from their social networks at work; parents and elders pass away; and health problems begin impeding interaction (Hughes, Waite, Hawkley, & Cacioppo, 2004). At such times, persisting strong ties can play a critical role in successful adaptation to changing circumstances. Conversely, if the structure of women’s egocentric network influences their testosterone levels, as the results above suggest, it may also confound testosterone’s linkages with sexual health. Consistent with this argument, for instance, previous research indicates that having a kin-centered network also inhibits older women’s partnering patterns (Das et al., 2012). This as-yet-unexplored social confounding possibility implies that at least some of the previously published findings of testosterone’s effects on sexuality might be spurious—such that both the purported cause (hormones) and effect (sexual behaviors) are instead driven by social context (Bachmann et al., 2002). It also adds a new dimension to the ongoing debate over testosterone therapy and the medicalization of women’s sexual issues (Traish, Feeley, & Guay, 2009). Current work is examining, for instance, whether endogenous testosterone is more a mediator than a cause of women’s sexual desire and capacity, such that the ultimate roots of the latter lie more in a competitive and/or sexualized social context. In other words, this sex steroid might be a biological mechanism through which interpersonal situations marked by heightened sexual signaling elevate arousal. If so, psychosocial interventions may arguably have effectiveness comparable to testosterone therapy, while avoiding potential iatrogenic harm. More generally, the proximal behavioral and psychological traits the neuroendocrine literature has thus far tied to hormones have second-order linkages with a range of factors potentially important for late-life well being. Thus, more research is needed on the role of hormones in the structure of the life course, whether through actions and attitudes that induce negative health transitions and other “turning points” (Crosnoe & Elder, 2002), or social resources that buffer against ill health (Smith & Christakis, 2008). The argument here is not to reduce such patterns solely to baseline hormonal differences. At a given stage of the life trajectory, as the social modulation results above indicate, these levels are themselves products of relational patterns. Rather, findings suggest a need for conceptual models that can accommodate the complex interplay of social and neuroendocrine factors in shaping the life cycle. There were several limitations to this study. As noted, although cross-lagged analysis is widely used to explore direction of association in longitudinal data, the approach does not disaggregate between- and within-subjects effects (Berry & Willoughby, 2017). The more rigorous ALT-SR specification could not be used due to model nonconvergence. Experimental evidence would lend confidence to the causal role of testosterone in the formation and maintenance of social ties. Participants were aged 62–90 years at Wave 2, ages at which not simply testosterone levels but also their linkages with social behaviors may be very different than among younger populations. Generalizability may accordingly be limited. However, hormone-sociality patterns in late life remain understudied, despite having major implications for healthy aging, as argued above. Models adjusted only for a small range of clearly exogenous factors. Variables like negative behaviors and mental health were not controlled because they were conceptualized as mediators of testosterone-sociality connections, such that their inclusion would have amounted to overadjustment and inappropriate attenuation of estimates. However, potential omitted-variable bias remains a concern. Moreover, linkages may well vary across (be moderated by) categories of indicators that were controlled (such as ethnicity)—patterns that could not be examined in the current study due to sample-size limitations. IPA weights were designed to adjust for between-wave attrition (noninterview). However, they could not take into account measure-specific missingness patterns. Given the newness of the topic, values were not imputed. Similarly, the five-year lag between waves, while well suited to detecting stable linkages, may have blinded the analyses to important temporal inflections in effects as well as intervening processes. Finally, given only two waves of data, psychobehavioral mediation of testosterone-sociality links, implicit in the conceptual model, could not be examined. Only baseline connections could be established. More research is therefore needed on precisely why testosterone lowers women’s tie strength, and is in turn lowered by their kin-rich networks. Conclusion Longitudinal data from a nationally representative U.S. sample suggested that rather than following a uniform pattern, women’s endogenous testosterone has distinct associations with specific dimensions of their social networks. Consistent with social modulation of hormones, a higher kin proportion in one’s egocentric network, arguably a stable compositional feature, negatively predicted women’s testosterone levels. No inverse effect was found. In contrast, results for tie strength were consistent with hormonal regulation of women’s sociality. Both perceived support from one’s friends and family, and closeness to network members, were negatively influenced by testosterone, but not vice versa. Findings have implications for social confounding of testosterone-sexuality linkages, for medicalization of women’s sexuality, as well as for potential hormonal roots of women’s late-life social resources. Conceptual models that can accommodate the complex interplay of social and neuroendocrine factors in shaping the life course are needed. Funding None reported. Conflict of Interest None reported. Author contributions All work for this study—including conception, analysis and writing—was done by A. Das. Acknowledgments The author thanks the editor and two anonymous reviewers for their thorough and insightful comments. References Archer , J . ( 2006 ). Testosterone and human aggression: An evaluation of the challenge hypothesis . Neuroscience and Biobehavioral Reviews , 30 , 319 – 345 . doi: 10.1016/j.neubiorev.2004.12.007 Google Scholar Crossref Search ADS PubMed WorldCat Bachmann , G. , Bancroft , J. , Braunstein , G. , Burger , H. , Davis , S. , Dennerstein , L. ,… Traish , A .; Princeton . ( 2002 ). Female androgen insufficiency: The Princeton consensus statement on definition, classification, and assessment . Fertility and Sterility , 77 , 660 – 665 . doi: 10.1016/S0015-0282(02)02969-2 Google Scholar Crossref Search ADS PubMed WorldCat Berry , D. , & Willoughby , M. T . ( 2017 ). On the practical interpretability of cross-lagged panel models: Rethinking a developmental workhorse . Child Development , 88, 1186–1206. doi: 10.1111/cdev.12660 WorldCat Bouman , A. , Heineman , M. J. , & Faas , M. M . ( 2005 ). Sex hormones and the immune response in humans . Human Reproduction Update , 11 , 411 – 423 . doi: 10.1093/humupd/dmi008 Google Scholar Crossref Search ADS PubMed WorldCat Cauley , J. A. , Gutai , J. P. , Kuller , L. H. , LeDonne , D. , & Powell , J. G . ( 1989 ). The epidemiology of serum sex hormones in postmenopausal women . American Journal of Epidemiology , 129 , 1120 – 1131 . doi: 10.1093/oxfordjournals.aje.a115234 Google Scholar Crossref Search ADS PubMed WorldCat Coleman , J. S . ( 1988 ). Social capital in the creation of human capital . American Journal of Sociology , 94 , S95 – S120 . doi: 10.1086/228943 Google Scholar Crossref Search ADS WorldCat Cornwell , B. , Schumm , L. P. , Laumann , E. O. , & Graber , J . ( 2009 ). Social networks in the NSHAP Study: Rationale, measurement, and preliminary findings . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i47 – i55 . doi: 10.1093/geronb/gbp042 Google Scholar Crossref Search ADS WorldCat Cornwell , E. Y. , & Waite , L. J . ( 2009 ). Measuring social isolation among older adults using multiple indicators from the NSHAP study . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i38 – i46 . doi: 10.1093/geronb/gbp037 Google Scholar Crossref Search ADS WorldCat Crosnoe , R. , & Elder Jr, G. H . ( 2002 ). Successful adaptation in the later years: A life course approach to aging . Social Psychology Quarterly , 65 , 309 – 328 . doi: 10.2307/3090105 Google Scholar Crossref Search ADS WorldCat Curran , P. J . ( 2000 ). A latent curve framework for the study of developmental trajectories in adolescent substance use . In J. S. Rose , L. Chassin , C. C. Presson , & S. J. Sherman (Eds.), Multivariate applications in substance use research: New methods for new questions (pp. 1 – 4 ). Mahwah, NJ : Lawrence Erlbaum Associates . Google Preview WorldCat COPAC Das , A. , Laumann , E. O. , & Waite , L. J . ( 2012 ). Sexual expression over the life course: Results from three landmark surveys . In J. DeLamater & L. Carpenter (Eds.), Sexualities over the life course: Emerging perspectives (pp. 236 – 259 ). New York : NYU Press . Google Preview WorldCat COPAC Das , A. , & Sawin , N . ( 2016 ). Social modulation or hormonal causation? Linkages of testosterone with sexual activity and relationship quality in a nationally representative longitudinal sample of older adults . Archives of Sexual Behavior , 45 , 2101 – 2115 . doi: 10.1007/s10508-016-0815-2 Google Scholar Crossref Search ADS PubMed WorldCat Foster , K. R. , Wenseleers , T. , & Ratnieks , F. L . ( 2006 ). Kin selection is the key to altruism . Trends in Ecology & Evolution , 21 , 57 – 60 . doi: 10.1016/j.tree.2005.11.020 Google Scholar Crossref Search ADS PubMed WorldCat Gavrilova , N. , & Lindau , S. T . ( 2009 ). Salivary sex hormone measurement in a national, population-based study of older adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i94 – 105 . doi: 10.1093/geronb/gbn028 Google Scholar Crossref Search ADS WorldCat Geary , D. C . ( 2000 ). Evolution and proximate expression of human paternal investment . Psychological Bulletin , 126 , 55 – 77 . doi: 10.1037//0033-2909.126.1.55 Google Scholar Crossref Search ADS PubMed WorldCat Gettler , L. T. , & Oka , R. C . ( 2016 ). Aging US males with multiple sources of emotional social support have low testosterone . Hormones and Behavior , 78 , 32 – 42 . doi: 10.1016/j.yhbeh.2015.10.003 Google Scholar Crossref Search ADS PubMed WorldCat Granger , D. A. , Kivlighan , K. T. , Fortunato , C. , Harmon , A. G. , Hibel , L. C. , Schwartz , E. B. , & Whembolua , G. L . ( 2007 ). Integration of salivary biomarkers into developmental and behaviorally-oriented research: Problems and solutions for collecting specimens . Physiology & Behavior , 92 , 583 – 590 . doi: 10.1016/j.physbeh.2007.05.004 Google Scholar Crossref Search ADS PubMed WorldCat Gröschl , M. , Wagner , R. , Rauh , M. , & Dörr , H. G . ( 2001 ). Stability of salivary steroids: The influences of storage, food and dental care . Steroids , 66 , 737 – 741 . doi: 10.1016/s0039-128x(01)00111-8 Google Scholar Crossref Search ADS PubMed WorldCat Halpern , C. T. , Udry , J. R. , & Suchindran , C . ( 1998 ). Monthly measures of salivary testosterone predict sexual activity in adolescent males . Archives of Sexual Behavior , 27 , 445 – 465 . doi: 10.1023/A:1018700529128 Google Scholar Crossref Search ADS PubMed WorldCat Hughes , M. E. , Waite , L. J. , Hawkley , L. C. , & Cacioppo , J. T . ( 2004 ). A short scale for measuring loneliness in large surveys: Results from two population-based studies . Research on Aging , 26 , 655 – 672 . doi: 10.1177/0164027504268574 Google Scholar Crossref Search ADS PubMed WorldCat Katz , S . ( 1983 ). Assessing self-maintenance: activities of daily living, mobility, and instrumental activities of daily living . Journal of the American Geriatrics Society , 31 , 721 – 727 . doi: 10.1111/j.1532–5415.1983.tb03391.x Google Scholar Crossref Search ADS PubMed WorldCat Kolenikov , S. , & Angeles , G . ( 2009 ). Socioeconomic status measurement with discrete proxy variables: Is principal component analysis a reliable answer ? Review of Income and Wealth , 55 , 128 – 165 . doi: 10.1111/j.1475-4991.2008.00309.x Google Scholar Crossref Search ADS WorldCat Kozloski , M. J. , Schumm , L. P. , & McClintock , M. K . ( 2014 ). The utility and dynamics of salivary sex hormone measurements in the National Social Life, Health, and Aging Project, Wave 2 . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69 ( Suppl. 2 ), S215 – S228 . doi: 10.1093/geronb/gbu123 Google Scholar Crossref Search ADS WorldCat Krackhardt , D . ( 1992 ). The strength of strong ties: The importance of philos in organizations . In N. Nohria & R. Eccles (Eds.), Networks and organizations: Structure, form, and action (pp. 216 – 239 ). Boston, MA : Harvard Business School Press . Google Preview WorldCat COPAC Lykken , D. T. , & Tellegen , A . ( 1996 ). Happiness is a stochastic phenomenon . Psychological Science , 7 , 186 – 189 . doi: 10.1111/j.1467–9280.1996.tb00355.x Google Scholar Crossref Search ADS WorldCat Muthén , L. K. , & Muthén , B. O . ( 1998–2012 ). Mplus user’s guide , (7th ed.). Los Angeles, CA : Muthén & Muthén . Google Preview WorldCat COPAC O’Muircheartaigh , C. , English , N. , Pedlow , S. , & Kwok , P. K . ( 2014 ). Sample design, sample augmentation, and estimation for Wave 2 of the NSHAP . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69 ( Suppl. 2 ), S15 – S26 . doi: 10.1093/geronb/gbu053 Google Scholar Crossref Search ADS WorldCat Pluchino , N. , Carmignani , A. , Cubeddu , A. , Santoro , A. , Cela , V. , & Alcalà , T. E. ( 2013 ). Androgen therapy in women: For whom and when . Archives of Gynecology and Obstetrics , 288 , 731 – 737 . doi: 10.1007/s00404-013-2969-7 Google Scholar Crossref Search ADS PubMed WorldCat Pollet , T. V. , van der Meij , L. , Cobey , K. D. , & Buunk , A. P . ( 2011 ). Testosterone levels and their associations with lifetime number of opposite sex partners and remarriage in a large sample of American elderly men and women . Hormones and Behavior , 60 , 72 – 77 . doi: 10.1016/j.yhbeh.2011.03.005 Google Scholar Crossref Search ADS PubMed WorldCat Portes , A . ( 1998 ). Social capital: Its origins and applications in modern sociology . Annual Review of Sociology , 24 , 1 – 24 . doi: 10.1146/annurev.soc.24.1.1 Google Scholar Crossref Search ADS WorldCat Qato , D. M. , Schumm , L. P. , Johnson , M. , Mihai , A. , & Lindau , S. T . ( 2009 ). Medication data collection and coding in a home-based survey of older adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i86 – i93 . doi: 10.1093/geronb/gbp036 Google Scholar Crossref Search ADS WorldCat Smith , K. P. , & Christakis , N. A . ( 2008 ). Social networks and health . Annual Review of Sociology , 34 , 405 – 429 . doi: 10.1146/annurev.soc.34.040507.134601 Google Scholar Crossref Search ADS WorldCat Smith , S. , Jaszczak , A. , Graber , J. , Lundeen , K. , Leitsch , S. , Wargo , E. , & O'Muircheartaigh , C . ( 2009 ). Instrument development, study design implementation, and survey conduct for the national social life, health, and aging project . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i20 – i29 . doi: 10.1093/geronb/gbn013 Google Scholar Crossref Search ADS WorldCat Stata Corp . ( 2014) . STATA Release 14.0 . College Station, TX : Stata Press . Google Preview WorldCat COPAC Traish , A. M. , Feeley , R. J. , & Guay , A. T . ( 2009 ). Testosterone therapy in women with gynecological and sexual disorders: A triumph of clinical endocrinology from 1938 to 2008 . The Journal of Sexual Medicine , 6 , 334 – 351 . doi: 10.1111/j.1743-6109.2008.01121.x Google Scholar Crossref Search ADS PubMed WorldCat Udry , J. R . ( 2000 ). Biological limits of gender construction . American Sociological Review , 65 , 443 – 457 . doi: 10.2307/2657466 Google Scholar Crossref Search ADS WorldCat Umberson , D . ( 1987 ). Family status and health behaviors: Social control as a dimension of social integration . Journal of Health and Social Behavior , 28 , 306 – 319 . doi: 10.2307/2136848 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M . ( 2012 ). Testosterone and sexual desire in healthy women and men . Archives of Sexual Behavior , 41 , 1471 – 1484 . doi: 10.1007/s10508-012-9946-2 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M. , Goldey , K. L. , & Kuo , P. X . ( 2011 ). The Steroid/Peptide Theory of Social Bonds: Integrating testosterone and peptide responses for classifying social behavioral contexts . Psychoneuroendocrinology , 36 , 1265 – 1275 . doi: 10.1016/j.psyneuen.2011.06.001 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M. , Steiger , J. , & Goldey , K. L . ( 2015 ). Effects of gendered behavior on testosterone in women and men . Proceedings of the National Academy of Sciences of the United States of America , 112 , 13805 – 13810 . doi: 10.1073/pnas.1509591112 Google Scholar Crossref Search ADS PubMed WorldCat Vermeulen , A . ( 1976 ). The hormonal activity of the postmenopausal ovary . The Journal of Clinical Endocrinology and Metabolism , 42 , 247 – 253 . doi: 10.1210/jcem-42-2-247 Google Scholar Crossref Search ADS PubMed WorldCat Waite , L. , & Das , A . ( 2010 ). Families, social life, and well-being at older ages . Demography , 47 ( Suppl ), S87 – 109 . doi: 10.1353/dem.2010.0009 Google Scholar Crossref Search ADS PubMed WorldCat Weuve , J. , Tchetgen Tchetgen , E. J. , Glymour , M. M. , Beck , T. L. , Aggarwal , N. T. , Wilson , R. S. ,… Mendes de Leon , C. F . ( 2012 ). Accounting for bias due to selective attrition: The example of smoking and cognitive decline . Epidemiology , 23 , 119 – 128 . doi: 10.1097/EDE.0b013e318230e861 Google Scholar Crossref Search ADS PubMed WorldCat Weuve , J. , Proust-Lima , C. , Power , M. C. , Gross , A. L. , Hofer , S. M. , Thiébaut , R. ,… Dufouil , C .; MELODEM Initiative . ( 2015 ). Guidelines for reporting methodological challenges and evaluating potential bias in dementia research . Alzheimer’s and Dementia , 11 , 1098 – 1109 . doi: 10.1016/ j.jalz.2015.06.1885 Google Scholar Crossref Search ADS PubMed WorldCat Williams , S. R. , Pham-Kanter , G. , & Leitsch , S. A . ( 2009 ). Measures of chronic conditions and diseases associated with aging in the National Social Life, Health, and Aging Project . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 , i67 – i75 . doi: 10.1093/geronb/gbn015 Google Scholar Crossref Search ADS WorldCat Wingfield , J. C. , Hegner , R. E. , Dufty Jr, A. M. , & Ball , G. F . ( 1990 ). The “challenge hypothesis”: Theoretical implications for patterns of testosterone secretion, mating systems, and breeding strategies . The American Naturalist , 136 , 829 – 846 . doi: 10.1086/285134 Google Scholar Crossref Search ADS WorldCat © The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journals of Gerontology Series B: Psychological Sciences and Social Sciences Oxford University Press

Network Connections and Salivary Testosterone Among Older U.S. Women: Social Modulation or Hormonal Causation?

Loading next page...
 
/lp/oxford-university-press/network-connections-and-salivary-testosterone-among-older-u-s-women-C1AW2BdWaC

References (54)

Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1079-5014
eISSN
1758-5368
DOI
10.1093/geronb/gbx111
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives This study examined potentially bidirectional connections of older U.S. women’s salivary testosterone with their social network connections. Methods Data were from the 2005–2006 and 2010–2011 waves of the National Social Life, Health and Aging Project (NSHAP), a national probability sample of older U.S. adults. Autoregressive cross-lagged panel models tested linkages of women’s testosterone with their social networks. Results Consistent with recent biological theory suggesting social modulation of hormones, a higher kin proportion in one’s egocentric (person-centered) network, arguably a stable compositional feature, negatively predicted women’s testosterone levels. In contrast, findings for tie strength were consistent with hormonal regulation of women’s sociality—with both perceived support from friends and family, and closeness to network members, negatively influenced by testosterone. Discussion Rather than being a static and exogenous biological factor, older women’s testosterone levels seem partly an outcome of their social context. Implications for sexual health and hormone therapy are discussed. However, this androgen also influences dimensions of their intimate networks critical to successful aging. Findings suggest the need for social scientists to engage with the neuroendocrine literature, which offers suggestions on linkages of hormones with specific network patterns. Challenge hypothesis, Hormones, Kin proportion, Social evolution, Tie strength A range of studies links higher testosterone to specialized “life-history strategies” centered around prolific short-term and nonmonogamous partnering (Archer, 2006; Wingfield, Hegner, Dufty, & Ball, 1990). Traits facilitating these effects include risk taking and impulsivity, and are accompanied by a range of “antisocial” behaviors, reactive aggression, as well as lower empathy. Consistent with this diminished prosocial capacity, recent cross-sectional evidence from the nationally representative NHANES (National Health and Nutrition Examination Survey) indicates negative linkages of men’s testosterone with cooperative, supportive relationships beyond the intimate partnership (Gettler, 2016). In other words, the well-documented underinvestment in monogamous partners or in parenting care, among men with elevated testosterone, may be paralleled by a lack of investment in strong social relationships in general (Archer, 2006). While theory predicts similar patterns among women, their testosterone-sociality linkages remain unexplored. Moreover, causal direction in associations of testosterone with social relationships remains unclear. Both clinical and an older sociological literature appear to assume hormones are exogenous to social processes—such that any connection between the two is in the form of hormonal causation (Archer, 2006; Halpern, Udry, & Suchindran, 1998; Udry, 2000). In contrast, recent studies indicate behavioral and relational modulation of testosterone levels, suggesting that close ties may prospectively lower testosterone (Archer, 2006; Pollet, van der Meij, Cobey, & Buunk, 2011; van Anders, Steiger, & Goldey, 2015). In turn, a large and growing literature indicates women’s network ties have consistent associations with a range of better outcomes in late life, including physical and mental health (Smith & Christakis, 2008). Thus, predictors of these crucial social resources become important to explore. While studies have yielded a large number of “risk factors,” none have examined hormones. Moreover, testosterone also has deep physiological linkages with multiple health dimensions in late life (Bouman, Jan Heineman, & Faas, 2005; Cauley, Gutai, Kuller, Ledonne, & Powell, 1989). If this sex steroid does indeed influence network outcomes, therefore, it may also confound previously published associations of social ties with behaviors and health conditions (Smith & Christakis, 2008). Conversely, as discussed below, any influence of a woman’s egocentric (person-centered) web of relationships on her testosterone levels would have serious implications for the vast literature on testosterone’s apparent effects on sexual health (Bachmann et al., 2002; Pluchino et al., 2013). The current study filled these gaps by examining linkages of older women’s salivary testosterone with their social ties, as well as potential causal direction in such associations. Specifically, it used data from the first two waves of the National Social Life, Health and Aging Project (NSHAP)—a national probability sample of older U.S. adults—along with autoregressive cross-lagged (ARCL) panel models (Curran, 2000) to simultaneously address reciprocal influences of women’s testosterone on their social relationships and vice versa. The latter included one structural indicator of a nurturant relational context—the proportion of kin in one’s egocentric network—as well as two indicators of tie strength: perceived support from friends and family, and closeness to network members. It was conjectured that kin proportion, as a stable compositional feature, would be driven by a range of factors other than hormones. Thus, any linkage with testosterone would more likely follow a social modulation than hormonal causation pattern. In contrast, closeness to and perceived support from one’s social alters (i.e., network members) could take either form. Two research questions were examined: (a) do older women’s elevated testosterone levels predict their network kin proportion and tie strength? And (b) are these effects bidirectional? Testosterone and Social Relationships Explanatory models of testosterone-prosociality linkages derive from the “challenge hypothesis.” The original version of this conceptual framework concentrated on species differences in mating and social propensities. Among monogamous species, it also addressed temporal changes in them, catalyzed by testosterone (Wingfield et al., 1990). Briefly, the model proposes that this sex hormone may facilitate a trade-off between mating and parenting effort. Thus, rather than stable relationships and parental care, those with higher testosterone may invest in short-term and promiscuous mating strategies (Archer, 2006). These patterns may begin early in life, such that an individual is gradually channeled into a sexual career marked by more risky and frequent partnering. Behaviors associated with such an orientation may include a range of antisocial activities, all of which can be viewed as putting shorter-term goals before longer-term ones—as also aggression, dominance, status striving, and a general lowering of social empathy. Rather than having a “localized” association with sexuality, in other words, high testosterone might be linked more generally to life trajectories marked by less norm conformity and higher interpersonal conflict. As noted, clinical as well as an older sociological literature appears to assume hormonal causation with these associations, such that testosterone lowers prosocial propensities and heightens promiscuity (Archer, 2006; Halpern et al., 1998; Udry, 2000). In contrast, a newer “social modulation” model suggests the inverse—that is, that “exposure” to a stable and high-quality partnership, and constrained sexual opportunities, may prospectively lower testosterone (Archer, 2006; Das & Sawin, 2016; Pollet et al., 2011). Similarly, behaviors such as wielding power may also raise testosterone (van Anders et al., 2015), while participating in emotionally supportive relationships may lower it (Gettler & Oka, 2016). Per evolutionary logic, continuously high testosterone levels and their associated behaviors are suboptimal, since they elevate survival costs (from increased sexual competition and aggression) as well as reproductive ones (through less involvement in offspring, which are thus less likely to survive and pass on ancestral genes). In monogamous species, such costs are controlled by maintaining testosterone at lower baseline levels, and making this hormone responsive to environmental cues (Wingfield et al., 1990). Far from being a static and exogenous biological factor, in other words, testosterone levels are in large part a dynamic outcome of a changing social context. Thus, for instance, testosterone rises when competition (especially over potential mates) or interaction with sexually desirable partners is imminent. Human mating systems have been described as involving mild polygamy, together with high parental care, such that testosterone shows substantial ecological responsiveness (Archer, 2006; Geary, 2000). More constrained exposure to arousal-inducing social signals, therefore, may prospectively lower testosterone. In other words, while testosterone’s linkages with sexualized life trajectories and associated behaviors seem clear, causal direction at a given stage of the life course does not. Human literature on the challenge hypothesis has largely focused on men, with far fewer studies examining women’s patterns (Archer, 2006). A new generation of scholarship, rooted in feminist biological theory, is beginning to fill these gaps (van Anders, 2012). Importantly, it finds that while there is similarity across genders in linkages of testosterone to behaviors, there is also considerable variation. These differences may apply not simply to the relative strength of associations, but also to the specific behaviors or ecological factors tied to this sex steroid among women or men. Evolved functions, social roles, and/or sociocultural experiences related to femininity and masculinity are all speculated to play a role. For instance, certain activities (e.g., handshake grip strength) may be testosterone-enhancing for men but not women, because only the former have been taught to value them as competitive (van Anders, Goldey, & Kuo, 2011). Beyond the challenge hypothesis, a new Steroid/Peptide Theory of Social Bonds (S/P Theory) addresses these same testosterone-prosociality-partnership interconnections—uniting testosterone and oxytocin in a single model, and explicitly focusing on women’s patterns (van Anders et al., 2011, 2015). The framework divides intimacy into sexual (sexual contact between individuals that may be oriented around pleasure, reproduction, power, etc.) and nurturant (warm loving contact between individuals that enhances social bonds) subtypes. Both types increase oxytocin. However, only sexual intimacy is linked to elevated testosterone. Nurturant intimacy, in contrast, may have negative associations with it. Importantly, such an orientation of testosterone along a nurturance–competition gradient, in this theoretical model, applies to social relationships in general, beyond the pair bond. Current biological literature also assumes any social or relational modulation is due to spontaneous psychobiological downregulation of testosterone in nurturant, noncompetitive contexts where empathy and dampening of aggression are more needed (Gettler & Oka, 2016). In turn, evolutionary kin selection theory (also knows as the “inclusive fitness” model) strongly indicates networks rich in relatives are more likely than any other to provide precisely such an environment (Foster, Wenseleers, & Ratnieks, 2006). The model, one of the main evolutionary explanations for altruism, predicts behavior to maximize the fitness of one’s relatives, even at a cost to one’s own survival and reproduction. However, a longstanding network literature indicates a second potential mechanism through which such kin-centered contexts can influence testosterone—social control of antisocial (and hence testosterone-enhancing) behaviors. Embeddedness in a dense network of “stakeholders,” such as concerned family members, can serve as a crucial source of control over one’s counter-normative activities (Coleman, 1988; Portes, 1998; Umberson, 1987). Indeed, such relational inhibition may be gender-specific—such as enforcement of norms that proscribe older women’s but not men’s sexuality (Das, Laumann, & Waite, 2012). If so, such network composition may also dampen the former’s testosterone—although, as a structural feature of one’s web of relationships, it is less likely than tie strength to be influenced by testosterone levels. Overall, then, rather than a gender-neutral framework, building theory on hormone-sociality linkages requires careful attention to factors that may be specifically important to one gender or the other. Despite these clear and potentially conflicting theoretical predictions, no empirical analyses yet exist on linkages of women’s testosterone with their social ties. Accordingly, the current study examined three conjectures. To test social modulation, it was hypothesized that: Hypothesis 1.Women’s higher network kin proportions will predict their lower testosterone. Hypothesis 2. Women’s stronger social ties will predict their lower testosterone. Similarly, the inverse hormonal causation conjecture led to the following hypothesis: Hypothesis 3. Women’s elevated testosterone will predict their weaker social ties. Method Participants Data were from the first two waves of the U.S. National Social Life, Health, and Aging Project (NSHAP), fielded in 2005–2006 and 2010–2011, respectively. NSHAP is a national area probability sample of community residing older adults with an oversampling of Blacks and Hispanics, designed specifically to query bio-social patterns in late life. The first wave of NSHAP comprised 3,005 participants born between 1920 and 1947 (aged 57–85 years at the time of interview), with a response rate of 75.5%. In the second wave (which had an overall response rate of 76.9%), 2,261 of these individuals (including 1,184 women) were reinterviewed. The total analytic sample for the current study was comprised of 1,162 of these women. As described below, IPA weights were used to adjust for across-wave attrition. Procedure An in-person questionnaire was administered in either English or Spanish by a field interviewer in the participant’s home. Most interviewers were experienced personnel given further training by NORC at the University of Chicago, and remained with the project throughout the survey period. In addition to self-report, data included assessments of physical and sensory function, height and weight, and salivary, blood, and vaginal mucosal samples—all collected at the time of interview by nonmedically trained interviewers. Participant consent was obtained prior to interview. Institutional Review Boards at the Division of the Social Sciences and NORC at the University of Chicago approved data collection procedures (O’Muircheartaigh, English, Pedlow, & Kwok, 2014; Smith et al., 2009). Reanalysis of NSHAP secondary data was approved by a McGill University Research Ethics Board. Table 1 shows the summary statistics for all variables used in the analyses. For all self-reports, question wording was identical across waves. Table 1. Descriptive Statistics for Variables Used in Analyses Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Note: All estimates are weighted to account for differential probabilities of selection and differential nonresponse. Design-based standard errors are given in parentheses. SES = Socioeconomic status. aContinuous variable. bDichotomous variable. cOrdinal variable. dIn hourly units, on 24-hr scale. eSame as in Wave 1. fTime invariant covariate. Only Wave 1 values relevant. Open in new tab Table 1. Descriptive Statistics for Variables Used in Analyses Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Wave 1 Wave 2 Mean SE N Range Mean SE N Range Social ties Kin proportion in networka 0.65 (0.01) 1,160 0–1 0.64 (0.01) 1169 __e Perceived support from friends and familya 0.09 (0.04) 1,066 −3.31–1.44 0.22 (0.04) 1131 −3.05–1.62 Closeness to network membersa 3.20 (0.02) 1,159 1–4 3.12 (0.02) 1168 __e Salivary testosterone Testosterone (log)a 3.72 (0.02) 879 −0.69–5.39 3.49 (0.04) 983 −1.05–5.99 Time invariant covariates Agea 67.46 (0.25) 1,175 57–85 __f __f __f __f Ethnicity Whiteb 0.81 (0.02) 1,169 0–1 __f __f __f __f Blackb 0.10 (0.02) 1,169 0–1 __f __f __f __f Hispanic/otherb 0.08 (0.02) 1,169 0–1 __f __f __f __f Time varying covariates Marital statusb 0.59 (0.02) 1,175 0–1 0.40 (0.02) 1175 __e Network sizec 3.82 (0.06) 1,175 0–5 3.98 (0.04) 1173 __e SES (standardized)a −0.01 (0.06) 1,175 −2.35–6.34 −0.08 (0.05) 1175 −2.32–5.49 Hormone supplementsb 0.51 (0.02) 1,168 0–1 0.50 (0.02) 1160 __e Saliva collection timed 15.01 (0.19) 1,082 1–24 15.47 (0.14) 1018 __e Waist size (inches)a 36.26 (0.23) 1,144 23–60 37.28 (0.24) 1137 18–64 Diagnosed conditionsc 1.45 (0.05) 1,175 0–9 2.22 (0.06) 1175 __e Poor functional healthc 0.53 (0.05) 1,175 0–13 3.13 (0.20) 1175 __e Note: All estimates are weighted to account for differential probabilities of selection and differential nonresponse. Design-based standard errors are given in parentheses. SES = Socioeconomic status. aContinuous variable. bDichotomous variable. cOrdinal variable. dIn hourly units, on 24-hr scale. eSame as in Wave 1. fTime invariant covariate. Only Wave 1 values relevant. Open in new tab Measures: Social Ties Three wave-specific indicators were included in this group. First, kin proportion in network was derived from NSHAP’s egocentric (respondent-centered) social network data, and ranged continuously from 0 to 1 (Cornwell, Schumm, Laumann, & Graber, 2009). Respondents were asked to enumerate individuals with whom they discussed important matters—a standard name generator in network studies (Waite & Das, 2010). This “network roster” allowed up to five people to be named. They were then probed to characterize their relationship with each nominated network member, as one of eighteen distinct types. The kin proportion variable was simply the ratio of family relationships to all ties. Second, perceived support from friends and family combined four Likert scales—of being able to open up to and rely on friends and on family. The questions were included in the main part of the interview, and were drawn from the 2002 Health and Retirement Study (Cornwell & Waite, 2009). Since these items only contained three categories each—ranging from 0 (hardly ever or never) to 2 (often)—summary indexes based on Cronbach’s α were considered suboptimal. Instead, principal component analysis based on a polychoric correlation matrix was used (Kolenikov & Angeles, 2009). The first principal component had a Wave 1 eigenvalue of 2.28 and explained 57% of the total variance in the items, more than twice as much as the next component. The corresponding Wave 2 figures were 2.22% and 56%, respectively. Accordingly, this component was adopted as the measure of perceived support. Finally, the network roster data yielded another measure of tie strength: closeness to network members. Participants rated each such social alter on a scale of 1 (“not very close”) to 4 (“extremely close”). The summary measure was an average of these alter-specific scores. Measures: Salivary Testosterone Testosterone was derived from saliva samples taken during the biomeasure collection portion of the in-home interview (Gavrilova & Lindau, 2009; Kozloski, Schumm, & McClintock, 2014). Participants were encouraged not to eat during the interview, even though eating and dental care are reported to have no effect on the replicability of steroid levels in saliva (Gröschl, Wagner, Rauh, & Dörr, 2001). Following previously validated protocols (Granger et al., 2007), passive drool was used to collect whole unstimulated saliva, which was then frozen until assay. Salivary enzyme immunoassays were conducted using commercially available kits. Wave 1 assays were by Salimetrics Laboratories (State College, PA); in Wave 2, specimens were sent to Dresden LabService GmbH (Dresden, Germany) for duplicate assays using identical assay kits from Wave 1. Due to extreme right skew, testosterone values—originally in pg/mL—were log transformed for analysis. Measures: Control Variables This block of self-reports was comprised of both time varying and time invariant covariates, and was controlled in all models. A first such measure was a participant’s age (in years), entered linearly as a continuous variable. This variable was conceptualized as time invariant (and included only the Wave 1 indicator) since it was perfectly correlated with time, with Wave 2 age a perfect linear function of the preceding value. Ethnicity was indexed through a set of dummy variables for Black, Hispanic, and Other, with non-Hispanic White as the reference. Time-varying covariates (measured at each wave) included, first, a dichotomous indicator of marital status. In addition, network size (total number of alters enumerated in the network roster) was included to avert confounding of testosterone’s linkages with tie strength or kin composition by the sheer size of one’s egocentric network. A next time varying covariate was a standardized summary index for a participant’s socioeconomic status—comprised of a participant’s wave-specific self-reported household wealth, household income, educational attainment, perceived household income relative to acquaintances and to American families in general. Cronbach’s alpha at Wave 1 was 0.75 and at Wave 2 was 0.76. A dichotomous measure indicated current consumption of any sex hormone supplements—included to net out any pharmacological influences on endogenous testosterone. NSHAP collected a complete log of currently used medications during the in-home interview, by direct observation using a computer-based log. The Multum® drug database, based on the hierarchical classifications of the American Hospital Formulary Service, was used for coding drug names (Qato, Schumm, Johnson, Mihai, & Lindau, 2009). A continuous indicator for saliva collection time (in hourly units, on a 24-hr scale) was also controlled, given long-established circadian rhythms in older women’s testosterone (Vermeulen, 1976). Similarly, waist size (in inches) was included to net out potential linkages of this sex steroid with visceral adiposity (Kozloski et al., 2014). Physical health indicators included one’s total number of diagnosed conditions. Participants were asked about any lifetime diagnosis of a range of medical conditions, of which nine—heart attack, arthritis, ulcers, asthma, stroke, hypertension, diabetes, cancer, and enlarged prostate—were combined into a single score (Williams, Pham-Kanter, & Leitsch, 2009). Poor functional health was indexed by a continuous additive score for impaired Activities of Daily Living (Katz, 1983). Specifically, difficulties with routine everyday activities such as dressing, bathing, eating, getting in or out of bed, and using the toilet were queried through Likert scales ranging from 0 (no difficulty) to 3 (unable to do). These were then summed up to create the additive score. Statistical Analysis Three models were run, respectively examining linkages of women’s testosterone with their network kin proportion (Table 2), with perceived support from family and friends (Table 3), and with closeness to network members (Table 4). Analyses used an ARCL approach (Curran, 2000). Thus, cross-lagged paths were specified to estimate reciprocal influences for all time varying factors—the four substantively important variables, as well as all time varying covariates. In addition, for each such variable, autoregressive paths examined dependency of Wave 2 on Wave 1 values. Finally, all Wave 2 measures were regressed on time invariant covariates. Supplementary Figure 1 illustrates the model. Table 2. Autoregressive Cross-Lagged Model for Kin Proportion in Network, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 2. Autoregressive Cross-Lagged Model for Kin Proportion in Network, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Kin proportion in network (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Kin proportion in network (Wave 1) 0.46** −0.21** (0.03) (0.08) Testosterone (log) (Wave 1) −0.01 0.41** (0.02) (0.04) Time invariant covariates Age 0.01* −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.05 0.10 (0.03) (0.13) Hispanic/other 0.10** 0.11 (0.04) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.05 −0.06 (0.05) (0.13) Wave 2 0.03 0.01 (0.02) (0.04) Network size Wave 1 0.01 −0.01 (0.01) (0.02) Wave 2 −0.04** 0.00 (0.01) (0.03) SES (standardized) Wave 1 −0.02* −0.01 (0.01) (0.05) Wave 2 −0.03 −0.06 (0.02) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 3. Autoregressive Cross-Lagged Model for Perceived Support from Friends and Family, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 3. Autoregressive Cross-Lagged Model for Perceived Support from Friends and Family, and Salivary Testosterone, among Older U.S. Women: Coefficients (standard errors) Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Perceived support from friends and family (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Perceived support from friends and family (Wave 1) 0.35** 0.04 (0.03) (0.03) Testosterone (log) (Wave 1) −0.18** 0.41** (0.07) (0.04) Time invariant covariates Age −0.02** −0.01* (0.01) (0.00) Ethnicity (ref: white) Black −0.28* 0.10 (0.11) (0.13) Hispanic/other −0.36** 0.10 (0.13) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.48** −0.06 (0.18) (0.13) Wave 2 −0.17** 0.01 (0.06) (0.04) Network size Wave 1 0.11** −0.01 (0.03) (0.02) Wave 2 0.17** 0.00 (0.03) (0.03) SES (standardized) Wave 1 0.03 −0.01 (0.06) (0.05) Wave 2 0.02 −0.06 (0.06) (0.05) CFI 0.90 RMSEA 0.05 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 4. Autoregressive Cross-Lagged Model for Closeness to Network Members and Salivary Testosterone among Older U.S. Women: Coefficients (standard errors) Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab Table 4. Autoregressive Cross-Lagged Model for Closeness to Network Members and Salivary Testosterone among Older U.S. Women: Coefficients (standard errors) Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Closeness to network members (Wave 2)a Testosterone (log) (Wave 2)a Autoregressive and cross-lagged effects Closeness to network members (Wave 1) 0.37** 0.02 (0.03) (0.07) Testosterone (log) (Wave 1) −0.08** 0.41** (0.03) (0.04) Time invariant covariates Age −0.01** −0.01* (0.00) (0.00) Ethnicity (ref: white) Black 0.17** 0.10 (0.04) (0.13) Hispanic/other −0.17** 0.10 (0.06) (0.13) Time varying covariates Marital status (ref: unmarried) Wave 1 0.03 −0.05 (0.09) (0.13) Wave 2 0.02 0.00 (0.03) (0.04) Network size Wave 1 0.03* −0.01 (0.01) (0.02) Wave 2 −0.07** 0.00 (0.02) (0.03) SES (standardized) Wave 1 0.00 −0.01 (0.03) (0.05) Wave 2 −0.02 −0.06 (0.03) (0.05) CFI 0.92 RMSEA 0.04 N 1,162 Note: In addition to those shown, analyses incorporated the following time-varying covariates: hormone supplements, saliva collection time, waist size, diagnosed conditions, and poor functional health. The model also included cross-lagged and autoregressive effects for all time varying covariates, as well as influences of time invariant measures on these factors. Results are available on request. Estimates were weighted to adjust for differential probabilities of selection and differential nonresponse. Between-wave attrition was handled through inverse-probability-of-attrition weights. Design-based standard errors are given in parentheses. Bold represents the significant at least p <.05. CFI = Comparative fit index; RMSEA = Root mean square error of approximation; SES = Socioeconomic status. aContinuous outcome. Estimates are ordinary least squares coefficients. *p < .05; **p < .01. Open in new tab It is noted that ARCL models do not disaggregate between- and within-subjects effects. A newer autoregressive latent trajectory specification with structured residuals (ALT-SR) allows extraction of within-person cross lagged effects (Berry & Willoughby, 2016). Per reviewer recommendation, the latter was tried. Convergence could not be achieved—arguably due to the number and diversity of time varying covariates. Hence, inferences were based on the more conventional and widely used ARCL method. Another analytic issue was selective attrition—that is, nonreinterview of the Wave 1 participant, whether through mortality or nonresponse, due to factors also influencing covariates. To take this selection process into account, pooled logistic regression was used to fit a predictive model for attrition across waves. Covariates included a participant’s demographic and health attributes (age, education, gender, race, self-rated physical and mental health, and number of health conditions diagnosed over the lifetime). Based on predicted probabilities from this model, stabilized inverse-probability-of-attrition (IPA) weights were created (Weuve et al., 2012, 2015). Thus, participants with characteristics associated with a lower probability of continuation were assigned larger weights, “compensating” for their underrepresentation in the second wave. Finally, all analyses were weighted by the product of these IPA weights with Wave 1 population weights that adjusted for the intentional oversampling of Blacks and Hispanics and incorporated a non-response adjustment based on age and urbanicity (O’Muircheartaigh et al., 2014). IPA weights were generated with the Stata 14.1 statistical package (Stata Corp, 2014). All other analyses were conducted in Mplus Version 7.1 (Muthen & Muthen, 1998–2012). Standard errors were adjusted for sample stratification (sampling strata independently) and clustering (sampling individuals within each of 100 primary sampling units). Estimation was through a robust weighted least squares estimator using a diagonal weight matrix. Results Tables 2–4 respectively show results from cross-lagged analyses for linkages of testosterone with the kin proportion in one’s close personal network, with perceived support from friends and family, and with closeness to network members. Given the range of estimates, the focus below is on those conceptually important for pattern inference. Since each model had a mix of variable types, unstandardized coefficients and standard errors are presented. Autoregressive effects remained substantively identical across models; hence, each is stated once. Table 2: Testosterone and Kin Proportion in Network Both autoregressive effects in this model were significant—that of Wave 1 on Wave 2 kin proportion (Coeff. = 0.46, p < .01), and that of testosterone across waves (Coeff. = 0.41, p < .01). In other words, there seems to be considerable temporal constancy in both of these factors. However, even net of this apparent stability, and consistent with social modulation of testosterone, the cross-lagged effect of Wave 1 kin proportion on Wave 2 testosterone also reached significance (Coeff. = −0.21, p < .01). Contrary to hormonal causation arguments, however, the same was not true of the inverse effect, of Wave 1 testosterone on Wave 2 kin proportion. As with other biological processes, then, testosterone may have stochastic as well as stable components—with at least the former open to social influence (Lykken & Telegen, 1996). From a longer temporal perspective, the stable component may also be subject to such influence. Early life factors may imprint behavioral patterns that produce and maintain differential testosterone levels (Archer, 2006)—a conjecture that could not be tested with the NSHAP sample of older adults. Table 3: Testosterone and Perceived Support from Friends and Family In contrast to objective kin proportion, subjective perception of support from friends and family at Wave 1 did not have a cross-lagged effect on women’s Wave 2 testosterone. Rather, consistent with hormonal causation, it was Wave 1 testosterone that negatively affected Wave 2 perceived support (Coeff. = −0.18, p < .01). Older women with higher levels of this sex steroid, in other words, do seem to experience more subjective isolation even in their most proximal social ties. As with testosterone and kin proportion, Wave 1 perception of support also had a positive autoregressive effect on Wave 2 values (Coeff. = 0.35, p < .01). Table 4: Testosterone and Closeness to Network Members As with all of the other substantively important variables, the autoregressive effect of Wave 1 on Wave 2 closeness to one’s social network members was positive (Coeff. = 0.37, p < .01). Cross lagged effects supported a hormonal causation inference, consistent with Table 3 but contrary to Table 2 patterns. Specifically, Wave 1 testosterone had a negative effect on Wave 2 closeness to network (Coeff. = −0.08, p < .01), while the inverse linkage failed to reach significance. Discussion The inclusion of biological indicators in population-representative and longitudinal social surveys offers an unprecedented opportunity to explore proximal connections of life course patterns with hormonal or other biological processes. The present study, the first to use a national probability sample to examine potential causal direction in linkages of testosterone with social networks, extended this growing literature. To recall, theory deriving from the challenge hypothesis indicates that testosterone may be negatively associated with having strong, nurturant and kin-centered social ties. While men’s linkages have been explored, women’s have not—a gender discrepancy that holds across the human literature (van Anders et al., 2015). Causal direction also remains unclear, with rival conceptual models suggesting inverse effects. Moreover, patterns may vary across network dimensions. While a high kin proportion, as a stable compositional feature, would more likely influence women’s testosterone than vice versa (Hypothesis 1), the latter’s associations with tie strength could take either a similar social modulation (Hypothesis 2) or a hormonal causation (Hypothesis 3) form. Per conjecture, cross-lagged effects for kin connections indicated a sociality-to-hormones direction: while high Wave 1 proportions of relatives in one’s egocentric network had a negative effect on Wave 2 testosterone, the inverse was not true. In contrast to this clear social modulation finding, those for tie strength were fully consistent with hormonal causation (i.e., Hypothesis 3 rather than 2). For both perceived support from friends and family, and closeness to one’s network members, it was Wave 1 testosterone that negatively influenced Wave 2 relational outcomes rather than vice versa. Findings add to the emerging evidence that far from being linked solely to men’s competitive propensities and aggression—the predominant focus in the first generation of studies based on the challenge hypothesis—testosterone has distinct and nuanced implications for women’s sociality. In turn, these social factors may be critical for healthy aging. If testosterone negatively influences women’s tie strength, for instance, it also arguably lowers availability of resources channeled through such relationships. Network theorists have long argued that deep social connections rich in mutual trust can enhance a person’s perceived agency, lower discomfort with uncertainty, and reduce resistance to change (Krackhardt, 1992). Late life can be a time of high uncertainty, when individuals experience fundamental changes in the structure of both their families and their broader social network. Children leave home; retirement uproots individuals from their social networks at work; parents and elders pass away; and health problems begin impeding interaction (Hughes, Waite, Hawkley, & Cacioppo, 2004). At such times, persisting strong ties can play a critical role in successful adaptation to changing circumstances. Conversely, if the structure of women’s egocentric network influences their testosterone levels, as the results above suggest, it may also confound testosterone’s linkages with sexual health. Consistent with this argument, for instance, previous research indicates that having a kin-centered network also inhibits older women’s partnering patterns (Das et al., 2012). This as-yet-unexplored social confounding possibility implies that at least some of the previously published findings of testosterone’s effects on sexuality might be spurious—such that both the purported cause (hormones) and effect (sexual behaviors) are instead driven by social context (Bachmann et al., 2002). It also adds a new dimension to the ongoing debate over testosterone therapy and the medicalization of women’s sexual issues (Traish, Feeley, & Guay, 2009). Current work is examining, for instance, whether endogenous testosterone is more a mediator than a cause of women’s sexual desire and capacity, such that the ultimate roots of the latter lie more in a competitive and/or sexualized social context. In other words, this sex steroid might be a biological mechanism through which interpersonal situations marked by heightened sexual signaling elevate arousal. If so, psychosocial interventions may arguably have effectiveness comparable to testosterone therapy, while avoiding potential iatrogenic harm. More generally, the proximal behavioral and psychological traits the neuroendocrine literature has thus far tied to hormones have second-order linkages with a range of factors potentially important for late-life well being. Thus, more research is needed on the role of hormones in the structure of the life course, whether through actions and attitudes that induce negative health transitions and other “turning points” (Crosnoe & Elder, 2002), or social resources that buffer against ill health (Smith & Christakis, 2008). The argument here is not to reduce such patterns solely to baseline hormonal differences. At a given stage of the life trajectory, as the social modulation results above indicate, these levels are themselves products of relational patterns. Rather, findings suggest a need for conceptual models that can accommodate the complex interplay of social and neuroendocrine factors in shaping the life cycle. There were several limitations to this study. As noted, although cross-lagged analysis is widely used to explore direction of association in longitudinal data, the approach does not disaggregate between- and within-subjects effects (Berry & Willoughby, 2017). The more rigorous ALT-SR specification could not be used due to model nonconvergence. Experimental evidence would lend confidence to the causal role of testosterone in the formation and maintenance of social ties. Participants were aged 62–90 years at Wave 2, ages at which not simply testosterone levels but also their linkages with social behaviors may be very different than among younger populations. Generalizability may accordingly be limited. However, hormone-sociality patterns in late life remain understudied, despite having major implications for healthy aging, as argued above. Models adjusted only for a small range of clearly exogenous factors. Variables like negative behaviors and mental health were not controlled because they were conceptualized as mediators of testosterone-sociality connections, such that their inclusion would have amounted to overadjustment and inappropriate attenuation of estimates. However, potential omitted-variable bias remains a concern. Moreover, linkages may well vary across (be moderated by) categories of indicators that were controlled (such as ethnicity)—patterns that could not be examined in the current study due to sample-size limitations. IPA weights were designed to adjust for between-wave attrition (noninterview). However, they could not take into account measure-specific missingness patterns. Given the newness of the topic, values were not imputed. Similarly, the five-year lag between waves, while well suited to detecting stable linkages, may have blinded the analyses to important temporal inflections in effects as well as intervening processes. Finally, given only two waves of data, psychobehavioral mediation of testosterone-sociality links, implicit in the conceptual model, could not be examined. Only baseline connections could be established. More research is therefore needed on precisely why testosterone lowers women’s tie strength, and is in turn lowered by their kin-rich networks. Conclusion Longitudinal data from a nationally representative U.S. sample suggested that rather than following a uniform pattern, women’s endogenous testosterone has distinct associations with specific dimensions of their social networks. Consistent with social modulation of hormones, a higher kin proportion in one’s egocentric network, arguably a stable compositional feature, negatively predicted women’s testosterone levels. No inverse effect was found. In contrast, results for tie strength were consistent with hormonal regulation of women’s sociality. Both perceived support from one’s friends and family, and closeness to network members, were negatively influenced by testosterone, but not vice versa. Findings have implications for social confounding of testosterone-sexuality linkages, for medicalization of women’s sexuality, as well as for potential hormonal roots of women’s late-life social resources. Conceptual models that can accommodate the complex interplay of social and neuroendocrine factors in shaping the life course are needed. Funding None reported. Conflict of Interest None reported. Author contributions All work for this study—including conception, analysis and writing—was done by A. Das. Acknowledgments The author thanks the editor and two anonymous reviewers for their thorough and insightful comments. References Archer , J . ( 2006 ). Testosterone and human aggression: An evaluation of the challenge hypothesis . Neuroscience and Biobehavioral Reviews , 30 , 319 – 345 . doi: 10.1016/j.neubiorev.2004.12.007 Google Scholar Crossref Search ADS PubMed WorldCat Bachmann , G. , Bancroft , J. , Braunstein , G. , Burger , H. , Davis , S. , Dennerstein , L. ,… Traish , A .; Princeton . ( 2002 ). Female androgen insufficiency: The Princeton consensus statement on definition, classification, and assessment . Fertility and Sterility , 77 , 660 – 665 . doi: 10.1016/S0015-0282(02)02969-2 Google Scholar Crossref Search ADS PubMed WorldCat Berry , D. , & Willoughby , M. T . ( 2017 ). On the practical interpretability of cross-lagged panel models: Rethinking a developmental workhorse . Child Development , 88, 1186–1206. doi: 10.1111/cdev.12660 WorldCat Bouman , A. , Heineman , M. J. , & Faas , M. M . ( 2005 ). Sex hormones and the immune response in humans . Human Reproduction Update , 11 , 411 – 423 . doi: 10.1093/humupd/dmi008 Google Scholar Crossref Search ADS PubMed WorldCat Cauley , J. A. , Gutai , J. P. , Kuller , L. H. , LeDonne , D. , & Powell , J. G . ( 1989 ). The epidemiology of serum sex hormones in postmenopausal women . American Journal of Epidemiology , 129 , 1120 – 1131 . doi: 10.1093/oxfordjournals.aje.a115234 Google Scholar Crossref Search ADS PubMed WorldCat Coleman , J. S . ( 1988 ). Social capital in the creation of human capital . American Journal of Sociology , 94 , S95 – S120 . doi: 10.1086/228943 Google Scholar Crossref Search ADS WorldCat Cornwell , B. , Schumm , L. P. , Laumann , E. O. , & Graber , J . ( 2009 ). Social networks in the NSHAP Study: Rationale, measurement, and preliminary findings . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i47 – i55 . doi: 10.1093/geronb/gbp042 Google Scholar Crossref Search ADS WorldCat Cornwell , E. Y. , & Waite , L. J . ( 2009 ). Measuring social isolation among older adults using multiple indicators from the NSHAP study . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i38 – i46 . doi: 10.1093/geronb/gbp037 Google Scholar Crossref Search ADS WorldCat Crosnoe , R. , & Elder Jr, G. H . ( 2002 ). Successful adaptation in the later years: A life course approach to aging . Social Psychology Quarterly , 65 , 309 – 328 . doi: 10.2307/3090105 Google Scholar Crossref Search ADS WorldCat Curran , P. J . ( 2000 ). A latent curve framework for the study of developmental trajectories in adolescent substance use . In J. S. Rose , L. Chassin , C. C. Presson , & S. J. Sherman (Eds.), Multivariate applications in substance use research: New methods for new questions (pp. 1 – 4 ). Mahwah, NJ : Lawrence Erlbaum Associates . Google Preview WorldCat COPAC Das , A. , Laumann , E. O. , & Waite , L. J . ( 2012 ). Sexual expression over the life course: Results from three landmark surveys . In J. DeLamater & L. Carpenter (Eds.), Sexualities over the life course: Emerging perspectives (pp. 236 – 259 ). New York : NYU Press . Google Preview WorldCat COPAC Das , A. , & Sawin , N . ( 2016 ). Social modulation or hormonal causation? Linkages of testosterone with sexual activity and relationship quality in a nationally representative longitudinal sample of older adults . Archives of Sexual Behavior , 45 , 2101 – 2115 . doi: 10.1007/s10508-016-0815-2 Google Scholar Crossref Search ADS PubMed WorldCat Foster , K. R. , Wenseleers , T. , & Ratnieks , F. L . ( 2006 ). Kin selection is the key to altruism . Trends in Ecology & Evolution , 21 , 57 – 60 . doi: 10.1016/j.tree.2005.11.020 Google Scholar Crossref Search ADS PubMed WorldCat Gavrilova , N. , & Lindau , S. T . ( 2009 ). Salivary sex hormone measurement in a national, population-based study of older adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i94 – 105 . doi: 10.1093/geronb/gbn028 Google Scholar Crossref Search ADS WorldCat Geary , D. C . ( 2000 ). Evolution and proximate expression of human paternal investment . Psychological Bulletin , 126 , 55 – 77 . doi: 10.1037//0033-2909.126.1.55 Google Scholar Crossref Search ADS PubMed WorldCat Gettler , L. T. , & Oka , R. C . ( 2016 ). Aging US males with multiple sources of emotional social support have low testosterone . Hormones and Behavior , 78 , 32 – 42 . doi: 10.1016/j.yhbeh.2015.10.003 Google Scholar Crossref Search ADS PubMed WorldCat Granger , D. A. , Kivlighan , K. T. , Fortunato , C. , Harmon , A. G. , Hibel , L. C. , Schwartz , E. B. , & Whembolua , G. L . ( 2007 ). Integration of salivary biomarkers into developmental and behaviorally-oriented research: Problems and solutions for collecting specimens . Physiology & Behavior , 92 , 583 – 590 . doi: 10.1016/j.physbeh.2007.05.004 Google Scholar Crossref Search ADS PubMed WorldCat Gröschl , M. , Wagner , R. , Rauh , M. , & Dörr , H. G . ( 2001 ). Stability of salivary steroids: The influences of storage, food and dental care . Steroids , 66 , 737 – 741 . doi: 10.1016/s0039-128x(01)00111-8 Google Scholar Crossref Search ADS PubMed WorldCat Halpern , C. T. , Udry , J. R. , & Suchindran , C . ( 1998 ). Monthly measures of salivary testosterone predict sexual activity in adolescent males . Archives of Sexual Behavior , 27 , 445 – 465 . doi: 10.1023/A:1018700529128 Google Scholar Crossref Search ADS PubMed WorldCat Hughes , M. E. , Waite , L. J. , Hawkley , L. C. , & Cacioppo , J. T . ( 2004 ). A short scale for measuring loneliness in large surveys: Results from two population-based studies . Research on Aging , 26 , 655 – 672 . doi: 10.1177/0164027504268574 Google Scholar Crossref Search ADS PubMed WorldCat Katz , S . ( 1983 ). Assessing self-maintenance: activities of daily living, mobility, and instrumental activities of daily living . Journal of the American Geriatrics Society , 31 , 721 – 727 . doi: 10.1111/j.1532–5415.1983.tb03391.x Google Scholar Crossref Search ADS PubMed WorldCat Kolenikov , S. , & Angeles , G . ( 2009 ). Socioeconomic status measurement with discrete proxy variables: Is principal component analysis a reliable answer ? Review of Income and Wealth , 55 , 128 – 165 . doi: 10.1111/j.1475-4991.2008.00309.x Google Scholar Crossref Search ADS WorldCat Kozloski , M. J. , Schumm , L. P. , & McClintock , M. K . ( 2014 ). The utility and dynamics of salivary sex hormone measurements in the National Social Life, Health, and Aging Project, Wave 2 . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69 ( Suppl. 2 ), S215 – S228 . doi: 10.1093/geronb/gbu123 Google Scholar Crossref Search ADS WorldCat Krackhardt , D . ( 1992 ). The strength of strong ties: The importance of philos in organizations . In N. Nohria & R. Eccles (Eds.), Networks and organizations: Structure, form, and action (pp. 216 – 239 ). Boston, MA : Harvard Business School Press . Google Preview WorldCat COPAC Lykken , D. T. , & Tellegen , A . ( 1996 ). Happiness is a stochastic phenomenon . Psychological Science , 7 , 186 – 189 . doi: 10.1111/j.1467–9280.1996.tb00355.x Google Scholar Crossref Search ADS WorldCat Muthén , L. K. , & Muthén , B. O . ( 1998–2012 ). Mplus user’s guide , (7th ed.). Los Angeles, CA : Muthén & Muthén . Google Preview WorldCat COPAC O’Muircheartaigh , C. , English , N. , Pedlow , S. , & Kwok , P. K . ( 2014 ). Sample design, sample augmentation, and estimation for Wave 2 of the NSHAP . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69 ( Suppl. 2 ), S15 – S26 . doi: 10.1093/geronb/gbu053 Google Scholar Crossref Search ADS WorldCat Pluchino , N. , Carmignani , A. , Cubeddu , A. , Santoro , A. , Cela , V. , & Alcalà , T. E. ( 2013 ). Androgen therapy in women: For whom and when . Archives of Gynecology and Obstetrics , 288 , 731 – 737 . doi: 10.1007/s00404-013-2969-7 Google Scholar Crossref Search ADS PubMed WorldCat Pollet , T. V. , van der Meij , L. , Cobey , K. D. , & Buunk , A. P . ( 2011 ). Testosterone levels and their associations with lifetime number of opposite sex partners and remarriage in a large sample of American elderly men and women . Hormones and Behavior , 60 , 72 – 77 . doi: 10.1016/j.yhbeh.2011.03.005 Google Scholar Crossref Search ADS PubMed WorldCat Portes , A . ( 1998 ). Social capital: Its origins and applications in modern sociology . Annual Review of Sociology , 24 , 1 – 24 . doi: 10.1146/annurev.soc.24.1.1 Google Scholar Crossref Search ADS WorldCat Qato , D. M. , Schumm , L. P. , Johnson , M. , Mihai , A. , & Lindau , S. T . ( 2009 ). Medication data collection and coding in a home-based survey of older adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i86 – i93 . doi: 10.1093/geronb/gbp036 Google Scholar Crossref Search ADS WorldCat Smith , K. P. , & Christakis , N. A . ( 2008 ). Social networks and health . Annual Review of Sociology , 34 , 405 – 429 . doi: 10.1146/annurev.soc.34.040507.134601 Google Scholar Crossref Search ADS WorldCat Smith , S. , Jaszczak , A. , Graber , J. , Lundeen , K. , Leitsch , S. , Wargo , E. , & O'Muircheartaigh , C . ( 2009 ). Instrument development, study design implementation, and survey conduct for the national social life, health, and aging project . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 ( Suppl. 1 ), i20 – i29 . doi: 10.1093/geronb/gbn013 Google Scholar Crossref Search ADS WorldCat Stata Corp . ( 2014) . STATA Release 14.0 . College Station, TX : Stata Press . Google Preview WorldCat COPAC Traish , A. M. , Feeley , R. J. , & Guay , A. T . ( 2009 ). Testosterone therapy in women with gynecological and sexual disorders: A triumph of clinical endocrinology from 1938 to 2008 . The Journal of Sexual Medicine , 6 , 334 – 351 . doi: 10.1111/j.1743-6109.2008.01121.x Google Scholar Crossref Search ADS PubMed WorldCat Udry , J. R . ( 2000 ). Biological limits of gender construction . American Sociological Review , 65 , 443 – 457 . doi: 10.2307/2657466 Google Scholar Crossref Search ADS WorldCat Umberson , D . ( 1987 ). Family status and health behaviors: Social control as a dimension of social integration . Journal of Health and Social Behavior , 28 , 306 – 319 . doi: 10.2307/2136848 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M . ( 2012 ). Testosterone and sexual desire in healthy women and men . Archives of Sexual Behavior , 41 , 1471 – 1484 . doi: 10.1007/s10508-012-9946-2 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M. , Goldey , K. L. , & Kuo , P. X . ( 2011 ). The Steroid/Peptide Theory of Social Bonds: Integrating testosterone and peptide responses for classifying social behavioral contexts . Psychoneuroendocrinology , 36 , 1265 – 1275 . doi: 10.1016/j.psyneuen.2011.06.001 Google Scholar Crossref Search ADS PubMed WorldCat van Anders , S. M. , Steiger , J. , & Goldey , K. L . ( 2015 ). Effects of gendered behavior on testosterone in women and men . Proceedings of the National Academy of Sciences of the United States of America , 112 , 13805 – 13810 . doi: 10.1073/pnas.1509591112 Google Scholar Crossref Search ADS PubMed WorldCat Vermeulen , A . ( 1976 ). The hormonal activity of the postmenopausal ovary . The Journal of Clinical Endocrinology and Metabolism , 42 , 247 – 253 . doi: 10.1210/jcem-42-2-247 Google Scholar Crossref Search ADS PubMed WorldCat Waite , L. , & Das , A . ( 2010 ). Families, social life, and well-being at older ages . Demography , 47 ( Suppl ), S87 – 109 . doi: 10.1353/dem.2010.0009 Google Scholar Crossref Search ADS PubMed WorldCat Weuve , J. , Tchetgen Tchetgen , E. J. , Glymour , M. M. , Beck , T. L. , Aggarwal , N. T. , Wilson , R. S. ,… Mendes de Leon , C. F . ( 2012 ). Accounting for bias due to selective attrition: The example of smoking and cognitive decline . Epidemiology , 23 , 119 – 128 . doi: 10.1097/EDE.0b013e318230e861 Google Scholar Crossref Search ADS PubMed WorldCat Weuve , J. , Proust-Lima , C. , Power , M. C. , Gross , A. L. , Hofer , S. M. , Thiébaut , R. ,… Dufouil , C .; MELODEM Initiative . ( 2015 ). Guidelines for reporting methodological challenges and evaluating potential bias in dementia research . Alzheimer’s and Dementia , 11 , 1098 – 1109 . doi: 10.1016/ j.jalz.2015.06.1885 Google Scholar Crossref Search ADS PubMed WorldCat Williams , S. R. , Pham-Kanter , G. , & Leitsch , S. A . ( 2009 ). Measures of chronic conditions and diseases associated with aging in the National Social Life, Health, and Aging Project . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64 , i67 – i75 . doi: 10.1093/geronb/gbn015 Google Scholar Crossref Search ADS WorldCat Wingfield , J. C. , Hegner , R. E. , Dufty Jr, A. M. , & Ball , G. F . ( 1990 ). The “challenge hypothesis”: Theoretical implications for patterns of testosterone secretion, mating systems, and breeding strategies . The American Naturalist , 136 , 829 – 846 . doi: 10.1086/285134 Google Scholar Crossref Search ADS WorldCat © The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

The Journals of Gerontology Series B: Psychological Sciences and Social SciencesOxford University Press

Published: Sep 15, 2019

There are no references for this article.