Abstract
The relational turbulence model argues that periods of transition in romantic relationships are ripe for upheaval due to heightened relational uncertainty and interference from partners (Solomon & Knobloch, 2004). This research explores communicative and physiological manifestations of relational turbulence during the transition to the empty-nest phase of marriage. Participants completed surveys about their marriage, engaged in a videotaped conflict interaction, and provided saliva samples that were tested for cortisol levels. Multi-level modeling results indicated that relational uncertainty predicts avoidant conflict behaviors but not approach conflict behaviors, and interference from partners predicts indirectness, topic avoidance, withdrawal, and criticism in conflict interaction between empty-nest spouses. Results also revealed that indirectness and withdrawal were positively associated with increased cortisol following conflict. In addition, indirectness, topic avoidance, and withdrawal during conflict interaction corresponded with a more rapid decrease of cortisol following the episode, whereas criticism and demandingness were associated with an increase in cortisol.
Notes
[1] Although some medical conditions (e.g., cancer, diabetes, endocrine disease, hypertension) and medications (e.g., steroidal medications, hormonal birth control, antidepressants) have been shown to interfere with cortisol levels, participants were not pre-screened for these conditions for two reasons. First, given the challenges already inherent in obtaining a sample of married couples who had recently entered the empty-nest phase of marriage, adding extensive exclusion criteria based on health conditions and medications would have been too prohibitive to recruitment. In a study that examined cortisol in college undergraduates, extensive pre-screening resulted in the exclusion of approximately 50% of interested participants (Floyd et al., Citation2007), which likely would have been much greater in a population of aging empty-nest couples. Second, if participants did have a medical condition or were taking medications that affect cortisol, their cortisol samples would not be completely invalid, they would simply add to measurement error (e.g., Floyd & Riforgiate, Citation2008). Thus, pre-screening for the various health conditions and medications that can interfere with cortisol seemed too restrictive given the relatively modest impact these variables might have on the data.
[2] The researcher attended specialized training for analyzing salivary cortisol at the Salimetrics Laboratory in State College, PA. The analyses were conducted in an endocrinology laboratory at the researcher’s institution under the supervision of a faculty member with expertise in conducting cortisol analysis.
[3] We conducted a series of post-hoc analyses to assess the indirect effects of relational uncertainty and interference from partners on cortisol increases and decreases. We constructed structural equation models in which either relational uncertainty or interference from partners predicted one of the conflict message features, which in turn predicted either cortisol increase or cortisol decrease. To evaluate the indirect effects in our models, we conducted bootstrap analysis with 2,000 bootstrap samples and 95% bias-corrected confidence intervals. Given the nonindependence in our data, we ran the models separately for males and females. Results indicated that none of the indirect effects were statistically significant. A few of the tests approached significance. Specifically, in the models where criticism was the mediator of effects on cortisol decrease, the standardized, indirect effects for partner interference approached significance for both males (β = 0.29, p = 0.06) and females (β = 0.28, p = 0.07). In addition, in the models for females where topic avoidance was the mediator of effects on cortisol increase, the standardized, indirect effects for relational uncertainty (β = 0.09, p = 0.13) and partner interference (β = 0.11, p = 0.13) were marginally nonsignificant. Finally, in the male model where topic avoidance mediated effects on cortisol decrease, the standardized, indirect effect for partner interference (β = 0.13, p = 0.12) also approached significance. Given the small sample size in this study, we are reluctant to dismiss the possibility of indirect effects for relational uncertainty and interference from partners out of hand. Future studies should attempt to retest these associations in a larger sample that has sufficient power to detect small and medium sized effects.