ABSTRACT
Theoretical models of depression posit that, under stress, elevated trait rumination predicts more pronounced or prolonged negative affective and neuroendocrine responses, and that trait rumination hampers removing irrelevant negative information from working memory. We examined several gaps regarding these models in the context of lab-induced stress. Non-depressed undergraduates completed a rumination questionnaire and either a negative-evaluative Trier Social Stress Test (n = 55) or a non-evaluative control condition (n = 69), followed by a modified Sternberg affective working memory task assessing the extent to which irrelevant negative information can be emptied from working memory. We measured shame, negative and positive affect, and salivary cortisol four times. Multilevel growth curve models showed rumination and stress interactively predicted cortisol reactivity; however, opposite predictions, greater rumination was associated with blunted cortisol reactivity to stress. Elevated trait rumination interacted with stress to predict augmented shame reactivity. Rumination and stress did not significantly interact to predict working memory performance, but under control conditions, rumination predicted greater difficulty updating working memory. Results support a vulnerability-stress model of trait rumination with heightened shame reactivity and cortisol dysregulation rather than hyper-reactivity in non-depressed emerging adults, but we cannot provide evidence that working memory processes are critical immediately following acute stress.
Acknowledgement
The authors thank Northwestern undergraduate and post-baccalaureate research assistants John S. Gaffney, Grace C. Galloway, Amy Seungmin Lee, Claire E. Maby, Julia S. Retzky, Michael R. Sladek, and Kelsey A. Thompson for their many hours invested in data collection and thoughtful discussion. The authors thank Susan Mineka and Emma K. Adam for logistic support of the project.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Suzanne Vrshek-Schallhorn http://orcid.org/0000-0003-2825-200X
Notes
1 To ensure that the two start times did not influence results, we conducted an additional cortisol model with start time (early, 1300 h = 0, late, 1530 h = 1) as a covariate, including all interactions necessary to partial its influence out of the primary effect of interest, Rumination × Stress × Time2. The primary effect remained significant, t(233) = 2.63, p = 0.0090, the simple main effect of start time (representing the influence of start time on intercept/baseline cortisol for the control group, when effects of stress condition are partialed out) was significant and was consistent with expected lower values for later start times, b = −0.2281, SE(b) = 0.1123, t(233) = −2.03, p = 0.0434, and interactions involving start time (e.g., indicating stress reactivity) were not significant, ps ≥ 0.4713.
2 Results for the brooding subscale of rumination were similar but not identical to those for the full scale. Without covariates, brooding's relationship to reactivity under stress (i.e. Brooding Rumination × Stress × Time2) was significant for cortisol, p = .017, but not for shame, p = .2516, NA, p = .2923, or PA, p = .7650. When subclinical depression symptoms and its higher order interactions were covaried, this interaction approached significance for cortisol, p = .0593, and for shame, p = .0505, but remained non-significant for NA, p = .1516, and PA, p = .7438.
3 When covarying gender and necessary higher order interactions, the pattern of results for cortisol reactivity (Rumination × Stress × Quadratic time) remained the same, p = 0.0158, as did shame, p = 0.1650 (but p = 0.0253 with depression symptoms covaried). Similarly, the pattern of results remained the same when covarying past MDD and all necessary higher order interactions: Rumination predicted cortisol reactivity, p = 0.0017 (p = 0.0082 with DID covaried), but not shame reactivity, p = 0.1214 (p = 0.0157 with subclinical depression symptoms also covaried).