Marriage and health: His and hers.

The evidence base is narrower than it appears. The review cites 64 articles, but the physiological claims rest heavily on one cohort of ~90 newlywed couples from the first author's lab at Ohio State. That single sample generates citations across at least seven core findings — immune function, endocrine function, cardiovascular responses, the superiority of negative over positive behavior, gender differences in reactivity, persistence of physiological changes, and behavioral vs. self-report data. These appear as separate publications (Kiecolt-Glaser 1993, Malarkey 1994, Kiecolt-Glaser 1996), creating an impression of independent replication that does not exist. The total number of independent samples with concurrent behavioral coding and physiological measurement is about four, covering ~190 couples. The review does not flag this concentration.

The newlywed studies used stringent exclusion criteria and a hospital research unit that controlled diet, activity, and caffeine — genuine strengths for internal validity. But the resulting sample (healthy, childless, first-marriage newlyweds, overwhelmingly not maritally distressed) limits generalizability. Only 3% of newlyweds scored in the distressed range, yet the health claims are about distressed marriages.

Every interaction study measures physiology during or within minutes after a 10–40 minute lab discussion. The health claims require that these acute responses, repeated over years, cause disease. No cited study spans this extrapolation.

The review uses causal language — "produces," "direct effects," "enhanced" — to describe associations from studies where no variable was experimentally manipulated. Hostile behavior and physiological arousal were measured in the same minutes of the same interaction; the direction of influence is unknown. Physiological arousal could drive hostile behavior just as plausibly as the reverse.

The apparent convergence of evidence is inflated by treating reanalyses of the same sample as independent findings. Three of the six most-cited physiological studies draw on the same 90 newlywed couples. Six separate citations look like six independent findings. They are not.

The 25-fold association between unhappy marriages and major depression (Weissman 1987) comes from a cross-sectional study that cannot distinguish whether depression causes people to rate their marriages negatively or the reverse. It is unadjusted — controlling for other variables would likely reduce it. The review presents it as evidence of a "potent risk factor" without these qualifications.

Some reported effect sizes are genuinely large — hostile interactions and marital dissatisfaction accounted for half the variance in women's blood pressure in one study of 43 hypertensive couples. But with sample sizes of 26–90 couples, these point estimates carry wide uncertainty, and confidence intervals are never discussed.

Every physiological finding comes from couples arguing for 10 to 40 minutes in a laboratory with blood draws and blood pressure cuffs attached. The paper's central claim is that these acute measurements reveal how marriage affects long-term health. That requires an enormous inferential bridge — from a half-hour lab argument to years of cardiovascular disease, immune suppression, or wound healing — and no cited study spans it.

The claim that immune and endocrine changes are "large enough to be clinically significant" for wound healing and infectious disease bridges separate bodies of evidence. The wound healing studies measured responses to academic exam stress, not marital conflict. The infectious disease evidence comes from general interpersonal difficulties, not specifically marriage. No study traces the path from marital conflict through physiological change to a clinical health outcome in the same people.

The claim that women's greater physiological reactivity during marital conflict "reverses" the usual sex difference in stress reactivity compares across entirely different studies, populations, labs, and years. No study measured the same people's responses to both a marital and a non-marital stressor.

The theoretical explanation for gender differences — that women's more relationally interdependent self-construals make them more physiologically reactive — has never been tested by measuring self-construal and physiological reactivity in the same participants. The authors acknowledge this, but the framework receives several pages of development, giving it more weight than its empirical status warrants.

The abstract states that negative marital dimensions have "direct influences on cardiovascular, endocrine, immune, neurosensory, and other physiological mechanisms." The body text is considerably more cautious, acknowledging missing links, tenuous connections, and the need for studies that don't yet exist. The abstract omits these qualifications entirely.

The conclusion that negative behavior predicts physiology better than positive behavior may be partly an artifact of the research paradigm. Every interaction study asked couples to discuss a disagreement — a design that maximizes negative behavior and suppresses positive behavior. The authors note this but still present the asymmetry as a substantive finding about marriage.

The review treats acute physiological responses measured over minutes as proxies for the chronic burden needed to cause disease over years. This extrapolation is the foundation of every health claim, but it is never defended. The word "direct" in "direct influences on health mechanisms" conflates a verified short-term co-occurrence with an unverified long-term consequence.

The most obvious alternative explanation — that personality traits like neuroticism independently drive both negative marital behavior and physiological reactivity, without the behavior causing the physiology — is not addressed. The concurrent measurement design cannot distinguish whether hostile behavior causes arousal or arousal causes hostile behavior.

The paper discloses NIH funding and was published in Psychological Bulletin after a 15-month review process with revision. The search strategy is described in general terms, which is more transparent than many narrative reviews of the era. The paper explicitly acknowledges the missing causal chain, tenuous links between acute responses and health, underrepresentation of distressed couples, and the conflict-paradigm bias against detecting positive-behavior effects.

However, there is no conflict of interest statement. The first author is reviewing a body of literature in which her own lab produced a large share of the key findings, and this dual role is never acknowledged. No open data or reproducible search documentation is provided. The abstract and conclusion carry causal framing that the body text's own caveats do not support.

Kiecolt-Glaser is both the first author and the primary evidence source. The physiological claims rest overwhelmingly on studies from her own lab — at least six of the key marital interaction papers. Only a handful of independent labs contributed marital interaction data with physiological outcomes during this period. The review never flags this concentration.

Two high-impact statistics do heavy rhetorical work on thin foundations. The 250%/50% mortality differential between unmarried men and women comes from Ross 1990, itself a review-of-reviews drawing on data from decades with different gender roles. The 25-fold depression risk comes from a single brief epidemiological overview. Neither figure is qualified in the text.

Discrepant findings are handled asymmetrically. When a study finds no relationship between conflict behavior and physiology (Miller 1999), the explanation offered is that the conflict discussion was too short and the experimenter didn't help enough — essentially, that other labs didn't run the study as intensively as Kiecolt-Glaser's lab did.

The paper has no commercial conflicts — all funding is from NIH grants. The first author is reviewing her own laboratory's work. Seven of the paper's twenty core claims draw substantially on a single cohort of ~90 newlywed couples studied at Kiecolt-Glaser's Ohio State lab. These appear as separate citations but are reanalyses of the same data. The review covers 1990–1999, the exact decade in which her lab produced most of its marital interaction-physiology studies. The conceptual framework places "direct influences on biological systems" at the center — the specific contribution of her research program.