Gettler · Proceedings of the National Academy of Sciences of the United States of America 2011 · prospective cohort study · n=624

Longitudinal evidence that fatherhood decreases testosterone in human males.

Cited 568 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective longitudinal cohort study

PubMed 21911391 · doi:10.1073/pnas.1105403108 · record verified 2026-08-26

What was done

Researchers tracked a representative cohort of 624 single, nonfather men in the Philippines (mean age 21.5 ± 0.3 years at baseline in 2005) over a 4.5-year follow-up period. Salivary waking and evening testosterone levels were measured at baseline and follow-up to test whether initial testosterone levels predicted becoming a partnered father, whether becoming a father was associated with longitudinal testosterone changes compared to remaining single nonfathers, and whether daily childcare interaction related to testosterone concentrations.

What was found

Men with high waking testosterone at baseline were more likely to become partnered fathers by the 4.5-year follow-up (P < 0.05). Men who transitioned to partnered fatherhood experienced significant median declines in waking (-26%) and evening (-34%) testosterone, which were significantly greater than declines in single nonfathers (P < 0.001). Fathers reporting 3 hours or more of daily childcare had significantly lower testosterone at follow-up than uninvolved fathers (P < 0.05).

Why it matters

These longitudinal findings help resolve whether lower testosterone in fathers is a cause or consequence of fatherhood, demonstrating that high testosterone predicts mating success but subsequently declines with parenting and active caregiving, consistent with evolutionary trade-offs between mating and parental investment.

Limits

The study was conducted in a single population of young men in the Philippines, which may limit generalizability across different cultures, older ages, or diverse socioeconomic settings. Absolute testosterone concentrations, effect estimates with confidence intervals, and confounding variables like sleep or health changes were not reported in the abstract, and childcare duration was self-reported.

Cited by