1938: An Anthropological Experiment Begins
In fall 1938, Harvard health director Arlie Bock raised an unusual question: medicine studies sickness; who studies thriving? Funded by the W.T. Grant Foundation, he selected 268 'promising' sophomores and began tracking them. Thus began the Harvard Study of Adult Development, popularly the Grant Study.
A year later, Sheldon Glueck started a parallel project tracking 456 14-year-old boys from inner-city Boston tenements. The two cohorts were later merged — a Harvard elite and a Boston working-class arm sharing one dataset for 80+ years, a rarity in social science.
The protocol: surveys every two years, medical records every five, in-person interviews every fifteen, autopsies thereafter. JFK was rumored to be an early subject; the investigators have never confirmed identities.
Four Directors, One 'Grand Question'
Directors: Arlie Bock (1938–1954) → Charles McArthur → George Vaillant (1972–2003) → Robert Waldinger (2003–present). Vaillant, a psychiatrist, distilled the archive into three books: Adaptation to Life (1977), Aging Well (2002), Triumphs of Experience (2012). Waldinger's 2015 TED talk 'What makes a good life?' has 40+ million views; his 2023 book The Good Life (with Marc Schulz) sealed the public framing.
Waldinger's one-line summary: 'Good relationships keep us happier and healthier. Period.'
Headline Finding: Beats Cholesterol
The most-cited result appears in Aging Well (2002) and Waldinger et al. (2007, Psychosomatic Medicine): relationship satisfaction at 50 predicted physical health at 80 better than cholesterol levels did. Men with rich close relationships at 50 functioned better cognitively and physically at 80; isolated men showed steeper cognitive decline and earlier chronic disease.
Vaillant put it more provocatively in Triumphs (2012): 'In 75 years of data, happiness comes down to five letters — love. That's it.' Sounds like poetry; the regression analyses support him.
Predictors of Wellbeing at 80
Relative strengths below synthesize Vaillant 2002/2012 and Waldinger 2007/2023. Read every row alongside the caveat that the sample is white male.
| Predictor of wellbeing at 80 | Evidence strength | Notes & caveats |
|---|---|---|
| Warm relationships (satisfaction at 50) | Very strong | Beats cholesterol for physical health (Waldinger 2007) |
| Mature defenses (sublimation, altruism) | Strong | Vaillant's hierarchy — detailed in companion piece #288 |
| Absence of alcohol dependence | Very strong (negative) | Single most destructive variable; predicts divorce, job loss, early death |
| Non-smoking, regular exercise | Strong | Standard medical variables; confirmed broadly outside Grant |
| Years of education | Moderate | Larger effect in Glueck (working-class) arm |
| Income, social class | Weak to moderate | Marginal effect after basic needs are met |
| IQ, college grades | Weak | Undergrad GPA was nearly uncorrelated with elder wellbeing |
| Childhood adversity | Moderate (buffered) | Effect is dampened when 'warm relationships' exist later |
The Grant data sit between Kahneman & Deaton's (2010) $75k threshold and Killingsworth's (2021) smoother curve — closer to 'money is a weak predictor once basic needs are met.'
Alcohol — the Single Most Destructive Variable
Vaillant emphasized in The Natural History of Alcoholism Revisited (1995) that alcohol dependence bundles most tightly with divorce, unemployment, depression, and premature death in the sample. Crucially, dependence more often preceded depression than followed it — drinking began, relationships fractured, depression arrived.
Given OECD-high alcohol consumption in many Asian work cultures, the finding lands hard: alcohol used as a 'social lubricant' may corrode the very relationships that protect long-term health.
Is Loneliness Like Smoking?
Waldinger emphasizes another finding: loneliness at 50 was associated with faster cognitive decline and elevated cardiovascular risk. Schulz, Waldinger et al. (2022, Frontiers in Psychology) reported social connection in the 8th–9th decades predicts cognitive vitality.
Consistent with Holt-Lunstad et al. (2010), whose meta-analysis equated social isolation's mortality risk to smoking 15 cigarettes/day. Causality runs both ways, but the Grant longitudinal arc leans toward loneliness → worse health more than the reverse.
Honest Limits: Whose Life Was Measured
The study's biggest weakness is its sample: Grant cohort = 268 white male Harvard sophomores (1938–1942); Glueck cohort = 456 white male Boston adolescents. No women, no non-white participants, no non-Western cultures. Vaillant himself flags this throughout Triumphs.
Do the conclusions replicate? Partially:
- Wisconsin Longitudinal Study (1957–): 10,000+ Wisconsin high-school grads — relationship quality strongly predicts elder wellbeing.
- Whitehall II (1985–): 10,000 British civil servants — social integration predicts mortality and cognitive decline, but workplace grade acts as a stronger variable than in Grant.
- MIDUS (1995–): US midlife longitudinal — relationship quality predicts cortisol, inflammation, immune markers.
- KLoSA (Korea, 2006–): 10,000+ Koreans aged 45+ — contact frequency with children/spouse/friends strongly predicts depression and physical health. Single-person elder households are most vulnerable.
Direction is consistent; effect sizes and mediators vary by culture and class. In Whitehall II, job control is a much larger variable than in Grant.
Korean Context: 33.4% Single-Person Households
Korea's 2023 census: 33.4% of households are single-person; over 2 million are elder-only. KLoSA 2020 reported one in five Korean elders 'often felt lonely in the past week.' Korea's OECD-leading elder suicide rate is not unrelated.
The Grant prescription doesn't translate verbatim, but the implication is clear: elder loneliness is a medical risk factor, in the same category as smoking or hypertension. As Waldinger says, the 'relationship pension' must start contributions long before 50.
Conclusion: 'Love' as a Regression Coefficient
Vaillant's five-letter summary isn't poetry — it's a coefficient. Ask a 50-year-old, 'Is there someone you can call when things truly go wrong?' Those who answer yes are, statistically, healthier and happier at 80. Even after acknowledging sample limits, this conclusion has survived in six-plus independent longitudinal studies.
One call today. One meal together. Remembering one person's name. The variable Grant Study measured for 85 years turned out to be small and daily.
