Stronger ties and a sense of purpose went with longer lives, while men's close friendships shrank

The size of the gap

In 2010 Holt-Lunstad, Smith and Layton pooled 148 studies that measured people's social ties and then counted deaths over the years that followed¹. Together the studies covered 308,849 people, followed for about 7.5 years on average¹. Those with stronger social relationships had 50% higher odds of still being alive at the end¹.

That is odds of survival, not years of life. 50% higher odds does not mean living 50% longer¹. It is also a pool of observational studies. Illness can shrink a person's circle as well as the other way round, so a man who stops going out because he is sick shows up as someone with weak ties who then died¹. The authors note that most studies followed community samples, and that health at the start did not change the result¹. It is large, and it held across almost every way of measuring a tie.

How the ties were measured mattered. Measures that combined several kinds of tie showed the biggest gap, 91% higher odds across 30 results¹. Taking part in social life showed 52%, the size or density of a person's network 45%, feeling supported 35% and being married 33%¹. Two measures made no significant difference, simply not living alone and the support people said they had received¹. Sharing a roof with someone is not the same as having people, and the numbers treat it that way.

Purpose ran the same way

Ties are one thing a person has. A reason to get up in the morning is another, and it turned up in the death records too. In 2006 the US Health and Retirement Study asked 6,985 Americans over 50 to rate their sense of purpose in life, as the average of 7 statements scored 1 to 6². The researchers then tracked deaths to 2010, and 776 people died in that time².

Against those with the top score of 6, those scoring below 3 had 2.43 times the risk of dying². 83 of those 300 people died, against 17 of the 422 with the top score². The risk climbed as purpose fell. Those in the low and middle bands had about 1.7 times the risk, and only the band just under the top showed no clear difference².

The model allowed for a long list of things that could have explained the gap away. Age, sex, education, smoking, exercise, alcohol, weight, chronic illness, depression, anxiety, optimism and social participation were all in it². That last one matters here. Purpose kept its link with death even after the model took account of how much people took part in social life². Heart and circulatory deaths were 2.66 times as likely in the lowest group, against everyone scoring 5 or more². Cancer deaths showed no link².

The obvious worry is the same one as with social ties. People who are already ill may feel less purpose. Leaving out deaths in the first year only weakened the result modestly, to 2.24². The limitations section of the paper says the first 2 years, but its methods and table say the first year². Purpose was measured once, though, so the study cannot show that raising it would add years². And 1,142 people who did not return the questionnaire, or answered fewer than 3 of its 7 questions, were left out, so the authors cannot rule out selection bias².

A second US study points the same way across a wider age range. Hill and Turiano followed 6,163 adults aged 20 to 75 for 14 years³. Each step up in purpose, one standard deviation, went with a 15% lower risk of death, and the link did not differ by age³. Its purpose measure had only 3 questions, with a reliability of 0.36, and 91% of the sample was white³. Two studies with different flaws landing on the same side count for more than either one alone.

Fewer close friends, most of all for men

If ties and purpose go with longer lives, the next question is how many people have them. For American men the answer has got worse. In 1990, 55% of men and 41% of women said they had six or more close friends, not counting relatives⁵. In a 2021 survey of 2,019 adults it was 27% of men and 24% of women⁴⁵. The share of men with no close friends at all went from 3% to 15%, and of women from 2% to 10%⁵.

Women's circles shrank too, but less. The gap shows more clearly in who gets support. In the week before the 2021 survey, 41% of women had received emotional support from a friend, against 21% of men⁴. Among men with female friends it was 28%, and among men without, 16%⁴.

Some caution is due. The 1990 survey was by telephone and the 2021 survey was online, and the 2021 report notes that people may report more friends to a live interviewer⁴. Each person also decided for themselves what counts as a close friend⁴.

A different kind of data does not rely on anyone counting friends. The American Time Use Survey covered 219,368 people from 2003 to 2020, and found time alone rose and time with friends fell for both sexes⁶. Women spent 57 minutes a day with friends in 2003, 33 in 2019 and 18 in 2020, and men 63, 36 and 21⁶. For people aged 15 to 24 it went from 153 minutes to 67, then 43⁶.

The survey counts only time spent in person, and it collected no data for 52 days in 2020⁶. So the 2020 figures say as much about the pandemic as about friendship⁶. The fall before that does not. Young people already spent less than half as much time with friends in 2019 as in 2003⁶.

What was tested against loneliness

So what helps a lonely person? The obvious answer is more contact, and the trials do not back it. In 2011 Masi, Chen, Hawkley and Cacioppo pooled 50 studies of loneliness programs, 20 of them randomised trials⁷. They grouped the trials by what each program tried to change.

Programs that worked on unhelpful thoughts about other people cut loneliness most, with an effect size of 0.598 across 4 trials⁷. Adding social support helped a little, at 0.162 across 12 trials⁷. Programs that offered more chances to meet people, or taught social skills, made no clear difference. In the skills trials loneliness rose very slightly⁷.

This lines up with the survival data. Not living alone made no significant difference to survival in the 148 studies¹, and more chances to meet people made no clear difference to loneliness here⁷. Being near people is not the thing that counts. How a person reads the people around him seems closer to it. A man who expects to be turned away can sit in a full room and leave with nothing.

The overall numbers are modest. Across all 20 randomised trials the effect was 0.198⁷. On the UCLA Loneliness Scale that is about 1.6 points, from 41.17 down to 39.58⁷. Studies without a randomised control group showed larger effects, which is a reason to trust the trials over them⁷. The strongest group rests on only 4 trials, and the weakest on 2 each⁷. 11 of the 20 randomised trials were with people aged 60 or older, and 7 included only women⁷. The evidence comes mostly from older adults and women, not from the men in the friendship surveys.

A later review found psychological therapy helped, but not the thinking approach above the rest. In 2021 Hickin and colleagues pooled 28 randomised trials of psychological therapy for loneliness, with cognitive behaviour therapy the most common, and found an effect of 0.43⁸. Programs built on cognitive behaviour therapy did no better than the others, and social skills training came out close to it, unlike in the 2011 review⁸. Some of those trials were small, and some lost up to 45% of their participants⁸. It is the best evidence we found, and it is thinner than the size of the problem deserves.

How they found it

• The 2010 review pooled 148 studies that measured people's social ties and then counted deaths over the years that followed. Each bar pools the results that used one kind of measure, and the overall figure was 50%.
• Purpose was the average of 7 statements rated 1 to 6, from a standard wellbeing scale, asked of 6,985 Americans over 50 in 2006. Deaths were tracked to 2010.
• The purpose model allowed for age, sex, education, race, marital status, smoking, exercise, alcohol, weight, daily functioning and chronic illness, and for depression, anxiety, hostility, optimism, mood and social participation.
• The 2021 friendship survey asked 2,019 US adults in May 2021 and set the answers against a Gallup survey from 1990. The American Time Use Survey covered 219,368 people from 2003 to 2020.
• The 2011 loneliness review pooled 50 studies and each study design separately. Across all 20 randomised trials the effect was 0.198, about 1.6 points on the UCLA Loneliness Scale.
• A 2021 review of 28 randomised trials of psychological therapy for loneliness, with cognitive behaviour therapy the most common, found an effect of 0.43.

What it cannot tell you

• Odds of survival, not years of life. 50% higher odds does not mean living 50% longer.
• The survival and purpose studies are observational. Illness can shrink a person's circle or sense of purpose as well as the other way round, and purpose was measured only once. Neither measures addiction or recovery.
• 1,142 people who did not return the purpose questionnaire, or answered fewer than 3 of its questions, were left out. In the 14-year purpose study, the 3-question measure had a reliability of 0.36, and 91% of the sample was white.
• The 1990 friendship survey was by telephone and the 2021 survey online, and each person decided what counts as a close friend. The time-use survey counts only time in person and collected no data for 52 days in 2020.
• Only 4 loneliness trials in the strongest group, and 2 each in the weakest. 11 of the 20 randomised trials were with people aged 60 or older, and 7 included only women.
• Some trials in the 2021 loneliness review were small, and some lost up to 45% of their participants.

In P-Minus

Lonely is one of the eight feelings you can tag when you log an urge in P-Minus, and when you close each day. P-Minus also asks you to name one person to reach when your plan is not enough, and to write now what you will say. The help button then shows that card.

Sources

1. Holt-Lunstad J, Smith TB, Layton JB. Social relationships and mortality risk: a meta-analytic review. PLoS Med. 2010;7(7):e1000316.
2. Alimujiang A, Wiensch A, Boss J, Fleischer NL, Mondul AM, McLean K, et al. Association between life purpose and mortality among US adults older than 50 years. JAMA Netw Open. 2019;2(5):e194270.
3. Hill PL, Turiano NA. Purpose in life as a predictor of mortality across adulthood. Psychol Sci. 2014;25(7):1482-6.
4. Cox DA. The state of American friendship: change, challenges, and loss. Washington: Survey Center on American Life; 2021.
5. Cox DA. Men's social circles are shrinking. Survey Center on American Life; 2021.
6. Kannan VD, Veazie PJ. US trends in social isolation, social engagement, and companionship, nationally and by age, sex, race/ethnicity, family income, and work hours, 2003-2020. SSM Popul Health. 2023;21:101331.
7. Masi CM, Chen HY, Hawkley LC, Cacioppo JT. A meta-analysis of interventions to reduce loneliness. Pers Soc Psychol Rev. 2011;15(3):219-66.
8. Hickin N, Käll A, Shafran R, Sutcliffe S, Manzotti G, Langan D. The effectiveness of psychological interventions for loneliness: a systematic review and meta-analysis. Clin Psychol Rev. 2021;88:102066.

Sources checked October 2026.