The finding sounds almost too perfectly made for people who love books. Among 3,635 Americans over 50, those who read books were less likely to die during the following 12 years than people who did not. Reading newspapers and magazines did not show the same consistent pattern.
There was also a striking gap in the raw survival curves: at the point when 20 per cent of each group had died, book readers reached that point 23 months later than non-readers.
That is the source of the “nearly two years” claim. It is important, however, that the 23-month figure was unadjusted. It should not be read as a promise that books add two years to a person’s life. This is a report on observational research, not health advice.
This is one study, not settled consensus. Its value lies in the unusual comparison between books and periodicals, and in how carefully the researchers tried to test that association. Its limits lie in everything an observational study cannot fully separate.
The study began with one ordinary question
The paper appeared in Social Science & Medicine in 2016. Avni Bavishi, Martin Slade and Becca Levy of the Yale School of Public Health used data from the US Health and Retirement Study, a long-running national study of adults over 50.
Participants were asked how many hours they had spent reading books during the previous week. They separately reported time spent reading newspapers and magazines. The researchers placed book reading into three groups: none, up to three and a half hours per week, and more than three and a half hours.
The final cohort contained 3,635 people and 34,496 person-years of observation. Average follow-up was 9.49 years, with survival information available for up to 12 years. During that period, 27.4 per cent of the sample died.
The underlying Health and Retirement Study is designed to represent older Americans and has interviewed participants every two years since 1992. That gives the Yale analysis a stronger base than a casual online poll. It does not turn the reading question into an experiment.
Books produced the clearest association
After accounting for age, sex, race, education, wealth, marital and employment status, self-rated health, eyesight, depression and several diagnosed conditions, people in the middle book-reading group had a 17 per cent lower mortality hazard than non-book readers. Those in the highest group had a 23 per cent lower hazard.
Combining the book readers, the adjusted result was a 20 per cent lower risk of death during the follow-up period. The association also remained when the researchers divided the sample by sex, education, wealth and baseline health.
Periodicals were different, but the title-sized version needs a small correction. Newspapers and magazines were not wholly unrelated to survival. The strongest periodical readers showed a weaker protective association, while the lower reading group did not. Book reading was significantly more strongly associated with survival at both reading levels.
So “books, not newspapers or magazines” captures the contrast, but not a literal zero-versus-something result. The paper found a more consistent and substantially stronger pattern for books.
The two-year number is real, but it is the raw number
The study compared the unadjusted survival curves at the point when 20 per cent of a group had died. Non-book readers reached that point after 85 months. Book readers reached it after 108 months. The difference was 23 months.
Once the full set of measured differences between readers and non-readers was included, the survival association remained statistically strong. But the adjusted gap at that same point was four months, not 23. The adjusted model is also where the researchers obtained the overall 20 per cent lower mortality hazard.
Those figures answer different questions. The 23 months describes what happened in the two groups before adjustment. The hazard ratio estimates the association after measured differences were taken into account. Combining the adjusted controls with the unadjusted two-year figure makes the result sound more certain than the paper supports.
I think the finding becomes more interesting, not less, when those numbers are kept separate. A large raw difference tells us that book reading marks a distinctive kind of life. The adjusted result suggests that familiar advantages such as wealth and education did not explain all of it.
The proposed explanation was “deep reading”
The Yale team suggested that books demand a slower and more sustained form of attention. A reader has to remember characters or arguments, connect distant parts of a text and remain inside the same structure for longer than a typical newspaper item.
The researchers also tested cognition as a possible mediator. Their model suggested that changes in cognitive scores statistically carried the association between book reading and survival. They proposed a second route through empathy, social perception and emotional intelligence, especially because fiction made up much of book reading in the wider population.
These are plausible explanations, but the study did not observe a biological mechanism and did not assign people to read books. A mediation model can organise a possible pathway in the data. It cannot establish that this was the process extending anyone’s life.
The paper also could not distinguish fiction from non-fiction, print from ebooks, or reading from listening to an audiobook. “Book reading” was a broad self-reported category built from one week at baseline.
Statistical adjustment cannot remove every difference
Book readers in the sample were more likely to be women, better educated and wealthier. The models adjusted for those differences, along with a substantial list of health measures. That is careful work, but adjustment is limited to factors that were measured well enough to include.
Health can shape reading as readily as reading might shape health. Someone experiencing subtle decline may find sustained reading harder before that decline appears in a diagnosis or a self-rating. Personality, access to books, social routine and the stability of leisure time may also influence both reading and later outcomes.
There is also a measurement problem hidden by the long follow-up. Participants reported reading during the previous week, then survival was followed for as long as 12 years. The analysis treats that brief snapshot as a useful marker of a more durable habit. For many people it probably was. For some, it will not have been.
None of this makes the association disappear. It explains why “linked to living longer” is more accurate than “made people live longer”.
Later evidence makes the story less tidy
A useful comparison arrived in 2023. Researchers analysed older adults from 15 European countries, including Denmark and Sweden, using the Survey of Health, Ageing and Retirement in Europe. In that six-year prospective study, frequent reading was associated with several measures of wellbeing, daily functioning and cognitive impairment.
It was not associated with all-cause mortality after the study’s corrections. That later analysis grouped books, magazines and newspapers together, so it was not a direct replication of Yale’s book-specific test. Still, its null mortality finding is a reason not to turn one American cohort into a general longevity rule.
A 2022 systematic review of longitudinal research also found that leisure activities were generally associated with better cognitive outcomes in older adults, while stressing that results varied by activity, population and cognitive measure. Most of the evidence remained observational.
The Yale finding therefore sits in an intriguing but incomplete evidence base. Books may demand a kind of sustained engagement that periodicals do not. Book reading may also identify people whose health, resources and routines already make that engagement possible. Both can be true at once.
The clean claim is not that a book grants extra time. It is that, in one carefully analysed cohort, the people who kept returning to books had more of it.