The result sounds made for a library poster. Among 3,635 Americans aged 50 and older, people who read books were less likely to die during the following 12 years than people who read none. Those reporting more than three and a half hours of book reading a week had a 23 percent lower mortality hazard after adjustment for age, sex, race, wealth, education, health and other measured differences.
A second number gave the finding its staying power. At the point when 20 percent of each group had died, book readers reached that survival milestone 23 months later than nonreaders. But that 23-month comparison came from the raw, unadjusted curves. It is not evidence that picking up a novel adds nearly two years to a person’s life.
This is not medical advice. It is a careful reading of an observational study and what it does, and does not, show.
This is one study, not settled consensus. The association persisted after extensive statistical adjustment, but the researchers did not randomly assign people to read. Book reading may have contributed to the survival difference, may have marked other advantages that the model did not capture, or both.
The study began with a one-week snapshot
The paper appeared in Social Science & Medicine in 2016. Lead author Avni Bavishi, Martin Slade and Becca Levy of Yale used data from the US Health and Retirement Study, a nationally representative longitudinal survey 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 team divided book reading into three groups: none, up to three and a half hours a week, and more than three and a half hours.
The final cohort contributed 34,496 person-years of observation. Follow-up averaged 9.49 years and extended to 12 years for some participants. During that period, 27.4 percent of the full sample died. The long-running national cohort is a stronger base than a casual readership survey, but the exposure measure remained a single self-reported week at baseline.
The 23 percent and 23 months answer different questions
During follow-up, 33 percent of non-book readers died, compared with 27 percent of book readers, an unadjusted difference of six percentage points. In the statistical models, people reading up to three and a half hours weekly had a 17 percent lower mortality hazard than nonreaders. Those above three and a half hours had a 23 percent lower hazard.
When the two book-reading groups were combined, readers had a 20 percent lower adjusted mortality hazard over the follow-up period. The model included age, sex, race, education, wealth, marital and employment status, self-rated health, eyesight, depression and several diagnosed conditions. The pattern also remained when the sample was divided by sex, education, wealth and baseline health.
The survival-time comparison was different. The unadjusted curves reached 80 percent survival after 85 months among nonreaders and after 108 months among book readers. That is the 23-month difference in the headline. In the fully adjusted curves, the gap at the same survival point was four months, not 23.
Both findings are real features of the analysis, but they should not be fused into one causal claim. The 23 percent is the adjusted relative hazard for the highest reading group. The 23 months is a descriptive gap between unadjusted groups. What persisted after adjustment was the broader association between book reading and lower mortality, not an adjusted promise of 23 extra months.
Books produced a stronger pattern than periodicals
The comparison with newspapers and magazines was one of the paper’s most interesting choices. Periodical reading did not show the same consistent dose-response pattern. The highest periodical-reading group had a weaker protective association, while the lower group did not. At both levels, book reading was significantly more strongly associated with survival than periodical reading.
That does not mean newspapers and magazines were useless, nor that every book was beneficial. The hierarchy belongs to this model, this outcome and this older US cohort. The study measured hours, not subject matter, reading difficulty, literary quality or whether a person found the experience engaging.
It also could not separate print from ebooks, fiction from nonfiction or reading with the eyes from listening to an audiobook. “Book reading” was a broad category. The paper’s discussion noted that fiction accounted for much of book reading in the wider population, but the participant survey itself did not identify what each person read.
Cognition was a possible pathway, not a proven mechanism
The Yale team proposed that books encourage “deep reading”, meaning sustained attention to characters, arguments and relationships across a long text. Following a book can require readers to hold earlier information in mind, connect it with later passages and remain inside one structure for longer than a typical periodical item.
The researchers tested cognition as a statistical mediator. Changes in cognitive scores appeared to carry part of the association between book reading and survival. They also discussed empathy, social perception and emotional intelligence as possible routes, particularly for fiction.
Those explanations remain hypotheses. A mediation analysis can show that a proposed pathway fits the observed data, but it cannot establish a biological chain in which reading changed cognition and cognition then extended life. The study did not measure such a chain experimentally.
A related ScienceBlog report on mentally engaging sedentary activities described a later study linking reading and similar activities with cognitive performance in 397 older adults. That is compatible with cognitive engagement mattering, but it is another observational association, not a demonstration that books prevent death.
Adjustment cannot erase every difference between readers and nonreaders
Book readers in the Yale sample were more likely to be women, better educated and wealthier. The models adjusted for those differences and a substantial set of health measures. That is careful epidemiology, but statistical adjustment can account only for variables that were measured, measured accurately and modeled appropriately.
Reverse causation remains plausible. Health can shape reading as readily as reading might shape health. Someone experiencing subtle cognitive, visual or physical decline may find sustained reading harder before that change appears in a diagnosis or self-rated health score. Personality, access to books, social routines and reliable leisure time could influence both reading and later survival.
The baseline measure creates another limit. Participants recalled the previous week’s reading, and the analysis used that brief report as a marker of a habit over many years. It may have represented a durable routine for many people. For others, one unusually busy or quiet week could have placed them in the wrong group.
These problems do not make the association disappear. They explain why “was associated with lower mortality” is accurate and “made people live longer” is not.
Later evidence makes the longevity story less tidy
A useful comparison arrived in 2023. Researchers followed older adults in 15 European countries for six years. In that prospective study, frequent reading was associated with several measures of well-being, daily functioning and cognitive impairment, but not with all-cause mortality after the study’s corrections.
The European analysis grouped books, newspapers and magazines together, so it was not a direct replication of Yale’s book-specific test. Its null mortality result still gives a good reason not to turn one American cohort into a universal longevity rule. A 2022 systematic review of longitudinal research likewise found generally favorable associations between leisure activities and later cognitive outcomes, while stressing variation across activities, populations and measures.
Questions about personal health and longevity belong with a qualified clinician, not with a correlation about leisure reading. Nothing in this study turns books into a treatment or substitutes them for established medical care.
The clean conclusion is narrower and more defensible. In one large cohort, older adults who reported reading books had a lower mortality hazard over as many as 12 years, and measured differences in wealth, education, health and gender did not explain the whole association. Whether books helped create that difference remains unresolved. The readers had more time, but the study cannot tell us how much of that time came from reading.