The Fear Everyone Shares
Ask anyone what they dread most about dying, and the answers cluster around the same themes: pain, loneliness, the terror of ceasing to exist. Movies depict death as anguish and medical systems treat it as the worst possible outcome. Terror management theory, one of psychology's most cited frameworks, holds that awareness of mortality drives everything from religious belief to consumer spending — the premise is so deeply embedded in Western culture that it barely registers as an assumption.
In 2017, a team from four universities stopped asking people what they imagined dying would feel like and instead measured what dying people actually wrote.
What 2,616 Blog Posts Revealed
Amelia Goranson and colleagues at the University of North Carolina at Chapel Hill identified 25 blogs written by people dying of cancer or amyotrophic lateral sclerosis. The selection criteria were strict: the author had to be the patient, not a caretaker; the blog had to span at least three months with at least ten posts; and the author had to have died during the course of writing it, with deaths confirmed by obituaries. The resulting dataset contained 2,616 posts spanning a median of 57 weeks before death.
For comparison, the team recruited 45 healthy participants on Amazon's Mechanical Turk and asked them to imagine they had been diagnosed with terminal cancer, then write a blog post keeping in mind they had only a few months left to live.
Using the Linguistic Inquiry and Word Count algorithm, a validated tool that quantifies emotional content by tallying positive and negative affect words, the researchers found that healthy people imagining death used significantly more negative-emotion words (2.25% of total words) than the patients who were actually dying (1.70%), a difference corresponding to a Cohen's d of 0.40. Independent human coders who rated a subset of posts blind to condition confirmed the pattern on the Positive and Negative Affect Schedule: patients scored 2.00 on negative affect versus 2.36 for the forecasters (d = 0.41) and 2.65 on positive affect versus 2.43 (d = 0.23). A replication sample of 75 additional coders produced even larger differences, with d = 0.91 for negative affect and d = 0.50 for positive affect.
The longitudinal finding was the one nobody anticipated: as patients drew closer to death, their use of positive-emotion words increased significantly (b = −0.14, p = .026), while their negative-emotion words did not increase at all (p = .839). The trend held after controlling for total word count and number of posts, ruling out the possibility that the act of writing more frequently was itself responsible for the shift.
The Condemned Confirmed It
A second study extended the analysis to 396 death-row inmates executed in Texas between 1982 and 2013, whose last statements were recorded by the state Department of Justice, and compared their words with imagined last words from 117 healthy participants and with 188 poems written by death-row inmates who were not facing imminent execution.
The inmates' final words contained far less negative affect (2.61% of total words) than either their own poetry (5.12%) or the forecasters' imagined statements (7.00%), and they also contained more positive-affect words (9.23%) than the forecasters produced (6.37%). The statistical effect for negative affect was substantial (η²p = .075).
What replaced the expected dread? Religion and social connection. Inmates facing execution spoke about God, family, and love at significantly higher rates than people imagining the same scenario, and mediation analyses showed these meaning-making categories partially explained why real last words were more positive and less negative than imagined ones.
The Strongest Counterargument
The most serious objection is self-presentation bias. Blog posts are public performances, last statements are delivered before witnesses, and dying people may craft their words for an audience while expressing peace they do not genuinely feel. A patient's final blog entry may be a gift to the readers who followed the journey, and an inmate's last words may be rehearsed for family members sitting in the gallery.
This critique is hard to dismiss entirely, but three pieces of evidence constrain it. First, LIWC captures word-level patterns that are difficult to consciously manipulate, and its positive-affect scores correlate with self-reported well-being even when writers are not deliberately communicating emotion. Second, the positivity trend in patients' blogs increased week by week over many months, and a deliberate performance for an audience would not be expected to grow steadily more positive as the writer's physical condition deteriorates. Third, Rabkin (2009) tracked terminal cancer patients through structured clinical interviews rather than public writing and found that positive mood scores remained equivalent to or higher than negative mood throughout the dying process, consistent with the blog data but collected through a method less susceptible to self-presentation.
What We Didn't Prove
Twenty-five blogs is not twenty-five thousand, and the bloggers were self-selected, likely more educated and reflective than average, with 80% being American. The death-row sample is overwhelmingly male, disproportionately composed of racial minorities, and drawn from a single state, so whether these findings extend to people dying of dementia, dying suddenly, or dying of old age in nursing homes remains genuinely unknown.
The study measures language, not internal emotional states — a person can write positive words while feeling despair, and a person can feel peace without ever writing the word "love." The trained coders improve on the automated algorithm by capturing context, but they are still interpreting text from the outside rather than measuring neurological experience.
A contrasting trajectory exists at longer time scales: the Rush Memory and Aging Project, which tracked 1,971 older adults for up to 22 years, found that purpose in life began declining 11 years before death and depression symptoms worsened 9 years out. The Goranson study captured only the final months, raising the possibility that dying becomes emotionally more positive very late, after years of gradual decline.
One number that deserves attention: approximately 2.5 million Americans die each year, and a substantial share of end-of-life spending is shaped by the assumption that dying is uniformly terrible. If that assumption is systematically wrong, it distorts advance directives, care decisions, and the willingness of physicians to initiate honest conversations about prognosis.
The Bottom Line
An analysis of terminally ill patients' blog posts and death-row inmates' final words found that people approaching death express more positive and less negative emotions than healthy people imagine. Patients' positivity increased as death grew closer. Independent coders, automated linguistic analysis, and replication samples all confirmed the pattern, which was driven in part by meaning-making through religion and social connection.
What You Can Do
If you or someone you care about is facing a terminal diagnosis, know that the emotional trajectory may differ from what you fear. Research suggests that focusing on relationships and meaning aligns with how people naturally adapt as death approaches. For advance care planning, base decisions on what matters most to you rather than on assumptions about how terrible the end will feel. Talk to people who have spent time with the dying. Read their own accounts. And when you catch yourself imagining the worst, recognize that your brain is doing exactly what the healthy forecasters in this study did: overestimating the dread and underestimating the grace.