Older adults who felt lonelier saw their memory and thinking abilities decline faster, over as long as a decade. Researchers saw this in 3,082 participants in aging studies at Rush University Medical Center in Chicago. Seeing family and friends regularly didn’t explain it. Comparing people of the same race, the lonelier ones still declined faster after the team counted how many children, relatives and friends each person saw at least once a month.

The study, published in The Journals of Gerontology: Series B in August 2025, was led by Qing Lin and senior author Cassandra Morrison, an assistant professor of psychology at Carleton University in Ottawa, with colleagues at McGill University, Belmont University and Rush.

Morrison told ScienceBlog the results don’t show which came first. “Because we averaged loneliness scores across participants’ visits, we cannot confidently establish that loneliness preceded cognitive decline,” she wrote.

Eighteen tests, once a year

The participants came from four studies run by the Rush Alzheimer’s Disease Center. One of them, the Memory and Aging Project, recruits mainly from retirement communities across northeastern Illinois. Everyone included was at least 55 at the start, had at least two cognitive assessments, and identified as Black or White.

Each year they sat a battery of 18 cognitive tests covering five areas:

  • episodic memory (memory for events)
  • semantic memory (memory for facts and words)
  • working memory (holding something in mind for a few seconds)
  • perceptual speed (how quickly people take in simple information)
  • visuospatial ability (how well they judge shapes and space)

Loneliness was measured at each visit with a modified version of the de Jong-Gierveld emotional loneliness subscale. Morrison described its items to ScienceBlog as capturing experiences “such as emptiness, feeling abandoned, and missing close relationships.” The researchers also rolled the scores into one overall measure. People were followed for about six years on average, and the analysis stopped at 10 years.

Across the whole sample, higher loneliness went with lower scores overall and with a steeper decline over time on every one of the six measures. The models accounted for age, sex, years of education, and whether a person already had mild cognitive impairment or dementia at the first visit.

Feeling lonely and seeing people

The phrase in the title about friends and relatives refers to one of the study’s extra checks. The Rush studies measure social network size by asking how many children, close relatives and close friends a person has, and how many of them they see at least once a month.

When the researchers added network size to their comparisons of more and less lonely groups, the results “remained unchanged in terms of effect size and significance.” The paper’s supplement defines the measure the same way, as the number of children, family members and friends seen at least once a month.

The authors connect this to a distinction researchers draw between loneliness, a feeling that one’s relationships fall short of what one wants, and social isolation, an objective shortage of contact. A person can feel lonely despite frequent visitors, or feel well connected with only a few close ties.

ScienceBlog covered the other side of that distinction in August, in a report on a 30,421-person analysis that found social isolation linked to poorer cognitive trajectories even when loneliness was held to its expected course. Taken together, the two studies suggest that how connected people feel and how connected their lives are in practice carry somewhat separate signals.

A score that blends every visit

Each person’s loneliness score was an average of every visit. That means how someone felt near the end of the study, possibly after their memory had already started to slip, still counted toward whether they were grouped as lonely.

Morrison acknowledged the problem directly. Adjusting for diagnosis “does help control for cognitive status at the beginning of the study,” she wrote, but “this does not eliminate the possibility that preclinical cognitive changes influenced loneliness throughout follow-up.”

“It is possible that loneliness contributes to cognitive decline, but it is also possible that early cognitive changes affect social interactions, relationships, and feelings of loneliness,” she added. She told ScienceBlog that the findings can be read as evidence that loneliness and cognitive decline “are linked and influence one another.”

The groups also differed in cognitive health. More than a third of the lonelier White participants, 35 percent, were classed as having mild cognitive impairment or dementia, compared with about a fifth of the less lonely White participants. The researchers reran the analysis on the 2,264 participants classed as cognitively healthy and reported the same results.

The authors list other gaps themselves. They did not account for socioeconomic status, which is tied to both loneliness and cognition, or for hearing and vision loss, which can cut people off socially and also affect thinking. The paper also notes that the analysis was not preregistered, meaning the plan was not filed publicly before the data were examined.

Stronger among White participants

The study’s most unexpected result concerned race. The authors had predicted that loneliness would be more harmful for Black participants, who face greater exposure to many risk factors for isolation. They concluded the reverse: that the link was stronger among White participants.

The picture underneath is mixed. In the paper’s estimates, the difference in rate of decline between lonelier and less lonely White participants was larger than among Black participants on global cognition, semantic memory and working memory, though no model tests that difference directly.

A separate test that looked only at overall score levels found loneliness weighed more heavily for White participants on four measures. Even so, lonelier Black participants declined faster than less lonely Black participants on all six measures, while among White participants the difference was not significant on perceptual speed or visuospatial ability.

A third model, which compared the races without regard to loneliness, found that White participants declined faster than Black participants on all six measures, a point Morrison stressed in her reply. However, across the four groups, less lonely White participants and lonelier Black participants did not differ significantly on global cognition, episodic memory or semantic memory. The authors list the steeper White decline as a limitation of the race comparison, and say the effects of loneliness may take longer than the study’s 10 years to show among Black participants, whose decline was slower.

The White groups were also older at the start, so the authors reran the comparison on a subset matched for age. There, the episodic memory differences lost significance, which they say suggests age played a part, and lonelier Black participants declined faster than less lonely White participants on global cognition and semantic memory. The authors offer possible explanations for the race result, including extended support networks and coping strategies in Black communities, but they did not test them.

A different answer on Chicago’s South Side

An earlier Rush study drew a different picture. In the Chicago Health and Aging Project, a 2023 analysis of 7,696 older residents of four adjacent South Side neighborhoods, 64 percent of them Black, the authors reported that lonely participants declined faster in both racial groups.

Their own numbers lean the other way from the Rush finding. Lonely and non-lonely Black participants declined at 0.075 and 0.060 standard deviation units per year on global cognition, against 0.075 and 0.071 for White participants. The paper’s own supplementary figure describes the White rates as similar.

In models split by race and sex, the paper reports that the association was statistically significant among Black men and women but not among White men and women. The formal test for a difference by race was not significant.

Asked why the two studies differ, Morrison pointed to differences in measurement and analysis. The South Side study used a single yes or no question, whether a person had felt lonely much of the time in the past week, while her team used a multi-item scale. “Our measure may thus capture a broader range of emotional experiences associated with loneliness which may provide a more nuanced assessment of individual differences than a single question,” she wrote.

Her team also used more cognitive tests across more domains. She was careful not to claim the question was settled, writing that “we cannot determine which factors explain the discrepancy without further research using comparable measures across cohorts.”

A European study that found no faster decline

Not every large study finds that loneliness speeds decline at all. A 2026 analysis of 10,217 adults aged 65 to 94 in the Survey of Health, Ageing and Retirement in Europe found that people who reported high loneliness at the start had poorer memory. Their memory did not, however, decline any faster between 2012 and 2019.

Its design was narrower in some ways and stricter in others. It tested only word recall and measured loneliness once, at the start, with three questions. Unlike the Rush analysis, it adjusted for depression.

The study left out 13,675 people, among them those under 65, those who missed a wave or lacked a starting loneliness score, and those with a history of dementia or trouble with daily activities. The people left out had more health conditions, worse memory and more loneliness than those it kept, though the authors say age differences may explain this.

The length of follow-up is described two ways. The paper calls it a six-year period while dating it 2012 to 2019 and plotting results out to year seven, and the publisher’s press release says seven years.

What the findings support

Across these studies, one result is consistent: lonelier older adults tend to score lower on cognitive tests. Whether loneliness also speeds decline is less settled, and the answer shifts with the measures used and the people studied.

The Rush analysis adds two specific points. In a large Illinois sample tracked with a detailed battery of tests, the gaps in decline between more and less lonely participants held after a check for how many family members and friends they saw each month. And its race finding, which runs against the authors’ own prediction and against the leaning of the South Side study’s estimates, remains an open question.

In the Rush data, feeling lonely went along with faster decline on thinking tests, even after the researchers counted how many family members and friends people saw each month. Which came first is still open, and Morrison thinks it may run both ways. “Both processes could occur simultaneously and even reinforce one another,” she wrote.