Ageing is often described as though the mind moves in only one direction. Names become harder to retrieve, distractions are more costly and a task that once felt immediate can take longer. Some of those changes are real. Processing speed and parts of working memory commonly become less efficient across adulthood.
But that is not the whole story of a mind, and a three-year online study offers a useful reason to resist treating it as one.
The study, published in Scientific Reports, analysed 3,966 adults aged 19 to 94 who completed repeated assessments through the BrainHealth Project. On average, their scores rose over time on a composite measure that combined cognitive clarity, emotional wellbeing and social connection.
That finding is encouraging. It is also easier to misunderstand than it first appears. The study did not show that every cognitive skill becomes better with age, that all 3,966 people were followed for a full three years, or that an online programme was proven to cause the gains. What it did show is subtler: a broad, multidimensional measure of brain health can improve across adulthood, including later adulthood, even while particular abilities remain vulnerable to decline.
What the BrainHealth Index actually measures
The central measure was the BrainHealth Index, or BHI, developed at the Center for BrainHealth at the University of Texas at Dallas. According to the centre’s description of the index, it is intended to create a personal benchmark across several contributors to brain health rather than diagnose disease or rank one person against a population norm.
The online assessment took roughly 60 to 70 minutes, although participants could spread it across two weeks. Its objective cognitive tasks examined strategic attention, reasoning, abstraction, memory, innovation and processing speed. Questionnaires added information about sleep, mood, anxiety, stress, resilience, happiness, purpose, social support, social engagement, physical fitness and growth mindset.
Those inputs were organised into three factors. Clarity included cognitive performance and supporting functions such as sleep. Emotional Balance covered mood, stress, anxiety, resilience and productivity. Connectedness described a person’s perceived connection to other people and to a sense of purpose.
This breadth is not a technical footnote. It explains how a composite brain-health score can rise without implying that every mental ability rose with it. A person might process information a little more slowly while sleeping better, feeling more emotionally steady, strengthening relationships or finding greater purpose. Another might improve on a reasoning task while reporting no social change. Both trajectories can affect the same overall index.
3,966 supplied a trajectory, but 478 reached year three
The researchers identified 4,190 possible participants, then excluded withdrawals and records that did not meet the analysis requirements. The final group of 3,966 had completed the baseline assessment and at least the six-month follow-up. That is what allowed each person to contribute a change over time rather than a single snapshot.
Assessments were offered every six months. Participation then narrowed at each later wave: 2,501 people completed the one-year assessment, 1,663 reached 18 months, 1,153 reached two years, 764 reached 30 months and 478 had completed the three-year assessment by the data cutoff.
Calling it a three-year study is accurate because the observation window extended that far. It should not be read to mean that every member of the 3,966-person analysis was observed for three years. The later estimates rest on a much smaller set of people who continued returning.
That attrition matters because long-term participants may differ from those who stop. They may be more motivated, more comfortable with online testing, healthier or simply have more time. Statistical models can use the available repeated measurements, but they cannot guarantee that the people who remained represent everyone who began.
Average scores rose across the available assessments
Across the follow-up period, the researchers reported monotonic increases in the group averages for the overall BHI and for Clarity, Emotional Balance and Connectedness. In plain language, the estimated average trajectory continued upward rather than rising once and then immediately fading.
Participants who began in the lowest scoring quartile generally showed the steepest gains. Improvement also appeared in the higher starting quartiles, which is important because the lowest group had the most room to rise. Even so, regression toward the average can make unusually low initial scores look better on repeat testing, and the researchers acknowledged that possibility.
The pattern appeared across age groups. Older age was not a barrier to a positive trajectory on the composite. Nor did the association between platform use and improvement differ significantly by age or gender. Those results are consistent with retained capacity for change across the adult lifespan.
They do not mean everyone improved. A rising average can include people who gained substantially, people who barely changed and people who declined. The paper presents a population-level pattern in a participating cohort, not a promise about an individual future.
More engagement travelled with greater gains
After assessment, participants could use a set of online tools intended to support change. These included strategy-based cognitive training, habit-building modules, educational resources and quarterly virtual coaching. Use varied substantially.
People with higher utilisation tended to record larger improvements in the overall index, while the lowest-use group changed least. That dose-like pattern makes the programme a plausible contributor. It does not, however, prove that the programme produced the difference.
The project was registered at ClinicalTrials.gov, but the analysis was open-label and single-arm. There was no randomized untreated group against which to compare the participants. People decided how deeply to engage, so motivation, health, education, spare time and confidence with technology could influence both utilisation and later scores.
Repeated exposure to the tests may also improve familiarity. Self-reported mood or connection can change with expectations, and the study overlapped with a period of unusual social disruption associated with the COVID-19 pandemic. Without a control group, the observed improvement cannot be cleanly divided into programme effects, practice effects, outside changes and participant selection.
Improvement is not the same as reversing cognitive ageing
The results fit a view of adulthood in which loss and growth can occur at the same time. Vocabulary, accumulated knowledge and emotional regulation may remain stable or strengthen while speed, divided attention or working memory become more demanding. A composite can capture some of that balance, but it should not blur the distinction between its components.
The National Institute on Aging’s guidance on cognitive health similarly describes a mixed evidence base. Structured trials have found that certain kinds of training can benefit targeted abilities, while evidence for lasting cognitive effects from many commercial games and general activities remains incomplete.
Earlier ScienceBlog coverage of strategy-based training in adults over 55 reported randomized evidence and changes in brain blood flow measured with MRI. That work helps make later-life improvement biologically plausible. It does not turn the present online cohort into a randomized experiment, and the 3,966-person study itself did not scan every participant’s brain.
The phrase “the brain’s capacity for positive change” therefore needs restraint. The new study measured behaviour, performance and self-reported experience. Its results are consistent with lifelong plasticity, but they are not direct evidence that a specific neural circuit changed in every participant.
Social connection belongs in a brain-health measure
One of the study’s more valuable choices was to treat connectedness as part of brain health rather than as pleasant background scenery. Thinking does not occur apart from sleep, stress, purpose and relationships. A person can score well on a short cognitive task and still be struggling in ways that shape daily function.
Recent ScienceBlog reporting on social isolation, loneliness and cognitive decline highlighted another necessary distinction. Isolation describes the objective scarcity of contact, while loneliness is the felt gap between the relationships someone has and the relationships they need. They overlap, but one does not perfectly stand in for the other.
The Connectedness factor relied on perceived support, engagement and purpose, making it sensitive to the lived quality of connection. That can reveal meaningful change that a memory score alone would miss. It can also make the index harder to interpret as a purely cognitive outcome. Improvement in Connectedness may be valuable in its own right, but it is not equivalent to faster processing or stronger recall.
That is not a flaw so much as a statement about what the BHI is designed to be: a broad health indicator. The problem arises only if readers translate that indicator into the narrower claim that biological cognitive ageing has been reversed.
A hopeful result with important boundaries
The sample was not a miniature version of the whole adult population. It was self-selected through online advertising, word of mouth and a brain-health registry. About three quarters of participants were women. The cohort was highly educated and predominantly white, especially among those older than 65. Only 100 participants were 18 to 25, while 1,814 were older than 65.
Participants were generally healthy enough to function independently. People with diagnosed neurodegenerative disease and certain functionally impairing strokes, concussions or brain injuries were excluded. The findings should not be presented as evidence that the platform treats dementia or restores abilities lost to neurological disease.
There is also a commercial-interest boundary worth knowing. The University of Texas System has filed a patent application covering the BrainHealth Index and aspects of the platform, and several study authors are named as inventors. A patent does not invalidate a result, but disclosure helps readers judge why independent replication is especially useful.
The most defensible conclusion is both more modest and more humane than the promise of a permanently improving brain. Across a large online cohort, adults who returned for follow-up showed average gains in a measure combining cognitive clarity, emotional balance and connectedness. Those gains appeared across ages, including later life.
Specific cognitive abilities can still decline. Some participants may not have improved. The platform’s causal role remains uncertain, and only a fraction of the starting cohort had reached three years. Yet the trajectory challenges the fatalistic idea that ageing closes every route to positive change. A mind can become slower in one respect while becoming steadier, better supported or more purposeful in another. That is not a loophole in the meaning of brain health. It may be one of its most realistic definitions.