Hundred days of cognitive training enhance broad cognitive abilities in adulthoodfindings from the COGITO study

Picture an older adult sitting down at a computer screen — not to check email, not to browse, but to practice twelve cognitive tasks for the hundredth day in a row. One hour a day. Over a hundred consecutive days. The question worth asking is: what does that actually buy you? Not in terms of getting better at the specific tasks on the screen — that's almost guaranteed. The harder question is whether those gains lift something broader; whether the mind you carry out the door after a hundred sessions is genuinely sharper in ways that matter. That question is exactly what Schmiedek and colleagues built the COGITO study to answer. The scientific problem behind that question has a name: transfer. It's the difference between getting better at a memory game and getting better at memory. Most brain-training studies show that people improve on trained tasks, and that finding is almost trivially expected. The interesting test is whether practice on one task improves performance on entirely different tasks and whether those improvements reflect changes in broad cognitive abilities rather than narrow task-specific tricks. Schmiedek and colleagues frame this using a hierarchical taxonomy of cognition. Narrow skills sit at the bottom — the particular processes a specific task demands. Broad abilities sit above them, including things like working memory, episodic memory, perceptual speed, and fluid reasoning. These broad abilities are defined by the variance shared across multiple heterogeneous tasks. And at the top sits general intelligence. The key insight is that demonstrating transfer at the level of a broad ability is a much stronger claim than showing improvement on a single untrained test, because an ability captures what's common to many tasks, not the idiosyncrasies of any one. To test this properly, the COGITO study committed to a scale that most cognitive-training research has never approached. One hundred and one younger adults and one hundred and three older adults completed, on average, about one hundred and one one-hour daily practice sessions each. The full range ran from eighty-seven to one hundred and nine sessions. They practiced twelve different computerized tasks every day across three domains: six perceptual speed tasks, three working memory tasks, and three episodic memory tasks. The tasks were deliberately varied in format and content. Perceptual speed tasks included rapid two-choice decisions — odd versus even numbers and consonants versus vowels — and comparison tasks requiring fast judgments about whether two stimuli were identical, using digit strings, consonant strings, and three-dimensional objects. Episodic memory tasks required memorizing lists of thirty-six nouns, number-noun pairs, and the positions of objects in a grid. Working memory tasks included an alpha span, a numerical memory updating task, and a spatial three-back. In the alpha span, participants judged whether a number beneath each presented letter matched that letter's ordinal position within the letters seen so far. These weren't games designed to feel easy. A central design feature was individualized difficulty. Before training began, each participant's accuracy was measured for every task, and presentation and masking times were calibrated so predicted performance stayed in a challenging range — above a minimum but below an upper level that varied by task. This kept each person working at the edge of their ability. Transfer was then assessed using multiple tests per ability domain administered in ten pre-training and ten post-training sessions, along with paper-and-pencil measures and Raven's matrices. Critically, the analysis used confirmatory factor models and latent difference score methods rather than raw score comparisons. This design strips away task-specific noise and examines change at the level of shared ability. The within-training gains arrived as expected, and they were large. Effect sizes for practiced working memory tasks were substantial in both age groups. For the alpha span, the effect size equaled 1.08 in younger adults and 1.36 in older adults. For the numerical memory updating task, the effect size was 1.20 in younger adults and 0.98 in older adults. For the figural n-back, the effect size was 0.90 in younger and 1.46 in older adults. Episodic memory gains were also solid, with object-position memory reaching an effect size of 0.82 in younger and 1.00 in older adults. Notably, age differences in practice gains were mixed — for some tasks, older adults actually showed larger effects than younger adults. But those are the expected results. The real test was transfer. The most robust finding was transfer to working memory at the latent factor level — that is, transfer to the shared dimension underlying multiple different working memory tests that participants had never practiced. Younger adults showed a reliable latent working memory gain with an effect size equal to 0.36. Older adults showed essentially the same thing, with an effect size equal to 0.31. Both were statistically reliable. This is the finding that matters most: not just that people scored higher on individual untrained tests, but that the underlying working memory ability — extracted across multiple measures — genuinely improved. For younger adults, the gains extended further. Episodic memory showed a large latent transfer effect, with an effect size equal to 0.52. Fluid reasoning — the kind of abstract problem-solving measured by Raven’s matrices — showed a small but reliable latent transfer effect, with an effect size equal to 0.19. Perceptual speed showed no reliable latent transfer in either group. For older adults, latent-level transfer was largely confined to working memory. They did not show reliable gains in episodic memory or fluid reasoning at the latent factor level, though at the individual task level they showed meaningful improvements: an effect size of 0.54 on Raven's matrices and an effect size of 0.50 on a word-pairs memory test. Now, a skeptic might propose that all of this just reflects better mood, higher motivation, or a more positive self-concept after a hundred days of feeling successful. Schmiedek and colleagues address this directly through the pattern of correlations among the latent change factors themselves. If training effects were driven by general motivation or placebo-like enthusiasm, you'd expect changes in one domain to correlate equally with changes in all other domains. That's not what happened. The two working memory latent change factors — one from practiced tasks and one from transfer tasks — were so tightly linked they were fixed to a correlation of one in the combined model, indicating a single general working memory change factor. The episodic memory transfer latent change factor correlated with the episodic memory practiced latent change factor at zero point five-eight — but its correlation with working memory latent change was only zero point two-five and not significant. These domain-consistent patterns are hard to explain with motivation alone. They indicate that training improved specific cognitive processes, not just general engagement. On the age differences, Schmiedek and colleagues are careful. Both groups showed latent working memory transfer with nearly identical effect sizes, which is itself meaningful — older adults' cognitive systems remained plastic enough to show genuine ability-level change. The restriction of episodic memory and reasoning transfer to younger adults likely reflects differences in cognitive plasticity, though the authors note a methodological wrinkle: COGITO kept difficulty fixed at levels set from pre-training rather than adapting dynamically throughout. It's possible that older participants were operating at more plasticity-inducing difficulty later in practice than the design captured. The paper points to neural evidence suggesting mechanisms including changes in dopaminergic systems and experience-dependent white-matter reorganization, but these interpretations are held lightly. So, back to that older adult on day one hundred. What does the science say they actually gained? Large improvements on the practiced tasks themselves — effect sizes often exceeding one. A reliable ability-level improvement in working memory, with an effect size equal to 0.31 at the latent factor level, comparable to what younger adults showed. Meaningful gains on specific transfer tasks including Raven's matrices and word-pairs memory. What the study cannot yet tell us is how long any of this lasts, how well it extends to real-world intellectual demands beyond the laboratory, or what the minimum effective dose looks like. What makes the COGITO findings stand apart from most prior training research isn't the scale of the practice effects on trained tasks — those were expected. It's the demonstration that change propagated upward through the cognitive hierarchy, from narrow practiced skills to broad latent abilities, in a pattern that respects domain boundaries. That's a stronger claim. And for a field that has spent decades debating whether brain training does anything at all, it's a meaningful place to stand. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.