Your Brain Can Learn Without Practice, New Study Reveals

For more than a century, the science of skill acquisition has rested on a deceptively simple assumption: to get better at something, you have to practice it, repeatedly and extensively. From William James’s writings on habit formation in the nineteenth century to the landmark texture discrimination experiments of the 1990s, researchers have documented how visual abilities sharpen over days or weeks of deliberate daily training. A new study published in iScience now delivers a striking challenge to that orthodoxy, showing that under the right conditions, the brain can produce substantial perceptual improvements with no practice at all between an initial training session and a test days later.

The research, led by Ylka Kolken and Peter De Weerd of Maastricht University together with Nitzan Censor of Tel Aviv University and colleagues, revisits one of the most influential paradigms in visual neuroscience: the texture discrimination task, or TDT, first devised by Dov Karni and Dov Sagi in 1991. In the classic version of this task, participants fixate on the center of a screen and judge whether a briefly flashed letter is a T or an L, while simultaneously reporting the orientation of a faint, texture-defined target in the periphery. Performance is measured by the stimulus onset asynchrony, or SOA, the time gap between the target and a masking pattern that participants need to reach a fixed accuracy level of 81.6 percent. Lower SOA thresholds mean sharper perception.

Over the past decade, a series of studies from Censor’s group had already shaken the field by showing that after a single initial training session, just five near-threshold reminder trials per day could match the gains produced by hundreds of daily practice trials. The proposed mechanism was memory reconsolidation: when a memory is reactivated, it briefly becomes labile and can be modified or strengthened before being stored again. Brief reminders, the argument went, could repeatedly reopen the consolidation window, allowing learning to proceed efficiently without massive repetition. But a nagging question remained: under what circumstances do reminders actually work?

Kolken and colleagues suspected the answer lay in attention. The dual-task design of the standard TDT forces participants to divide attention between a central letter and a peripheral texture, and decades of work have shown that divided attention carries a steep cost. In a classic experiment, Sabine Kastner and colleagues found that peripheral discrimination performance plummeted from 86 percent to 45 percent correct when a demanding central task was performed simultaneously. If learning in the dual-task TDT is therefore suboptimal, starved of attentional resources, then reminders might only be beneficial when the original encoding was weakened by that divided attention. To test this, the team compared learning in the standard dual-task TDT with a single-task version in which the central letter discrimination was removed entirely, leaving only the peripheral target behaviorally relevant.

The experiment followed a clean three-arm design. In both task versions, participants were assigned to a full training condition with daily sessions across five days, a reminder condition with training on days one and five plus five tiny reminder trials on the intermediate days, or a control condition with no practice whatsoever between the first and fifth days. In the dual-task TDT, with thirty participants evenly split across conditions, the results replicated the reminder literature precisely. Full training and reminders produced nearly identical, substantial reductions in SOA threshold, while the control group barely improved. Bayesian analyses confirmed that the reminder and full-training groups were statistically indistinguishable, and both clearly outperformed the controls.

Then came the surprise. When the central fixation task was removed and fixation was enforced by an eye tracker that aborted any trial in which gaze strayed more than 1.5 degrees from center, day-one thresholds collapsed to around 62 milliseconds, compared with roughly 180 milliseconds in the dual-task version. The difficulty of the classic task, it turned out, was mainly attentional rather than sensory. Because this near-ceiling performance left no room for improvement, the researchers made the stimulus harder, coarsening the texture spacing and shrinking the target from three elements to two, and ran the full three-condition experiment again with thirty new participants.

The outcome upended expectations. In the single-task TDT, all three groups, including the controls who did nothing between day one and day five, improved by similar amounts. A block-by-block analysis ruled out rapid within-session learning as the explanation: performance at the start of day five was significantly better than performance at the end of day one, a gain of roughly 36 milliseconds in SOA threshold that reflected genuine between-session consolidation. In other words, a single session of fully attended training was enough to set in motion a learning process that unfolded over several days entirely offline, with no reminders and no additional practice. Full daily training offered no advantage whatsoever over doing nothing at all.

The team also ruled out a seductive alternative explanation: cheating with the eyes. Because attention and gaze are tightly linked, participants in the dual-task TDT might have quietly shifted their fixation toward the anticipated peripheral target, adopting a compromise gaze position to boost performance. Eye-tracking recordings during the dual-task version showed this was not happening. Ninety percent of trials fell within 1.2 degrees of the fixation point on day one, average gaze position did not shift toward the target between days one and five, and there was no consistent direction of drift across participants. The T/L task, long assumed to guarantee fixation, apparently did exactly that.

The broader implication is that attentional context, the configuration determining how many spatial locations must be attended, is a key factor shaping which route perceptual learning takes. When attention is divided, as in the classic dual-task TDT, the initial memory trace is weak and needs to be actively reactivated through reminders to consolidate into lasting improvement. When attention is fully available, a single session produces a trace strong enough to consolidate on its own, unfolding over multiple days without any external prompting. This framework may also explain why a recent study by Zhu and Zhang found no reminder benefit in a dual-task TDT, and why Chen and Op de Beeck found reminder gains to be smaller than full training for complex-object learning: in both cases, the attentional resources available for encoding may have differed from the original reminder studies.

The findings carry real practical weight. Training programs for vision rehabilitation, sports, and skill learning almost universally assume that daily practice blocks are essential, implicitly assuming that offline learning cannot extend beyond a single day. This study suggests that assumption deserves scrutiny: improvements can continue for days after a single well-attended session, and the true time course of this latent consolidation, whether it proceeds gradually or concentrates in the first night of sleep, remains an open question the authors flag for future work with intermediate test points. The study is not without limitations, including the inability to fully record eye movements in the single-task condition and a residual trend toward lower single-task thresholds, but the core message stands. Practice may still make perfect, but the brain, given the right conditions, appears remarkably capable of finishing the job on its own.

Subject of Research: Attentional context determines whether reminder trials or offline consolidation drive visual perceptual learning

Article Title: No practice and still perfect: Different routes for promoting offline gains in perceptual learning

Article References: Kolken, Y., Roberts, M. J., Naseri, P., De Martino, F., Censor, N., & De Weerd, P. (2026). No practice and still perfect: Different routes for promoting offline gains in perceptual learning. iScience, 29(11), Article 117839. https://doi.org/10.1016/j.isci.2026.117839

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117839

Keywords: visual perceptual learning, texture discrimination task, memory reconsolidation, attention, offline learning, memory consolidation, eye tracking, divided attention, perceptual thresholds, neuroscience, skill learning, iScience