A new study may help settle a long-running debate: Does overloading visual memory interfere with the way we read? Your brain may predict the next word before you consciously read it, but does this ability still work when your attention is divided? A single-author study published in January 2026 found that predictable words were still processed faster even when participants’ visual memory was occupied by another task. The findings suggest that sentence context continues to guide reading in the background, even when the mind is busy.How the Brain Predicts Words While Reading. Image by MagnificNote: This article is intended for general information and educational purposes. It summarizes scientific research in accessible language for a broad audience and is not an official scientific press release.While reading a sentence, you may already be anticipating what comes next. Decades of research have shown that words fitting a strongly constraining context are generally read faster, and sometimes skipped more often, than less predictable words. But how much mental capacity does this process require?A study by Dr. Aine Ito of the National University of Singapore offers new insight into this question. Published in Memory & Cognition in January 2026, the research examined whether readers continue to benefit from predictable sentence contexts while simultaneously holding unrelated spatial information in memory.Across two preregistered experiments, word-predictability effects remained robust under the specific visuospatial working-memory load used in the study. The findings suggest that sentence context can continue to facilitate word processing even when visual and spatial memory resources are occupied by another task.What the Researcher InvestigatedThe study investigated whether context-based word prediction depends on visuospatial working memory, the system that temporarily maintains information about locations, spatial sequences, shapes, and visual patterns.One theory proposes that readers sometimes anticipate upcoming information through “event simulation.” They may construct mental representations of the situations described by a sentence, including some of their visual features. If these representations contribute directly to prediction, an additional demand on visuospatial working memory might weaken the usual advantage for predictable words.Alternatively, prediction in relatively simple sentences may rely primarily on non-visual linguistic information. In that case, a separate spatial-memory task could make reading more demanding without substantially altering the predictability effect.How the Study Was ConductedThe study comprised two preregistered experiments involving native English speakers.Experiment 1: Eye-TrackingThe final sample included 81 participants with a mean age of 22. They silently read 88 target sentences while an eye tracker recorded gaze duration, the time spent fixating on a target word, and skipping probability.Each target word appeared in either a highly predictive or an unpredictable context. For example, “air” was highly predictable in the sentence beginning, “In polluted cities, you can barely breathe the air…”In the memory-load condition, participants first saw 12 blue squares. Five squares turned red one at a time, and participants had to remember their locations in the correct order. They then read a sentence, answered a comprehension question, and reproduced the spatial sequence. In the control condition, they read without completing the memory task.Experiment 2: Self-Paced ReadingThe second experiment included 125 participants with a mean age of 26. In a web-based moving-window task, they revealed each sentence one word at a time by pressing a key.The experiment used the same predictable and unpredictable sentence contexts and the same five-location spatial-memory task. This allowed the researcher to test whether the eye-tracking results would also appear when participants controlled the presentation of each word.Why the Experimental Design MattersPrevious studies of working memory and predictive language processing have frequently used the visual-world paradigm, in which participants listen to language while viewing objects on a screen.This method is useful for studying how people combine visual and linguistic information, but it makes visuospatial working memory difficult to isolate. Participants must process the objects and their locations while interpreting the language they hear.The current study instead examined sentence-based predictability without a concurrent visual scene. By combining text reading with a separate spatial-memory task, the experiments helped distinguish the demands of maintaining spatial information from those involved in using linguistic context.Key FindingsBased on the results of the two experiments, the author identified the following key findings:1. Predictable Words Were Read FasterAcross both experiments, highly predictable words were processed faster than unpredictable words.In the eye-tracking experiment, average gaze duration was 228 milliseconds for predictable target words and 247 milliseconds for unpredictable words: a difference of 19 milliseconds. This reproduced the established predictability effect: words that fit strongly constraining contexts generally require less immediate processing time.2. Predictable Words Were Skipped More OftenIn Experiment 1, participants skipped predictable target words 39% of the time, compared with 37% for unpredictable words.Although the numerical difference was small, the analysis showed a significant overall effect of predictability on skipping probability. Strongly expected words were therefore slightly more likely to be processed without receiving a direct fixation.3. Memory Load Changed Overall Reading BehaviorThe visuospatial memory task affected how participants read overall.Average gaze duration decreased from 242 milliseconds without the additional task to 233 milliseconds under memory load. Skipping probability increased from 34% to 42%.Rather than slowing down, participants appeared to adopt a faster reading strategy while holding the spatial sequence in memory. The author suggests that they may have tried to reach the recall task sooner, before the sequence became more difficult to remember. This interpretation was not tested directly.4. The Predictability Benefit Persisted Under LoadIn Experiment 1, there was no statistically significant interaction between word predictability and visuospatial memory load.The additional task made participants read faster and skip more words overall, but it did not significantly reduce the relative advantage for predictable words. The predictability effect remained present whether or not participants were holding a spatial sequence in memory.5. Adjusted Reading Times Supported the Same PatternThe initial analysis in Experiment 2 found a smaller predictability effect under memory load. However, participants also pressed the key substantially faster overall in the load condition.To account for this difference, the researcher conducted a follow-up analysis comparing target-word reading times with reading times immediately before the target. This relative reading-time analysis found effects of predictability and memory load but no significant interaction between them.The author therefore interpreted the original interaction as likely reflecting faster overall reading under load, which may have masked the predictability effect in the raw reading times.The Author’s ConclusionsThe author concludes that visuospatial working-memory resources were not necessary for computing the probabilities of upcoming words in the reading tasks used in the study. According to Dr. Ito, context-based prediction during reading was largely unaffected by the availability of visuospatial working-memory resources.The author suggests that predictions in these experiments may have been generated primarily from non-visual information, such as linguistic representations. She also notes that the observed predictability effects may reflect both advance prediction and easier integration after a word appears, because the reading measures used in the study cannot fully separate these processes.Dr. Ito emphasizes that the results apply to the specific sentence contexts and visuospatial memory task examined in the experiments. The study did not test whether other components of working memory are required for language prediction.Limitations Identified by the AuthorThe author notes that many of the highly predictive sentences contained words strongly associated with the target. In the sentence about polluted cities, for example, “polluted” and “breathe” provide strong cues to “air.” According to Dr. Ito, these local associations may support prediction without requiring detailed visual event simulation.The author suggests that visuospatial working memory could play a greater role when predictions depend on broader discourse, connections among multiple events, speaker characteristics, or social context.Dr. Ito also discusses the initial difference between the two experiments. Experiment 1 showed no interaction between predictability and memory load, whereas the initial analysis in Experiment 2 did. Although the follow-up relative reading-time analysis supported the same overall pattern as Experiment 1, the author emphasizes the importance of replication using other experimental methods, including auditory presentation or brain-activity measures.ConclusionThe brain appears to process predictable words on a kind of autopilot, even when visual memory is busy with another task. In both experiments, participants continued to read predictable words faster despite having to remember visual information at the same time.According to Dr. Ito, this suggests that visual and spatial memory plays only a limited role in using sentence context to anticipate upcoming words. However, the author notes that longer stories or socially complex information may involve different mental processes and require further study.The information in this article is provided for informational purposes only and is not medical advice. For medical advice, please consult your doctor.ReferenceIto, A. (2026). Predictability effects during reading comprehension are not modulated by a visuospatial working memory load. Memory & Cognition. https://doi.org/10.3758/s13421-026-01857-7The post Reading on Autopilot: New Study Shows How the Brain Predicts Words appeared first on CogniFit Blog: Brain Health News.