Outsmarted by Ayumu: What a Chimpanzee Taught Us About Memory
Discover the famous Ayumu experiment, understand why chimps outperform humans at this task, and explore what this reveals about the evolution of memory.
In 2007, a young chimpanzee named Ayumu stunned the scientific world. At Kyoto University's Primate Research Institute, Ayumu demonstrated a remarkable ability to memorize the positions of numbers flashed briefly on a screen - outperforming adult humans consistently and dramatically. The research, published in the journal Current Biology by Inoue and Matsuzawa, made international headlines and challenged long-held assumptions about human cognitive superiority.
This finding opened new windows into the evolution of memory and forced scientists to reconsider what makes human cognition unique. Rather than being uniformly superior to other primates in all cognitive domains, humans appear to have traded certain immediate memory abilities for the capacity for language and abstract thought. The 'Chimp Test' based on this research has since become one of the most popular cognitive tests on the internet, offering a humbling and fascinating way to explore the limits of human spatial working memory.
The story of Ayumu is part of a broader research program at Kyoto University that has studied chimpanzee cognition for over 40 years. Professor Tetsuro Matsuzawa, who has led this work since the late 1970s, has documented a wide range of cognitive abilities in chimpanzees, including tool use, numerical competence, and face recognition. The memory research with Ayumu, however, remains the most striking demonstration that chimpanzees can outperform humans in a well-controlled laboratory setting.
The Ayumu Phenomenon: Primate Memory Research
Professor Tetsuro Matsuzawa and his student Sana Inoue at Kyoto University trained chimpanzees to touch numbers 1-9 in ascending order on a touchscreen. The critical experimental manipulation was the masking procedure: after the subject touched '1,' all other numbers were replaced by white squares. The chimps had to remember where each remaining number had appeared and continue clicking them in order. This is essentially a test of spatial working memory with sequential retrieval demands.
Inoue and Matsuzawa (2007) compared the performance of three young chimpanzees (including 5-year-old Ayumu) with adult human university students. When numbers were displayed for 650 milliseconds, both species performed comparably. But when display time was reduced to 430 milliseconds and then to just 210 milliseconds - faster than a single eye fixation - Ayumu's performance remained high (approximately 80% correct for 5 numbers at 210ms) while human performance dropped dramatically (approximately 40% correct). This was a controlled, well-replicated finding that has withstood scrutiny from the scientific community.
The 'cognitive tradeoff hypothesis,' proposed by Matsuzawa (2009), suggests that chimpanzees retained an ancestral form of eidetic-like immediate memory that humans traded away during the evolution of language. As the human brain reallocated neural resources to support the complex symbolic and syntactic processing required for language, the rapid visuospatial encoding system that chimpanzees possess may have been diminished. This hypothesis is consistent with the observation that human children (who are still developing language) perform better on the task than adults.
Silberberg and Kearns (2009) raised methodological questions about the original study, noting that the chimpanzees had thousands more practice trials than the human participants. However, subsequent research by Inoue and Matsuzawa (2009) addressed this by providing extensive training to human participants, who improved but still could not match Ayumu's performance at the fastest speeds. Cook and Wilson (2010) further confirmed the chimpanzee advantage using modified protocols. The weight of evidence supports a genuine species difference in rapid spatial memory encoding.
Neuroimaging research helps explain the mechanism. Sliwa and Freiwald (2017) used fMRI in macaque monkeys and found specialized neural circuits for rapid visuospatial encoding in primate parietal and prefrontal cortex. In humans, these same regions are heavily co-opted for language processing (the inferior parietal lobule, for instance, is critical for both spatial cognition and reading). This neural 'competition' may explain why language development correlates with declining performance on rapid spatial memory tasks in human children.
Key Research Findings
- Young chimpanzees (especially Ayumu) consistently outperform adult humans at rapid spatial memory tasks when display times are 210ms or less (Inoue & Matsuzawa, 2007)
- The performance gap narrows substantially at slower presentation speeds (650ms+), suggesting humans can compensate with additional encoding time
- Human children aged 5-6 perform better than adults on similar tasks, paralleling the pattern seen in young chimps versus older chimps
- Extensive training improves human performance significantly but does not eliminate the gap with chimpanzees at the fastest speeds (Inoue & Matsuzawa, 2009)
- The cognitive tradeoff hypothesis suggests humans traded rapid eidetic memory for language capacity during evolution (Matsuzawa, 2009)
How the Chimp Test Works
Our chimp test replicates the core challenge of the Inoue and Matsuzawa experiments. Numbers appear at random positions on a grid, and you must remember their locations after they disappear. The fundamental cognitive demand is the same: encode the spatial positions of ordered items and retrieve them sequentially.
The test starts with 4 numbers and progressively increases difficulty by adding one number per level. Unlike the original research, which used a fixed number of items at varying display times, our version uses untimed viewing (you control when masking occurs by clicking '1') with increasing item counts. You also get 3 lives, allowing for mistakes without immediately ending the test. This makes the test more accessible while still pushing toward the limits of human spatial working memory.
The 8x5 grid provides 40 possible positions for numbers, ensuring that at all difficulty levels the positions are well-separated and do not form trivially obvious patterns. Random placement also means that each attempt is unique, preventing you from memorizing specific configurations.
How the Test Works
- 1Numbers (starting with 4) appear at random positions on the 8x5 grid
- 2You click '1' to begin, which hides all remaining numbers behind blank tiles
- 3You must click the remaining positions in ascending order (2, 3, 4...)
- 4Correct completion of all numbers advances you to the next level (one more number)
- 5Mistakes cost one of your 3 lives; the game ends when all lives are lost
Why Chimps Beat Humans (And What Affects Your Performance)
Understanding why chimps excel at this task illuminates the factors that affect human performance. The chimpanzee advantage appears to stem from a fundamentally different encoding strategy - one that captures the entire visual scene at once rather than processing items sequentially. Several factors modulate human performance on this task.
Eidetic vs. Sequential Encoding
Chimps appear to use a 'snapshot' or eidetic-like memory that captures the entire display simultaneously. Humans typically encode items sequentially, reading numbers one by one. This sequential strategy is slower and becomes increasingly error-prone as the number of items grows. Matsuzawa (2009) argued that this difference reflects a fundamental divergence in cognitive architecture between the two species.
Language Interference
Humans automatically engage in subvocalization - silently naming the numbers ('one, two, three...'). This verbal encoding process is slower than pure visuospatial encoding and may interfere with the spatial memory trace. Chimps, lacking productive language, encode the positions purely as spatial information. Dual-task experiments in humans show that articulatory suppression (repeating 'the, the, the' to block subvocalization) can paradoxically improve performance on spatial tasks.
Display Duration
The chimpanzee advantage is most pronounced at very short display times (210ms). At longer display times (650ms+), humans can use their sequential encoding strategy to build a complete representation. In our test, display time is self-paced, which gives humans their best chance at matching chimpanzee-level performance.
Developmental Changes
Human children aged 5-6 perform better than adults on rapid spatial memory tasks (Inoue & Matsuzawa, 2007). This age effect parallels the development of language, suggesting that as language networks mature and monopolize neural resources, eidetic-like spatial memory diminishes. Haber (1979) documented that eidetic imagery is more common in children than adults.
Practice Effects
Humans can improve significantly with practice. Inoue and Matsuzawa (2009) showed that trained human participants substantially improved their scores, though chimps maintained their advantage at the fastest speeds. In our test, repeat players often report meaningful improvement over their first several sessions.
Attention and Arousal
Full, focused attention during the viewing period is critical. Any distraction - even a brief glance away from the screen - drastically impairs spatial encoding. The brief presentation times in the original experiments leave no room for attentional lapses, which partially explains the human disadvantage under those conditions.
Strategies to Improve at the Chimp Test
While you may never match Ayumu's speed, these strategies can help you push closer to the limits of human spatial memory. The key insight from the primate research is that humans struggle because they try to process numbers verbally rather than spatially. Strategies that encourage spatial, holistic encoding tend to be most effective.
Gestalt Perception
Try to see the numbers as a pattern, shape, or constellation rather than as individual items. This approach mimics the holistic processing that chimps appear to use naturally. Research on expert perception by Chase and Simon (1973) showed that chess masters perceive board positions as meaningful patterns rather than individual pieces - the same principle applies here.
Minimize Verbalization
Resist the urge to 'read' or name the numbers. Focus on their positions as visual-spatial information, not linguistic symbols. Some players find it helpful to hum or quietly vocalize a monotone sound to block the automatic naming response, similar to the articulatory suppression technique used in cognitive psychology experiments.
Chunking by Spatial Region
Group numbers by their location on the grid (top-left cluster, middle row, bottom-right area). This reduces the number of independent items you need to track from n individual positions to 2-3 spatial groups, well within working memory capacity. Gobet et al. (2001) showed that spatial chunking is one of the most effective strategies for visuospatial memory tasks.
Eye Movement Strategy
Practice a consistent scan pattern before clicking '1.' Some players trace a path from the lowest number upward, building a motor plan that can be executed from memory. Hayhoe and Ballard (2005) showed that eye movements are tightly linked to motor planning, so rehearsing the scan path may create a motor memory trace that supplements your visual memory.
Relaxed, Receptive Focus
Paradoxically, trying too hard can impair performance on spatial memory tasks. Anxiety and effortful processing engage the prefrontal cortex, which may interfere with the rapid, automatic encoding that this task demands. A relaxed, receptive attentional state allows faster and more holistic visual encoding. This is consistent with research on choking under pressure (Beilock & Carr, 2001).
Progressive Speed Training
Practice at difficulty levels slightly beyond your comfort zone. Ericsson, Krampe, and Tesch-Romer (1993) demonstrated that deliberate practice - structured training just beyond current ability, with feedback - is the most effective route to expert performance. Gradually increasing the number of items trains your spatial encoding system to work faster and more efficiently.
How You Compare: Population Statistics
Human performance on the chimp test varies widely based on practice, strategy use, and individual differences in spatial working memory capacity. For reference, Ayumu can reliably complete sequences of 9 numbers at display times where most humans struggle with 5. In our self-paced version (where you control the display time by clicking '1'), human performance is generally higher than in the timed version used in the original research.
Your score represents the highest level you completed successfully before losing all 3 lives. Because you can retry a failed level (at the cost of a life), your score reflects your peak spatial working memory capacity under optimal conditions rather than your average performance.
| Ranking | Score Range | Percentile |
|---|---|---|
| Chimp-Level Memory | 12+ numbers | Top 1% |
| Excellent | 9-11 numbers | Top 10% |
| Above Average | 7-8 numbers | Top 30% |
| Average | 5-6 numbers | Top 50% |
| Below Average | Under 5 numbers | Bottom 50% |
References
- Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled performance: What governs choking under pressure? Journal of Experimental Psychology: General, 130(4), 701-725.
- Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55-81.
- Cook, P., & Wilson, M. (2010). Do young chimpanzees have extraordinary working memory? Psychonomic Bulletin & Review, 17(4), 599-600.
- Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363-406.
- Gobet, F., Lane, P. C. R., Croker, S., Cheng, P. C.-H., Jones, G., Oliver, I., & Pine, J. M. (2001). Chunking mechanisms in human learning. Trends in Cognitive Sciences, 5(6), 236-243.
- Inoue, S., & Matsuzawa, T. (2007). Working memory of numerals in chimpanzees. Current Biology, 17(23), R1004-R1005.
- Inoue, S., & Matsuzawa, T. (2009). Acquisition and memory of sequence order in young and adult chimpanzees (Pan troglodytes). Animal Cognition, 12(Suppl 1), S59-S69.
- Matsuzawa, T. (2009). Symbolic representation of number in chimpanzees. Current Opinion in Neurobiology, 19(1), 92-98.
- Silberberg, A., & Kearns, D. (2009). Memory for the order of briefly presented numerals in humans as a function of practice. Animal Cognition, 12(2), 405-407.
- Sliwa, J., & Freiwald, W. A. (2017). A dedicated network for social interaction processing in the primate brain. Science, 356(6339), 745-749.
Ready to Test Yourself?
Put your knowledge into practice with our Chimp Test test.
Try the Chimp Test TestExplore More Articles
The 250 Millisecond Window: What Happens Between Seeing and Doing
Explore the neuroscience behind human reaction time, learn what affects your reflexes, and discover evidence-based strategies to improve your response speed.
Miller's Magic Number: Why You Can Only Remember 7 Things at Once
Discover how your working memory functions, explore the research on memory capacity, and learn techniques to expand your ability to remember sequences.