Mental calculation speeds surpass known limits for high-level cognition

By analyzing data from international mental calculation competitions, this study reveals that high-level cognitive processing can sustain information rates exceeding 215 bit/s, significantly surpassing the previously accepted limit of 10 bit/s.

The Gist

Imagine your brain as a super-highway. For decades, scientists thought this highway had a strict speed limit sign: 10 bits per second. They believed that no matter how smart you were, your conscious mind could only process about 10 tiny pieces of information every second. It's like trying to drive a Ferrari through a narrow, winding country lane; you just can't go fast without crashing.

But a new study by researchers in Munich suggests that sign might be wrong. They found that when humans are pushed to their absolute limit, that highway can actually handle 215 bits per second. That's not just speeding; that's a rocket ship.

Here is how they figured it out, explained simply:

The Test: Mental Math Olympians

The researchers didn't just ask regular people to do math. They looked at the world's best mental calculators—people who compete in international tournaments to add, multiply, or find square roots of huge numbers in their heads.

Think of these people as the "Formula 1 drivers" of the brain. They aren't naturally gifted geniuses; they are regular people who have trained their brains for years, much like an athlete trains their muscles.

The Experiment: Measuring the "Data Stream"

The team treated the brain like a computer. They measured three things for every calculation:

1. Reading the numbers: How much data comes in?
2. Doing the math: How much "thinking" happens?
3. Saying the answer: How much data goes out?

They used a concept called Entropy (a fancy word for "how many possibilities there are"). If you have to guess a number between 1 and 10, that's low entropy. If you have to guess a specific number between 1 and a trillion, that's high entropy.

The Big Discovery

When they crunched the numbers, they found that for short, intense bursts of calculation, these experts were processing information at 215 bits per second.

To put that in perspective:

• The Old Limit (10 bits/s): Like reading a book one word at a time, pausing to think about every single word.
• The New Limit (215 bits/s): Like watching a high-definition movie where you understand every scene, every line of dialogue, and every visual detail instantly.

The researchers broke this speed down into two parts:

• 125 bits/s for "Seeing": The speed at which they can take in the numbers.
• 90 bits/s for "Calculating": The speed at which their brain executes the math algorithms.

Why Can They Go So Fast?

You might wonder, "How do they do it? Do they have a supercomputer in their head?"

The answer is practice and strategy.

• The Mental Abacus: Many of these experts use a technique called the "Mental Abacus." Imagine a physical abacus (a frame with beads) in your mind. Instead of writing down numbers, they visualize beads sliding back and forth. Their brain treats the math like a physical movement, which is much faster than trying to "think" through the logic step-by-step.
• Shortcuts: They don't calculate everything from scratch. They have memorized thousands of small math facts (like a massive lookup table) so they can skip steps. It's like knowing the answer to a math problem because you've seen it a thousand times, rather than solving it for the first time.

The Catch: It's Hard to Sustain

While they can hit these rocket-ship speeds, they can't keep it up forever. The study found that as the task gets longer (like adding 200 numbers instead of 15), the speed drops.

Think of it like a sprinter vs. a marathon runner. These mental calculators are sprinters. They can hit 215 bits per second for a few seconds or minutes, but they can't maintain that pace for hours. The brain gets tired, and the risk of making a mistake goes up, so they have to slow down to be safe.

The Bottom Line

This study changes how we view human potential. It suggests that the "10 bits per second" limit isn't a hard wall of our biology; it's more like a default setting for our daily lives. When we are highly trained, focused, and motivated, our brains can unlock a hidden "turbo mode" that processes information 20 times faster than we ever thought possible.

So, the next time you feel like your brain is slow, remember: you might just be driving in the slow lane. With enough training, you might be able to switch to the fast lane.

Technical Summary

1. Problem Statement

Current cognitive science models, notably a recent study by Zheng and Meister (2025), posit a universal limit for high-level human information processing at approximately 10 bit/s. This limit is thought to reflect the complexity of conscious decision-making under ordinary circumstances. However, it remains unclear whether this limit holds when the human mind is pushed to its absolute performance limits through intense training and specific cognitive tasks. The authors aim to test the upper bounds of human cognitive speed by analyzing mental calculation, a domain characterized by standardized tasks, explicit algorithms, and measurable performance metrics.

2. Methodology

The authors analyzed data from international mental calculation competitions (e.g., Mental Calculation World Cup, Japanese Soroban competitions) and record-setting attempts. They focused on four disciplines: addition, multiplication, calendar date calculation, and square root extraction.

Key Analytical Framework:

• Entropy-Based Measurement: Instead of measuring raw speed, the authors quantified task complexity using Shannon entropy (bits). They calculated the information content required at three distinct stages:
  1. Input/Extended Input: The entropy of the visual numbers presented. "Extended input" accounts for digits read multiple times during the algorithm (common in square roots and multiplication).
  2. Algorithmic Processing: The entropy of the internal decisions and interim results generated by the mental algorithm (e.g., carrying over numbers, partial products).
  3. Output: The entropy of the final reported answer.
• Scaling Laws: They plotted calculation time ($T$) against task entropy ($H$) to determine scaling relationships, specifically testing for power-law dependencies ($T \propto H^\alpha$).
• Algorithmic Modeling: The authors modeled specific strategies used by experts (e.g., the "mental abacus" for addition, "cross-multiplication" for multiplication, and the "duplex method" for square roots) to estimate the number of internal steps and the associated entropy. They ruled out the possibility that competitors rely on massive pre-memorized lookup tables (which would reduce processing time to near-zero) by demonstrating the combinatorial explosion required for such tables (e.g., $10^{13}$ options for multi-digit addition).
• Comparison: Results were compared against data from memory competitions to distinguish between algorithmic execution and memory encoding/decoding.

3. Key Contributions

• Redefining Cognitive Limits: The study provides empirical evidence that high-level cognition can operate at rates far exceeding the previously accepted 10 bit/s limit.
• Decomposition of Cognitive Load: The authors successfully separated cognitive processing into distinct components (perception, algorithmic execution, and motor output), allowing for a granular analysis of where information is processed.
• Validation of Hick's Law in Complex Tasks: They demonstrated that reaction times in complex mental calculations scale with entropy according to Hick's Law (reaction time $\propto$ entropy), with exponents close to unity ($H^{1.29}$ for extended input and $H^{1.14}$ for algorithmic steps), confirming that decision complexity drives time even in highly practiced skills.
• Distinction from Memory: The study clarifies that mental calculation relies on algorithmic execution rather than the "mnemonic image" strategies used in memory competitions, resulting in significantly higher information throughput.

4. Key Results

• Peak Information Rates:
  • The total information processing rate peaks at >215 bit/s for the shortest, most intense calculation tasks (e.g., summing 15 four-digit numbers in ~1.77 seconds).
  • This rate is sustained for minutes, dropping to ~30 bit/s for longer tasks (e.g., 10-minute disciplines), but remains significantly higher than the 10 bit/s baseline.
• Component Breakdown (Peak Performance):
  • Extended Input (Visual Perception): ~125 bit/s.
  • Algorithmic Execution: ~90 bit/s.
  • Output (Motor/Reporting): <15 bit/s.
• Scaling Laws:
  • Calculation time scales as a power law with entropy. When accounting for "extended input" (re-reading digits) and "algorithmic entropy" (interim decisions), the data across different disciplines collapses into a single universal power law ($R^2 > 0.9$), indicating a fundamental cognitive constraint.
  • The exponents for time vs. entropy are approximately 1.29 (extended input) and 1.14 (algorithmic), slightly higher than the linear prediction of Hick's Law, likely due to error-avoidance strategies in longer tasks.
• Comparison with Memory:
  • Mental calculation is 2 to 4 times faster than entropy-matched memory tasks.
  • Memory tasks require mapping digits to "mnemonic images" (encoding/consolidation), which adds a significant cognitive overhead compared to the direct algorithmic execution used in calculation.

5. Significance

• Challenging the "10 bit/s" Dogma: The findings suggest that the 10 bit/s limit applies to untrained or ordinary conscious decision-making. However, with extensive training and concentration, the human brain can sustain processing speeds >20 times higher than this limit.
• Nature of Expertise: The results support the view that expert mental calculation is a learned skill involving optimized algorithms and working memory management, rather than an innate, singular gift. The regularity of performance across hundreds of competitors suggests these high speeds are a reproducible physiological capability.
• Cognitive Architecture: The study implies that the human mind possesses a high-bandwidth channel for conscious computation that is usually bottlenecked by the need for error-free performance in complex, novel situations. In highly practiced, structured environments, this bottleneck is removed.
• Implications for AI and Neuroscience: Understanding how humans achieve such high bit rates through algorithmic optimization and parallel processing (e.g., mental abacus) offers insights for designing more efficient cognitive architectures and AI systems that mimic human expert performance.

In conclusion, the paper demonstrates that the human mind is capable of operating at ~215 bit/s under optimal conditions, fundamentally revising our understanding of the speed limits of high-level cognition.