/ Research

Feynman's Counting Experiment — Do We Think in Different Codes?

3 November 2026 · CognitionType Research Lab

Try this. Close your eyes and count silently to 60 at a steady rhythm, one number per second. Now do it again, but this time say the alphabet out loud while you count. Then try a third time while reading a paragraph of this article.

For most people, one of those combinations falls apart almost immediately. For some, it is the talking. For others, it is the reading. A few people find that both are fine, and many find that neither is.

If you have seen this experiment shared online, it probably came with a neat conclusion: some people think in words and some think in pictures, and this little test sorts you into your box. The original story, told by the physicist Richard Feynman, is more interesting than that. It also says a lot less about fixed "types" than the viral version suggests.

What was Richard Feynman's counting experiment

The account comes from a chapter titled "It's as Simple as One, Two, Three," published in Feynman's 1988 book What Do You Care What Other People Think?, as told to his friend Ralph Leighton. It describes something Feynman did as a graduate student at Princeton around 1940.

What set him off was a psychology paper he called "kind of dumb." Its author had timed his feverish wife as she counted to 60, found that she counted faster when her temperature was higher, and concluded that the brain's time sense depended on a chemical reaction involving iron. Feynman thought the chain of reasoning was weak, but he found the underlying question irresistible: when you count at an even rate, what sets that rate, and what can change it?

So he started counting. His first silent count to 60 took 48 seconds. He was not bothered that it wasn't exactly a minute:

"The problem was not to count for exactly one minute, but to count at a standard rate."

Over repeated trials he got 49, 48, 47, 48, 49, 48, 48. He then ran up and down the stairs at the Princeton graduate college to raise his heart rate and counted again. Nothing changed. Simply sitting and guessing when a minute had passed, by contrast, was wildly inconsistent.

Why Feynman could read while counting but could not talk

Feynman then counted while doing ordinary chores. He could fill in a laundry form while counting. He found that he could read newspaper articles with no effect on his rate at all.

What he could not do was speak. When friends saw him sprinting up and down the stairs and asked what he was doing, he found he couldn't answer without losing the count. Typing from a book also seemed to slow him down, but on closer inspection his rate wasn't changing. When he hit a difficult word, the count simply paused:

"My counting rate wasn't slowing down; rather, the counting itself was being held up temporarily from time to time."

At breakfast, he reported all this to the other students. One of them was John Tukey, who would go on to become one of the most influential statisticians of the twentieth century. Tukey did not believe him:

"I don't believe you can read, and I don't see why you can't talk. I'll bet you I can talk while counting to myself, and I'll bet you you can't read."

Feynman read a book while counting and hit his usual 48 seconds, then summarised what he had read. Tukey then counted while reciting "Mary had a little lamb" and other nonsense aloud, and hit his own time exactly. When they compared notes, they found that Tukey was not counting in words at all. In Feynman's words, he was "visualizing a tape with numbers on it going by." Tukey could talk because his count wasn't using his voice, and he couldn't read because his count was using his eyes.

Is the counting experiment real science or just an anecdote

It is an anecdote. Two people tested themselves, informally, more than eighty years ago. There is no published, controlled replication of the Feynman–Tukey comparison as such, and claims online that it "proves" people are verbal or visual thinkers go well beyond what two graduate students could show over breakfast.

But the logic Feynman stumbled onto is exactly the logic of a major line of cognitive research. In 1968, the Canadian psychologist Lee Brooks published a study showing that two tasks interfere most when they compete for the same kind of mental resource. Two verbal tasks clash, and so do two spatial tasks. A verbal task paired with a spatial one clashes far less. This "selective interference" approach became one of the main tools psychologists use to work out how a task is being done inside someone's head.

Six years later, Alan Baddeley and Graham Hitch built that idea into their model of working memory. Their model includes a phonological loop, a short-lived speech-based store kept alive by silent rehearsal, and a visuospatial sketchpad for images and spatial layouts. A standard way to knock out the phonological loop is articulatory suppression: making someone repeat an irrelevant word aloud. Feynman's inability to talk while counting is articulatory suppression, discovered by accident.

Research on counting itself backs this up. Robert Logie and Baddeley proposed in 1987 that keeping track of a running count relies on the phonological loop. Later work by Logie, Ken Gilhooly and Valerie Wynn found that mental addition was disrupted by concurrent speech tasks, but not by concurrent hand movements. For a verbal counter like Feynman, that is exactly the expected result.

Feynman's real conclusion — test what people do, not what they say

What Feynman took away from the experiment was not "I am a word person and Tukey is a picture person." It was a point about method:

"What goes on in different people's heads when they think they're doing the same thing — something as simple as counting — is different for different people."

And second, that you can find this out without relying on self-report. You don't have to ask a person how they count, he wrote; you observe what they can and can't do while they count.

That first point has held up remarkably well. Russell Hurlburt at the University of Nevada, Las Vegas, has spent decades sampling people's inner experience at random moments using a beeper. In a 2008 study with Christopher Heavey, inner speech, inner seeing, unsymbolised thinking, feelings and sensory awareness each showed up in roughly a quarter of sampled moments overall. But individuals varied enormously. Some people's minds were dominated by one form; others barely used it at all.

People also misjudge their own inner experience. In a 2018 study led by Vincent Brouwers with Hurlburt, 16 students were beeped while silently reading short stories. Many theorists assume readers hear a voice narrating the text. Only about 3 percent of samples actually involved that kind of inner narration. Visual imagery was frequent. We explored this gap between assumed and actual inner speech in Do I Have an Inner Monologue?.

Where Feynman overstated it — the test can be beaten

Feynman went further than the evidence, though. He called the test "absolute," with "no way to beat it; no way to fake it." That assumes each person has one fixed way of counting. His own story shows that they don't.

Look closely at what happened with the laundry. Feynman could write numbers on a form while counting, but he could not count his socks, because his verbal "counting machine" was already busy. So he improvised. He arranged the socks in geometrical patterns, a pair in each corner of a square, and read off the total visually. He then extended the method to newspaper columns, grouping lines into clusters of 3, 3, 3 and 1 to make 10, and counting more than a hundred lines while his silent count ran on.

In other words, the man held up as the archetypal verbal counter invented a visual counting strategy the moment the verbal one was occupied. He was not stuck in one code. He switched codes to fit the job.

The research literature shows the same flexibility. In 1996, Robert Logie, Sergio Della Sala and colleagues tested 251 adults on classic markers of verbal short-term memory, such as the word-length effect and the phonological similarity effect. Across the group, both effects appeared as expected. But when 40 participants were retested, the effects were unreliable at the individual level. The authors suggested that people were using different strategies, and changing them from session to session.

Do some people really think in words and others in pictures

Partly, yes. Differences in how strongly people use different internal formats are real and measurable. In 2020, Hettie Roebuck and Gary Lupyan published the Internal Representations Questionnaire, which identified four distinct but related tendencies: internal verbalisation, visual imagery, orthographic imagery (seeing words as written text), and the ability to manipulate mental representations.

At the extremes, the differences affect performance. In 2024, Johanne Nedergaard and Lupyan compared people who reported very little inner speech with people who reported a near-constant inner voice. The low-inner-speech group did worse at remembering lists of words and at judging whether picture names rhymed. But when only participants who said they had repeated the words aloud were compared, the memory gap disappeared. The researchers noted that these differences "may often be masked because people with anendophasia use alternative strategies."

The same is true at the visual end. Adam Zeman's team at the University of Exeter coined the term aphantasia in 2015 for people with no voluntary visual imagery. Large recent samples put full aphantasia at around 1 percent of people. Someone with aphantasia could not count the way Tukey did, but would very likely find another way to count. That is the real lesson. People differ in which codes come most easily, but most of us are mixtures, and we reach for whatever code is free.

Feynman's coloured equations and why explanations fail

The chapter ends with a detail that is often left out. Feynman wrote that when he looked at equations, he saw the letters in colour: light-tan j's, violet-bluish n's, dark brown x's. That description is consistent with grapheme-colour synaesthesia, which Julia Simner and colleagues found in about 1 percent of people when they screened a large sample with objective tests. Feynman was never tested, so it is best treated as a striking self-report rather than a diagnosis.

His reason for mentioning it was teaching. He pointed out that every idea is explained in terms of earlier ideas, which eventually rest on something as basic as counting — "which can be so different for different people." While lecturing on Bessel functions, he wondered "what the hell it must look like to the students." That may be the most practical insight in the whole story. A clear explanation in one person's code can be noise in another's.

Thinking in codes through a cognitive profile lens

The Feynman–Tukey story maps neatly onto three of the seven dimensions in the CognitionType framework.

Memory and sequencing. Feynman's counting lived in the part of working memory that holds and rehearses verbal sequences. That is why speech knocked it out and reading did not. If you lose your place when someone talks to you mid-task, you may be leaning heavily on that same verbal rehearsal system. Our guide to working memory covers how much of daily life runs through it.

Visual processing. Tukey's moving tape of numbers, and Feynman's own sock squares, show visual processing doing sequencing work. Some people naturally hold lists, dates and plans as spatial layouts or images. For them, visual clutter is often more disruptive than background chatter.

Attention and rhythm. The most underrated finding in the story is that Feynman's rate stayed steady and only paused when a hard word "needed more brains." Keeping an internal rhythm, and noticing when attention has been briefly pulled away, is a skill of its own. Simon Grondin and colleagues have shown that explicit counting makes time estimates more precise than intuitive guessing, which matches Feynman's observation that his unaided sense of a minute was "very irregular."

Seen this way, the question is not "Are you a word person or a picture person?" It is which systems carry your thinking in a given task, and what happens when two jobs compete for the same one.

How to find out which code you count in

Run the experiment yourself. Time a silent count to 60 a few times to find your baseline. Then count while reciting something aloud, and again while reading. Whichever one breaks your rhythm points to the code you were using. Try it with several kinds of tasks, not just counting, because the answer can change.

Notice when you switch. If one approach is blocked, see what you do instead. Feynman's sock patterns show that flexibility is often more telling than a single preference.

Design around interference, not labels. If talking breaks your concentration, protect quiet time for sequence-heavy work. If visual clutter derails you, clear the screen rather than the room. Avoid "learning styles" thinking. Harold Pashler and colleagues found no good evidence that teaching to a preferred style improves learning, so use these insights to reduce conflict between tasks rather than to limit how you learn.

Look at your broader profile. If you want a structured view of how memory, visual processing and attention work together in your own mind, CognitionType is a useful tool for understanding your own cognitive profile across seven dimensions. It complements formal assessment rather than replacing it.

What the counting experiment really teaches

Feynman did not prove that humanity splits into verbal and visual thinkers. He showed something subtler: two brilliant people, doing what looked like the same simple task, were running it on completely different machinery. Neither could have guessed this from the inside.

More than eighty years on, the research agrees with the heart of that insight and corrects the edges. People differ in the codes they think in. Those codes can be detected by what interferes with them. And most minds switch codes more often than their owners realise.


CognitionType is an informational assessment, not a clinical diagnosis. If you suspect a specific condition such as a learning difference, attention difficulty or working memory difficulty, we encourage you to seek a formal evaluation from a qualified professional.

Discover your own cognitive profile across 7 dimensions.

Take the free assessment