The First Machine That Could Teach

The Evolution of EdTech — Part 1 of 13

By Clarence Stephen | March 2026

2,400 words | 10 min read

In the spring of 1924, a psychology professor named Sidney Pressey walked into the American Psychological Association's annual meeting carrying a contraption the size of a portable typewriter. It had a small window, four keys, and a mechanical counter hidden in the back. It looked like something a patent office would politely decline.

He called it a "machine for intelligence testing." What he had actually built — though the full implications would take decades to appreciate — was the first device in human history designed to teach a human being without another human being in the room.

That sentence deserves a moment. For roughly five thousand years of recorded civilization, education had required a person. A teacher, a tutor, a parent, a priest, a monk bent over manuscripts by candlelight. Socrates needed an audience for his dialogues. The one-room schoolhouse needed its exhausted instructor juggling eight grade levels at once. The constant across every culture, every century, every pedagogical philosophy was the presence of another human mind.

And then there was Pressey's box.

How It Worked

The machine was elegant in its simplicity. A typewritten multiple-choice question appeared in the window — something like: "To help the poor debtors of England, James Oglethorpe founded the colony of (1) Connecticut, (2) Delaware, (3) Maryland, (4) Georgia."

The student pressed one of four keys. If they chose correctly, the machine advanced to the next question. If they chose incorrectly, the question stayed. The student had to keep trying until they got it right.

This was not testing. This was teaching. And in that distinction lay something that modern learning scientists would spend the next century validating: immediate feedback is among the most powerful forces in human learning. When you know instantly whether you are right or wrong, your brain encodes the correct information more deeply. When you must actively retrieve the right answer rather than passively receive it, the memory consolidates.

Pressey received a patent in 1928 and a second for an improved version in 1930. He published papers describing how the machine handled drill material, automatically dropping questions from the rotation once a student answered them correctly twice in succession. He was, without knowing the terminology, inventing spaced repetition and mastery-based learning — concepts that would not be formally named for decades.

The Industrial Revolution of the Mind

What made Pressey's vision radical was not the machine itself but the philosophy animating it. In a 1932 paper, he argued that education stood on the verge of an "industrial revolution" — that just as machines had transformed manufacturing, they could transform learning. The drudgery of drill and practice could be offloaded to machines, freeing teachers to do what only humans could do: inspire, mentor, guide.

"The first thing that the coming industrial revolution in education will do," Pressey wrote, "is free the teacher from the drudgery of routine drill and information-fixing."

That sentence was written in 1932. Swap "industrial revolution" for "AI transformation" and it could headline a venture capital pitch in 2025. The man was not wrong. He was early by a century.

But the revolution never came — not on his timeline. The Great Depression crushed funding for educational experimentation. Schools could barely afford chalk, let alone mechanical teaching devices. Pressey's machines were manufactured in small quantities, tested in a handful of classrooms, and quietly shelved. By the mid-1930s, he had largely abandoned the project. He would later reflect with considerable bitterness that the educational establishment simply was not ready for what he had built.

A familiar refrain, as this series will demonstrate.

Enter Skinner

Two decades later, the idea resurfaced in the hands of a far more famous — and far more polarizing — psychologist.

In the fall of 1953, B.F. Skinner visited his daughter Deborah's fourth-grade classroom at Shady Hill, a private school in Cambridge, Massachusetts. What he observed appalled the man who had spent his career studying how behavior is shaped by consequences. Some students had finished their work and sat fidgeting with boredom. Others struggled in silence, falling further behind with every passing minute. The teacher, however skilled, could not possibly attend to thirty individual learning trajectories simultaneously.

Skinner saw the problem in the precise terms of operant conditioning: the classroom violated nearly every principle of effective learning. Feedback was delayed by days. Reinforcement was inconsistent. Individualization was nonexistent.

He went home and built a teaching machine.

Skinner's device was more sophisticated than Pressey's. Paper discs fitted inside a rectangular wooden box with a hinged metal lid. The student read a question, composed an answer — not multiple choice, because Skinner believed constructed responses produced deeper learning than recognition — and slid a lever to reveal the correct answer. Right answers advanced; wrong answers prompted study of the correct response before moving on.

The key principles were deceptively simple:

  1. Small steps. Material was broken into tiny increments ("frames"), so students almost always succeeded.
  2. Active responding. Students produced answers rather than selecting them.
  3. Immediate feedback. Every response was confirmed or corrected instantly.
  4. Self-pacing. Each student moved at their own speed. No one waited. No one was abandoned.

Skinner called this "programmed instruction," and he was characteristically unrestrained about its implications. In his landmark 1954 paper "The Science of Learning and the Art of Teaching," he argued that the typical classroom was a pedagogical disaster — that it was frankly astonishing students learned anything at all given how badly the environment was designed.

The Boom

By the late 1950s, teaching machines were a cultural phenomenon. Time, Life, and Newsweek ran features. Companies rushed to manufacture devices. In 1958, Skinner installed a dozen machines in a self-study room at Harvard, where they were used to teach an undergraduate natural sciences course. Students using the machines performed as well as or better than those in traditional lectures.

The movement exploded. By 1963, over 60 companies were producing teaching machines or programmed textbooks. The U.S. military — perpetually hungry for efficient training methods — invested heavily. The Air Force, Navy, and Army developed programmed instruction curricula for everything from electronics repair to foreign language acquisition.

For a brief, shining moment, it seemed like Pressey's industrial revolution had finally arrived.

The Bust

It had not.

By the late 1960s, the teaching machine movement was in collapse. The reasons were numerous and instructive — a cautionary tale that would repeat itself, with unsettling precision, in every subsequent wave of educational technology.

The machines were boring. Programmed instruction, with its rigid frame-by-frame lockstep, turned out to be mind-numbingly tedious. The small-step approach ensuring students were "right" 95% of the time meant bright students felt patronized while struggling students still struggled — just with better posture.

The content was bad. In the rush to capitalize on the craze, publishers churned out poorly designed programs. Writing effective programmed instruction turned out to be extraordinarily difficult — far harder than writing a traditional textbook. Most attempts were neither engaging enough to hold attention nor structured enough to produce learning.

Teachers resisted. Despite assurances that machines would liberate teachers, many educators saw an existential threat. The American Federation of Teachers was skeptical. Numerous teachers felt that machines reduced education to rote memorization, stripping away everything that made the profession meaningful.

The research was mixed. Early studies showed promising results, but as more rigorous evaluations accumulated, the picture grew muddier. Programmed instruction worked well for factual recall but faltered with complex reasoning, creativity, or the higher-order thinking that matters most.

By 1970, the teaching machine was essentially dead as a commercial product.

What They Got Right

Here is what is remarkable about Pressey and Skinner, viewed from a century's distance: they were not wrong about the problems. They were not even wrong about most of the solutions. They were wrong about the technology.

A mechanical device with paper discs and typewritten questions could not deliver the kind of responsive, adaptive, endlessly patient instruction they envisioned. The idea was right. The medium was not.

Consider what they were trying to build:

If that reads like the pitch deck for every AI tutoring startup in 2025, that is because it is. The vision Pressey articulated in 1924 and Skinner refined in 1954 is, almost point for point, the vision that Khan Academy, Duolingo, and a hundred others are pursuing today — only now with large language models, adaptive algorithms, and screens in every pocket instead of mechanical counters and paper discs.

The Lesson

Every wave of educational technology follows the same arc: wild optimism, rapid adoption, inevitable disappointment, and then slow, partial integration. The teaching machine era established this cycle. It also established something more durable: the aspiration. The belief that technology could make education more personal, more effective, more equitable.

Pressey died in 1979, largely forgotten outside academic psychology. Skinner died in 1990, famous for many things but rarely credited for his educational innovations. Neither lived to see the personal computer, the internet, or artificial intelligence fulfill even a fraction of what they had imagined.

But every adaptive learning algorithm, every spaced-repetition flashcard app, every AI tutor that patiently waits for a student to construct an answer before offering feedback — every one of these is a direct descendant of a wooden box with four keys, presented at a psychology conference in 1924 by a professor who believed machines could teach.

He was right. It just took a hundred years for the machines to catch up to the idea.

Next in the series: "Television Was Going to Save Education" — how the most powerful communication technology of the 20th century was going to revolutionize the classroom. Spoiler: it didn't.

Sources

Clarence Stephen is the founder of Readify, an AI-powered literary platform. Yale B.Sc. Physics. Previously: Tiger Global, Morgan Stanley. Learn more at ireadifybooks.com

What could AI do for your classroom?

Readify uses local AI to adapt reading material to each student's level — generating illustrations, simplifying vocabulary, and tracking comprehension in real time. No student data leaves the building.

Bring Readify to Your School Try the Platform

Liked this article?

Get the full Evolution of EdTech series delivered to your inbox, plus weekly insights on AI in education.

Subscribe to Readify Branches