What Is the Zone of Proximal Development?
The zone of proximal development (ZPD) is the range of tasks a learner can't yet do alone but can do with help. Below it sit the things you can already do on your own. Above it sit the things that stay out of reach even with a guide. The psychologist Lev Vygotsky argued that teaching is most useful when it aims at that middle band.
His own definition, from Mind in Society (1978), is the one most textbooks quote. The zone, he wrote, is:
"the distance between the actual developmental level as determined by independent problem solving and the level of potential development as determined through problem solving under adult guidance or in collaboration with more capable peers."
It helps to picture three zones, one inside the other:
- What you can do alone. You solve it, explain it or perform it without anyone's help. Practice here feels good and teaches you little that's new.
- What you can do with help. You get there with a hint, a demonstration, a worked example or a partner who knows more. This is the zone of proximal development.
- What you can't do yet, even with help. The gap is too wide for a hint to bridge. Help at this level mostly produces copying.
The zone also moves. Once you can do alone what used to need help, the edge has shifted outward. Vygotsky put the idea in one line in Thought and Language: "What the child can do in cooperation today he can do alone tomorrow."
Where the Idea Came From: Vygotsky and the Problem With Test Scores
Lev Semyonovich Vygotsky was born in 1896 in Orsha, in what is now Belarus, and did most of his work as a psychologist in Moscow in the late 1920s and early 1930s. He died of tuberculosis in 1934, at 37. The zone of proximal development comes from the last years of that short career.
The Russian phrase is zona blizhaishego razvitiya. "Proximal" was the translators' choice. A more literal rendering is "zone of nearest development," which arguably says it better: it's the development that's closest to happening.
Vygotsky used the idea partly to criticize how children were tested. The intelligence tests of his day measured what a child could do unaided, and that number was treated as the child's level. He asked his readers to imagine two ten-year-olds who both tested at a mental age of eight. Same score. With a little assistance, one could handle problems up to a twelve-year-old's level, while the other got only as far as a nine-year-old's. Their zones differed, he argued, and that mattered more for teaching them than the score they shared.
The same goes for adults: a test of what you can do alone shows where you are, but what to study next depends on how far the right help could take you.
Most English readers met the idea decades later, through Mind in Society: The Development of Higher Psychological Processes (Harvard University Press, 1978). Vygotsky didn't write it as a book: it's a selection of his writings edited by Michael Cole, Vera John-Steiner, Sylvia Scribner and Ellen Souberman.
ZPD vs Scaffolding: The Word Vygotsky Never Used
"Scaffolding" is so tied to the ZPD that many people assume it's Vygotsky's term. It isn't. The term is usually traced to a 1976 paper by David Wood, Jerome Bruner and Gail Ross, "The role of tutoring in problem solving," published in the Journal of Child Psychology and Psychiatry.
In that study, Ross tutored 30 children aged three to five as they built a pyramid from 21 interlocking wooden blocks, a task too hard for them to manage alone. Watching the tutor, the authors listed six functions of good scaffolding:
| Function | What the helper does | What it looks like when you study alone |
|---|---|---|
| Recruitment | Gets the learner interested in the task | Start from a question you actually want answered |
| Reduction in degrees of freedom | Simplifies the task until it's manageable | Take one subproblem or one section at a time |
| Direction maintenance | Keeps the learner moving toward the goal | Write the goal at the top of your notes |
| Marking critical features | Points out what matters | Highlight the one sentence that carries the argument |
| Frustration control | Keeps struggle from turning into giving up | Set a time limit before you allow yourself a hint |
| Demonstration | Models a solution for the learner to imitate | Study one worked example, then try a fresh problem |
What separates scaffolding from ordinary help? Reviewing a decade of studies in 2010, Janneke van de Pol, Monique Volman and Jos Beishuizen found three features that kept coming up:
- Contingency. The help is matched to what the learner is doing right now: more when they struggle, less when they don't.
- Fading. The help is gradually withdrawn.
- Transfer of responsibility. The learner ends up doing the work on their own.
Fading matters because help that suits a beginner can get in the way later. Cognitive load researchers call this the expertise reversal effect. Slava Kalyuga, Paul Ayres, Paul Chandler and John Sweller described it in 2003: techniques that work well for novices can lose their benefit, or even backfire, for more experienced learners. A worked example that saves a beginner from overload is just extra reading for someone who already knows the steps.
Alexander Renkl and colleagues tested one way to fade that help in experiments published in 2002. Students studied worked examples in which the solution steps were removed one at a time, starting from the last, until they were solving whole problems themselves. That "backward fading" helped them more than the usual pairing of one full example with one problem to solve, at least on problems like the ones they had practiced.
A 2017 meta-analysis by Brian Belland and colleagues pooled 144 studies of computer-based scaffolding in science, technology, engineering and math education and found an average effect of g = 0.46 on cognitive outcomes, a moderate effect.
Who Can Help: Teachers, Peers and the "More Knowledgeable Other"
Vygotsky's definition names two kinds of helper: "adult guidance" and "more capable peers." You'll often see this helper called a "more knowledgeable other," or MKO. That label is a later shorthand from education writing, not a phrase from his definition.
Don't skip the peer half. Someone a little ahead of you still remembers what was confusing, which an expert may have forgotten. Tutoring research shows the benefit runs both ways. In a 1982 meta-analysis of school tutoring programs, Peter Cohen, James Kulik and Chen-Lin Kulik found that tutored students outperformed comparison students in achievement, with an average effect size of 0.40, and that the tutors themselves gained too, at 0.33. That second number is the protégé effect: explaining something to someone a step behind you is a way of learning it yourself.
One-to-one tutoring is the classic picture of the ZPD in action, though it works less well than its most quoted number suggests. Benjamin Bloom reported in 1984 that tutored students scored about two standard deviations above students in a conventional class, which would put the average tutored student ahead of 98% of the comparison group. When Kurt VanLehn reviewed the tutoring research in 2011, he found a smaller effect for human tutors, d = 0.79, and almost the same, d = 0.76, for computer tutors that coach a student step by step.
The helper doesn't have to be in the room. For a self-learner, the "more capable" partner can be:
- A worked example in a textbook, which demonstrates the move you can't make yet.
- A transcript or a glossary, which turns a lecture or a text that's slightly too fast into one you can follow.
- Other readers' highlights. Someone who read a hard article before you and marked the passages that matter has done some of the "marking critical features" for you. On web pages, Glasp's browser extension has a Community tab whose Popular view lists the passages two or more readers highlighted, and you can browse more of them in Glasp's community.
Zone of Proximal Development Examples
Each row shows one learner working on one skill. The middle column is the zone.
| Learner | Can do alone | Can do with help (the ZPD) | Out of reach for now | A scaffold that fits |
|---|---|---|---|---|
| Six-year-old learning to read | Sounds out short, familiar words | Reads a new picture book when an adult prompts the tricky words | Reads a chapter book alone | Shared reading, pointing at each word |
| Adult learning Spanish | Orders food and follows a slow podcast | Follows a native-speed podcast with a transcript | Follows a fast-paced film with no subtitles | A transcript first, then the audio alone |
| Beginner programmer | Writes a loop by following a tutorial | Finds the bug in their own code when a mentor asks guiding questions | Designs a large system from scratch | Questions like "What did you expect this line to print?" |
| Student reading a hard book | States a chapter's main claim | Reconstructs the chapter's argument with a study guide | Critiques the book against the rest of the field | Chapter questions and other readers' highlights |
| Junior doctor | Takes a patient's history | Reaches a diagnosis while thinking aloud with a senior colleague | Manages a rare, complex case alone | Case discussion and checklists |
Language learning has its own version of this. Stephen Krashen's input hypothesis proposed that we acquire a language by understanding input pitched a little beyond our current level, which he wrote as "i + 1."
In a 2000 study, Hu Hsueh-chao and Paul Nation gave learners a story in which 80%, 90%, 95% or all of the words were familiar. At 80%, none of the 16 readers understood it adequately, and the authors estimated that most learners need about 98% coverage to read fiction without help. A 2023 replication found no sharp line at 98%: understanding simply rose with the share of words known. A text where you know nearly every word and need a glossary for the rest is in your zone, and one where you look up every fifth word is not. Our guide on how to learn a language with AI covers ways to find input at that level.
How Hard Should Learning Feel? What the Research Says
Vygotsky described the zone in terms of help, not difficulty. For anyone studying alone, though, the practical question is how hard it should feel.
The cognitive psychologist Janet Metcalfe studied how people choose what to study, and called her answer the region of proximal learning, a name that echoes Vygotsky's. In her experiments, learners skipped what they already knew and went first for the unlearned items they felt closest to learning. Experts reached for harder items than novices did, and people moved on to harder items when they had more time. In a small 2003 experiment with Nate Kornell, giving most of the study time to medium-difficulty items produced better recall (32%) than giving it to the hardest ones (25%).
In a 2019 paper in Nature Communications, Robert Wilson, Amitai Shenhav, Mark Straccia and Jonathan Cohen showed mathematically that for learning algorithms that improve by gradient descent on two-choice tasks, learning was fastest when the training error rate was about 15.87%, or roughly 85% correct. They called it "The Eighty Five Percent Rule." It held in simulations of artificial neural networks and in a model of perceptual learning in animals. Under other assumptions about noise, the best accuracy came out nearer 82% or 75%.
Nobody has shown that human students learn best at exactly 85%, and the authors note that the rule still has to be generalized beyond two-choice tasks and the algorithms they studied. Still, if you're getting nearly everything right, the material is too easy, and if you're getting most of it wrong, it's too hard.
This fits Robert and Elizabeth Bjork's research on desirable difficulties: some struggle makes learning stick, as long as it's struggle you can overcome. The ZPD supplies the condition. A difficulty is desirable when it sits inside your zone, where effort plus a little help gets you over it.
Use these signals to judge where you are:
| What you notice | Where you probably are | What to do |
|---|---|---|
| You get almost everything right, quickly | Below the zone | Raise the level, put the notes away, add a time limit |
| You get most of it right with effort, and a hint gets you the rest | In the zone | Stay here and cut back on hints over time |
| Hints don't help, and you're copying answers | Above the zone | Go back to the prerequisite, or find an easier source |
How to Find Your Own Zone When You Study Alone
Teachers find a student's zone by watching how much help they need. When you study alone, you have to run that test on yourself. This routine works for a chapter, a paper or a problem set.
- Try it cold. Before you reread anything, answer the end-of-chapter questions, explain the main idea on a blank page, or attempt the problems. That's your "alone" level. Wrong answers here aren't wasted: in experiments by Nate Kornell and colleagues (2009), an attempt made before seeing the answer improved learning even when it failed.
- Sort what you missed into two piles. "With a hint" is anything you got partly right, or understood the moment you saw the answer. "Not yet" is anything that still doesn't make sense after you read the solution. The first pile is your zone.
- Spend your help on the first pile. Find the smallest piece of help that gets you past each item: one worked example, one paragraph of explanation, one definition. Highlight the passage that did it so you can find it again.
- Send the second pile back a step. When a topic lands in "not yet," a prerequisite is usually missing. Study that first, or find an easier source on the same topic.
- Fade the help. A day or two later, try the same kinds of items with the help out of sight. Anything you can now do alone has left your zone, which is the goal.
- Move the edge. When the "with a hint" pile shrinks, raise the level: harder problems, a more advanced text, or less time.
A few tools make the piles easier to keep. With Glasp's web highlighter, you can mark up web pages and PDFs and add a note to any highlight, such as "needed a hint" or "not yet," so every item in a pile stays next to its source. For recorded lectures, YouTube Summary shows the full transcript next to the video, a standard scaffold for a talk that moves too fast. And when you want a hint, Glasp's AI chat answers from your own saved highlights and links each answer to the highlights it drew on.
AI Tutors and the ZPD: Help That Never Fades
An AI model can explain almost anything at almost any level, at any hour. In Vygotsky's terms, that's a more capable partner who never leaves. The risk is in the other half of scaffolding: an AI that answers every question in full never fades.
The clearest evidence comes from a 2025 study in PNAS by Hamsa Bastani and colleagues, run with nearly 1,000 students at a high school in Turkey. With a standard GPT-4 chat open, students did 48% better on practice problems, and with a tutor version that gave hints instead of answers they did 127% better. On the exam, taken without AI, the standard GPT-4 group scored 17% below the control group. The hint-giving tutor largely avoided that harm.
Read through the ZPD, the plain chatbot did the work inside the students' zone for them, and they learned less than classmates who did that work themselves.
A Harvard trial published the same month shows the other side. Gregory Kestin and colleagues built a physics tutor around research-based teaching practices, including step-by-step scaffolding. The median learning gain of students who used it was more than double that of students in an active learning class, and they got there in less time. We cover both trials in detail in Do AI Tutors Actually Work?.
You can make a general-purpose AI behave more like a scaffold by telling it how to help:
- "Don't give me the answer. Ask me one question at a time."
- "If I'm stuck after two tries, give me the smallest hint you can."
- "Before you explain, ask me what I already know about this."
- "Quiz me on this without my notes, then tell me which answers needed a hint."
Then do the fading yourself. Close the chat and try the next problem without it.
Common Misconceptions About the Zone of Proximal Development
"The ZPD is a fixed level a person has." In everyday use, the zone depends on the task and on the help available, and it moves as you learn. Your zone in statistics might be multiple regression while your zone in French is still ordering coffee.
"It's the same as teaching to someone's learning style." The ZPD is about what a learner can do with and without help, not about a preference for pictures or sound.
"The ZPD applies to any task, at any age." Vygotsky developed it to describe children's development, and some scholars think that's where it belongs. Seth Chaiklin argued in 2003 that the popular reading, any task a person can do with help, strays from Vygotsky's narrower meaning, which concerned psychological functions that are maturing as a child moves from one stage of development to the next. In his words, the zone "is not concerned with the development of skill of any particular task, but must be related to development." Using it for adult self-study, as this article does, extends the idea beyond what Vygotsky claimed.
Frequently Asked Questions
What is the zone of proximal development in simple terms?
The zone of proximal development is the set of things you can't do alone yet but can do with help. Below it is what you can already do; above it is what's still out of reach. Vygotsky argued that teaching works best when it aims at this middle zone.
What is an example of the zone of proximal development?
A child who can sound out short words but can't yet read a new picture book alone can read it when an adult prompts the tricky words. That shared reading sits in the child's zone. For an adult, it might be following a native-speed podcast in a new language with the help of a transcript.
What is the difference between the ZPD and scaffolding?
The ZPD is the gap between what a learner can do alone and what they can do with help. Scaffolding is the help itself, when it's adjusted to the learner and gradually withdrawn. The term "scaffolding" is usually traced to Wood, Bruner and Ross (1976), not to Vygotsky.
What does "more knowledgeable other" mean?
A "more knowledgeable other" (MKO) is whoever provides the help: a teacher, a parent, a peer or, in modern use, a book or a tool. Vygotsky's own wording was "adult guidance" or "collaboration with more capable peers." The MKO label came later.
Who created the zone of proximal development?
Lev Vygotsky, a Soviet psychologist who worked in the 1920s and early 1930s and died in 1934. The idea reached most English readers through Mind in Society (1978), an edited collection of his writings.
How do you find your zone of proximal development?
Try the material alone first, then note which mistakes you can fix with a small hint and which you can't fix even with the answer in front of you. The first group is your zone. Work there with as little help as you need, then test yourself again without help a day or two later.
Can AI act as a more knowledgeable other?
It can, when it behaves like a scaffold. An AI that asks questions and gives small hints keeps you working inside your zone. One that hands over full answers does the work for you. In one 2025 trial, students who practiced with plain GPT-4 did 17% worse on the exam once the AI was gone.
What are the criticisms of the zone of proximal development?
A common one is that it's vague: a zone is hard to measure precisely. Scholars also disagree about how far the idea reaches beyond child development, where Vygotsky applied it. Even so, it's a good guide to choosing what to study and how much help to accept, and that choice matters more now that AI offers help at all hours.
Learn at the Edge, Then Let Go of the Help
In practice, working in your zone of proximal development comes down to a habit. Find what you can almost do, get the smallest amount of help that gets you there, then practice without the help until you don't need it, and move on to the next edge.
To keep that loop in one place, use Glasp to save the passages and transcript lines that got you unstuck, along with your notes and your Kindle highlights. When you need a hint, ask Glasp's AI chat, which answers from those sources, then put it away and try again alone.
