Miller’s Law (7±2): An S-Tier Behavioral Designer’s Guide

Ask a designer to name a psychology “law” and most reach for the same one: Miller’s Law, the magical number seven. Menus should have seven items. Navigation bars, seven links. Bullet lists, seven points. It is the most quoted number in design, repeated in style guides, design-school lectures, and product reviews as if George Miller had carved it into stone.

It is also the wrong lesson pulled from the right paper.

Miller never said interfaces should cap at seven of anything. He did not even believe seven was a hard limit. He opened his 1956 paper by joking that the number had been “persecuting” him, half-wondering aloud whether it was a Pythagorean coincidence. What he actually found is far more useful than a magic number, and almost nobody quotes it: the mind does not count items, it counts chunks. And you get to decide how much fits by deciding how to group it.

That single shift separates a product people can hold in their heads from one they bounce off of. Group a process well and sixteen digits feel like four easy steps. Group it badly and four steps feel like sixteen. This guide is about the discovery underneath the slogan, why the number itself is mostly wrong, and how to design for the bottleneck that is actually there.

Speed Run Notes

• Miller’s Law is misremembered as “seven items.” The real finding is that working memory holds a few chunks, and chunking is the design lever, not the number.
• The number is a decoy. Cowan’s 2001 reanalysis puts the real limit closer to four chunks; Miller himself treated “seven” as rhetorical, not a measurement.
• Chunking is elastic. With recoding, one man pushed his digit span from 7 to 79; chess masters hold 20+ pieces in roughly 5 chunks. Expertise is better chunking.
• The “menus need 7 items” rule is a myth. A visible menu is recognition, not recall, so it never enters the working-memory bottleneck Miller measured.
• It is not a Core Drive. It is a capacity law sitting underneath the Octalysis Core Drives: motivation decides whether someone wants to act; chunking decides whether the information fits so they can.
• White Hat: chunk the user’s complexity so it fits. Black Hat: flood working memory until people give up and accept the default.

Table of Contents

Author Credibility: Yu-kai Chou

Yu-kai Chou created the Octalysis Framework after studying gamification since 2003 — years before the term entered mainstream vocabulary. As a Human-Systems Architect & Behavioral Designer, his framework has been applied by LEGO, Microsoft, Porsche, Coca-Cola, Salesforce, and MrBeast, impacting over 1.5 Billion Users.

Chou has taught the Octalysis methodology at Harvard, Stanford, Yale, Tesla, Google, BCG, and IDEO.

His work has been cited by Harvard, Stanford, MIT, Forbes, Wall Street Journal, Wired, US Department of Energy, NIST, NSF, NCBI, US Department of Education, ClinicalTrials.gov, and Google Scholar — with 3,700+ more academic publications. Explore his books here.

What Is Miller’s Law?

Miller’s Law states that the number of objects an average person can hold in immediate (short-term) memory is about seven, plus or minus two. The phrase comes from psychologist George A. Miller’s 1956 paper, “The Magical Number Seven, Plus or Minus Two,” published in Psychological Review. It is one of the most cited papers in the history of psychology, and one of the most consistently misread.

The popular version stops at the number. The real paper is about a bottleneck and a workaround. Miller wrote at the moment Claude Shannon’s information theory was sweeping through psychology, and he framed the mind as a communication channel with a measurable capacity. He noticed that across very different tasks, such as judging the pitch of a tone, the position of a dot, or the saltiness of a solution, people could reliably sort stimuli into only about seven categories before they started making mistakes. He called this the limit on “absolute judgment.” Separately, he noticed that immediate memory also seemed capped at around seven items. The coincidence of the two sevens is what gave the paper its playful title.

Here is the part that matters for anyone building a product, a curriculum, or a game. Miller argued that absolute judgment is limited by the amount of information, measured in bits, while immediate memory is limited by the number of items, regardless of how much information each item carries. That distinction is the whole ballgame. If memory counts items rather than bits, then you can smuggle far more information past the limit by packing more into each item. The technical name Miller gave that packing is the chunk.

So the honest one-sentence version of Miller’s Law is not “seven items.” It is this: working memory holds a small, fixed number of chunks, and the size of a chunk is up to you. The designer’s job is not to count to seven. It is to build better chunks. Everything else in this guide follows from that one correction.

It is worth pausing on the word “law,” because it sets the wrong expectation. Miller’s Law is not a law in the sense that gravity is a law. It is a well-established empirical regularity, a finding that shows up reliably across people and tasks, with real exceptions and a moving boundary. Calling it a law gives it a false air of precision that fueled the “exactly seven” cult in the first place. Treat it the way you would treat a strong rule of thumb backed by seventy years of evidence: trustworthy as a constraint to design within, dangerous as a number to obey literally. The constraint is genuine. The decimal is decoration.

The Number Is a Decoy: What Miller Actually Discovered

Strip the famous number away and three findings remain, each more useful than “seven.”

Memory is item-bound, not information-bound

A random binary string like 1 0 0 1 1 1 0 0 1 0 has ten items. Recode it into pairs and read those pairs as base-four or as familiar two-digit numbers, and you have shrunk ten items into five, then into two or three. The information did not change. The number of memory slots it consumes did. Miller’s own example was binary digits recoded into octal, where a trained subject could repeat far more of them than the raw binary count should allow. The principle is identical to reading a long number as a date, a price, and an area code instead of a wall of figures. Recoding is free capacity, and it is the only kind you can manufacture on demand.

The seven that started it: absolute judgment

The first half of Miller’s paper is not about memory at all. It is about perception. When people are asked to label a single tone by its loudness, or a single line by its length, accuracy collapses once the number of categories climbs past about six or seven. Below that they almost never err; above it the mistakes pour in. This is the limit on absolute judgment, and it looks like a hard wall. Then Miller pointed out the escape. The wall only holds for one dimension at a time. Judge a stimulus that varies in pitch and loudness and duration at once, and total capacity climbs well past seven, because each dimension carries its own small budget and they add together. The lesson designers should take is not “limit everything to seven.” It is the opposite: when one channel is saturated, open another. A label plus a color plus a position communicates more than any one of them pushed to its limit.

Chunking moves the ceiling

Because the limit is counted in chunks and a chunk can be trained to hold more, the practical capacity of memory is not fixed at all. It expands with knowledge. This is the finding the slogan throws away, and it is the one with the most leverage for design, education, and expertise. Miller saw all three of these results in 1956, and he was explicit that the number itself was the least interesting part. The reason the number survived while the mechanism got lost is simple. “Make seven items” is an instruction you can follow without thinking. “Build better chunks” forces you to understand the material first. The easy lesson crowded out the true one.

Chunking: The Real Lever

A chunk is a unit of meaning that the mind treats as one thing. The letters F, B, and I are three chunks if you read them cold. “FBI” is one chunk because you recognize the agency. The chunk did not get smaller; your knowledge made three items collapse into one. That collapse is the entire mechanism behind every clever thing memory can do.

How a chunk gets built

Chunks are not found, they are made, and they get made in two ways. The first is recognition: when a pattern already lives in your long-term memory, incoming material snaps to it instantly, the way “FBI” or “2026” or a familiar chord does. The second is construction: when no pattern exists, you can deliberately impose one, which is what every mnemonic system does. Turning “4492” into “the year I started, plus a near-mile pace” forges a link between new digits and old knowledge. Both routes share the same requirement, which is meaning. You cannot chunk what you do not understand, and that is why the same screen looks like signal to an expert and noise to a beginner. The expert is not trying harder. The expert is recognizing.

Chunks are personal, which is why one-size design fails

A chunk only exists relative to a particular person’s knowledge, and that has a sharp design consequence. The string “1492” is one chunk to almost any adult and four loose digits to a young child. “ATP” is one chunk to a biologist and three letters to everyone else. There is no objective chunk size you can engineer for, because the same content carves into different numbers of chunks depending on who is reading it. This is the deeper reason the “seven items” rule fails: it pretends the items are fixed when the unit of measurement itself is subjective. The practical move is to design for the chunks your actual audience already owns. Use the conventions, groupings, and vocabulary they recognize, so material arrives pre-chunked rather than forcing them to build chunks on the spot. A finance dashboard for analysts can lean on terms that would overwhelm a consumer app, not because analysts have bigger memories, but because they walk in with the chunks already loaded. Know your reader’s existing patterns and you can pour far more through the same narrow channel.

The man who remembered 79 digits

The cleanest demonstration is a 1980 study by K. Anders Ericsson, William Chase, and Steve Faloon, published in Science. Their subject, an undergraduate runner known as SF, started with an ordinary digit span of about 7. Over roughly two years and more than 230 hours of practice, he pushed it to 79 digits. He did not grow a bigger memory. He was a competitive runner, so he started hearing strings of digits as race times. “3492” became “3 minutes 49.2 seconds, near world-record mile pace.” Four items became one chunk. He stacked chunks into groups, and groups into supergroups, until a string that should have been impossible fit through a normal-sized bottleneck. When the experimenters switched from digits to letters, his span collapsed straight back to about six. The skill was the chunking system, not the storage. SF had not upgraded the hardware. He had written better software for it.

Why chess masters see more

The same pattern shows up in expertise. In their 1973 paper “Perception in Chess,” Chase and Simon showed master players a real game position for five seconds, then asked them to reconstruct it. Masters placed twenty or more pieces almost perfectly; novices managed a handful. The obvious conclusion, that masters have better memories, is wrong. When the same pieces were scattered randomly across the board, the masters’ advantage vanished. They performed no better than beginners. Their memory was not larger. They had thousands of meaningful board patterns stored in long-term memory, so a real position arrived pre-chunked: not twenty pieces but five familiar structures. Take away the meaning and you take away the chunks. This is why ten thousand hours of deliberate practice changes what a person can perceive, not just what they can do.

The phone number in your pocket

You use this every day without naming it. A phone number is not ten digits. It is an area code, a prefix, and a line number, which is three chunks, formatted with dashes or spaces so your eye groups them for you. Credit cards travel in groups of four. Software license keys come dashed into blocks. The formatting is not decoration. It is the designer doing the chunking on the reader’s behalf so the reader does not have to. Strip the dashes out of a phone number and watch how much harder it becomes to read back, even though not a single digit changed.

Herbert Simon returned to the question directly in his 1974 Science article, “How Big Is a Chunk?” His estimate landed in the same neighborhood as Miller’s for the number of chunks, but with a crucial addition: moving a chunk into long-term memory takes time, on the order of several seconds each. Chunks are not free. They have to be built, and building them is the slow, deliberate work that separates a novice from an expert. The capacity limit is real. The escape hatch is learning, and learning is not instant.

Why “Seven” Is Wrong: Cowan’s Magical Number Four

If the number was always shaky, what is the honest figure? The most careful modern answer comes from Nelson Cowan’s 2001 paper in Behavioral and Brain Sciences, pointedly titled “The Magical Number 4 in Short-Term Memory.” Cowan combined decades of studies that controlled for the tricks people use to inflate their scores, including silent rehearsal, grouping, and long-term knowledge, then asked how much pure capacity remains when you strip those away. The answer was about four chunks, not seven.

Cowan’s argument is not that Miller was sloppy. It is that Miller’s seven included the rehearsal and chunking people do automatically, so it measured performance rather than raw capacity. When you block rehearsal, for example by having someone repeat an irrelevant word while they try to hold the list, the ceiling drops to three or four almost every time. Seven was the practiced score. Four is closer to the bare machine running with no help.

For a designer, the gap between four and seven is not academic. It is the difference between assuming people can track seven things in their head and discovering the hard way that they can track three. If you are designing for the moment a user is mid-task, distracted, on a phone, with no chance to rehearse, plan for four. Treat seven as a best case that requires quiet attention you will almost never get. The conservative number is the one that protects the experience, because real use never happens in the calm of a memory lab.

This is also why arguing about the exact figure misses the point. Whether the true limit is four, five, or seven, it is small and it is counted in chunks. Those two facts carry every design implication worth having. The decimal point does not.

What Miller Got Right

It is fashionable to dunk on the number, so it is worth being precise about how much Miller got right, because it is most of it.

He correctly identified that immediate memory has a sharp, low ceiling. That was a genuinely surprising claim in 1956, when the dominant behaviorist mood avoided talking about internal capacity at all. He correctly separated the capacity of memory from the capacity of perception, two limits that look similar and have different causes. Most importantly, he identified recoding, which we now call chunking, as the fundamental way humans beat the limit, and he called it “the very lifeblood of the thought processes.” That sentence has aged better than the title. Half a century of research on expertise, skilled memory, and learning is, in a real sense, an elaboration of it.

There is a bigger reason the paper mattered. It landed in the same window as Chomsky’s work on language and Newell and Simon’s work on problem-solving, and together they helped pull psychology out of pure behaviorism and into the study of the mind as an information processor. Miller did not just measure a limit. He gave the emerging field of cognitive science one of its founding demonstrations that internal capacity is real, measurable, and worth modeling. He was also honest about his uncertainty in a way the slogan erased. He never claimed seven was a law of nature; he flagged it as a recurring number that demanded explanation, and the explanation he offered turned out to be the durable contribution. The man was more careful than his most famous sentence. We are the ones who flattened him into a number.

Where Miller’s Law Falls Apart

The science holds up. The folklore built on top of it does not. Three failures account for most of the damage.

The “seven items per menu” myth

The single most expensive misuse is the belief that on-screen menus, navigation bars, or option lists should be limited to seven items because of working memory. This is a category error. Miller measured recall, which means holding items in your head with nothing to look at. A menu is recognition: the options are right there on the screen. You do not store them; you scan them. Nothing enters the working-memory bottleneck, so the seven limit simply does not apply. A restaurant menu can list forty dishes. A settings page can show thirty rows. Capping them at seven to honor a misread paper throws away clarity for a rule that was never about visible choices. The real navigation question, how fast people can decide among options, belongs to Hick’s Law, not Miller’s, and the cure there is good categorization, not amputation.

The number that will not sit still

The second failure is treating seven as a constant when the data never supported one. Seven digits, six letters, five words, four chunks under interference: the figure moves with the material and the conditions every time it is measured. Any design rule pinned to a specific count inherits that instability. “Show no more than seven” is false precision dressed as rigor. The trustworthy version drops the number and keeps the shape. Keep the count of independent things small, and where it cannot be small, group them.

Expertise quietly breaks the ceiling

The third failure is forgetting that capacity is elastic. A novice and an expert looking at the same dashboard are not under the same load, because the expert sees chunks where the novice sees clutter. A pilot reads an instrument panel as a few patterns; a passenger sees dozens of dials. Design a system to the novice’s tiny capacity and you can cripple the expert with oversimplification and endless click-throughs. Design it to the expert’s chunked fluency and you lose the novice on day one. The limit is not one number to design against. It is a moving target that depends on who is looking, which means progressive disclosure and learnable structure beat any fixed cap.

What’s Really Happening Inside the Brain

Miller described a limit. Later researchers built the machine that produces it. The most influential account is the working-memory model proposed by Alan Baddeley and Graham Hitch in 1974, which replaced the older idea of a single short-term “store” with a small system of specialized components. A central executive directs attention. A phonological loop holds verbal and acoustic material, which is the inner voice you use to keep a phone number alive by repeating it. A visuospatial sketchpad does the same for images and locations. Each subsystem has its own tiny capacity, which is why you can rehearse a number and picture a route at the same time without much interference: different components, different budgets.

This model explains the wobble in Miller’s number. Verbal spans depend partly on how fast you can say the items. You can hold about as many words as you can rehearse in roughly two seconds, which is why short words give longer spans than long ones, an effect known as the word-length effect. The limit is not a fixed number of boxes. It is a decaying trace kept alive by rehearsal, racing against time. Chunking wins that race by reducing how many separate traces you have to refresh. For the full architecture, see the breakdown of Baddeley’s Working Memory Model.

Sitting on top of this hardware is the practical cost of using it. When the number of chunks a task demands exceeds the few your working memory can hold, the surplus does not get processed. It gets dropped, and the experience registers as confusion or fatigue. That overflow is exactly what Cognitive Load Theory formalizes for learning and instruction. Miller’s Law tells you the bottleneck exists. The working-memory model tells you how it is built. Cognitive load tells you what it costs when you ignore it. The three together turn a curiosity about memory into a budget you can design against.

Miller’s Law vs. Other Theories

Miller’s Law is often lumped in with a cluster of “limit” findings. They are siblings, not duplicates, and keeping them straight is what turns trivia into design judgment.

vs. Hick’s Law

Hick’s Law is about the time it takes to decide among visible options, and it scales with the number of choices. Miller’s Law is about how many things you can hold in mind when the options are not in front of you. Hick governs the menu you are looking at. Miller governs the steps you have to remember after the menu closes. A checkout flow can have many fields on screen, which is a Hick problem solved by grouping and good labels, while still failing if it asks you to remember a code from a previous screen, which is a Miller problem solved by carrying the code forward so nothing has to be held.

vs. Cognitive Load Theory

Cognitive Load Theory takes Miller’s bottleneck and turns it into a teaching method. It splits load into three kinds: the inherent difficulty of the material, the waste added by poor presentation, and the productive effort of building real understanding. Miller’s Law names the constraint; cognitive load gives you the budget categories for spending against it. Where Miller says “few chunks,” cognitive load says “so cut the wasteful load and let the learner’s slots go to the material that matters.”

vs. The Paradox of Choice

The Paradox of Choice is about the anxiety and regret that too many options create at the moment of decision. Miller’s Law is upstream of feeling. It is the raw capacity limit that makes large sets hard to compare in the first place. Choice overload is partly an emotional response; Miller’s bottleneck is the mechanical reason you cannot weigh twenty options against each other in your head without writing them down.

vs. Fitts’s Law

Fitts’s Law is the most distant cousin. It predicts the physical time to move to a target based on its size and distance, so it governs the body, not the mind. Together the three UX laws map the cost of an interface end to end: Miller for what you must remember, Hick for what you must decide, Fitts for what you must reach.

Miller’s Law in the Real World

Once you read capacity in chunks, you start seeing the law everywhere good design quietly respects it.

Onboarding and forms

The reason multi-step onboarding outperforms one giant form is pure Miller. A single screen with twenty fields is twenty things competing for a four-chunk budget. Break it into four steps of five related fields and each step fits, with a progress indicator carrying the state so nothing has to be held between screens. The total amount of work is identical. The felt effort drops because no single moment overflows. The best checkout flows go further and carry every value forward automatically, so the user never has to remember a thing they typed two screens ago. One-time passcodes are the cautionary version: a six-digit code that arrives in one app and must be retyped in another sits right at the edge of capacity, which is exactly why the better implementations now autofill it for you.

Education and instruction

Good teaching is chunking made visible. A skilled instructor introduces a few concepts, lets them consolidate into a single chunk through practice, then treats that chunk as one unit in the next lesson. Phone numbers, musical phrases, and grammar rules are all taught this way, with small groups first, then groups of groups. A curriculum that dumps fifteen new terms in one session is not demanding; it is broken, because fourteen of them fall out of the bottleneck before they can be encoded. The fix is not less content. It is better grouping and enough spacing for chunks to set, which is also why spaced repetition and worked examples beat a single dense lecture for almost any complex skill.

Interfaces and information design

Dashboards, tables, and reports live or die on grouping. A wall of forty metrics is unreadable; the same forty organized into five labeled sections of related metrics becomes scannable, because the section headers do the chunking. White space, dividers, and headings are not aesthetic flourishes. They are working-memory tools that tell the eye where one chunk ends and the next begins. The visible-options caveat still holds: the goal is not to hide things down to seven, but to group them so the few chunks the eye assembles each carry more. A pricing page with three named tiers beats the same features listed as a flat wall, not because three is magic, but because three labeled chunks are something a shopper can actually hold and compare.

Consider a concrete before and after. A settings screen lists eighteen toggles in one undivided column: notifications, privacy, billing, appearance, language, accessibility, and a dozen more, all flat. A user hunting for one option scans the whole wall every time, because nothing tells them where to look. Now group the same eighteen into four labeled sections of four or five related toggles each. The number of options has not dropped, but the search collapses, because the user holds four section names instead of eighteen toggles, then drills into the one that matters. This is the entire art of information design compressed into one move: never reduce the content to fit memory, restructure it so each thing the user holds carries more. The wall becomes a map the moment you add the chunk boundaries.

Product and game design

Games are relentless chunkers because they have to teach complex systems to players who will quit if overwhelmed. A new mechanic is introduced alone, drilled until it becomes automatic, then combined with the next. By the late game a player juggles systems that would be incomprehensible if shown on day one, because each one has consolidated into a single chunk. The same logic powers progress structures. A long goal feels unmanageable, while the same goal split into chunked milestones feels like a series of wins. That motivational use of chunking connects directly to the open-loop tension of the Zeigarnik Effect, where an unfinished chunk stays active in the mind and pulls the player back.

Healthcare and other high-stakes work

The places where overload is most dangerous are the ones that adopted chunking earliest. Aviation runs on checklists precisely because no pilot can hold a forty-step pre-flight sequence in working memory under pressure; the checklist externalizes the load and chunks it into labeled phases. Surgery followed the same logic when the World Health Organization’s surgical safety checklist cut complications by grouping dozens of verifications into a few timed pauses. Medicine teaches the “rule of threes” for handoffs and the SBAR format (Situation, Background, Assessment, Recommendation) for the same reason: a nurse passing a patient to the next shift cannot rely on memory, so the format pre-chunks what must be communicated. None of this is about limiting information. It is about packaging a large amount of it so a tired human under stress, operating near the four-chunk floor, can still process every piece. When the cost of a dropped chunk is a crashed plane or a missed dose, designers stop treating Miller’s Law as trivia and start treating it as a safety system.

How to Apply Miller’s Law with the Octalysis Framework

Here is the reframe that makes Miller’s Law operational for behavioral design. Miller’s Law is not a Core Drive. It is a capacity law that sits underneath all eight of them. The Octalysis Framework maps the eight Core Drives that decide whether a person wants to act. Miller’s Law decides something prior and just as decisive: whether the information you put in front of that person fits through the working-memory bottleneck well enough for them to act at all. Motivation without legibility creates a spectator. Motivation plus a load the mind can carry creates someone who moves.

Two levers multiply. The Core Drives set how badly someone wants the outcome. Chunking sets whether the path to it is light enough to walk. Get the motivation right but overflow the bottleneck and people abandon a thing they genuinely wanted. Lower the load and the same motivation finally converts. The clearest tell of a designer who understands this: when engagement is high but completion is low, they look at cognitive load before they touch the incentives.

Chunking powers Core Drive 2 (CD2): Development & Accomplishment. Progress is only motivating when it is legible, and legibility is chunking. A skill tree, a level structure, an onboarding checklist: each one takes an overwhelming long-term goal and breaks it into a handful of chunks the player can hold, track, and feel themselves completing. A progress bar is Miller’s Law turned into encouragement, because it chunks an invisible distance into a visible few.

Chunking serves Core Drive 3 (CD3): Empowerment of Creativity & Feedback. Feedback only empowers if the user can hold it in mind while they act on it. Drown them in twelve simultaneous signals and none of it lands. Group the feedback into a few clear chunks and it becomes usable, so the player can actually reason about what to try next instead of freezing.

The dark edge runs through Core Drive 8 (CD8): Loss & Avoidance and Core Drive 6 (CD6): Scarcity & Impatience. Overload is a Black Hat tactic. A cancellation flow that buries the exit among twenty competing options, a pricing page engineered so the plans cannot be held in mind and compared, a terms screen written to exceed anyone’s capacity so they click “accept” out of fatigue: each one weaponizes the bottleneck against the user. The tell is whether the complexity serves the person or defeats them. The ethical fork is simple. White Hat design chunks the user’s complexity so they can act; Black Hat design manufactures complexity so they cannot track what is being done to them. If removing structure would help the user and you added it anyway, that is the line.

Take a concrete case. Imagine a savings app that wants people to set up automatic transfers. The motivation is there, since people genuinely want to save, which is Core Drive 4 (CD4): Ownership & Possession doing its work. But the setup screen asks for nine decisions at once: amount, frequency, source account, destination, start date, round-up rules, goal name, category, and notification settings. Nine independent things blow the four-chunk budget, and the abandonment rate is high despite real desire. Chunk it into three steps, “how much,” “from where,” and “toward what,” carry each answer forward, and default the rest. Nothing about the motivation changed. The load did, and completion climbs. That is Miller’s Law and Octalysis working as one system rather than two.

Practical Steps for Applying Miller’s Law

1. Count the chunks, not the elements. For any screen or step, ask how many independent things the user must hold in mind at once. Visible, glanceable options do not count; only what must be remembered does. Aim for that number to land around four, not seven.
2. Group before you cut. When a set is too large, your first move is to organize it into labeled categories, not to delete items. A menu of thirty grouped into five sections is fine. Amputating it to seven to obey a misread rule destroys value.
3. Carry state forward so nothing has to be held. Every time a flow makes a user remember a code, a choice, or a value from a previous screen, you are spending their scarce capacity. Pass it along automatically. The cheapest chunk is the one the user never has to keep.
4. Format the raw data into chunks for them. Dash the phone numbers, space the card digits, break the long ID into blocks. Do the grouping on the user’s behalf instead of handing them an undivided string.
5. Teach in chunks, then chunk the chunks. Introduce a few units, let them consolidate with practice, then build on them as single pieces. Resist the urge to front-load everything; spacing is what lets a chunk move into long-term memory.
6. Design for distraction, not for the lab. Your user is rarely calm and focused. Assume rehearsal is impossible and plan for the four-chunk floor, so the experience holds up on a phone, mid-task, with a child tugging a sleeve.
7. Let experts chunk their own way. Use progressive disclosure and shortcuts so novices get simplicity and experts get density. One fixed level of complexity cannot serve both, because their chunk sizes are different.

Miller’s Law Was the Beginning, Not the End

The lasting irony of Miller’s Law is that its fame rests on the one part its author cared least about. The number seven made it quotable. Chunking made it true. Seventy years of work on memory, expertise, and learning grew out of the idea Miller buried in the middle of the paper, that the mind beats its own limits by recoding many small things into a few meaningful ones.

For anyone who designs experiences, that is the takeaway worth keeping. You are not designing against a magic number. You are designing for a small budget of chunks, with the power to decide how much each chunk carries. Spend that budget on what matters, group the rest, and carry the state so the user never has to. Do that and complexity stops being the enemy. It becomes something you can package, hand over, and make feel simple, which is the whole job. Miller’s Law is one model among many for that job, and the full set lives in the Behavioral Framework Library.

Frequently Asked Questions About Miller’s Law

What is Miller’s Law in simple terms?

Miller’s Law is the finding that people can hold only a small number of items, famously about seven plus or minus two, in short-term memory at once. The more important half is that memory counts chunks of meaning rather than raw items, so grouping information lets far more of it fit through the same limit.

Is the “magical number seven” actually correct?

Not as a fixed limit. The number slides with the material, landing near seven for digits, six for letters, and five for words, and Nelson Cowan’s 2001 reanalysis put true capacity, with rehearsal blocked, closer to four chunks. Miller himself treated seven as a striking coincidence, not a measured law.

What is a chunk in Miller’s Law?

A chunk is a unit of information the mind treats as a single thing. “FBI” is one chunk to someone who knows the agency but three separate letters to someone who does not. Knowledge is what fuses items into chunks, which is why experts can hold so much more in their field.

Does Miller’s Law mean menus should have seven items?

No. This is the most common misapplication. Miller measured recall, where you hold items in your head with nothing to look at. A menu is recognition, since the options are visible and you scan rather than remember them. Visible lists can be much longer than seven; the real limit on menu length comes from decision time, which is Hick’s Law.

What is the difference between Miller’s seven and Cowan’s four?

Miller’s seven measured how many items people could repeat back when they were free to rehearse and group them, so it captured trained performance. Cowan’s four measured raw capacity with those aids blocked. Both describe the same bottleneck; four is the honest floor to design for, and seven is a best case that rarely survives real conditions.

How can I increase working-memory capacity?

You cannot grow the raw number of slots, but you can dramatically expand what fits by building better chunks. That means deep practice and knowledge in a domain, since expertise turns clusters of details into single recognizable units, the mechanism that let one trained subject recall 79 digits.

Who was George Miller?

George A. Miller (1920 to 2012) was an American cognitive psychologist and a founder of the cognitive-science movement. His 1956 paper, “The Magical Number Seven, Plus or Minus Two,” helped shift psychology away from strict behaviorism toward the study of internal mental processes such as memory and attention.

How do designers use Miller’s Law correctly?

By counting the things a user must hold in mind, not the things on screen, and keeping that number small through grouping, formatting, and carrying state forward. The goal is never to delete options down to seven; it is to organize complexity into a few meaningful chunks so the user’s limited memory is never the failure point.

References

• Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97.
• Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114.
• Ericsson, K. A., Chase, W. G., & Faloon, S. (1980). Acquisition of a memory skill. Science, 208(4448), 1181–1182.
• Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55–81.
• Simon, H. A. (1974). How big is a chunk? Science, 183(4124), 482–488.
• Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The Psychology of Learning and Motivation (Vol. 8, pp. 47–89). Academic Press.
• Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The Psychology of Learning and Motivation (Vol. 2, pp. 89–195). Academic Press.
• Miller, G. A., Galanter, E., & Pribram, K. H. (1960). Plans and the Structure of Behavior. Holt, Rinehart & Winston.
• Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285.
• Baddeley, A. D. (1994). The magical number seven: Still magic after all these years? Psychological Review, 101(2), 353–356.

Related Reading

• Baddeley’s Working Memory Model — the architecture that produces Miller’s bottleneck.
• Cognitive Load Theory — how to manage the limited chunk budget while teaching.
• Hick’s Law — the decision-time law that governs visible menus, often confused with Miller’s.
• Fitts’s Law — the physical-movement cousin in the trio of UX laws.
• The Paradox of Choice — the emotional cost of option overload, downstream of the capacity limit.
• Yu-kai Chou’s books — the full Octalysis approach to motivation and behavioral design.

WOULD YOU LIKE YU-KAI CHOU TO WORK WITH YOUR ORGANIZATION?