You meet these on the way up to everything else. Each carries the same four fields —
the claim, its origin, its evidence, and its limits and misuse — because a principle quoted
without its limits is how a law becomes a superstition.
House rule: every principle ships with its misuse.
No exceptions, and especially not Miller.
6 documented — A named, findable source says this.4 traditional — Long-standing practice, with no single citable origin.1 folk — Widely repeated, and the evidence does not support it as it is used.
Confidence
STR·01
Fitts's Law
documented
The time to acquire a target is a function of its distance and its size.
T = a + b·log₂(2D/W)
OriginPaul Fitts, 1954 — a motor-control study of pointing, decades before pointers; imported into HCI in the 1970s–80s (Card, English & Burr).
EvidenceOne of the most replicated results in human–machine studies; the reason screen edges and corners are "infinitely deep" targets, why the macOS menu bar lives at the edge, and why 44pt/48dp touch minimums exist.
Limits & misuseIt models pointing, not deciding, reading, or gesturing; it says nothing about which target is correct. "Make everything bigger" is not Fitts — it's a misreading with a real cost, since every enlarged target pushes another one farther away.
Why documented: Fitts 1954 is a published motor-control study, replicated for seventy years.
Choice time grows with the logarithm of the number of equally likely options.
T = b·log₂(n+1)
OriginW. E. Hick (1952) and Ray Hyman (1953) — reaction-time experiments on stimulus–response choice.
EvidenceHolds cleanly for simple, practiced, equally weighted choices — which is why command menus and remote controls benefit from pruning.
Limits & misuseThe most over-quoted law in design. It does not apply to reading a menu of unfamiliar options (that's visual search, roughly linear), and it never said "fewer options are always better" — a well-grouped fifty beats a random eight. Citing Hick to delete features is rhetoric wearing a formula.
Why documented: Hick 1952 and Hyman 1953 are published reaction-time experiments.
Immediate memory holds about seven chunks — said Miller, about a specific laboratory task, once.
OriginGeorge Miller, "The Magical Number Seven, Plus or Minus Two" (1956) — a review of short-term memory span; the paper's real gift is chunking.
EvidenceChunking is robust and genuinely useful: phone-number grouping, IBAN spacing, code formatting. Later work (Cowan, 2001) puts the practical span nearer four.
Limits & misuseThe card exists for this field. "Menus must have seven items" appears nowhere in Miller — a navigation menu is on screen, and recognition is not recall. Miller himself spent years bemused that a paper about memory span became a law about menu length. The Institute keeps the principle because its misuse is the best available lesson in how citations decay into folklore.
Why folk: Miller published a real result about a specific task, then protested the generalisation himself. "Seven items in a menu" is not what the paper says.
People spend most of their time on other sites: they arrive carrying expectations your interface didn't set.
OriginJakob Nielsen (2000) — an argument about convention, not cognition.
EvidenceThe empirical spine of design systems and platform conventions; the reason the cart lives top-right and search looks like search.
Limits & misuseRead as "never innovate," it would have banned pull-to-refresh and pinch-zoom. The honest reading: convention is a budget — spend your novelty where it pays, and pay the toll of teaching whenever you do.
Why traditional: Nielsen’s formulation is an observation about expectation, not a measured law.
The interface must keep the human's model and the machine's state in agreement, continuously and legibly.
OriginNielsen's first heuristic (1994); Norman's gulf of evaluation (1986/1988) is the deeper statement.
EvidenceIts violations fill the basement: every mode error is this principle, unpaid.
Limits & misuseVisibility is not verbosity — a system narrating everything buries the one state change that mattered. The skill is salience proportional to consequence.
Why documented: One of Nielsen’s ten heuristics, published and stable since 1994.
0.1s feels instantaneous; 1s keeps the flow of thought; 10s is the limit of attention — after that, people leave.
OriginResponse-time limits synthesized by Nielsen (1993) from earlier work (Miller 1968; Card et al. 1991).
EvidenceThe quiet constitution behind spinners, skeletons, optimistic UI, and progress bars — each is a different answer to which threshold you're about to miss.
Limits & misuseThe numbers are bands, not physics; context moves them (a payment may be granted patience a search never gets). And faking progress to soothe the 10s band is a loan from the basement.
Why documented: The 0.1 / 1 / 10 second thresholds are Card, Robertson & Mackinlay and Nielsen.
The perceivable properties of a thing should announce what can be done with it.
OriginJ. J. Gibson (1979) coined it for ecological psychology — what an environment offers an animal; Norman (1988) imported it, then spent years clarifying that design means perceived affordances and later offered "signifiers" for the visible layer. The drift is part of the entry.
EvidenceThe Norman door downstairs; flat design's discoverability crisis and the partial return of borders and shadows measure the cost of hiding the announcement.
Limits & misuse"It affords clicking" applied to anything clickable makes the word mean nothing. Affordance is a relationship, not a coat of paint — and citing Gibson while meaning Norman is this stairwell's favorite pedantry, indulged exactly once, here.
Why documented: Gibson 1979 coined it, Norman 1988 imported it, Norman 1999 corrected the drift. All three are citable, which is why the drift is documentable.
The strongest instruction is the impossible action: design so the wrong thing cannot be done.
OriginNorman (1988) on constraints; the manufacturing lineage is poka-yoke (Shingo).
EvidenceThe ATM that returns your card before the cash (Floor 5's classic — the machine holds the money hostage against your forgetfulness); SIM-card notches; plugs that only fit one way.
Limits & misuseConstraints protect; they also imprison. A forcing function on a rare-but-legitimate action becomes the workaround culture that defeats it — the taped-over sensor is a constraint's obituary. The Hawaii alert is what its absence costs.
Why traditional: A design practice with a long lineage rather than one founding paper.
Show people their options; don't make them remember the incantation.
OriginNielsen heuristic #6, standing on well-established memory research: recognition is reliably easier than recall.
EvidenceThe GUI's whole victory over the command line, replayed today in command palettes — which win precisely by turning recall (remember the command) into recognition (type two letters, see the options).
Limits & misuseFor experts, recall is faster — keyboard shortcuts exist because recognition has a reading tax. The principle sets the default for the newcomer, not the ceiling for the practiced; interfaces that never let you graduate to recall are politely condescending forever.
Why documented: Nielsen heuristic, and the recognition-over-recall memory result predates it.
Every task has an irreducible complexity; the only question is who absorbs it — the system or the person.
OriginLarry Tesler, at Xerox PARC and after — the conservation of complexity.
EvidenceEvery "simple" interface conceals machinery that ate the complexity on the user's behalf: one-tap payment stands on a mountain of absorbed difficulty. The converse is the form that makes you do the parsing the system declined to.
Limits & misuseNot all complexity is conserved — some is invented, and should be deleted, not moved. And absorbing complexity has a price the principle doesn't mention: the system that hides everything also hides what went wrong (see: every inscrutable error the basement collects).
Why traditional: Tesler’s law of conservation of complexity is an industry formulation, widely attributed and not published as a study.
A reader can only look at one thing first. Design decides which, or the accident does.
OriginNo single source, which is why this entry is marked traditional. The perceptual half is citable — Treisman and Gelade (1980) showed that certain attributes are processed preattentively, found in a glance and without search, which is why size and colour separate an element from its neighbours before anyone has read a word. The design half is older and unattributed: the typographic tradition of the display line, the manuscript initial, and the newspaper front page, which solved this problem for centuries before it had a name.
EvidenceThe sixty-seven published grids are an argument about this: a grid exists to make the ordering repeatable rather than renegotiated per page. Downstairs, the Hawaii alert is what happens when the most consequential control on a screen is given no more weight than its neighbours.
Limits & misuse"Make it stand out" applied to six things at once produces a screen with no hierarchy at all, only noise — the second focal point costs more than the first one gained. The devices are also not interchangeable: colour alone fails for a reader who cannot distinguish it, which is why 1.4.1 exists, and scale alone fails once everything has already been enlarged. Hierarchy is a ranking, and a ranking with no bottom is not one.
Why traditional: Treisman and Gelade (1980) is citable for the mechanism, and is cited here. "Visual hierarchy" as a design rule is a practice tradition with no origin document — it is taught everywhere and attributed nowhere, and the honest label for that is traditional rather than documented.