Norman Corrected Himself, and Almost Nobody Followed
Gibson coined affordances, Norman borrowed the word in 1988, and in 1999 he corrected himself in print. The correction never took, and flat interfaces are where the confusion shows.
Date
Category
Writer
In December 2013, reporting on a beta of iOS 7.1, Fast Company noted that Apple had added an optional accessibility setting called Button Shapes. Switching it on drew a rounded rectangle behind the text of controls that were, in the shipped design, just words. An operating system had shipped a preference for making buttons look like buttons.
That switch is the clearest artefact I know of a confusion that runs through thirty-five years of interface writing, that Donald Norman created, that Donald Norman publicly corrected in 1999, and that almost nobody in the field followed him on. The correction is short, it is in print, and it is more precise than the thing it corrects.
What Gibson actually wrote
James J. Gibson introduced the noun in chapter 8 of The Ecological Approach to Visual Perception, published in 1979. The chapter is titled The Theory of Affordances and it opens with the definition:
"The affordances of the environment are what it offers the animal, what it provides or furnishes, either for good or ill."
Gibson is explicit that he made the word up. The verb "to afford" is in the dictionary; the noun is his, coined because he needed a term that pointed at the environment and the animal at once. His formulation is that an affordance is "equally a fact of the environment and a fact of behavior", and that it is "both physical and psychical, yet neither". That is not mysticism. It is a deliberate refusal of the subject and object split.
Two further properties matter for anything that follows. First, affordances are relational and must be "measured relative to the animal" rather than in the units of physics: a surface is "stand-on-able" if it is horizontal, flat, extended and rigid relative to the size and weight of the creature standing on it. A table affords support to a cat and not to a horse, and nothing about the table has changed.
Second, and this is the property that gets lost: an affordance "does not change as the need of the observer changes". It does not appear when you notice it and disappear when you do not. Gibson holds that affordances are perceived directly, because ambient light carries information specifying them, and that when the information is misleading, perception is wrong. His example is the visual cliff: the glass affords support, the optical information says otherwise, and the infant will not crawl.
Engelbart’s prototype mouse. A wooden box with one button, and nothing about it suggests what it does. Photograph by Michael Hicks, CC BY 2.0, via Wikimedia Commons.
What Norman did with it in 1988
In The Psychology of Everyday Things, published in 1988, Norman brought the word into design. His definition, on page 9, is not Gibson's:
"the term affordance refers to the perceived and actual properties of the thing, primarily those fundamental properties that determine just how the thing could possibly be used."
The phrase "perceived and actual" is the whole problem in four words. Gibson's affordance is an actual property of an animal-environment relation, full stop, and the perceptual information about it is a separate thing that may be present, absent or wrong. Norman's definition welds the two together, which means a single word now has to do the work of both, and designers reach for it whenever they mean either. The same page gives the examples that made the concept famous and travelled further than the definition: "Plates are for pushing. Knobs are for turning. Slots are for inserting things into. Balls are for throwing or bouncing."
Those sentences are good design writing. They are also where the drift starts, because a flat rectangle on a screen is not a plate and cannot be pushed.
The correction, in his own words
Norman issued the correction twice. The first time was a message to the CHI-WEB mailing list on 28 November 1998, archived at Berkeley, in which he wrote that he had introduced the term to design in The Psychology of Everyday Things and that "I should have used the term 'perceived affordance,' for in design, we care much more about what the user perceives than what is actually true."
The second time was the article proper: "Affordance, Conventions and Design", published in Interactions in May 1999, pages 38 to 43. It opens by handing the word back to its owner. Gibson invented it, Norman writes, "to refer to the actionable properties between the world and an actor (a person or animal). To Gibson, affordances are relationships." Then the admission:
"My fault: I was really talking about perceived affordances, which are not at all the same as real ones."
He goes further than an apology. He commits to a revision: "POET was about 'perceived affordance.' When I get around to revising POET, I will make a global change, replacing all instances of the word 'affordance' with the phrase 'perceived affordance.'"
The article's real contribution is the separation of four things designers habitually merge. Affordances are properties of the world: "Affordances reflect the possible relationships among actors and objects: they are properties of the world." Perceived affordances are what the user believes is possible, and Norman's blunt design rule is that "The designer cares more about what actions the user perceives to be possible than what is true." Constraints come in three kinds, of which physical is the most literal: "For example, it is not possible to move the cursor outside the screen: this is a physical constraint." And conventions are cultural constraints: arbitrary, artificial and learned, slow to be adopted and slow to go away. Of a convention he writes that "The choice of action is arbitrary: there is nothing inherent in the devices or design that requires the system to act in this way."
The sentence that should have ended the argument is his instruction at the close: "Please don't confuse affordance with perceived affordances. Don't confuse affordances with conventions."
The four things, laid side by side
Term | Source | What it actually is | Can a designer add it? |
|---|---|---|---|
Affordance | Gibson 1979, chapter 8 | A relation between an environment and an animal, measured relative to that animal, invariant across the observer's changing needs | Only by changing the physical or computational properties of the thing, for example by making a region genuinely clickable |
Perceived affordance | Norman 1999 | What the user believes can be done, whether or not it can | Yes. This is almost all of what visual interface design does |
Perceptual information about an affordance | Gaver 1991 | The available evidence specifying an affordance, separable from the affordance itself | Yes, and Gaver's point is that it can be present, absent or false independently |
Convention | Norman 1999 | A cultural constraint, arbitrary and learned, slow to arrive and slow to leave | No single designer can add one. You can only adopt or break an existing one |
William Gaver had already done half this work eight years earlier. In "Technology Affordances", presented at CHI'91 in New Orleans and published in the proceedings at pages 79 to 84, he states that "Affordances per se are independent of perception. They exist whether the perceiver cares about them or not, whether they perceived or not", and builds a four-cell scheme from the two variables: perceptible affordance, hidden affordance, false affordance, and correct rejection. A false affordance is the case where the information suggests an action the world does not support. That is a precise name for a decorative element that looks tappable, and it existed in the literature in 1991.
Why the wrong version stuck
I can see four reasons, and none of them is that designers are careless.
The correction was published in the wrong place for the audience
The 1988 book sold for decades and is on reading lists. The correction was a six-page column in a magazine and an email to a mailing list. A book cannot be recalled. By 1999 the incorrect definition had eleven years of circulation and a generation of teaching behind it.
The promised revision changed the word, not the practice
Norman said he would replace every instance of "affordance" with "perceived affordance" in a revision of the book. A global find and replace does not undo a concept that has already been absorbed. Readers who learned "affordance" in 1988 did not re-learn it because a later edition printed two words instead of one.
The wrong version is more useful in a meeting
"This button affords clicking" is a sentence that wins arguments. "This button carries perceptual information suggesting a click target, and the click target is a real affordance created by the event handler" does not. The imprecise version compresses a true claim and a false one into a word that sounds technical.
Norman himself gave up on it
In 2008, in Interactions volume 15 issue 6, he published "Signifiers, not affordances". He defines a signifier as "some sort of indicator, some signal in the physical or social world that can be interpreted meaningfully", notes that since 1988 the term affordance "has been widely used and misused. The result has been confusions and a goldmine for academic scholars", and concludes with an instruction: "Designers of the world: Forget affordances provide signifiers." That is a third vocabulary in twenty years, from the same author. A field being asked to correct itself twice will usually correct itself zero times.
The flat button, concretely
Here is what the confusion costs, measured. In September 2017 Kate Moran published an eyetracking study for Nielsen Norman Group, Flat UI Elements Attract Less Attention and Cause Uncertainty. Seventy-one general web users each saw one of two versions of nine websites and were given a task on each, for example being asked to reserve a hotel room and to say when they had found where they would click. The versions differed in the strength of their clickability signifiers, not in their information.
On average, participants spent 22 per cent more time on the pages with weak signifiers and made 25 per cent more fixations on them. Six of the nine page pairs produced different gaze patterns. On one page, 9 of the 24 participants who saw the weak-signifier version stopped reading at a subheading and never looked at the text link below it at all.
Moran's earlier piece from November 2015, "Long-Term Exposure to Flat Design", describes the coping behaviour: users "hover the mouse over elements hoping for dynamic clickability signifiers, or click experimentally". One participant put it as "I just start clicking and praying that it works."
Now apply Norman's 1999 vocabulary to that finding, because it dissolves the whole debate. A flat text-only button has the same real affordance as a raised one: the click handler exists, the region is live, the world supports the action. What differs is the perceived affordance, which is near zero, and the convention, which says that underlined blue text and rounded filled rectangles are actionable. When a team says flat design "removes affordances", it is wrong twice over: it removed no affordance at all, and it removed perceptual information and leaned harder on convention instead. That is a defensible trade in a product people use daily and a poor one on a page a stranger sees once.
Norman anticipated the exact error in 1999: "Far too often I hear graphical designers claim that they have added an affordance to the screen design when they have done nothing of the sort." He also gives the case where the claim would be legitimate, which designers never quote: restricting where characters can be typed in a form, so that some regions genuinely accept input and others genuinely do not. "This would be a real use of affordances."
What I had wrong, and what is still open
I taught the 1988 definition for years, in the form "an affordance is what an object suggests you can do with it". That sentence is Gibson inverted. Gibson's affordance does not suggest anything; it exists whether or not it is noticed, and suggestion is a property of the information, not of the relation. The word I wanted for suggestion was Gaver's "perceptible affordance", available since 1991, or Norman's "signifier", available since 2008. I had both and used neither.
What I cannot establish: whether Norman ever did make the global replacement he promised. The revised edition of the book appeared in 2013 under a different title and does introduce signifiers, but I have not been able to check its text against his 1999 commitment, and I am not going to claim he kept or broke a promise on the basis of secondary accounts. If he did keep it, the revision replaced a word he had by then already abandoned, which would be its own small comedy.
Sources
James J. Gibson, "The Theory of Affordances", chapter 8 of The Ecological Approach to Visual Perception, Houghton Mifflin, Boston, 1979. cs.brown.edu/courses/cs137/2017/readings/Gibson-AFF.pdf
Donald A. Norman, The Psychology of Everyday Things, Basic Books, New York, 1988, page 9, for the definition of affordance and the "plates are for pushing" examples, as quoted in the Interaction Design Foundation's glossary chapter on affordances. ixdf.org/literature/book/the-glossary-of-human-computer-interaction/affordances
Donald A. Norman, "Affordance, Conventions and Design", Interactions, May 1999, pages 38 to 43, ACM. Full text as circulated. cseweb.ucsd.edu/~goguen/courses/271sp03/jnd.html
Donald A. Norman, "Affordance, Conventions and Design (Part 2)", jnd.org, giving the original publication details as Interactions, May 1999, pages 38 to 43, and the commitment to replace every instance of "affordance" in a revision. jnd.org/affordance-conventions-and-design-part-2/
Donald A. Norman, message to the CHI-WEB list, 28 November 1998, the earlier version of the same correction. courses.ischool.berkeley.edu/i213/s99/Readings/norman-on-affordances.txt
Donald A. Norman, "Signifiers, not affordances", Interactions, volume 15, issue 6, 17 November 2008. jnd.org/signifiers-not-affordances/
William W. Gaver, "Technology Affordances", Proceedings of CHI'91, New Orleans, 28 April to 2 May 1991, ACM, New York, pages 79 to 84. lri.fr/~mbl/Stanford/CS477/papers/Gaver-CHI1991.pdf
Kate Moran, "Flat UI Elements Attract Less Attention and Cause Uncertainty", Nielsen Norman Group, 3 September 2017. Eyetracking study, 71 participants, nine paired pages. nngroup.com/articles/flat-ui-less-attention-cause-uncertainty/
Kate Moran, "Long-Term Exposure to Flat Design: How the Trend Slowly Decreases User Efficiency", Nielsen Norman Group, 8 November 2015. nngroup.com/articles/flat-design-long-exposure/
Chris Dannen, "Apple Takes A Step Backwards From Flat Design", Fast Company, 16 December 2013, on the Button Shapes setting in iOS 7.1. fastcompany.com/3023552/apple-takes-a-step-backwards-from-flat-design
Related reading
Latest Articles.
© Moc Mallen Studio
Thoughts, ideas, and perspectives on design, simplicity, and creative process.
