A Story of Affordance

“Affordance” is one of the terms designers use most often. Yet Donald Norman, who first introduced it to design, has spent more than twenty years saying it is being misused. I looked into its precise meaning and traced the commotion surrounding it.

A history of the commotion

The 1920s: Gestalt psychology—demand character and invitation character

In The principles of gestalt psychology, Kurt Koffka says that we perceive the meaning or value of objects immediately, as we perceive color. Objects, he writes, have a character that demands action.

Each thing tells us what it is…. A fruit says “Eat me”; water, “Drink me”; thunder, “Fear me”; and a woman, “Love me.” … They tell us what we should do with them.

(The example of perceiving “Love me” upon seeing a woman seems inappropriate to a twenty-first-century reader.)

Another Gestalt psychologist, Kurt Lewin, coined Aufforderungscharakter, translated into English as “invitation character.”

1979: James Gibson’s The Ecological Approach to Visual Perception—possibilities for action

Perceptual psychologist James J. Gibson proposed a new concept derived from the Gestalt psychologists’ ideas, coining the word affordance to express it.

He describes the key difference between their concept and his as follows:

Unlike the Gestalt concept, an affordance does not change when the observer’s needs change. The observer may fail to perceive it, or may perceive it yet not act on it, depending on present needs. But the affordance itself is invariant and always there. —J. J. Gibson, The ecological approach to visual perception

Although first used in 1966, the word became widely known through his 1979 book The ecological approach to visual perception. Gibson introduces it this way:

The verb “afford” is in the dictionary, but the noun “affordance” is not. I made it up…. If a terrestrial surface is nearly horizontal rather than sloping, flat rather than convex or concave, sufficiently extended relative to an animal’s size, and rigid relative to its weight, that surface affords support…. It is stand-on-able: the animal can stand on four feet or two. It is therefore walk-on-able and run-over-able. Unlike water or a swamp, it is not sink-into-able…. The four properties—horizontal, flat, extended, rigid—are physical properties of the surface. But whether it affords support for a particular species depends on the animal…. An affordance therefore cannot be measured in physical terms alone.

Gibson’s concept can be summarized as follows.

Affordances are not inherent properties of the environment; they exist in the relationship between an agent—an animal or person—and the environment. “Sit-on-ability,” for example, is not an essential property of a chair. It exists only when the chair’s shape and size suitably match the agent’s body and skeletal structure. An affordance between an adult and a chair may not exist between a child and the same chair. A wall affords climbing to an ant, but not to a human.
Affordances exist whether agents perceive them or not, and whether they want them or not. Even if an agent does not realize they can sit on a chair, the affordance remains. It also remains when the agent does not want to sit.

1988: Donald Norman’s The Design of Everyday Things—perceived and actual properties

About ten years later, cognitive psychologist and designer Donald Norman introduced affordances to designers through The design of everyday things.

Norman, born in 1935, encountered the concept through Gibson, born in 1904, and sought to apply his understanding of it to design. But the meaning he initially presented differed considerably from Gibson’s.

Norman introduces affordances as follows:

There is already a psychology of materials and objects: the study of affordances. The term refers to the perceived and actual properties of a thing, especially the fundamental properties that determine how it could possibly be used.

Despite explicitly referring to an established field, Norman’s account differed from the concept used in that field—the Gibsonian school—and caused confusion for decades.

Gibson emphasizes that an affordance is not a physical property, a central point in his school of thought. Norman, however, calls affordances “actual properties of a thing,” which can make them sound inherent in objects. For Gibson, size, rigidity, and flatness are physical properties, but whether they allow an agent to stand on an object—the affordance of standing—is a property of the relationship between object and agent.
Gibson’s affordances exist independently of whether agents perceive them. By calling them “perceived and actual properties,” Norman mixed up the existence of an affordance with its perception.

1998: Donald Norman’s The Invisible Computer—distinguishing real and perceived affordances

Another ten years later, Norman published The invisible computer. In Korean it was retitled Human-Centered Design, making it sound like a different book. (Korean publishing has a long tradition of taking excessive liberties with original titles.)

Here Norman acknowledges his mistake and reintroduces the concept.

An affordance is not a property but a relationship between an object and an organism acting on it…. Psychologist James J. Gibson coined the term while studying human visual perception. In The design of everyday things, I appropriated and extended it to design. In designing objects, real affordances matter less than perceived ones. It is perceived affordances that tell users what actions they can perform and how…. Distinguishing real from perceived affordances is crucial. Design concerns both, but perceived affordances determine usability. I did not make this distinction clear enough in The Design of Everyday Things. The term has since been widely misused, and I have spent much time trying to correct this. Designers say, “I added an affordance by putting a shadow around the icon.” However well-intentioned, such misuse makes me shudder. I imagine J. J. Gibson rising from his grave, glaring at me again, theatrically removing his hearing aid, and lying back down in disgust.

According to Norman, when their arguments grew heated, Gibson would say his piece and conspicuously switch off his hearing aid. Gibson died in 1979, so unfortunately Norman could not have discussed the 1988 book with him directly. Perhaps he was recalling their earlier heated debates.

1999: Donald Norman’s “Affordance, Conventions, and Design”—affordances in digital interfaces are meaningless

Distinguishing affordances from perceived affordances seems relatively easy with physical products. Digital interfaces are the problem. In 1999, Norman reportedly watched CHI-Web participants discussing things like “putting affordances on buttons” until he finally lost control, shouted “No!” and brought the discussion to a halt. This led to Affordances, conventions, and design, published in the May–June 1999 issue of Interactions.

What I introduced in The Design of Everyday Things was “perceived affordance.” In a revised edition, I intend to find every occurrence of “affordance” and replace it with “perceived affordance.” … Please do not confuse affordances with perceived affordances.

But Norman’s distinction between real and perceived affordances in computer interfaces introduced further confusion.

In screen-based graphical interfaces, the designer can control only perceived affordances. Computer systems already have inherent physical affordances…. You can point at, touch, click, or look at every pixel. Most of these affordances have little significance for application design…. Similarly, saying a visual element “affords clicking” is wrong. You can click it, of course, but you can click anywhere on the screen. It may tell users where to click, but that is convention or feedback, not an affordance…. I hear “I added an affordance to the screen” far too often. What was added was a visual hint about a possible action, not an affordance. It is neither a real nor a perceived affordance.

For Norman, real affordances include:

A monitor is a flat, rigid surface near your face, so you can point to a spot on it.
You can press the mouse button even when the computer is off.

But he says we should not describe an onscreen button as affording clicking. Two different claims are mixed together here.

A visual hint showing that an onscreen button can be clicked should not be called an affordance.
A virtual action involving an onscreen button, rather than a physical action, should not be called an affordance.

The first claim is sound. The second deserves more thought.

What would Gibson have thought? Unfortunately, he died in 1979, just before computers became widespread and computer displays became common in perceptual psychology research. He said little about computer interfaces. But he wrote extensively about pictures, photographs, and movies—artificial displays on two-dimensional surfaces rather than actual spaces. Those views let us infer something about how he might have assessed Norman’s distinction.

On photographs, Gibson writes:

A photograph is a surface in its own right while also displaying information about something else…. We can distinguish the picture’s surface from the surfaces depicted within it…. Ask, “How far away does the tree in the picture look?” and someone might answer, “About thirty paces.” Ask, “How far away does the picture look?” and they might answer, “Oh, about four paces.” … A duality of information produces a duality of experience. —The ecological approach to visual perception

He calls a tree depicted in a photograph a virtual object, and the perspective discovered in the Renaissance artificial perspective. Virtual objects in photographs can sometimes fool observers into perceiving affordances, but not if observers can freely turn their heads and walk around the picture. This is why trick art requires viewing from a precisely specified position. Gibson argues that outside laboratories, which often fix the observer’s position and head, people and animals are not fooled by virtual objects or artificial perspective. (See Gibsonian categorization of vision.) They can move their bodies and heads, look around, or try touching the object, discover that it is not real—that the affordance suggested by the picture does not actually exist—and thus cannot perform an “action” on the virtual object.

What happens if we extend this discussion to computer displays?

A display itself is a rigid plane, but it also presents information about something else. It contains virtual buttons. From Gibson’s perspective, the crucial difference from a photograph is that these virtual buttons exhibit virtual movement in response to the agent’s actions. Whatever its appearance, an onscreen button affords being pressed to operate it. Visual features suggesting pressability—shading consistent with light from above, a strong border separating it from the background panel, and so on—reveal that affordance. Through these features, the agent perceives it. Pressing the button changes its appearance to suggest being depressed and triggers an operation, letting the agent know that the affordance was perceived correctly. In the virtual environment, the button not only looks pressable; it really can be pressed. In one sense the screen can be pressed—the virtual button—and in another it cannot—the flat, rigid physical panel. To borrow Gibson’s phrase, there is a duality of experience.

I think this interpretation fits Gibson’s perspective better. He argues that visual perception exists to help organisms discover environmental affordances and act appropriately. What matters is whether the perception–action loop is connected or broken. Pictures and movies break that loop; a GUI connects it to a greater degree. VR interfaces may make still better use of ecological perception, including ambulatory vision.

To summarize, Norman’s claim that virtual actions on onscreen buttons cannot count as affordances does not fit Gibson’s concept.

2008: Donald Norman—signifiers, not affordances

About ten more years pass. In 2008, Norman publishes two articles on affordances:

Affordances and design
Signifiers, not affordances

The first is almost a republication of his 1999 CHI-Web mailing-list post, so its timing and content differ little from the discussion above.

Let us therefore focus on the second.

Time to reflect. As time passes and technology changes, computing expands from the individual to groups, society, and culture. Communication skills become as important as computing skills…. A central design principle was perceived affordance: it tells us what to do in a situation we have never encountered. This works for unfamiliar objects. But unfamiliar situations? People, social groups, cultures? The answer is both the same and different. Perceived affordances, constraints, and conceptual models still matter, but something else is needed. We should consider trails and behaviors. We often discover what to do through other people’s actions, their absence, or the traces they leave…. A signifier is a kind of mark: a signal about the physical or social world that can be meaningfully interpreted. Sometimes an unintended by-product rather than a deliberate design, it nevertheless conveys important information…. Let us use “signifier” instead of “affordance.” … Gibson’s affordances need not be perceived or known; they simply exist. What I described when introducing the term to design in 1988 was perceived affordance. Since then, the term has been both widely used and widely misused…. The perceived part of an affordance is a signifier…. Designers of the world: forget affordances and use signifiers.

Despite repeated efforts, the confusion keeps growing, and people continue using “affordance” to mean all sorts of things. Norman keeps trying. This time he proposes a new term, signifier, to replace it. The article makes two main proposals:

First, use the frequently misused “affordance” in its original sense, and use “signifier” for what people have been calling affordances.
Second, expand the focus beyond objects to principles that apply to groups, society, and culture.

How widely has the first proposal been accepted? As I write in 2021, thirteen years have passed, but unfortunately almost everyone around me still uses “affordance” in their own varied senses. “I added an affordance to the button” seems the most common misuse. I have yet to hear “signifier” used in Norman’s proposed sense in an actual conversation.

What about the second proposal? I think it is meaningful and connects well with Gibson’s research. Gibson not only includes other animals and people in his concepts of environment and affordance; he identifies them as its most important elements. Describing affordances, he writes:

The richest and most elaborate affordances are provided by other animals and people…. They differ from ordinary objects, and infants quickly learn to distinguish plants and other objects from animals and humans…. Behavior affords behavior. Psychology and the social sciences elaborate this basic fact. Sexual, parental, fighting, cooperative, economic, and political behavior all concern what one person affords another, or how those affordances are misperceived. —The ecological approach to visual perception

2013: Donald Norman—clarifying affordances, perceived affordances, and signifiers

2013 saw the twenty-fifth-anniversary revised and expanded edition of The design of everyday things. Its organization, content, and examples changed so much that it is almost a new book. I recommend rereading it even if you have read an earlier edition.

Here he distinguishes the concepts as follows.

Affordances are possibilities for interaction between people and the environment. Some are perceptible; others are not.
Perceived affordances often function as signifiers, but can be ambiguous because an affordance can sometimes be misperceived.
Signifiers communicate something, particularly what actions are possible and how to perform them. A signifier must be perceivable; otherwise it is not working.

So eager is he for designers to say “signifier” instead of “affordance” that he includes a fictional conversation.

Designer: People are not using our application correctly.
Mentor: Could you explain?
Designer: We show a list of recommended restaurants. Swipe left or right to see other recommendations. Swipe up on a restaurant to see its menu, or down to see whether friends are already there. People find the restaurant lists, but seem unable to use the menus or friend-finding features.
Mentor: Why do you think that is?
Designer: I am not sure. Should we add affordances? Perhaps arrows and labels?
Mentor: Good idea. But why call them affordances? Aren’t the affordances already there?
Designer: Oh, yes. But they are not visible.
Mentor: Exactly. You need signals about the affordances.
Designer: Right. But isn’t that what I just said?
Mentor: No. You called arrows and labels affordances, though they are only signals and add no action possibilities. It would be better to call them signifiers.
Designer: I see. Then why do designers worry about affordances at all? We should focus more on signifiers.
Mentor: Wisely said. Communication is the key to good design, and signifiers are the key to communication.
Designer: Ah, I see my confusion now. Signifiers are what send signals. Everything is clear.
Mentor: That is how deep insights work. Once understood, they become obvious.

2015: William M. Mace—a Gibsonian assessment of Norman’s contribution

As noted, Gibson died in 1979, but his book seems to have become a classic. In 2015, it was republished in a “Classic Edition” collection alongside other distinguished books. Professor William M. Mace, who wrote the introduction, assesses Norman as follows:

After Gibson’s book appeared in 1979, the concept of affordance, to borrow internet terminology, went viral. This was largely due to Donald Norman’s influence in interaction design. Norman’s account somewhat skewed the meaning, but he and his colleagues quickly corrected it, acknowledged Gibson’s contribution, and clarified the various uses derived from the original meaning. —William M. Mace, The ecological approach to visual perception, Classic Edition

It is politely put, but I get the impression there is more he might have said. Deeper philosophical differences underlie the long commotion over affordances.

Gibson’s ecological approach and Norman’s classical cognitivism

Why did Gibson switch off his hearing aid in the middle of debates with Norman? According to Norman, merely mentioning “internal representation” or “information processing” made Gibson grimace.

Gibson criticized classical cognitivism. The concept of affordance itself arose in that critique. Norman, however, is a classical cognitivist. A concept proposed to criticize classical cognitivism became famous through a classical cognitivist’s interpretation and spread worldwide in distorted form. That must have felt strange to Gibsonians.

I would like to discuss the difference between these perspectives another time, using more practical examples.

Closing thoughts

The long commotion began with Norman’s initial misrepresentation of Gibson’s affordances. Norman worked hard to set things right, asking people to use “affordance” in Gibson’s original sense and introducing “signifier” instead. But explaining that original sense created further confusion. Perhaps this is partly inevitable: affordance is a concept from the ecological approach, which Norman does not embrace.

In any case, Norman wanted to restore affordances to Gibson’s meaning and have designers focus more on signifiers. Following that wish, here is a short account.

What is an affordance?

A possibility of action between an agent and its environment. If translating the term into Korean, “action possibility” is more apt than “action inducement.” Agents include animals and people. The environment includes other agents, who are the most important objects of perception. This connects with Norman’s later concept of social signifiers.
It is not inherent in the environment but exists in the relationship between agent and environment. A chair’s “sit-on-ability” is not an essential property of the chair; it exists only when its shape and size fit the agent’s body and skeleton. An affordance available to an adult may not be available to a child. A wall affords climbing to an ant, but not a human.
It exists whether the agent perceives or wants it or not. Even if the agent fails to realize that they can sit on a chair, the affordance exists, and it remains when they do not want to sit.
“Adding an affordance” would properly mean introducing a previously unavailable possibility for action. But usually “adding a feature” already says that, so there seems little reason to say “adding an affordance” except in unusual cases.

What is a signifier?

It communicates something, particularly which actions are possible and how to perform them. It must be perceivable; otherwise it is not functioning.
It can be deliberately designed or arise accidentally. A visible trail naturally formed by many people walking in the mountains is also a signifier.
When an affordance users need to notice is not sufficiently apparent, design and add a good signifier. We may have been saying “adding an affordance,” but what we were adding was a signifier.

What should you do when someone says “I added an affordance to the button”?

If you know them well, gently suggest “signifier.”
If you do not, correcting them may offend unnecessarily, so mentally substitute “signifier” yourself. Try to speak as clearly as possible while doing your best to understand others even when their language is looser.