by Don Norman
•Perceived affordances often act as signifiers, but they can be ambiguous.
•Signifiers signal things, in particular what actions are possible and how they should be done. Signifiers must be perceivable, else they fail to function.
In design, signifiers are more important than affordances, for they communicate how to use the design. A signifier can be words, a graphical illustration, or just a device whose perceived affordances are unambiguous. Creative designers incorporate the signifying part of the design into a cohesive experience. For the most part, designers can focus upon signifiers.
Because affordances and signifiers are fundamentally important principles of good design, they show up frequently in the pages of this book. Whenever you see hand-lettered signs pasted on doors, switches, or products, trying to explain how to work them, what to do and what not to do, you are also looking at poor design.
AFFORDANCES AND SIGNIFIERS: A CONVERSATION
A designer approaches his mentor. He is working on a system that recommends restaurants to people, based upon their preferences and those of their friends. But in his tests, he discovered that people never used all of the features. “Why not?” he asks his mentor.
(With apologies to Socrates.)
DESIGNER
MENTOR
I’m frustrated; people aren’t using our application properly.
Can you tell me about it?
The screen shows the restaurant that we recommend. It matches their preferences, and their friends like it as well. If they want to see other recommendations, all they have to do is swipe left or right. To learn more about a place, just swipe up for a menu or down to see if any friends are there now. People seem to find the other recommendations, but not the menus or their friends? I don’t understand.
Why do you think this might be?
I don’t know. Should I add some affordances? Suppose I put an arrow on each edge and add a label saying what they do.
That is very nice. But why do you call these affordances? They could already do the actions. Weren’t the affordances already there?
Yes, you have a point. But the affordances weren’t visible. I made them visible.
Very true. You added a signal of what to do.
Yes, isn’t that what I said?
Not quite—you called them affordances even though they afford nothing new: they signify what to do and where to do it. So call them by their right name: “signifiers.”
Oh, I see. But then why do designers care about affordances? Perhaps we should focus our attention on signifiers.
You speak wisely. Communication is a key to good design. And a key to communication is the signifier.
Oh. Now I understand my confusion. Yes, a signifier is what signifies. It is a sign. Now it seems perfectly obvious.
Profound ideas are always obvious once they are understood.
MAPPING
Mapping is a technical term, borrowed from mathematics, meaning the relationship between the elements of two sets of things. Suppose there are many lights in the ceiling of a classroom or auditorium and a row of light switches on the wall at the front of the room. The mapping of switches to lights specifies which switch controls which light.
FIGURE 1.6.Signifiers on a Touch Screen. The arrows and icons are signifiers: they provide signals about the permissible operations for this restaurant guide. Swiping left or right brings up new restaurant recommendations. Swiping up reveals the menu for the restaurant being displayed; swiping down, friends who recommend the restaurant.
Mapping is an important concept in the design and layout of controls and displays. When the mapping uses spatial correspondence between the layout of the controls and the devices being controlled, it is easy to determine how to use them. In steering a car, we rotate the steering wheel clockwise to cause the car to turn right: the top of the wheel moves in the same direction as the car. Note that other choices could have been made. In early cars, steering was controlled by a variety of devices, including tillers, handlebars, and reins. Today, some vehicles use joysticks, much as in a computer game. In cars that used tillers, steering was done much as one steers a boat: move the tiller to the left to turn to the right. Tractors, construction equipment such as bulldozers and cranes, and military tanks that have tracks instead of wheels use separate controls for the speed and direction of each track: to turn right, the left track is increased in speed, while the right track is slowed or even reversed. This is also how a wheelchair is steered.
All of these mappings for the control of vehicles work because each has a compelling conceptual model of how the operation of the control affects the vehicle. Thus, if we speed up the left wheel of a wheelchair while stopping the right wheel, it is easy to imagine the chair’s pivoting on the right wheel, circling to the right. In a small boat, we can understand the tiller by realizing that pushing the tiller to the left causes the ship’s rudder to move to the right and the resulting force of the water on the rudder slows down the right side of the boat, so that the boat rotates to the right. It doesn’t matter whether these conceptual models are accurate: what matters is that they provide a clear way of remembering and understanding the mappings. The relationship between a control and its results is easiest to learn wherever there is an understandable mapping between the controls, the actions, and the intended result.
Natural mapping, by which I mean taking advantage of spatial analogies, leads to immediate understanding. For example, to move an object up, move the control up. To make it easy to determine which control works which light in a large room or auditorium, arrange the controls in the same pattern as the lights. Some natural mappings are cultural or biological, as in the universal standard that moving the hand up signifies more, moving it down signifies less, which is why it is appropriate to use vertical position to represent intensity or amount. Other natural mappings follow from the principles of perception and allow for the natural grouping or patterning of controls and feedback. Groupings and proximity are important principles from Gestalt psychology that can be used to map controls to function: related controls should be grouped together. Controls should be close to the item being controlled.
FIGURE 1.7.Good Mapping: Automobile Seat Adjustment Control. This is an excellent example of natural mapping. The control is in the shape of the seat itself: the mapping is straightforward. To move the front edge of the seat higher, lift up on the front part of the button. To make the seat back recline, move the button back. The same principle could be applied to much more common objects. This particular control is from Mercedes-Benz, but this form of mapping is now used by many automobile companies. (Photograph by the author.)
Note that there are many mappings that feel “natural” but in fact are specific to a particular culture: what is natural for one culture is not necessarily natural for another. In Chapter 3, I discuss how different cultures view time, which has important implications for some kinds of mappings.
A device is easy to use when the set of possible actions is visible, when the controls and displays exploit natural mappings. The principles are simple but rarely incorporated into design. Good design takes care, planning, thought, and an understanding of how people behave.
FEEDBACK
Ever watch people at an elevator repeatedly push the Up button, or repeatedly push the pedestrian button at a street crossing? Ever drive to a traffic intersection and wait an inordinate amount of time for the signals to change, wondering all the time whether the detection circuits noticed your vehicle (a common problem with bicycles)? What is missing in all these cases is feedback: some way of letting you know that the system is working on your request.
Feedback—communicating the results of an action—is a well-known concept from the science of control and information theory. Imagine trying to hit a target with a ball when you cannot see the target. Even as simple a task as picking up a glass with the hand requires feedback to aim the hand properly, to grasp the glass, and to lift it. A misplaced hand will spill the conte
nts, too hard a grip will break the glass, and too weak a grip will allow it to fall. The human nervous system is equipped with numerous feedback mechanisms, including visual, auditory, and touch sensors, as well as vestibular and proprioceptive systems that monitor body position and muscle and limb movements. Given the importance of feedback, it is amazing how many products ignore it.
Feedback must be immediate: even a delay of a tenth of a second can be disconcerting. If the delay is too long, people often give up, going off to do other activities. This is annoying to the people, but it can also be wasteful of resources when the system spends considerable time and effort to satisfy the request, only to find that the intended recipient is no longer there. Feedback must also be informative. Many companies try to save money by using inexpensive lights or sound generators for feedback. These simple light flashes or beeps are usually more annoying than useful. They tell us that something has happened, but convey very little information about what has happened, and then nothing about what we should do about it. When the signal is auditory, in many cases we cannot even be certain which device has created the sound. If the signal is a light, we may miss it unless our eyes are on the correct spot at the correct time. Poor feedback can be worse than no feedback at all, because it is distracting, uninformative, and in many cases irritating and anxiety-provoking.
Too much feedback can be even more annoying than too little. My dishwasher likes to beep at three a.m. to tell me that the wash is done, defeating my goal of having it work in the middle of the night so as not to disturb anyone (and to use less expensive electricity). But worst of all is inappropriate, uninterpretable feedback. The irritation caused by a “backseat driver” is well enough known that it is the staple of numerous jokes. Backseat drivers are often correct, but their remarks and comments can be so numerous and continuous that instead of helping, they become an irritating distraction. Machines that give too much feedback are like backseat drivers. Not only is it distracting to be subjected to continual flashing lights, text announcements, spoken voices, or beeps and boops, but it can be dangerous. Too many announcements cause people to ignore all of them, or wherever possible, disable all of them, which means that critical and important ones are apt to be missed. Feedback is essential, but not when it gets in the way of other things, including a calm and relaxing environment.
Poor design of feedback can be the result of decisions aimed at reducing costs, even if they make life more difficult for people. Rather than use multiple signal lights, informative displays, or rich, musical sounds with varying patterns, the focus upon cost reduction forces the design to use a single light or sound to convey multiple types of information. If the choice is to use a light, then one flash might mean one thing; two rapid flashes, something else. A long flash might signal yet another state; and a long flash followed by a brief one, yet another. If the choice is to use a sound, quite often the least expensive sound device is selected, one that can only produce a high-frequency beep. Just as with the lights, the only way to signal different states of the machine is by beeping different patterns. What do all these different patterns mean? How can we possibly learn and remember them? It doesn’t help that every different machine uses a different pattern of lights or beeps, sometimes with the same patterns meaning contradictory things for different machines. All the beeps sound alike, so it often isn’t even possible to know which machine is talking to us.
Feedback has to be planned. All actions need to be confirmed, but in a manner that is unobtrusive. Feedback must also be prioritized, so that unimportant information is presented in an unobtrusive fashion, but important signals are presented in a way that does capture attention. When there are major emergencies, then even important signals have to be prioritized. When every device is signaling a major emergency, nothing is gained by the resulting cacophony. The continual beeps and alarms of equipment can be dangerous. In many emergencies, workers have to spend valuable time turning off all the alarms because the sounds interfere with the concentration required to solve the problem. Hospital operating rooms, emergency wards. Nuclear power control plants. Airplane cockpits. All can become confusing, irritating, and life-endangering places because of excessive feedback, excessive alarms, and incompatible message coding. Feedback is essential, but it has to be done correctly. Appropriately.
CONCEPTUAL MODELS
A conceptual model is an explanation, usually highly simplified, of how something works. It doesn’t have to be complete or even accurate as long as it is useful. The files, folders, and icons you see displayed on a computer screen help people create the conceptual model of documents and folders inside the computer, or of apps or applications residing on the screen, waiting to be summoned. In fact, there are no folders inside the computer—those are effective conceptualizations designed to make them easier to use. Sometimes these depictions can add to the confusion, however. When reading e-mail or visiting a website, the material appears to be on the device, for that is where it is displayed and manipulated. But in fact, in many cases the actual material is “in the cloud,” located on some distant machine. The conceptual model is of one, coherent image, whereas it may actually consist of parts, each located on different machines that could be almost anywhere in the world. This simplified model is helpful for normal usage, but if the network connection to the cloud services is interrupted, the result can be confusing. Information is still on their screen, but users can no longer save it or retrieve new things: their conceptual model offers no explanation. Simplified models are valuable only as long as the assumptions that support them hold true.
There are often multiple conceptual models of a product or device. People’s conceptual models for the way that regenerative braking in a hybrid or electrically powered automobile works are quite different for average drivers than for technically sophisticated drivers, different again for whoever must service the system, and yet different again for those who designed the system.
Conceptual models found in technical manuals and books for technical use can be detailed and complex. The ones we are concerned with here are simpler: they reside in the minds of the people who are using the product, so they are also “mental models.” Mental models, as the name implies, are the conceptual models in people’s minds that represent their understanding of how things work. Different people may hold different mental models of the same item. Indeed, a single person might have multiple models of the same item, each dealing with a different aspect of its operation: the models can even be in conflict.
Conceptual models are often inferred from the device itself. Some models are passed on from person to person. Some come from manuals. Usually the device itself offers very little assistance, so the model is constructed by experience. Quite often these models are erroneous, and therefore lead to difficulties in using the device.
The major clues to how things work come from their perceived structure—in particular from signifiers, affordances, constraints, and mappings. Hand tools for the shop, gardening, and the house tend to make their critical parts sufficiently visible that conceptual models of their operation and function are readily derived. Consider a pair of scissors: you can see that the number of possible actions is limited. The holes are clearly there to put something into, and the only logical things that will fit are fingers. The holes are both affordances—they allow the fingers to be inserted—and signifiers—they indicate where the fingers are to go. The sizes of the holes provide constraints to limit the possible fingers: a big hole suggests several fingers; a small hole, only one. The mapping between holes and fingers—the set of possible operations—is signified and constrained by the holes. Moreover, the operation is not sensitive to finger placement: if you use the wrong fingers (or the wrong hand), the scissors still work, although not as comfortably. You can figure out the scissors because their operating parts are visible and the implications clear. The conceptual model is obvious, and there is effective use of signifiers, affordances, and constraints.
FIGURE 1.8.Junghans
Mega 1000 Digital Radio Controlled Watch. There is no good conceptual model for understanding the operation of my watch. It has five buttons with no hints as to what each one does. And yes, the buttons do different things in their different modes. But it is a very nice-looking watch, and always has the exact time because it checks official radio time stations. (The top row of the display is the date: Wednesday, February 20, the eighth week of the year.) (Photograph by the author.)
What happens when the device does not suggest a good conceptual model? Consider my digital watch with five buttons: two along the top, two along the bottom, and one on the left side (Figure 1.8). What is each button for? How would you set the time? There is no way to tell—no evident relationship between the operating controls and the functions, no constraints, no apparent mappings. Moreover, the buttons have multiple ways of being used. Two of the buttons do different things when pushed quickly or when kept depressed for several seconds. Some operations require simultaneous depression of several of the buttons. The only way to tell how to work the watch is to read the manual, over and over again. With the scissors, moving the handle makes the blades move. The watch provides no visible relationship between the buttons and the possible actions, no discernible relationship between the actions and the end results. I really like the watch: too bad I can’t remember all the functions.
Conceptual models are valuable in providing understanding, in predicting how things will behave, and in figuring out what to do when things do not go as planned. A good conceptual model allows us to predict the effects of our actions. Without a good model, we operate by rote, blindly; we do operations as we were told to do them; we can’t fully appreciate why, what effects to expect, or what to do if things go wrong. As long as things work properly, we can manage. When things go wrong, however, or when we come upon a novel situation, then we need a deeper understanding, a good model.
