Harold Thimbleby, 2003
There are two basic types of electric house light switch: toggle and push (strictly, push to change). Both types are used on a wall in the interaction laboratory at UCLIC. Here is a photograph of them:
The photograph shows what look like four toggle switches and something else, second from the left (which we will get to later).
A typical toggle switch has a 'toggle' that can be in one of two possible positions, up or down. The light a toggle switch controls can also be in one of two states, on or off, so there are two natural ways to connect the switch to the lights. One way is 'toggle up/light on and toggle down/light off', and the other way is 'toggle up/light off and toggle down/light on'. (There are two more ways, but then the two states of the switch would not control the two states of the light: these ways would be considered a wiring fault if they ever occurred in practice.)
In the UK, the conventional choice is that 'down' is on, and 'up' is off. In the USA, the convention is reversed. If you know the convention, you can look at a toggle switch and decide what to do, just by knowing whether you want the light on or off. You can also diagnose problems: thus, if the toggle is down but the light is off, there is a fault (or you are in the wrong country, or both).
The toggle switch exhibited here is mounted on the wall on its own (on the far left). The toggle switch physically behaves so that when the bottom of the toggle is pushed, the top of the toggle pops out and vice versa. Thus when the toggle superficially looks like it is 'up' it is actually 'down.' We easily learn affordance conventions like this. Older-style toggle switches have a 'dolly' that is more clearly up or down, but which probably don't look so stylish.
In contrast, a push switch changes its state each time it is pushed. In the examples here, there is no perceptual difference between the two sorts of switch. Are the affordances the same, then? The natural way for a push switch to control a light is that alternate pushes switch the light on, and the other pushes switch it off.
The push switches exhibited here are mounted on the wall in a single bank of three (on the right). If a user wishes to change the state of a light, using a push switch means they have only one action, namely a push. This is simpler; the user has half the number of actions to choose from to achieve any goal, whether to switch the light on or off.
A common problem with a push switch is when there is a delay in the light (or other controlled device, such as a TV) responding to the user's actions: if the user is not sure they have pushed or pushed hard enough, they may push again, and of course this action may return the light to its initial state, which the user wanted to change. This problem happens in our room: some of the lights are slow to come on when switched on! In contrast to a toggle switch, a user cannot look at a push switch and, from the visible switch state alone, know whether the light should be (or is about to be after a delay) on or off. With the lights here, for a few seconds the user may think they've switched the lights on, but the lights do not show it: so is the user wrong, or the lights slow? A user may not know.
Sometimes a single light is wired to two or more switches that each separately control it. This is typical for stair lighting, where there are often switches at the top and bottom of stairs. The natural mapping of one switch in a pair is changed when the other is used. If one toggle is changed, the other switch has essentially changed from a US to a UK orientation or vice versa. Now a user cannot look at either switch alone and know whether the light should be on or off. Using a push switch for this situation is therefore preferable, since it simpifies the user interface by removing this confusion.
(There is another switch in our room that controls the lights in this way: there is another entrance to this room, and it is important to be able to switch the lights on and off from either door, for instance so you can enter the room, switch the lights on, and leave by the other door.)
The bank of three switches (on the right) are all push switches. Each switch controls a different bank of ceiling lights in this room. The fluorescent lights in this room have a delay: when they are switched on, they do not come on immediately. There are three banks of lights, and the mapping from switches to banks, though presumably fixed, is not defined. Suppose your task is to switch on the front bank of lights (regardless of other aspects of room lighting). You might start by pressing the leftmost switch. If a bank of lights goes out, you should press the switch again. If nothing happens, you need to wait. If the wrong bank of lights comes on after a delay (how long should you wait?), you must press it again. If at this stage you have not got the front bank of lights on, you must try another push switch. If you are unlucky, you will have to try the third switch too. If you are really unlucky, perhaps you did not wait long enough when you pushed a switch a second time, and you will now have to start again, but wait longer after each push.
In our room, the orientation of the bank of three switches is East-West, but the orientation of the banks of lights themselves is North-South, missing another opportunity for simpler affordances. The order of the switches gives little clue to the permutation of their mapping to the banks, though probably the middle switch controls the middle bank (why?) -- but if you are not familiar with this room, you can't easily distinguish the middle light bank anyway.
A single push switch and light combination is arguably easier to use than a toggle switch, and the advantages are more apparent when two (or more) switches control a single light. However, when several push switches control several lights, as here, the user interface becomes harder. If there had been a row of toggle switches, we still wouldn't be able to deduce the switch-to-lightbank mapping, but we would know whether banks were on or off. For example if our task was to switch the front bank of lights on, and two switches were in the on state and one off, it would be obvious what to do: namely, change the single off switch to its on state.
More advanced light switches may include additional indicators. They could have a light on them to show whether they are in the on or off state. This would be useful for stair lights and push switches, since then a user could reliably turn lights off -- a safety precaution when replacing a faulty light bulb. Unfortunately, lights on switches have other purposes: they are very useful to help locate a switch in the dark, so they are usually on when the controlled light is off!
There is another switch here that can be seen in the photograph, second from left, that we have not discussed: a keyswitch, which is not intended for normal use. A key is required, which turns off the lighting circuit altogether. When this is done, the emergency lighting should come on, as if there had been a power cut. Here, if you switch off the circuit the lights come on! Clearly this is a feature for expert users, and its physical design, requiring a special key, ensures that only experts can use it.
If the user interface to the lighting in our room had been better designed, there would have been less to notice, less to say, less to learn, and less fun. The whole topic of switch design is fascinating: for example, an emergency switch might designed to be 'turn to switch on' and 'push to switch off' (since pushing is the easiest thing to do with an emergency switch). Note how design very rapidly gets more complicated as more lights or more switches are added. Even a row of three switches controlling just three lights is easy to get 'wrong.'
Such a simple systems raise many fascinating 'why?' questions ... and what would you have done?
Further reading: see Don Norman's classic Design of Everyday Things (Basic Books, 1988), where he has a rant about light switch design.