In an experimental set-up called dichotic listening, subjects hear two voices at once over a set of headphones. They must listen carefully to one of the voices and repeat each word that it says as the words are spoken, a task called shadowing. This voice is called the attended voice, while the other voice is unattended. Subjects become quite good at the shadowing task after a few minutes, repeating the attended speech quite accurately. However, after a few minutes of shadowing, subjects have no idea what the unattended voice was saying, or even if it was speaking English or not. They are able to report basic features, such as whether the voice was male or female, but are unable to remember anything about the content of the speech.

Dichotic listening provides compelling evidence for limits on attention. Broadbent, a researcher who performed some of the first dichotic listening experiments, theorized that our mind can be conceived as a radio receiving many channels at once. Each channel contains distinct sensory perceptions, as in the two auditory events in the dichotic listening task. Because our attention is limited, it is difficult to spread attention thin over several channels at once. In fact, we only have enough resources to effectively attend one channel at once. Therefore, we need some mechanism to limit the information that we take in. Broadbent's filter theory fills this role. The filter processes all the stimuli we are exposed to, and decides which to send through the one attended channel.

The filter can be directed by top-down or bottom-up influences. Top- down influences include a person's own intentions and expectations. If I am trying to read a book, then my intention to read will direct my attention to the words on the page, constituting a top-down influence. Bottom-up influences, by contrast, are directed by stimuli in the world that "catch" our attention. If someone taps me on the shoulder while I'm reading, the tap will direct my attention away from the book and toward that person, constituting a bottom-up influence.

The cocktail-party effect is a combination of top-down and bottom-up influences on attention. Imagine you are at a party where many conversations are going on at once. You can tune out other voices and pay attention only to the one that interests you (this voice is the attended channel). Yet, if someone at the next table mentions your name, your attention will be suddenly drawn to that conversation. In the dichotic listening task, subjects almost always noticed if the unattended voice mentioned their name, even if they couldn't report anything else that it said. The cocktail-party effect also works for other words of personal importance, such as the name of your favorite restaurant or a movie that you just saw, or the word "sex."

This effect is bottom-up in that it is driven by a stimulus in the world; we do not intend to direct our attention to that other voice or conversation, but it is drawn there nonetheless. However, it also depends on top-down influences in that it hinges on what is important or familiar to a certain person. Stimuli are more likely to grab attention in this way if they have been previously primed--that is, if they have been thought about recently or often. So, we can see in the cocktail effect that the channel filter can be directed by a combination of top-down and bottom-up influences.

Sometimes we may need to pay attention to more than one channel at a time. In some cases, if the stimuli are different enough, we may be able to pay attention to two things at once; we will examine this further in the section on interference. However, in cases where the stimuli are similar--for example, a friend's voice and Oprah on television--we must rapidly switch between the two channels. This is possible because of sensory store, a very short-term form of memory that stores what we hear and see, regardless of whether it was attended or not, for a period of a few seconds. So, you can listen to the friend channel for a few seconds, and then you can switch to the Oprah channel and retrieve her last few words from sensory store. However, this is difficult, and most people find that they miss some information from each channel, so you may find yourself often asking your friend to repeat herself.