Assumption- the sensory world is too rich for complete analysis. Attention is like a spotlight. Everywhere else is darkness but we are able to focus our attention on an area.
An example: a cocktail party phenomenon. Lots of people talking to each other and noisy. You get involved with another individual. It’s hard to hear them from background noise. You strain to hear them. You put forth a lot of effort and narrow your attention on that person. We are not really aware of the other conversations going on around us.
Lab version: Cherry - 1953 realized by using stereophonic recordings through headphones he could create a cocktail party atmosphere. He created a dichotic listening task. Subjects, via headphones, hear too different messages, one in each ear. You are told to shadow one of these messages. Repeat the message out loud as they hear it. Doing this task required effort, and assumed to require full attention. In general people do this pretty well but report that it requires a lot of attention. Subjects claim to not really hear the message in the unshadowed ear and display poor memory for its content afterwards.
Is the unattended message not even being processed by the cognitive system?
Cherry did many variations on this experiment.
Cherry tried changing the language of the unshadowed message halfway through. Afterwords subjects were asked if they noticed anything while shadowing the message. Subjects reported not noticing the language change.
Cherry tried playing the message backwards halfway through. Playing it backwards doesn’t sound like language anymore and disrupts the rhythm and flow of language. Once again subjects noticed no changes.
Subjects did notice when the gender of the speaker in unshadowed ear changed half way through the experiment.
Subjects noticed when beeps and tones were played on top of the unshadowed message.
Why do some thins get through and not others?
Broadbent used a split-span study instead of dichotic listening. Instead of having participants focus on one ear only, Broadbent wanted to force participants to hear three digits being spoken in both ears presented in simultaneous pairs every half second. Total of six digits presented to participant. Participant was asked to repeat what they heard.
Subjects reported the numbers by ear. Left ear numbers then right ear. Why would they not do them by order?
Variation: He then forced people to do it by pairs. Subjects could do it but reported it being much much much harder than doing it by ear.
In 1958, Broadbent published Perception and Communication . It caused people to rethink entirely that attention was a mental state that required a whole new way of viewing. He proposed viewing the mind as a serial communication channel. The mind can only do one thing at a time because it has to cope with more information that it can possibly cope with. The mind does this through the use of a selective filter to block unwanted input.
(I) Filter Model
The attentional filter can be tuned to block out stimuli according to certain physical characteristics. Between the senses and filter, a buffer can briefly hold unanalyzed information so it can be analyzed later. Sounds a lot like an icon. Icon had not yet been discovered yet.
The buffer can explain Broadbent’s split-span data: since it’s hard to do two things at once, one set of numbers is held in the buffer and simply read off the second set after reporting the first set.
The filter can be tuned to block a person’s voice by physical characteristics (e.g. pitch). This explains why a change in language or backwards language doesn’t get through. It’s still the same pitch. This is why changes in gender, or tones and beeps, do get through. There are changes in pitch.
However, people tend to do more than one thing at once.
Broadbent claims that attending to two things at once (parallel processing) is an illusion. Example: Watching tv, talking on the phone, solving a crossword puzzle.
Broadbent would claim you are not focusing on all at the same time. We’re actually just switching quickly back and forth across tasks (multiplexing).
Moray - played the same word list 35 times in the unshadowed ear. Subjects still had poor memory for the words afterwords.
Moray did a variation where he would play the subjects own name in the unshadowed ear. About a third of the subjects heard it. It’s only a third but still this is problematic for the filter. This is the other half of the cocktail party effect.
More bad news….
Triesman (1960) – dichotic listening task, with shadowing. Then halfway through, she switched stories in each ear. Many subjects switched to shadowing the message in the unshadowed ear (violating instructions of experiment). It suggests that they were following a story or some other factors at work.
(2) Attenuation model (Triesman) 1960
Retains the idea of a buffer and a filter. But the filter plays a more active role. Before this theory, the filter was pretty dumb. It isn’t just looking for pitch or physical characteristics. She suggests that filter can be sensitive to a variety of characteristics such as info that has most recently entered awareness.
Some inputs have privileged status: information that always enters awareness. (examples: sounds of ones name, cry of one’s child).
So context and prior knowledge affect the filter is tuned. Also called the “leaky filter” model.
Broadbent and Triesman are the only stimuli that receive attention are processed. (Both are examples of early selection models – stimuli that aren’t attended to drop out of the system entirely.)
Late Selection Models
All inputs are perceptually analyzed. Then the cognitive system decides what will enter consciousness.
Pertinence Model (Deutsch & Deutsch 1963; Norman 1968)
All sensory input is processed (pattern recognition). Only pertinent (important) input enters awareness.
Can this idea be proven?
Gray and Wedderburn 1960 – Subjects presented with word lists and receive shocks for a particular words. Over time subjects should become conditioned to those words. Then they participate in a dichotic listening task. Shadow one ear. The shocking words were occasionally played in the unshadowed ear. Subjects have no memory of having heard the “shock” words. G and W realized that people may not consciously be aware but that they may show signs in body. Subjects were also wearing galvanic skin response (GSR) monitors. They measured skin conductance (sweating). GSR activity spiked when the “shock” words were played. The spikes show that everything was analyzed but never entered into conscious awareness.
Attention is not a fixed capacity. There might be no limit on how much information that people can pick up at any point in time. Crucial variables: practice which causes a given task to require less attention. (learning to drive is an example).
Empirical demonstration: (Spelke, Hirst, Neisser 1976)
Trained college students to read and take dictation at the same time. Initially students did terrible. Would lose track of words etc etc. It required a lot of effort and required a semester of practice but were able to do the task eventually.
Norman and Bobrow (1975)
Performance on a task is limited by the amount of mental effort that we can expend on it. Performance deteriorates when we divide our attention because several processes are competing for the same limited resources. Distinguished kinds of tasks. Data limited tasks were tasks where performance is limited by memory ability or the quality of the stimulus. These are cases where additional attention will not performance. (example: long-distance eavesdropping).
Resource limited tasks are tasks where performance will improve if more resources are applied to the task. An example is turning off the car radio when learning to drive. Practicing tasks will lower the cognitive resources that each requires.