Chapter 04: Paying Attention
- Selective Attention: The focusing of cognitive resources on one or a small number of tasks to
the exclusion of others
- Dichotic Listening Task:
A person listens to an audiotape over a set of headphones. On the tape are different
messages, recorded so as to be heard simultaneously in opposite ears. Participants in a
dichotic listening task typically are played two or more different messages and asked
to “shadow” – that is, to repeat aloud- one of them.
Information presented at a rapid rate (~150 words/minute) so the shadowing task is
At the end of the task, participants are asked what information they remember from
either message – the attended message or the unattended message.
Sometimes the tapes are recorded so that both messages are heard in both ears- called
binatural presentation –and some researchers have used it in addition to dichotic
The person must concentrate on the message to be shadowed. Because the rate of
presentation of information is so fast, the shadowing task is difficult and requires a
great deal of mental resources.
Fewer resources are available to process information from the nonshadowed,
- Cherry (1953) demonstrated in a classic study that people can, with few errors, shadow a
message spoken at a normal to rapid rate.
When researchers later questioned these participants about the material in the
unattended message, they could nearly always report accurately whether the message
contained speech or noise, and if speech, whether the voice was that of a man or a
When the unattended message consisted of speech played backwards, some
participants reported noticing that some aspect of the message, which they assumed
to be normal speech, was vaguely odd.
- Participants could not recall the content of the unattended message or the language in which
it was spoken.
In one variation of the procedure, the participants did not notice the language switch
from English to German of the unattended message.
Participants in another experiment (Moray, 1959) heard prose in the unattended
message and a short list of simple words in the unattended message. They failed to
recognize the occurrence of most words in the unattended message.
- Broadbent (1958) proposed a filter theory of attention, which states that there are limits on
how much information a person, can attend to at any given time. The person uses an
attentional filter to let some information through and block the rest.
- The filter is based on some physical (in this particular example, basic acoustic) aspect of the
attended message: the location of its source or its typical pitch or loudness, for instance.
Only material that gets past the filter can be analyzed later for meaning.
The filter selects information for later processing.
- The meaning from an unattended message is not processed. Broadbent’s filter theory maintains that the attentional filter is set to make a selection
of what message to process early, typically before the meaning of the message is
According to Broadbent’s model, it should not be possible to recall any of the meaning
of an unattended message.
- Broadbent believed instead that what is limited is the amount of information we can process
at any given time.
Two messages that contain little information, or that present information slowly can
be processed simultaneously.
For example: A participant may be able to attend simultaneously to more than
one message if one repeats the same word over and over again, because it
would contain little information. In contrast, messages that present a great deal
of information quickly overcome the filter; fewer of them can be attended to at
The filter thus protects us from “information overload” by shutting out
messages when we hear too much information to process all at once.
- Moray (1959) – “Cocktail Party Effect”
Shadowing performance is disrupted when one’s own name is embedded in either the
attended or the unattended message. Moreover, the person hears and remembers
hearing his name.
- Filter theory predicts that all unattended messages will be filtered out- - that is, not processed
for recognition or meaning – which is why participants in dichotic listening tasks can recall
little information about such messages.
- Cocktail Party effect shows that people sometimes do hear their own name in an unattended
message or conversation, and hearing their name will cause them to switch their attention to
the previously unattended message.
Moray concluded that only “important” material can penetrate the filter set up to block
Participants did not always hear their name in the unattended channel: When
not cued in advance to be vigilant, only 33% of the participants ever noticed
Alternative explanation for the name recognition finding is that the shadowing
task does not always take 100% of one’s attention.
Attention occasionally lapses and shifts to the unattended message.
During these lapses, name recognition occurs.
- Treisman (1960) Effect of semantics/ Attentional Leakage
She played participants two distinct messages, each presented to a different ear, and
asked the participants to shadow one of them. At a certain point in the middle of the
messages, the content of the first message and the second message was switched so
that the second continued the story-line of the first and vice versa.
Immediately after the two messages “switched ears”, many participants reported one
or two words from the “unattended ear”.
Treisman reasoned that participants might be basing their selection of which message
to attend to at least in part on the meaning of the message – a possibility that filter
theory does not allow for.
- Wood and Cowan (1995)
Participants perform a dichotic listening task. Two of the groups shadowed an excerpt
from The Grapes of Wrath (read very quickly, at a rate of 175 words/minute) in the
attended channel (always presented to the right ear) and were presented with an
excerpt from 2001: A Space Odyssey in the unattended channel. Five minutes into the task, the speech in the unattended channel switched back
to backward speech for 30 seconds.
Two groups differed only in how long the “normal” speech was presented after
the backward speech: 2 ½ minutes for one group; 1 ½ minutes for the other.
Third control group of participants heard an unattended message with no
Wood and Cowan first asked whether the people who noticed the backward speech in
the unattended message showed a disruption in their shadowing of the attended
The answer was a clear yes.
Wood and Cowan counted the percentage of errors made in shadowing and noted that
the percentage rose to a peak during the 30 seconds of the backward-speech
Effect was especially dramatic for those people who reported noticing the
Participants who did report hearing backward speech made noticeably
more errors, which peaked 10 to 20 seconds after the backward speech
Control participants showed no rise in their shadowing errors, nor did most of
the participants who did not report noticing the backward speech.
Wood and Cowan concluded that the attentional shift to the unattended message was
unintentional and completed without awareness.
Based this conclusion on the facts that detection of the backward speech
interrupted and interfered with shadowing and that error rates peaked in a
uniform time for all participants who noticed the backward speech.
Believed that the participants who noticed the backward speech had their
attention “captured” by the backward speech, which led to poorer performance
on the main shadowing task.
- Conway, Cowan, and Bunting (2001) showed that research participants who detect their
name in the unattended message are those who have a lower working-memory span.
20% of participants with high working-memory spans detected their names in the
unattended channel, compared with 65% of participants with low working-memory
A lower working-memory capacity means less ability to actively block the unattended
People with low working-memory spans are less able to focus.
- Proposed by Anne Treisman (1960)
Instead of considering unattended messages completely blocked before they could be
processed for meaning (as in filter theory), Treisman argued that their “volume” was
“turned down.” [Some meaningful information in unattended messages might still be
available, even if hard to recover]
- Incoming messages are subjected to three kinds of analysis:
1. The message’s physical properties, such as pitch or loudness
2. Linguistic, a process of parsing the message into syllables and words
3. Semantic, processing the meaning of the message.
- Words that have subjective importance (such as your name) or that signal danger (“Fire!”
“Watch out!”) have permanently lower thresholds they are recognizable even at low
volumes Words or phrases with permanently lower thresholds require little mental effort by
the hearer to be recognized.
- The context of a word in a message can temporarily lower its threshold.
If a person hears “The dog chased the…,” the word cat is primed – that is, especially
ready to be recognized.
Even if the word cat were to occur in the unattended channel, little effort would
be needed to hear and process it.
Hearing the previous words in a sentence primed the participants to detect and
recognize the words that followed, even when those words occurred in the unattended
- According to Treisman (1964), people process only as much as is necessary to separate the
attended from the unattended message.
If the two messages differ in physical characteristics, then we process both message
only to this level and easily reject the unattended message.
If the two messages differ only semantically, we process both level of meaning and
select which message to attend to based on this analysis.
Processing for meaning takes more effort
Messages not attended to are not completely blocked but rather weakened in much
the way that turning down the volume weakens an audio signal from a stereo.
Parts of the message with permanently lowered thresholds (“significant”
stimuli) can still be recovered, even from unattended messages.
Attenuation Theory Filter Theory
• Allows for many • Allows for only one
different kinds of kind of analysis for all
analyses of all messages
• Unattended messages messages, once
are weakened but the processed for physical
informatio they characteristics, are
contain is still discarded and fully
- Proposed by Deutsch and Deutsch (1963) and later elaboarated and extended by Norman
Theory holds that all messages are routinely processed for at least some aspects of
Attentional selection occurs after this routine processing.
- Filter theory hypothesizes a bottleneck – a point at which the processes a person can bring to
bear on information are greatly limited – at the filter.
- Late-selection theory also describes a bottleneck, but locates it later in the processing, after
certain aspects of the meaning have been extracted. All material is processed up to this point, and information judged to be most
“important” is elaborated more fully. This elaborated material is more likely to be
retained; unelaborated material is forgotten.
- Message’s “importance” depends on many factors, including its context and the personal
significance of certain kinds of content.
Also relevant is the observer’s level of alertness: At low levels of alertness (such as
when we are asleep), only very important messages (such as the sound of our
newborn’s cry) capture attention.
At higher levels of alertness, less important messages (such as the sound of a
television program) can be processed.
- Attentional system functions to determine which of the incoming messages it’s the most
important; this message is the one to which the observer will respond.
Attention, Capacity and Mental Effort
- Bottle neck Analogy: The smaller diameter of the bottle’s neck relative to the diameter of the
bottle’s bottom reduces the rate at which liquid can pass through. The wider the neck, the
faster the contents can pass through.
- Daniel Kahneman (1973) viewed attention as a set of cognitive processes for categorizing and
The more complex the stimulus, the harder the processing, and therefore the more
attentional resources are required.
People have some control over where they direct their mental resources, however:
they can often choose what to focus on and where to allocate their attentional
Many factors influence this allocation of capacity, which itself depends on the
extent and type of mental resources available.
The availability of mental resources, in turn, is affected by the overall level of
arousal, or state of alertness.
- Kahneman argued that one effect of being aroused is that more cognitive resources are
available to devote to various tasks.
The level of arousal also depends on a task’s difficulty.
We are less aroused while performing easy asks, than we are when performing
more difficult tasks.
We bring fewer cognitive resources to easy tasks, which, fortunately, require fewer
resources to complete.
- Arousal affects our capacity (the sum total of our mental resources) for tasks.
- Allocation policy is affected by an individual’s enduring dispositions, momentary intentions,
and evaluation of the demands on one’s capacity
We pay more attention to things we are interested in, are in the mood for, or have
- In Kahneman’s view, attention is part of what the layperson would call “mental effort.” The
more effort expended, the more attention we are using.
- Greater effort or concentration results in better performance on some tasks- those that
require resource-limited processing, performance of which is constrained by the mental
resources or capacity allocated to it. (Example, taking a midterm)
- On some other tasks, one cannot do better no matter how hard one try.
Example: Trying to detect a dim light or a soft sound in a bright and noisy room.
Even if you concentrate as hard as you can on such a task, your vigilance still may not
help you detect the stimulus
Performance is said to be data limited, meaning that it depends entirely on the
quality of the incoming data, not on mental effort or concentration. Automaticity and the Effects of Practice
- Over time, the attentional capacity required for a given task decreases.
- What affects the capacity a task requires?
The difficulty of the task
Individuality’s familiarity with the task
Practice is thought to decrease the amount of mental effort a task requires
The Stroop Effect
- A series of colour bars or colour words printed in conflicting colours.
- Participants were asked to name, as quickly as possible, the ink colour of each item in the
When shown bars, they did so quickly, with few errors and apparently little effort.
When the items consisted of words that named colours other than that of the ink in
which the item was printed, participants stumbled through these lists, finding it
difficult not to read the word formed by the letters.
- Automatic one that requires no attention and cannot be inhibited.
- The most common theoretical account of the Stroop effect is that for literate adults reading is
automatic, and its products, which conflict with the ink colour on a Stroop trial, interfere with
a participant’s ability to name the ink colour.
Automatic versus Attentional (Controlled) Processing
- Posner and Snyder (1975) offered three criteria for cognitive processing to be called
automatic processing: (1) it must occur without intention; (2) it must occur without
involving conscious awareness; (3) it must not interfere with other mental activity.
- Schneider and Shiffrin (1977) examined automatic processing of information under well-
controlled laboratory conditions
Asked participants to search for certain targets, either letters or numbers, in different
arrays of letters or numbers called frames
Example: A participant might be asked to search for the target J in an array of
four letters: B M K T (Note: this trial is “negative”, in the sense that the target is
not present in the frame.)
Previous work had suggested that when people search for targets of one type (such as
numbers) in an array of a different type (such as letters), the task is easy.
Numbers against a background of letters seem to “pop out” automatically.
The number of nontarget characters in an array, called distractors, makes little
inference if the distractors are of a different type from the targets.
Finding a specific letter against a background of other letters seems much harder.
Searching for J among the stimuli R J T, is easier than searching for the J among
the stimuli G K J L T, both of which are more difficult than finding a J amongst
When the target and the distractors are of the same type, the number of
distractors does make a difference.
Two conditions in the experiment:
Set of target letters or numbers, called the memory set, consisted of one
or more letters or numbers, and the stimuli in each frame were also
letters or numbers. Targets in one trial could become distractors in
The task was expected to be hard and to require concentration and
Consistent-mapping condition Target memory set consisted of numbers and the frame consisted of
letters, or vice versa. Stimuli that were targets in one trial were never
distractors in other trials.
The task was expected to require less capacity.
In addition, experimenters varied three other factors to manipulate the attentional
demands of the task
Frame size- the number of letters and numbers presented in each display
Number was always between 1 and 4
Slots not occupied by a letter or number contained a random dot
Frame time – the length of time each array was displayed
Varied from ~ 20 milliseconds to 800 milliseconds.
Memory set- the number of targets the participant was asked to look for in each
In the consistent mapping condition, thought to require only automatic processing
(because the targets and distractors were not the same type of stimuli), participants’
performance varied only with the frame time, not with the number of targets searched
for (memory set) or the number of distractors present (frame size)
Participants were just as accurate in searching for one as for four targets and in
searching among one, two, or four items in a frame. Accuracy depended only on
the length of time the frames were displayed.
In the varied-mapping condition, thought to require more than automatic processing
(because the targets and distractors could both be letters, or both numbers, and
because targets on one trial could become distractors on another), participants’
performance in detecting the target depended on all three variables: memory set size,
frame size, and frame time.
Distinguished two kinds of processing:
Used for easy tasks, and with familiar items
Operates in parallel (meaning it can operate simultaneously with other
Does not strain capacity limitations
Occurred in the consistent-mapping condition
Used for difficult tasks and ones that involve unfamiliar processes
Operates serially (with one set of information processed at a time)
Under conscious control
Use with non-routine or unfamiliar tasks.
Occurred in the varied-mapping condition
- Divided Attention: the ways in which a cognitive processor allocates cognitive resources to
two or more tasks that are carried out simultaneously
- Spelke, Hirst and Neisser (1976) study Two Cornell University students were recruited as participants. Five days a week, for
17 weeks, working in 1 hour sessions, these students learned to write words dictated
while they read short stories. Their reading comprehension was periodically tested.
After 6 weeks of practice, their reading rates approached their normal speeds.
By the end of 6 weeks, their scores on the reading comprehension tests were
comparable whether they were only reading stories (and thus presumably giving the
reading task their full attention) or reading stories while writing down dictated words.
Further investigation revealed that participants could also categorize the dictated
words by meaning and could discover relations among the words without sacrificing
reading speed or comprehension.