Please enable JavaScript.
Coggle requires JavaScript to display documents.
MODELS OF ATTENTION :PENCIL2: P+C2 LECTURE 1 - Coggle Diagram
MODELS OF ATTENTION
:PENCIL2:
P+C2 LECTURE 1
Attention
It is the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought (1) .
It implies withdrawal from some things to deal effectively with other (2)” - William James, 1890. The Principals of Psychology, chapter XI (Attention).
(1) - Information is selected for processing
(2) - Some information is suppressed or not processed.
Selective Attention
“The ability to focus on that which is important to the task at hand while ignoring or suppressing task-irrelevant information.”
The Gist...
When do we attend?
Early or late selection?
Pre or post processing such as identification or recognition?
How far down the system do unattended elements get?
How do we attend?
Structure or capacity models?
Can you not attend to something because of the structural limitations of the system or because the system has limited capacity to pay attention?
Early selection
Broadbent's Filter Model
Early selection:
The cognitive system selects the information it is attending to early.
Focuses on that information and doesn’t process any other information that you are not attending to - it’s just discarded.
The information not attended to is filtered out before any processing such as identification.
Structural not capacity limitations
Information is passed from one system to another system.
What limits the information passed on between systems is a structural limitation - a filter.
They are built in, not due to processing power limitations of the system.
What’s not happening, and what happens in other models is that you fail to pay attention just because you don’t have enough processing power.
You only have so much capacity to pay attention, and anything beyond that means that you don’t process or you don’t see - you don’t pay attention to the stimuli.
Bottleneck Model
Store for physical characteristics for all stimuli (Sensory store)
Selection of “to be attended to…” from the store
Filter that stops information that you’re not attending to from passing onto the cognitive, perceptual system.
Processing limited to "attended to" material
Perceptual / semantic analysis
This is where meaning is extracted.
Predictions of the model
All-or-nothing processing
.
When a filter is focused on one channel there should be no real processing beyond basic physical elements of the other channel.
No “divided attention”
Evidence from Cherry's work.
Dual-tasking
...is actually doing two things, one after another.
Dual tasking performance involves channel switching.
Switching the filter costs time (300-500ms)
Dichotic listening (shadowing)
Cherry (1953)
Two inputs, one into each ear.
The attended input, the information that you need to focus on and that you need to shadow.
Participants asked to actually say the words that were coming into their ear.
The other input coming into the other ear is ignored, and participants are supposed to not pay attention to that channel.
What’s important is how much information participants could give about the unattended channel; the ignored input.
How much could they recall?
Unattended speech
What CAN be reported
Whether the signal is voice or noise
Gender of the speaker
If the signal/speech becomes the same in both channels
Basic physical characteristics
What CAN'T be reported
The content of the message (even if simple, e.g. single word is repeated a number of times)
If the signal is speech or reversed speech
Cannot tell the language of the speech
Any changes in the language
Meaning
Conclusions
Unattended material is not processed beyond the physical characteristics.
No semantic processing, e.g. meaning extracted or identification.
Only material in the selected channel (ear) is processed.
Can recall semantic information from the attended channel.
Selection is early, prior to processing.
Evidence for processing in the unattended channel
Some individuals notice their name in the unattended channel.
Said the individual's name and then given an instruction in the unattended channel.
E.g. “Emily, change channel.” or “Emily, put up your hand.”
The attended channel was just the usual stuff that they had to shadow.
Around 30% recalled the information they had been given just after their name had been said in the unattended channel (Moray, 1959; Wood & Cowan, 1995)
Suggests that information from the unattended channel is being processed.
Physiological response to words in the non-attended channel that were pre-associated with electric shocks.
Conditioned people to fear a stimulus (certain words)
When they heard the word, they got an electric shock.
When they said words that had been associated with electric shocks in the unattended channel, participants actually showed the fear response.
Measured fear response using the Galvanic Skin Response (a measure of how much you sweat).
Same response even extended to synonyms (Corteen & Dunn 1974; Von Wright et al., 1975)
Shows some processing of meaning in the unattended channel.
Opposes Broadbent
.
Participants follow a coherent message when switched to “unattended” channel
(Treisman, 1960)
If you started a sentence in the attended channel and then continued that sentence in the unattended channel, participants switched across the channels to understand the full, coherent message.
Treisman's Modified Model: "Attenuated, not eradicated' - Bottleneck Model
Unattended information is not filtered out, but attenuated (turned down)
The signal from the unattended channel is considerably weaker/weakened.
However, some stimuli are “louder” than others
Primed stimuli such as your name
Words associated with electric shock
Processed even when unattended.
Note: the attenuation still happens early
Prior to processing of perceptual/meaning processing
Still a structural model
The attenuator is the new “part” in the model (replacing the filter)
Late selection
Deutsch & Deutsch (1963)
All incoming information is processed to the highest level
All stimuli’s physical characteristics are processed, it is names and categorised (Duncan, 1980)
“The meaning of all current stimuli are extracted in parallel and without interference.” (Kahneman, 1973)
In early filter models, only physical characteristics are processed in parallel.
Selection occurs at the level of response/awareness.
Structural and Late
Everything goes through processing
Selection/Filtering happens after processing
New bottleneck is before response.
Still a structural model.
Evidence
(Lackner and Garrett, 1972)
Needs to be strong evidence of processing of unattended messages
Participant uses material from Unattended Channel to resolve ambiguous sentences.
(Attended channel): “The spy
put out
the torch as our signal to attack.”
(Unattended channel): “The spy
extinguished
the torch in the window.” OR (Unattended channel): “The spy
displayed
the torch in the window.”
Participants' paraphrases reflected the content of the unattended channel.
Stroop
Two elements of the stimuli (word and colour)
Both elements are encoded (processed)
Both elements compete for response
One an automatic response (reading)
One a controlled response (naming or counting)
Is the unattended information remembered because…
(1) The filter is leaky?
Unattended information is identified/processed
(2) The filter is slipping?
The attention system ‘slips’ and attends to material (and processes) it shouldn’t.
The idea of the filter says that slipping is compatible with early selection models.
Lachter et al., (2004)
says that evidence for the filter leaking comes from experiments that did not ensure this information was truly unattended.
Perhaps not a true overlap of channels?
Capacity Models
No structure bottleneck
The cognitive system has a limited amount of processing capacity.
“Paying attention” is the equivalent of “investing energy”
So the limiting factor is not a structural one (filter/attenuator) (early/late)
It is how much you have to “spend” in terms of resources.
Kahneman (1972)
Any task demands a processing capacity (“juice”)
This is dependent on…
...difficulty of task
...individual expertise on task
Task performance is dependent on the allocation of capacity to task.
This is dependent on…
...enduring dispositions (habits and preferences)
...momentary (need right now)
...evaluation of demands on capacity
Evidence:
Dual-task: doing two things at once
Do task 1 on its own
Do task 2 on its own
Do tasks 1 and 2 together
Drop in performance (accuracy/RT)
Dual Task Decrement
Dual Task Evidence
Individuals can learn to do two things at once
E.g. Reading prose whilst writing to diction
Initially impossible but 6 weeks of practice lead to competence (
Spelke, Hirst and Neisser, 1976
)
Treisman and Davies (1973) had participants monitor:
Two visual screens
Two auditory inputs (on each ear)
One visual/one auditory input
All conditions showed dual task decrement
Two visual screens and two auditory inputs had larger dual task decrements (worse performance) compared to the condition with one visual and one auditory input.
Modality specific resources?
Role of Arousal
Too much - reduced focus, more irrelevant details noticed.
Too little - no real motivation
Increased levels can counteract capacity limits.
So we have evidence for and against
Structural models that filter/attenuate.
Pre-processing (Early, Broadbent and Treisman)
Post processing (Late, Deutsch and Deutsch)
Resource/Capacity models
These all have limitations and evidence issues
The next step is perhaps a model that combines the above.
Perceptual Load Theory
Lavie:
The system is limited (as in early selection)
But will process everything until it runs out of capacity.
So when there is low demand on capacity, more of what is seen is processed.
-Increased chance of the non-relevant unattended elements processed.
So when there is a high demand on capacity, less of what is seen is processed.
Non-relevant unattended parts filtered out.
You can have early and late selection within your system
When that happens is dependent on the demands of the attended stimuli.
Stimuli...
“Is “z” or “x” present in these stimuli?”
Low load tasks
Little demand for perceptual processing.
E.g. 'z'
Compatibility effects:
Press Key 1 if “z” is present, Key 2 if “x” is present at the fixation point (centre)
Incompatible distractors reduce performance (increase reaction time).
RT is slowest for Incompatible
, then Neutral, then compatible
Interference from distractor leads to response competition.
High load tasks
Increased demand for perceptual processing.
E.g. 'mvznsk'
Compatibility effects:
Same task procedure
Different stimuli
RT is slowest for Incompatible
, then Neutral, then compatible
If irrelevant stimuli (distractors) are processed
Prediction from the model…
-
Increased load should decrease compatibility effect
Results...
Dependent measure: RT for incongruent – RT for congruent
Higher Reaction Time
More interference by incongruent (the incongruent takes longer)
Low or no difference
No interference
Low perceptual load (single letter stimuli)
Strong compatibility effect
High RT for incongruent (distractor has distracted) compared to congruent
High perceptual load (multiple letters)
No real compatibility effect
No real difference in RT between congruent/incongruent.
All perceptual capacity is taken up by processing the stimuli, no capacity left to notice distractors.
Low load means late filtering
More of the scene processed
“Broad focus of attention”
High load means system filters out irrelevant stimuli early (not processed)
Less of the scene processed
"Narrow focus of attention"
Further results...
Stimuli...
What is the colour of the vertical line?
Which of the lines (vertical and horizontal) is longer
Did you see...?
(a) a small triangle (b) a small circle (c) a small square (d) a large square
People who were given the second task tend to not notice the small square in the bottom left hand corner of the final slide, because the length of line comparison task that they are doing is much more difficult.
Inattentional Blindness
Cartwright-Finch & Lavie (2006)
Inattentional Blindness
Not noticing task irrelevant elements in the visual scene
E.g. Basketball and Gorillas etc.
Increased load
Reduced participants noting irrelevant elements of stimuli.
High perceptual load reduces the compatibility effect, stops you being distracted, and increases Inattentional Blindness.
The Role of Working Memory
Perceptual load reduced interference
But what about general control processes beyond the perceptual system?
Even when distractors are processed they seldom gain control of behaviour.
Cognitive control processes (Working Memory) maintains a clear distinction between task-relevant and task-irrelevant information.
What happens to interference when we vary the load on Working Memory?
Initially, the participant was given a set of digits to remember.
Low WM load: 0, 1, 2, 3, 4
High WM load: different sequence each time.
Then asked to categorise faces (pop star/politician) with:
Congruent names
Incongruent names
Asked to report probe digit.
High WM load - increases the impact of the incongruent pictures/names
Larger difference in RT between congruent and incongruent compared to Low WM load.
With the low working memory task, you are less likely to be distracted by the incongruent names than in the high working memory task.
A task that takes up a lot of working memory increases the ability of the stimuli to interfere.
High WM load increased interference
High perceptual load reduces interference
Overloading the passive system reduces interference.
Overloading the active one increases interference.