Michael Smith only
Breeuna09
ch4_attention_sp17.pptAttention chapter 4
Coglab- attentional blink
“Everyone knows what attention is. It is the taking
posession by the mind, in clear and vivid form, of one out
of what seem several simultaneously possible objects or
trains of thought. Focalization, concentration, of
consciousness are of its essence.”
“…withdrawal from some things in order
to deal effectively with others…”
What is attention?
-- William James (1890)
We can certainly introspect
about attention (à la James):
You know what it's like to think hard, or to
have your mind wander...
You know what it's like to have something
grab your attention...
You probably know how to tune out signals...
You can drive and listen
to the radio at once...
"A pool of mental effort that is selective, shiftable, and divisable..."
But we're scientists, so we're
going to study attention
a little more carefully...
You could probably derive a reasonable
introspective definition of attention
Attention:
ability to focus on specific stimuli/locations
aka
the interface between memory systems
Attention determines what information
receives further analysis…
Why do we have attention?
You are constantly bombarded
by information, but most does
not enter consciousness.
MANY ASPECTS
Many approaches
Attention as Info Processing
- Late 1800s/early 1900s – introspection
- E.g., describe how paying attn affects clarity of patch
- 1920s – 50s we lost our minds
- 1950s – renewed interest in attn research bc WW2 – pilots
- E.g., dichotic listening task –
present different msgs to L & R ears
Let's start with selective attention
in processing auditory signals
Auditory signals were originally preferred because ears don't
move... dichotic listening was the method of choice.
Cherry (1953): One of the first dichotic listening studies
Asked subjects to shadow the message coming to
one ear and ignore the other. Subjects did very well
shadowing the attended ear.
However, they did not realize foreign languages
were included…
Did not realize a word was repeated 35 times...
Cherry also asked what they remembered about the message presented to unattended ear…
People could recall basic physical attributes:
(male versus female voice)
(loud versus quiet)
We commonly refer to
a bottleneck in
information processing.
Many studies have been
dedicated to finding the
locus of this bottleneck.
Central assumption: attention is a limited-capacity system. We cannot process all information at once.
Broadbent (1958) Filter Model
Sensory memory holds info for fraction of second &
passes it to filter
Filter – ids msg that is being attended based on physical chars
- Pitch, tone of voice, speed, accent
- Determines 1 msg that will receive further processing
- ALL other msgs filtered out
After selection, stimuli are shunted along a limited-capacity channel
Detector – process info from attended msg to determine higher level chars
- Processes ALL info that enters it bc only imp/attended info let in
- Output sent to STM…..then maybe LTM
Flow diagram of Broadbent’s filter model of attention.
Selective
Early Selection Model: filtering
before meaning is determined
“Early selection” bc filter operates at an early stage in information flow
Filter model – since all of unattended msgs filter out,
we should not be conscious of info in unattended ear, but….
Moray (1959): The cocktail party effect
Some information (e.g., the subject's name)
“sneaks through” the unattended ear and
is recognized.
A Couple of Problems…
What does this say about filtering?
Treisman’s
Attenuation
Model of
Attention
The unattended
message
must receive
some semantic
analysis. Cannot
be purely
pre-attentive…
Participants told to attend just to one ear but the unattended
message got through too - attention shifted across ears according
to word meaning
Treisman’s Attenuation Model of Attention (leaky filter)
Sensory register
Attenuator – like Broadbent’s filter. Analyzes info in terms of:
physical properties, AND language (syllables/words), AND meaning
All stimuli in sensory memory receive some meaningful analysis (via LTM)
But ONLY as far as necessary to id attended msg
Unattended msgs (weaker) are id’d & sent to ….
Dictionary Unit – contains words stored in memory, each w a threshold for being activated
word w low threshold detected if presented softly, obscured
word w high threshold need strong signal to be detected
Final output determined by DU
In Treisman's theory, pre-attentive analysis
must be almost as complete as the attentive analysis.
So what use is the pre-attentive analysis?
Another theory anyone?
MacKay (1973)
Broadbent's and Treisman's were both early
selection theories -- attention selects a subset
of information in sensory memory, allowing
passage into STM.
Mackay proposed a late selection theory
“late” = info is processed to level of meaning
BEFORE an attended msg is selected
Mackay’s experiment
Ambiguous sentence:
They were throwing stones at the bank.
river bank? or place to save money?
Shadowed the sentence in the attended ear.
Simultaneously, a biasing word
was presented to the unattended ear
river or money
At test picked closest meaning to shadowed sentence:
- They threw stones toward the side of the river
- They threw stones at the savings and loan yesterday
What happened
Mackay cont’d
- Meaning of biasing words (i.e., river, $) affected subj choice of sentence…
- ….even though subjs unaware of hearing biasing words!
- So, meaning of biasing words must have been processed after all
- Conclude: most incoming info is processed to level of meaning before its selected for further processing…….well, according to late selection theorists….
A difference between the early- and late-selection approaches to selective attention is the characteristics of the messages that are used to accomplish selection. Early selection (Broadbent’s approach) is based on physical characteristics. Late selection (Makay’s approach) is based on meaning. Treisman’s attenuation model falls in between these two because selection can be based on physical characteristics, meaning, or both.
State of debate?
- Research looked at:
- Types of info used for selecting a msg to attend to
- Physical chars, meaning
- WHEN selection happens (early, late)
- Debate still NOT solved….
- Results depend on task and type of stimuli used
So….. research shifted to WHAT CONTROLS attention…..
PROCESSING CAPACITY & PERCEPTUAL LOAD
Ability to selectively attend can depend both on distracting stimulus & nature of the task
How do people ignore distractors while trying to focus?
- Processing capacity – amount of info people can handle. LIMITED!
- Perceptual load – related to task difficulty
- Low load tasks – use only small amount of cap. Easy, well-practiced
- High load tasks – use more processing cap. Difficult, not well-practiced
Lavie’s Load Theory of Attention
- Distractors will ONLY slow down processing in LOW-load tasks
- low-load, there is spare capacity, so resources are available to process irrelevant
- high-load, all processing capacity is already being used, no resources are left over to process irrelevant stimuli
- No effect on performance
So, when doing a difficult task,
you are LESS likely to be distracted….
Irrelevant Stimuli
- Ignoring irrelevant stimuli is a function of:
- load (high, low) AND
- How powerful irrelevant stimulus is
- You may be better able to ignore some things….
- E.g., sirens
Another classic example is the Stroop Effect
in color naming
John Ridley Stroop
Part 1: Please read the
following words aloud,
as quickly as possible.
BE LOUD!
GREEN
ORANGE
RED
PURPLE
BLUE
GREEN
ORANGE
RED
PURPLE
BLUE
Again, read the
words aloud,
as quickly
as possible.
Now, please
name
the color
of these objects,
as quickly
as possible.
GRUEEN
ORANGE
MOTRED
PURPLE
BLODUE
Again, please
name
the color
of the printing,
as quickly
as possible.
GREEN
ORANGE
RED
PURPLE
BLUE
pretty funny,
isn’t it?
Why?
- Harder to name colors of words vs. color of shapes
- Words cause a competing response with colors and slow RT d…o…w…n
- Task irrelevant stimuli (color words) are powerful here bc reading words is highly practiced & so automatic that it’s difficult NOT to read them
Data from Stroop (1935)
Get your own
Stroop-effect
t-shirt!
Based on failures of selective attention, Kahneman
(1973) proposed a capacity theory of attention
Everyday activities that divide attention
(e.g., driving a car while talking to a friend)
seem inconsistent with filter theories.
The level of demand that signals/tasks
require seems more important.
Yay! Another theory!
Kahneman
suggests that
attention is
a limited pool
of energy
that we try
to concentrate
or divide
optimally
Kahneman's model
predicts we can
do multiple
tasks if we do
not exceed capacity.
Allocation of capacity
is flexible
and under some
strategic control.
Trying to read while someone talks to you…
Various factors determine whether dual tasks will exceed capacity, such as...
Driving & Car radio…
Task Difficulty
Task Similarity
Amount of cognitive juice needed affects
how well you can multi-task
For support, Kahneman cites a study by…
Posner & Boies (1971) -- dual-task
Subjects did 2 simultaneous tasks:
letter-matching (A-A vs A-B)
Right index finger = same
Right middle finger = different
tone detection
Left index finger = "I heard a tone"
Timing is everything
Switching attention
between tasks - even
easy ones - has a
cognitive “cost”….
for each add’l task you take on, you lose ability to do each one optimally
Think about cell
phones...
# = tone
Time to detect
Tone during trial
ATTENTION AS SELECTION
Overt & Covert Attention
Types of Attention
- Overt – shifting attn from 1 place to another by MOVING the eyes
- E.g., finding Waldo
- Covert – shifting attn from 1 place to another while keeping eyes stationary
Shifting attention
- 2 factors determine how people shift attn by moving eyes:
- Bottom-up processing – based on physical chars of stimulus
- Top-down processing – based on cognitive factors such as observer’s knowledge about scenes, objects
Scanning based on
Stimulus Salience!!
- Bottom up processing depends on properties of a stimulus
- Stimulus salience – physical chars of stimulus
- Color, contrast, movement
- Can influence attn
- Find all the blondes
in the pic
Attentional Capture
- Attentional capture – when attn due to stimulus saliency causes involuntary shift of attn
- e.g., loud noise, bright light, fast movement, potential evil master
Parkhurst (2002)
- Subj viewed saliency maps
- 1st fixations were assoc w highly salient areas
- After that, scanning was influenced by T-D processes (goals, experience)
Do you know what
this is?
Saliency based on Cog Factors
- Where we look isn’t determined ONLY by saliency, but also…
- MEANING
- Large variation in how
people view scenes…..
this is due to individual diffs
in T-D processing
- Scene schemas
- Vo & Henderson (2009) – subj looked longer at things that seem out of place.
- Means attn is being affected by their knowledge
Saliency based on Cog Factors
- Knowledge of scenes can help guide our attention
- E.g., subj more likely to detect stop signs @ intersections (Shinoda et al, 2001)
- Used regularities in the enviro to determine where to look for stop signs…..
- This is yet ANOTHER eg
of top-down
processing
Scanning based on
Task Demands
- Most tasks require attn to different places as task unfolds
- Eye movements determined by task
- Linked to action
COVERT ATTENTION
Directing Attention WITHOUT Eye Movements
Attention to Location
- Posner et al (1978) –
precuing method, covert
- 80% valid precuing
- Subj reacted more quickly when attn was focused where target would appear
- Shows we process info more effectively at the place where our attention is directed
- Attn like a spotlight – improves processing when directed tw a specific location
Attention to Objects
- Same object advantage – when attn is directed to 1 place on an object, attn spreads to other places on that object
DIVIDED ATTENTION
Can we attend to more than one thing at a time?
We can inadvertently pay attn
to 2 things at once – task & distractor
What about intentionally dividing attn?
Divided attn – distribtution of attn among 2+ tasks
Play game & listen to convo
Drive & listen to music, think, talk
Listen to 2 convos at once?
Practice Makes Perfect!
- Schneider & Shiffrin (1977)
- Subj held target stimuli in memory (#s)
- Determine if target was present among distractor (letters)
- Start 55% accurate
- After 900 trials, 90%
- Under consistent
mapping
Practice dedicated to a task reduces the
capacity required by that task.
A distinction of…
Automatic vs. Controlled
Processes
Automatic (or overlearned) processes
require little/no attention; can be carried out
in parallel with other processes
Remember me?
Controlled processes require attention;
are carried out
in serial manner
driving a standard-shift car...
serving a tennis ball...
using chopsticks...
typing...
reading...
We've all had the experience of being lousy at something, then getting better with practice…
Attention and Practice
We’re relatively skilled at some divided attn tasks:
Listening in class & taking notes
Walking & talking
Watching tv & doing crossword
This sounds really, really hard…
A classic experiment….
Spelke, Hirst, & Neisser (1976)
At the same time, they read a book out loud..
Subjects listened to messages AND typed them.
Spelke et al.’s findings:
After 17 weeks, people could perform
both tasks almost perfectly!
I just knew
you could do it!
Is automaticity always learned?
Does it ever come naturally?
Work by Treisman and her colleagues
suggests that some automatic processes
are “built-in” to the human brain…
“pop-out effects” in visual search…
(e.g., Treisman & Gelade, 1980)
Feature Search – Color
Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
And again -- Can you find a blue square?
unique targets are detected very fast
RTs not affected by # of distractor items
“yes” and “no” responses equally fast
indicates parallel search
Feature Search
unique features seem to “pop out” of the display
Conjunction Search
Now can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
Again -- Can you find a blue square?
conjunctions of features are detected rather slowly
RTs increase with # of distractors
“no” responses are slower than “yes” responses (2:1 ratio)
indicates serial search
Conjunction Search
no more “pop out”
Why do pop-out effects occur?
According to Treisman's feature integration
theory, basic features of objects (e.g., color)
are automatically detected.
Conjunction search requires features to be detected
(automatic) and then combined into objects. This requires
attention, and is done serially (controlled).
More about that….
- According to Treisman’s Feature Integration Theory, binding happens in 2 stages when processing image
- Preattentive stage – before we’re conscious of object
- Objects analyzed into separate features – shape, color, movement
Proof?
Treisman & Schmidt (1982) – flashed display below
Illusory conjunctions – combos of features from different stimuli
Showed that features exist independently, like scrabble tiles
Focused attention stage
Attention important in combining features to create perception of whole objects
Follow up – told subj to pay attn to ONLY shapes -eliminated IC
Studies with Balint’s subj –
parietal damage; trouble
focusing atten on
indiv objects
Another example of
Illusory Conjunctions
"Certain aspects of visual processing seem to be
accomplished simultaneously (for the entire field at once)
and automatically (without attention being focused on any
one part of the visual field). Other aspects of visual
processing seem to depend on focused attention and are
done serially, or one at a time, as if a mental spotlight
were being moved from one location to another.“
-- Anne Treisman (1986)
Treisman’s view – The spotlight metaphor
Schneider & Shiffrin(1977)
• Subjects saw a series of 20 rapidly (2.5 sec)
presented frames. Each had 4 stimulus locations
with either letters, digits, or dots. 1 – 4
stimuli could appear on any frame.
• Each subject had a set of targets to remember
and search for (either 1, 2, 3, or 4 targets).
• Frames were shown for variable durations,
from 40 – 800 milliseconds.
However, NOT everything can me made automatic…
sample frame
Consistent mapping
Improvement in
performance with
practice. The arrow
indicates when
the task had
become automatic.
Task: identify a target
from the memory set
if one was presented
The task entailed either
consistent or varied mapping
In CM, targets and distractors came from
different categories (digits vs. letters).
In VM, targets and distractors came from
the same category.
Also, in VM, targets in one trial
might be distractors in another…
Varied mapping condition for Schneider and Shiffrin’s (1977) experiment. This is more difficult than the consistent mapping condition because all the characters are letters and also because a character that was a distractor on one trial (like the T) can become a target on another trial, and a character that was in the memory set on one trial (like the P) can become a distractor on another trial.
Same task
as before
Results?
Under VM, never automatic…..always controlled processing
Distractions while Driving
- Dingus et al (2006) – cams in car & front & rear windows
- 82 crashes, 771 near crashes
- 80% crashes & 67% of near
crashes driver was inattentive 3s before
- Most distracting activity was CELL PHONE
- Redelmeier & Tibshirani (1997) – crash risk 4x higher if driver using cell
- Hands-free offered
NO advantage!
More….
- Strayer & Johnston (2001) – simulated driving task. Press brakes ASAP when see red light (notice what I did there?)
- While talking on phone, missed 2x more red lights
- Increased time to press brakes
- Even for hands-free again!
A little more….
- Nationwide Ins (2008)
- Majority of subj thought they were good drivers while talking on phone
- 45% had been hit/nearly hit by someone talking on phone
- Everyone knows phone use is dangerous….but we don’t think it applies to us!!
- Hanowski et al (2009) –
truckers 23x more likely
to cause crash if texting
Why?
- Talking on phone uses cog resources that would otherwise be used for driving
- Anything that distracts attn can degrade driving
- Phones
- GPS
- Voice activated apps, voice text, email
- Talking/texting is likely using MORE resources than people think
- Sometimes situation requires ALL resources IMMEDIATELY!
WHAT HAPPENS WHEN WE DON’T ATTEND?
So….
Inattentional Blindness
- We can be unaware of clearly
visible stimuli if we aren’t directing attn
- Attention affects perception!
- E.g., gorilla study (Simons & Chabris)
- 46% failed to notice gorilla
- E.g., looking in a
store window
Change
Detection
- Attention is important in
detecting change
- Levin & Simons (1997) – subj saw vids in which some detail changed in EVERY shot
- Only 10% noticed
- Follow-up – showed same subj the same film again AND told them there would
be changes in objects,
body position, or clothing
- Subj id’d fewer than ¼ of changes!
Change Detection
- Classic study by Simons & Levin (1998)
https://www.youtube.com/watch?v=FWSxSQsspiQ
https://www.youtube.com/watch?v=vBPG_OBgTWg
