Week 7 Flashcards
Limitations in visual attention under time pressure :
RSVP
• RSVP = rapid serial visual presentation
• Only one location, but time-pressured
• Letters, digits, words, etc visually displayed, in a single
location, one after the other at a rapid rate.
• Typically about 100 ms per item
• Ps may be asked to look out for certain targets; asked at
the end of list about items (around 15 items).
• Conceptual processing even at this fast rate of
presentation
• Post-target intrusions common
Q X Y Ps report that the X was blue
Two-target RSVP - The attentional blink (AB)
Ps ignore the digital distractors and report the letters.
The Attentional Blink - 2 targets
• If Ps must detect 2 targets in the RSVP stream,
there is a decrement in reporting T2 when it
occurs a few hundred ms after T1
• Blink extends out to about T1 + 6
• Blink is not typically observed for the T1 + 1 item
(Lag-1-sparing)
• T1 & T2 processed as one event when T2 occurs
immediately after T1?
Attentional blink and the ‘lag effect’
The time taken with the probability of getting T2 correct, given T1 correct
The Attentional Blink
• AB is found if T1 and T2 defined in the same way ( 2 digits) or differently (red letter vs. digit)
• Making T1 easier to identify/report reduces the AB
• AB reflects demands of selecting & identifying T1
• But: AB occurs if Ps have to merely detect T1 without
reporting its identity
• AB not just a recall problem: Interference is observed when a recognition test of targets is used.
• AB is strongly influenced by whether or not items precede and follow T1 and T2
• or T2 is extremely brief
• Adjacent items serve as pattern masks that curtail processing of the targets
• Pattern masks compete with targets to engage perceptual processing
• They are commonly used to ensure that brief stimuli do not reach awareness
• Ps say they didn’t see the masked word, but it may be processed sufficiently
to affect responses in tasks
The AB - putting it all together
• Multiple sources of the AB? No account explains all results (Dux & Marois, 2009)
• Evidence for “resource depletion” – capacity/structural limitations on the
number of targets identified
• Effects of T1 difficulty (more difficult à bigger AB)
• Attentional control mechanisms for selecting targets and rejecting
distractors have a role
• Discriminability of targets and distractors important
• Debate continues about role of distractors in the AB
Attentional limitations under time pressure
• Effects under time pressure = stress testing the system to find
its limits (what aspects of the task are challenging under time pressure?)
• The AB evidence suggests that consolidation of targets in WM is an
operation that can’t be done for more than one target (or target chunk) at a time
• And it may be associated with competitive or inhibitory effects for other stimuli
From RSVP to task switching
• The AB (as an example) is thought to show a “hard limit” on human
cognitive capacities
• But we have seen that issues of task control are important
• how people stop responding to RSVP distractors
• how Ps switch between demands to remember targets and ignore
distractors
• Many researchers interested in task control in its own right – the
task switching paradigm
• Domain called attentional control (also EXECUTIVE control)
Task switching
• Ps incur a time (and perhaps error) cost when they switch from
one simple task to another
• Switching is an executive control operation
• Contributes to dual task performance (e.g., Attentional Blink)
• Task set: Preparation to perform one task rather than another
• Involves selecting, linking, enabling “modules” for task
components (e.g., perception, response selection).
• Links to irrelevant modules must be disabled.
• Switch costs arise in establishing the appropriate task set &
disengaging an inappropriate set
Task-switching costs are not a new thing…
• The basis for Henry Ford’s Highland Park, Michigan plant: • Model T Ford - durable and affordable • Moving assembly line • Interchangeable parts • Workers at specific locations in the line • Still the approach used today
Task switching paradigm
• Arthur T. Jersild (1927): block of trials with only a repeated
task, block of trials where two tasks are alternated, latter
takes longer.
• Typical procedure involves AABB, AAABBB, and so on.
• Task practice, fatigue, are matched over switch trials and
same-task trials
• Usually simple responses to digits or letters
• Observe a severe “switch cost” of up to several hundred ms
when the task changes
• Even though the task changes are regular and thus
predictable
The task switching phenomenon
• Same vs. different stimuli (Jersild, 1927, & later research):
• Small switch costs if both stimuli and responses different –
e.g., number task alternating with word task
• Major cost incurred when Ps change the task and thus the
response selection rule for a common stimulus configuration
• Trivial accounts of switch costs:
• Do Ps occasionally forget what task is next?
• No. Don’t see evidence of occasional slow trials
• Instead a general slowing over the RT distribution (Fagot
1994)
Basic findings in task switching
• Rogers & Monsell, 1995, AABB paradigm:
• Stimuli – digit-letter pair, e.g., G7, B2
• A Digit task – right button for odd digit, left for even
• B Letter task – right button for vowel, left for
consonant
• Task cue: Stimulus appears in one of 4 boxes. Perform
letter task for top boxes, number task for lower boxes.
Rogers and Monsell paradigm
Letter task
Vowel (R) or consonant (L)?
Digit task
Odd (R) or even (L)?
Task alternation: Letter Letter Digit Digit Letter Letter Digit Digit
Switch trial? N Y N Y N Y N
Switch trials can be predicted; sequence is regular.
Record accuracy and latency (Reaction Time) for responses for switch vs. non-switch
trials
Classic results of Rogers & Monsell
Mean RTs for 2 days, about 880 trials per day
• Large improvement on day 2, especially switch trials
• Large switch cost (even though task switches were predictable)
• Similar performance for letter and digit tasks (similar difficulty)
Practice and task difficulty effects
• Practice
• Costs are reduced but not eliminated with practice of separate tasks &
switching tasks
• Task difficulty:
• Switching TO the easier task incurs the greater cost
• e.g., from colour naming to word naming in Stroop (Allport et al, 1994).
BLUE RED GREEN YELLOW red yellow green yellow
TASK SWITCH
• What does this result suggest about the roles of relevant task
engagement vs. irrelevant task disengagement in switch costs?
Easier to switch to the more difficult task
• Suggests that DISENGAGING from the prior
task is a major factor!
• The difficult task requires effort and
concentration, so it’s difficult to disengage
from a difficult task to engage in another task
Causes of task switching effects?
• Rogers & Monsell, 1995. Emphasis on Task Set Re-configuration
(TSR) prior to the switched task
• Can maintain 2 task sets for different tasks & stimuli.
• But with different responses for the same stimuli, need to change task
set when task changes
• Supported by manipulations of preparation time:
Task-cuing paradigm (Sudevan & Taylor, 1987).
• Odd-even judgments (task 1) and greater than/less than 5 judgments (task 2)
• P is informed by a cue before the trial which task to perform.
• Cue to digit interval varied (400 to 4000 ms).
• Switch cost decreased as interval increased to 2-3 sec.
Rogers & Monsell- switching
• Large switch cost (even though task switches were predictable)
• Cuing reduces task-switch costs but does not
eliminate them
Causes of task switching effects
• But is it preparation time or delay from the last trial that
reduces the switch cost?
• These factors were confounded in initial studies
• Meiran (1996): Varied cue-to-next-stimulus interval and last-responseto-cue interval.
• A short cue-to-stimulus interval led to a large switch cost even if there
was a long delay from the last trial.
• Implicates active preparation, not just decay of effects of prior
trial
• Involving both disengaging the past task and engaging the coming task
Preparation vs. interference from last trial
• Can adequate preparation remove the switch cost?
• No, there is always a residual cost
• So task preparation is not the whole story
• Residual costs are exogenous effects - “stimulus driven”
• Can’t do any decision-making for the next trial until the
stimulus has been presented and identified
• Deciding on the response for the current stimulus
may be affected by the response associated with the
stimulus for the alternative task
Theories of the switch cost
Theories differ in:
1. The role of active preparation - endogenous
factor
2. The role of interference effects from prior task
and task set that dissipate passively
3. Whether exogenous factors play a role
Summary of task switching theories
- Disengagement theory: Proactive interference from “task set inertia” (TSI).
Allport et al. 1994.
• TS decays after several minutes, impairs adoption of alternative task set.
• But evidence supports role of active preparation - Endogenous + exogenous factors: Rogers & Monsell.
• Endogenous component (preparation by P) requires time: Switch cost reduced as
preparation time increases.
• The residual cost that cannot be removed by adequate preparation is due to an
exogenous component of task set that must be triggered by the arrival of a
suitable stimulus. - Endogenous only theory: de Jong (2000).
• The residual cost arises because Ps do not prepare adequately on every trial.
Effects of practice on control: Automaticity
• Kahneman’s capacity theory: Over-learned tasks
become automatic & consume fewer resources.
• Substantial practice:
• Improves performance
• Reduces task effort
• Facilitates re-structuring and co-ordination of concurrent
tasks
Automaticity every day & in the lab
• With practice many tasks can be performed with little effort
and concurrently with certain other tasks
• Driving a car
• Riding a bicycle
• Classic demonstration in the lab – KEY RESULT:
Shiffrin & Schneider (1977)
• Ps given a memory set - targets to be searched for in the
upcoming trial
• For example, letters ‘G’ & ‘M’
• A trial was a rapid series of 20 square displays
• Fixation dot at centre
• Then each display had 4 letters around the centre
• On each trial there was one or zero targets, the
remainder distractors
Shiffrin & Schneider paradigm
• Ps had to press a key as soon as they saw a target, or press another key
for “no target” at end of trial
• S & S varied the display (“frame”) duration (40 ms +), number of items on each frame,
nature of display items (e.g., letters vs. non-letters), number of items in the memory set.
• Key independent variables for automaticity:
1. Size of memory set = 2 vs. 4 letters
2. Relationship between memory set and display items throughout trial
block
• Categorical (“consistent mapping”): targets came from one set
• Mixed (“varied mapping”): targets and distractors came from same set
• So in mixed condition a letter, say B, could be a target on one trial and a
distractor on another trial
• but in categorical condition B could only be a target OR a distractor
(not both) throughout the block
Shiffrin & Schneider key results
In the categorical condition, Ps learn the targets and respond as soon as they see one.
• After practice, there is no effect of memory set size (2 vs. 4) - evidence of automaticity.
• In the mixed condition, it is not possible to use a simple response rule (see a B, respond
target).
• There is always a cost of having a larger memory set, even after nearly 10,000 trials!
Characterising automatic processes
- Without awareness
- Without conscious deliberation / obligatory
- Without expenditure of resources
- Fast
- Rigid/habitual
• (Shiffrin & Schneider found that Ps had great difficulty in categorical
condition when the assignment of items was reversed.)
• E.g.:
• automatically reading a word that is looked at (Stroop task)
Theories of automaticity
• Logan (1988)
• Automaticity based on knowledge acquisition, is not all-or-none
• Separate memory traces for each encounter with a stimulus.
• Practice leads to storage of information about the stimulus and
how to respond to it.
• Rapid retrieval of relevant information about stimulus
• Automaticity is memory retrieval – a single step direct-access
retrieval of a past solution.
• In the absence of practice, thought and application of rules is
required.
Cautions about automaticity
Problems with traditional criteria for automaticity:
• Many “automatic” processes do not meet all of the criteria
for automaticity
• Capacity & interference:
• Even well-practised tasks, that can be performed without
conscious awareness, can still be impacted by task load.
Awareness and automaticity
• Can have lack of awareness in intentional tasks
• e.g., using cue for retrieval from memory
• awareness of cue but not retrieval processes
• Can also have awareness of many aspects of tasks that are highly
routinised
• Mistakenly taking a routine turn at an intersection does not mean that
the action is “unconscious”
• Rather, the link with intention is lost
• Level of control critical:
• Automatic performance can be reduced with additional demands
• Automaticity depends on the situation.
Working memory
[also called short-term or primary memory]
Attention meets memory • What is attended to goes to Working Memory (WM)….and vice versa! • WM is the domain of conscious thought • WM involved in making decisions and initiating actions based on plans and in response to environmental input • WM involved in directing attention
Working Memory - Alan Baddeley
Working Memory (WM)
• Material disappears after a few seconds if not refreshed
• Limited capacity (7…ish):
• Information is displaced by new information.
• In neuropsychology WM is distinguished from STM
Compared to… Long-term memory (LTM)?
• More permanent traces, vast capacity
• Forgetting through interference; decay also?
• Continual interplay between WM & LTM in perceiving, speaking & action.
Baddeley’s WM mode
Central Executive
-Phonological loop
-Visuo-spatial sketchpad
-Episodic Buffer
• CE co-ordinates activity of “subordinate” systems that store information
• We will discuss the CE, Phonological loop and the V-S sketchpad
• Episodic buffer - recent (and contentious) addition to model; links
information across visual, verbal, and spatial domains
The central executive
• CE: An attention controller that is an interface between
WM systems and long term memory (LTM)
• does not have its own storage capacity
• Functions (there may be others!)
• Co-ordination of the subsidiary WM systems
• Control of encoding & retrieval strategies
• Switching of attention
• Mental manipulation of material held in the slave systems
Investigating the central executive
• Random number/letter generation thought to require the
CE
• Under pressure of concurrent CE demands sequences become
less random.
• Baddeley & colleagues have shown interference between
random number/ letter generation and:
• Playing chess, reasoning
• Problem solving
• Generating items from semantic categories
• Mental arithmetic
The phonological loop
• Maintains verbal, sequential information in a phonological (soundbased) code
• 2 components
• Verbal store (“inner ear”)
• Subvocal articulatory rehearsal process (“inner voice”)
• Information decays after about 2 sec, unless maintained by rehearsal.
• Articulatory rehearsal also may be used to enter information into the
store.
• Examples of tasks:
• Remember password, phone number, recipe, instructions, in the short term
• Digit span
Phonological loop: 4 key effects
- A: Phonological similarity effect
- B: Irrelevant (unattended) speech effect
- C: Word length effect
- D: Concurrent articulation effect
- Also called “articulatory suppression” has
Evidence A: Phonological similarity effect
• Conrad, 1964. Most confusions in immediate serial (ordered)
recall for letters with similar sounding names
• Conrad & Hull: more errors in serial recall for B, G, V, P, T than Y, H, W, K, R
• Baddeley, 1966: Serial recall of phonologically similar vs. dissimilar 5-item
sequences presented auditorily.
• mad, man, mat, cap, cad, can, cat, cap (similar pool)
• cow, day, bar, few, hot, pen, sup, pit (control pool)
• % sequences correct: Control = 82%, Similar = 9.6%
• Similar results for visual presentation
• Recalling order information is the problem
• Sound not spelling - effect found with caught, sort, taut, etc
Interpretation
• Confusions among phonologically similar items indicate use of
phonological representations.
• Absence of phonological similarity effect suggests that Ps have
abandoned a phonological coding strategy
• Distinguish from long term memory effects
• LTM usually shows semantic not phonological confusions.
• e.g., confuse movies, books with similar themes
Evidence B: Irrelevant speech effect
• Sometimes called “unattended speech effect”
• Speech impairs serial verbal recall of visually presented material.
• Nonwords, Arabic and backward speech interfere: Not
dependent on meaning of material.
• Music sometimes interferes
• White noise has no effect
• Intensity of auditory stimulus not important
• Baddeley’s interpretation:
• Obligatory access of speech-like input to phonological store
• Corruption of trace, added noise, rather than phonological confusion effect
• Can get impairment for materials that don’t produce a phonological
similarity effect
Evidence C: Word length effect
• Immediate memory span declines with the spoken duration of the list items.
• Baddeley, Thompson & Buchanan, 1975.
• Auditory lists, 4 - 8 items, items listed for Ps to refer to
• Short: sum, hate, harm, wit, bond, yield, etc.
• Long: association, opportunity, representative, organization, individual, etc.
• Sequences correct: Short words Long words
56% 20%
• Controlled number of syllables - coerce, typhoon, voodoo, etc. vs. ember,
wiggle, phallic
• Lower accuracy for items with longer-duration pronunciations
Interpreting the word length effect
• Reflects the speed of subvocal rehearsal and hence the
rate of refresh of the memory trace.
• Less time to rehearse the word
• Rehearsal involves central programming of speech but
not output
• patients without vocalisation show rehearsal
• but not those with impaired speech programming.
Word length effect: Alternative views
• Cowan: delays at output associated with the longer
articulation time of long items
• More items forgotten in the time taken to say
“representative” than in the time taken to say “dog”
• But studies show word length effects even with probed
recall.
• Was the item “representative” in position 3?
Evidence D: Concurrent articulation
• Also called articulatory suppression:
• Repeating la-la, the, hiya etc, adversely affects serial recall & abolishes the word
length effect.
• Phonological similarity effect removed by articulation with visual but not
auditory presentation
• Interpretation:
• Eliminates subvocal rehearsal
• Impairs phonological recoding of visual material
• What components of articulation are important? Role of irrelevant speech?
• hearing speech is not a major factor
• silent articulation interferes
• but not non-speech actions (chewing, etc)
What is the loop for?
• Comprehension of oral and written language? Only
as back up.
• Brain injured patients with severely impaired verbal
serial recall can understand and read most sentences
• Vocabulary learning – children, learning a new
language
The visuo-spatial sketchpad
• Visuo-spatial rather than verbal encoding of material
• Tasks: Corsi tapping task (Neuropsych), memory for un-nameable
shapes/patterns, navigation and tracking
• Concurrent speech does not interfere with V-S tasks
• Brandimonte & Hitch, 1992
• speech impairs verbal encoding of shapes (e.g., “umbrella”)
• but does not affect visuo-spatial encoding of shapes
• Baddeley suggests that V-S memory is based on a visuo-spatial code that
supports imagery
Nature of V-S sketchpad
• Visual vs. spatial representations
• Consistent with imagery research, Baddeley recently
has divided V-S sketchpad into:
• Visual cache stores visual patterns
• Inner-scribe – spatially based rehearsal (e.g., of
movement sequences)
• Doesn’t require visual input
• Spatial rehearsal performed by blind Ps
Function of Visuo-Spatial Sketchpad
- Planning and execution of spatial tasks
- e.g., in sport, driving, etc.
- Manipulating visual images
- Keeping track of changes in the visual perceptual world.
- Maintaining orientations in space and directing movement.
- Comprehending certain verbal information (in navigation etc.)
The Baddeley model: Conclusions
• The model describes and organises some key facts about human
short term memory
• Limitations of memory
• Coding and modality effects
• Insufficient development of how the stores interact and the
interplay of WM and long-term memory
• The Baddeley model may not be important in 50 years time
• But the evidence gathered by Baddeley and the fruits of
research stimulated by the model will always be important!
Visual working/short-term memory
• Memory buffer that allows the retention of visual information, for
a short time period, when it is out of view. Holds 3-4 items*
• Gives rise to a coherent and continuous representation of the
visual world, that would otherwise be disrupted by objects being
occluded or saccadic eye movements etc
• Allows relevant visual information to be used when undertaking
goal-directed behavior and thoughts
• What gives rise to the capacity limits of working memory?
• Slot model (Luck and Vogel, 1997): Fixed number of object
representations that can be held in memory at one time; once at this limit,
no other items can be held in memory
• Resource model (Alvarez & Cavangh, 2004): Limited supply of a
representational medium is continuously distributed between objects; items
that receive more resource are stored with less noise.
• Complexity of objects influencing VSTM capacity favours resource model
• Still under debate.
Neural Substrates of VSTM/VWM
• Specific vs distributed?
• VWM load detected across the brain
(Naughtin et al, 2016, Cerebral Cortex)
