Multi-tasking and cognitive capacity Flashcards
multi-tasking and cognitive capacity
Even when we do just one task, limits to cog capacity
All process take time (e.g. memory, retrieval, decision making)
Limits to input any one process (e.g. syntactic parsing) can handle
Representational/storage capacity limited (E.g. WM)
Capacity limits become even more obvious when resources must be shared between tasks, i.e.
Have to get more than one task done in certain time
At least some tasks time-critical (cant wait), so must either (try to) do them simultaneously/switch back and forth between them
examples of real-life multi-tasking
YouTube video
Cooking and ironing and baby-monitoring and phone-answering and door-bell
Other professions
limitations of multi-tasking and cog capacity
Theoretically (what is global computational architecture of mind/brain)?
Practically (efficiency, risk – ‘human error’ as source of accidents)
what do multi-tasking demands include?
When we try to do tasks simultaneously:
Comp for shared resources – dual-task interference
When try to switch between tasks, overhead inc:
Set-shifting/task-switching costs
Retrospective memory
Prospective memory
Monitoring for trigger conditions
Remembering meaning of ‘trigger’
Other demands on executive control
Planning, scheduling, prioritising, coordinating 2 task streams
Trouble-shooting, problem-solving when things go wrong/unexpected conditions arise
So multi-tasking not single competence and executive control processes critical
what happens if you drive and use a phone?
Use of handset while driving now illegal in UK
Use of hands-free phone not illegal per se, but can be prosecuted if use concurrent with accident
Think can drive competently while speaking on phone, esp. if using hands-free
Wrong!
Epidemiological studies show increased accidents (relative risk similar to driving at legal limit for alcohol)
Observational studies show delayed braking etc
Exp studies show impaired braking, detection of potential hazards, etc, esp. in young drivers
Straver, Drews and Crouch (2006) - study on relative risk of cell phone use and drinking
Driver in simulator, follows pacer car in slow lane of interstate for 15 minutes, tries to maintain distance, pacer car brakes occasionally
Baseline v alcohol v casual talk on hand-held/hands-free mobile
Mobile users: Slower reactions, more tail-end collisions, slower recovery
Alcohol: More aggressive driving (closer following, harder breaking)
No sig diffs between effects of talking on hand-held and hands-free
further driving simulator studies - just talking on hands-free phone
Reduced (50%) anticipatory glances to safety critical locations
Reduced (50%) later recognition memory of objects in driving env
Reduced (50%) amplitude of P300 to onset of brake light in car being followed
Increased p(unsafe lane change)
does talking to passenger have same effect as being on phone
Passengers sensitive to driver’s load
Passengers help spot hazards, turn-offs etc
measuring dual-task interference in lab
Just 2 tasks, designed for measurement and manip
Typically measures perf
On tasks A and B alone – should get 100% perf
On A and B combined
Does perf on each task deteriorate when other must also be performed?
Must measure perf on both tasks: P may be able to trade off perf of A and B
what are the sources of dual-task interference?
Slower and less accurate perf in dual-task conditions might be attributable to:
1.Comp for use of specialised domain-specific resources
Parts of body
Brain ‘modules’: Specific processes/representations
- Comp for use of general-purpose processing capacity
‘Central processor?’
Pool of general-purpose processing resource
- Limited cap of executive control mechanisms that set up and manage flow of info through system, and/ sub-optimal control strats
comp for domain-specific resources
2 continuous speech inputs cannot simultaneously be understood/repeated (but both can be monitored for target word meaning)
Perf spatial tracking task interferes with use of visual imagery to remember stuff – both use visuo-spatial WM
Where 2 tasks need to use same perceptual processes/same response mechanisms/same ‘central’ translation/coord processes, should expect dual-task interference (unless info rate low enough to switch use of that resource between tasks)
Doesn’t concurrent perf of any pair of tasks seems to result in some interference (relative to perf of either task by itself?)
comp for general-purpose processor?
Analogous to standard late 20th century digital computer, with CPU
Plausible account of brain architecture?
Others identified consciousness with ‘central processor’ – assumed to be required to pattern recognition, access to memory, decision making, action selection, awareness etc
comp for general-purpose resource pool
Kahneman (1975) proposed pool of general purpose resource/effort shared among concurrent tasks
Capacity of general purpose resource might vary:
Over people
Within people, over states of alertness: Level of ‘sustained attention’ (=available ‘resource/effort’)
Diminishes with boredom/fatigue
Increased with time of day (apart from post-lunch dip), presence of moderate stressors (e.g. noise, heat, but only up to a point), emotional arousal (up to a point), conscious effort
whose capacity is shared by any 2 tasks - central processor/resource pool
Sum of capacity demands doesn’t exceed available total
= No interference
Sum does exceed
= Interference
Increasing difficulty of one task should reduce the capacity available for other task
Hard to know (a priori - relating to or denoting reasoning or knowledge which proceeds from theoretical deduction rather than from observation or experience) how much capacity a given task should use
So, test theory by using pairs of tasks for which it seems obvious that each would require all/most of central capacity
Allport, Antonis and Reynolds (1972) - case of demanding tasks combined w/o interference
Uni of Reading Y3 music students (competent pianists) – tasks:
Sight-read Grade 2 (easier)/ grade 4 (harder) piano pieces
Shadow prose from Austen novel (easier)/text on Old Norse (harder)
Relatively little practice: 10 x 1 min shadowing (to reach criterion of 2 trials with no omissions), 2 x 1 min sight-reading, 7 x 1 min dual task
Exp: two sessions of 2 x 1 min dual task for each combo of easy and hard, and 1 min sight-reading/shadowing alone (order balanced)
Rate of shadowing and number of shadowing errors no diff w/ and w/o concurrent sight-reading
Concurrent shadowing also didn’t increase sight-reading errors
More shadowing errors for harder text, and more sight-reading errors for Grade 4 pieces (difficulty manips effective), but (by session 2):
Shadowing perf not influenced by difficulty of music piece
Sight-reading perf not influenced by difficulty of prose shadowed
further examples of demanding tasks w/o interference
Tasks combined w/o apparent inference:
Schaffer (1975): skilled visual copy-typing can be combined with shadowing of prose w/o interference
One task insensitive to difficulty of other
North (1977): continuous tracking and digit key task
Identify digit with key press
Identify digit before present one
Classify successive pair of digits
No effect of difficulty of digit task on tracking delays
what is the claim of no central general-purpose processor/resource?
Pairs complex input-output translation tasks can be combined w/ little/no interference if they use non-overlapping ‘modules’ (diff input codes/modalities, action systems, central representations/networks)
Idea of general-purpose ‘central processor’ seems unnecessary
Wickens’ (1984) (incomplete) summary diagram of domain-specific cognitive models
But, even when tasks use completely diff modules, some interference may arise due to coord and control demands
i. e. as a consequence of load on (specialised) executive processes
e. g. Allport at al. error in one task briefly slows other
what shouldn’t you do when driving?
Driving/navigation (visuo/spatial input –> hand/foot responses) and
Conversation (speech input meaning speech production)
Require use of diff ‘modules?’
Yes and no – use diff input and output modalities, but both require construction of ‘mental model’ (for driving: representation of route, goals, progress, meaning of road-signs, interp of observed events, prediction of potential hazards etc.)
Particularly obvious how construction of mental model for driving can be interfered with by conversation if ask driver to think about visuo-spatial array/imagine movements
the importance of practice
Tasks which cannot be combined w/o interference become easier to combine with practice
E.g. changing gear while driving
Spelke, Hirst and Neisser (1976): after 85h practice at
Reading storied (for comprehension) at same time as writing to dictation (6 weeks)
Reading storied concurrent with writing category of spoken words (11 weeks)
Some Ps showed little dual-task interference
Because
Practicing one task automates it: reduces need for ‘executive’ control of constituent processes
Practicing combining tasks develops optimal control strats for combining particular task pair
what is Broadbent’s objection to Allport-type exps?
With pairs continuous tasks like shadowing and sight-reading there is:
Some predictability in input (i.e. can anticipate)
Substantial lag between input and output (so must be temporary storage in WM of input and/output)
So, could be central processor switching between 2 tasks (time-sharing) – while processor services one task, input/output for other task could be stored in WM buffers
Test: processor-switching should be revealed if we use concurrent tasks w/ v. small lags between input and output, and next stimulus unpredictable
i.e. reaction time tasks
what is the psych refractory period/PRP?
2 choice reaction-time tasks, stimulus onsets separated by variable, v. short, interval (‘stimulus onset asynchrony’/SOA)
Schumacher et al. (2001)
Task 1 (auditory-vocal): low/medium/high tone
Task 2 (visual-manual): o–/-o-/–o –> index/middle/ring finger
A PRP effect, robust to practice, occurs even when stimuli and responses for 2 tasks in diff modalities
Pashler (1990) – capacity limitation must be in central translation processes
Perceptual processes –> central translation –> response execution
Pashler’s theory
Pashler’s theory: response selection is bottleneck: can be perf for only one task at time
If second stimulus meanwhile arrives and identified, must wait until response selection mechanisms is free
other PRP explanations
RT task pairs where no PRP interference observed:
Repeat spoken letter name and move lever in direction of arrow (Greenwald and Shulman, 1973)
Repeat spoken word and visual same/diff (McLeod and Posner, 1984)
Saccadic eye movement and auditory/manual (Pashler et al., 1993)
All cases with very natural/practiced input-output mapping
When observed, PRP effect could arise not from structural bottleneck, but from cautious control strat Ps adopt to avoid producing response to second stimulus first (Meyer and Kieras, 1997, computational model)
If subject trained to adopt more liberal strat, PRP effect disappears (Schumacher et al., 2001)
Continues to be debated whether dual task costs in PRP paradigm due to structural bottleneck (response selection mechanism)/to ‘soft’ capacity limits attributable to control strats
Even if former, response-selection mechanism not ‘general-purpose processor’
conclusions
Much interference between simultaneously perf tasks can be explained as due to comp for specialised resources
Although idea of general-purpose processing capacity for high-level cognition (=consciousness) has been popular, little ev for it
Cases of dual-task interference previously interpreted as due to comp for limited capacity of ‘general-purpose central-processor’/even generic response-selection mechanism, may be better explained in terms of
Comp for domain-specific resources
And/or
Capacity of (specialised) executive control mechanisms
And/or
Use of inefficient control strats
