E N D
1. Task Switching Rogers & Monsell (1995) Monsell (2003)
2. Overarching Cause Cognitive processes require control processes to organize them, but these executive mechanisms are to put it mildly poorly understood. (Rogers & Monsell, 1995, p. 207)
The task-switching paradigm tries to directly tap into these executive mechanisms
3. PRP = doing two tasks on each trial
Task Switching = doing one of two tasks on each trial
Switching tasks causes our responses to be:
Slower
Less Accurate
Is this true even when given sufficient time to prepare for the change in task?
4. Task-Set Definition: the set of processes and responses that are necessary to complete the current task
5. Task-Set Definition: the set of processes and responses that are necessary to complete the current task
Task-sets can be activated by:
the stimulus
e.g. Stroop task GREEN BLUE
= endogenous
6. Task-Set Definition: the set of processes and responses that are necessary to complete the current task
Task-sets can be activated by:
the stimulus
e.g. Stroop task GREEN BLUE
= endogenous
or by intentional / voluntary control
e.g. 100m sprinter
= exogenous
7. Clinical Implications Frontal lobe damage can result in:
Mild once rewarded for responding to a particular stimulus attribute, the patient is unable to stop responding in that way
Severe (utilization behavior) unable to stop complete action patterns triggered by everyday stimuli, e.g. scissors, light-switches, etc.
8. History - Jerslid (1929) Present P w/ a list of items
P repeats same task for all items or alternates on each item
Some alternations had no effect
Add 3 to a #, Write antonym, Add 3, Antonym
Others had large effects
Add 3, Subtract 6, Add 3, Subtract 6
Not much other research until mid 90s
9. Modern Paradigm Train Ps on 2 simple tasks
Then, do a block in which Ps sometimes alternates tasks, sometimes repeats tasks
Task normally determined by:
Alternating runs: n trials of each task
n = 2: A A B B A A B B A A
Task-cuing: cue appears before or w/ stimulus
Switch & non-switch trials in same block
10. Switch non-switch = switch cost
Predictions: long enough R-S interval & predictable task sequence ? no switch cost
15. Method R-S interval: 150, 300, 450, 600, 1200ms
R-S interval random (unblocked) in Expt 2, blocked in Expt 3 & 4.
Practice: 200 trials on each task individually
Experimental: ? 2,000 trials (in 2 sessions)
16. Crosstalk
19. Conclusions Task-set needs to be reconfigured when changing tasks, which takes time.
Only ? 1/3 of the switch cost eliminated by preparation.
Participants inability to use a preparatory interval to get themselves into the state they would have been in had they just performed the same task appears to be a robust phenomenon. (Rogers & Monsell, 2003, p. 223)
20. Explanation Stimulus cue completion hypothesis:
Completion of the reconfiguration is trigerred only by, and must wait upon, the presentation of a task-associated stimulus (Rogers & Monsell, 1995, p. 224)
Endogenous processes can do some of the work, but exogenous processes needed to complete it.
21. Experiment 6 Alternate tasks after 4 repetitions:
A A A A B B B B A A A A B B B B
23. Overall Effects Switch cost / task-repetition benefit
RTs longer on switch trials than on repetition trials (200ms), and error rate higher
Preparation effect
Switch cost reduced if P has advance knowledge of task
Residual cost
Preparation effect levels off at 600ms of prep time without the switch cost ever entirely disappearing
Mixing cost
Even after many repetitions of one task, RT is still slower than in one-task situation
24. Switch Cost Sources Task-set reconfiguration (TSR) = mental gear changing. Can include:
Shifting attention b/w attributes
Retrieving S-R mappings
Adjusting response criteria
Inhibit prior task-set
25. Language Goschke (2000): saying an irrelevant word in preparation interval eliminated preparation effect relative to naming the next task or saying nothing
Others found repeating irrelevant word throughout task hurts more on switch trials
Language effects may disappear w/ practice as representation of task instruction become procedural rather than verbal
26. Discussion Switching between tasks obviously has consistent and robust effects
These costs should be considered in human-machine interface design (e.g. ATC)