Frequent change can be good for you:
ERP evidence that task switch probability affects cognitive control
Michael Bersick (1), Doreen Nessler (1), Ray Johnson, Jr. (2), and David Friedman (1)
Mixing Costs
Baseline for cuerelated ERPs
1a
300
250
250
200
200
150
150
100
100
50
50
0
0
Pz
cue onset
500
4
?
4
1000 ms
more/less than 5?
odd/even?
1b
[1] Monsell & Mizon (2006). JEP: HPP (32), 493-516.
[2] Wylie & Allport (2000). Psychological Research (63), 212-233.
[3] Schneider & Logan (2006). JEP: General (135), 623-640.
MFN
Pz
5a
Cue-related P3
500 1000 ms
-1000 -500 0
Cz
7a
500 1000 ms
Fz
8
7b
0 500 1000 ms
RT 200 400 ms
Informative cue, unpredictable
Informative cue, predictable
Uninformative cue, unpredictable
Uninformative cue, predictable
Cz
6b
target onset
Fig. 4: (b) Informative cues preceding
equiprobable, unpredictable switches
elicited a parietal, P3-like effect relative to
pre-switch trials. (a) This effect was
absent
for
informative
cues
on
equiprobable, predictable switch trials.
Fig. 5: (a) and (b) Informative cues
preceding
both
predictable
and
unpredictable rare switches elicited much
larger amplitude P3s than those present in
the equiprobable condition (cf. Fig 4).
Cz
6a
Cz
Pz
5b
4b
Fig. 6: After uninformative
cues, the decrease in the
amplitude of the target P3
was larger for unpredictable
(b) than for predictable (a)
pre-switch and switch trials.
Fig. 7: (a) and (b): Target
P3s to rare switch and preswitch stimuli.
Fig. 8: After uninformative
cues, response conflict, as
indicated by the amplitude of
the MFN, was larger for
unpredictable than predictable, equiprobable switches.
Target P3
Rare switch
(b) Unpredictable
Rare Switch
Uninformative cue
Equiprobable
4a
Pz
Predictable, informative cue
Unpredictable, informative cue
Predictable, uninformative cue
Unpredictable, uninformative cue
Fig. 3: Informative cues reduced switch costs.
In addition, when switches were equiprobable
switch costs after uninformative cues were
significantly reduced.
Single task
Pre-switch
Switch
(a) Predictable
Switch Costs
Equiprobable
Rare Switch
Fig 2: Informative cues reduced mixing
costs. Mixing costs decreased dramatically
after uninformative cues in the equiprobable
condition, but only when task order was
predictable.
Informative cue
-1000 -500 0
Informative
References
300
Equiprobable
Target onset
300 800 1300
0
Cue onset
Baseline for
target-related ERPs
Uninformative
350
0 500 1000 ms
target onset
Discussion
EEG recording:
62 sintered Ag/AgCl electrodes; referred to averaged mastoids; continuous DC100Hz; 500 Hz sampling rate; Fig. 1a: ERP epochs.
350
Rare switch
Participants:
Fifteen young adults (19 to 30 years; M = 24.6; SD = 3.8).
Design:
Digit (not 5) required response: more/less than 5? or odd/even?
pure blocks: one task; mixed blocks: two tasks.
Factors:
(1) Ratio of switch (S) to stay or repeat (R) trials
equiprobable; switch after 0, 1, or 2 trials: a-bb-aaa
rare switch 1:3; switch after 2, 3, or 4 trials: aaa-bbbb-aaaaa
(2) Cue-status: informative, uninformative (Fig. 1b);
(3) Predictability Status: predictable, unpredictable
Mixing costs = pre-switch RTs in mixed blocks - RTs in pure blocks
Switch costs = switch RTs – pre-switch RTs mixed blocks
Results
The cost of switching between two tasks is reduced by
providing a cue warning of an impending switch. Cognitive
control accounts of this effect claim that an informative cue
allows for task-set reconfiguration prior to the actual switch
[1]. Nevertheless, the fact that switch costs remain even after
a long preparatory interval suggests that different task sets
can interfere with each other [2]. However, such proactive
interference might only be a problem when successive task
sets must be reactivated. If switching is frequent, then
keeping both task-sets active is a good strategy, assuming
that little cost accrues from doing so. We examined how
switch probability, switch predictability, and informative-cue
availability interact to change the ease with which
participants can maintain and access different task sets.
Equiprobable
Methods
Introduction
1 Cognitive Electrophysiology Laboratory, NY State Psychiatric Institute, NY, NY;2 Department of Psychology , Queens College/CUNY, Flushing,
NY
Participants seem to utilize predictive information to actively maintain both task sets
when doing so aids performance. Decreased mixing and switch costs for predictable,
equiprobable stimuli, the decrease of cue-related P3 amplitude from rare to
equiprobable switches, and the absence of any P3 to informative cues for equiprobable,
predictable switches all suggest that cue information is increasingly irrelevant to
performing the task as switches become more probable and predictable. Although
response conflict was highest after uninformative cues, smaller target P3s for predictable
vs. unpredictable equiprobable switches were observed, suggesting that predictability
allowed processing resources to be conserved. Participants may also have created a more
global behavioral set encompassing both tasks [3]. Switch costs in the equiprobable
condition might be best described as relatively “pure” costs of switching between active
task sets rather than as proactive interference.