Task Switching Higher-Level Cognition Oct 7, 2008
Monsell, 2003 Task Switching Basic phenomena Reaction time (RT) cost for switching tasks Higher error rate Diminished if prepared for switch Slower RT in alternating- than repetitivetask blocks
Review TRENDS in Cognitive Sciences Vol.7 No.3 March 2003 (a) Trial Cue (50, 650, or 1250 ms) Predictable task sequence Random task sequence 8 Stimulus (until response) 6 8 1 3 8 4 2 7 9 1 8 2 Task: odd/even or high/low Cue: color or shape (b) 1000 Errors (%) Mean correct RT (ms) 900 800 700 600 500 6.0 4.0 2.0 0.0 Predictable Random 50 650 1250 1000 900 800 700 600 500 6.0 4.0 2.0 1 2 3 4 0.0 1 2 3 4 Position in run
Changing tasks Switch cost: time used by task-set reconfiguration (TSR) TSR might involve: inhibition of prior task s rules activation of new task s rules Cue: Partial vs. all-or-none TSR (debated)
(a) (b) 900 TRENDS in Cognitive Sciences reparation effect and residual cost. (a) In this experiment (Ref. [13], Exp. 3), the stimulus is a character pair that contains a digit and/or a letter. The tasks were t digit as odd/even, or the letter as consonant/vowel. The task changed predictably every two trials and was also cued consistently by location on the screen (r n subjects). (b) The time available for preparation (response stimulus interval) varied between blocks. Increasing it to, 600 ms reduced switch cost (the effect ), but compared with non-switch trials there was little benefit of any further increase, which illustrates the residual cost of switching. (Data redrawn wi from Ref. [13].) s.trends.com G7 #E Letter task (switch) Letter task (non-switch) 4A Digit task (switch) L9 Digit task (non-switch) Task: cued, classify odd/even or consonant/vowel Mean correct RT (ms) 850 800 750 700 650 600 Switch trial Non-switch trial 0 500 1000 1500 Response stimulus interval (ms) Preparation time
Interference Task-set inertia Specific cases: larger switch cost when switching to stronger (easier) task Counterintuitive; Not explained by TSR Associative retrieval of task set Slower on task 2 with stimuli from task 1
Neuroimaging PFC (medial, lateral) and sometimes other regions: more active on switch than nonswitch trials Task updating, active maintenance Hard to locate executive area, as activity would be seen in executive module and the areas it modulates
Neuroimaging BOLD during preparation for switching: Left inferior frontal junction; pre-sma Activity in response to cue or TSR?
Neuroimaging TMS to pre-sma increases RT on taskswitching trials Executive control module or the target of control?
Further research Many unanswered questions Preparation effect Language interference Task-set intertia
Reynolds et al., 2006 Computational and neural mechanisms of task switching Switch costs present even with preparation Suggests interaction between priming and task updating/active maintenance in PFC
Why do switch costs exist? Though active maintenance in PFC is optimal, only engaged on a subset of trials All-or-none task-switch reconfiguration (TSR) Slow trials: interference from associative learning from previous stimuli Investigate with: Model, fmri Random-cueing task-switching paradigm
Model Connectionist architecture (LEABRA) Associative learning (Hebbian) DA gating signal Presence: Active maintenance Absence: Mechanistic source of task-switch costs
No Decay Prepared DA (AM) No DA (No AM) Decay Unprepared
Results 1150 1050 950 850 750 650 (A) 550 Reaction Time (msec) task repeat task switch Response Repeat bars = beh O = model Response Alternation Greater switch cost on R-R than R-A trials Interference from task switch with same response Weight to PFC from hidden layer (B) 0.6 0.58 0.56 0.54 0.52 0.5 To Current Task Unit To Inappropriate Task Unit Task Repeat Task Switch T-S trials: Smaller difference between unit activation more competition for response, increased RT
Switch-cost difference 1600 Response Time (msec) (A) 1400 1200 1000 800 600 Task Repeat Task Switch bars = beh O = model Fastest Trials Prepared (AM) Slowest Trials Unprepared (No AM) Small vs. Large switch costs between prepared and unprepared trials
Lateral PFC activity (BOLD) 0.6 % Signal Change (B) 0.4 0.2 0 0.2 Fastest Trials Slowest Trials bars = beh O = model Delay Period (Prepared) (Unprepared) Target Period Double dissociation Delay: Fast use AM Target: Unprepared = more processing (TSR)
Rubinstein et al., 2001 Executive Control of Cognitive Processes in Task Switching Exist 2 contrasting theoretical proposals in task switching: Interference from prior task Delay for reconfiguration of rules for new task (TSR) by executive control processes Results of this paper suggest the latter
Existing models Theories of how task switching might be mediated by separable executive control processes for
Existing models: ATA Attention-to-Action model (Norman & Shallice, 1986) Action Schemas (AS): skill routines Contention Scheduling (CS): handles conflicts in AS; quickly assesses priorities w/o exec control Supervisory attentional system (SAS): topdown; supersedes CS to better accommodate goals (slower)
Existing models: FLE Frontal-Lobe Executive model (Duncan, 1986) Goal list: current ranked intentions Mean-ends analysis: like SAS Action structures: procedural knowledge
Existing models: SRD Strategic Response-Deferment model (Meyer & Kieras, late 1990s) Executive-Process Interactive Control (EPIC) Perceptual, cognitive, & motor processes interfaced with working memory Simultaneous-tasks procedure
Review: TS Experiments Jersild (1927) Longer RT for more complex rules in both task alternation and repetition Spector & Biederman (1976) Allport et al. (1994) Visual cues lessen switching-time costs More proactive interference for tasks with greater similarity (sometimes) (TSI)
Review: TS Experiments Rogers & Monsell (1995) Some answers to unexplained Allport et al. (1994) findings Current paper: testing a model of executive control for task switching is warranted
Model Executive Control Processes -» goal shifting T task cuing rule activation stimulus rule task discriminability complexity cuing Stimulus stimulus identification response selection movement production Response Task Processes
Experimental Design Successive-task procedure: Tasks do not temporally overlap Can adjust sequence of stimuli and processes Examine how executive control processes enable task switching 1. Visual pattern classification 2. Arithmetic problem solving
Task manipulations Tasks: Familiar/unfamiliar Rule complexity: Low/high Visual cues (task modality): Present/absent
Experiment 1 Executive control and task processes: empirically dissociated; affected separately by different factors Visual pattern classification (~WCST) Sort by: shape, size, shading, or numerosity B o o o o
Results Discriminability affects RT, not switchingtime costs RTs: alternating blocks > repetitive blocks Increased rule complexity = increased RT in alternating-task blocks
Experiment 2 Executive control mediated by goal shifting and rule activation Arithmetic problems Low complexity: +, - High complexity: x, Cue: present between 2 numbers, or not
10000] H Alternating-Task Blocks Repetitive-Task Blocks 10000i 13 Alternating-Task Blocks Repetitive-Task Blocks in 8000 Reaction Mean Time 6000 4000 2000 low high Rule Complexity 2000 present Task Cuing absent
Experiments 3 & 4 Due to rule activation, switching-time costs may be asymmetric in relation to familiarity of tasks A to B B to A
Experiment 3 Arithmetic: high & low complexity, familiarity 12000i 10000' Alternating-Task Blocks Repetitive-Task Blocks 2000 <n E 1500 H Low Current-Task Familiarity High Current-Task Familiarity <u E 1000 c u 500 2000 low Rule Complexity high low high Rule Complexity
Experiment 4 Pattern classification: low/high complexity, stimulus discrimination 4000- H Alternating-Task Blocks Repetitive-Task Blocks 4000i B Alternating-Task Blocks Repetitive-Task Blocks E 3000 low high Rule Complexity low Stimulus high Discriminability
Overall Results Switching-time costs Increase (additively) with rule complexity Decrease with presence of task cue Asymmetrical time costs for task switch Familiar to Unfamiliar > Unfam to Fam Functional independence of basic processes (e.g., discriminating visual stimuli)
Their Model Accounts for: Behavioral results Counterintuitive Stroop effects from Allport et al. (1994)
Fun Fact