Advance preparation and stimulus-induced interference in cued task switching: further insights from BOLD fMRI
Introduction
Task switching is commonly held to be an appropriate paradigm for the exploration of executive control (e.g. Monsell, 2003a). Controlled information processing enables behavior that goes beyond inflexible stimulus-driven S–R association. Successful behavior in task switching would fail if subjects based their performance on fixed S–R mappings while ignoring task-specific contextual constraints on action. In compliance with ‘contextual constraints’, subjects are able to mentally structure the configuration of potentially available objects-for-action and potential response options (i.e. a task set is formed) in accordance with an internally represented goal.
Studies on brain-damaged patients (Milner, 1963; Stuss & Benson, 1986), non-human primates (Fuster, 1980, Goldman-Rakic, 1987, Miller and Cohen, 2001; Milner, 1963, Stuss and Benson, 1986), and brain-imaging studies (Brass & von Cramon, 2002; Nagahama et al., 1999, Passingham et al., 2000; Seitz & Binkofski, 2000; Toni, Rushworth, & Passingham, 2001) suggest an important role of the prefrontal cortex (PFC) subserving this kind of goal-directed behavior. Furthermore, studies that measure PFC and parietal cortex simultaneously usually show that the PFC acts in concert with the parietal cortex in order to implement control (Chafee & Goldman-Rakic, 1998; Dove, Pollmann, Schubert, Wiggins, & von Cramon, 2000; Sohn, Ursu, Anderson, Stenger, & Carter, 2000).
The task switching paradigm captures two core features of goal-directed behavior: flexibility and anticipatory control. Flexibility is realized by introducing frequent changes of the relevant goal (i.e. the task to be performed), which is operationalized by the independent variable task transition (task switch versus task repeat). Anticipatory control comes into play when the upcoming task can be prepared in advance, which is operationalized by the independent variable preparation interval with either a short interval (no advance preparation) or a long interval (advance preparation).
With this very basic design the present functional imaging study aimed at investigating the nature of advance task preparation and its relevance for flexibly switching between alternative tasks. This issue is central for understanding the basic cognitive mechanisms underlying task switching performance and has been causing severe controversy among theorists (Altmann, 2003, Monsell, 2003a, Monsell, 2003b).
One basic empirical finding is that switching from one task to another task as compared to repeating the same task impairs behavioral performance. These ‘switch costs’ are supposed to reflect the need for a stronger engagement of control to enable a task switch. Or, in other words, switch costs reflect elevated control being necessary to counteract the tendency to repeat the previously performed task. The two scenarios outlined below make different assumptions about how this perseverative tendency is mediated.
Another important finding is that switch costs are often reduced with prolonged preparation intervals (e.g. Meiran, 1996; Rogers & Monsell, 1995). This observation suggests that a task switch can be prepared in advance. Two different explanations for this switch cost reduction are being contrasted in the present study.
According to the first scenario, the system tends to perseverate because the previously adopted task set is persisting over time into the next trial. Thus, establishing the competing task set in a current switch trial requires additional time-consuming control effort because proactive interference from the persistently activated, now misleading task set has to be overcome. With sufficient preparation time, this same process can be finished in advance of target presentation. As proactive interference has already been overcome during the preparation interval, it is no longer slowing down appropriate task implementation after the target has been presented. This notion to some degree resembles the concept of ‘advance task set re-configuration’ (Meiran, 1996; Rogers & Monsell, 1995).
According to the second scenario, a previously adopted task set is dissipating rapidly before the next trial is presented. Thus, it is not the persistently activated previous task set that causes interference in a current switch trial. Alternatively, as recent studies are suggesting, interference might be induced by the target stimulus itself which is retrieving the previous task set from memory (Allport & Wylie, 2000; Waszak et al., 2003, Waszak et al., in press; Wylie & Allport, 2000). However, when every new trial starts with a neutral task set because interference is induced only after the target has been presented, there is nothing which can be done during preparation but biasing the initially neutral task set in the direction of the currently instructed task set—and this is equal for both switch and repeat trials. It is therefore not immediately clear why advance task preparation being equally engaged for both trial types should have a benefit that is differently stronger for switch trials compared to repeat trials as being indicted by reduced switch costs. A solution for this paradox is that a task switch can benefit differentially from advance preparation because the target-associated previous task set loses its potential to gain a misleading influence during task implementation. This is because the instructed task set has time to settle into a stable state, thus becoming resistant against later disruption (see also Koch & Allport, submitted for publication). A discussion of related notions can be found in other recent publications (Gilbert & Shallice, 2002; Goschke, 2000; Wylie, Javitt, & Foxe, 2003; Yeung & Monsell, 2003).
By measuring the subjects’ behavioral performance, the involvement of a task preparation process can be inferred only indirectly from the beneficial impact it has during the subsequent task implementation. Functional MRI can be used to obtain a more direct record of the ongoing preparation process by measuring the correlated blood-oxygenation-level-dependent (BOLD) activation. In the present study we were measuring BOLD activation to distinguish between the two theoretical scenarios sketched above, which are both equally compatible with the reduction of behavioral switch costs.
We realized an explicitly cued task switching procedure (i.e. an unpredictable task cue indicated the current task) and introduced a long preparation interval of 2000 ms (CTI-2000) and a short preparation interval of 100 ms (CTI-100). An accumulating number of previous fMRI studies using cued task switching procedures did not find elevated BOLD activation for switch trials compared to repeat trials with long preparation intervals (Brass & von Cramon, 2002; Braver, Reynolds, & Donaldson, 2003; Dove, 2000; Luks, Simpson, Feiwell, & Miller, 2002).
This result intuitively appears to be incompatible with the notion that establishing the instructed task set in switch trials is facing interference from the persistently activated, now misleading previous task set. The advance resolution of this interference should require more control effort during preparation than merely refreshing the task set in repeat trials. This should be paralleled by stronger BOLD activation for switch compared to repeat trials.
We suggest to explain this rather unexpected result in terms of the alternative scenario outlined above, which accounts for reduced behavioral switch costs without assuming that between-task interference is being resolved during preparation. This account predicts reduced or even absent additional switch-related control effort when advance preparation is possible, both during task preparation and during task implementation. Hence, being easily compatible with the absence of enhanced BOLD activation at long preparation intervals.
Furthermore, this account predicts that interference caused by the target-induced N − 1 task set impairs task implementation specifically when advance preparation is not possible. Thus, high switch-related control demands with a short preparation interval should be reflected by enhanced activation in switch trials compared to repeat trials. This is exactly the pattern Dove et al. (2000) observed for several frontal and parietal brain areas realizing a CTI of 0 s.
Different from previous fMRI studies we realized a short and a long CTI condition within the same subjects which allows to draw stronger conclusions regarding the comparison of switch-related BOLD effects for different CTIs. Moreover, as being delineated below, a fine-grained analysis of the temporal structure of the trial-related BOLD response was intended to bring about further theoretical constraints.
As argued above, determining the pattern of switch-related BOLD activation for CTI-100 and CTI-2000 can reveal important information. However, it would be even more informative to know, whether brain areas that are engaged in the CTI-2000 condition are involved in cue-related and/or target-related processing. Unfortunately, the analysis of fMRI time courses notoriously faces problems of decomposing a trial-related BOLD response into sub-components associated with separate within-trial events (cue and target) when the events are not spaced generously or the event order is not counterbalanced. This problem also holds for the standard method based on multiple linear regression (Friston et al., 1998) which we applied for detecting relevant activations within the whole-brain volume. Any effect we observe for CTI-2000 thus always reflects the sum of effects caused by cue-related and target-related processing.
To gain at least partial information about the composition of the trial-related BOLD response at CTI-2000 we implemented a novel temporal analysis of trial-averaged time courses (Ruge, Brass, Lohmann, & von Cramon, 2003). This method allows to decide whether a brain area of interest is generally activated cue-related and/or target-related in terms of present–absent judgements.
However, if a brain area turns out to be engaged both in the cue phase as well as the target phase, even for this method, the 2000 ms CTI is too short to determine the relative quantitative contributions of the respective cue-related and target-related BOLD sub-components. In this case, the analysis can nevertheless tell that an activation of unspecified strength is associated with both the cue and additionally with the target.
Similarly, potential activation differences between switch and repeat trials at CTI-2000 can not differentially be assigned either to the cue period or to the target period. With regard to the hypothesis that the CTI-2000 condition would not differ between switch and repeat trials, this does not pose a major limitation. Furthermore, it should be noted that, though BOLD activation at CTI-2000 displays the sum of potential cue-related and target-related sub-components, activation effects that are associated with only one sub-component are still detected by both the standard regression-based whole-brain analysis as well as our additional temporal analysis.
Section snippets
Subjects
We measured 22 subjects who all gave written informed consent to participate in the present study. Four subjects were excluded due to movement artifacts, all during the second experimental block (see below). The mean age of the remaining 18 subjects was 25.5 (range 21–35), 10 were female. No subject had a history of neurological disorder, major medical disorder, or psychiatric disorder. All subjects were right-handed as assessed by the Edinburgh Inventory (Oldfield, 1971).
Experimental procedure
We adopted a spatial
Behavioral data
As mentioned in the methods section the scanning session consisted of two different experimental blocks of which only the block with ‘bivalent responses’ is of interest in the present paper. The sequence of blocks was balanced across subjects implicating that the bivalent response block was either the first one or the second one to be performed by the subjects. In order to check for potential transfer effects we included the between-subjects variable block sequence into the analysis.
We computed
Discussion
The aim of this study was to further elaborate how advance task preparation contributes to the flexibility of human behavior observed in task switching situations. Two different scenarios were sketched in the introduction which led to diverging predictions how advance task preparation should be reflected in BOLD activation. While both accounts agree that advance task preparation essentially means to establish the currently appropriate task set prior to the presentation of the target stimulus,
Conclusion
We presented an interpretation of the typical pattern of results obtained in fMRI studies of explicitly cued task switching which had been partly unexplained so far and which has often been thought to be incompatible with the typical pattern of behavioral results. The theoretical perspective we adopted is compatible with recent behavioral findings which (1) indicate that advance task preparation does not necessarily incorporate the re-configuration of a persistently activated previous task set
Acknowledgements
This work was supported by the German-Israeli Foundation for Scientific Research and Development (GIF) Grant G-635-88.4/1999. The authors wish to thank Birte Forstmann for continuous and forwarding discussions about the topic and two anonymous reviewers for very helpful and constructive comments.
References (58)
Task switching and the pied homunculus: where are we being led?
Trends in Cognitive Sciences
(2003)- et al.
When the same response has different meanings: Recoding the response meaning in the lateral prefrontal cortex
Neuroimage
(2003) - et al.
Neural mechanisms of transient and sustained cognitive control during task switching
Neuron
(2003) - et al.
The preparation and execution of self-initiated and externally-triggered movement: a study of event-related fMRI
Neuroimage
(2002) - et al.
Prefrontal cortex activation in task switching: an event-related fMRI study
Cognitive Brain Research
(2000) - et al.
The roles of timing and task-order during task-switching
Neuroimage
(2002) - et al.
Event-related fMRI: Characterizing differential responses
Neuroimage
(1998) - et al.
Task switching: a PDP model
Cognitive Psychology
(2002) Deconvolution of impulse response in event-related BOLD fMRI
Neuroimage
(1999)- et al.
Separate visual pathways for perception and action
Trends in Neuroscience
(1992)
Lipsia: A new software system for the evaluation of functional magnetic resonance images of the human brain
Computerized Medical Imaging and Graphics
Evidence for anterior cingulate cortex involvement in monitoring preparatory attentional set
Neuroimage
Component processes in task switching
Cognitive Psychology
Task switching
Trends in Cognitive Sciences
Task-set reconfiguration processes do not imply a control homunuculus: Reply to Altmann
Trends in Cognitive Science
Transient neural activity in the medial superior frontal gyrus and precuneus time locked with attention shift between object features
Neuroimage
The assessment and analysis of handedness: The Edinburgh Inventory
Neuropsychologia
Subprocesses of performance monitoring: A dissociation of error processing and response competition revealed by event-related fMRI and ERPs
Neuroimage
Task-switching and long-term priming: Role of episodic S–R bindings in task-shift costs
Cognitive Psychology
Task-Switching, stimulus-response bindings, and negative priming
The role of the frontal cortex in task preparation
Cerebral Cortex
Decomposing components of task preparation with functional MRI
Journal of Cognitive Neuroscience
Neural circuits subserving the retrieval and maintenance of abstract rules
Journal of Neurophysiology
Randomized event-related experimental designs allow for extremely rapid presentation rates using functional MRI
Neuroreport
Matching patterns of activity in primate prefrontal area 8a and parietal area 7ip neurons during a spatial working memory task
Journal of Neurophysiology
Selective averaging of rapidly presented individual trials using fMRI
Human Brain Mapping
Lokalisierung an internen Kontrollprozessen beteiligter Hirngebiete mithilfe des Aufgabenwechselparadigmas und der ereigniskorrelierten funktionellen Magnetresonanztomographie
Statistical parametric maps in functional imaging: A general linear approach
Human Brain Mapping
The prefrontal cortex: Anatomy, physiology, and neuropsychology of the frontal lobe
Cited by (76)
Dissociating strategy-dependent and independent components in task preparation
2014, Neuropsychologia