2 Theoretical and empirical background

2.1 Interference processing in dual tasks

2.1.1 The dual-task paradigm

↓3

The limited ability to perform two tasks simultaneously has been extensively studied in cognitive psychology. The common principle of these studies is to present two tasks more or less simultaneously and to measure behavioral performance costs associated with dual-task processing. However, the applied paradigms also vary depending on the theoretical backgrounds and research questions of the respective authors. Baddeley (1998), for example, frequently used continuous secondary tasks like visual tracking to investigate the properties of assumed sub-systems in human working memory. Other authors like Kahnemann (1973) also mostly applied rather complex component tasks and investigated the resource allocation of their participants to these tasks in order to understand the dynamics of the assumed resource limitations.

↓4

The paradigm of the Psychological Refractory Period (PRP) was established by authors assuming a structural processing limitation in the cognitive system (Pashler, 1994; Welford, 1952). In comparison to the other dual-task paradigms, the PRP-paradigm has the advantage of being very precise with respect to the definition of the ongoing tasks and the underlying processing stages. As will be seen later, these properties also form the basis for the precise investigation of the neural mechanisms involved in dual-task processing.

In the PRP paradigm (see Figure 1), two stimuli (S1, S2) are presented with varying stimulus onset asynchronies (SOA) and participants are required to respond to both stimuli with distinct motor responses (R1, R2). Usually, participants are required to respond to the stimuli according to the presentation order, thus giving priority to S1. The most important finding with the PRP-paradigm is that the processing times for the second of the two tasks (Task 2) increase with decreasing SOA while processing times for the first task (Task 1) are widely unaffected by the SOA manipulation. Telford (1931) first observed this behavioral effect and called it the effect of the Psychological Refractory Period, in analogy to the neural refractory period which relates to the inability of a neuron to elicit two action potentials in short succession.

This idea of the PRP was further specified by Welford (1952) and more recently by Pashler (1994). These authors related the PRP effect to a processing bottleneck inherent to the cognitive system. Assuming that information processing can be divided into several processing stages (Sternberg, 1969), the bottleneck assumption postulates that certain processing stages can proceed in parallel in two temporally overlapping tasks whereas other stages are capacity-limited and can only be processed serially. Accordingly, at high temporal

↓5

Figure 1: The central bottleneck model in the PRP paradigm (Pashler, 1994). Two stimuli (S1, S2) are presented with different temporal overlaps (stimulus onset asynchrony, SOA) and participants are required to respond with two motor responses (R1, R2). The central bottleneck model assumes that at short SOAs response selection (RS) is temporally interrupted in task 2 until task 1 has finished. Stimulus perception (P) and initiation of the motor response (MR) can be processed in parallel in both tasks.

overlap, the two tasks of a dual-task situation compete for access to these capacity-limited processing stages. This competition for attentional processing capacities at a putative processing bottleneck in dual-task situations is also called dual-task interference. The bottleneck model relates the increased reaction times in Task 2 (RT2) to a temporal interruption of Task 2 processing during the processing of the bottleneck stage in Task 1.

According to the so-called “first-come first-served” principle (Pashler, 1994), the task that reaches the capacity-limited processing stage first – usually Task 1 - gains access to this processing stage and the processing of the other task is delayed. Accordingly, RT1 is usually widely unaffected by the SOA manipulation whereas the duration of the interruption in Task 2 and the resulting RT2 depend on the temporal overlap of the two tasks.

↓6

Note that there exists an ongoing debate about the location and the robustness of the dual-task bottleneck within the information processing stream. The original central bottleneck model by Pashler (1994) postulated that the bottleneck is located at the response selection stage (McCann & Johnston, 1992; Pashler, 1994; Schubert, 1999). However, there is also some evidence for capacity limitations at a perceptual (Arnell & Duncan, 2002; Hein & Schubert, 2004; Marois & Ivanoff, 2005) or motor stage (Karlin & Kerstenbaum, 1968; Meyer & Kieras, 1997; Schumacher et al., 2001). Although the studies of the present dissertation do not aim at resolving this debate, evidence for the consistency of the response selection bottleneck may be drawn from all three presented studies.

Importantly, Pashler (1994) assumed that the transition from Task 1 bottleneck processing to Task 2 bottleneck processing is passive. That is, as soon as Task 1 has finished, Task 2 processing automatically continues without the recruitment of additional control processes. This assumption implies an important prediction for the neural implementation of dual-task processing. If no additional control processes are involved in dual-task processing, one might not expect additional brain regions to be involved in dual-task processing compared to the processing of the component single tasks. Previous empirical findings regarding this prediction will be reviewed in chapter 2.2.3. In addition, the presence of dual-task-specific brain activity will also be tested empirically in Study 1 of this dissertation. As will be outlined in the next section, more recent dual-task models conquer this view of passive bottleneck processing. These models postulate the involvement of active cognitive control mechanisms that coordinate the dual-task processing stream.

2.1.2 Cognitive control in the dual-task paradigm

Recent dual-task models assume the involvement of additional processes related to the active control of the processing stream at and before the bottleneck (De Jong, 1995; Meyer & Kieras, 1997; Logan & Gordon, 2001; Sigman & Dehaene, 2006). This idea was initially formulated to explain for the finding that participants can voluntarily switch the processing order in a dual task and do not solely depend on the presentation order (De Jong, 1995, Meyer & Kieras, 1997). The influence of strategic and voluntary control in dual-task processing was then further investigated and tested in computational dual-task models.

↓7

According to these models, the control of the task order in a dual-task situation includes the planning and coordination of the appropriate sequence of actions in two tasks prior to stimulus presentation and during bottleneck processing itself. Recently, Sigman and Dehaene (2006) proposed a computational model for PRP situations where the serial processing of the two tasks at the bottleneck originates from several control mechanisms. First, an attentional task setting mechanism is involved in the planning and coordination of the appropriate action sequence. In addition, an attentional switching mechanism enables the task processes in the second task to proceed after the first task has passed the bottleneck. This is very similar to the conceptions by Logan and Gordon (2001) and also Meyer and Kieras (1997) proposed similar mechanisms in their production-rule-based EPIC (Executive-Process/Interactive Control) architecture (see also Luria & Meiran, 2003). Taken together, the postulated control mechanisms serve the flexible, goal-directed behavior required to deal with the interference of two tasks in a dual-task situation.

Although the proposed computational models provide a detailed description of the involved control mechanism, there is only few direct evidence from cognitive psychology for such control processes (but see De Jong (1995) or Luria & Meiran (2003) for exceptions).

However, various neuroimaging studies tested the basic assumption of these active control models, namely that additional processing requirements are involved in dual-task situations compared to single-task ones. Additional processing requirements should be reflected in additional effort in the brain in dual-task situations compared to single-task situations. The finding of increased brain activity during dual-task processing would provide converging evidence for the involvement of additional active control processes in the processing of dual tasks and may be difficult to reconcile with the assumption of a passive processing bottleneck (Pashler, 1994). It would also provide the basis for more detailed investigations regarding the nature of these control mechanisms. A candidate brain region for the neural implementation of active control mechanisms in dual tasks may be the lateral prefrontal cortex which has been consistently associated with cognitive control mechanisms in single tasks (Miller & Cohen, 2001; Norman & Shallice, 1986; Passingham, 1993).

2.2 The functional role of the lateral prefrontal cortex (lPFC)

2.2.1 The lPFC and cognitive control

↓8

Several authors have argued that cognitive control mechanisms are the key function of the lPFC (Miller & Cohen, 2001; Norman & Shallice, 1986, Passingham, 1993; Petrides, 2000). Cognitive control describes the ability to coordinate thoughts and actions in accordance with internal goals in order to elicit coordinated and purposeful behavior (Fuster, 1989; Miller & Cohen, 2001; Koechlin, Ody, & Kouneiher, 2003; Norman & Shallice, 1986). As we are permanently confronted with multiple options for behavior, it is important that we control which behaviors are executed in which order and which behaviors are not executed at all. On the one hand, we have to flexibly overcome reflexive and automatic behavior that interferes with intended goal-directed behavior. On the other hand, we also need to maintain and coordinate multiple relevant information streams according to our internal goal hierarchies. These two mechanisms – the inhibition of interfering prepotent response tendencies including the attentional re-focussing on relevant information as well as the maintenance and coordination of multiple relevant information streams as it is necessary in dual-task processing – both constitute important aspects of cognitive control.

The crucial role of the lPFC for cognitive control is supported by ample empirical evidence from single-cell recordings in monkeys, neuropsychological patients and neuroimaging studies (for the neuroanatomical landmarks of the lPFC, see Figure 2).

In single-cell recordings in monkeys, lateral prefrontal neurons have been shown to have the capability to maintain relevant information (die Pellegrino & Wise, 1991; Fuster & Alexander, 1971; Goldman-Rakic, 1987; Kubota & Niki, 1971), even in the face of distracting information (Miller, Erickson, & Desimone, 1996). At the same time, lateral prefrontal neurons are highly flexible and can adopt various task rules (Bunge, Kahn, Wallis, Miller, & Wagner, 2003; Muhammad, Wallis, & Miller, 2006; Wallis, Anderson, & Miller, 2001) even including contingencies on a higher-order level (Shima, Isoda, Mushiake , & Tanji, 2007). These properties of the lateral prefrontal neurons form the basis for flexible, goal-directed behavior.

↓9

In accordance with these findings in monkeys, human patients with lateral prefrontal lesions show severe deficits in daily-life behaviors and experimental paradigms involving cognitive control. This includes deficits in the maintenance of information against distractor

Figure 2: Cytoarchitectonic map by Brodmann (1909). Lateral view. Lateral frontal cortex is colored. red: motor cortex; orange: premotor cortex; blue: prefrontal cortex. Numbers refer to the Brodmann areas. (taken from Barbas, Ghashghaei, Rempel-Clower, & Xiao, 2002).
The lateral prefrontal cortex is part the frontal lobes which comprise the most anterior part of the cerebral hemispheres. Identification and classification of subregions within the frontal lobes are based on morphological features like surface landmarks and microscopic analyses of the constituent neurons resulting in cytoarchitectonic maps. The depicted cytoarchitectonic map by Brodmann (1909) is one of the most widely accepted. Three primary functional subregions of the frontal lobes can be identified on the caudal-to-rostral axis of the lateral frontal surface: motor cortex, premotor cortex and prefrontal cortex. In addition, medial frontal cortex has its own subdivisions, interacting strongly with the lateral frontal regions. The primary motor cortex (Brodmann’s area (BA) 4) is the smallest and most homogeneous of these regions, mainly stretching along the central sulcus. Rostral to BA4, the lateral premotor cortex (BA 6) extends along the precentral sulcus and gyrus. Often, BA 8 (frontal eye fields) and BA44 (pars opercularis) are also counted to the lateral premotor cortex. All cortical regions anterior to the premotor cortex are called the prefrontal cortex (PFC). The PFC may be subdivided into several subregions: (1) dorsolateral PFC (dlPFC, BA 9/46), (2) ventrolateral PFC (vlPFC, BA 45/47), (3) anterior PFC (aPFC, BA 10), (4) orbitofrontal Cortex (OFC, BA 11/12/13/14/47), (5) medial PFC (mPFC, BA 24/32). The dlPFC and the vlPFC can be anatomically separated as the neural substrate dorsal and ventral to the inferior frontal sulcus (IFS), respectively. It will be referred to these two subregions stretching along the three frontal gyri (superior, middle and inferior frontal gyrus), when using the term “lateral prefrontal cortex”.

interference (Chao & Knight, 1995) as well as the inability to flexibly switch between task representations (Aron, Monsell, Sahakian, & Robbins, 2004; Stuss, Floden, Alexander, Levine, & Katz, 2001). The latter is also exemplified in the Wisconsin Card Sorting Test (WCST; Grant and Berg, 1948), a prominent test for lateral prefrontal functioning. In the WCST, subjects are required to place the top card of a card deck under one of four target cards according to a sorting rule. The sorting rule, however, is only indicated implicitly by the experimenter’s feedback about the correctness of the current response. After ten consecutive correct responses the sorting rule changes unbeknownst to the subject. Patients with lateral prefrontal damage are frequently unable to use the feedback of the experimenter in order to switch to the newly relevant sorting rule. They perseverate on the previous rule, unable to flexibly change their behavior according to the new context (Barceló & Knight, 2000; Stuss et al., 2000). Patient studies like this support the importance of lateral prefrontal regions for cognitive control. Note, however, that the WCST is a very complex task consisting of various cognitive components so that no direct inference about the functionality of the lPFC can be made solely based on such neuropsychological findings. Importantly, the involvement of the lPFC in specific components of the WCST was further specified in neuroimaging studies that showed increased lPFC activity particularly related to rule shifts in the WCST (Monchi, Petrides, Petre, Worsley, & Dagher, 2001; Konishi et al., 1998; Lie, Specht, Marshall, & Fink, 2006). Thus, the flexible switching between task rules seems to be the crucial component associated with the lateral prefrontal cortex in the WCST.

↓10

Similarly, a vast amount of fMRI studies showed that the lPFC is related to cognitive control in various task situations. In particular, this includes the maintenance of task rules in the face of distraction (de Fockert, Rees, Frith, & Lavie, 2001; Sakai & Passingham, 2003) and the control of interference processing in single tasks like the Stroop task (Banich et al., 2001; Zysset, Müller, Lohmann, & von Cramon, 2001), in task switching (Braver, Reynolds, & Donaldson, 2003; Dove, Pollmann, Schubert, Wiggins, & von Cramon, 2000), the Simon Task (Fan, Flombaum, McCandliss, Thomas, & Posner, 2003; Liu, Banich, Jacobson, & Tanabe, 2004), the Flanker paradigm (Casey et al., 2000; Hazeltine, Poldrack, & Gabrieli, 2000) and incompatibly mapped choice reaction tasks (Schumacher & D’Esposito, 2002; Schumacher, Elston, & D’Esposito, 2003). Common to all these paradigms is the requirement to suppress prepotent response tendencies interfering with the required responses and thus flexibly switch to another stimulus dimension or stimulus-response mapping. Thus, all these paradigms tag cognitive control as defined above, supporting the view that the lPFC is crucially involved in cognitive control.

The neural mechanisms of cognitive control, in particular the roles of conflict monitoring (Botvinick, Braver, Barch, Carter, & Cohen, 2001) and top-down attentional control (Hopfinger, Buonocore, & Mangun, 2000) are recently debated. The role of the lPFC in cognitive control in such interference situations can be understood as the biasing of task processing in posterior brain regions to resolve interference (Badre & Wagner, 2004; Miller & Cohen 2001). Importantly, there exists recent evidence for the presence of conflict-contingent amplification of activity in task-relevant regions compared to task-irrelevant regions (Egner & Hirsch, 2005). Egner and Hirsch (2005) used a variant of the Stroop paradigm and compared so-called high-control compared to low-control situations. High-control situations were related to conflict adaptation effects (see also Kerns et al., 2004) for recent Stroop trials preceded by an incongruent Stroop trial (high control) compared to precedence by a congruent Stroop trial (low control). Reduced behavioral interference effects for incongruent trials in high-control situations was associated with cortical amplification of activity in task-relevant sensory brain regions. Even more importantly, reduced behavioral interference effects were also associated with increases in functional coupling beween task-relevant regions and the lPFC. Accordingly, the top-down modulation of task-relevant regions exerted by the lPFC seems to be an important mechanism for dealing with interference between relevant and irrelevant task representations.

As outlined above, cognitive control is also important in dual tasks, controlling the processing associated with the putative bottleneck. In dual-task situations, however, two stimuli are both relevant for subsequent behavior within the same task trial. This renders the task situation more complex as both task representations are relevant and attention to the second stimulus must also be present to some degree so that the second task can still be performed correctly. Both task streams have to be maintained and coordinated so that they both can find expression in behavior. Although there is an emerging literature on the involvement of the lPFC in dual–task processing, only little is known about how general this involvement is and how exactly the lPFC exerts cognitive control in such dual-task situations. The specification of the functional role of the lPFC in cognitive control in dual tasks is the primary aim of this dissertation.

2.2.2 lPFC involvement in dual tasks - Is the whole more than the sum of its parts?

↓11

Early neuroimaging studies on dual-task processing in the lPFC (Adcock, Constable, Gore, & Goldman-Rakic, 2000; Bunge, Klingberg, Jacobsen, & Gabrieli, 2000; D’Esposito et al., 1995; Goldberg et al., 1998; Jaeggi et al., 2003; Just et al., 2001; Klingberg, 1998; Koechlin, Basso, Pietrini, Panzer, & Grafman, 1999; Smith et al., 2001) were mainly concerned with the question whether there exist regions in the lPFC related to additional processing requirements in dual tasks or not. In most of these studies, rather complex tasks like mental rotation, semantic categorization or the reading span test were administered either separately, as single tasks, or temporally overlapping, as dual tasks. The comparison of the activity changes in dual-task blocks with the sum of the single-task blocks was used as an indicator whether the “whole is more than the sum of its parts” (Duncan, 1979), that is, whether additional dual-task-specific activity is elicited in dual tasks or not. Whereas some of these studies did not find dual-task-specific activity in the lPFC (Adcock et al., 2000; Bunge et al., 2000; Klingberg, 1998; Smith et al., 2001) others found either increased dual-task activity in the lPFC compared to the single tasks (Jaeggi et al., 2003) or even additional lPFC regions involved in dual-task blocks that were not involved supra-threshold in single-task blocks (D’Esposito et al., 1995; Koechlin et al., 1999).

These divergent results on the involvement of the lPFC in dual-task processing might have several reasons. Most likely, the complexity of the applied component tasks leaves many degrees of freedom in the way the dual task is performed by the participants in the different studies - with more or less additional processing requirements being involved. In addition, the rather complex component tasks in these studies might already involve cognitive control processes associated with lPFC regions that overlap with the potential dual-task-related control regions. Accordingly, no additional dual-task-related activity could show up in the comparison of the dual-task blocks with the single-task blocks.

More recent fMRI studies on the functional neuroanatomy of dual-task processing therefore applied better controlled dual tasks using the PRP paradigm (Collette et al., 2003; Erickson et al., 2005a; Schubert & Szameitat, 2003). As outlined above, the advantage of the PRP paradigm is that the applied choice reaction tasks are well-defined and that there exist precise assumptions regarding the underlying processing stream. When comparing the activity changes in the lPFC associated with two choice reaction tasks performed as single versus dual tasks, clear evidence for a dual-task-specific involvement of the lPFC was found (Schubert & Szameitat, 2003; Szameitat et al., 2002). The consistently activated dual-task-related regions were located in regions around the inferior frontal sulcus (IFS) and in regions of the middle frontal gyrus (MFG). Schubert & Szameitat (2003) related this activity to mechanisms of interference control in dual tasks. In addition, Erickson et al. (2005a) showed that dual-task-related regions in the lPFC are not just related to differences in task preparation between dual-task and single-task blocks. In their study, Erickson et al. (2005a) mixed dual-task and single-task trials within the same task blocks such that task preparation was identical for both task types. Still, increased activity in the lPFC, particularly in the left posterior lPFC and the bilateral IFG was found in dual-task trials compared to single-task trials. Applying the well-controlled PRP paradigm, these studies show consistently that the lPFC is involved in dual-task processing.

↓12

However, for several reasons, no unequivocal inferences can be made from these studies with respect to the underlying cognitive mechanisms associated with the dual-task-related lPFC activity.

First, it remains unclear which type of interference is reflected in the additional activity in the lPFC. As outlined above, interference can emerge at different processing stages – related to perception, response selection or the motor response. The studies above do not provide unequivocal evidence that additional processing requirements reflected in the increased lPFC activity are related to the processing of a response selection bottleneck. These studies used either overlapping stimulus modalities (Erickson et al., 2005) or overlapping response modalities (Schubert & Szameitat, 2003). Therefore, the obtained dual-task-related activity changes in the lPFC might also be related to the resolution of perceptual or motor interference instead. In chapter 2.2.3, this possibility will be discussed in detail. In Study 1 of this dissertation, these alternative sources of dual-task-related activity in the lPFC will be eliminated by using component tasks without any overlap on a perceptual or motor level. The finding of dual-task-related activity in such a task situation could then be attributed to the processing associated with the response selection bottleneck.

Second, comparing dual-task blocks with single-task blocks may reflect any cognitive difference between these types of blocks. Besides the outlined mechanism of task order control, this may also include differences in working memory load1 or divided attention which are clearly more demanding in dual-task blocks. In order to understand the functionality of the dual-task-related lPFC regions in more detail, the underlying cognitive mechanisms need to be identified. For this purpose, in all three studies of this dissertation, parametric fMRI designs are used (Braver et al., 1997). By manipulating the difficulty of specific cognitive functions within different dual-task blocks, specific hypotheses with respect to the functionality of the lPFC in dual-task processing can be tested (see also chapter 2.2.4). In addition, in Study 2 different control functions distinguishing dual tasks from single task were manipulated within the same experiment, giving the possibility to compare the localization of these functions within the lPFC.

↓13

Third, focusing exclusively on the lPFC when investigating the functional neuroanatomy of dual-task processing might provide an incomplete picture of the involved neural mechanisms. As outlined above, the lPFC interacts strongly with posterior regions to control task processing according to internal goals in single tasks. The investigation of the functional integration between lPFC and posterior task-relevant regions in dual-task situations is the aim of Study 3.

2.2.3 Types of interference processing in the lPFC (Study 1)

As outlined above, the studies reported so far do not indicate which type of interference is associated with the dual-task-related lPFC activity. The involvement of processing related to perceptual or motor interference cannot be excluded. Some more recent dual-task neuroimaging studies addressed this question (Dux et al., 2006; Herath et al., 2001; Jiang, 2004). Jiang (2004) and Herath et al. (2001) investigated this by using variants of the PRP paradigm to parametrically manipulate certain aspects of dual-task processing. Both studies found increased activity in high-interference dual-task situations at short SOAs compared to long SOAs in a region in the posterior lPFC (plPFC). However, Jiang (2004) found the SOA effect exclusively for conditions with both stimuli presented in the periphery of a circular display, thus requiring the simultaneous allocation of attention in space. Accordingly, Jiang (2004) concluded that this region in the plPFC is related to the resolution of perceptual interference rather than to the resolution of cognitive bottleneck interference in dual tasks. In the study by Herath et al. (2001), however, stimuli in different perceptual modalities (visual, somatonsensory) were presented. The finding of an SOA-related activation in the plPFC in that study excludes the possibility that exclusively perceptual interference is related to plPFC. In addition, it is rather unlikely that interference at a cognitive level is associated with the activity in that study, as simple detection tasks were used which seem not to require response selection and therefore do not interfere at a cognitive level of processing (Schubert, 1999). Consequently, Herath and colleagues associated this region with the processing of interference at a motor level. Although these studies excluded alternative accounts for dual-task-related lPFC activity due to their parametric manipulations of dual-task interference, no direct evidence for interference at a central response selection stage was provided so far. Only by excluding an overlap of stimulus- and response modalities of the component tasks, direct evidence for an association of dual-task-related lPFC activity to the processing of the response selection bottleneck can be obtained Study 1 of this dissertation applied such an approach, using non-overlapping modality pairings for the component tasks2. In addition, the degree to which these non-overlapping modality pairings are compatible with each other was manipulated in Study 1. By comparing dual-tasks with modality-compatible (e.g. visual-manual and auditory vocal) and modality-incompatible (e.g. visual-vocal and auditory manual) tasks as component tasks, an additional manipulation of the degree of central task interference was introduced (Hazeltine, Ruthruff, & Remington, 2006). As shown by Hazeltine and colleagues (2006), different pairings of non-overlapping stimulus-response modality pairings may differ with respect to content-dependent interference. That is, athough there is no overlap in perceptual or motor processing, interactions between the task-related central codes may differ. This would indicate that central interference is not generic, but depends on the task contents.

To investigate the neural effects of such a manipulation, we used individually determined regions of interest in the IFS, obtained from the dual-task vs. single-task-task contrast. In so far, Study 1 can provide crucial information regarding the question whether the processing of central bottleneck interference is associated with the lPFC during dual-task processing.

2.2.4 Neural implementation of dual-task-related cognitive control in the lPFC (Study 2)

↓14

Surprisingly little is known about the neural implementation of control processes that are involved in dual-task processing. As outlined above, recent dual-task models assume that active control processes are involved in the coordination of the dual-task stream by setting task priorities and switching between the two task streams (Logan & Gordon, 2001; Meyer & Kieras, 1997; Sigman & Dehaene, 2006). These mechanisms of task order control serve the optimal task performance despite of the involved bottleneck and may contribute essentially to the dual-task-specific lPFC activity that was found previously.

To my knowledge, only the two studies by Szameitat and colleagues (2002; 2006) addressed the question whether there are neural correlates of dual-task-related control processes in the lPFC. Szameitat et al. (2002) used a version of the PRP paradigm and compared the fMRI signals between dual-task blocks with different demands on task order control. In their dual-task paradigm, participants performed a visual-manual and an auditory-manual choice reaction task in every trial. The two tasks were presented in dual-task blocks with either random temporal order of the two component tasks or in blocks with fixed order. Within random-order blocks, the task order of the two component tasks changed randomly from trial to trial. Accordingly, participants needed to re-arrange and control the processing order permanently in order to perform the dual tasks in the correct temporal order. The increased demands on the computational processes related to task order control led to increased reaction times and error rates in random-order compared to fixed-order blocks (see also De Jong, 1995; Luria & Meiran, 2003). Even more importantly, when comparing the activity changes in random-order and fixed-order blocks, Szameitat et al. (2002) found an extended fronto-parietal network with bilateral activation foci in the lPFC. The lPFC activation was mainly located in regions surrounding the left and right IFS extending from anterior to posterior portions of this sulcus and dorsally into the MFG. These activation foci overlapped closely with the activation foci obtained when subtracting the signal changes in single-task blocks from those in dual-task blocks as indicated in an additional analysis of the same study. Szameitat et al. (2002) concluded that these dual-task-related regions in the lPFC are associated with the control of the task order in dual-task situations.

Even stronger evidence for this conclusion comes from their event-related study (Szameitat et al., 2006), where task order control was manipulated within the same task blocks. In detail, Szameitat et al. (2006) compared the activity changes in so-called same-order and so-called different-order dual-task trials. While in same-order trials the processing order of the two component tasks in a given trial N (e.g., visual then auditory task) was identical to trial N-1, the order of the two component tasks was reversed between trial N and N-1 in different-order trials. According to the assumption that mechanisms of task order control may rely on the episodic trace of the task order in the previous trial, task order control difficulty was expected to be increased in different-order compared to same-order trials (De Jong, 1995; Luria & Meiran, 2003). As expected, the processing times and the error rates were elevated in different- compared to same-order dual-task trials and, even more importantly, these differences were associated with two activation peaks in the lPFC. These were located in the right MFG and along the left IFS overlapping with the activity peaks from Szameitat et al. (2002). Thus, there is strong evidence that task order control is one cognitive mechanisms that is associated with dual-task-related regions in middle and posterior portions of the lPFC.

↓15

However, the attribution of dual-task-related lPFC activity to task order control processes was recently objected to by other authors (Jiang, Saxe, & Kanwisher., 2004). Jiang et al. (2004) noted that the demands to maintain additional task set components in working memory might cause the additional lPFC activity in dual-task compared to single-task situations. According to this argument, task order control might not be the only factor underlying dual-task-specific lPFC activity. The maintenance of additional task set components may also be crucial. If this is the case, it would be important to investigate whether both functions, task order control and task set maintenance, use overlapping or non overlapping neural substrate in the lPFC. In Study 2, we manipulated task order control and task set maintenance orthogonally to identify the contribution of both functions to lPFC activity during dual-task performance.

2.2.5 Interaction of the lPFC with other brain regions during dual-task processing (Study 3)

The understanding how the lPFC interacts with other task relevant regions is crucial for the understanding functionality of the lPFC in dual-task processing. As Miller & Cohen (2001) stated, an essential function of the lPFC is the biasing of signals to other brain regions to guide the flow of activity along neural pathways in accordance with internal goals.

Of specific importance for the understanding of this top-down control is the pattern of connectivity of the lPFC subregions with the rest of the human brain (Barbas et al., 2002; Petrides & Pandya, 1999). All PFC regions, including the lPFC, have distinct patterns of cortical connectivity with other regions throughout the brain (ses also Fuster, 1989 and Goldman-Rakic, 1987). These connections are mostly reciprocal and enable prefrontal regions to integrate various types of information and to exert a top-down influence on other regions in order to coordinate information processing across a wide range of the central nervous system (Desimone & Duncan, 1995; Miller & Desimone, 1994).

↓16

It was outlined above (chapter 2.2.1) that the top-down modulation of task-relevant regions has been shown to be an important mechanism for dealing with interference between relevant and irrelevant task representations in single-task situations (Egner & Hirsch, 2005). Other studies using measures of functional connectivity also support the importance of top-down control for flexible, goal-directed behavior (Abe et al., 2007; Erickson, Ringo Ho, Colcombe, & Kramer, 2005b; Gazzaley, Rissmann, & D’Esposito, 2004).

However, for dual tasks where two task representations are simultaneously relevant there is no evidence yet whether and how the lPFC interacts with task-relevant regions. According to the computational dual-task models by Sigman & Dehaene (2006) or Logan and Gordon (2002), so-called attentional task setting mechanisms might be crucial for the observed behavioral pattern present in the PRP-effect. In Study 3, the interactions of the lPFC with posterior task-relevant regions as well as the activity pattern in these task-relevant regions associated with the PRP effect were investigated. For this purpose a localizer approach (see chapter 3.2.2) and functional connectivity measures (see chapter 3.2.3) were applied in Study 3.

2.3  Summary

A predominant function of the lPFC is cognitive control – the ability to coordinate thoughts and actions in accordance with internal goals in order to elicit coordinated and purposeful behavior. Dual-task processing is unique in that sense, as two internal goals are simultaneously relevant and need to be coordinated in order to successfully perform both tasks. The reviewed literature shows (1) that the lPFC is involved dual-task processing,
(2) that different types of dual-task interference are related to the lPFC and (3) that active mechanisms of task order control might underlie dual-task-related activity in the lPFC. However, there are several open questions with respect to the exact functional role of the lPFC in dual-task processing which will be addressed in the present dissertation. These concern the generality of previous findings, the dissociability of different control processes in the lPFC and the interaction of the lPFC with posterior task-relevant brain regions.

↓17

Before summarizing the specific research questions and results of the three studies of this dissertation, I will give a short overview on the applied methods in these studies.


Footnotes and Endnotes

1  This is not the case for the Erickson et al. (2005) study

2  Only recently, after the publication of Study 1 of this dissertation, additional evidence for the association of the plPFC with the response selection bottleneck was provided by Dux et al. (2006). These authors also used non-overlapping modalities in their component task of a PRP paradigm applying rapid time-resolved fMRI acquisition. In particular, they tested the serial postponement prediction of the central bottleneck model (Pashler, 1994). Serial postponement relates to the idea that response selection in Task 2 is delayed at short SOAs as long as response selection is ongoing in Task 1. This delay is longer, the longer response selection takes in Task 1. In contrast, at long SOAs no such effect of the duration of response selection in Task 1 on RT2 should be present. Dux et al. measured signal latencies in plPFC regions that were related to response selection at short and long SOAs. The comparison of trials with slow versus fast RT1 revealed a differential pattern at short compared to long SOA consistent with the serial postponement prediction. That is, plPFC peak latency depended on RT1 speed at short SOA with prolonged activity for slow RT1. No similar effect was found at long SOA. This result further supports that the plPFC is associated with the central processing bottleneck characterized by the inability to perform two decisional processes at the same time.



© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 4.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated:
12.06.2008