Abstract

When people perform two tasks simultaneously, the tasks are often executed slower and with more errors than when they are carried out as single tasks. This is called dual task interference. With functional magnetic resonance imaging (fMRI), we show that concurrently performed visual and somatosensory reaction time (RT) tasks engage almost identical volumes of cortical and subcortical motor structures. Moreover, dual RT tasks engaged additional cortical regions that are not activated by the component RT tasks had they been performed as single tasks. When the inter-stimulus interval was <300 ms, the first task interfered with the second, and a field in the right inferior frontal gyrus (RIFG) appeared with activity correlated with the increased RT to the second stimulus. This activation was spatially distinct from the cortical activity of the main effect of dual task performance. Thus, the performance of single RT tasks, dual RT tasks and dual RT tasks that interfere differ psychophysically and in the brain structures subserving these tasks. A short occupancy of the common motor structures can explain the interference effect. The increased activity of the RIFG correlated with the interference effect is very likely to be a specific outcome of situations where two concurrent tasks interfere with each other. The brain appears to recruit the RIFG for a subsequent (delayed) response when there is interference between dual tasks.

Introduction

Experimental conditions in which the performance of one task affects the performance of another has been the focus of much research for many decades (Telford, 1931; Welford, 1952). For example, in dual reaction time (RT) tasks people need to make separate responses to two stimuli presented in rapid succession. Here, when the performance is measured relative to the single tasks, the reaction times and probability for making errors in responding to the second stimulus (RT2) of the pair that required responses is increased as a function of the inter-stimulus interval (Welford, 1952) (Fig. 1). Such deterioration of performance, called dual task interference, is present in all dual tasks with short inter-stimulus intervals (ISIs) (Telford, 1931; Welford, 1952; Pashler, 1994,1999; Meyer and Kieras, 1997). Dual task interference is defined as increased RT2 and increased error probability.

The neurophysiological basis of dual task interference is unknown. One of the neurobiological hypotheses that attempt to address the physiological basis of dual task interference is the cortical field hypothesis (CFH) (Roland and Zilles, 1998). This posits that if two different brain tasks make use of the same cortical fields, the tasks cannot be performed simultaneously. Thus, a large population of neurons cooperatively engaged in a particular computation may not be available to perform a concurrent task. Understanding whether such a population of neurons comprising a cortical field is locked to one task for a certain time, during which the field is unavailable for other tasks, thus leading to interference, is crucial to our understanding of the functional organization of the cerebral cortex. Notwithstanding the CFH or the lack of other physiologic hypotheses, however, recently there have been a number of neuroimaging studies of dual tasks. Some of these studies found prefrontal and parietal activations that were described as specifically related to the performance of a dual task (D'Esposito et al., 1995; Corbetta, 1998; Koechlin et al., 1999; Dove et al., 2000), while others failed to show additional cortical activity associated with dual tasks (Passingham, 1996; Goldberg et al., 1998; Klingberg, 1998; Adcock et al., 2000; Bunge et al., 2000).

Dual tasks imply division of attention between two tasks (Braun, 1998). Interference may or may not be an additional, specific, physiological phenomenon. To date, no study has distinguished the neural correlates of division of attention between two simultaneously performed tasks from those of interference that may occur between them. If there is a distinction between the cortical structures engaged in division of attention in dual tasks and those being active in dual task conditions that interfere, this would substantiate that these are two different physiological phenomena. Furthermore, if differences in activation of cortical areas were associated with different psychophysical results, for example a differential effect on RT1 and RT2, this would further corroborate that the division of attention and interference are indeed two different neurophysiological phenomena.

Using a behavioral paradigm of single and dual reaction time tasks, we investigated a number of hypotheses. Firstly, we tested whether two RT tasks, when performed simultaneously, would occupy the same motor cortical regions, thus leading to interference as proposed by the CFH. Next, we tested whether dual task performance would require engagement of additional cortical regions in excess of the regions that would be active in each of the component tasks if they were performed as single tasks, thus leading to interference. Thirdly, we investigated whether the neural correlates of performing two simultaneous tasks could be dissociated from those of interference and whether there is cortical activity specific to the interference, suggesting that the neural correlates of dual task performance and those of interference are distinguishable from one another.

Subjects underwent fMRI while performing single RT tasks to visual and somatosensory signals or to a combination of these in two dual task situations with short and long ISIs (Fig. 1). They responded to the visual stimuli by pressing a button with the middle finger, and to the somatosensory stimuli by pressing it with the index finger, of the right hand. Our behavioral paradigm did not require the task-related activity to be retained over time, so that the cortical activity specific to dual tasks is not contaminated by short-term memory-related cortical activations. This has been a major confounding factor in all previously reported dual task imaging paradigms (D'Esposito et al., 1995; Klingberg and Roland, 1997; Corbetta, 1998; Klingberg, 1998; Koechlin et al., 1999; Adcock et al., 2000; Bunge et al., 2000; Dove et al., 2000). Elimination of this factor therefore allowed us to clearly delineate and compare brain activations specifically associated with single tasks, and dual RT tasks with and without interference respectively.

Materails and Methods

Subjects, Design and the Imaging Procedures

Ten healthy right handed adults (age 22–32 years; three females and seven males) gave informed written consent to participate in the experiment, which was approved by the local Ethics Committee of the Karolinska Hospital, Stockholm. The functional imaging sessions consisted of four tasks (each block lasted 30 s, mean trial frequency = ~0.8 Hz) and two baseline conditions, which were presented as blocks in a pseudorandom order across sessions. The somatosensory stimulus was a blunt stylus that made a painless, 2 mm indentation on the tip of the left index finger. The impulses were generated by a computer-controlled solenoid and transmitted to the bore of the magnet through a string system made of Kevlar®. The transmission system had a constant delay of 10 ms, and the stimulus had a rise time of 6 ms that was compensated for by the software. The visual stimulus was a green LED (625 nm, 5 V, diameter 3 mm, intensity 8 mcd) that was transmitted through fiber optic cables to a panel mounted on the head coil, ~10 cm away from the eyes. In each modality, the stimulus duration was 50 ms. We maintained a constant background illumination of 0.5 cd/m2 during scanning experiments. The stimuli were driven and the responses collected by ERTS® software (Berisoft Corp., Frankfurt, Germany). Behavioral data were analyzed with JMP® software (SAS Inc., USA). All subjects were familiarized with the experimental tasks before scanning.

Behavioral Tasks

The experimental conditions were RT tasks involving brief signals, namely, (i) single visual RT task; (ii) single somatosensory RT task; (iii) dual long ISI (1150, 1200, 1250 ms) visual/somatosensory RT task; (iv) dual short ISI (200, 250, 300 ms) visual/somatosensory RT task; (v) control; and (vi) rest. The number of stimuli (24) and of motor responses (24) were matched between the single and the dual RT tasks. In dual RT tasks, trials of paired signals with different ISIs were presented randomly and with equal probability. As such, the stimuli had a mean frequency of 0.8 Hz. This was similar to the mean stimulus frequency of the single RT tasks. The mean intertrial interval was ~1150 ms (range 1000–1300 ms) in all tasks. Subjects responded to the stimuli by pressing a pre-specified button on a fiber optic response devise — with the right index finger for the somatosensory signals and with the right middle finger for the visual signals. This way, we maintained very high stimulus– response (S–R) compatibility during the experiment. The RT task was to press the appropriate button as quickly and accurately as possible when a signal was detected. In the control condition, our volunteers received 24 visual and 24 somatosensory stimuli within a 30 s block. They were required to attend to the stimuli, but were asked to make no responses to them. Consequently, the control condition perceptually matched each of the single RT tasks of either modality. In the rest condition the volunteers had their eyes closed and were asked to think of nothing in particular. During all tasks (except for the rest condition) subjects fixated on a red LED (565 nm, 5 V, diameter 3 mm, intensity 8 mcd) that was placed <1° lateral to the green LED. We expected that this would minimize eye movements. For each subject the RT data are the averages of 384 trials that spanned ~8 min per condition across 16 scanning sessions (i.e. 30 s block/task/session × 16 sessions). RT data were sorted as correct hits [(yes)|stimulus], misses [(no)|stimulus] and false alarms [(yes)|no stimulus) or (yes)|stimulus with RT < 125 ms, where such responses were considered as anticipatory responses] (Gescheider, 1997) for each modality. Dual task trials that recorded only one response per trial (i.e. two stimuli) were classified as misses. The RT data from such trials where subjects missed one or the other stimulus were excluded from the calculations of mean RTs. Following the scanning experiment, all subjects were debriefed and we recorded their assessment of task difficulty and other comments.

Imaging

The scanning experiments were conducted on a clinical 1.5 T scanner (Signa Horizon Echospeed, GE Medical Systems) equipped with a head coil. T2*-weighted gradient-echo, echo-planar image volumes with blood oxygenation level-dependent (BOLD) contrast were collected with the pulse sequence; TE = 60 ms, TR = 5000 ms, flip angle 90°, 64 × 64 matrix, field of view (FOV) 240 mm, voxel size 3.4 × 3.4 × 5.0 mm3. Each functional scanning session lasted 270 s, and comprised the four experimental task blocks and the baseline conditions randomly presented. The block duration was 30 s. Each functional volume consisted of 21 contiguous axial slices. A total of 864 such volumes were obtained from each subject in 16 sessions that lasted ~85 min. Additionally, a highresolution, three-dimensional gradient-echo T1-weighted anatomic image volume of the whole brain was obtained using the three-dimensional SPGR sequence; TE =4.2 ms, TR= 13 ms, FOV= 240 mm; flip angle 40°, voxel size 0.96 × 0.96 × 2.0 mm3.

Data Processing and Statistical Analysis

Processing of reconstructed functional echo-planar data was carried out using the routines provided in the SPM99 software (Wellcome Department of Cognitive Neurology, UK; http://www.fil.ion.ucl.ac.uk/spm). For each subject, functional data were first corrected for motion artifacts using a least squares approach and a six-parameter rigid body spatial transformation. (The estimated motion parameters were consistently <0.5 mm translation in any direction and <0.15° rotation in any plane for all subjects, and thus we did not have to exclude any scans from the analysis.) Then, the high-resolution anatomical image and the functional images were co-registered. Following this, first the anatomical brain volume and subsequently the functional volumes were stereotactically normalized into the space of Talairach and Tournoux (Talairach and Tournoux, 1988) using the MNI T1152average template. This procedure involved the estimation of the optimum 12-parameter affine registration that is followed by the estimation of non-linear deformations using a combination of three-dimensional discrete cosine transform basis functions. The normalized functional volumes were spatially smoothed using a three-dimensional Gaussian kernel of 6 mm at full width half maximum (FWHM). Time series data were then temporally smoothed with a Gaussian filter of 4.0 s FWHM. Condition specific effects were estimated for single subjects using a fixed effects General Linear Model with a delayed boxcar waveform as implemented by SPM99. Realign parameters were estimated and the global effects were included as covariates in the linear model. Areas of significant cortical activity were specified by appropriate linear combinations of parameters estimated in the linear model. Following the first level of (within-subject) analysis, a random effects analysis (Friston et al., 1999) was performed for each effect of interest. This procedure allows between-subject variability to be taken into account and therefore inferences to made about the population in general.

Statistical inferences were drawn at the cluster level rather than at the voxel level, such that the omnibus probability is .05. This is achieved by using distributional approximations from the Theory of Gaussian Fields (Worsley, 1995) where voxel level intensity was first thresholded at Z = 3.25 (P < 0.0005) and then setting the extent threshold appropriately such that the omnibus P value (P < 0.05) is preserved. In the analysis of intersection and union of activation patterns, the respective contrasts were first orthogonalized with respect to each other.

Cytoarchitectonic Mapping

The activations within the primary motor cortex and the right inferior frontal gyrus were examined for localization within cytoarchitectonic areas 4a, 4p (Geyer et al., 1996) and Area 44 (Amunts et al., 1999) respectively. Briefly, T1-weighted MRI scans (1.5 T Siemens Magnetron SP: three-dimensional FLASH pulse sequence) were obtained from 10 post-mortem brains that had been fixed in formalin or Brodian's solution. The brains were then embedded in paraffin and sectioned at 20 μm. The sections were stained for cell bodies. The areas 44, 45, 4a and 4p were defined with a observer-independent cytoarchitectonic technique and were corrected for distortions and shrinkage (Schormann and Zilles, 1998; Schleicher et al., 1999). Borders of cytoarchitectonic areas were transferred to corresponding digitized sections and reconstructed three dimensionally. Individual brains and their cytoarchitectonic areas were then transferred to the standard anatomical format of the human brain atlas (Roland et al., 1994). Corresponding areas from the different brains were superimposed in three-dimensional space and population maps were calculated for each area (Roland and Zilles, 1998). These maps describe, for each voxel, how many brains have a representation of a particular cytoarchitectonic area in the identical anatomical format. Voxels representing a brain were allocated to a particular cytoarchitectonic area if they had a higher probability of belonging to this area than to the others. We answered the question of whether a particular cytoarchitectonic area was activated by taking the intersection of that area and the cluster of functional activation. Since the activation fields originally lay in the Talairach space that corresponded to the MNI T1152average template, we transformed the functional images to the standard HBA space by using a set of 12-parameter affine transforms. This was then followed by a set of non-linear deformations that resulted in a nearest neighbor interpolation of the activation clusters.

Results

Behavior

All behavioral data presented here were obtained during the scanning sessions and were consistent across all subjects. In the analysis of mean RTs for the dual tasks, we excluded the data from those trials where one of the responses was incorrect or missed. There was a significant main effect [oneway analysis of variance (ANOVA)] of the ISI on the RT in both modalities, F(17,162) = 9.62, P < 0.001. In the post hoc analysis of RT data, we were able to show that for the long ISI dual task, the RT1 in each modality was significantly longer than the single RTs whereas the RT2 was no different from the single RTs [Tukey– Kramer Honestly Significant Difference (HSD) test for multiple comparisons of means, P < 0.05]. During the dual task condition with the short ISI, the reaction time to the second of the two stimuli (RT2) was significantly prolonged (Tukey–Kramer HSD test for multiple comparisons of means, P < 0.05) as compared with that of the dual task with the long ISI, irrespective of whether the second stimulus was somatosensory or visual (Fig. 2A). The RT2 increased linearly when the ISI decreased below 300 ms (Fig. 2C). The RT1 in the short ISI dual task condition was also significantly longer than the single RTs (Tukey–Kramer HSD test for multiple comparisons of means, P < 0.05). Thus, it could be shown that while the RT1 in all dual tasks increased significantly irrespective of the ISI, the RT2 only increased as a function of ISIs of <300 ms.

In the error analysis, we showed that there is a significant main effect (one-way ANOVA) of the ISI on the probability of subjects missing (see Materials and Methods) responses, F(9,90) = 2.47, P < 0.01, and that this is due to the dual task condition with the short ISI having significantly more misses (Tukey– Kramer HSD test for multiple comparisons of means, P < 0.05) than had the other conditions (Fig. 2B). Thus, in the dual task condition with the short ISI, there was a strong interference effect behaviorally, while both dual tasks also showed strong effects on RT1 that is not due to interference. The error rates did not correlate with increased RT2. At the post-scanning debriefing, all subjects reported that the long ISI dual task was much more difficult. The interference effect persisted even after many repetitions of psychophysical test sessions that the subjects underwent before scanning. (See Table 1 for RT and performance data.)

Brain Activity

Apart form the reaction time tasks of the scanning experiment, all subjects were also exposed to a control condition (see methods) during the scanning, in which they were stimulated with combined somatosensory and visual stimuli sequences similar to the dual tasks, but were not required to respond. Contrasting the brain activity of the RT tasks to that of the control task enabled us to dissect out the neural activity of the motor components of the RT tasks. When the BOLD signal from the single visual RT condition was contrasted with this control condition, there were widespread activations mainly in cortical motor areas and subcortical motor structures (Table 2). Similarly, in contrast to the control, the single somatosensory condition mainly activated motor regions (Table 2). In fact, to a very large extent, both the visual and the somatosensory single RT conditions engaged almost identical gray matter tissue in the cortical motor areas, which included areas 4a and 4p, the supplementary motor area (SMA) and the cingulate motor areas (CMA). The caudate nuclei, the putamina, the globus pallidus, and the anterior nuclei and the ventral lateral nuclei of the thalami (Fig. 3A,B) were also engaged by both RT tasks. The intralaminar nuclei of the thalami (Kinomura et al., 1996) too were engaged almost equiterritorially in the single visual and somatosensory RT tasks. The volume of the cortex engaged by both single RT tasks (intersection: motor structures common to both RT tasks) was 82% of the volume that was engaged by either (union: motor structures activated by one or the other RT tasks) the visual or the somatosensory RT task (Fig. 3A–D). Thus, taking the union of these two cerebral activations [(visual RT – control) ∪ (somatosensory RT – control)], we predicted that that the neuronal populations corresponding to the union (red in Fig. 3A,B, showing the union of single RT task motor activity) should also be engaged when the visual and the somatosensory RT tasks were performed concurrently. This indeed was the case, as can be seen in Figure 3C,D (red, showing the motor areas activated by the actual performance of a dual RT task). The dual task condition in which the somatosensory and visual stimuli were separated with a long ISI activated the very same cortical motor areas and subcortical motor structures (Table 3), as was predicted by the union of the cortical activation patterns of the single RT tasks of visual and somatosensory modalities.

However, there were additional regions activated in this dual task condition, as seen in Figure 3C,D (green). In order to examine whether these additional cortical activations could be attributed to dual task performance per se, we contrasted the BOLD signals of the dual task RT conditions with those of the single RT conditions (i.e. shortISIdual + longISIdual – somsingle – vissingle). This contrast revealed two bilateral cortical activations associated with the dual task performance. These activations were located in posterior superior parts of the prefrontal cortex that lines the precentral sulcus, including the frontal eye fields (Paus, 1996), and the cortex lining the middle part of the intraparietal sulcus, including the adjacent part of the supramarginal gyrus (Fig. 3E and Table 4). The increased activity in these regions did not correlate with the increased RT1 of either dual task. Similarly, these areas did not correlate with the increased RT2 or decreased performance in short ISI dual task

In order to test our hypothesis that dual task interference is associated with specific activity in cortical regions, we contrasted the BOLD signal from the dual RT task with short ISI with that of the dual task with long ISI (short ISI – long ISI). This contrast yielded a large single field of cortical activation in the posterior part of the right inferior frontal gyrus (RIFG) (Fig. 3F and Table 5). All subjects had a BOLD signal increase in this region when the dual task that produced interference was contrasted with the dual task that showed no interference. The RIFG activation was spatially distinct from the activations of the dual RT tasks – single RT tasks main effect contrast (Fig. 3E). We were able to locate part of this cluster within the cytoarchitectonically defined area 44 but not within area 45, nor within motor areas 4a and 4p (Geyer et al., 1996; Amunts et al., 1999) (see Materials and Methods). The activation volume overlapped with the part of the population map having the highest likelihood (P > 0.6) of representing area 44 (Amunts et al., 1999) of the right hemisphere (Fig. 3).

To confirm that the activity within this cluster is specifically associated with interference, we performed a further analysis. First, for each subject and for each experimental condition, we calculated the adjusted BOLD signal change relative to the perceptually matched control, taking the active RIFG field as a volume of interest (VOI). A one-sample t-test of the BOLD signal changes within this VOI, corresponding to each task, showed that the activity within the cluster is indeed increased (P < 0.01,with corrections for multiple comparisons) only for the task that produced interference, namely the short ISI dual task condition (Fig. 4AG). None of the other areas/VOIs associated with dual task main effect, namely the cortices of the precentral sulcus and the superior parietal sulcus, had this profile of specificity that was only associated with short ISI dual tasks that interfere. Next, we correlated the changes of the adjusted BOLD signal to the mean RT2 increase in each subject. This showed that the activity within the cluster is positively correlated (Pearson's r2 = 0.70, P < 0.02) with the mean increase of reaction time to the second stimulus (RT2) of the pair in the short ISI dual task (Fig. 4B). The RT2 increase in the short ISI dual task did not regress with increased BOLD signal changes representing the sensory or motor activations, or with the dual task main effect activations that we found in this study. Similarly, the BOLD signal changes within the RIFG, and the VOIs corresponding to the dual task main effect and motor activity, regressed neither with the error probabilities nor with the increased RT1 of either dual task.

Discussion

The results show that performance of dual RT tasks activates cortical regions in excess of those activated by the performance of component single tasks. The motor structures and the intralaminar thalamic nuclei engaged in the dual tasks were those predicted to be activated by the component visual and somatosensory RT tasks of the dual task. We have reported, for the first time, that dual task interference is specifically associated and correlated with increased activity in a cortical field located within the right inferior frontal gyrus.

The statistical differences of RTs between different tasks in both visual and somatosensory modalities and the error probabilities recorded while the subjects were being scanned could be replicated outside the scanner for all the subjects. Furthermore, the reaction times and error probabilities in our behavioral data conform to previous psychophysical results on dual RT task performance (Telford, 1931; Welford, 1952; Pashler, 1994).

The behavioral results showed that in either modality, the RTs for the first stimulus (RT1) were significantly longer in both dual RT task conditions than in the single RT tasks. In the single RT tasks, the average ISI was 1200 ms, which is comparable to the average inter-trial interval (~1150 ms) and the ISIs (i.e. 1150, 1200 and 1250 ms) in the dual RT tasks with long ISI. As such, it is unlikely that the RT1 increase seen in the long ISI dual task relates to the close temporal proximity of stimuli in that dual task. On the other hand, an RT2 increase was seen only in the short ISI dual task. As was mentioned, the RT2 increased linearly as the ISI was decreased below 300 ms (Fig. 2C). This illustrates that the increases of RT1 and the RT2 (i.e. interference) seen in dual RT tasks are two different psychophysical phenomena emanating from dual task situations. The RT1 increments of dual tasks indicate a process of additional synaptic activity associated with all dual task situations. This conforms to the previously shown evidence that when the attention needs to be divided between two concurrent processes there is a performance cost which is elicited as an increased RT (LaBerge, 1973; Godefroy and Rousseaux, 1996). The dual RT task differed from a single RT task in two aspects only: the necessity to divide attention between modalities and the option of having to switch between two different S–R associations. Thus, the longer RTs for the first stimulus of a pair in the dual task (RT1) are likely to be due to either one of these possibilities. The two dual tasks with long and short ISI shared the need to divide attention between modalities and motor sets, and the switching between conditional S–R associations, although there was a theoretical possibility that the attention switch between the two modalities and the motor sets may have been faster in the short ISI dual condition. In any case, the significantly longer RTs for the second stimulus (RT2) and the increased error probabilities were only seen in the short ISI dual condition. These are the hallmarks of dual task interference. The increased RT2 could signify additional synaptic activity, or a delay caused by occupancy of the neuronal populations engaged in reacting to the first stimulus, when the ISI is <300 ms.

Performance of dual RT tasks during scanning showed additional cortical areas that were significantly more active as compared to single RT tasks. This fits well with the RT1 increase that suggested additional synaptic activity. These brain activations were located bilaterally in the superior frontal cortex, including the frontal eye fields, the intraparietal sulcus (IPS) and the supramarginal gyri. Consequently, our finding of brain activity associated with dual RT task performance contradicted the view that dual task performance only extends and enhances the brain activity of the tasks that constitute the dual task, and do not recruit and sustain cortical areas in excess of those required by the component tasks (Klingberg and Roland, 1997; Goldberg et al., 1998; Klingberg, 1998; Adcock et al., 2000; Bunge et al., 2000). The finding of a main effect of additional cortical activity associated with dual task performance conforms to the previous neuroimaging evidence (D'Esposito et al., 1995; Corbetta, 1998; Koechlin et al., 1999; Dove et al., 2000) to the effect that dual tasks do in fact need additional cortical activity. The lack of additional cortical activations related to dual task performance in some of the earlier neuroimaging studies can be reconciled by the fact that the dual tasks that were neuroimaged in these studies had varying degrees of complex behavioral paradigms. The cognitively complex single tasks (Goldberg et al., 1998; Klingberg, 1998; Adcock et al., 2000; Bunge et al., 2000) in these paradigms by themselves were of such complexity that they activated the cortical gray matter areas in the parietal cortex and the frontal cortex that we have shown to be associated with dual tasks. Thus, the dual tasks in these experiments may simply have enhanced and extended the activations that would have been caused by the component tasks, had these been performed separately as single tasks.

The single visual and single somatosensory RT tasks engaged almost identical cortical gray matter tissue in the cortical motor regions of 4a, 4p, SMA, CMA, the basal ganglia and the parts of the thalamus receiving the output from the activated parts of the basal ganglia. This is in close agreement with the hypothesis that interference between concurrently performed tasks is due to the engagement of the same motor systems (Passingham, 1996; Roland and Zilles, 1998). Previously, brain activities associated with RT tasks of different modalities were found to converge on these motor regions (Naito et al., 2000). In our study, motor structures including the 4a, 4p, CMA, SMA, basal ganglia, and the ventral lateral and the ventral anterior nuclei of thalami were predicted to be active, and subsequently found to be active, in the coding and the execution of index finger flexion in response to the somatosensory stimuli and of long finger flexion in response to visual stimuli in the dual RT tasks. Only the dual task condition in which the ISI were <300 ms produced interference. In this condition, the RT2 increased linearly as the ISI was progressively shortened below 300 ms. The most parsimonious physiological interpretation would be that it would take some time for the motor structures to change from the coding and control of the index finger flexion to the long finger flexion and vice versa. If the large common engagement of the motor areas 4a, 4p, SMA, CMA and the subcortical motor structures (Fig. 3A–D) imply that ~80% of the neural populations engaged in the visual RT and the somatosensory RT of the dual task are identical, a possibility that is supported in the literature (Lamarre et al., 1983; Salinas and Romo, 1998; Naito et al., 2000), this could be one of the causes of the interference. The linear increase of the RT2 as the ISI decreased could signify that the neuronal population(s) at one or more of these motor structures needed an approximately fixed time interval to change the coding of a response to a visual stimulus to coding the response to a somatosensory stimulus (Fu et al., 1995). Such occupancy could lead to the recruitment of additional neuronal populations in situations where interference tends to occur. This type of interference would be in accordance with the predictions of the cortical field hypothesis (Roland, 1993; Roland and Zilles, 1998). The consequence of the interference mechanism, of whatever nature it is, as shown by the behavioral results, is that, at least for RT tasks with short cross-modal ISIs, the brain either misses the response to the second stimulus or makes a delayed response.

We found a cortical region located in the right inferior frontal gyrus that was specifically associated with interference. The activity within this cortical field is highly correlated (Fig. 4B) with the increased RT for the second stimulus. This is as if the brain needs to recruit the RIFG cortical field when dual tasks interfere, when interference (i.e. increased RT2) occurs due to the occupancy of the same neuronal populations by the computation of the RT task 1. This appears to happen only when the ISI is <300 ms. The temporal certainty of the stimuli (Georgopoulos et al., 1981; Lecas et al., 1986), i.e. that one stimulus is immediately followed by another of a different modality, albeit with a random ISI, is much more deterministic in the short ISI condition than in the long ISI condition. The responses required by the dual tasks are conditional in that the subjects had to choose a specific arbitrary response for the visual and somatosensory stimuli. The conditional associations were similar in both dual tasks, and had very high S–R compatibility, which is known to reduce dual task interference (Georgopoulos et al., 1981). Thus, the conditional associations of responses in the dual tasks cannot account for the increased activity in the RIFG cortical field.

One of the possibilities for the increased RIFG activity in the short ISI dual task is that it may be related to what has been loosely termed the ‘task difficulty’. The objective measure of task difficulty is to document increased RTs and/or decreased accuracy of the performance (Grady et al., 1996; Barch et al., 1997; Winstein et al., 1997). However, a number of reasons disqualify task difficulty as the cause of increased RIFG activity in our experiment. Firstly, all subjects reported that the long ISI dual task was more difficult than the short ISI dual task. Secondly, if dual RT tasks are more difficult than single RT tasks, both RT1 and RT2 should be prolonged. This was not the case. Furthermore, the long ISI dual task did not produce error rates that were significantly higher than those of the single task conditions whereas the short ISI dual task did. We have thus demonstrated that the long and the short ISI affect RT1 and RT2 differentially. Moreover, we have shown that each of these psychophysically distinct conditions was associated with different sets of cortical activations. Meanwhile, some authors have found increased activity of the prefrontal cortex (PFC) associated with the increased task difficulty (Grady et al., 1996; Barch et al., 1997; Sunaert et al., 2000), while others have not found any increased activity (D'Esposito et al., 1995; Winstein et al., 1997; Paus et al., 1998) or have found decreased activity (Goldberg et al., 1998) in the prefrontal cortex with regard to task difficulty. D'Esposito (D'Esposito, 1995) in particular reported that the increased PFC activity is not associated with task difficulty in dual tasks. In the studies that found increased activity in the PFC as a main effect of task difficulty (Grady et al., 1996; Barch et al., 1997; Sunaert et al., 2000), the centers of masses of their activations lay at least 18 or 13 mm (Table 6) away from that of the RIFG. Although the center of mass in our study is clearly localized to the cytoarchitectonically defined area 44, we cannot exclude that some of this activity also engaged the ventral premotor area as well (Fig. 3F,G).

The neural correlates error detection and performance monitoring that could be important in dual task performance are currently localized mostly to the cingulate cortex (Badgaiyan and Posner, 1998; Carter et al., 1998,1999; Bush et al., 2000; Kiehl et al., 2000) and are unlikely to contribute to the increased RIFG activity in our experiment.

Along with the views of the human brain as a massive parallel processor, presently there is a widespread interest in the possibility that synchrony of neural firing could serve as a code for linking features common to a given behavior (Crick, 1984; Van Essen et al., 1995; Singer, 1999) and that such temporal synchrony between groups of neurons will be an adequate mechanism to produce independent coherent states of cognition. However, if temporal synchrony between groups of neurons can lead to coherent behaviors, such synchrony ought to occur in parallel and not be affected by concurrent processes.

The psychophysical finding of performance costs associated with division of attention and the dual task interference has long been considered a serious limitation of the parallel processing capability of the brain. What neurobiological advantages such limitations confer on the organism remain unknown. The findings that the performance of dual RT tasks engage the gray matter of motor structures that are shared between the component tasks, and that there is at least one cortical region that is specifically associated with interference, asserts and strengthens such fundamental questions about parallel processing as a general-purpose mechanism in the human brain on a more physiological framework. This is because, if the brain were truly a parallel processor, the shared cortical fields should not lead to interference, nor should there be interference-specific cortical activity. Consequently, we see that the findings of this study have raised the need to reconcile the notion of computational parallel processing with the dual task interference phenomenon.

Notes

This research was supported by a grant from the VOLVO Foundation of Sweden.

Address correspondence to Priyantha Herath, Division of Human Brain Research, Department of Neuroscience, Karolinska Institute, Stockholm 171 77, Sweden. Email: priyantha.herath@neuro.ki.se.

Table 1

Mean RT and error rates from the scanning experiments

 som som1_long som2_long som1_short som2_short vis vis1_long vis2_long vis1_short vis2_short 
som1_*, som2_*, vis1_*, vis2_* indicate the data for somatosensory and visual components within the dual task trials when they appeared as the first or the second stimulus. vis and som stand for the respective single RT conditions. 
Mean RT ± SD 245.1 ± 22.8 348.4 ± 37.6 294.0 ± 34.7 385.27 ± 2.0 426.9 ± 83.7 241.9 ± 26.3 346.4 ± 31.5 244.7 ± 28.9 383.8 ± 60.5 380.4 ± 89.9 
% error rates ± SD  5 ± 1  2 ± 1  6 ± 2  7 ± 2  9 ± 1  4 ± 1  4 ± 1  4 ± 1  7 ± 2  9 ± 1 
 som som1_long som2_long som1_short som2_short vis vis1_long vis2_long vis1_short vis2_short 
som1_*, som2_*, vis1_*, vis2_* indicate the data for somatosensory and visual components within the dual task trials when they appeared as the first or the second stimulus. vis and som stand for the respective single RT conditions. 
Mean RT ± SD 245.1 ± 22.8 348.4 ± 37.6 294.0 ± 34.7 385.27 ± 2.0 426.9 ± 83.7 241.9 ± 26.3 346.4 ± 31.5 244.7 ± 28.9 383.8 ± 60.5 380.4 ± 89.9 
% error rates ± SD  5 ± 1  2 ± 1  6 ± 2  7 ± 2  9 ± 1  4 ± 1  4 ± 1  4 ± 1  7 ± 2  9 ± 1 
Table 2

Brain regions associated with single RT task performance in visual and somatosensory modalities.

Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
bPrimary motor cortex, specifically, area 4a and 4p. 
cPostcentral gyrus. 
Visual – control 26.384 (0.0001)a –32 –28  58 5.19 L M1b 
   –4  –6  46 4.97 L SMA/CMA 
  –34 –30 –48 4.55 L postCGc 
 14.648 (0.0001)a –30 4.53 L putamen 
  –14 –12 4.45 L VA thalamus 
  –26  –4 4.42 L putamen 
 2.456 (0.0001)a –38 –30 4.48 L transverse temporal gyrus 
  –50 –30  10 3.73 L superior temporal gyrus 
 9.360 (0.0001)a  26 4.48 R globus pallidus 
   28 4.42 R putamen 
   16  –2 4.32 R caudate nucleus 
Somatosensory – control 55.464 (0.0001)a –38 –22  52 4.92 L M1b 
   –2 –22  46 4.73 L SMA/ CMA 
  –42 –26  34 4.58 L post CGc 
 2.400 (0.01)a  56 –34  30 3.85 R supramarginal gyrus 
Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
bPrimary motor cortex, specifically, area 4a and 4p. 
cPostcentral gyrus. 
Visual – control 26.384 (0.0001)a –32 –28  58 5.19 L M1b 
   –4  –6  46 4.97 L SMA/CMA 
  –34 –30 –48 4.55 L postCGc 
 14.648 (0.0001)a –30 4.53 L putamen 
  –14 –12 4.45 L VA thalamus 
  –26  –4 4.42 L putamen 
 2.456 (0.0001)a –38 –30 4.48 L transverse temporal gyrus 
  –50 –30  10 3.73 L superior temporal gyrus 
 9.360 (0.0001)a  26 4.48 R globus pallidus 
   28 4.42 R putamen 
   16  –2 4.32 R caudate nucleus 
Somatosensory – control 55.464 (0.0001)a –38 –22  52 4.92 L M1b 
   –2 –22  46 4.73 L SMA/ CMA 
  –42 –26  34 4.58 L post CGc 
 2.400 (0.01)a  56 –34  30 3.85 R supramarginal gyrus 
Table 3

Brain regions activated by dual tasks with and without interference compared with the perceptual match

Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z core Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
bThese large clusters included the primary motor cortex, specifically, areas 4a and 4p. 
cWithin a trial, ISIs were 200, 250 and 300 ms. 
dWithin a trial, ISIs were 1150, 1200 and 1250 ms. 
Short ISIc – control 11.05 (0.0001)a –32 –26 54 5.34 L M1b 
  7.08 (0.0001)a –26  –4 –4 5.31 L putamen 
  –26 –10 12 5.08 L globus pallidus 
  –14  –4 –4 4.19 L ventral anterior thalamus 
  8.72 (0.0001)a  22 –10 12 4.86 R globus Pallidus 
  0.632 (0.002)a  52 –36 30 4.29 R superior temporal gyrus 
   46 –34 36 4.25 R supramarginal gyrus 
  0.552 (0.005)a –28 –24 –4 4.20 L ventral posterior thalamus 
  0.800 (0.001)a  –4 58 4.05 L SMA/CMA 
Long ISId – control 13.184 (0.0001)a –36 –28 50 5.69 L M1b 
  –46 –30 48 5.18 L supramarginal gyrus 
  0.376 (0.0001)a –10 –14 4.32 L ventral anterior thalamus 
  –28 4.26 L putamen 
  0.664 (0.001)a  28  12 4.09 R putamen 
Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z core Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
bThese large clusters included the primary motor cortex, specifically, areas 4a and 4p. 
cWithin a trial, ISIs were 200, 250 and 300 ms. 
dWithin a trial, ISIs were 1150, 1200 and 1250 ms. 
Short ISIc – control 11.05 (0.0001)a –32 –26 54 5.34 L M1b 
  7.08 (0.0001)a –26  –4 –4 5.31 L putamen 
  –26 –10 12 5.08 L globus pallidus 
  –14  –4 –4 4.19 L ventral anterior thalamus 
  8.72 (0.0001)a  22 –10 12 4.86 R globus Pallidus 
  0.632 (0.002)a  52 –36 30 4.29 R superior temporal gyrus 
   46 –34 36 4.25 R supramarginal gyrus 
  0.552 (0.005)a –28 –24 –4 4.20 L ventral posterior thalamus 
  0.800 (0.001)a  –4 58 4.05 L SMA/CMA 
Long ISId – control 13.184 (0.0001)a –36 –28 50 5.69 L M1b 
  –46 –30 48 5.18 L supramarginal gyrus 
  0.376 (0.0001)a –10 –14 4.32 L ventral anterior thalamus 
  –28 4.26 L putamen 
  0.664 (0.001)a  28  12 4.09 R putamen 
Table 4

Cortical regions associated with the main effect of dual task performance given by shortISIdual + longISIdual – somsingle – vissingle

Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
Dual tasks – single tasks 5.352 (0.0001)a –34 –52 44 5.04 L intraparietal sulcus 
  –30 –46 34 4.41 L intraparietal sulcus 
  –26 –58 48 3.66 L intra parietal sulcus 
 2.976 (0.0001)a –32  –8 46 4.06 L pre central sulcus 
  –20 –22 44 3.86 L precentral sulcus 
  –24 –14 50 3.78 L precentral sulcus 
 8.696 (0.0001)a  36  –2 30 4.42 R precentral sulcus 
   34 40 4.18 R precentral sulcus 
 2.584 (0.0001)a  48 –38 30 4.41 R supramarginal gyrus 
Comparison Cluster size in cm3 at (omnibus P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
Dual tasks – single tasks 5.352 (0.0001)a –34 –52 44 5.04 L intraparietal sulcus 
  –30 –46 34 4.41 L intraparietal sulcus 
  –26 –58 48 3.66 L intra parietal sulcus 
 2.976 (0.0001)a –32  –8 46 4.06 L pre central sulcus 
  –20 –22 44 3.86 L precentral sulcus 
  –24 –14 50 3.78 L precentral sulcus 
 8.696 (0.0001)a  36  –2 30 4.42 R precentral sulcus 
   34 40 4.18 R precentral sulcus 
 2.584 (0.0001)a  48 –38 30 4.41 R supramarginal gyrus 
Table 5

Brain regions specifically associated with dual task interference

Comparison Cluster size in cm3 at omnibus (P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
Short ISI – long ISI 2.75 (0.01)a 46 26 3.85 R inferior frontal gyrus 
  52 28 3.66 R inferior frontal gyrus 
Comparison Cluster size in cm3 at omnibus (P < 0.05) Coordinates of the peaks within the cluster Voxel level Z score Anatomical location of the peak activity 
Z scores correspond to voxels of peak activity whose locations are given as stereotaxic coordinates referring to the atlas of Talairach and Tournoux. 
aThe actual omnibus significance level for the cluster. 
Short ISI – long ISI 2.75 (0.01)a 46 26 3.85 R inferior frontal gyrus 
  52 28 3.66 R inferior frontal gyrus 
Table 6

Comparison between activations previously associated with task difficulty and the activations specific to interference.

Study Coordinates of prefrontal peak activity attributed to task difficulty (main effects or interactions) Distance (mm) from the peak activity associated with interference (i.e. from 46, 6, 26) 
 X Y Z  
(Grady et al.,1996) 26 42 32 41 
(Barch et al.,1997) 48 19 23 13.5 
 50 19 27 
(Sunaert et al.,2000) 45 21 15 18.6 
Study Coordinates of prefrontal peak activity attributed to task difficulty (main effects or interactions) Distance (mm) from the peak activity associated with interference (i.e. from 46, 6, 26) 
 X Y Z  
(Grady et al.,1996) 26 42 32 41 
(Barch et al.,1997) 48 19 23 13.5 
 50 19 27 
(Sunaert et al.,2000) 45 21 15 18.6 
Figure 1.

 Scheme showing the organization of single and dual RT tasks in the behavioral paradigm: S the stimulus of the single task, R the response of single task; S1, S2, R1, R2, the first and the second stimuli and the responses respectively of the dual RT tasks, separated by the ISI. In the dual task with the short ISI, the ISIs were 200, 250 and 300 ms. In the dual task with the long ISI, the ISIs were 1150, 1200 and 1250 ms. RT1 and RT2 are the RTs for the first and the second stimuli of the dual task.

Figure 1.

 Scheme showing the organization of single and dual RT tasks in the behavioral paradigm: S the stimulus of the single task, R the response of single task; S1, S2, R1, R2, the first and the second stimuli and the responses respectively of the dual RT tasks, separated by the ISI. In the dual task with the short ISI, the ISIs were 200, 250 and 300 ms. In the dual task with the long ISI, the ISIs were 1150, 1200 and 1250 ms. RT1 and RT2 are the RTs for the first and the second stimuli of the dual task.

Figure 2.

 Behavioral results. (A) The mean RTs and SDs for the somatosensory and visual modalities of different task situations. 1_long and 1_short indicate the RTs for the first stimulus in the dual task with long and short ISI respectively; 2_long and 2_short indicate the RTs for second stimulus. (i) The RTs for the first stimulus of both modalities in the long ISI dual task were significantly longer (P < 0.05, Tukey–Kramer HSD test for multiple comparisons) than the single RTs of the respective modalities. (ii) There were no significant differences (P > 0.05, Tukey–Kramer HSD) between the single RTs and the RT to the second stimulus in the long ISI dual task condition in either visual or somatosensory modality. (iii) RTs for both the first and the second stimulus in the short ISI dual task were significantly longer (P < 0.05, Tukey–Kramer HSD) than the single task RTs of both modalities. (B) The percentage of errors (% hits) as a function of different task conditions. Error bars are the SDs. The probabilities of producing errors were significantly higher (P < 0.05, Tukey–Kramer HSD) for both modalities in the short ISI dual task condition. Within either modality 96% of all errors (equally distributed across the subjects) were true misses. (C) The linear increase of RTs (and SDs) for the second stimulus (RT2) of the short ISI dual task as compared with the RT2 of the long ISI dual task. Error bars show SDs. The RT increments for the second stimulus in short ISI condition were statistically significant (P < 0.03, one-sample t-test) at the ISIs (200, 250, 300 ms) that we tested for.

Figure 2.

 Behavioral results. (A) The mean RTs and SDs for the somatosensory and visual modalities of different task situations. 1_long and 1_short indicate the RTs for the first stimulus in the dual task with long and short ISI respectively; 2_long and 2_short indicate the RTs for second stimulus. (i) The RTs for the first stimulus of both modalities in the long ISI dual task were significantly longer (P < 0.05, Tukey–Kramer HSD test for multiple comparisons) than the single RTs of the respective modalities. (ii) There were no significant differences (P > 0.05, Tukey–Kramer HSD) between the single RTs and the RT to the second stimulus in the long ISI dual task condition in either visual or somatosensory modality. (iii) RTs for both the first and the second stimulus in the short ISI dual task were significantly longer (P < 0.05, Tukey–Kramer HSD) than the single task RTs of both modalities. (B) The percentage of errors (% hits) as a function of different task conditions. Error bars are the SDs. The probabilities of producing errors were significantly higher (P < 0.05, Tukey–Kramer HSD) for both modalities in the short ISI dual task condition. Within either modality 96% of all errors (equally distributed across the subjects) were true misses. (C) The linear increase of RTs (and SDs) for the second stimulus (RT2) of the short ISI dual task as compared with the RT2 of the long ISI dual task. Error bars show SDs. The RT increments for the second stimulus in short ISI condition were statistically significant (P < 0.03, one-sample t-test) at the ISIs (200, 250, 300 ms) that we tested for.

Figure 3.

 Predicted and actually activated cortical regions in the dual RT tasks. (A,B) The regions that we predicted would be concurrently activated in dual task situations. The prediction was obtained from a union of the orthogonalized activation maps of each of the single tasks contrasted with the perceptual control, i.e. (visual RT – control) ∪ (somatosensory RT– control). The results are displayed by overlaying the active regions of each of these contrasts on the mean anatomical MR image. Orange represents the activations specific to somatosensory single RT task; blue, the activations of the visual single RT task; and red, the regions that are common to both these RT tasks (intersection). (C,D) The cortical activations representing the actual performance of a dual task, i.e. that in the dual RT task with long ISI. Orange represents the union of cortical areas that are active in one or the other single RT tasks; green, the areas exclusively activated by the dual RT task; and red, the areas predicted by the intersection of activations in the single RT tasks. Activation maps shown here are the results of random effects analysis for each contrasts. Prior to union or intersection, each of them were thresholded such that the resulting t-maps had a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity. Each of the clusters shown here has an omnibus P < 0.05. They are displayed by overlaying on the mean anatomical MR image at Talairach coordinates Z = 11 and Z = 49. The left hemisphere is displayed to the right. (E) Cortical regions that were significantly more active in dual task performance than in the single task performance, i.e. the contrast shortISIdual + longISIdual – somsingle – vissingle. As shown, dual task performance activated a set of fronto-parietal regions bilaterally. The resulting t-maps were thresholded at a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity. Each of the clusters shown here had an omnibus P < 0.05. They are displayed by overlaying on the mean anatomical MR image at Talairach coordinates Z = 49. The color scale is arbitrary, where yellow indicates a higher Z score. The left hemisphere is displayed on the right. Cortical activation specific to interference: (F) shows the right inferior frontal gyrus (RIFG) activation resulting from short ISI dual task – long ISI dual task contrast. The cluster is the result of a random effects analysis. It was thresholded at a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity and spatial extent >200 voxels, yielding an omnibus P < 0.02 for the whole cluster and is displayed by overlaying on the mean anatomical MR image. (G) The contour (purple) of the RIFG activation overlaid on the cytoarchitectonically defined population map of area 44 of the right hemisphere on the standard brain of the HBA (Talairach coordinate X = 44). The population map indicates the number of post-mortem brains having area 44 represented in each voxel. Red in the arbitrary color scale (which ranged from blue to red) here represents the highest likelihood (P > 0.60) of the voxels representing area 44. Dark blue represents the voxels belonging to only one of the 10 post-mortem brains (P = 0.10). The RIFG activation overlapped ~20% of the cytoarchitectonically defined area 44.

Figure 3.

 Predicted and actually activated cortical regions in the dual RT tasks. (A,B) The regions that we predicted would be concurrently activated in dual task situations. The prediction was obtained from a union of the orthogonalized activation maps of each of the single tasks contrasted with the perceptual control, i.e. (visual RT – control) ∪ (somatosensory RT– control). The results are displayed by overlaying the active regions of each of these contrasts on the mean anatomical MR image. Orange represents the activations specific to somatosensory single RT task; blue, the activations of the visual single RT task; and red, the regions that are common to both these RT tasks (intersection). (C,D) The cortical activations representing the actual performance of a dual task, i.e. that in the dual RT task with long ISI. Orange represents the union of cortical areas that are active in one or the other single RT tasks; green, the areas exclusively activated by the dual RT task; and red, the areas predicted by the intersection of activations in the single RT tasks. Activation maps shown here are the results of random effects analysis for each contrasts. Prior to union or intersection, each of them were thresholded such that the resulting t-maps had a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity. Each of the clusters shown here has an omnibus P < 0.05. They are displayed by overlaying on the mean anatomical MR image at Talairach coordinates Z = 11 and Z = 49. The left hemisphere is displayed to the right. (E) Cortical regions that were significantly more active in dual task performance than in the single task performance, i.e. the contrast shortISIdual + longISIdual – somsingle – vissingle. As shown, dual task performance activated a set of fronto-parietal regions bilaterally. The resulting t-maps were thresholded at a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity. Each of the clusters shown here had an omnibus P < 0.05. They are displayed by overlaying on the mean anatomical MR image at Talairach coordinates Z = 49. The color scale is arbitrary, where yellow indicates a higher Z score. The left hemisphere is displayed on the right. Cortical activation specific to interference: (F) shows the right inferior frontal gyrus (RIFG) activation resulting from short ISI dual task – long ISI dual task contrast. The cluster is the result of a random effects analysis. It was thresholded at a Z score of >3.25 (P < 0.0005 uncorrected) for the peak intensity and spatial extent >200 voxels, yielding an omnibus P < 0.02 for the whole cluster and is displayed by overlaying on the mean anatomical MR image. (G) The contour (purple) of the RIFG activation overlaid on the cytoarchitectonically defined population map of area 44 of the right hemisphere on the standard brain of the HBA (Talairach coordinate X = 44). The population map indicates the number of post-mortem brains having area 44 represented in each voxel. Red in the arbitrary color scale (which ranged from blue to red) here represents the highest likelihood (P > 0.60) of the voxels representing area 44. Dark blue represents the voxels belonging to only one of the 10 post-mortem brains (P = 0.10). The RIFG activation overlapped ~20% of the cytoarchitectonically defined area 44.

Figure 4.

 The graphs show the specificity (A) of activity within the RIFG cluster for the short ISI dual task that produced interference and the correlation (B) of this specific activity with the increased mean RTs for the second stimulus of a pair of signal in this task situation.

Figure 4.

 The graphs show the specificity (A) of activity within the RIFG cluster for the short ISI dual task that produced interference and the correlation (B) of this specific activity with the increased mean RTs for the second stimulus of a pair of signal in this task situation.

References

Adcock RA, Constable RT, Gore JC, Goldman-Rakic PS (
2000
) Functional neuroanatomy of executive processes involved in dual-task performance.
Proc Natl Acad Sci USA
 
28
:
3567
–3572.
Amunts K, Schleicher A, Burgel U, Mohlberg H, Uylings HB, Zilles K (
1999
) Broca's region revisited: cytoarchitecture and intersubject variability.
J Comp Neurol
 
20
:
319
–341.
Badgaiyan RD, Posner MI (
1998
) Mapping the cingulate cortex in response selection and monitoring.
NeuroImage
 
7
:
255
–260.
Barch DM, Braver TS, Nyström LE, Forman SD, Noll DC, Cohen JD (
1997
) Dissociating working memory from task difficulty in human prefrontal cortex.
Neuropsychologia
 
35
:
1373
–1380.
Braun J (1998) Divided attention: narrowing the gap between brain and behavior. In: The attentive brain (Parasuraman R, ed.). Cambridge, MA: MIT Press.
Bunge SA, Klingberg T, Jacobsen RB, Gabrieli JD (
2000
) A resource model of the neural basis of executive working memory.
Proc Natl Acad Sci USA
 
28
:
3573
–3578.
Bush G, Luu P, Posner MI (
2000
) Cognitive and emotional influences in anterior cingulate cortex.
Trends Cogn Sci
 
4
:
215
–222.
Carter CS, Braver TS, Barch DM, Botvinick MM, Noll D, Cohen JD (
1998
) Anterior cingulate cortex, error detection, and the online monitoring of performance.
Science
 
280
:
747
–749.
Carter CS, Botvinick MM, Cohen JD (
1999
) The contribution of the anterior cingulate cortex to executive processes in cognition.
Rev Neurosci
 
10
:
49
–57.
Corbetta M (
1998
) Frontoparietal cortical networks for directing attention and the eye to visual locations: identical, independent, or overlapping neural systems?
Proc Natl Acad Sci USA
 
3
:
831
–838.
Crick F (
1984
) Function of the thalamic reticular complex: the searchlight hypothesis.
Proc Natl Acad Sci USA
 
81
:
4586
–4590.
D'Esposito M, Detre JA, Alsop DC, Shin RK, Atlas S, Grossman M (
1995
) The neural basis of the central executive system of working memory.
Nature
 
378
:
279
–281.
Dove A, Pollmann S, Schubert T, Wiggins CJ, Cramon DYv (
2000
) Prefrontal cortex activation in task switching: an event-related fMRI study.
Brain Res Cogn Brain Res
 
9
:
103
–109.
Friston KJ, Holmes AP, Price CJ, Büchel C, Worsley KJ (
1999
) Multisubject fMRI studies and conjunction analyses.
NeuroImage
 
10
:
385
–396.
Fu QG, Flament D, Ebner TJ (
1995
) Temporal encoding of movement kinematics in the discharge of primate primary motor and premotor neurons.
J Neurophysiol
 
73
:
836
–854.
Georgopoulos AP, Kalaska JF, Massey JT (
1981
) Spatial trajectories and reaction times of aimed movements: effects of practice, uncertainty, and change in target location.
J Neurophysiol
 
146
:
725
–743.
Gescheider GA (1997) Psychophysics: the fundamentals. Hillsdale, NJ: Lawrence Erlbaum Associates.
Geyer S, Ledberg A, Schleicher A, Kinomura S, Schormann T, Burgel U, Klingberg T, Larsson J, Zilles K, Roland PE (
1996
) Two different areas within the primary motor cortex of man.
Nature
 
382
:
805
–807.
Godefroy O, Rousseaux M (
1996
) Divided and focused attention in patients with lesion of the prefrontal cortex.
Brain Cogn
 
30
:
155
–174.
Goldberg TE, Berman KF, K F, Ostrem J, Horn JDV, Esposito G, Mattay VS, MGold J, Weinberger DR (
1998
) Uncoupling cognitive workload and prefrontal cortical physiology: a PET rCBF study.
NeuroImage
 
7
:
296
–303.
Grady CL, Horwitz B, Pietrini P, Mentis MJ, Ungerleider LG, Rapoport SI, Haxby JV (
1996
) Effect of task difficulty on cerebral blood flow during perceptual matching of faces.
Hum Brain Mapp
 
4
:
227
–239.
Kiehl KA, Liddle PF, Hopfinger JB (
2000
) Error processing and the rostral anterior cingulate: an event-related fMRI study.
Psychophysiology
 
37
:
216
–223.
Kinomura S, Larsson J, Gulyas B, Roland PE (
1996
) Activation by attention of the human reticular formation and thalamic intralaminar nuclei.
Science
 
271
:
512
–515.
Klingberg T (
1998
) Concurrent performance of two working memory tasks: potential mechanisms of interference.
Cereb Cortex
 
8
:
593
–601.
Klingberg T, Roland PE (
1997
) Interference between two concurrent tasks is associated with activation of overlapping fields in the cortex.
Cogn Brain Res
 
6
:
1
–8.
Koechlin E, Basso G, Pietrini P, Panzer S, Grafman J (
1999
) The role of the anterior prefrontal cortex in human cognition.
Nature
 
399
:
148
–151.
LaBerge D (1973) Identification of two components of the time to switch attention: a test of a serial and a parallel model of attention. In: Attention and Performance IV (Kornblum S, ed.), pp. 71–85. New York: Academic Press.
Lamarre Y, Busby L, Spidalieri G (
1983
) Fast ballistic arm movements triggered by visual, auditory, and somesthetic stimuli in the monkey. I. Activity of precentral cortical neurons.
J Neurophysiol
 
50
:
1343
–1358.
Lecas JC, Requin J, Vitton N (
1986
) Changes in neuronal activity of the monkey precentral cortex during preparation for movement.
J Neurophysiol
 
56
:
1680
–702.
Meyer DE, Kieras DE (
1997
) A computational theory of executive cognitive processes and multiple-task performance: Part 1. Basic mechanisms.
Psychol Rev
 
104
:
3
–65.
Naito E, Kinomura S, Geyer S, Kawashima R, Roland PE, Zilles K (
2000
) Fast reaction to different sensory modalities activates common fields in the motor areas, but the anterior cingulate cortex is involved in the speed of reaction.
J Neurophysiol
 
83
:
1701
–1709.
Pashler H (
1994
) Dual-task in simple tasks: data and theory.
Psychol Bull
 
2
:
220
–244.
Pashler HE (1999) The psychology of attention. Cambridge, MA: MIT Press.
Passingham RE (
1996
) Attention to action.
Phil Trans R Soc Lond B
 
351
:
1473
–1479.
Paus T (
1996
) Location and function of the human frontal eye-field: a selective review.
Neuropsychologia
 
34
:
475
–483.
Paus T, Koski L, Caramanos Z, Westbury C (
1998
) Regional differences in the effects of task difficulty and motor output on blood flow response in the human anterior cingulate cortex: a review of 107 PET activation studies.
NeuroReport
 
9
:
R37
–47.
Roland PE (1993) Brain activation. New York: Wiley-Liss.
Roland PE, Graufelds C, Wåhlin J, Ingelman L, Andersson M, Ledberg A, Pedersen J, Åkerman S, Dabringhaus A, Zilles K (
1994
) Human brain atlas: for high-resolution functional and anatomical mapping.
Hum Brain Mapp
 
2
:
1
–12.
Roland PE, Zilles K (
1998
) Structural divisions and functional fields in the human cerebral cortex.
Brain Res Brain Res Rev
 
26
:
87
–105.
Salinas E, Romo R (
1998
) Conversion of sensory signals into motor commands in primary motor cortex.
J Neurosci
 
18
:
499
–511.
Schleicher A, Amunts K, Geyer S, Morosan P, Zilles K (
1999
) Observerindependent method for microstructural parcellation of cerebral cortex: a quantitative approach to cytoarchitectonics.
NeuroImage
 
9
:
165
–177.
Schormann T, Zilles K (
1998
) Three-dimensional linear and nonlinear transformations: an integration of light microscopical and MRI data.
Hum Brain Mapp
 
6
:
339
–347.
Singer W (
1999
) Time as coding space?
Curr Opin Neurobiol
 
9
:
189
–194.
Sunaert S, Van Hecke P, Marchal G, Orban GA (
2000
) Attention to speed of motion, speed discrimination, and task difficulty: an fMRI study.
NeuroImage
 
11
:
612
–623.
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the human brain. New York: Thieme.
Telford CW (
1931
) The refractory phase of voluntary and associative responses.
J Exp Psychol
 
XIV
:
1
–36.
Van Essen DC, Anderson CH, Olshausen BA (1995) Dynamic routing strategies in sensory, motor and cognitive processing. In: Large scale neuronal theories of the brain (Koch C, Davis JL, eds). Cambridge, MA: MIT Press.
Welford AT (
1952
) The ‘psychological refractory period’ and the timing of high-speed performance — a review and theory.
Br J Psychol
 
43
:
2
–19.
Winstein CJ, Grafton ST, Pohl PS (
1997
) Motor task difficulty and brain activity: investigation of goal-directed reciprocal aiming using positron emission tomography.
J Neurophysiol
 
77
:
1581
–1594.
Worsley K (
1995
) Testing for a signal with unknown location and scale in a stationary Gaussian random field.
Ann Statist
 
23
:
608
–639.