Abstract

The role of the prefrontal cortex remains controversial. Neuroimaging studies support modality-specific and process-specific functions related to working memory and attention. Its role may also be defined by changes in its influence over other brain regions including sensory and motor cortex. We used functional magnetic imaging (fMRI) to study the free selection of actions and colours. Control conditions used externally specified actions and colours. The prefrontal cortex was activated during free selection, regardless of modality, in contrast to modality-specific activations outside prefrontal cortex. Structural equation modelling (SEM) of fMRI data was used to test the hypothesis that although the same regions of prefrontal cortex may be active in tasks within different domains, there is task-dependent effective connectivity between prefrontal cortex and non-prefrontal cortex. The SEM included high-order interactions between modality, selection and regional activity. There was greater coupling between prefrontal cortex and motor cortex during free selection and action tasks, and between prefrontal cortex and visual cortex during free selection of colours. The results suggest that the functions of the prefrontal cortex may be defined not only by selection-specific rather than modality-specific processes, but also by changing patterns of effective connectivity from prefrontal cortex to motor and sensory cortices.

Introduction

Our understanding of the role of the prefrontal cortex depends upon two complementary approaches to characterize brain function. The first is to define regional specialization for functions such as working memory and executive processes (Goldman-Rakic, 1987a; Smith and Jonides, 1999; D'Esposito et al., 2000; Petrides, 2000). Recent models have integrated mnemonic and executive functions (Duncan, 2001; Miller and Cohen, 2001) because individual neurons change their mnemonic function according to context and encode abstract rules (Rao et al., 1997; Asaad et al., 1998, 2000). The second approach is to characterize interactions within distributed neural networks. The function of a region is limited by its intrinsic and extrinsic anatomical connectivity (Passingham et al., 2002) and the changing efficacy of these connections as a function of the behavioural state (McIntosh and Gonzalez-Lima, 1994; McIntosh et al., 1994; Horwitz et al., 1999).

A specialized role for the dorsolateral prefrontal cortex has been proposed in which it biases neuronal excitation in other areas, leading to the selection of specific neuronal representations of sensory stimuli, memories or actions (Miller, 1999, 2000). For example, the attentional modulation of extrastriate cortical activity depends on the integrity of prefrontal cortex (Barcelo et al., 2000) and attention modulates topographic activation within visual cortex (Brefczynski and DeYoe, 1999) and auditory cortex (Knight et al., 1981; Grady et al., 1997; Alho et al., 1999). Moreover, prefrontal cortex may activate a specific non-prefrontal neuronal assembly even in the absence of a corresponding stimulus (Kastner et al., 1999; Tomita et al., 1999). Competitive bias may also occur between specific neuronal representations. This has been demonstrated for visual objects' representations (Chelazzi et al., 1993, 1998; Desimone and Duncan, 1995; Desimone, 1999), under presumed prefrontal or parietal influences. Similar mechanisms may be the basis of the selection of neural representations of words (Frith, 2000), items from memory (Rowe et al., 2000, 2001) or actions (Deiber et al., 1991; Frith et al., 1991). In the motor domain, we propose that it is the neuronal representations of the abstract goals or targets of movement that are selected, and neuronal encoding of this type of representation has been identified in motor and premotor cortex (Alexander and Crutcher, 1990; Shen and Alexander, 1997; Kakei et al., 1999; Graziano et al., 2002).

We used fMRI to study the selection processes mediated by prefrontal cortex, during selection of actions and colours. Selection was either by the subjects' own choice or specified to them. We compared free selection with externally specified responses, because free selection exemplifies ‘top-down’ influences distinct from ‘bottom-up’ stimulus-driven processes or arbitrary stimulus–response associations. We used statistical parametric mapping (SPM) to identify modality-specific and modality-independent regions of activation associated with free selection. Our first hypothesis was that free selection (choice) of responses would be associated with activation of the same regions of dorsolateral prefrontal cortex regardless of task modality.

Within prefrontal cortex, there may be regions that cause the selection of remote neuronal representations regardless of modality. If so, there must be modality-specific effective connections with dependent areas outside the prefrontal cortex, such as sensory or motor cortex. There may also be modality-specific subdivisions within prefrontal cortex, with regional activation according to selection within a given modality.

The influence of the prefrontal cortex over remote non-prefrontal target areas was formulated in terms of effective connectivity changes (Friston et al., 1997). We used structural equation modelling (SEM; Horwitz et al., 1999; Buchel and Friston, 2000) to measure changes in effective connectivity among prefrontal, motor and visual cortices during performance of actions and colour based tasks. This involved the use of moderator variables which were the same as psychophysiological interaction effects. We extended the use of SEM to incorporate high order interactions between selection, modality and regional activity. Our second hypothesis was that different task demands would be reflected in different patterns of effective connectivity between a common prefrontal region and modality-specific non-prefrontal areas. More specifically, we proposed that free selection enhances modality-specific effective connectivity from prefrontal cortex to dependent regions in motor and visual cortex.

Materials and Methods

Subjects

Twelve right-handed volunteers participated in the study (aged 23–38 years, mean 30 years, seven men), after providing written informed consent. The subjects were recruited from a departmental register of volunteers. None had a history of neurological or psychiatric illness, and they took no regular medication. The study was approved by the joint ethics committee of the National Hospital for Neurology and Neurosurgery (UCLH NHS Trust) and the Institute of Neurology (UCL), London.

Behavioural Paradigm

The experiment used an epoch-based factorial design, with four similar trial types separated by rest periods, presented in pseudorandom order. The four trial types are illustrated in Figure 1. Trials were based on selection of finger movements or colours. The finger movements and colours were either freely selected by subjects or specified by the experimenter. The four principal tasks were performed for 26 s, following a brief written instruction cue. Rest periods between tasks lasted 8–20 s.

Figure 1.

Schematic representation of the four tasks. In all conditions, a large cue box was presented above four small response boxes, each one corresponding to a finger of the right hand. In condition EC (externally specified colour), the cue box filled with one of four vivid colours, to indicate the necessary response. One of four coloured response boxes was also filled with this colour, and the subject pressed a button with the corresponding finger. In FC (freely selected colour), the subject freely chose one of four colours, and pressed the finger indicated by the response box of the chosen colour. In EA (externally specified action), the direction of an arrow in the cue box specified the button press with one of four fingers. In FA (freely selected action), the arrow pointed to a neutral location, and the subject freely chose which button to press. Cues and response boxes were presented together for 1.5 s, every 3 s.

Figure 1.

Schematic representation of the four tasks. In all conditions, a large cue box was presented above four small response boxes, each one corresponding to a finger of the right hand. In condition EC (externally specified colour), the cue box filled with one of four vivid colours, to indicate the necessary response. One of four coloured response boxes was also filled with this colour, and the subject pressed a button with the corresponding finger. In FC (freely selected colour), the subject freely chose one of four colours, and pressed the finger indicated by the response box of the chosen colour. In EA (externally specified action), the direction of an arrow in the cue box specified the button press with one of four fingers. In FA (freely selected action), the arrow pointed to a neutral location, and the subject freely chose which button to press. Cues and response boxes were presented together for 1.5 s, every 3 s.

Externally specified action (EA):

an arrow was presented every 3 s (duration 1.5 s) in an instruction box in the top half of the screen. Four smaller boxes were presented simultaneously below the instruction box, representing four lightly sprung buttons that could be pressed by the fingers of the right hand. The arrow had one of four orientations (9, 11, 1 and 3 o'clock) specifying the action of button press by one of the four fingers, by direct visuomotor mapping. The subjects used the specified finger to press the button.

Free selection of action (FA):

the arrow in the button box pointed to 12 o'clock. Subjects chose one of the response buttons to press. They were asked to make a fresh choice on each trial, regardless of previous choices.

Externally specified colour (EC):

the instruction box was filled with a single colour (red, yellow, green or blue). The four response boxes were each filled with one of the colours, in different arrangements on each trial, and subjects pressed the button that matched the instruction colour.

Free selection of colour (FC):

the instruction box contained four colours. The four response boxes were each filled with one of the colours, in a different order on each trial. The subjects were required to choose one of these four colours, then press the button represented by the response box of the chosen colour. Again, subjects were asked to make a fresh choice on each trial.

Subjects were pre-trained on all tasks on the day of scanning. They were instructed to maintain their gaze on the top instruction box throughout the experiment. This was confirmed by simultaneous infrared oculographic monitoring in four subjects (ASL 5000). The mean reaction time for each subject in each of the four conditions was entered into a two-way repeated measures analysis of variance (ANOVA), with factors of selection (free versus externally specified) and modality (action versus colour), using SPSS8.0 for Windows.

Functional Imaging

Subjects lay supine with their heads fixed by firm foam pads. Task instructions, trial stimuli and the recording of latency from stimulus to button press were controlled by an Apple Macintosh 7600 computer operating Cognitive Interface software (Cogent, Wellcome Department of Imaging Neuroscience, London, UK). The four lightly sprung buttons were mounted under subjects' fingertips, on a moulded splint that supported a comfortable neutral hand position.

Functional imaging used

\(T_{2}^{{\ast}}\)
-weighted echo-planar MRI throughout 28 min continuous whole brain imaging (2 T Siemens Vision MRI scanner, repeat time 3650 ms, echo time 40 ms, 64 × 64 × 40 voxels, 3 mm isotropic resolution). The first five images were discarded to allow steady state magnetization. Statistical parametric mapping software was used for image processing and analysis (SPM99, http://www.fil.ion.ucl.ac.uk/spm). The images were realigned to the mean image by rigid body transformation, sinc interpolated in time to correct for phase shift during volume acquisition and transformed to normal anatomic space (Talairach and Tournoux, 1988) using the Montreal Neurological Institute template by linear and smoothly non-linear transformations (Friston et al., 1995a). For individual subject analyses the data were smoothed spatially with a Gaussian kernel of 6 mm at full width half maximum. High-resolution T1-weighted images were acquired for all subjects to facilitate anatomical localization of activations (repeat time 11 ms, echo time 4 ms, TI 1000 ms, 256 × 224 × 176 voxels, 1 mm isotropic resolution).

Statistical Parametric Mapping

A general linear model was applied voxel by voxel to the functional data (Friston et al., 1995b, 1996; Worsley et al., 1996), using box-car covariates for the epochs EA, FA, EC, FC and rest, and a transient covariate for the task instructions. The covariates were convolved with a canonical haemodynamic response function. Residual motion effects were modelled by inclusion of the first temporal derivative of the realignment parameters, representing scan-to-scan movement. A band-pass filter was applied prior to parameter estimation, including a low frequency cut-off of 120 s, and a higher frequency filter corresponding to the canonical haemodynamic response function.

Voxel-wise parameter estimates were derived for each covariate, in a subject-specific fixed effects model. Contrast images of interest were calculated for each subject, including: the main effects of task performance versus rest [EA + EC + FA + FC − 4 × rest]; the main effect of action versus colour [(EA + FA) − (EC + FC)]; colour versus action [(EC + FC) − (EA + FA)]; free selection versus external specification [(FA + FC) − (EA + EC)]; external specification versus free selection [(EA + EC) − (FA + FC)]; the interactions between selection context and the modality [(EA + FC) − (EC + FA)] and [(EC + FA) − (EA + FC)]; and the simple effects of free selection versus external specification were calculated within each modality [FC − EC] and [FA − EA]. An SPM{F} map was also calculated for each subject that included the four experimental conditions and rest in the F-contrast.

For each contrast, the 12 subject specific contrast images were entered into a two-way one-sample t-test, to create a SPM{t}-statistic image, with 11 degrees of freedom. This two-step procedure is equivalent to a random effects analysis. It enables inferences to be extended to the general population from which the subjects were drawn (Friston et al., 1999). To accommodate greater inter-subject anatomic variability inherent in group analyses contrast images were smoothed by a second Gaussian kernel FWHM 8 mm prior to the second level t-test. This is equivalent to a single smoothing step of 10 mm (the root sum of squares of the sequential smoothing kernels).

For the contrasts of main effects of selection and modality, results are tabulated and illustrated for voxels for which P < 0.05, corrected for whole brain multiple comparisons. Where indicated in the text and tables, reduced search volumes were used for specific contrasts in view of prior anatomically constrained hypotheses. For interactions between modality and selection, reduced search volumes were defined within the SPM99 results graphical user interface, as 10 mm spheres centred on either the hand knob or previous functional localization of colour sensitive areas (see results section for loci). The reduced search volume for the prefrontal cortex could not be specified in this way. Instead, a mask image was created using MRIcro software (http://www.psyc.nott.ac.uk/∼/mricro.html ) that included precentral, superior and middle frontal gyri bilaterally with total search volume 170 cm3. In addition, to convey the extent of activation trends for principal contrasts, some results of interest are presented at a lower threshold, P < 0.001 uncorrected.

Structural Equation Modelling (SEM)

A hypothesis-driven approach was taken to estimate the changes in effective connectivity within a specified model, which included bilateral prefrontal, parietal and prestriate cortex and the left motor cortex. The method was similar to that used by Buchel and Friston (1997) and Rowe et al. (2001), enabling a systems level analysis of regional interactions while at the same time acknowledging that the connections may be indirect or polysynaptic. Our model is shown schematically in Figure 2a. The model included the time-course of the first eigenvariate of the adjusted BOLD signal for each region (Buchel and Friston, 1997; Rowe et al., 2001).

Figure 2.

(A) The model included anatomic interconnections among prefrontal (PF), parietal (PA), prestriate (PS) and left motor cortex (M). The left side of each figure represents the left hemisphere. The basic anatomic connections are given as solid lines. The path coefficients were estimated in subject-specific models (see Materials and Methods). The values in this figure show the group mean path coefficient for each connection of the model. (B) For each connection between source S and target region T, we proposed a potential influence of modality (AvC, action versus colour, thick solid line) or selection (FvE, free versus external, thin solid line) on the degree of coupling. Selection may also affect the degree to which modality changes the influence of the source on the target region (double line). (C) These influences were modelled using interaction covariates. (DH) The results of the structural equation modelling, each figure representing different interactions between selection, modality and regional brain activation. Values shown are the group mean path coefficients for each significant connection (shown as solid lines). Connections present in the model but not significantly modulated by each context are shown as dashed lines. D, E, F represent the second-order interactions between regional activity and the tasks: coupling greater during action than colour (D), greater during colour than action (E) and greater during free selection than specified responses (F). G and H illustrate the third-order interactions between regional activity: greater coupling during free selection of action (G) or during free selection of colour (H). Although the basic connections, second- and third-order interactions are illustrated separately, they are calculated simultaneously within a single model for each subject.

Figure 2.

(A) The model included anatomic interconnections among prefrontal (PF), parietal (PA), prestriate (PS) and left motor cortex (M). The left side of each figure represents the left hemisphere. The basic anatomic connections are given as solid lines. The path coefficients were estimated in subject-specific models (see Materials and Methods). The values in this figure show the group mean path coefficient for each connection of the model. (B) For each connection between source S and target region T, we proposed a potential influence of modality (AvC, action versus colour, thick solid line) or selection (FvE, free versus external, thin solid line) on the degree of coupling. Selection may also affect the degree to which modality changes the influence of the source on the target region (double line). (C) These influences were modelled using interaction covariates. (DH) The results of the structural equation modelling, each figure representing different interactions between selection, modality and regional brain activation. Values shown are the group mean path coefficients for each significant connection (shown as solid lines). Connections present in the model but not significantly modulated by each context are shown as dashed lines. D, E, F represent the second-order interactions between regional activity and the tasks: coupling greater during action than colour (D), greater during colour than action (E) and greater during free selection than specified responses (F). G and H illustrate the third-order interactions between regional activity: greater coupling during free selection of action (G) or during free selection of colour (H). Although the basic connections, second- and third-order interactions are illustrated separately, they are calculated simultaneously within a single model for each subject.

The specific coordinates for the seven regions comprising the model were taken from the contrast of ‘all tasks versus rest’, including the middle frontal gyrus corresponding to: area 46 of prefrontal cortex (Rajkowska and Goldman-Rakic, 1995) at xyz = −44, 28, 30 and 46 38 28; left motor cortex, xyz = −44 −24 62; ventrolateral prestriate cortex, to include colour sensitive regions of human prestriate cortex (Hadjikhani et al., 1998; Zeki and Bartels, 1999; Bartels and Zeki, 2000; Tootell and Hadjikhani, 2001), xyz = −36, −70, −12 and 40 −78 −14; and intraparietal cortex, xyz = −30 −56 40 and 28 −54 44. Regions were defined as spheres (5 mm radius) centred on these coordinates, including all voxels that exceeded P < 0.001 in the SPM{F} for all conditions. These voxels will have been activated in association with some or all of the experimental factors, but the inclusion voxels was unbiased by any particular contrast. Premotor cortex was not modelled separately, despite activation around the intersection of the superior frontal sulcus and precentral sulcus, because it was not clearly distinguishable from the frontal eye-fields (Petit et al., 1997; Luna et al., 1998). Although subjects were instructed to look only in the upper cue box, it is possible that smaller eye-movements within the cue box may have differentiated colour and action tasks and therefore resulted in differential activations in this area. Our model did not include interhemispheric connections, since the combination of these with reciprocal frontal–parietal connections would have created a loop that may destabilize model fitting. Our priority was to include interactions with parietal cortex within each hemisphere.

The coupling induced by the visuomotor tasks (versus rest) may differ with context, illustrated in Figure 2b. The context may be defined by the modality (colour versus action) or by the nature of selection (freely versus externally specified). The factorial design speaks to the role of interactions. There may be interactions between modality and selection in determining activation within a voxel. There may also be psychophysiological interactions (PPIs), defined by Buchel and Friston (1997) as an interaction between the psychological context and the physiological coupling between brain regions.

At one level, there may be an interaction between one psychological contextual factor (modality or selection) and the physiological coupling. In our SEM, the contextual factor entered explicitly as a designed variable, encompassing the period of visuomotor task performance and the following rest period, illustrated in Figure 2c. In this model, the effect of visuo-motor task performance enters vicariously through the haemodynamic changes it evokes in regions in the model. For implementation in the SEM, we proposed that where two regions are anatomically connected, the degree of coupling between source and target region may change with the modality context (bold arrows) and with the selection context (thin arrows). These effects were incorporated into the model by the inclusion of an ‘psychophysiological interaction covariate’ as a potential influence on a target area, illustrated in Figure 2c. The interaction covariates are calculated as the product of a mean-corrected vector of BOLD activation in the source area (we used the time-course of the first eigenvariate of the adjusted BOLD response for the region of interest) and a mean-corrected vector defining the contexts of selection or modality (normalized and convolved by a canonical haemodynamic response function).

A higher level psychophysiological interaction may also exist. That is, the effect of modality on inter-regional physiological coupling may itself interact with the selection context (Fig. 2b, double arrow). This third-order interaction was incorporated into the model by an additional ‘psycho-psychophysiological interaction covariate’ (the product of the source activity and a vector describing the interaction between modality and selection) with the potential to influence the target area (Fig. 2c, double arrow).

The SEM was implemented with the SPM99 Toolbox under Matlab 5.3, using an iterative maximum likelihood algorithm (Higham, 1993) to calculate path coefficients by which the implied covariances best match the observed variance–covariance structure of the empirical data. The influence that one variable (regional activation or contextual moderator variable) exerts upon another is quantified by a path coefficient. This represents the magnitude and direction of the influence, in the range −1 to +1. A path coefficient of +1 indicates a unit response (in units of standard deviation) of the dependent variable (target area) to a unit change in the explanatory variable (source area or contextual moderator variable). After optimal model fit for each subject, the path coefficients were treated as dependent variables in one sample t-tests. This second-level analysis allowed us to investigate the consistency of each modelled context-dependent modulation of inter-regional coupling across the group, testing the null hypothesis that a path coefficient was zero.

Results

Behavioural Data

The mean response latencies for each of the four conditions are shown in Figure 3. There was no overall effect of modality (main effect of modality: F = 0.8, df 1,11, n.s.), but there was a significant interaction between free selection and modality (F = 15, df 1,11, P < 0.01). Post hoc paired t-test comparison of response latencies between FA and FC conditions confirmed a significant difference (t = 2.74, df = 11, P < 0.05). It should be noted that the tasks were ‘blocked’ in paced epochs so that during the condition of freely selected actions, subjects could have anticipated the next trial and chosen their responses in advance of the visual cues. The response latency does not therefore necessarily include delay due to selection processes. In contrast, on colour free selection trials, the arrangement of colours was varied from trial to trial. The subjects were asked to select between colours following the presentation of the response boxes. With such selection, the response latency would be expected to be increased relative to externally specified colour trials, due to post-stimulus selection processes (as shown by Fig. 3). Alternatively, subjects might have chosen a colour ahead of the response boxes, and then waited for the response boxes to appear in order to know which corresponding colour to press. This can still be considered to be free selection of colours, within blocks, but this strategy is not supported by the behavioural data. It is also possible that subjects disregarded task instructions, and chose a finger to press regardless of colours. This is unlikely, since in this case response latencies would have been similar to those in the free selection of action condition, and they were, in fact, higher. We conclude that subjects were selecting between alternative responses, in both modalities.

Figure 3.

Group mean response latency (ms) from the presentation of the stimuli to button press, for each condition (error bars: between subjects SE).

Figure 3.

Group mean response latency (ms) from the presentation of the stimuli to button press, for each condition (error bars: between subjects SE).

Analyses of Regional Specialization

Table 1 details those areas associated with performance of all tasks versus rest, including primary visual and motor areas, prestriate and premotor areas, parietal and frontal association cortex. There were significant differences between the free selection tasks and those in which the targets were externally specified (+FC +FA −EC −EA). The free selection of responses was associated with greater activity of the middle frontal gyrus (Brodmann area 46) of the prefrontal cortex. Details are listed in Table 2 and illustrated by Figure 4. Contrasts for simple main effects of selection context within each modality revealed similar peak activations in the left middle frontal gyrus (Brodmann area 46) of the prefrontal cortex for both free selection of actions [FA − EA], and bilaterally for free selection of colours [FC − EC], listed in Table 2. The right middle frontal gyrus peak activation for the simple main effect of selection within action tasks was below threshold for whole brain correction of multiple comparisons. However, in view of the hypothesis regarding a generic role of prefrontal cortex in selection, correction was made within a reduced search volume that included precentral, superior and middle frontal gyri bilaterally (search volume 170 cm3). The right peak activation was significant within this reduced search volume, P < 0.05. The formal interaction analysis between modality and selection was also not significant here (see below).

Figure 4.

SPM(t) maps at P < 0.05 (corrected for whole brain comparisons) showing bilateral prefrontal cortical activation greater with free selection of responses than with externally specified responses, overlaid on the group mean structural images. Top row, parasagittal slice x = −36 (left hemisphere), coronal slice y = 32, middle row parasagittal slice x = 44 (right hemisphere), coronal slice y = 40. The high statistical threshold applied in the top two rows enables clear anatomical localization of the peaks of selection associated activation. However in view of the hypothesized role of the prefrontal cortex in free selection, a lower threshold has been applied to the surface rendered images in the bottom row (P < 0.001 uncorrected) to convey the extensive bilateral frontal activations. Note the absence of motor or visual cortex activation for this contrast even at reduced statistical threshold.

Figure 4.

SPM(t) maps at P < 0.05 (corrected for whole brain comparisons) showing bilateral prefrontal cortical activation greater with free selection of responses than with externally specified responses, overlaid on the group mean structural images. Top row, parasagittal slice x = −36 (left hemisphere), coronal slice y = 32, middle row parasagittal slice x = 44 (right hemisphere), coronal slice y = 40. The high statistical threshold applied in the top two rows enables clear anatomical localization of the peaks of selection associated activation. However in view of the hypothesized role of the prefrontal cortex in free selection, a lower threshold has been applied to the surface rendered images in the bottom row (P < 0.001 uncorrected) to convey the extensive bilateral frontal activations. Note the absence of motor or visual cortex activation for this contrast even at reduced statistical threshold.

Table 1

Contrast of all tasks versus rest, including peaks for which family-wise error P < 0.05 in bold (corrected for whole brain comparisons, t > 10.6) and P < 0.001 (uncorrected, t > 4.03)

Region
 
P(FW)
 
P(FDR)
 
t
 
xyz
 

 

 
R frontal pole 1.000 0.003 4.98 36 50 22 
L frontal pole 1.000 0.006 4.34 −40 52 20 
R inferior frontal gyrus (BA 45) 0.907 0.001 6.25 54 16 
R middle frontal gyrus (BA 46) 0.997 0.002 5.30 48 40 26 
L middle frontal gyrus (BA 9) 0.356 0.001 8.18 −44 28 30 
R superior frontal sulcus 0.126 0.000 9.63 36 10 60 
Supplementary motor area 0.007 0.000 12.82 −4 −2 60 
Supplementary motor area 0.322 0.001 8.33 22 42 
R precentral sulcus 0.251 0.001 8.71 62 12 30 
L precentral sulcus 0.029 0.000 11.14 −44 0 42 
L precentral gyrus 0.133 0.000 9.58 −38 −6 50 
L precentral gyrus 0.616 0.001 7.26 −34 −14 56 
L central sulcus 0.824 0.001 6.59 −54 −20 40 
L central sulcus 0.953 0.001 5.99 −44 −26 62 
L central sulcus 1.000 0.004 4.82 −30 −24 74 
L postcentral sulcus 0.101 0.000 9.85 −38 −32 48 
R intraparietal sulcus 0.001 0.000 15.75 30 −66 50 
R intraparietal sulcus 0.013 0.000 12.04 44 −42 44 
L intraparietal sulcus 0.250 0.001 8.72 −42 −40 52 
L intraparietal sulcus 0.429 0.001 7.89 −30 −56 40 
L sup parietal cortex 0.213 0.001 8.95 −14 −76 56 
R prestriate cortex 0.480 0.001 7.70 40 −78 −14 
L prestriate cortex 0.947 0.001 6.03 −36 −70 −12 
L prestriate cortex 0.984 0.002 5.67 −28 −82 20 
R striate cortex 0.036 0.000 10.92 4 −84 −12 
R striate cortex 0.571 0.001 7.40 14 −92 10 
L striate cortex 0.710 0.001 6.97 −4 −100 
R inferotemporal gyrus 0.181 0.001 9.19 42 −62 −24 
L SII 0.644 0.001 7.17 −54 −22 18 
L SII 0.977 0.002 5.76 −44 −2 
R cerebellar hemisphere 0.010 0.000 12.33 30 −46 −26 
L cerebellar hemisphere 0.026 0.000 11.25 −34 −64 −30 
R cerebellar hemisphere 0.133 0.000 9.58 16 −68 −24 
Vermis 0.184 0.001 9.17 −6 −78 −24 
Dentate 1.000 0.005 4.47 −10 −34 −34 
L thalamus 0.110 0.000 9.77 −28 −20 
L putamen 0.619 0.001 7.25 −26 10 −4 
R insula 1.000 0.007 4.20 30 22 −2 
L insula
 
1.000
 
0.007
 
4.31
 
−36
 
12 
Region
 
P(FW)
 
P(FDR)
 
t
 
xyz
 

 

 
R frontal pole 1.000 0.003 4.98 36 50 22 
L frontal pole 1.000 0.006 4.34 −40 52 20 
R inferior frontal gyrus (BA 45) 0.907 0.001 6.25 54 16 
R middle frontal gyrus (BA 46) 0.997 0.002 5.30 48 40 26 
L middle frontal gyrus (BA 9) 0.356 0.001 8.18 −44 28 30 
R superior frontal sulcus 0.126 0.000 9.63 36 10 60 
Supplementary motor area 0.007 0.000 12.82 −4 −2 60 
Supplementary motor area 0.322 0.001 8.33 22 42 
R precentral sulcus 0.251 0.001 8.71 62 12 30 
L precentral sulcus 0.029 0.000 11.14 −44 0 42 
L precentral gyrus 0.133 0.000 9.58 −38 −6 50 
L precentral gyrus 0.616 0.001 7.26 −34 −14 56 
L central sulcus 0.824 0.001 6.59 −54 −20 40 
L central sulcus 0.953 0.001 5.99 −44 −26 62 
L central sulcus 1.000 0.004 4.82 −30 −24 74 
L postcentral sulcus 0.101 0.000 9.85 −38 −32 48 
R intraparietal sulcus 0.001 0.000 15.75 30 −66 50 
R intraparietal sulcus 0.013 0.000 12.04 44 −42 44 
L intraparietal sulcus 0.250 0.001 8.72 −42 −40 52 
L intraparietal sulcus 0.429 0.001 7.89 −30 −56 40 
L sup parietal cortex 0.213 0.001 8.95 −14 −76 56 
R prestriate cortex 0.480 0.001 7.70 40 −78 −14 
L prestriate cortex 0.947 0.001 6.03 −36 −70 −12 
L prestriate cortex 0.984 0.002 5.67 −28 −82 20 
R striate cortex 0.036 0.000 10.92 4 −84 −12 
R striate cortex 0.571 0.001 7.40 14 −92 10 
L striate cortex 0.710 0.001 6.97 −4 −100 
R inferotemporal gyrus 0.181 0.001 9.19 42 −62 −24 
L SII 0.644 0.001 7.17 −54 −22 18 
L SII 0.977 0.002 5.76 −44 −2 
R cerebellar hemisphere 0.010 0.000 12.33 30 −46 −26 
L cerebellar hemisphere 0.026 0.000 11.25 −34 −64 −30 
R cerebellar hemisphere 0.133 0.000 9.58 16 −68 −24 
Vermis 0.184 0.001 9.17 −6 −78 −24 
Dentate 1.000 0.005 4.47 −10 −34 −34 
L thalamus 0.110 0.000 9.77 −28 −20 
L putamen 0.619 0.001 7.25 −26 10 −4 
R insula 1.000 0.007 4.20 30 22 −2 
L insula
 
1.000
 
0.007
 
4.31
 
−36
 
12 

Values given include probability of family-wise error (FW), false discovery rate (FDR), t-value and xyz coordinates (standardized Talairach coordinates in millimetres using the MNI-ICBM 152 template).

Table 2

Contrast of free selection versus externally specified task performance, including peaks for which family-wise error P < 0.05 (corrected for whole brain comparisons, t > 10.6)

Region
 
P(FW)
 
P(FDR)
 
t
 
xyz
 

 

 
Combined colour and action tasks [FA + FC − EA − EC]       
    Left middle frontal gyrus (BA 46) 0.009 0.001 13.44 −46 32 28 
    Right middle frontal gyrus (BA 46) 0.040 0.001 11.59 44 40 32 
Colour tasks only [FC − EC]       
    Left middle frontal gyrus (BA 46) 0.012 0.004 13.00 −50 30 28 
    Right middle frontal gyrus (BA 46) 0.012 0.004 13.01 44 38 32 
Action tasks only [FA − EA]       
    Left middle frontal gyrus (BA 46) 0.002 0.000 15.20 −40 36 30 
    Inferior parietal cortex 0.014 0.001 12.79 54 −44 50 
    Right middle frontal gyrus BA 46
 
SVC
 
0.001
 
7.42
 
42
 
40 32 
Region
 
P(FW)
 
P(FDR)
 
t
 
xyz
 

 

 
Combined colour and action tasks [FA + FC − EA − EC]       
    Left middle frontal gyrus (BA 46) 0.009 0.001 13.44 −46 32 28 
    Right middle frontal gyrus (BA 46) 0.040 0.001 11.59 44 40 32 
Colour tasks only [FC − EC]       
    Left middle frontal gyrus (BA 46) 0.012 0.004 13.00 −50 30 28 
    Right middle frontal gyrus (BA 46) 0.012 0.004 13.01 44 38 32 
Action tasks only [FA − EA]       
    Left middle frontal gyrus (BA 46) 0.002 0.000 15.20 −40 36 30 
    Inferior parietal cortex 0.014 0.001 12.79 54 −44 50 
    Right middle frontal gyrus BA 46
 
SVC
 
0.001
 
7.42
 
42
 
40 32 

Values given include probability of family-wise error (FW), false discovery rate (FDR), t-value and xyz coordinates (standardized Talairach coordinates in millimetres using the MNI-ICBM 152 template). SVC, small volume correction; see text.

There were modality differences in the patterns of task-related activation. The colour-based tasks, compared with the action-based tasks [(+EC +FC −EA −FA)] were associated with significantly greater activation of ventral prestriate cortex, in the colour-sensitive areas identified by Zeki as human V4 [28, −78, −14, t = 9.71; −30, −76, −16, t = 7.63; P < 0.05 corrected within reduced search volume 10 mm radius from the colour sensitive region identified by Bartels and Zeki (2000)]. At a reduced threshold (P < 0.001, uncorrected), there were additional activations in anterior prestriate cortex, identified as V4a (−30, −52, −6, t = 9.15; −34, −52, −22, t = 6.43) (Zeki and Bartels, 1999; Bartels and Zeki, 2000). There was no significant additional activation in the region of the left motor or premotor cortex. The action-based tasks, compared with the colour-based tasks [(+EA +FA −EC −FC)], were associated with additional activation (thresholded at P < 0.001, uncorrected) of the supplementary motor area (−2, −6, 68, t = 5.33), SII (−66, −4, 4, t = 5.49; 60, −8, 6, t = 5.56), right ventral prefrontal (46, 42, 8, t = 4.90) and right sensorimotor cortex (24, −30, 74, t = 4.77). There was no significant additional activation in the visual cortex.

There were also interactions between selection and modality. These are shown in Figure 5 [contrasts (+EC −EA −FC +FA) and (−EC +EA +FC −FA)]. The difference in activation between freely selected and externally specified colour tasks was significantly greater than the difference between freely selected and externally specified action tasks in ventral prestriate cortex [−40, −78, −10, t = 4.64, P < 0.05, corrected within reduced search volume 10 mm radius from the colour-attention region identified by Bartels and Zeki (2000)]. At reduced threshold (P < 0.001 uncorrected) there were additional interactions in striate cortex, xyz = −1, −90, −10 (t = 4.47) and xyz = −6, −82, 14 (t = 5.17); ventral prestriate cortex, xyz = −34, −88, −14 (t = 5.75) and 22, −90, −6 (t = 4.72); and the left frontal cortex xyz = −40, 6, 44 (t = 7.70) and −52, 26, 30 (t = 7.30). In these areas of the left prefrontal cortex in which there was a weak interaction between modality and selection [(FC − EC] − [FA − EA)], there was also a significant simple main effect of action selection [(FA − EA)]. In other words, there was more colour-selection activation but the same regions were also activated during action-selection: there were no regions identified in which there was colour specificity for selection processes.

Figure 5.

SPM(t) maps at P < 0.001 (uncorrected) showing areas of interaction between selection context (free selection more than externally specified) and modality. Top row: striate and prestriate cortex were most activated during the free selection of colours (SPM images are overlaid on the group mean structural images; parasagittal plane x = −6, coronal slice y = −90). Bottom row: the left sensorimotor cortex was most activated during selection of actions (parasagittal plane x = −36, coronal slice y = −26). See text for significance of activations corrected for multiple comparisons within defined search volumes.

Figure 5.

SPM(t) maps at P < 0.001 (uncorrected) showing areas of interaction between selection context (free selection more than externally specified) and modality. Top row: striate and prestriate cortex were most activated during the free selection of colours (SPM images are overlaid on the group mean structural images; parasagittal plane x = −6, coronal slice y = −90). Bottom row: the left sensorimotor cortex was most activated during selection of actions (parasagittal plane x = −36, coronal slice y = −26). See text for significance of activations corrected for multiple comparisons within defined search volumes.

There was the opposite interaction between selection and modality in sensorimotor cortex, xyz = −36, −26, 60 (t = 4.40, P < 0.05 corrected within reduced search volume, 10 mm sphere centred on the ‘hand knob’), indicating that the difference in activation between freely selected and externally specified action tasks was greater than the difference between freely selected and externally specified colour tasks in that area. There was no significant interaction between action and free selection in the prefrontal cortex even at reduced threshold. There were no significant cross-modal interactions, such as visual cortical activation enhanced by free selection action.

Analyses of Effective Connectivity

Figure 2 illustrates the results of the structural equation modelling, with group mean path coefficients for the principal connections and moderator variables. Each of the principal connections had significant positive path coefficients (Fig. 2a) indicating covariance due to performance of the visuomotor tasks (also called ‘stationary coupling’). The modelled influence of prefrontal cortex on motor and visual cortex is significant even in the presence of indirect influence via parietal cortical inputs to motor and visual cortex.

The modality of responses induced differences in inter-regional coupling, indicated by the significant path coefficients for the second-order interaction covariates (fig. 2d–e). Performance of action tasks, versus colour tasks, induced greater coupling between prefrontal cortex and the left motor cortex. Conversely, the performance of the colour tasks (versus action tasks) induced greater coupling between prefrontal cortex and prestriate cortex bilaterally, and between parietal and prestriate cortex on the right.

The free selection of responses (versus externally specified responses) was associated with greater coupling between the left prefrontal cortex and the left motor cortex, and between prefrontal and parietal cortices (Fig. 2f).

There were significant third-order interactions between the selection context, the modality of the task and inter-regional coupling. Although the free selection of responses did not overall increase coupling between prefrontal and prestriate cortex (second-order interaction), free selection enhanced the modality effect (third-order interaction) on the left for action and bilaterally for colour (Fig. 2g–h). This indicates that the task-related coupling between prefrontal and prestriate cortex was greatest for freely selected colour responses. Coupling between the parietal and prefrontal cortex was similarly greatest during freely selected performance in colour tasks. Conversely, there was a positive effect of the interaction between action tasks and free selection on the coupling between parietal to motor cortex on the left: parietal to motor cortical coupling was greatest during free selection of action. The third-order interaction between free selection, actions and prefrontal-motor activation was not significant. However, the combination of second-order interactions above indicates that the highest coupling between prefrontal and motor cortex was during free selection of actions.

Discussion

Activity within Prefrontal Cortex

The free selection of responses was associated with greater activation of the bilateral dorsal prefrontal cortex, compared with performance of similar externally specified responses. This occurred for the selection of both actions and colours. When selecting between potential responses, the critical aspect of the task for prefrontal cortical activation might be the ‘freedom’ of choice to make ‘willed actions’ (Frith et al., 1991). However, we propose that it is the demands on selection processes that determine prefrontal cortical activity. Activity of dorsal prefrontal cortex occurs during selection among motor responses with hands, feet or mouths (Deiber et al., 1991; Frith et al., 1991; Hyder et al., 1997; Spence et al., 1997, 1998). The activation of area 46 on the middle frontal gyrus is of particular significance in studies of response selection, because lesion studies suggest that it is critical for the selection of action based on representational memory (Goldman-Rakic, 1987a,b; Hoshi et al., 2000). It might be argued that prefrontal cortical activation in our free selection tasks was due to working memory for previous selections. However, disruption of prefrontal cortical function by transcranial magnetic stimulation impairs response selection even in the absence of working memory demands (Hadland et al., 2001).

There is also evidence for dorsal prefrontal activation when subjects select responses other than manual movements: random number generation (Jahanshahi et al., 2000); verbal response selection (Buckner et al., 1995; Klein et al., 1995; Phelps et al., 1997; Thompson-Schill et al., 1997, 1998; Desmond et al., 1998; Ojemann et al., 1998); and selection of items from working memory (Rowe et al., 2000). Moreover, the greater the choice (Thompson-Schill et al., 1997; Desmond et al., 1998) or conflict amongst potential responses (George et al., 1994; Carter et al., 1995; Taylor et al., 1997), the greater the activation. The selection of one response from a set of possible responses may be distinguished from the generation of that set. The activation of the dorsal prefrontal cortex is associated with the selection process, not the generation of possible responses (Bunge et al., 2002).

There were no significant interactions between selection and modality on the middle or superior frontal gyri. This result suggests that the same parts of the middle frontal gyrus mediate both action- and colour- based selection. Our results may reflect the spatial resolution of fMRI and group analysis that may average any fine detail of subspecialized areas present within individual subjects. This study is insensitive to such fine-grained spatial organization of function. However, previous functional imaging studies of working memory have also suggested that it is executive processes rather than modality of remembered items that determines the degree of activation of dorsal prefrontal cortex (Owen, 1997; Owen et al., 1998; Smith and Jonides, 1999; D'Esposito et al., 2000)

Activity within Motor, Visual and Parietal Cortex

Frith (2000) proposed that the selection of a target item from a visual scene is comparable to motor and verbal response selection, whether the visual stimuli remain present or are held in working memory. In contrast to motor and verbal selection, the attentional mechanisms of visual selection have been extensively studied at regional and cellular levels. Specific aspects of the visual scene are encoded by neuronal activity in occipital and temporal cortex, including retinotopy, colour states and information about object form or identity. Electrophysiological studies indicate that cells in inferotemporal cortex exhibit competitive interactions amongst different neuronal representations (Luck et al., 1997; Chelazzi et al., 1998; Reynolds et al., 1999). Bias among competing representations has been proposed as a central mechanism of visual attention (Desimone and Duncan, 1995; Desimone, 1998).

It is suggested that choice amongst potential action or colour responses is also mediated by biased activation of action representations or colour representations, respectively. There was greater activity in visual cortex during colour tasks, possibly attributable to an asymmetry in the experimental design (colours were presented only during colour tasks). However, there was even greater activation when colours were freely selected (Fig. 5). This is consistent with top-down modulation of striate and prestriate cortex during free selection of responses. There was also greater activation in motor cortex when actions were freely chosen rather than externally specified (Fig. 5), even though a button press was made on every trial. Attentional bias among action representations may therefore also mediate the selection of voluntary actions.

If the selection of action is mediated by bias among specific neuronal representations, one must consider the nature of these action representations in the motor system. They could represent the muscular activity required for movement of a finger; the anticipated sensory consequences of an intended action; the location or spatial arrangement of the target of action in either egocentric or allocentric coordinates; or more complex integrated schemata (Norman and Shallice, 1980). It is possible that multiple types of representation are activated in parallel (Jeannerod, 1997).

We believe that it is primarily the abstract goal of action that is selected, rather than the physical properties of the movement. In this respect, the action tasks differed from colour tasks, even though all tasks included an equal number and type of button presses. Evidence for this comes from electrophysiological studies in non-human primates. Cells tuned to the target of response (final limb position) rather than the required movements (limb movement direction or muscular action) are found in dorsal premotor and primary motor cortex (Alexander and Crutcher, 1990; Shen and Alexander, 1997; Kakei et al., 1999; Graziano et al., 2002). Moreover, the target specificity of primary motor cortical cells is not conditional on the task rule,. in contrast to responses in prefrontal cells (Hoshi et al., 1998).

The parietal cortex was also activated when freely selecting colours and actions (Fig. 4) albeit to a lesser extent than the prefrontal cortex. Significant changes were only seen for the free selection of action, in the right inferior parietal lobe (Table 2). Response selection has previously been associated with parietal activation in some studies (Deiber et al., 1991; Spence et al., 1997, 1998; Thompson-Schill et al., 1997; Ojemann et al., 1998; Jahanshahi et al., 2000; Rowe et al., 2000; Rowe and Passingham 2001) but not consistently (Frith et al., 1991; Spence et al., 1998). However, in contrast to prefrontal cortex, parietal activations were not independent of modality and the location of activation varied more widely between studies, including superior-, inferior-, medial- and intra-parietal cortex. The role of the parietal cortex in selection of actions is therefore distinct from the role of the prefrontal cortex. The parietal cortex has also recently been implicated in attentional modulation of representations of non-spatial features of visual objects, including colour (Liu et al., 2003). The left intraparietal and medial parietal cortex were activated at the time of attentional shift towards the colour dimension in complex visual displays. However, unlike prefrontal cortex and fusiform gyrus, the parietal cortex was not activated during sustained colour attention in the study by Liu et al. (2003).

Effective Connectivity between Regions

We used structural equation modelling to determine whether selection context and modality would affect the interregional coupling among prefrontal, motor and visual cortex. Analyses of connectivity are complementary to analyses of regional specific activations, and inferences about regional coupling should not be made based on observation of similar parameter estimates in different regions. For example, Stephan et al. (2003) showed similar anterior cingulate activation during both letter and visuospatial decisions, but distinct patterns of effective connectivity from cingulate cortex to left inferior frontal and right intraparietal cortex according to modality.

Our model simplified the complexity of cortico-cortical anatomic connections in order to test specific hypotheses at the level of interacting cortical systems. In humans, attentional bias of cortical processing of sensory stimuli depends on prefrontal cortical afferents (Alho et al., 1994; Barcelo et al., 2000). Moreover, functional imaging studies show that selective attention to sensory stimuli is associated with modulatory fronto-temporal and fronto-occipital interactions (Buchel and Friston, 1998; Alho et al., 1999; Kastner et al., 1999; Hopfinger et al., 2000, 2001; Kastner and Ungerleider, 2000).

The SEM analysis supports existence of modulatory inputs from prefrontal cortex to multiple modality-specific regions. The free selection of colours was associated with the greatest influence of prefrontal cortex on visual cortex, indicated by the significant influence of the third-order interaction between selection, modality and prefrontal activity bilaterally. The second-order effect of modality on prefrontal-visual cortical coupling (whether freely selected or externally specified) was also significant bilaterally, but coupling was not overall greater in free selection than in externally specified tasks. The presentation of colours only during colour-based tasks may have contributed to the modality-dependent increase in inter-regional coupling. However, it would not be sufficient to explain the interaction with selection context. The coupling between prefrontal and motor cortex was greatest during the free selection of action tasks, even though a button was pressed on every trial in both colour-based and action-based tasks. Although there was no third-order interaction between selection, modality and regional activity, both modality and selection positively modulated this coupling.

Although we have suggested that prefrontal cortex provides ‘top-down’ modulation of visual cortex, it is possible that other areas also do so. One candidate region is the parietal cortex, which is directly connected with extrastriate cortex (Andersen et al., 1990) and motor cortex (Petrides and Pandya, 1984). A previous study of parietal to premotor connectivity using similar SEM analysis of fMRI data failed to show significant changes when subjects attended to predetermined actions (Rowe et al., 2001). The current SEM results suggest that parietal connectivity to motor cortex may contribute to selection of actions when chosen freely on each trial. Moreover, the dorsal prefrontal to visual cortical interactions are significantly stronger during free selection of colour, even in the presence of an ‘indirect’ influence via sequential fronto-parietal and parietal-visual cortical interactions. The prefrontal influence on visual cortical activity shown here and by Barcelo et al. (2000) is likely to be mediated by polysynaptic connections. Structural equation modelling is insensitive to the potential complexity of cortico-cortical connections, restricting its use to testing specific hypotheses at a systems level of analysis, rather than at a synaptic or cellular level.

Conclusions

The same areas of prefrontal cortex were activated in association with free selection of both actions and colours. An SEM analysis suggests that free selection of responses is mediated by changes in effective connectivity between the prefrontal cortex and modality specific areas, over and above the influence of parietal cortex. While there is functional specialization within prefrontal cortex, we suggest that the function of prefrontal cortex is also characterized by changing effective connectivity to multiple, remote, modality-specific regions.

This work was supported by the Wellcome Trust.

References

Alexander GE, Crutcher MD (
1990
) Neural representations of the target (goal) of visually guided arm movements in three motor areas of the monkey.
J Neurophysiol
 
64
:
164
–178.
Alho K, Woods DL, Algazi A, Knight RT, Naatanen R (
1994
) Lesions of frontal cortex diminish the auditory mismatch negativity.
Electroencephalogr Clin Neurophysiol
 
91
:
353
–362.
Alho K, Medvedev SV, Pakhomov SV, Roudas MS, Tervaniemi M, Reinikainen K, Zeffiro T, Naatanen R (
1999
) Selective tuning of the left and right auditory cortices during spatially directed attention.
Brain Res Cogn Brain Res
 
7
:
335
–341.
Andersen R, Asanuma C, Siegel R (
1990
) Corticocortical connections of anatomically and physiologically defined subdivisions within the inferior parietal lobule.
J Comp Neurol
 
296
:
65
–113.
Asaad WF, Rainer G, Miller EK (
1998
) Neural activity in the primate prefrontal cortex during associative learning.
Neuron
 
21
:
1399
–1407.
Asaad WF, Rainer G, Miller EK (
2000
) Task-specific neural activity in the primate prefrontal cortex.
J Neurophysiol
 
84
:
451
–459.
Barcelo F, Suwazono S, Knight RT (
2000
) Prefrontal modulation of visual processing in humans.
Nat Neurosci
 
3
:
399
–403.
Bartels A, Zeki S (
2000
) The architecture of the colour centre in the human visual brain: new results and a review.
Eur J Neurosci
 
12
:
172
–193.
Brefczynski JA, DeYoe EA (
1999
) A physiological correlate of the ‘spotlight’ of visual attention.
Nat Neurosci
 
2
:
370
–374.
Buchel C, Friston KJ (
1997
) Modulation of connectivity in visual pathways by attention: cortical interactions evaluated with structural equation modelling and fMRI.
Cereb Cortex
 
7
:
768
–778.
Buchel C, Friston KJ (
1998
) Dynamic changes in effective connectivity characterized by variable parameter regression and Kalman filtering.
Hum Brain Mapp
 
6
:
403
–408.
Buchel C, Friston K (
2000
) Assessing interactions among neuronal systems using functional neuroimaging.
Neural Netw
 
13
:
871
–882.
Buckner RL, Raichle ME, Petersen SE (
1995
) Dissociation of human prefrontal cortical areas across different speech production tasks and gender groups.
J Neurophysiol
 
74
:
2163
–2173.
Bunge S, Hazeltine E, Scanlon M, Rosen A, Gabrieli J (
2002
) Dissociable contributions of prefrontal and parietal cortices to response selection.
Neuroimage
 
17
:
1562
–1571.
Carter CS, Mintun M, Cohen JD (
1995
) Interference and facilitation effects during selective attention: an H215O PET study of Stroop task performance.
Neuroimage
 
2
:
264
–272.
Chelazzi L, Miller EK, Duncan J, Desimone R (
1993
) A neural basis for visual search in inferior temporal cortex.
Nature
 
363
:
345
–347.
Chelazzi L, Duncan J, Miller EK, Desimone R (
1998
) Responses of neurons in inferior temporal cortex during memory-guided visual search.
J Neurophysiol
 
80
:
2918
–2940.
Deiber MP, Passingham RE, Colebatch JG, Friston KJ, Nixon PD, Frackowiak RS (
1991
) Cortical areas and the selection of movement: a study with positron emission tomography.
Exp Brain Res
 
84
:
393
–402.
Desimone R (
1998
) Visual attention mediated by biased competition in extrastriate visual cortex.
Philos Trans R Soc Lond B Biol Sci
 
353
:
1245
–1255.
Desimone R (
1999
) Visual attention mediated by biased competition in extrastriate visual cortex. In: Attention, space and action (Humphreys G, Duncan J, Treisman A, eds), Vol. 1, pp. 13–30. Oxford: Oxford University Press.
Desimone R, Duncan J (
1995
) Neural mechanisms of selective visual attention.
Annu Rev Neurosci
 
18
:
193
–222.
Desmond JE, Gabrieli JDE, Glover GH (
1998
) Dissociation of frontal and cerebellar activity in a cognitive task: evidence for a distinction between selection and search.
Neuroimage
 
7
:
368
–376.
D'Esposito M, Postle BR, Rypma B (
2000
) Prefrontal cortical contributions to working memory: evidence from event-related fMRI studies.
Exp Brain Res
 
133
:
3
–11.
Duncan J (
2001
) An adaptive coding model of neural function in prefrontal cortex.
Nat Rev Neurosci
 
2
:
820
–829.
Friston KJ, Ashburner J, Poline J-B, Frith CD, Frackowiak RSJ (
1995
a) Spatial registration and normalisation of images.
Hum Brain Mapp
 
2
:
165
–189.
Friston KJ, Frith CD, Turner R, Frackowiak RS (
1995
b) Characterizing evoked hemodynamics with fMRI.
Neuroimage
 
2
:
157
–165.
Friston KJ, Holmes A, Poline JB, Price CJ, Frith CD (
1996
) Detecting activations in PET and fMRI: levels of inference and power.
Neuroimage
 
4
:
223
–235.
Friston KJ, Buchel C, Fink GR, Morris J, Rolls E, Dolan RJ (
1997
) Psychophysiological and modulatory interactions in neuroimaging.
Neuroimage
 
6
:
218
–229.
Friston KJ, Holmes AP, Worsley KJ (
1999
) How many subjects constitute a study?
Neuroimage
 
10
:
1
–5.
Frith C (
2000
) The role of the prefrontal cortex in the selection of action, as revealed by functional imaging. In: Control of cognitive processes: attention and performance XVIII (Monsell S, Driver J, eds). Cambridge: MIT Press.
Frith CD, Friston K, Liddle PF, Frackowiak RS (
1991
) Willed action and the prefrontal cortex in man: a study with PET.
Proc R Soc Lond B Biol Sci
 
244
:
241
–246.
George MS, Ketter TA, Parekh P, Rosinsky N, Ring H, Casey BJ, Trimle MR, Horwitz B, Herscovitch P, Post SM (
1994
) Regional brain activity when selecting a response despite interference: an H215O PET study of the Stroop and emotional stroop.
Hum Brain Mapp
 
1
:
194
–209.
Goldman-Rakic P (
1987
a) Circuitry of primate prefrontal cortex and regulation of behaviour by representational memory. In: Handbook of physiology (Plum F, ed.), pp. 373–417. Bethesda, MD: American Physiological Society.
Goldman-Rakic PS (
1987
b) Motor control function of the prefrontal cortex.
Ciba Found Symp
 
132
:
187
–200.
Grady C, VanMeter J, Maisog J, Pietrini P, Krasuski J, Rauschecker J (
1997
) Attention related modulation of activity in primary and secondary auditory cortex.
Neuroreport
 
11
:
2511
–2516.
Graziano M, Taylor C, Moore T (
2002
) Complex movements evoked by microstimulation of precentral cortex.
Neuron
 
34
:
841
–851.
Hadjikhani N, Liu AK, Dale AM, Cavanagh P, Tootell RB (
1998
) Retinotopy and color sensitivity in human visual cortical area V8.
Nat Neurosci
 
1
:
235
–241.
Hadland K, Rushworth MFS, Passingham RE, Jahanshahi M, Rothwell J (
2001
) Interference with performance of a response selection task that has no working memory component: an rTMS comparison of the dorsolateral prefrontal and medial frontal cortex.
J Cogn Neurosci
 
13
:
1097
–1108.
Higham NJ (
1993
) Optimization by direct search in matrix computations.
SIAM J Matrix Anal Appl
 
1993
:
317
–333.
Hopfinger JB, Buonocore MH, Mangun GR (
2000
) The neural mechanisms of top-down attentional control.
Nat Neurosci
 
3
:
284
–291.
Hopfinger JB, Woldorff MG, Fletcher EM, Mangun GR (
2001
) Dissociating top-down attentional control from selective perception and action.
Neuropsychologia
 
39
:
1277
–1291.
Horwitz B, Tagamets MA, McIntosh AR (
1999
) Neural modeling, functional brain imaging, and cognition.
Trends Cogn Sci
 
3
:
91
–98.
Hoshi E, Shima K, Tanji J (
1998
) Task-dependent selectivity of movement-related neuronal activity in the primate prefrontal cortex.
J Neurophysiol
 
80
:
3392
–3397.
Hoshi E, Shima K, Tanji J (
2000
) Neuronal activity in the primate prefrontal cortex in the process of motor selection based on two behavioral rules.
J Neurophysiol
 
83
:
2355
–2373.
Hyder F, Phelps EA, Wiggins CJ, Labar KS, Blamire AM, Shulman RG (
1997
) ‘Willed action’: a functional MRI study of the human prefrontal cortex during a sensorimotor task.
Proc Natl Acad Sci USA
 
94
:
6989
–6994.
Jahanshahi M, Dirnberger G, Fuller R, Frith CD (
2000
) The role of the dorsolateral prefrontal cortex in random number generation: a study with positron emission tomography.
Neuroimage
 
12
:
713
–725.
Jeannerod M. (
1997
) The cognitive neuroscience of action. New York: Blackwell.
Kakei S, Hoffman D, Strick P (
1999
) Muscle and movement representations in the primary motor cortex.
Science
 
285
:
2136
–2139.
Kastner S, Ungerleider L (
2000
) Mechanisms of visual attention in the human cortex.
Annu Rev Neurosci
 
23
:
315
–341.
Kastner S, Pinsk MA, De Weerd P, Desimone R, Ungerleider LG (
1999
) Increased activity in human visual cortex during directed attention in the absence of visual stimulation.
Neuron
 
22
:
751
–761.
Klein D, Milner B, Zatorre R, Meyer E, Evans A (
1995
) The neural substrates underlying word generation: a bilingual functional imaging study.
Proc Natl Acad Sci USA
 
92
:
2899
–2903.
Knight RT, Hillyard SA, Woods DL, Neville HJ (
1981
) The effects of frontal cortex lesions on event-related potentials during auditory selective attention.
Electroencephalogr Clin Neurophysiol
 
52
:
571
–582.
Liu T, Slotnick SD, Serences JT, Yantis S (
2003
) Cortical mechanisms of feature based attentional control.
Cereb Cortex
 
13
:
1334
–1343.
Luck SJ, Chelazzi L, Hillyard SA, Desimone R (
1997
) Neural mechanisms of spatial selective attention in areas V1, V2, and V4 of macaque visual cortex.
J Neurophysiol
 
77
:
24
–42.
Luna B, Thulborn KR, Strojwas MH, McCurtain BJ, Berman RA, Genovese CR, Sweeney JA (
1998
) Dorsal cortical regions subserving visually guided saccades in humans: an fMRI study.
Cereb Cortex
 
8
:
40
–47.
McIntosh A, Gonzalez-Lima F (
1994
) Structural equation modeling and its application to network analysis in functional brain imaging.
Hum Brain Mapp
 
2
:
2
–22.
McIntosh AR, Grady CL, Ungerleider LG, Haxby JV, Rapoport SI, Horwitz B (
1994
) Network analysis of cortical visual pathways mapped with PET.
J Neurosci
 
14
:
655
–666.
Miller EK (
1999
) The prefrontal cortex: complex neural properties for complex behavior.
Neuron
 
22
:
15
–17.
Miller E (
2000
) The prefrontal cortex and cognitive control.
Nat Rev Neurosci
 
1
:
59
–65.
Miller EK, Cohen JD (
2001
) An integrative theory of prefrontal cortex function.
Annu Rev Neurosci
 
24
:
167
–202.
Norman D, Shallice T (
1980
) Attention to action: willed and automatic control of behaviour. San Diego, CA: University of California Center for Human Information Processing.
Ojemann JG, Buckner RL, Akbudak E, Snyder AZ, Ollinger JM, McKinstry RC, Rosen BR, Petersen SE, Raichle ME, Conturo TE (
1998
) Functional MRI studies of word-stem completion: reliability across laboratories and comparison to blood flow imaging with PET.
Hum Brain Mapp
 
6
:
203
–215.
Owen AM (
1997
) The functional organization of working memory processes within human lateral frontal cortex: the contribution of functional neuroimaging.
Eur J Neurosci
 
9
:
1329
–1339.
Owen AM, Stern CE, Look RB, Tracey I, Rosen BR, Petrides M (
1998
) Functional organization of spatial and nonspatial working memory processing within the human lateral frontal cortex.
Proc Natl Acad Sci USA
 
95
:
7721
–7726.
Passingham RE, Stephan KE, Kotter R (
2002
) The anatomical basis of functional localization in the cortex.
Nat Rev Neurosci
 
3
:
606
–616.
Petit L, Clark VP, Ingeholm J, Haxby JV (
1997
) Dissociation of saccade-related and pursuit-related activation in human frontal eye fields as revealed by fMRI.
J Neurophysiol
 
77
:
3386
–3390.
Petrides M (
2000
) The role of the mid-dorsolateral prefrontal cortex in working memory.
Exp Brain Res
 
133
:
44
–54.
Petrides M, Pandya DN (
1984
) Projections to the frontal cortex from the posterior parietal region in the rhesus monkey.
J Comp Neurol
 
228
:
105
–116.
Phelps EA, Hyder F, Blamire AM, Shulman RG (
1997
) fMRI of the prefrontal cortex during overt verbal fluency.
Neuroreport
 
8
:
561
–565.
Rajkowska G, Goldman-Rakic PS (
1995
) Cytoarchitectonic definition of prefrontal areas in the normal human cortex: I. Remapping of areas 9 and 46 using quantitative criteria.
Cereb Cortex
 
5
:
307
–322.
Rao SC, Rainer G, Miller EK (
1997
) Integration of what and where in the primate prefrontal cortex.
Science
 
276
:
821
–824.
Reynolds JH, Chelazzi L, Desimone R (
1999
) Competitive mechanisms subserve attention in macaque areas V2 and V4.
J Neurosci
 
19
:
1736
–1753.
Rowe JB, Passingham RE (
2001
) Working memory for location and time: maintenance versus response selection in prefrontal cortex.
Neuroimage
 :
77
–86.
Rowe JB, Toni I, Josephs O, Frackowiak RS, Passingham RE (
2000
) The prefrontal cortex: response selection or maintenance within working memory?
Science
 
288
:
1656
–1660.
Rowe J, Friston K, Frackowiak R, Passingham R (
2001
) Attention to action: specific modulation of cortico-cortical interactions in humans.
Neuroimage
 
17
:
988
–998.
Shen L, Alexander GE (
1997
) Neural correlates of spatial sensory-to-motor transformation in primary motor cortex.
J Neurophysiol
 
77
:
1171
–1194.
Smith EE, Jonides J (
1999
) Storage and executive processes in the frontal lobes.
Science
 
283
:
1657
–1661.
Spence SA, Brooks DJ, Hirsch SR, Liddle PF, Meehan J, Grasby PM (
1997
) A PET study of voluntary movement in schizophrenic patients experiencing passivity phenomena (delusions of alien control).
Brain
 
120
:
1997
–2011.
Spence SA, Hirsch SR, Brooks DJ, Grasby PM (
1998
) Prefrontal cortex activity in people with schizophrenia and control subjects. Evidence from positron emission tomography for remission of ‘hypofrontality’ with recovery from acute schizophrenia.
Br J Psychiatry
 
172
:
316
–323.
Stephan KE, Marshall JC, Friston KJ, Rowe JB, Ritzl A, Zilles K, Fink GR (
2003
) Lateralised cognitive processes and lateralised task control in the human brain.
Science
 
301
:
384
–386.
Talairach J, Tournoux P (
1988
) A co-planar stereotactic atlas of the human brain. Stuttgart: Thieme.
Taylor S, Komblum S, Lauber E, Minoshima S, Koeppe R (
1997
) Isolation of specific interference processing in the stroop task: PET activation studies.
Neuroimage
 
6
:
81
–92.
Thompson-Schill SL, D'Esposito M, Aguirre GK, Farah MJ (
1997
) Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation.
Proc Natl Acad Sci USA
 
94
:
14792
–14797.
Thompson-Schill SL, Swick D, Farah MJ, D'Esposito M, Kan IP, Knight RT (
1998
) Verb generation in patients with focal frontal lesions: a neuropsychological test of neuroimaging findings.
Proc Natl Acad Sci USA
 
95
:
15855
–15860.
Tomita H, Ohbayashi M, Nakahara K, Hasegawa I, Miyashita Y (
1999
) Top-down signal from prefrontal cortex in executive control of memory retrieval.
Nature
 
401
:
699
–703.
Tootell RB, Hadjikhani N (
2001
) Where is ‘dorsal V4’ in human visual cortex? Retinotopic, topographic and functional evidence.
Cereb Cortex
 
11
:
298
–311.
Worsley KJ, Marrett S, Neelin P, Vandal AC, Friston KJ, Evans AC (
1996
) A unified statistical approach for determining significant signals in images of cerebral activation.
Hum Brain Mapp
 
4
:
58
–73.
Zeki S, Bartels A (
1999
) The clinical and functional measurement of cortical (in)activity in the visual brain, with special reference to the two subdivisions (V4 and V4 alpha) of the human colour centre.
Philos Trans R Soc Lond B Biol Sci
 
354
:
1371
–1382.

Author notes

1Wellcome Department of Imaging Neuroscience, Institute of Neurology, London WC1N 3BG, UK, 2Danish Research Centre for Magnetic Resonance, 2650 Hvidovre, Denmark, 3Department of Psychology, University of Newcastle upon Tyne, Newcastle NE1 7RU, UK, 4Neuroimaging Laboratory, Fondazione Santa Lucia, 00179 Rome, Italy and 5Department of Experimental Psychology, University of Oxford, Oxford OX1 3UD, UK