Cognitive Brain Research 16 (2003) 309–322 www.elsevier.com/locate/cogbrainres

Research report A ttention and movement-related motor cortex activation: a high- density EEG study of spatial stimulus–response compatibility P. Praamstraa,* , R. Oostenveld b

aBehavioral Brain Sciences Centre, University of Birmingham, Birmingham B15 2TT, UK bDepartments of Neurology and Medical Physics, University Medical Centre Nijmegen, Nijmegen, The Netherlands Accepted 25 November 2002

Abstract

Visual spatial attentional activation of motor areas has been documented in single cell neurophysiology and functional imaging studies of the brain. Here, we investigate a candidate event-related brain potential representing visuospatial attentional activity in motor areas of the cortex. The investigation aimed to elucidate the neural origin and the functional characteristics of this brain potential, which has been labelled N2cc and is typically observed in spatial stimulus–response compatibility tasks. High-density EEG was recorded in 10 subjects while they performed a Simon-type spatial stimulus–response compatibility task and a control task where the same stimuli were assigned to Go–Nogo response alternatives. The N2cc showed a time course parallel to the posteriorly distributed , associated with visuospatial selection. Scalp distribution and current source density reconstructions allowed a spatial separation of N2pc and centrally distributed N2cc and were compatible with a source for the N2cc in the lateral premotor cortex. Comparisons across tasks demonstrated that the N2cc depends on bilateral response readiness, ruling out an exclusively attentional interpretation. Instead, the activity appears associated with visuospatial attentional processes that serve the selection and suppression of competing responses, in accord with a function of the dorsal premotor cortex in response selection. Together, the results consolidate the N2cc as a new ERP component relevant to the investigation of visuospatial motor processes.  2002 Elsevier Science B.V. All rights reserved.

Theme: Motor systems and sensorimotor integration

Topic: Cortex

Keywords: Stimulus–response compatibility; Visual spatial attention; Visuomotor processing; Event-related potentials; Premotor cortex

1 . Introduction model developed by Kornblum et al. posits that when stimulus and response sets share certain attributes, the Behavioral research into the processes that link sensory presentation of a stimulus can elicit an automatic activation input to motor output has made abundant use of ex- of the corresponding response. perimental approaches that manipulate the spatial relation Studies pertaining to the neural basis of spatial S-R between stimulus and response. With spatially congruent compatibility effects have addressed, among other ques- stimulus–response pairings, responses are typically faster tions, whether there is neural activity that can be taken to and more accurate than with spatially incongruent pairings express automatic response activation [11,52,68]. In a [18]. According to Kornblum et al. [31], spatial stimulus– single-cell study of monkey primary motor cortex, Riehle response (S-R) compatibility effects represent one par- et al. [52] acknowledged that this entails an unresolved ticular instance of a broader set of S-R compatibility ambiguity: early neuronal activation co-varying with effects that depend on the presence of shared features stimulus location can be interpreted either as a visuospatial between stimulus and response. The dimensional overlap representation or as an automatic activation of the spatially congruent response. This ambiguity exemplifies a more *Corresponding author. Tel.: 144-121-414-7211; fax: 144-121-414- general issue in sensorimotor neurophysiology, i.e. how to 4897. distinguish between neural activity representing the E-mail address: [email protected] (P. Praamstra). stimulus that instructs an action from that associated with

0926-6410/02/$ – see front matter  2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0926-6410(02)00286-0 310 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 the action itself (e.g. Refs. [4,5,32,57]). It is, furthermore, tasks with a vertical arrangement of stimuli and response compounded by the problem that motor cortex neuronal hands, to ensure that stimulus asymmetries and shifts of activity following an instruction stimulus may also repre- attention are confined to the vertical axis and not picked up sent the reorientation of selective spatial attention [32]. by the LRP derivation, which measures only electric While similar ambiguities arise in human EEG studies, potential differences between homologous left-right scalp scalp-recorded event-related potentials (ERPs), in particu- electrode sites. This procedure dissociates stimulus and lar the lateralised readiness potential (LRP), have been attention-related EEG activity from response-related activi- successfully exploited to elucidate spatial S-R compatibili- ty, and has helped to establish the notion that spatial ty effects. A number of relevant ERP studies have sug- stimulus attributes can elicit automatic response activation gested that spatial attentional orientation, in the context of [14,60,62]. a spatial S-R compatibility task, elicits attention-related As already stated, the present study is concerned with motor cortex activation that can be identified in lateralised the presumably attention-related EEG activity above the ERPs [44,49,63]. The present study sought to further motor cortex that is subtracted out when using a vertical clarify the generation and functional characteristics of the S-R arrangement. There are two, closely related, dis- presumed attention-related activity. Firstly, because atten- advantages to the elimination of this activity from consid- tion-related activation of motor structures has been pro- eration when studying spatial S-R compatibility. Firstly, posed to play an important role in spatial S-R compatibility eliminating attention-related EEG activity ignores the effects [53,56]. Secondly, in order to consolidate the possible role of attention and attention-related motor cortex attention-related motor area activation as an ERP com- activation in the explanation of spatial S-R compatibility ponent relevant to the investigation of visuospatial motor effects. There is evidence that spatial S-R compatibility functions. We will first describe the relation of the atten- effects crucially depend on an attention shift towards the tion-related motor cortex activity to the movement-related location of a stimulus presented outside the attentional LRP and to the N2pc, an ERP component associated with focus [53,56]. According to the premotor theory of atten- visual spatial selection. tion, shifts of attention necessarily involve activity in Movement-related EEG activity in reaction time tasks is sensorimotor circuits. Thus, to the extent that these circuits typically isolated by subtracting the activity recorded coincide or interact with the sensorimotor structures that above the motor cortex ipsilateral to the response hand support the directional motor responses under study, from the activity recorded contralaterally. This subtraction attention-related motor activation needs to be taken into is performed for both response hands and the combined account in investigations of spatial S-R compatibility. signals form the LRP (for reviews see Refs. [8,16]). By the The second disadvantage of eliminating the attention- nature of its derivation, the LRP reflects the differential related N2pc is that this component’s overlap with the LRP activation of left and right motor cortices, providing a may be due not merely to volume conduction, but also, and sensitive measure of the point in time at which response more interestingly, to a shared neural substrate in the selection has progressed to a stage where side-specific motor cortex [44]. The extension of the N2pc from its motor activation is started. In addition, it is sensitive to occipital maximum towards central electrode sites is covert, automatic response tendencies elicited by salient commonly explained as a result of passive volume conduc- stimulus features [8]. Although the LRP derivation is tion [16,62]. In contrast, Wascher and Wauschkuhn [63] designed to separate movement-related from stimulus-re- noted that the N2pc, as seen in a spatial S-R compatibility lated EEG activity, it does not always prevent overlap. task, might reflect an interaction of attention and Using a visual stimulus display in which task-relevant movement-related processes, and thus be relevant to the stimuli are presented lateral from fixation, to create spatial explanation of compatibility effects. Praamstra and Plat S-R (in)compatibility, there will usually be a lateralisation [49] reported current source density analyses of the N2pc of visual evoked EEG responses, especially at electrode that showed current extrema located over occipital and sites overlying the occipital cortex. Moreover, when the central areas of the brain, suggesting that the central visual display contains multiple elements of which only extension of the N2pc might be due to concurrent activa- one is task-relevant, the visual response is followed by a tion of visual and motor areas. As a candidate motor area lateralised ERP component related to visuospatial selec- to generate this activity, they suggested the dorsal pre- tion. This attention-related ERP component has been motor cortex (PMd), a premotor area close to the motor labelled N2pc, related to its negative polarity, its occur- executional level but also known to exhibit visuospatial rence in the 200–300 ms latency range, and its distribution attention-related activity (e.g. Refs. [6,15,32]) and re- at posterior electrode sites contralateral to the location of garded as a site where visual and motor signals interact the stimulus [34,35]. The N2pc may extend from posterior [66,67]. electrode sites to central electrodes and precede or partially Summarising the above background, spatial S-R com- overlap the LRP [44,49,62,63]. To prevent this overlap, patibility tasks evoke a specific pattern of simultaneously some investigators have used spatial S-R compatibility occurring EEG potentials above posterior visual and P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 311 central motor areas. The posterior activity, occurring in isolation as N2pc, has been well characterised functionally [34,35] and anatomically [26]. With respect to the centrally distributed activity, from here on designated as N2cc (central contralateral), uncertainty remains as to its origin in motor areas (1), and its functional characterisation as visuospatial attentional activity (2). These points were addressed in the following way. To clarify the origin of the N2cc, the present investigation used high-density EEG for the recording of brain electrical activity in a standard spatial S-R compatibility task. To evaluate the presumed attentional nature of the N2cc, the spatial S-R compatibili- Fig. 1. The four stimuli used in spatial S-R compatibility task (2-CRT ty task was complemented with a Go–Nogo task, using the task) and Go–Nogo task. The letters ‘A’ and ‘B’ appeared within the same stimuli. We anticipated that visuospatial attentional permanently displayed frames left and right of the fixation cross and orientation to the stimulus instructing for a Nogo-response instructed for a left and right hand response, respectively (2-CRT task) or would yield an N2cc without subsequent movement ex- for Go and Nogo response alternatives (Go–Nogo task). ecution-related LRP. Such an isolated N2cc would both support an attentional interpretation of this component and facilitate the identification of its neural origin. Analyses focused on the N2cc and N2pc, as lateralised event-related effect [50]. The letter ‘A’ instructed for a left and ‘B’ for a potentials (L-ERPs) derived by means of contralateral– right hand response, irrespective of the side of presentation ipsilateral subtractions, while the midline N2 ERP com- relative to the fixation cross. AL, BR, AR, and BL will be ponent was interrogated as an additional source of in- used as shorthand for the (data from) conditions where the formation. A was presented left (AL: left-compatible), the B right (BR: right-compatible), the A right (AR: left-incompat- ible), and the B left (BL: right-incompatible) conditions. 2 . Methods Stimuli were presented for 200 ms in blocks of 121 trials, of which the first trial was always discarded. Each 2 .1. Subjects and behavioral tasks block contained an equal number of stimuli of each condition and stimuli were presented in a random order. Ten subjects (three women and seven men; age 42 to 60 The trial duration varied pseudo-randomly between 1750 years; mean 49) gave their informed consent to participate and 2250 ms. Before the experimental session, subjects in the study. The experimental procedures were approved performed one practice block. During the experiment by the department’s ethical review board. Nine subjects subjects were seated upright in a comfortable chair and were right-handed, one was left-handed. instructed to maintain eye fixation on the fixation cross. Tasks were a Simon-type spatial S-R compatibility task, Responses were made by pressing a button underneath the implemented in the same way as in Praamstra and Plat left and right index finger. In the main task, i.e. the 2-CRT [49], and a control task where the same stimuli were task, subjects responded to each stimulus according to the assigned to Go–Nogo response alternatives in order to response assignments described above. In the control task, dissociate attentional activity from movement-related ac- i.e. the Go–Nogo task, the same stimuli were presented, tivity (cf. Ref. [63]). Since S-R compatibility is a factor but subjects responded either only to the ‘A’ stimuli, using common to both tasks, we will refer to the first task as their left hand, or only to the ‘B’ stimuli, using their right two-choice response task (2-CRT) and to the second task hand. Blocks in which ‘A’ was the Go stimulus (and ‘B’ as Go–Nogo task. Subjects were instructed to respond to the Nogo stimulus) were alternated with blocks with ‘B’ as visual stimuli presented on a computer screen at 1 m the Go stimulus (and ‘A’ the Nogo stimulus). The hand viewing distance. The stimuli (see Fig. 1) consisted of two corresponding to the Nogo stimulus in a block was placed rectangular frames displayed permanently in white lines in the subjects’ lap. against a grey background (white 41.7 cd/m2 ; grey 7.7 The 2-CRT task consisted of four blocks. The Go–Nogo cd/m2 ). The frames subtended 0.9830.98 of visual angle task consisted of eight blocks, so that 2-CRT task, Go task and were positioned with their centre at a distance of 1.08 condition and Nogo task condition had equal numbers of from a central fixation cross. In each trial, a letter (‘A’ or trials to enable comparison of EEG data. The tasks were ‘B’) was presented in one, and a filler (three horizontal presented in one session of |3 h duration, including bars) in the other frame. The filler ensures that only minor electrode application. The order of the tasks was counter asymmetries occur within the latency window of exogen- balanced across subjects, such that one half of the subjects ous visual evoked responses [49,63]. The presence of a started with the 2-CRT task and the other half of the filler does not alter the basic spatial S-R compatibility subjects with the Go–Nogo task. 312 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322

2 .2. Physiological recording and data processing trials were analysed with repeated measures analysis of variance, using the MANOVA procedure of SPSS [42]. EEG was recorded continuously with Ag/AgCl elec- The analysis was performed separately for the 2-CRT task trodes from 130 standard locations according to the 10-5 and the Go–Nogo task and included compatibility (com- extension of the international 10-10 electrode system patible versus incompatible S-R relation) and hand (left [1,43], relative to a vertex reference. Electrodes were versus right hand) as within-subjects variables. For the mounted in a nylon cap. Active electrodes (http:/ Go–Nogo task only data from the Go condition were /www.biosemi.com) made skin preparation redundant. entered in the reaction time analysis. Electrode impedances were not measured. Eye movements ERP analyses were also performed with repeated mea- were monitored by bipolar horizontal and vertical EOG sures analysis of variance, as further specified in the derivations. EEG and EOG signals were amplified between Results. The Greenhouse–Geisser correction on probability 0.16 and 128 Hz by BioSemi Active-One amplifiers and values was applied where appropriate. Analyses were sampled at 512 Hz. For each subject and condition the data performed separately for 2-CRT task and Go–Nogo task, were averaged with reference to stimulus-onset to form but also across tasks, where appropriate. In the analyses ERPs. Recordings were re-referenced to an off-line recon- across tasks, the Go and Nogo conditions of the Go–Nogo structed average mastoids reference. Individual trials con- task are designated each as ‘tasks’, yielding three levels for taining artefacts were rejected before averaging. In two the variable task. Analysis of the N2pc and N2cc in L-ERP subjects, eye blink and eye movement artefacts were waveforms was guided by our previous work, in which treated using a common EOG artefact correction approach current source density (CSD) analysis of the N2pc dis- [22]. tribution demonstrated occipital and central current ex- Lateralised event-related potentials (L-ERPs), i.e. the trema, centered around the electrodes PO7/PO8 and C3/ attention-related N2pc/N2cc and the movement-related C4, respectively [49]. Here, we first evaluated whether the LRP, were derived by, firstly, computing difference po- same CSD distribution was found in the S-R task of the tentials between homologous electrodes contra and ipsila- present study. Subsequently, we selected the electrode teral to the side of movement and, secondly, averaging the pairs PO7/PO8 and C3/C4 as representative electrode difference potentials associated with left and right hand sites for the N2pc and the N2cc, respectively. Amplitudes movements. L-ERPs were computed separately for com- were measured as the mean value between 240 and 250 patible and incompatible conditions. For homologous ms, representing a narrow window around the peak latency electrodes over left and right motor cortex (C3 and C4), of the N2pc and N2cc. the L-ERP derivation reads L-ERP5[(C42 Further analyses of the event-related potential

C3)left-hand movement1(C32C4) right-hand movement ]/2 (for re- waveforms included the amplitude of the central N2 views see Refs. [8,16]). In several analyses, the N2pc/ component determined in a window around the peak N2cc was computed relative to the direction of attention latency in the grand averaged data (240–250 ms). These (i.e. the standard procedure for the attention-related N2pc), analyses attempted to elucidate the relation of the N2cc, instead of the side of movement. Note that if this computa- obtained through the L-ERP derivation, with the N2 in tion is performed across compatible and incompatible unsubtracted ERP waveforms. The following analyses conditions, movement-related activity is subtracted out were performed: (1) an analysis of the N2 at midline from the resulting L-ERP waveforms (leaving only some electrode sites, grouped as frontal (Fpz, AFz), frontocentral residual activity due to a reaction time difference between (Fz, FCz), central (Cz, CPz), parietal (Pz, POz), and conditions). occipital (Oz, Iz). (2) Comparisons of the N2 distribution L-ERP waveforms, as computed according to the above on the basis of statistical significance maps (standard t-map formula, combine the activity of both hemispheres. Where option in ERP analysis software package BrainVision). (3) information about the entire scalp distribution was re- A further regional analysis of the N2 in the central midline quired, we made subtractions of relevant conditions instead region and the left and right sensorimotor regions, pooling of applying the L-ERP derivation. For instance, computa- nine electrodes for each region. The central midline region tion of the N2pc/N2cc and LRP for the compatible included the electrodes FCz, Cz, CPz, FFC1h, FFC2h, condition requires the data set where an ‘A’ was presented FCC1h, FCC2h, CCP1h, CCP2h. The left sensorimotor left (AL: left-hand compatible condition) and the data set region was formed by the electrodes FC1, C1, FFC3h, where a ‘B’ was presented right (BR: right-hand compat- FCC3h, CCP3h, FC3, C3, FCC5h, CCP5h. Its right ible condition). A subtraction of the AL and BR data sets hemisphere counterpart included electrodes FC2, C2, reveals the respective contributions of each hemisphere to FFC4h, FCC4h, CCP4h, FC4, C4, FCC6h, CCP6h. attention and movement-related asymmetries, which would not be separately identifiable in the L-ERP waveforms. 2 .4. Source characterisation 2 .3. Data analysis Source analysis was performed on grand averaged L- Response accuracy and the reaction times of correct ERP and ERP distributions using a forward model based P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 313 on the boundary element method (BEM) and a standard 3 .2. Event-related potentials realistically shaped head model. The head model was constructed from a high quality MRI used as standard head 3 .2.1. L-ERPs in 2-CRT task in fMRI analysis software (SPM). The electrode positions Lateralised event-related potential waveforms for the according to the 10-5 system [43] were determined on the 2-CRT task, obtained by contralateral–ipsilateral subtrac- surface of this head model using a computer algorithm tions, were characterised by three successive lateralised which simulates normal electrode placement referenced to components, as illustrated in Fig. 2. The earliest deflection anatomical landmarks. More details on the volume conduc- at 150–160 ms (1) was due to contralateral–ipsilateral tor model can be found in Oostenveld et al. [44]. amplitude differences of visual evoked potentials, pre- The equivalent current dipole model was constructed by sumably related to the non-identical visual stimulation to moving all vertices of the triangulated brain surface of the left and right hemifield. The subsequent deflection peaking BEM model 10 mm inward. The surface was re-meshed at |245 ms (2) represents the N2pc/N2cc, with a maxi- with 2562 vertices evenly distributed over the complete mum at lateral occipital electrode sites PO7 and PO8. The surface. Subsequently three orthogonal dipoles were placed distribution of the N2pc/N2cc extended from occipital to on each vertex, resulting in a source model with 7686 frontal and central electrode sites, where it was immedi- dipoles. The dipole strength of this source model was ately followed by the LRP (3), representing movement- estimated using an L2-norm linear estimate (LE) with related activity. zero-order Tikhonov regularisation. The choice of the l To illustrate the relation of the lateralised N2pc/N2cc to parameter determining the weight of the regularisation was the basic ERP waveforms from which they are derived guided by the L-curve [24]. Maps were constructed by through subtraction, Fig. 3 shows the ERPs of the 2-CRT plotting the combined strength of the three orthogonal task grouped according to stimulus side. Thus, data from dipoles at each vertex divided by the vertex surface, the conditions where the left hemispace was attended (AL resulting in a representation of the current source density and BL) were averaged, as were data for the conditions (CSD). where the stimuli were presented to the right (AR and BR). The signal-to-noise ratio in the data of individual This involved averaging across compatible and incompat- subjects was not sufficient to perform source analysis on ible conditions, but this appeared justified by an analysis of individual subjects’ data. The use of a standard realistically the AL, BR, AR, BL waveforms at the N2pc/N2cc peak shaped head model gives results comparable to individual latency, for the electrodes marked in Fig. 3. This analysis head models [19]. However, the application of this model showed that the variable compatibility did not influence the to grand averaged data will result in a slightly blurred N2pc/N2cc effect (effect of compatibility by stimulus side representation of the true source activity, due to inter- by hemisphere F(1,9),1). In spite of the suggestion, in individual differences in functional anatomy. Fig. 3, that the left hemisphere contributes more to the

3 . Results

3 .1. Behavioral measures

The 2-CRT task yielded a robust compatibility effect (F(1,9)520.26, P,0.005), with faster responses in the compatible than in the incompatible condition (means6S.D.: 411624 vs. 449635 ms). Response side had no effect on response latency (main effect of response side F(1,9),1) and did not influence the compatibility effect (interaction response side and compatibility F(1,9), 1). The error rate was 3.8%. It was influenced by a significant compatibility effect (F(1,9)59.78, P,0.05), Fig. 2. Butterfly plot of lateralised event-related potentials (L-ERPs) due to more errors in the incompatible than in the recorded in the 2-CRT task, showing waveforms from a subset of electrodes in a superimposition of compatible and incompatible con- compatible condition. ditions. Movement-related activity (the LRP) is indicated by the horizon- In the Go–Nogo task, the difference in response latency tal bar [3] and plotted upwards for both conditions. It is preceded by the between compatible and incompatible condition was con- N2pc/N2cc [2] which in the compatible condition (black traces) has the siderably smaller than in the 2-CRT task (380646 vs. same polarity as the LRP, but has opposite polarity in the incompatible 392642 ms). This compatibility effect proved, neverthe- condition (red traces) where direction of attention and response side differ. Lateralised visual-evoked potentials [1] have for both conditions less, significant (F(1,9)522.44, P,0.005). The number of the same polarity as the subsequent N2pc/N2cc. The L-ERP waveforms errors was very low (1.1%) and consisted mainly of shown were recorded from the positions indicated in the electrode layout missing responses. by black dots, each referred to its contralateral homologue. 314 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322

where attention was directed to the left were subtracted from the waveforms for the conditions where it was directed to the right hemispace (ARBR2ALBL), in order to isolate and map the distribution of the N2pc/N2cc above each hemisphere. This analysis, reported below in Section 3.2.4, showed bilateral activation foci in oc- cipitotemporal and sensorimotor areas, supporting the use of different labels, i.e. N2pc and N2cc, for the components overlying these areas.

3 .2.2. L-ERPs:2-CRT vs. Go–Nogo task Based on the scalp distribution of the N2pc/N2cc complex described above and in Section 3.2.4 below, analyses of the N2pc and N2cc across tasks were per- formed on selected electrode pairs PO7/PO8 and C3/C4, as sensors close to the occipital (N2pc) and sensorimotor area (N2cc) activations, respectively. The waveforms recorded at these electrode sites are represented in Fig. 4. The initial analysis used a three-way ANOVA with task (3), compatibility (2) and electrode (2) as factors. Note that this analysis used absolute values, ignoring the opposite polarity of N2pc/N2cc in compatible and incom- patible conditions, which is merely contingent on the derivation of these potentials relative to the side of Fig. 3. ERPs of the 2-CRT task, showing the relation of the N2pc/N2cc movement. A striking difference between N2cc and N2pc, effect to the basic ERP waveforms. The vertical grey bar is placed at the apparent in the figure, is that the N2pc at PO7/PO8 is not peak latency of the N2cc/N2pc, which coincides with the centrally distributed N2 in the ERP waveforms. Continuous line, stimulus side left affected by the variable compatibility whereas the N2cc (AL1BL); dashed line, stimulus side right (AR1BR). The illustrated amplitude at C3/C4 is, that is, in the Go–Nogo task waveforms were recorded from the electrodes marked on the electrode conditions. This difference was expressed in significant layout. compatibility by electrode (F(1,9)515.69, P,0.01) and compatibility by electrode by task interactions (F(2,18)5 N2pc/N2cc than the right hemisphere, this difference 34.08, P,0.001). Looking separately at the N2pc, there is failed to reach significance (F(1,9)53.95, P50.078). a marked difference in amplitude of the N2pc in the Nogo Supported by this analysis, waveforms for the conditions task condition relative to the Go task condition and the

Fig. 4. N2pc and N2cc as recorded at representative electrode sites PO7/PO8 and C3/C4 in the 2-CRT task and the Go–Nogo task. Continuous line, compatible condition; dashed line, incompatible condition. P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 315

2-CRT task (effect of task F(2,18)58.10, P,0.05). The condition, was considerably overestimated. The analyses reduced amplitude of the N2pc in the Nogo task condition reported in this section attempted to further clarify the indicates that simultaneously with the spatial selection (of behavior of the N2cc by examining its relation with the N2 letter vs. filler), the featural information that distinguishes ERP component. Go and Nogo stimuli (letter ‘A’ vs. letter ‘B’) is already The N2cc above the motor cortex coincides with a processed and manifested in the N2pc. central N2 component, as illustrated earlier in Fig. 3. The Subsequent analyses compared N2pc (electrode pair N2 at midline electrodes is represented in Fig. 5 for the PO7/PO8) and N2cc (electrode pair C3/C4) for each task different tasks. It was evaluated at midline electrodes with separately, using simple effect analyses in the form of a three-way ANOVA with factors task (3), compatibility two-way ANOVAs with factors compatibility (2) and (2) and region (five levels: frontal to occipital) (see electrode (2). In the 2-CRT task, the amplitudes of N2pc Methods). The amplitude of the N2 differed significantly and N2cc were of similar magnitude in compatible and across tasks (F(2,18)522.57, P,0.005). As can be ob- incompatible condition (effect of compatibility F,1; served in Fig. 5, this was due to a higher amplitude in the interaction compatibility by electrode F(1,9)53.85, P5 2-CRT task and the Nogo task condition, compared to the 0.082). In the Go task condition, by contrast, the compat- Go task condition. Differences in distribution between ible and incompatible N2cc amplitudes differed, mani- tasks did not reach significance (task by region interaction; fested in a significant compatibility by electrode interaction F(8,72)52.60, P.0.05). Neither was there an effect of (F(1,9)545.85, P,0.001). This interaction was also sig- compatibility or significant interactions involving this nificant in the Nogo task condition (F(1,9)519.40, P, variable. Note that for the Nogo task condition the N2 is 0.01). However, the amplitude relation between compatible followed, at frontal electrode sites, by a pronounced and incompatible N2cc was reversed between Go and additional negative deflection, known as the Nogo–N2 Nogo task conditions. That is, in the Go task condition (e.g. Ref. [28]). Collapsing compatible and incompatible only a compatible N2cc developed, while in the Nogo task conditions of each task, the N2 distribution was further condition there was only a small incompatible N2cc. This inverse relationship means that in the Go–Nogo task the N2cc did not develop, or was severely attenuated, when spatial attention was directed towards the inactive side, i.e. the side associated with the Nogo response. The N2cc requires, in other words, readiness for a manual response ipsilateral to the direction of spatial attention. In summary, these analyses yield three main points. Firstly, the clearly distinct effects of the variable com- patibility at PO7/PO8 and C3/C4 support the separation of N2pc and N2cc. Secondly, the interactive effects of the direction of attention and side of active response on the presence and the appearance of the N2cc in the Go–Nogo task indicates that the N2cc does not solely reflect a visuospatial attentional orientation. Thirdly, the robust modulation of the N2pc between Go and Nogo task conditions indicates that the outcome of the visual dis- crimination guiding the Go–Nogo decision is already manifested in the ERP signal associated with the spatial selection of the task relevant information.

3 .2.3. N2 analyses The analyses of L-ERP waveforms in the previous section show that the Go–Nogo task does not produce the same N2cc as found in the 2-CRT task, which argues against an interpretation of the N2cc as purely visuospatial attentional activation. The analyses did not address, how- ever, the marked difference in the entire L-ERP waveform between the 2-CRT task and the Go–Nogo task. Note, in particular, that the N2cc in the Go task condition (compat- Fig. 5. Midline ERPs for the 2-CRT task and the Go–Nogo task ible condition) is not followed by a well-defined LRP, conditions. The represented data are averaged across compatible and probably due to temporal overlap between the N2cc and incompatible data sets. The closed arrowhead indicates the N2; the open the LRP. This implies that the N2cc amplitude, for this arrowhead marks the Nogo–N2. 316 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 explored using the full electrode array and statistical significance maps (t-maps). The difference in amplitude between Nogo and Go task conditions was shown to have a frontal maximum, centered around electrode sites Fz and AFz (t-values.6, P,0.001), indicating that the decision to withhold a response invokes predominantly frontal activa- tion. The N2 amplitude difference between 2-CRT task and Go task condition, by contrast, had a central maximum around Cz spreading to bilateral sensorimotor areas (t- values.3, P,0.05). The N2 amplitude difference between 2-CRT task and Go task condition, with a distribution above the midline and bilateral sensorimotor areas, represents information relevant to the interpretation of the N2cc that is not available in the L-ERP waveforms. It specifically raises the question to what extent the behavior of the N2cc in the two tasks that required a motor response (2-CRT task and Go task condition) can be explained in terms of amplitude variations of the N2. Recall that there was a suggestion of temporal overlap of N2cc and LRP in the Go task condition. This overlap cannot be resolved with analyses of the L-ERP waveforms, but a tentative decomposition might be derived from the lateralisation of the N2, measured peak-to-peak (relative to preceding positive peak). Thus, central scalp areas were divided in sen- sorimotor regions contralateral (MC) and ipsilateral (MI) to the stimulus side, and a midline region (MM), each covered by an array of nine electrodes. Analyses of the mean N2 amplitudes for each region, using a three-way Fig. 6. (A) Graphs representing the N2 amplitude (peak-to-peak) at ANOVA (task by region by compatibility), were performed electrodes overlying lateral sensorimotor areas and midline. MC, contrala- teral sensorimotor area; MI, ipsilateral sensorimotor area; MM, medial for the N2 amplitude measured peak-to-peak. The N2 peak sensorimotor area. Continuous line, 2-CRT task; dashed line, Go task was quantified as the mean amplitude in a fixed interval of condition. (B) N2 for the 2-CRT task (left panel) and the Go task 240–250 ms. Based on the grand average waveforms, the condition (right panel) at electrodes Cz, C3, and C4, reconfigured as preceding positive peak was quantified as the mean recording sites at the midline, ipsilateral, and contralateral to stimulation. amplitude from 180 to 190 ms for the 2-CRT task, and as Continuous line, compatible condition; dashed line, incompatible con- dition. The grey bars in the left panel indicate the N2 peak-to-peak the mean amplitude in the window 200–210 ms for the Go lateralisation. task condition. Fig. 6A represents the peak-to-peak N2 amplitudes for each region, separately for compatible and incompatible conditions. Comparing across tasks, the N2 is evidently of higher amplitude and has a steeper regional the N2 lateralisation, reflected in a task by region by differentiation in the 2-CRT task relative to the low compatibility interaction (F(2,18)512.63, P,0.01). amplitude N2 in the Go task condition. In more detail, N2 peak-to-peak amplitudes were also measured separ- there is an N2 amplitude difference (main effect of task ately for C3 and C4 electrode sites and Fig. 6B enables an F(1,9)510.46, P,0.025), which is not confined to the appreciation of the relation between this measure and the midline, but extends to lateral sensorimotor areas, especial- overall lateralisation of the N2, i.e. the contralateral– ly contralateral to stimulation (task by region interaction ipsilateral peak amplitude difference as used in the N2cc. F(2,18)521.63, P,0.01). Compatibility just failed to The waveforms of the compatible Go task condition reach significance as main effect (F(1,9)54.38, P50.066), illustrate that the small peak-to-peak lateralisation of the but interactions involving this variable proved informative. N2 by no means accounts for the overall lateralisation. The Whereas for the 2-CRT task the amplitude relation be- lateralisation in terms of N2 peak-to-peak amplitude tween regions remains the same in compatible and incom- contributed no more than 45617% to the N2cc amplitude patible conditions, the amplitude relation of ipsi (MI) and in the compatible condition, fitting the earlier raised contralateral regions (MC) reversed between compatible suspicion of an overlap of the N2cc by the LRP. The and incompatible conditions for the Go task condition. higher amplitude lateralisation of the N2 in the 2-CRT Thus, stimulus side prevailed in the 2-CRT task and task, on the other hand, more closely matched the am- response side prevailed in the Go task condition to explain plitude of the corresponding N2cc, accounting for P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 317

Fig. 7. (A) Scalp potential distribution and current source density reconstruction of the N2 around peak latency (240–250 ms). Data averaged across compatible and incompatible conditions of spatial S-R compatibility task, after adjustments to make the data uniform across data sets with respect to the direction of attention. Left column, left side view; right column, top view. Isopotential line spacing 0.5 mV. (B) Scalp potential distribution and current source density reconstruction of the N2pc/N2cc showing a clear separation of occipito-temporal (N2pc) and central (N2cc) activation. Left column, left side view; right column, top view. Potential distribution and current density reconstruction based on a subtraction of the data represented in Fig. 3, averaged over 240–250 ms. Isopotential line spacing 0.25 mV.

78624% and 72622% of the N2cc amplitude in compat- 2-CRT task. Midline and lateral contributions to the ible and incompatible condition, respectively. These values contralaterally emphasised N2 are, in fact, already sug- were each significantly higher than the value for the gested in the potential distribution of the N2. A current compatible Go task condition in separate two-tailed t-tests source density estimate was made on the brain surface, (t53.8, df59, P50.004 and t53.2, df59, P50.012). using all the data of the 2-CRT task by mirroring the AL To summarise, the analyses of the N2 peak-to-peak and BL data sets, as if stimulation was on the right side, amplitude elucidate the relation of N2cc and LRP in the and averaging these data with the AR and BR data sets. L-ERP waveforms, and suggest that their overlap can be The result, represented in Fig. 7A, shows the midline N2 resolved by defining the N2cc in terms of peak-to-peak activation and the lateralisation towards the left hemi- lateralisation of the N2. While the earlier analyses indi- sphere, contralateral to the attended hemispace. cated that the N2cc was attenuated when visuospatial The activity underlying the lateralisation of the N2 can attention is directed to the side associated with the Nogo be emphasised by applying current source density analysis response, the N2 analyses reveal a more generally reduced on the lateralised N2pc/N2cc, separating it from the amplitude of the N2 in bilateral and midline sensorimotor midline N2 activity. This is shown in Fig. 7B, where the areas for the Go task condition relative to the 2-CRT task. N2pc/N2cc potential distribution was constructed by Moreover, the lateralisation of the N2 showed quantitative subtracting the combined 2-CRT conditions where atten- and qualitative differences between Go task condition and tion was oriented to the left (AL and BL) from those where 2-CRT task, in terms of peak-to-peak amplitude, which attention was oriented to the right (AR and BR). This justify the inference that movement-related activity is not results in a current source density estimate showing that strongly represented in the N2/N2cc. Taken together, the the N2cc activity is characterised by bilateral foci in N2cc analyses on the L-ERP waveforms and the N2 sensorimotor areas. The activity related to the N2pc is analyses show that a substantial N2/N2cc developed only visible in bilateral occipitotemporal areas. in the 2-CRT task and that it cannot, therefore, be Note that these analyses were performed on grand explained in terms of visuospatial attentional activation or average data, using a standard realistic head model, aiming movement-related activity alone. The view that we will at separation of the main contributing sources of activity develop below proposes that the N2/N2cc represents rather than at accurate localisation. In terms of separation, activity of premotor areas associated with the selection and the source analysis of L-ERPs at the latency of the N2pc/ suppression of competing responses. N2cc and the source analysis of the N2 ERP converge in demonstrating separate occipito-temporal and central acti- 3 .2.4. Source analyses vation foci for the N2pc and N2cc, respectively. The Based on the N2 and N2cc analyses, combined midline central midline activation in the N2 analysis is interpreted and lateral sensorimotor area activation were anticipated in as activity from medial frontal cortex, separate from the the current source density reconstructions of the N2 in the N2cc activation foci at the lateral convexity. 318 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322

4 . Discussion motor areas at the lateral convexity, or simultaneous activation of medial and lateral motor structures. The N2 The goal of this study was to characterise early lateral- amplitude and distribution differences between the spatial ised motor area activation, represented in the N2cc, as S-R compatibility task and the Go task condition support recorded in a spatial S-R compatibility task. It was that both medial and lateral motor areas are activated and hypothesised that the N2cc reflects visuospatial attentional contributing to the N2 in the spatial S-R compatibility task. activity. Analyses of the N2cc and the N2 ERP component This inference was corroborated by the current source confirmed an N2cc distribution over motor areas of the density reconstructions (see Fig. 7). Although the current cortex, spatially contiguous to N2 activity overlying me- density reconstructions do not allow a precise identification dial frontal cortex. Comparison of spatial S-R compatibili- of the neural sources, it can be inferred from the dis- ty task and Go–Nogo task showed that the N2cc depends tribution of the N2 that the sources have a radial orienta- on bilateral response readiness, i.e. a competition between tion. For the contribution to the N2 that comes from the responses defined on opposite hands, ruling out a purely lateral convexity, this implies that the source is more likely visuospatial attentional role. As we will develop below, the in the lateral premotor cortex than in the primary motor found characteristics are best explained in terms of the cortex. The former, in contrast to the latter, is located on N2cc reflecting lateral premotor cortex involvement in the crown of the precentral gyrus, to the effect that response selection, which, broadly conceived, includes compound neural activity yields a radially oriented equiva- sensory-attentional neural activation contingent on the lent current dipole [20,51]. The primary motor cortex is motor significance of stimuli. predominantly located on the anterior wall of the central sulcus, and is represented by a tangentially orientated 4 .1. Neural origins of the N2cc dipole in movement-related EEG studies (e.g. Refs. [3,47]). Several earlier studies have reported data indicating that The joint activation of medial premotor and lateral the N2pc event-related potential, associated with visuospa- premotor cortex, proposed to underlie the N2/N2cc in the tial attentional selection, can demonstrate a characteristic 2-CRT task, is also found in neuroimaging studies involv- anterior extension to central and frontocentral electrode ing manipulations of spatial S-R compatibility [9,13]. sites [16,49,62,63]. This anterior extension was reproduced Relevant medial premotor areas include the pre-supple- here. In a dipole source simulation study [43], we have mentary motor area and the anterior cingulate cortex shown that this distribution is not likely explained by [9,13]. One study reported lateral premotor cortex activa- volume conduction, as proposed by other investigators tion, but without concomitant medial premotor activation [16,62]. The current source density analysis applied here [27]. The lateral premotor area that is most consistently resolves its distribution in an occipital and a central current implicated is the dorsal premotor cortex (PMd). Besides density focus, reproducing more basic analyses on data evidence from neuroimaging, there is a wealth of data from recorded with lower spatial resolution [49]. The current single-cell neurophysiology that relate the PMd to spatial density analyses thus provide support for our introduction attention and visual spatial sensorimotor processing of the label N2cc to designate the central extension of the [12,58,66,67]. This makes the PMd the most likely source N2pc. The identification of occipital and central current responsible for the lateralisation of the N2, yielding the density foci, while consistent across different studies and N2cc in contralateral–ipsilateral subtractions. study populations, does not entirely rule out the possibility that intervening parietal areas of cortex contribute to the 4 .2. Functional characterisation of the N2cc N2pc/N2cc complex. A parietal contribution, in fact, has been taken for granted by Wascher and co-workers, based Relevant evidence regarding the functional significance on their reading of scalp potential distributions very similar of the early motor area activation was obtained from a to our data [63–65]. While our results give reason to comparison between the 2-CRT task and the Go–Nogo question this interpretation, it should be pointed out that a task. Under the hypothesis that the N2cc reflects visuospa- recent investigation into the neural sources of the N2pc did tial attentional activity, we had anticipated a similar N2cc find evidence for parietal involvement, presumed to reflect in both tasks. However, in the Go–Nogo task N2cc-like the initiation of attention shifts [26]. This was identified activity failed to develop when attention had to be directed with MEG, however, and could only be reconstructed as a towards the hemispace corresponding to the (inactive) side possible contribution to the N2pc EEG potential on the of the body that was assigned the Nogo response. This basis of computations using the MEG information. rules out an interpretation of the N2cc purely in terms of Whereas the analysis of contralateral–ipsilateral subtrac- visual spatial attention. Apparently, a side-specific re- tions focused on the separation of N2pc and N2cc, looking sponse readiness is required for the N2cc. An attentional at the ERPs that form the basis of these subtractions, it is interpretation of the N2cc can therefore only be upheld if a more relevant to ask whether the midline N2 activation, modulation of visuospatial attention by response-related spreading to the hemisphere contralateral to stimulation, factors is accepted. represents activation of medial motor areas, activation of Two recent EEG studies on directing covert spatial P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 319 attention provide an opportunity for comparison [25,40]. make a response with the hand ipsilateral to the direction Both used centrally presented arrow cues that instructed of attention. However, this still yields only a small participants to attend covertly to either a left or right visual amplitude N2cc. Hence, we believe that a further ingredi- field location, in anticipation of stimuli presented there ent to any task eliciting a substantial N2cc, is a selection after a variable interval. Nobre et al. [40] found a between response alternatives defined at opposite hands. frontocentral hemispheric asymmetry as a function of cue Together, this set of conditions defines a combination of direction in the time window between 200 and 320 ms. attentional and response-related properties for the N2cc Moreover, this effect was continuous with a similar that would not seem incompatible with our hypothesis (see asymmetry maximal over posterior scalp sites, together Introduction and Ref. [49]) of the N2cc originating in the resembling a somewhat attenuated version of our N2pc/ dorsal premotor cortex (PMd). N2cc effects. Nobre et al. interpreted the frontocentral The following properties of the PMd stand out as most asymmetry as activation of premotor areas, as found in relevant. First is the context-dependent nature of PMd imaging studies of covert visual spatial attention (e.g. Refs. neuronal activity. That is, even early responses that appear [10,21,41]). Contrary to the results of Nobre et al., the sensory-attentional (e.g. given their temporal relationship study by Hopf and Mangun [25] did not yield evidence of with a stimulus or because the activation covaries with frontocentral asymmetries in the 200–300 ms latency stimulus location), are significantly influenced by the range. This absence may be related to the fact that cue- motor significance of the eliciting stimulus [4,12,15,58]. validity was only 50% (80% in Nobre et al.) and to a For instance, the activity of PMd cells may be modulated longer cue–target interval than Nobre et al. used. If this is by whether a stimulus only guides the orientation of the case, the combined results of these studies underscore attention or instructs a movement [15,67]. A second the point made above that an attentional interpretation of relevant aspect of PMd function is that it is specifically the N2cc needs to accommodate a strong influence of engaged in conditional motor tasks where visual spatial response-related factors. processing serves the selection of responses based on Before examining how the N2cc might combine vis- non-spatial stimulus characteristics [66]. Phrased different- uospatial attentional and response-related properties, we ly, it helps to inhibit location-based response tendencies should address the possibility that the early motor area [39,48,54] to support response choice on the basis of activation represented by the N2cc just reflects movement- arbitrary linkages between stimulus and response [45]. related activity. Whereas Wascher and Wauschkuhn [63], in Finally, very recent work demonstrated that the PMd can a seminal investigation of the activity that we designate as simultaneously encode multiple potential reach directions N2cc, tended towards an attentional interpretation, Wascher [7]. Analogous to processes in the frontal eye field where et al. [65] seem to regard the N2cc as an early deflection of saccade target selection occurs through suppressive inter- the movement-related LRP. Our results argue against such actions between representations of multiple potential an interpretation. Recall that in the compatible Go task targets [55], this could mean that the PMd’s participation condition (in contrast to the incompatible condition) there in response selection proceeds through similar competitive was a sizeable lateralisation expressed in the C3/C4 L- interactions between potential movements. ERP waveforms at the N2pc/N2cc latency, comparable in amplitude to the N2cc in the 2-CRT task. However, our 4 .3. N2pc and N2cc: visual selection and response analyses of the N2cc in terms of N2 peak-to-peak am- selection plitude revealed that this lateralisation was only for a small part explained by the N2, the main part being due to An unanswered question so far is why the N2cc and overlap with the movement-related LRP. In the 2-CRT N2pc coincide in time. If the N2cc reflects visuospatial task, by contrast, the same decomposition showed that it information processing for response selection, the temporal was the N2 peak-to-peak amplitude which contributed coincidence would suggest a parallel evolution of visual most to the overall lateralisation in the L-ERP waveform. spatial selection and response selection. One important We conclude that the N2cc, defined as a lateralisation of aspect of response selection in the spatial S-R compatibili- the N2 peak-to-peak, can be differentiated from the ty task is maintaining independent control of the direction movement-related LRP. Furthermore, the N2cc thus de- of spatial attention and manual response activation, to fined remains very small in the Go–Nogo task, even when prevent manual response selection to passively follow the attention is directed to the hemispace corresponding with direction of spatial attention. As discussed in the previous the side of the Go-response. section, the PMd is regarded as a pivotal structure for the Having ruled out a narrow attentional interpretation and central nervous system’s ability to dissociate attention, an explanation as movement-related activity, we can now gaze and limb movements in space [45,66]. An engage- list the conditions that determine the generation of the ment of the PMd, to prevent such motor system cross-talk, N2cc. A shift of visuospatial attention to an attention- would therefore seem appropriately timed when it occurred attracting stimulus is not sufficient to elicit an N2cc, simultaneously with the visuospatial selection or attention- irrespective of whether the stimulus requires a response or al focusing that is represented in the N2pc. not. What is minimally needed, in addition, is readiness to Is the role outlined here for the N2cc, including its 320 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 temporal relation to the N2pc, consistent with existing processing in PMd as here outlined, how should we view views on the function and the neurophysiological under- its relation to early LRP deflections, occurring in the same pinnings of the N2pc? The spatial filtering in visual time window, that investigators have identified as EEG cortical areas, reflected in the N2pc, is considered a correlates of automatic location-based response activation process under top-down control. This means that the [14,60,62]? Note that most investigations used a vertical attention system is primed for relevant stimulus features stimulus arrangement to prevent N2cc-like activity to the detection of which, through an initial feature analysis overlap the LRP (see Introduction). Typical values of the that sends information back re-entrantly to extrastriate early LRP deflections regarded as automatic response occipital cortex, leads to preferential processing of in- activation are substantially smaller than the amplitude of formation from a particular spatial location [17,36]. On the N2cc [14,60,62]. These early LRP deflections might this view of the spatial filtering process reflected in the therefore correspond to that part of the N2cc in L-ERP N2pc, the temporal coincidence of N2pc and N2cc can be waveforms that is not accounted for by the N2 peak-to- accommodated by assuming that the information from an peak amplitude. This would imply that automatic response initial feature analysis is not only used for visual spatial activation is more likely mediated by the primary motor selection and fed back to the extrastriate cortex, but is cortex than by the PMd, and that it should be viewed as already available for response selection processes as well. activity that escapes the control of location-based response This information being available to the motor system very tendencies attributed to the PMd. early is not a new notion [8]. Moreover, with respect to the The above account is probably incomplete. Spatial S-R PMd and its role in response selection, Crammond and correspondence can be expected to confer an initial Kalaska [12] emphasised that considerable processing of advantage on the compatible response in the competition sensory stimuli must take place before activation of PMd between two potential responses that we assume take place cells, as even attributes like spatial location are not in PMd and regard as reflected in the N2cc. Hence, represented unconditionally, but contingent on the motor location-based response activation, as well as its inhibition, instructional content of a stimulus. In short, it seems are an integral part of the response selection process conceivable that both the attentional focusing process represented in the N2cc. This is supported by the N2 reflected in the N2pc and the visual spatial response waveforms illustrated in Fig. 7B (left panel) for the 2-CRT processing reflected in the N2cc initially operate on partial task. At the electrode site overlying the motor cortex stimulus information that is accessed by both processes at contralateral to stimulation (MC), the waveforms for the same time, yielding temporally coincident N2pc and compatible and incompatible condition superimpose exact- N2cc deflections. ly up to a latency of more than 350 ms, i.e. well beyond Beyond the temporal coincidence of the N2pc and N2cc, the latency where movement-related motor cortex activa- there may be similarities between the processes that tion must have started for the compatible response. This underlie these components. Visuospatial selection in visual implies that the composition of neural activity underlying areas has been shown to involve inhibitory interactions the N2cc differs between compatible and incompatible between the neural responses to neighbouring visual conditions, with inhibitory activity being more strongly stimuli presented simultaneously [29,30,37]. The N2pc represented in the latter compared to the former condition. exhibits properties consistent with such a mechanism [36], A final point concerns the N2cc and disinhibition of and relevant inhibitory interactions may extend as far as automatic response activation under pathological condi- the opposite hemisphere [46], as required to explain the tions. Dysfunction of the PMd compromises a monkey’s N2pc [37]. As already referred to in the previous section, ability to suppress location-based response tendencies the PMd can simultaneously prepare for multiple potential [23,54]. We have attributed Parkinson’s disease patients’ movements [7]. The selection of an intended movement impaired performance on a spatial S-R compatibility task might be accomplished in the same fashion as the atten- also to dysfunction of the PMd, partly on the basis of an tional selection of a visual target in extrastriate areas or the enhanced N2cc [48,49]. Basal ganglia dysfunction in selection of a target for an eye movement in the frontal Parkinson’s disease leads to impaired focusing of neural eyefields [55], i.e. through inhibitory interactions between activity in basal ganglia and cortex [38]. Based on this multiple potential movements that are pre-activated in a pathophysiology, we have proposed that visuospatial atten- particular context. Relevant here, interhemispheric inhib- tional activation may not be kept sufficiently focused itory interactions between motor areas, to mediate the within the PMd and cause spurious activation of competition between movements of opposite hands, appear movement-related cells, manifested in an enhanced N2cc to be active in the time window of the N2cc, as measured [49]. Such a reconstruction appears consistent with evi- in choice response tasks using transcranial magnetic dence for a joint role of basal ganglia and PMd in stimulation (TMS) [33]. visuospatial motor processing [5,66] (see also Ref. [2]). A subdivision of the PMd in rostral and caudal parts, the 4 .4. N2cc and automatic response activation former more closely involved with spatial attentional/ mnemonic processes and the latter closer to the motor If the N2cc is related to the type of visuospatial motor executional level [6,59], should not be regarded as con- P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322 321 tradictory to this account, given that the direction of spatial following attentional vs. intentional cues, Exp. Brain Res. 95 (1993) attention can influence saccade as well as manual reach 15–27. trajectories in the absence of pathology [53,56,61]. [5] D. Boussaoud, I. Kermadi, The primate striatum: neuronal activity in relation to spatial attention versus motor preparation, Eur. J. Neurosci. 9 (1997) 2152–2168. [6] D. Boussaoud, Attention versus intention in the primate premotor 5 . Conclusions cortex, NeuroImage 14 (2001) S40–S45. [7] P. Cisek, J.F. Kalaska, Simultaneous encoding of multiple potential reach directions in dorsal premotor cortex, J. Neurophysiol. 87 This investigation aimed to elucidate the origin and (2002) 1149–1154. functional significance of early motor area activation in a [8] M.G.H. Coles, H.G.O.M. Smid, M.K. Scheffers, L.J. Otten, Mental spatial S-R compatibility task, represented in the N2cc chronometry and the study of human information processing, in: lateralised ERP component. The results justify our intro- M.D. Rugg, M.G.H. Coles (Eds.), Electrophysiology of Mind: duction of the label N2cc, since mere volume conduction Event-related Brain Potentials and Cognition, Oxford University Press, Oxford, 1995, pp. 86–131. of the N2pc towards central electrodes cannot account for [9] J.D. Connolly, M.A. Goodale, J.F.X. Desouza, R.S. Menon, T. Vilis, the task-related modulations in the waveform amplitude. A comparison of frontoparietal fMRI activation during anti-saccades This corroborates a claim to this effect that we made and anti-pointing, J. Neurophysiol. 84 (2000) 1645–1655. earlier on the basis of a dipole source simulation [42]. [10] M. Corbetta, F.M. Miezin, G.L. Shulman, S.E. Petersen, A PET Although the source analysis results, taken in isolation, do study of visuospatial attention, J. Neurosci. 13 (1993) 1202–1226. [11] D.J. Crammond, J.F. Kalaska, Modulation of preparatory neuronal not allow us to pinpoint the origin of the N2cc very activity in dorsal premotor cortex due to stimulus–response com- precisely, with converging evidence from neuroimaging patibility, J. Neurophysiol. 71 (1994) 1281–1284. and single cell neurophysiology it can be narrowed down [12] D.J. Crammond, J.F. Kalaska, Prior information in motor and with some confidence to the lateral premotor cortex, where premotor cortex: activity during the delay period and effect on the most likely area involved in visuospatial motor pro- pre-movement activity, J. Neurophysiol. 84 (2000) 986–1005. [13] P. Dassonville, S.M. Lewis, X.H. Zhu, K. Ugurbil, S.G. Kim, J. cessing is the PMd. Ashe, The effect of stimulus–response compatibility on cortical The view developed here reiterates insights from single motor activation, NeuroImage 13 (2001) 1–14. cell neurophysiology and monkey studies that the lateral [14] R. De Jong, C.-C. Liang, E. Lauber, Conditional and unconditional premotor cortex is implicated in response selection rather automaticity: a dual-process model of effects of spatial stimulus– than just the preparation of a selected motor response, and response correspondence, J. Exp. Psychol. Hum. Percept. Perform. attributes the N2cc a role in the selection and suppression 20 (1994) 721–750. [15] G. Di Pellegrino, S.P. Wise, Visuospatial versus visuomotor activity of competing responses, mirroring the selection and sup- in the premotor and prefrontal cortex of a primate, J. Neurosci. 13 pression of relevant vs. irrelevant/competing visuospatial (1993) 1227–1243. information reflected in the N2pc. The proposed view [16] M. Eimer, The lateralized readiness potential as an on-line measure acknowledges the operation of automatic, location-based, of central response activation processes, Behav. Res. Methods response tendencies. While we do not take issue, therefore, Instrum. Comput. 30 (1998) 146–156. [17] M. Eimer, Mechanisms of visuospatial attention: evidence from with a possible expression of such tendencies in EEG as event-related brain potentials, Visual Cogn. 5 (1998) 257–286. early deflections of the LRP, we have argued that they do [18] P.M. Fitts, C.M. Seeger, S-R compatibility: spatial characteristics of not exhaustively explain the N2cc, and that inhibitory stimulus and response codes, J. Exp. Psychol. 46 (1953) 199–210. activity recruited to control location-based response activa- [19] M. Fuchs, J. Kastner, M. Wagner, S. Hawes, J.S. Ebersole, A tion is also expressed in this component. standardized boundary element method volume conductor model, Clin. Neurophysiol. 113 (2002) 702–712. [20] C. Gerloff, N. Uenishi, M. Hallett, Cortical activation during fast repetitive finger movements in humans: dipole sources of steady- A cknowledgements state movement-related cortical potentials, J. Clin. Neurophysiol. 15 (1998) 502–513. A. van Oosterom and D. Stegeman advised on the [21] D.R. Gitelman, A.C. Nobre, T.B. Parrish, K.S. LaBar, Y.H. Kim, J.R. Meyer, M. Mesulam, A large-scale distributed network for source analysis. P. Praamstra was supported by MRC grant covert spatial attention: further anatomical delineation based on no. G9901257. stringent behavioural and cognitive controls, Brain 122 (1999) 1093–1106. [22] G. Gratton, M.G.H. Coles, E. Donchin, A new method for off-line removal of ocular artifact, Electroencephalogr. Clin. Neurophysiol. R eferences 55 (1983) 468–484. [23] U. Halsband, R.E. Passingham, Premotor cortex and the conditions [1] American Electroencephalographic Society, Guideline 13: guide- for movement in monkeys (Macaca mulatta), Behav. Brain Res. 18 lines for standard electrode position nomenclature, J. Clin. Neuro- (1985) 269–276. physiol. 11 (1994) 111–113. [24] P.C. Hansen, Rank deficient and discrete ill-posed problems, in: [2] M.A. Basso, R.H. Wurtz, Neuronal activity in substantia nigra pars Society for Industrial and Applied Mathematics: Monographs on reticulata during target selection, J. Neurosci. 22 (2002) 1883–1894. Mathematical Modeling and Computation, SIAM, Philadelphia, [3] K. Botzel,¨ H. Plendl, W. Paulus, M. Scherg, : is 1997. there a contribution from the supplementary motor area?, Elec- [25] J.-M. Hopf, G.R. Mangun, Shifting visual attention in space: an troencephalogr. Clin. Neurophysiol. 89 (1993) 187–196. electrophysiological analysis using high spatial resolution mapping, [4] D. Boussaoud, S.P. Wise, Primate frontal cortex: neuronal activity Clin. Neurophysiol. 111 (2000) 1241–1257. 322 P. Praamstra, R. Oostenveld / Cognitive Brain Research 16 (2003) 309–322

[26] J.-M. Hopf, S.J. Luck, M. Girelli, T. Hagner, G.R. Mangun, H. [48] P. Praamstra, B.-U. Kleine, A. Schnitzler, Magnetic stimulation of Scheich, H.-J. Heinze, Neural sources of focused attention in visual the dorsal premotor cortex modulates the Simon effect, NeuroReport search, Cereb. Cortex 10 (2000) 1233–1241. 10 (1999) 3671–3674. [27] M. Iacoboni, R.P. Woods, J.C. Mazziotta, Bimodal (auditory and [49] P. Praamstra, F.M. Plat, Failed suppression of direct visuomotor visual) left frontoparietal circuitry for sensorimotor integration and activation in Parkinson’s disease, J. Cogn. Neurosci. 13 (2001) sensorimotor learning, Brain 121 (1998) 2135–2143. 31–43. [28] E. Jodo, Y. Kayama, Relation of a negative ERP component to [50] R.W. Proctor, C.H. Lu, Referential coding and attention-shifting response inhibition in a Go/Nogo task, Electroencephalogr. Clin. accounts of the Simon effect, Psychol. Res. 56 (1994) 185–195. Neurophysiol. 82 (1992) 477–482. [51] J. Rademacher, V.S. Caviness, H. Steinmetz, A.M. Galaburda, [29] S. Kastner, P. De Weerd, R. Desimone, L.G. Ungerleider, Mecha- Topographic variation of the human primary cortices: implications nisms of directed attention in the human extrastriate cortex as for neuroimaging, brain mapping and neurobiology, Cereb. Cortex 3 revealed by functional MRI, Science 282 (1998) 108–111. (1993) 313–329. [30] S. Kastner, L.G. Ungerleider, Mechanisms of visual attention in the [52] A. Riehle, S. Kornblum, J. Requin, Neuronal correlates of sen- human cortex, Annu. Rev. Neurosci. 23 (2000) 315–341. sorimotor association in stimulus–response compatibility, J. Exp. [31] S. Kornblum, T. Hasbroucq, A. Osman, Dimensional overlap: Psychol. Hum. Percept. Perform. 23 (1997) 1708–1726. cognitive basis for stimulus–response compatibility—A model and [53] S. Rubichi, R. Nicoletti, C. Iani, C. Umilta, The Simon effect occurs taxonomy, Psychol. Rev. 97 (1990) 253–270. relative to the direction of an attention shift, J. Exp. Psychol. Hum. [32] M.A. Lebedev, S.P. Wise, Tuning for the orientation of spatial Percept. Perform. 23 (1997) 1353–1364. attention in dorsal premotor cortex, Eur. J. Neurosci. 13 (2001) [54] T. Sawaguchi, I. Yamane, K. Kubota, Application of the GABA 1002–1008. antagonist bicuculline to the premotor cortex reduces the ability to [33] L. Leocani, L.G. Cohen, E.M. Wassermann, K. Ikoma, M. Hallett, withhold reaching movements by well trained monkeys in visually Human corticospinal excitability evaluated with transcranial mag- guided reaching task, J. Neurophysiol. 75 (1996) 2150–2156. netic stimulation during different reaction time paradigms, Brain [55] J.D. Schall, D.P. Hanes, K.G. Thompson, D.J. King, Saccade target 123 (2000) 1161–1173. selection in frontal eye field of Macaque. I.Visual and premovement [34] S.J. Luck, S.A. Hillyard, Electrophysiological correlates of feature activation, J. Neurosci. 15 (1995) 6905–6918. analysis during visual search, Psychophysiology 31 (1994) 291– [56] B.M. Sheliga, L. Craighero, L. Riggio, G. Rizzolatti, Effects of 308. spatial attention on directional manual and ocular responses, Exp. [35] S.J. Luck, S.A. Hillyard, Spatial filtering during visual search: Brain Res. 114 (1997) 339–351. evidence from human electrophysiology, J. Exp. Psychol. Hum. [57] L. Shen, G.E. Alexander, Neural correlates of a spatial sensory-to- Percept. Perform. 20 (1994) 1000–1014. motor transformation in primary motor cortex, J. Neurophysiol. 77 [36] S.J. Luck, M. Girelli, M.T. McDermott, M.A. Ford, Bridging the (1997) 1171–1194. gap between monkey neurophysiology and human perception: an [58] L. Shen, G.E. Alexander, Preferential representation of instructed ambiguity resolution theory of visual selective attention, Cogn. target location versus limb trajectory in dorsal premotor area, J. Psychol. 33 (1997) 64–87. Neurophysiol. 77 (1997) 1195–1212. [37] S.J. Luck, L. Chelazzi, S.A. Hillyard, R. Desimone, Neural mecha- [59] S.R. Simon, M. Meunier, L. Piettre, A.M. Berardi, C.M. Segebarth, nisms of spatial selective attention in areas V1, V2, and V4 of D. Boussaoud, Spatial attention and memory versus motor prepara- Macaque visual cortex, J. Neurophysiol. 77 (1997) 24–42. tion: premotor cortex involvement as revealed by fMRI, J. Neuro- [38] J.W. Mink, The basal ganglia: focused selection and inhibition of physiol. 88 (2002) 2047–2057. competing motor programs, Prog. Neurobiol. 50 (1996) 381–425. [60] B. Sturmer,¨¨ H. Leuthold, E. Soetens, H. Schroter, W. Sommer, [39] L. Moll, H.G. Kuypers, Premotor cortical ablations in monkeys: Control over location-based priming in the Simon task: behavioral contralateral changes in visually guided reaching behavior, Science and electrophysiological evidence, J. Exp. Psychol. Hum. Percept. 198 (1977) 317–319. Perform., in press. [40] A.C. Nobre, G.N. Sebesteyn, C. Miniussi, The dynamics of shifting [61] S.P. Tipper, L.A. Howard, G. Houghton, Behavioral consequences visuospatial attention revealed by event-related potentials, Neuro- of selection from neural population codes, in: S. Monsell, J. Driver psychologia 38 (2000) 964–974. (Eds.), Attention & Performance XVIII: Control of Cognitive [41] A.C. Nobre, G.N. Sebesteyn, D.R. Gitelman, M.M. Mesulam, R.S.J. Processes, MIT Press, Boston, 2000, pp. 223–245. Frackowiak, C.D. Frith, Functional localization of the system for [62] F. Valle-Inclan,´ The locus of interference in the Simon effect: an visuospatial attention using positron emission tomography, Brain ERP study, Biol. Psychol. 43 (1996) 147–162. 120 (1997) 515–533. [63] E. Wascher, B. Wauschkuhn, The interaction of stimulus- and [42] R.G. O’Brien, M.K. Kaiser, MANOVA method for analyzing response-related processes measured by event-related lateralizations repeated measures designs: an extensive primer, Psychol. Bull. 97 of the EEG, Electroencephalogr. Clin. Neurophysiol. 99 (1996) (1985) 316–333. 149–162. [43] R. Oostenveld, P. Praamstra, The five percent electrode system for [64] E. Wascher, M. Reinhard, B. Wauschkuhn, R. Verleger, Spatial S-R high-resolution EEG and ERP measurements, Clin. Neurophysiol. compatibility with centrally presented stimuli: an event-related 112 (2001) 713–719. asymmetry study on dimensional overlap, J. Cogn. Neurosci. 11 [44] R. Oostenveld, P. Praamstra, D.F. Stegeman, A. Van Oosterom, (1999) 214–229. Overlap of attention and movement-related activity in lateralized [65] E. Wascher, U. Schatz, T. Kuder, R. Verleger,Validity and boundary event-related brain potentials, Clin. Neurophysiol. 112 (2001) 477– conditions of automatic response activation in the Simon task, J. 484. Exp. Psychol. Hum. Percept. Perform. 27 (2001) 731–751. [45] R.E. Passingham, The Frontal Lobes and Voluntary Action, Oxford [66] S.P. Wise, G. Di Pellegrino, D. Boussaoud, The premotor cortex and University Press, Oxford, 1993. nonstandard sensorimotor mapping, Can. J. Physiol. Pharmacol. 74 [46] R.J. Perry, S. Zeki, The neurology of saccades and covert shifts in (1996) 469–482. spatial attention. An event-related fMRI study, Brain 123 (2000) [67] S.P. Wise, D. Boussaoud, P.B. Johnson, R. Caminiti, Premotor and 2273–2288. parietal cortex: corticocortical connectivity and combinatorial [47] P. Praamstra, D.F. Stegeman, M.W.I.M. Horstink, A.R. Cools, computations, Annu. Rev. Neurosci. 20 (1997) 25–42. Dipole source analysis suggests selective modulation of the supple- [68] J. Zhang, A. Riehle, J. Requin, S. Kornblum, Dynamics of single mentary motor area contribution to the readiness potential, Elec- neuron activity in monkey primary motor cortex related to sen- troencephalogr. Clin. Neurophysiol. 98 (1996) 468–477. sorimotor transformation, J. Neurosci. 17 (1997) 2227–2246.