Abstract

Visual exploration is organized in sequences of saccadic eye movements that depend on both perceptual and cognitive context. Using functional magnetic resonance imaging, we studied the neural basis of sequential oculomotor behavior and its dependence on different types of memory by analyzing cerebral activity during performance of newly learned and familiar sequences of eye movements. Compared to a resting condition, both types of sequences activated a common fronto-parietal network, including frontal and supplementary eye fields, and several parietal areas. Within this network, newly learned sequences induced stronger activation than familiar sequences, probably reflecting higher attentional demands. In addition, specific regions were recruited for the performance of new sequences, including pre-supplementary eye fields, the precuneus and the caudate nucleus. This indicates that in addition to attentional modulation, novelty of saccadic sequences requires specific cortical resources, probably related to effortful sequence preparation and coordination as well as to spatial working memory. For familiar sequences, recalled from long-term memory, we observed specific right medial temporo-occipital activation in the vicinity of the boundary between the parahippocampal and lingual gyri, as well as an activation site in the parieto-occipital fissure. We conclude that neuronal resources recruited by the gaze system can change with the familiarity of the scanpath to be executed. This study is important to better understand how the brain implements memorized scanpaths for visual exploration and orienting.

Introduction

Visual exploration consists of a succession of saccades, which can follow stereotyped, even idiosyncratic, paths — scanpaths — that depend upon both visual features and cognitive components (Yarbus, 1967; Andrews and Coppola, 1999). Scanpaths take the form of programmed sequences of saccades rather than successions of individually programmed eye movements (Zingale and Kowler, 1987). They play an active role in perception, recognition and mental imagery of visual scenes, and are related to previously made scans — and thus experience and memory (Yarbus, 1967; Noton and Stark, 1971; Brandt and Stark, 1997; Liversdedge and Findlay, 2000). The present study aimed at investigating the neural basis for the spatio-motor component of scanpath memory. For that purpose, we tried to reduce as much as possible any form of episodic memory due to the recall of the visual scene. We tested whether the performance of well-practiced scanpaths — such as those used in familiar contexts — and new scanpaths — such as those used in unfamiliar environments — recruit similar brain resources. Thus, we tested whether procedural learning of a saccadic sequence modulates oculomotor brain activity.

Numerous studies have shown that practice of a manual sequence can improve performance and modify corresponding brain activity (Grafton et al., 1992; Jenkins et al., 1994; Hikosaka et al., 1996). The effect of practice is reflected at two levels that have to be controlled in studies of sequential movement-processing. First, a global learning phase is characterized by an improvement in the execution of any new sequences when the subject is trained to learn new sequences (Passingham, 1993; Hikosaka et al., 1995), and this is accompanied by modulation of the movement-related brain activation (Karni et al., 1995). Second, sequence-specific ‘procedural’ learning occurs with repeated execution of a single sequence. Familiar (i.e. over-practiced) and novel manual sequences are associated with different patterns of cerebral activation, presumably reflecting the degree of procedural learning and the use of different modes of memory (Jenkins et al., 1994; Doyon et al., 1996; Juepner et al., 1997a).

The neural basis of oculomotor sequence learning and memory has been investigated only in two previous positrons emission tomography (PET) studies. Petit et al. (Petit et al., 1996) investigated the recall from short-term memory of oculomotor sequences in darkness compared with endogenous saccades, while Kawashima et al. (Kawashima et al., 1998) concentrated on the initial learning phase of visually-guided sequences (serial reaction time paradigm, trials 3–9) compared with single reflexive saccades. Both studies reported activity in the pre-supplementary motor area (pre-SMA), the deep intraparietal sulcus (IPS) and the precuneus. However, the exact role of these regions in oculomotor sequence execution and memory remains to be defined. Activation in pre-SMA and parietal areas could be involved in either (i) the sequential nature of the task; (ii) and/or global learning, in which subjects become familiar with the experimental task itself; (iii) or the processing of new oculo-motor sequences. In the first two cases, activity in pre-SMA, precuneus and IPS should not be influenced by the familiarity of the performed sequence. In the third case, however, activity in these regions might be less pronounced — or even absent — during the execution of automatic sequences recalled from long-term memory as compared with the execution of unfamiliar sequences requiring short-term memory. To our knowledge, no study has investigated activation in these regions when the performed oculomotor sequences become familiar.

The goal of the present study was to investigate the effect of previous learning on the performance of saccadic sequences, using functional magnetic resonance imaging (fMRI). We compared brain regions activated during performance of familiar (repeated several hundred of times) as opposed to newly learned saccadic sequences. Our hypothesis was that the brain activation patterns should differ since (i) performing familiar sequences involves automatic processes and recall from long-term memory; and (ii) performing new sequences requires use of short-term memory and presumably resources for the implementation or consolidation of a procedure (Curran and Keele, 2000). Training for several days allowed subjects to become familiarized with a particular sequence, and ensured that by the time of the scans, the global task-learning was essentially finished. Thus, we could conclude that any difference observed in brain activation between the performance of new and familiar sequences was specifically related to the sequence order and thus to the type of memory used.

Materials and Methods

Behavioral Task: Subjects and General Design

Ten right-handed healthy female subjects, aged between 22 and 30 years, with normal vision and without history of neurological disorders, gave their informed consent to participate in this study. The study was in line with the Declarations of Helsinki and approved by the French National Ethical Committee. The experiment was divided into two parts: (i) a 4 day training period in which subjects were familiarized with the experimental paradigm and were extensively trained with a specific sequence; and (ii) a functional imaging session on day 5, in which subjects performed both familiar and new saccadic sequences.

A saccadic sequence consisted of five horizontal saccades to five possible target locations, indicated by empty green squares on a black screen (Fig.1). Target locations were separated by a 4° visual angle. The amplitude of saccades could therefore be 4, 8, 12 or 16°. Every sequence ended at the central target position. The squares indicating target locations remained on the screen to ensure comparable saccadic amplitudes for new and familiar sequences.

Every sequence was introduced by successively lighting one of the five target squares (a square filled with dim green light). Subjects were asked to track these targets with their eyes and to memorize the sequence at the same time. This ‘visually guided’ (V) part was performed four times for each sequence. Afterwards, subjects had to perform the sequence four times from memory (M condition). The succession of V and M conditions for a given sequence was named a ‘sequence set’ (see Fig. 1A,B). The frequency of saccades was fixed to one saccade every 930 ms (1.07 Hz), paced by an audio signal. Subjects were instructed to wait for the audio signal before executing the next saccade, which allowed us to control for response rate and thus made both types of sequence comparable. The rhythm was chosen on the basis of pilot studies to ensure comfortable performance. It is compatible with the rate of natural scanning saccades. In addition, the amplitudes chosen for the saccadic sequences were in the same range as natural saccadic amplitudes during scene exploration (Bahill et al., 1975; Andrews and Coppola, 1999).

Two kinds of sequence sets were tested: a ‘familiar’ (Fam) set that subjects extensively practiced the days before the main experiment in the scanner (>400 repetitions); and several unfamiliar sequence sets (New). New sequence sets presented during fMRI scanning differed from those during training, and each was presented only once to ensure its novelty. As a whole, sequences were balanced with respect to the number of leftward and rightward saccades and amplitudes (i.e. two saccades in one direction in three in the other for each sequence; no more than three saccades of the same amplitude in a sequence). Forty-five different New sequences were used in the whole experiment. The Fam sequence set was the one shown in Figure 1A for five subjects; the mirror sequence was presented for the five remaining subjects.

Learning and Psychophysical Measurements

Subjects underwent daily hour-long training sessions on four consecutive days. Every session consisted of three different training steps. First, subjects repeated an alternation of Fam and New sequence sets, in the way described above. This first training step was designed to familiarize subjects with the general task (‘learn to remember sequences of saccades’), and with the alternation of familiar and new sequence sets. In the second training step, the Fam sequence set was repeatedly performed alone, eight times in a row, to extensively train subjects in this sequence. In the third training step, subjects repeated the task with alternated Fam and New sets, but this time, no audio signal was given, and subjects were instructed during the memory part to execute within 20 s as many repetitions of the sequence as possible without error. This last training step provided us with a daily measure of the degree of learning (Kihlstrom, 1987; Passingham, 1993; Schlaug, 1994).

During training, horizontal and vertical eye movements were recorded, using an infrared reflector system that tracks the boundary between iris and sclera (IRIS, Skalar, Netherlands; optimal precision 1/30°, frequency range 0–100 Hz).

Eye Movements Analysis

Eye movement recordings during training were low-pass filtered and vertical eye movements were used to eliminate artifacts due to eye blinks. Horizontal saccades were defined as eye movements with an initial velocity of >20°/s (Zuber et al., 1965). Saccades were categorized as ‘large’ if they had an amplitude of >3°; the remaining ‘small saccades’ were interpreted as correction saccades. The latency of a saccade was defined as the delay between audio signal and saccade onset. Spatial accuracy was calculated as the difference between target location and the landing point of a saccade (before any correction saccade occurred). Each saccade that did not land within the square delimiting the correct target location in the sequence was considered an error. The error rate was calculated as the number of errors over the total number of saccades.

fMRI Design

Imaging was performed with a 3 T whole-body scanner (Bruker) equipped with a quadrature birdcage RF coil and a head-gradient coil insert designed for echo-planar imaging (EPI). Functional images were collected using a T2*-weighted gradient-echo EPI sequence (24 contiguous 5-mm-thick axial slices, TR = 5s, TE = 40 ms; voxel size: 3 × 4.7 × 5 mm3). High-resolution images (T1-weighted, voxel size: 1.5 × 2 × 1 mm3) were acquired for three-dimensional anatomical identification. Subjects lay supine and had their head fixed using foam rubber. Computer-controlled visual stimuli were displayed via an LED projector on a semi-transparent screen placed behind the scanner. Subjects looked at this screen through mirror glasses. Earphones protected subjects against the scanner noise and facilitated the detection of the audio trigger signal.

An experimental time series consisted of three familiar sequence sets executed alternatively with three different new sequence sets (see Fig. 1). As in the training period, three different New sequences were used within one experimental time series. Consecutive sets were separated by 10 s periods of a control condition. During control, subjects relaxed while keeping their eyes fixed on the central target. The visual display as well as the audio pacing remained the same as during the condition for memory-guided saccades; the only difference between this latter condition and the resting control was thus saccadic performance. A time series began with 20 s of rest (four images), followed by the acquisition of 64 images of the whole brain. Each subject performed three experimental time series.

Unfortunately eye movements could not be recorded online in our EPI coil with the required precision (i.e. minimum 1°). Indeed, the devices available either disturbed the fMRI images or showed spatial resolution worse than 3°. However, the extensive training made us confident that the subjects were performing well in the task. Just before the scan we asked subjects to execute one familiar sequence set and made sure that they had an explicit knowledge of the sequence. In addition, during debriefing after scanning, subjects did not feel they had performed differently in the scanner than during previous training.

fMRI Analysis

Data were analyzed with the non-commercial software SPM99 (Wellcome Department of Cognitive Neurology, London). After reconstruction, images were assessed to estimate translations and rotations. We checked that movement within a session did no exceed 4 mm/2° for every subject. In order to correct for this, and for between-sessions movement, all functional images were registered to the last image of the last session (the closest in time to the anatomical scan), using sinc-interpolation methods. Then, functional and anatomical images of every subject were spatially transformed into Talairach space (Montreal Neurological Institute template) (Evans et al., 1994) to allow a group analysis. Transformation parameters were calculated on anatomical images using sinc-interpolation method. Anatomical images were resampled to 1.5 × 1.5 × 1.5 voxels and functional images to 4 × 4 × 4 voxels.

These normalized functional images were then spatially smoothed using a discrete convolution with a Gaussian kernel of 8 × 8 × 8 mm full width at half maximum (FWHM). The first four images of each time series were discarded to ensure that a steady-state signal was reached. Pixel time series were temporally smoothed with a low-pass Gaussian kernel of 4 s FWHM to obtain better a priori information on the temporal variance/ covariance structure of the data and increase hemodynamic variance components with a neuronal basis, relative to other components (Friston et al. 1995b). Scaling factors were applied to obtain the same average signal in the all-time series, and the group analysis was based on a fixed effects model. Statistical pixel-by-pixel analysis was performed on a temporal basis model that assigned two regressors to each of the five conditions (Fam-V; Fam-M; New-V; New-M; Rest) (Friston et al., 1995a). Low-frequency changes over time were filtered out using a high-pass filter with a cut-off frequency of 1/180 Hz. F- and t-maps were calculated and superimposed on a mean anatomical image from the normalized individual three-dimensional anatomical images. Activations were taken as significant for report if they exceeded a Z-value of >4.5 (P < 0.05 after correction for multiple comparison). To obtain more precise information on the anatomical location of the functional regions of interest, individual analysis was performed for each subject on non-normalized images, using a spatial smoothing of 4 mm.

Results

Psychophysics

Familiar and New Sequence Performance in the Audio-triggered Conditions

On the last day of training, subjects repeated both familiar and new sequences with high performance (error rates ranging from 0 to 1% for familiar and from 0 to 8% for new sequences). At that time, there was no significant difference between the performance of familiar and new sequences with respect to spatial accuracy for both M- and V-conditions (paired t-test, P > 0.1). For the memory-guided condition, neither velocities nor latencies nor the total number of saccades differed between familiar and new sequences (paired t-test, P > 0.1), except for one subject who made more correction saccades for new sequences. The functional data of this subject were therefore excluded from the brain imaging analysis to avoid data misinterpretation by differential saccadic rates for familiar and new sequence sets. In the V-condition we observed significant (paired t-test, P < 0.01) increased latency, slower velocity and higher number of saccades for new compared with familiar sequences

Sequence-specific and Sequence-unspecific Learning Assessed in the No-trigger Condition

Data for the situation without trigger (training condition 3, see Materials and Methods) were analyzed for eight subjects, since these data sets were lost for two subjects for technical reasons. Learning effects were assessed on the basis of two parameters. First, we calculated the number of sequences that could be completed from memory within 20 s, since the speed of execution is supposed to be an indicator of the degree of learning (Passingham, 1993; Karni et al., 1995). Second, error rates were analyzed to estimate the difficulty of the tasks. The mean number of executed sequence repetitions within 20 s on day 1 of training ranged from 4 to 8 for familiar (corresponding to a saccadic frequency of 1–2 Hz) and from 2.2 to 6.2 for new sequences (0.55–1.55 Hz). Already at this stage of training sequence-specific learning effects were observed, leading to better performance for familiar than for new sequences. On day 4, the repetition rate ranged from 4.6 to 11.2 repetitions for familiar (1.15–2.8 Hz) and from 4 to 9 repetitions for new sequences (1 to 2.25 Hz). Figure 2 shows the mean number of completed sequences and error rates across the 4 days of training. Comparing days 1 and 4 revealed global learning effects that were expressed in a significant increase in the number of completed sequences from memory (Fig. 2A) and a decrease of error rates (Fig. 2B) for both familiar and new sequences (Wilcoxon, P ≤ 0.001). We refer to this type of learning effect as ‘global’, since the same significant tendency was observed for both familiar and new sequences. Moreover, sequence-specific learning was reflected in faster and more accurate performance of familiar compared with new sequences (Wilcoxon, P ≤ 0.001 for day 4).

Altogether these observations showed that at the end of the training period (i) subjects acquired all routines relevant to the task; (ii) sequence-specific learning had occurred and could be revealed in the triggerless task; and (iii) execution of familiar and new saccadic sequences in the audio-paced memory condition did not differ with respect to the number of saccades, latency or accuracy. We therefore concluded that during recall from memory every activation difference induced by familiar compared with new sequences could be attributable to the familiarity of the executed sequence, and not to sequence-non-specific skill learning or movement rate.

fMRI

For functional data analysis, we concentrated on the memory-guided condition. Audio and visual stimulation were the same during saccadic and resting control condition. Activation observed in sequence/rest contrasts thus can be attributable to saccadic sequence execution. In addition, any low-level interaction between audio-cueing and saccadic execution are comparable in new and familiar sequence conditions since the saccadic latencies did not differ. We did not analyze the V-condition, since V-New induced higher numbers of saccades and longer latencies than V-Fam. Thus, we would not have been able to distinguish between activation changes due to learning effects and those due to basic components of saccadic execution such as saccadic rate and rhythm. Indeed, rhythm and rate of movements are known to change cortical activation in motor and premotor areas (Rao et al., 1993; VanMeter et al., 1995; Schlaug et al., 1996).

The following report focuses mainly on group data. However, anatomical precision given in text or tables may rely on observations made on individual maps, as will be indicated. References to activation found in both group-averaged and individual data sets are based on statistical maps thresholded at P < 0.05 corrected for multiple comparison (Z > 4.5).

New Sequences–Resting Control

Execution of new sequences compared with resting control activated a cortical network of frontal, prefrontal and parietal areas (Table 1) that corresponds well with earlier described oculomotor networks (Anderson et al., 1994; O'Sullivan et al., 1995) [see Pierrot-Deseilligny et al. for a review (Pierrot-Deseilligny et al., 1995)].

Frontal Eye Fields (FEF).

In the dorsal part of the frontal lobe, we observed two different activation sites in the region of the frontal eye fields (Fig. 3). The first site was located bilaterally at the intersection of the precentral sulcus with the superior frontal sulcus (SFS). We refer to this medio-dorsal frontal activation site as the dorsomedial FEF (Luna et al., 1998; Lobel et al., 2000), since it is located in the medial part of the precentral sulcus. The second site was located in the precentral sulcus, extending into the precentral gyrus, and is referred to as the lateral FEF. The lateral and dorsomedial FEFs likely correspond to those described in other recent functional imaging studies of human oculomotor function (Petit et al., 1993, 1996; Lang et al., 1994; Paus, 1996; Luna et al., 1998; Heide et al., 1999). Individual analysis revealed that a dorsomedial FEF site was significantly activated in seven of nine subjects bilaterally and in the left hemisphere only in the two remaining subjects. The lateral site was significantly activated bilaterally in two subjects, only in the left hemisphere in six subjects (including the two showing unilateral left dorsomedial focus) and not at all in one subject. In addition, the group analysis showed a focus within the SFS anterior to the one identified as the dorsomedial FEF.

Supplementary Eye Fields (SEF).

On the dorsal medial wall, one activation site was localized in the descending branch of the paracentral sulcus, previously identified as an anatomical land-mark for the supplementary eye field (SEF) in humans (Luna et al., 1998; Grosbras et al., 1999). A further frontal activation site was observed more rostrally in the superior frontal gyrus, near the upper part of the cingulate sulcus (Fig. 3). This location has previoulsy been described as pre-SMA, a region related to ‘prefrontal’ functions such as selection of cognitive sets (Konishi et al., 1998) and distinct from the posterior SMA (SMA proper), which is dedicated more to motor/executive functions (Rizzolatti et al., 1996; Vorobiev et al., 1998). By analogy, we named the rostral medial-wall region activated by new sequence of saccades the ‘pre-SEF’ (Petit et al., 1996; Grosbras et al., 1998). The pre-SEF activation was highly significant (Z = 5.28) in the left hemisphere, and was observed by individual analysis in seven subjects. All but one of these subjects also showed two left FEF foci (dorsomedial + lateral). The corresponding focus in the right hemisphere reached significance in three of the seven subjects, but not in the group-averaged data (Z = 3.1). These three subjects, who exhibited significant right pre-SEF activation, also showed a right dorsomedial FEF activation site, and only one showed a right lateral FEF activation site. All subjects showing pre-SEF activation also had SEF activation in the same hemisphere. The mean distance between SEF and pre-SEF foci in the rostrocaudal (x) dimension was 19.5 mm (SD = 10.5).

Occipital and Parietal Cortex.

In the occipital lobe, the posterior part of the calcarine fissure was activated essentially in the left hemisphere.

Several activation sites were observed in the posterior parietal cortex. The most significant activation site was located in the anterior part of the intraparietal sulcus (IPS), deep in the sulcus, at the level of the posterior supramarginal gyrus, between Brodman areas (BA) 7 and 40. It presumably corresponds to the parietal oculomotor location activated in a number of oculo-motor paradigms, assumed to be the equivalent of area LIP in monkeys (Müri et al., 1996; Andersen, 1997; Petit et al., 1997; Luna et al., 1998). This IPS activation extended into a wide part of the superior and inferior parietal lobules (SPL and IPL), as well as into the anterior IPS. Moreover, several activation sites were observed in the parieto-temporo-occipital region, as, for example, at the junction between transverse occipital sulcus and IPS, earlier described as TRIPS (Orban et al., 1999).

Familiar–Rest

Contrasting brain activity during execution of familiar sequences with activity during Rest revealed activation sites in the dorsomedial and lateral FEF, similar to the activation sites for the contrast New–Rest, but with lower Z-scores (Table 1). No voxel in the medial wall (SEF) reached significance, but lowering the threshold showed that this region was nevertheless activated [Z = 3.77 for SEF (–12, –8, 60)], in line with the hypothesis that familiar and new sequence performance activated similar cortical networks. Individual analysis showed that the SEF was nevertheless significantly activated (Z > 4.5) bilaterally in two of nine subjects, and in the left hemisphere only in five subjects. In addition, we observed two activation sites in the left and right middle part of the superior frontal gyrus.

Parietal activation sites for familiar sequence execution were similar to those observed for new sequences, but, as in frontal regions, activity extends less for familiar sequences and reached significance only in the left dorsal middle part of the IPS (see Table 1).

The posterior calcarine sulcus was bilaterally activated. In addition, a focus in the medial temporo-occipital cortex was observed near the temporal horn of the lateral ventricle, between the most posterior part of the parahippocampal gyrus and the medio-anterior part of the lingual gyrus. This activation site was located at a place that Insauti et al. (Insauti et al., 1998) used to define the boundary between the occipital and temporal cortex. Since in most atlases the transition between the occipital and temporal cortex is described as progressive (Von Economo, 1929; Duvernoy, 1992), we cannot classify the observed activation site as occipital or temporal. Rather, we refer to it as medio-temporo-occipital.

Like in the New–Rest contrast, activation in frontal and parietal areas showed a higher Z-score and reached significance in more individuals in the left hemisphere than in the right hemisphere.

New–Familiar Sequences

Direct comparison between images acquired during reproduction of new sequences with those acquired during familiar sequences (Table 2) revealed significant (Z > 4.5) bilateral activation in a fronto-parietal network, already identified in both contrasts New–Rest and Familiar–Rest. Individual variability was observed in both intensity and extent of this activation. Activation of the dorsomedial FEF extended into the SFS, anterior to its junction with the precentral sulcus (see Fig. 4, center images). In contrast, activation in the lateral FEF site did not show any significant difference between the execution from memory of New and Familiar sequences (Z = 2.75, P > 0.05, uncorrected).

On the medial wall, no significant difference between new and familiar sequences was observed in the part identified as SEF (Z = 2.26 and Z = 2.16 for left and right, respectively; P > 0.05, uncorrected). The pre-SEF, however, was bilaterally significantly more activated during new compared with familiar sequence execution. This pre-SEF region was located very closely but dorsally to the prefrontal part of the cingulate sulcus, with little variability among individuals.

Several activation sites in the IPS were significantly more activated during execution of new compared with familiar sequences, extending again into both SPL (BA 7) and IPL (BA 40). In addition, activation was observed in sites that were not activated in the contrast New–Rest, including bilateral precuneus and left striatum (tail of caudate nucleus extending into the putamen).

Familiar–New

The contrast Familiar–New revealed significant differences in two small regions (see Table 2). The first activation site was the one observed in the medial temporo-occipital cortex of the right hemisphere. The second was seen in the upper (parietal) bench of the right parieto-occipital sulcus, extending medially toward the most posterior part of the cingulate gyrus. Individual analysis revealed that both sites were always activated in common, but reached significance in four of the nine subjects.

Discussion

In this study, we showed that extensive practice of an oculomotor sequence modifies the corresponding pattern of cerebral activation during recall. Execution of familiar and new sequences of saccades largely activated a common network of brain areas, namely FEF, SEF and several parietal areas. Within this network, activity was higher for new sequences, especially in the left hemisphere. Execution of new sequences, recalled from short-term memory, additionally activated regions such as the pre-SEF, precuneus and striatum that were not observed for familiar sequence execution. In contrast, familiar sequences, requiring recall from long-term memory, specifically involved a part of the right parieto-occipital sulcus, as well as a right medial temporo-occipital region. Familiarity thus clearly influences the cortical oculomotor network involved in sequential oculomotor behavior.

Behavioral Parameters

The global learning, characterized by improvement in performance of both familiar and new sequences, reflected the learning of task requirements (Navon, 1978; Passingham, 1993; Hikosaka et al., 1995; Doyon et al., 1996). It was achieved at the end of the training, allowing us to focus our observations on the sequence-specific recall of familiar compared with new sequences. The ability to perform the familiar sequence faster and with lower error rate than new sequences indicates that sequence-specific learning had occurred, leading to specific consolidation of the internal representation of the familiar sequence. Sequence-specific learning has been shown to involve parallel procedural processes (or implicit learning processes) — revealed by improvement in performance through practice without necessary awareness of skill acquisition — and explicit learning processes — involving conscious representation of the sequence (Perruchet and Amorim, 1992; Curran and Keele, 2000). Given the design of the task, subjects always had to explicitly recall a sequence, irrespective of whether the sequence was new or familiar. The comparison of the two types of sequences thus primarily tested for the effect of implicit procedural learning stage.

Audio-pacing during scanning corresponded to the slowest saccadic rate achieved during prior training without a trigger, and ensured that all subjects performed the task with a high level of performance and with the same rhythm for both types of sequence. All physical parameters were similar for familiar and new sequences tested during the two last days of practice. Thus, based on these observations and given the extensive training, we reasoned that even though we could not directly record subjects' behavior in the scanner, they performed as well during the actual scanning as during the prior training. All activation differences between familiar and new sequence execution from memory could therefore be interpreted in terms of sequence-specific learning and memory.

The Influence of Attention to Action in the Oculomotor Network: Fronto-parietal Regions

The first observation was greater global activity during execution of new compared with familiar sequences. It can be excluded that this higher activity was due to higher numbers of saccades, since the motor rate was the same for both sequences, as indicated by behavioral data. Previous studies of manual sequence learning have shown that greater activation in the network involved in sequence execution might be due, at least in part, to the greater general attentional load required by new sequences (Juepner et al., 1997a). Indeed, effort and concentration on performing the sequence from memory decreased as oculomotor sequences became more automatic, correlating with decreased cortical activity in dorsomedial FEF, SEF and IPS.

Activity in the lateral FEF did not decrease with sequence familiarity, pointing towards functional differences between lateral and dorsomedial FEF. The former might be recruited for more executive functions that are independent of the familiarity of the executed sequence, and that are not modulated by attention.

The observation of brain regions activated specifically for newly learned sequences suggests that those sequences require specialized brain resources independent of attention modulation of oculomotor performance. Analogously Juepner and colleagues (Juepner et al., 1997a) showed that paying attention to their action when executing an overlearned sequence of finger movements is not sufficient to explain the pattern of activity induced by newly learned execution. Other cognitive processes involved in newly learned sequences might explain specific patterns of activation.

Activity Specific of New Sequences: Pre-SEF, Striatum, Precuneus and FEF

The generation of precisely timed saccades, synchronized with audio-pacing, requires greater use of an internal memory store (Barnes and Donelan, 1999) than the generation of precise saccades in familiar sequences. Furthermore, the novelty of the sequence engages spatial working memory resources to keep online the spatial order of the sequence that are not necessary for familiar sequences. This can be reflected by the extension of frontal activation into the SFS (Haxby et al., 2000), as well as by extended posterior parietal cortex involvement (Andersen et al., 1993, 1997). Further research is needed to distinguish whether the activation observed in our paradigm is due to working memory, spatial attention or both. Despite the well-known right hemisphere dominance for visuo-spatial abilities, we observed a tendency for a left hemisphere dominance. This might be due to the sequential organization of visuo-spatial shifts in our paradigm. Such an interpretation is in line with lesion and transcranial magnetic stimulation studies showing the necessity of an intact left hemisphere for sequence execution from memory (Jason, 1983; Gaymard et al., 1993; Schluter et al., 1998).

Involvement of a part of the pre-SMA, which we named pre-SEF, during the execution of newly learned sequences is consistent with previous observations of pre-SMA activation associated with free selection of movements (Deiber et al., 1991) or with the early learning of sequential movements (Hikosaka et al., 1996; Nakamura et al., 1998; Sakai et al., 1998). Indeed, each step within new sequence execution necessitates selection and planning of the next appropriate saccade on the basis of an internal model for the whole sequence. Two previous PET studies reported activation of a part of the pre-SMA, corresponding to the pre-SEF, during the reproduction of an oculomotor sequence from short-term memory (Petit et al., 1996; Kawashima et al., 1998). These studies, however, did not allow a definition of the function of this area without ambiguity (see Introduction). Our results now emphasize the preponderant and specific role of the pre-SEF in the oculomotor preparation of new saccadic sequences.

The caudate nucleus was also specifically involved in execution of new sequences. In macaque monkeys, the pre-SMA (area F6) and caudate nucleus are known to be connected via the ventral thalamus (Rizzolatti et al., 1996), and both have been implicated in motor sequence learning (Hikosaka et al., 1995; Miyachi et al., 1997). Also, in humans, striatal involvement in new sequence organization and the early phase of procedural learning has been described by functional imaging studies (Juepner et al., 1997b; Rauch, 1997; Toni et al., 1998) and observations in patients with caudate abnormalities (Watkins et al., 1999) or Parkinson's disease (Jackson et al., 1995). We suggest that this basal ganglia–prefrontal loop is involved in the chronological organization of new eye movement combinations.

Similar to the pre-SEF and caudate nucleus, the medial parietal cortex was activated only in the New–Familiar contrast, in line with earlier findings (Petit et al., 1996). Activation in the precuneus has been observed during rule-changing (Nagahama et al., 1999). We suggest that it is important for organizing sequential behavior that follows a certain strategy, and/or to build up an internal representation of extrapersonal space (Ghaem et al., 1997; Maguire et al., 1998). In such a scenario, the precuneus might feed the prefrontal–caudate loop with motor and perceptual information of the sequence to select the next appropriate saccade.

Saccade selection furthermore seems to recruit the dorso-medial FEF, as suggested by specific FEF neuronal activation in monkeys during natural scanning (Burman and Segraves, 1994) or a laboratory task [see Tehovnik et al. for a review (Tehovnik et al., 2000)]. These studies suggested that the FEF is involved in the selection of saccadic parameters in a behavioral and visual context. In contrast, the pre-SEF, together with the caudate and precuneus, would be more crucial for the elaboration of motor plans in the context of a spatio-temporal set of movements. Similar sites of activation have been observed in early (but not late) phases of manual sequence learning, especially when the learning was explicit (Jenkins et al., 1994; Juepner et al. 1997; Sakai et al., 1998; Toni et al., 1998). Studies of implicit sequence learning did not show activation in those regions but, in contrast, did in the premotor cortex and cerebellum. Thus we suggest that functions subserved by the pre-SEF, precuneus and caudate when performing newly learned saccadic sequences (namely, programming, movement selection, working memory, build up of internal representation) might be related to the use of a declarative form of memory.

Areas Specialized for Familiar Sequences of Saccades

When familiar sequences were executed, no specific activation within the pre-SEF, precuneus, caudate and SFS could be observed, indicating that attention, spatial working memory and selection of action might have been reduced or inhibited. It might also indicate that explicit or declarative knowledge was switched off in favor of a more procedural executive mode when the sequence becomes well learned. The effect of a higher degree of procedural learning might thus be an attenuation of the use of explicit memory.

Some previous studies have shown that extensive practice of a particular sequence produces an increase in primary sensori-motor activation (Karni et al., 1995). However, we did not observe such an increase. This might be for several reasons. First, the system subserving procedural learning of skeleto-motor sequences might work differently from that for learning oculomotor sequences. Secondly, one week of practice might not be sufficient to produce sensible neural representation. Thirdly, this activity modulation could have been masked by the attentional modulation, since circuits for attention and for eye movement control do largely overlap (Nobre et al., 1997; Corbetta et al., 1998).

Interestingly, we observed regions specifically involved in the execution of familiar sequences, indicating that recall of saccadic sequences from long-term memory involves specific cortical networks different from those involved in the execution of unfamiliar sequences. One of these regions, located in the parieto-occipital sulcus, might correspond to an area described in monkeys that is implicated in the computation of extraretinal target coordinates and their associated eye movements (Galletti et al., 1995; Wise et al., 1997). The second activation site was observed at the boundary between the parahippocampal and lingual gyri, presumably related to the most anterior extrastriate activation found for shape processing (Corbetta et al., 1990), or to the medio-temporal involvement in spatial memory tasks requiring allocentric representation (Ghaem et al., 1997). This region has not been described before as being involved in oculomotor performance. We suggest that it could be implicated in long-term spatial memory.

In conclusion, this study provides the first evidence that the brain activation associated with the execution of oculomotor scanpaths depends on the familiarity of the executed sequence. Familiarity is reflected in the reduced activity in an oculomotor network common to all saccadic sequence performance, and in specific activation sites for the execution of either new or familiar sequences. The change from one oculomotor network to another depends on the degree of practice of the sequence, or, in other words, on the stage of procedural learning. We hypothesize that this change corresponds to a transition from an effortful to an automatic processing mode, liberating attentional and working memory resources for other behavioral needs (Schneider and Shiffrin, 1977). The two partially overlapping oculomotor networks described in this article are likely to provide a neurophysiological basis of memory for scanpaths in visual scene exploration and orienting (Noton, 1971).

Notes

The authors acknowledge Isabelle Israël for her help with the eye movement recording system and Kate Watkins for her help on the manuscript. M.-H. Grosbras was supported by a grant from the French Ministry of Education and Research.

Table 1

Network for sequences of saccades

Cerebral area Talairach coordinates Z n 
 x y z   
Group results. Actiation sites for Familiar/Control or for New/Control. SPM99 group analysis, P < 0.05, corrected for multiple comparisons. FEF, frontal eye field; Lat FEF, lateral FEF; SEF, supplementary eye field; IPS, intraparietal sulcus; SPL, superior parietal sulcus; IPL, intraparietal sulcus; POCS, postcentral sulcus; SMG, supramarginal sulcus; TROS, Transverse occipital sulcus; MOg, middle occipital gyrus; parahipp, parahippocampal gyrus; (6), Brodman area 6. Z, Z-score for the tested statistical contrasts. n, number of individual analysis in which the region showed a Z-score of >4.5. 
New/REST      
Dorsomedial FEF L (6) and SFS –28 –12 56 12 
 –32 –8 64 10.63  
 –32 52 10.13  
Dorsomedial FEF R (6) and SFS 24 –8 48  9.60 
Lat FEF L (6) –48 –12 52 
 –52 –8 48  
Lat FEF R (6) 36 –4 44  6.45 
SEF L (6) –8 –4 64  5.67 
SEF R(6) –16 64  4.71 
Pre-SEF L(6/8) –12 60  5.28 
Frontopolar (10) 48 –12  4.69 
Dorsal SPL (7) L –32 –52 72  5.40 
Post SPL/IPS (7) L –28 –76 32  4.99 
Ant IPS (7/40) L –28 –60 56 13.53 
Ant IPS 7/40) R 24 –56 56 11.21 
Ant IPS/PoCs (40) L –48 –32 44  6.95 
Ant IPS/PoCs (40) R 32 –28 40  5.67 
IPL (40) L –28 –56 32  5.97 
SMG (40) –44 –56 32  5.12 
TROs/IPS (40/39) –36 –76 20  5.75 
Calcarin s. L(17) –12 –72  4.73 
      
Familiar/REST      
Dorsomedial FEF L (6) –28 –16 52  8.82 
Dorsomedial FEF R (6) 24 –8 48  5.97 
Lat FEF L (6) –48 –12 52  4.79 
Frontopolar 52 –8  4.71 
SFG L(9) –16 40 52  4.63 
SFG R(9/10) 56 32  4.61 
Ant IPS/SPL (7) L –28 –60 48  6.42 
Ant IPS R (7/40) 24 –48 24  5.58 
SMG (40) L –44 –56 28  5.76 
Calcarin s. (17) L –12 –72  6.10 
Calcarin s. (17) R 12 –72  5.09 
Lingual/parahipp R 28 –60  8.56  
Cerebral area Talairach coordinates Z n 
 x y z   
Group results. Actiation sites for Familiar/Control or for New/Control. SPM99 group analysis, P < 0.05, corrected for multiple comparisons. FEF, frontal eye field; Lat FEF, lateral FEF; SEF, supplementary eye field; IPS, intraparietal sulcus; SPL, superior parietal sulcus; IPL, intraparietal sulcus; POCS, postcentral sulcus; SMG, supramarginal sulcus; TROS, Transverse occipital sulcus; MOg, middle occipital gyrus; parahipp, parahippocampal gyrus; (6), Brodman area 6. Z, Z-score for the tested statistical contrasts. n, number of individual analysis in which the region showed a Z-score of >4.5. 
New/REST      
Dorsomedial FEF L (6) and SFS –28 –12 56 12 
 –32 –8 64 10.63  
 –32 52 10.13  
Dorsomedial FEF R (6) and SFS 24 –8 48  9.60 
Lat FEF L (6) –48 –12 52 
 –52 –8 48  
Lat FEF R (6) 36 –4 44  6.45 
SEF L (6) –8 –4 64  5.67 
SEF R(6) –16 64  4.71 
Pre-SEF L(6/8) –12 60  5.28 
Frontopolar (10) 48 –12  4.69 
Dorsal SPL (7) L –32 –52 72  5.40 
Post SPL/IPS (7) L –28 –76 32  4.99 
Ant IPS (7/40) L –28 –60 56 13.53 
Ant IPS 7/40) R 24 –56 56 11.21 
Ant IPS/PoCs (40) L –48 –32 44  6.95 
Ant IPS/PoCs (40) R 32 –28 40  5.67 
IPL (40) L –28 –56 32  5.97 
SMG (40) –44 –56 32  5.12 
TROs/IPS (40/39) –36 –76 20  5.75 
Calcarin s. L(17) –12 –72  4.73 
      
Familiar/REST      
Dorsomedial FEF L (6) –28 –16 52  8.82 
Dorsomedial FEF R (6) 24 –8 48  5.97 
Lat FEF L (6) –48 –12 52  4.79 
Frontopolar 52 –8  4.71 
SFG L(9) –16 40 52  4.63 
SFG R(9/10) 56 32  4.61 
Ant IPS/SPL (7) L –28 –60 48  6.42 
Ant IPS R (7/40) 24 –48 24  5.58 
SMG (40) L –44 –56 28  5.76 
Calcarin s. (17) L –12 –72  6.10 
Calcarin s. (17) R 12 –72  5.09 
Lingual/parahipp R 28 –60  8.56  
Table 2

Comparison between Familiar and New sequences

Cerebral area Talairach coordinates Z n 
 x y z   
FEF, frontal eye field; Lat FEF, lateral FEF; SEF, supplementary eye field; IPS, intraparietal sulcus; SPL, superior parietal lobule; IPL, inferior parietal lobule; dIPS, dorsal IPS; V IPS, ventral IPS; POCS, postcentral sulcus; SMG, supramarginal sulcus; TROS, transverse occipital sulcus; Gom, median occipital gyrus; L and R, left and right hemisphere respectively; par-occ s., parieto-occipital sulcus; (6), Brodman area 6. 
New/Familiar 
Dorsomedial FEF /SFS (6) L –32 52  
Dorsomedial FEF /SFS (6) R 20 56 6.19 
Pre-SEF (6/8) L –12 56 4.57 
Pre-SEF (6/8) R 12 56 4.94 
Precuneus L –12 –64 64 5.29 
Precuneus L 12 –64 64 5.73 
dIPS and SPL (7) L –28 –60 56 7.50 
dIPS and SPL (7) R 20 –56 56 7.46 
dIPS and SPL (7) R 24 –44 40 6.89 
V IPS/SPL (7) L –28 –64 28 6.86 
Post SPL (7) R 28 –68 40 5.27 
Ant IPS (7/40) R 52 –16 36 4.96 
Ant IPS (7/40) R 44 –28 40 5.48 
IPS/angular 40) L –36 –48 48 7.03 
IPS/angular 40) R 28 –48 48 7.37 
IPS/SMG (40) L –40 –52 56 6.09 
IPS/SMG (40) L –48 –36 44 5.94  
IPS/SMG (40) R 32 –52 28 5.15  
IPS/SMG (40) R 28 –72 32 4.66  
Post SPL/par-occ s. (7/19) R 24 –72 36 5.39 
IPS/TROs. –28 –76 20 5.17 
Gom (19) –44 –72 16 5.45 
Putamen/caudate –24 –12 24 4.63 
 –28 24 4.49  
      
Familiar/Newly learned 
Parieto occ. R 24 –48 24 6.11 
Lingual/posterior parahipp. R 28 –60 4.85 
Cerebral area Talairach coordinates Z n 
 x y z   
FEF, frontal eye field; Lat FEF, lateral FEF; SEF, supplementary eye field; IPS, intraparietal sulcus; SPL, superior parietal lobule; IPL, inferior parietal lobule; dIPS, dorsal IPS; V IPS, ventral IPS; POCS, postcentral sulcus; SMG, supramarginal sulcus; TROS, transverse occipital sulcus; Gom, median occipital gyrus; L and R, left and right hemisphere respectively; par-occ s., parieto-occipital sulcus; (6), Brodman area 6. 
New/Familiar 
Dorsomedial FEF /SFS (6) L –32 52  
Dorsomedial FEF /SFS (6) R 20 56 6.19 
Pre-SEF (6/8) L –12 56 4.57 
Pre-SEF (6/8) R 12 56 4.94 
Precuneus L –12 –64 64 5.29 
Precuneus L 12 –64 64 5.73 
dIPS and SPL (7) L –28 –60 56 7.50 
dIPS and SPL (7) R 20 –56 56 7.46 
dIPS and SPL (7) R 24 –44 40 6.89 
V IPS/SPL (7) L –28 –64 28 6.86 
Post SPL (7) R 28 –68 40 5.27 
Ant IPS (7/40) R 52 –16 36 4.96 
Ant IPS (7/40) R 44 –28 40 5.48 
IPS/angular 40) L –36 –48 48 7.03 
IPS/angular 40) R 28 –48 48 7.37 
IPS/SMG (40) L –40 –52 56 6.09 
IPS/SMG (40) L –48 –36 44 5.94  
IPS/SMG (40) R 32 –52 28 5.15  
IPS/SMG (40) R 28 –72 32 4.66  
Post SPL/par-occ s. (7/19) R 24 –72 36 5.39 
IPS/TROs. –28 –76 20 5.17 
Gom (19) –44 –72 16 5.45 
Putamen/caudate –24 –12 24 4.63 
 –28 24 4.49  
      
Familiar/Newly learned 
Parieto occ. R 24 –48 24 6.11 
Lingual/posterior parahipp. R 28 –60 4.85 
Figure 1.

Experimental design. (A) Example for a visually guided (V) condition. Each of the five saccades is triggered by a visual target and an audio signal. After the fifth target presentation, the basic display remains unchanged for 300 ms, before the sequence is restarted. In the memory-guided (M) condition, only the basic display is presented, but saccades are again paced by the audio-signal. (B) Familiar sequence set. (C) General paradigm: in each fMRI session, the familiar sequence set (Fam) was performed three times in alternation with three different new sequences.

Figure 1.

Experimental design. (A) Example for a visually guided (V) condition. Each of the five saccades is triggered by a visual target and an audio signal. After the fifth target presentation, the basic display remains unchanged for 300 ms, before the sequence is restarted. In the memory-guided (M) condition, only the basic display is presented, but saccades are again paced by the audio-signal. (B) Familiar sequence set. (C) General paradigm: in each fMRI session, the familiar sequence set (Fam) was performed three times in alternation with three different new sequences.

Figure 2.

Behavioral results for memory-guided sequence performance. (A) Mean number of sequences (n = 8) completed within 20 s (without audio-trigger) as a function of day of practice. (B) Mean error rate (n = 8) as a function of day of practice. Error rates were calculated as the proportion of errors relative to the total number of saccades. Error bars indicate standard errors.

Figure 2.

Behavioral results for memory-guided sequence performance. (A) Mean number of sequences (n = 8) completed within 20 s (without audio-trigger) as a function of day of practice. (B) Mean error rate (n = 8) as a function of day of practice. Error rates were calculated as the proportion of errors relative to the total number of saccades. Error bars indicate standard errors.

Figure 3.

Cortical activation during performance of new and familiar sequences compared with rest. First row: group results (n = 9) superimposed on individual normalized anatomical images. Sulci overlaid: purple = paracentral; pink = central; green = precentral; blue = intraparietal; turquoise = superior frontal. Note on the left view the dissociation between the SEF and pre-SEF. Second row: three-dimensional rendering (some activated voxels deep in sulci are projected on the surface). Voxels with P < 0.05 corrected for multiple comparisons are represented with color scale relative to Z-scores. 1 = pre-SEF; 2 = dorsomedial FEF; 3 = SFS; 4 = SPL; 5 = IPS/superior marginal gyrus (SMG); 6 = precuneus; 7 = angular gyrus.; 8 = SEF; 9 = lateral FEF. Third row: three-dimensional rendering for the contrast Familiar/ rest.

Figure 3.

Cortical activation during performance of new and familiar sequences compared with rest. First row: group results (n = 9) superimposed on individual normalized anatomical images. Sulci overlaid: purple = paracentral; pink = central; green = precentral; blue = intraparietal; turquoise = superior frontal. Note on the left view the dissociation between the SEF and pre-SEF. Second row: three-dimensional rendering (some activated voxels deep in sulci are projected on the surface). Voxels with P < 0.05 corrected for multiple comparisons are represented with color scale relative to Z-scores. 1 = pre-SEF; 2 = dorsomedial FEF; 3 = SFS; 4 = SPL; 5 = IPS/superior marginal gyrus (SMG); 6 = precuneus; 7 = angular gyrus.; 8 = SEF; 9 = lateral FEF. Third row: three-dimensional rendering for the contrast Familiar/ rest.

Figure 4.

Cortical activation during performance of new compared with familiar sequences. Group-averaged data (n = 9; P < 0.05, corrected for multiple comparison) are superimposed on a three-dimensional rendering and on normalized average anatomical brain slices. The same labeling is used as for Figure 3.

Cortical activation during performance of new compared with familiar sequences. Group-averaged data (n = 9; P < 0.05, corrected for multiple comparison) are superimposed on a three-dimensional rendering and on normalized average anatomical brain slices. The same labeling is used as for Figure 3.

References

Andersen RA (
1997
) Multimodal integration for the representation of space in the parietal cortex.
Phil Trans R Soc Lond
 
352
:
1421
–1428.
Andersen RA, Snyder LH, Li C-H, Stricanne B (
1993
) Coordinate transformations of spatial information.
Curr Opin Neurobiol
 
3
:
171
–176.
Andersen RA, Snyder LH, Bradley DC, Xing J (
1997
) Multimodal representation of space in the posterior parietal cortex and its use in planning movements.
Annu Rev Neurosci
 
20
:
303
–330.
Anderson TJ, Jenkins IH, Brooks DJ, Hawken MB, Frackowiak RS, Kennard C (
1994
) Cortical control of saccades and fixation in man. A PET study.
Brain
 
117
:
1073
–1084.
Andrews TJ, Coppola DM (
1999
) Idiosyncratic characteristics of saccadic eye movements when viewing different visual environment.
Vision Res
 
39
:
2947
–2953.
Bahill AT, Adler D, Stark L (
1975
) Most naturally occurring human saccades have magnitude of 15 degrees or less.
Invest Ophthalmol
 
14
:
468
–469.
Barnes GR, Donelan SF (
1999
) The remembered pursuit task: evidence for segregation of timing and velocity storage in predictive oculomotor control.
Exp Brain Res
 
129
:
57
–67.
Brandt SA, Stark LW (
1997
) Spontaneous eye movements during visual imagery reflect the content of the visual scene.
J Cogn Neurosci
 
9
:
27
–38.
Burman DD, Segraves MA (
1994
) Primate frontal eye field activity during natural scanning eye movements.
J Neurophysiol
 
71
:
1266
–1271.
Corbetta M, Akbudak E, Conturo TE, Snyder AZ, Ollinger JM, Drury HA, Linenweber MR, Petersen SE, Raichle ME, Van Essen DC, Shulman GL (
1998
) A common network of functional areas for attention and eye movements.
Neuron
 
21
:
761
–773.
Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE (
1990
) Attentional modulation of neural processing of shape, color, and velocity in humans.
Science
 
248
:
1556
–1558.
Curran T, Keele SW (
2000
) Attentional and non-attentional forms of sequence learning.
J Exp Psychol Learn Mem Cogn
 
19
:
189
–202.
Deiber MP, Passingham RE, Colebatch F, Nixon P, Frackowiak RSJ (
1991
) Cortical areas and the selection of movement: a study with positron emission tomography.
Exp Brain Res
 
393
–402.
Doyon J, Owen AM, Petrides M, Sziklas V, Evans AC (
1996
) Functional anatomy of visuomotor skill learning in human subjects examined with PET.
Eur J Neurosci
 
8
:
637
–648.
Duvernoy HM (
1992
) Le cerveau humain. Surface, coupes sériées tri-dimensionnelles et IRM. Paris: Springer-Verlag.
Evans AC, Kamber M, Collins DL, Macdonald D (
1994
) An MRI-based probabilistic atlas of neuroanatomy. In: Magnetic resonance scanning and epilepsy (Shorvon D, Fish D, Andermann F, Bydder GM, Stefan H, eds), pp. 263–274. New York: Plenum.
Friston KJ, Friston KJ, Holmes AP, Worsley KJ, Poline JP, Frith CD, Frackowiak RSJ (
1995
) Statistical parametric maps in functional imaging: a general linear approach.
Hum Brain Mapp
 
2
:
189
–210.
Friston KJ, Holmes A., Poline J-B, Grasby PJ, Williams SC, Frackowiak RSJ, Turner R (
1995
) Analysis of fMRI time-series revisited.
NeuroImage
 
2
:
45
–53.
Galletti C, Battaglini PP, Fattori P (
1995
) Eye position influence on the parieto-occipital area PO (V6) of the macaque monkey.
Eur J Neurosci
 
7
:
2486
–2501.
Gaymard B, Rivaud S, Pierrot-Deseilligny C (
1993
) Role of the left and right supplementary motor areas in memory guided saccades sequences.
Ann Neurol
 
34
:
404
–406.
Ghaem O, Mellet E, Crivello F, Tzourio N, Mazoyer B, Berthoz A, Denis M (
1997
) Mental navigation along memorized routes activates the hippocampus, precuneus, and insula.
NeuroReport
 
8
:
739
–744.
Grafton ST, Mazziota JC, Presty S, Friston KJ, Frackowiak RSJ, Phelps ME (
1992
) Functional anatomy of human procedural learning determined with cerebral blood flow and PET.
J Neurosci
 
12
:
2542
–2548.
Grosbras M-H, Lobel E, Le Bihan D, Berthoz A, Leonards U (
1998
) Evidence for a pre-SEF in humans: a fMRI study.
NeuroImage
 
7
:
S988
.
Grosbras M-H, Lobel E, Van de Moortel P-F, Le Bihan D, Berthoz A (
1999
) An anatomical landmark for the supplementary eye field revealed with fMRI.
Cereb Cortex
 
9
:
705
–711.
Haxby JV, Petit L, Ungerleider LG, Courtney SM (
2000
) Distinguishing the functional roles of multiple regions in distributed neural systems for visual working memory.
NeuroImage
 
11
:
145
–156.
Heide W, Binkofski F, Posse S, Seitz R, Kömpf D, Freund HJ (
1999
) Cortical control of sequences of memory-guided saccades. In: Current oculomotor research (Becker W, ed.), pp. 223–232. New York: Plenum Press.
Hikosaka O, Rand MK, Miyachi S, Kae M (
1995
) Learning of sequential movements in the monkey: process of learning and retention of memory.
J Neurophysiol
 
74
:
1652
–1661.
Hikosaka O, Sakai K, Mikauchi S, Takino R, Sasaki Y, Putz B (
1996
) Activation of human pre-supplementary motor area in learning of sequential procedures: a functional MRI study.
J Neurophysiol
 
76
:
617
–621.
Insauti R, Insauti AM, Sobreviela MT, Salinas A, Martinez-Penuela JM (
1998
) Human medial temporal lobe in aging: anatomical basis of memory preservation.
Microsc Res Techn
 
43
:
8
–15.
Jackson GM, Jackson SR, Harrison J, Henderson L, Kennard C (
1995
) Serial reaction time learning and Parkinson's disease: evidence for a procedural learning deficit.
Neuropsychologia
 
33
:
577
–593.
Jason GW (
1983
) Hemispheric asymmetries in motor function: left hemisphere specialization for memory but not performance.
Neuropsychologia
 
21
:
35
–45.
Jenkins IH, Brooks DJ, Nixon P, Frackowiak RSJ, Passingham RE (
1994
) Motor sequence learning: a study with positron emission tomography.
J Neurosci
 
14
:
3775
–3790.
Juepner M, Stephan KM, Frith CD, Brooks DJ, Frackowiak RSJ, Passingham RE (
1997
) Anatomy of motor learning. I. Frontal cortex and attention to action.
J Neurophysiol
 
77
:
1313
–1324.
Juepner M, Stephan KM, Frith CD, Brooks DJ, Frackowiak RSJ, Passingham RE (
1997
) Anatomy of motor learning. II. Subcortical structures and learning by trial and error.
JNeurophysiol
 
77
:
1325
–1337.
Karni A, Meyer G, Jezzard P, Adams MM, Turner R, Ungerleider LG (
1995
) Functional MRI evidence for adult cortex plasticity during motor skill learning.
Nature
 
377
:
155
–158.
Kawashima R, Tanji J, Okada K, Sugiara M, Sato K, Kinomura S, Inoue K, Ogawa A, Fukuda H (
1998
) Oculomotor sequence learning: a positron emission tomography study.
Exp Brain Res
 
122
:
1
–8.
Kihlstrom JF (
1987
) The cognition unconscious.
Science
 
237
:
1445
–1457.
Konishi S, Nakajima K, Uchida I, Kameyama M, Nakahara K, Sekihara K, Miyashita Y (
1998
) Transient activation of inferior prefrontal cortex during cognitive set shifting.
Nature Neurosci
 
1
:
80
–84.
Lang W, Petit L, Höllinger P, Pietrzyck U, Tzourio N, Mazoyer B, Berthoz A (
1994
) A Positron Emission Tomography study of oculomotor imagery.
NeuroReport
 
5
:
921
–924.
Liversdedge SP, Findlay JM (
2000
) Saccadic eye movements and cognition.
Trends Cogn Sci
 
4
:
6
–14.
Lobel E, Kahane P, Leonards U, Grosbras MH, Lehericy S, LeBihan D, Berthoz A (
2000
) Localization of the human frontal eye fields: anatomical and functional findings from fMRI and intracerebral electrical stimulation. J Neurosurg (in press).
Luna B, Thurlborn KR, Strojwas MH, McCurtain BJ, Berman RA, Genovese CR, Sweeney JA (
1998
) Dorsal cortical regions subserving visually guided saccades in human: an fMRI study.
Cereb Cortex
 
8
:
40
–47.
Maguire EA, Frith CD, Burgess N, Donnett JG, O'Keefe J (
1998
) Knowing where things are: parahippocampal involvement in encoding object location in virtual large-scale space.
J Cogn Neurosci
 
10
:
61
–76.
Miyachi S, Hikosaka O, Miyashita K, Karadi Z, Rand MK (
1997
) Differential roles of monkey striatum in learning of sequential hand movement.
Exp Brain Res
 
115
:
1
–5.
Müri R, Iba-Zizen MT, Cabanis EA, Pierrot-Deseilligny C (
1996
) Location of the human posterior eye field with functional magnetic resonance imaging.
J Neurol Neurosurg Psychiatry
 
60
:
445
–448.
Nagahama Y, Okada T, Katsumi Y, Hayashi T, Yamauchi H, Sawamoto N, Toma K, Nakamura K, Hanakawa T, Konishi J, Fukuyama H, Shibasaki H (
1999
) Transient neural activity in the medial superior frontal gyrus and precuneus time locked with attention shift between object features.
NeuroImage
 
10
:
193
–199.
Nakamura K, Sakai K, Hikosaka O (
1998
) Neuronal activity in medial frontal cortex during learning of sequential procedures.
J Neurophysiol
 
80
:
2671
–2687.
Navon D (
1978
) Perception of misoriented words and letter strings.
Can J Psychol
 
32
:
129
–140.
Nobre AC, Sebestyen GN, Gitelman DR, Mesulam MM, Frackowiak RSJ, Frith CD (
1997
) Functional localization for visuospatial attention using positron emission tomography.
Brain
 
120
:
515
–533.
Noton D (
1971
) Scanpaths in eye movements during pattern perception.
Science
 
309
–311.
Noton D, Stark L (
1971
) Scanpaths in saccadic eye movements while viewing and recognizing patterns.
Vision Res
 
11
:
929
–942.
Orban GA, Sunaert S, Todd JT, VanHecke P, Marchal G (
1999
) Human cortical regions involved in extracting depth from motion.
Neuron
 
24
:
1
–20.
O'Sullivan EP, Jenkins IH, Henderson L, Kennard C, Brooks DJ (
1995
) The functional anatomy of remembered saccades: a PET study.
NeuroReport
 
6
:
2141
–2144.
Passingham RE (
1993
) The frontal lobes and voluntary action. Oxford: Oxford University Press.
Paus T (
1996
) Location and function of the human frontal eye field: a selective review.
Neuropsychologia
 
34
:
475
–483.
Perruchet P, Amorim MA (
1992
) Conscious knowledge and changes in performance in sequence learning: evidence against dissociation.
J Exp Psychol Learn Mem Cogn
 
18
:
785
–800.
Petit L, Orssaud C, Tzourio N, Salamon G, Mazoyer B, Berthoz A (
1993
) PET study of saccadic eye movements in humans: basal ganglia–thalamocortical system and cingulate cortex involvement.
J Neurophysiol
 
69
:
1009
–1017.
Petit L, Orssaud C, Tzourio N, Crivello F, Berthoz A, Mazoyer B (
1996
) Functional anatomy of a prelearned sequence of horizontal saccades in humans.
J Neurosci
 
16
:
3714
–3736.
Petit L, Orssaud C, Tzourio N, Mazoyer B (
1997
) Superior parietal lobule involvement in the representation of visual space: a PET review. In: Parietal lobe contributions to orientation in 3D space (Thiers P, Karnath H-O, eds), pp. 77–90. Heidelberg: Springer-Verlag.
Pierrot-Deseilligny E, Rivaud S, Gaymard B, Müri R, Vermesch AI (
1995
) Cortical control of saccades.
Ann Neurol
 
37
:
557
–567.
Rao SM, Binder JR, Bandettini PA, Haumecke TA, Yetkin FZ, Jesmanowicz A, Lisk LM, Morris GL, Estkowsky RTR, Wong EL, Haughton VM, Hyde JS (
1993
) Functional magnetic resonance imaging of complex human movements.
Neurology
 
43
:
2311
–2318.
Rauch SL (
1997
) Striatal recruitment during an implicit sequence learning task as measured by functional magnetic resonance imaging.
Hum Brain Mapp
 
5
:
124
–132.
Rizzolatti G, Luppino G, Matelli M (
1996
) The classic supplementary motor area is formed by two independent areas. In: Advances in neurology. Vol. 70. Supplementary sensorimotor area (Lüders HO, ed.), pp. 45–56. Philadelphia, PA: Lippincott-Raven.
Sakai K, Hikosaka O, Miyauchi S, Takino R, Sasaki Y, Putz B (
1998
) Transition of brain activation from frontal to parietal areas in visuo-motor sequence learning.
J Neurosci
 
18
:
1827
–1840.
Schlaug G (
1994
) Intersubject variability of cerebral activation in acquiring a motor skill: a study with positron emission tomography.
Exp Brain Res
 
98
:
523
–532.
Schlaug G, Sanes JN, Thargarav V, Darby DG, Jancke L, Edelman RR, Warach S (
1996
) Cerebral activation covaries with movement rate.
NeuroReport
 
7
:
879
–883.
Schluter ND, Rushworth MF, Nixon PD, Mills K, Passingham RE (
1998
) Temporary interference in human lateral premotor cortex suggests dominance for the selection of movements: a study using transcranial magnetic stimulation.
Brain
 
121
:
785
–799.
Schneider W, Shiffrin RM (
1977
) Controlled and automatic human information processing: I. Detection, search, and attention.
Psychol Rev
 
84
:
1
–66.
Tehovnik EJ, Sommer MA, Chou IH, Slocum WM, Schiller PH (
2000
) Eye fields in the frontal lobes of primates.
Brain Res Brain Res Rev
 
32
:
413
–448.
Toni I, Krams M, Turner R, Passingham RE (
1998
) The time course of change during motor sequence learning: a whole-brain fMRI study.
NeuroImage
 
8
:
50
–61.
VanMeter JW, Maisog JM, Zeffiro TA, Hallett M, Herscovitch P, Rapoport SI (
1995
) Parametric analysis of functional neuroimages: application to a variable-rate motor task.
NeuroImage
 
2
:
273
–283.
Von Economo C (
1929
) The cytoarchitectonics of the human cerebra cortex. London: Humphrey Milford/Oxford University Press.
Vorobiev V, Rizzolatti G, Matelli M, Luppino G (
1998
) Parcellation of human mesial area 6: cytoarchitectonic evidence for three separate areas.
Eur J Neurosci
 
10
:
2199
–2203.
Watkins KE, Gadian DG, Vargha-Khadem F (
1999
) Functional and structural brain abnormalities associated with a genetic disorder of speech and language
Am J Hum Genet
 
65
:
1215
–1221.
Wise SP, Boussaoud D, Johnson PB, Caminiti R (
1997
) Premotor and parietal cortex: corticocortical connectivity and combinatorial computations
Annu Rev Neurosci
 
20
:
25
–42
Yarbus AL (
1967
) Eye movements and vision. New York: Plenum Press.
Zingale CM, Kowler E (
1987
) Planning sequences of saccades.
Vision Res
 
27
:
1327
–1341.
Zuber BL, Stark L, Cook G (
1965
) Micro-saccades and the velocity– amplitude relationship for saccadic eye movements.
Science
 
150
:
1459
–1460.