Ocular fixation is the ability to keep the eyes stable on the object foveated. The object can be present (visual fixation) or absent. In the latter case, the fixation is memory-guided as the subject voluntarily holds his/her gaze on the remembered location of the object. Since Dodge (1907), visual fixation has been extensively studied, and several types of micro-phenomena have been described: tremor, drifts, and microsaccades (Leigh & Zee, 1999; Martinez-Conde, Macknik, & Hubel, 2004; Yarbus, 1967). Nonspecific to and fro saccadic eye movements, called square-wave jerks (SWJ) because of their particular pattern, have also been observed in normal subjects (Elidan, Gay, & Lev, 1984; Shallo-Hoffmann, Sendler, & Mühlendyck, 1990). Ohtsuka, Mukuno, Ukai, and Ishikawa (1986) proposed a continuum between microsaccades and SWJ that would vary just in amplitude and would reflect normal physiological instability of ocular fixation. The debate about the role of fixational eye movements in visual perception was revived in the last few years. Contrary to Kowler and Steinman (1979) but in line with Ditchburn (1980), an increasing number of studies showed that microsaccades are essential for vision. Particularly, electrophysiological recordings of neural responses in several visual areas of the monkey showed that microsaccades prevent the fading of visual perception by refreshing the retinal images (for a review, see Martinez-Conde et al., 2004). Cognitive processes can modulate fixational eye movements. The occurrence and amplitude of the different eye movements occurring during a fixation task have been found to increase with increasing attentional demands (Kowler, 1991; Shaffer, Krisky, & Sweeney, 2003). Similarly, the oculomotor instability in memory-guided fixation was found to be increased by comparison with visual fixation (Kowler, 1991; Leigh & Zee, 1999; Smyrnis et al., 2003). Recently, the amplitude and direction of microsaccades were found to be modulated by attention: The microsaccades would constitute an overt measure of covert attention shifts (Engbert & Kliegl, 2003; Hafed & Clark, 2002).

The suppression task is an attentional paradigm asking the subject to keep his eyes fixating the location of the fixation point, and ignoring distracters appearing in the periphery (i.e., suppressing or inhibiting eye movements toward them). In the suppression task, the fixation can be either visual or memory-guided. Fukushima, Fukushima, Morita, and Yamashita, (1990) reported a high rate of unwanted saccades in many of their schizophrenic patients performing a visual fixation task, in the presence of flashed distracters. Most saccades observed in these clinical studies were reflexive (i.e., showing express latencies between 80 and 120 ms). Smyrnis et al. (2003) studied normal subjects and found increased saccade frequency in a visually guided fixation task when distracters were flashed at the periphery, by comparison with a fixation task without distracters. Saccade suppression tasks have been used in a positron emission tomography (PET) study by Law, Svarer, Holm, and Paulson (1997), who reported brain activity mainly in the right hemisphere. Consistently, Tzelepi, Lutz, and Kapoula (2004) showed that EEG activity was enhanced on the right hemisphere in subjects asked to suppress saccade or vergence eye movements.

Another task that involves suppression or inhibition is the antisaccade paradigm (Hallett, 1978). In this task, after the extinction of a central fixation point (FP), a target appears in the periphery and the subject is instructed to look at the opposite side. As in studies of prosaccades, the temporal relationship between the offset of the FP and the onset of the target may be simultaneous (the target appears simultaneously with the FP offset), gap (a time interval is introduced between the FP offset and the target onset), or overlap (the FP remains on when the target appears). The task may also be memory-guided (e.g., see Kapoula, Bucci, Bernotas, & Zamfirescu, 2000). In the antisaccade task (mostly gap), a high rate of erratic prosaccades toward the stimulus was reported in subjects with either fixation and/or inhibitory abilities in immature children (Fischer, Biscaldi, & Gezeck, 1997a) or impaired in express saccade makers (Biscaldi, Fischer, & Stuhr, 1996; Cavegn & Biscaldi, 1996). Antisaccade deficit was also observed consecutively to a lesion of the dorsolateral prefrontal cortex (DLPFC) (Pierrot-Deseilligny et al., 2003a; Pierrot-Deseilligny, Rosa, Masmoudi, Rivaud, & Gaymard, 1991) or of the anterior cingulate cortex (Gaymard et al. 1998; Milea et al., 2003). Similar defect has been reported in Huntington's disease ( Lasker, Zee, Hain, Folstein, & Singer, 1987; Rothlind, Brandt, Zee, Codori, & Folstein, 1993), progressive supranuclear palsy ( Blin et al., 1995; Pierrot-Deseilligny, Rivaud, Pillon, Fournier, & Agid, 1989), and schizophrenia (for reviews, see Everling & Fischer, 1998; McDowell & Clementz, 2001). Brain imaging studies revealed a widely distributed neural network during the antisaccade performance (for a review, see Munoz & Everling, 2004).

The aim of the present study was to measure the ability of normal subjects to inhibit eye movements in a suppression task. In a gap paradigm, a distracter appeared in different locations in three-dimensional (3D) space, on the left or on the right, up or down, at close or at far. The movement to suppress could be a saccade, either lateral or vertical, or vergence, either convergence or divergence. We expected the suppression performance not to be perfect, and the errors to be distributed heterogeneously in 3D space, depending on spatial attentional allocations. Such expectations were partially based on prior studies showing shorter saccade latencies at close than at far (Bucci, Kapoula, Yang, Wiener-Vacher, & Bremond-Gignac, 2004; Yang, Bucci, & Kapoula, 2002), and differences in EEG cortical activation for saccades and vergence ( Kapoula, Evdokimidis, Smyrnis, Bucci, & Constantinidis, 2002; Tzelepi et al., 2004).

Eight healthy subjects (four women and four men) participated in the experiment. They were aged 22.6 ± 4.4 years (range = 19-31 years) and had 13.4 ± 1.2 years of education (range = 12-15 years). All were naïve to the purpose of the experiment and had never participated in eye-movement recordings. They gave their written consent and were paid for the present study, which was approved by the ethics committee and consistent with the Declaration of Helsinki.

A brief clinical examination was performed. The handedness, assessed by Edinburgh Handedness Inventory (Oldfield, 1971), was right dominant for the group of subjects (97.0 ± 5.6%, range = 87-100%; 100% indicating absolute right-handedness). Six of the eight subjects were right eye dominant. Subjects had no neuro-ophthalmological deficit and binocular vision without correction was normal. The proximal point of convergence was 58 ± 17 mm for the group of subjects. The amplitude of convergence, measured with a bar of prisms, was for the group of subjects 22.4 ± 6.5 diopters at 20 cm and 17.3 ± 7.8 diopters at 150 cm. The amplitude of divergence was 10.4 ± 3.9 diopters and 3.6 ± 1.3 diopters at 20 and 150 cm, respectively. These values were normal (von Noorden, 2002). Stereoscopic vision, assessed by the TNO (Test of Netherlands Organization; Richmond Products, Boca Raton, FL), was normal: median of 30 arcsec (range = 15-60").

Eye positions were recorded binocularly with a Skalar Iris infrared light eye-movement device (Skalar Medical b.v., The Netherlands). As we recorded successively horizontal and vertical eye movements, the sensors had to be clipped and fixed into the frame in alternatively the horizontal or the vertical direction. The system had a bandwidth of 100 Hz and was able to measure linearly within 3% horizontal eye movements up to 25° or vertical eye movements up to 16°. The optimal resolution was 2 arcmin. Data collection was performed using REX software (available online at http://www.tchain.com by Timothy C. Hain; Northwestern University Medical School, Chicago, IL). Eye position signals were stored on a disk for later off-line analysis. Each channel was sampled at 500 Hz. Another computer was used to monitor the triggering of the diodes.

For horizontal eye-movement recording, head movements were minimized by a bite board, which was attached to the table. For vertical eye-movement recording, the head was stabilized by both a chin rest and a forehead support.

The subjects sat facing a table (either horizontal or vertical) in which were embedded 2.9-mm diameter red light-emitting diodes (LEDs). The subjects performed four different conditions. As the task required suppression of eye movements (see Procedure), we may refer, in the following paragraphs, either to the distracter to ignore or to the movement to suppress for the description of the spatial arrangement. The amplitudes of vergence (to suppress) were calculated considering the mean interpupillar distance for the group of subjects (61 ± 2 mm).

For conditions 1 and 2, the horizontal table was adjusted at eye level, and the diodes were arranged on isovergence circles at three viewing distances, 20, 40, and 150 cm. Under all conditions, the initial fixation diode was in the primary position.

In condition 1 (see Figure 1A, left), we used six diodes. This setup is standard for the investigation of saccade and vergence in far and close space (see Bucci et al., 2004; Coubard, Daunys, & Kapoula, 2004; Kapoula, Isotalo, Müri, Bucci, & Rivaud-Péchoux, 2001; Yang & Kapoula, 2003). Three diodes were placed 150 cm away from the subject's eyes, one in the center, and two at ±10°. The required mean angle of vergence for fixating any of these diodes was 2.3°. The other three diodes were placed at 20 cm, one in the center and two at ±10°; the mean angle of vergence was 17.0°. The fixation stimulus was either the central diode at 20 cm or that at 150 cm. Thus, this condition called for six types of movements to suppress: saccades to the left or to the right at either 20 or 150 cm and convergence and divergence between the two central diodes at 20 and 150 cm. The amplitude of the required vergence to suppress was 14.7° (17.0-2.3°) for both convergence and divergence.

In condition 1, the FP could be far or close. It was of interest to study the ability to suppress eye movements when only one fixation point was required. Thus, we introduced a second condition using five diodes (see Figure 1A, middle). This setup is also standard for the investigation of saccade and vergence (see Takagi, Frohman, & Zee, 1995; Tam & Ono, 1994). The fixation stimulus was a central diode placed at 40 cm from the subject's eyes. The mean angle of vergence for fixating this diode was 8.7°. Four distracters could be lighted, one at a time at 10° left or right (for saccades), in the center at 20 cm (for convergence), or at 150 cm (for divergence). The amplitude of the required vergence to suppress was 8.3° (17.0-8.7°) for the convergence and 6.4° (8.7-2.3°) for the divergence.

In conditions 3 and 4 (Figure 1A, right), the subjects faced a vertical plane of LEDs. The fixation stimulus was a diode located centrally at the subject's eyes level. Two distracters could be lighted, one at a time at 7.5° up or down. Thus, there were two eye movements to suppress in these conditions: upward and downward saccades. The condition was performed either 40 cm (condition 3) or 150 cm (condition 4) from the subject's eyes.

To sum up, the subjects had to suppress 14 different types of eye movements: horizontal leftward or rightward saccades at a viewing distance of 20 cm (condition 1), 40 cm (condition 2), or 150 cm (condition 1); convergence with amplitude of either 8.3° (condition 2) or 14.7° (condition 1); divergence with amplitude of either 6.4° (condition 2) or 14.7° (condition 1); and upward or downward saccades at a viewing distance of either 40 cm (condition 3) or 150 cm (condition 4).

The experiment was conducted in a dark room. When all diodes were switched off, the luminance of the background of both the horizontal table and the vertical plane was 0 cd.m^–2. When switched on, the minimum luminance of each diode was 0.0016 cd.m^–2, which was well above the perceptual threshold.

In a distracter task (main condition; see Figure 1B), the central fixation diode was turned on for 2-2.5 s for each trial. The subject was instructed to fixate the stimulus as accurately as possible. Then, the fixation diode was switched off, and a distracter diode appeared 200 ms later (after the gap period). The subject was instructed to maintain his gaze on the location of the initial fixation stimulus (i.e., to fixate the remembered location of the initial fixation diode, no matter where the distracter appeared). The distracter diode stayed on for 1.5 s. A pilot study performed on ourselves revealed that the task was effortful and somewhat irksome, and was likely to induce tiredness. Thus, to maintain the subject's attention, we introduced a beep at the end of each trial, simultaneously with the distracter's offset. The beep indicated the 2-s pause, during which the subject was invited to blink systematically to prevent fatigue effects.

As a control condition, we used a no-distracter task (see Figure 1B) in which conditions were similar but no distracter appeared. The central fixation was turned on for 2-2.5 s. Then it was switched off and no distracter appeared. The instruction given to the subject was to fixate the remembered location of the fixation stimulus until a beep occurred. As in the distracter task, the beep indicated the end of the trial, and the subject was invited to blink systematically during the 2-s pause period.

The design is presented in Table 1. Each subject performed several blocks. Each block contained 30-60 trials preceded and followed by calibration (see the following section). In total, each subject completed 30 trials per experimental condition. In the distracter task (see Table 1, Distracter Task), each subject performed 7 blocks: 3 blocks in condition 1, 2 blocks in condition 2, and 1 block in conditions 3 and 4. The no-distracter task was performed in separated blocks. Indeed, it was shown that mixing trials aiming to assess inhibitory abilities (e.g., antisaccade trials) with other trials different in nature (e.g., prosaccade trials) interferes significantly on the performance of both types of trials (Weber, 1995). In the no-distracter task (see Table 1, No-Distracter Task), each subject performed 4 blocks, one for each condition. The experiment lasted about 1.5 hr in all for each subject, always with a pause between 2 blocks.

The order of presentation of the conditions was counterbalanced as closely as possible. Distracter location was randomized within each block.

For calibration, the subjects made a sequence of saccades at the beginning and at the end of each block of the distracter or no-distracter tasks. These movements helped to keep subjects alert and attentive.

In condition 1, the subject fixated the diodes switched on one after the other, at 0° (centrally) and 5°, 10°, 15° on the left and on the right, at 20, and at 150 cm. In condition 2, the subject fixated the diodes located at 0°, 5°, 10°, and 15° on the left and on the right at 40 cm. In conditions 3 and 4, the subject fixated the diodes located at 0°, 7.5°, and 15° upward and downward.

Each diode remained lighted on for 1.5 s, and the recording of eye movements was performed continuously and binocularly. A polynomial calibration was established by using fixation periods lasting at least 100 ms with an error staying below 0.5°. This method provided a very reliable calibration, as indicated by an average fixation error remaining below 1°.

We calculated two signals from the raw position records. A saccade or conjugate signal was calculated using the average of the calibrated left eye and right eye positions. A vergence or disconjugate signal was calculated by subtracting the right eye position from the left eye position, so convergence resulted in a positive deviation and divergence in a negative deviation. In the present study, we restricted the analysis on the conjugate signal for horizontal or vertical saccades, and the disconjugate signal for convergence and divergence. The rationale was to detect eventual movements toward or away from the direction or depth of the distracter. Erroneous movements in depth in response to direction distracters or movements in direction in response to depth distracters may happen but their analysis was beyond the scope of this study.

For each trial, we focused the analysis on a time window reduced to a period of 1 s (indicated by a dotted rectangle in Figure 1B) and including the successive parts as follows: the last 400 ms of the fixation period, the 200 ms of the gap period, and the first 400 ms of the test period, in which the subject was instructed to fixate the remembered location of the initial fixation stimulus. The last 400 ms of fixation were considered as the baseline. The test period was reduced to the first 400 ms, which is the standard upper latency limit in adults for slow regular saccades (Fischer, Gezeck, & Hartnegg, 1997b).

We rejected trials contaminated by blinks and those for which the fixation (the last 400 ms of fixation) was unstable. We considered the fixation as unstable when the amplitude of the position signal (either saccade or vergence) deviated over 0.5°. In all, 15.8% of the trials (range = 7.4-28.6%) were excluded from the analysis, which was consistent with a prior report (Coubard et al., 2004; 14.9% of rejection).

We grouped the data of all subjects and performed the following two analyses.

The first step of the analysis was to investigate whether the subjects produced erratic eye movements (i.e., movements detectable by standard velocity criteria). In the present study, we called pro-movements the erratic movements in the direction of the distracter (either pro-saccades for distracters located laterally or altitudinally or provergence for distracters located in depth). When the erratic movements were directed away from the distracter, we called them anti-movements (either antisaccades or antivergence).

In our experiment, the subjects were asked to fixate a remembered location. An important methodological issue was to determine whether the eye movements observed in our experiment were only inherent to fixational activity or related to the task demands. As mentioned in the Introduction, fixation activity has three main components: tremor, slow drifts, but also small saccades. Among the latter, SWJ are rapid, to and fro, saccadic eye movements, which form rectangular waves, with varying amplitude, frequency, and width. Shaffer et al. (2003) considered as SWJ saccades with an amplitude between 0.15 and 3° and an inter-saccades delay (time interval between the to and fro saccades) up to 400 ms (see also Discussion). Our criteria to distinguish eye movements related to our task from those occurring naturally during fixation were the following:

(i) Eye movements occurring after the onset of distracter were first detected by velocity crite-rion. The eye velocity of either conjugate or disconjugate signal was computed using a symmetrical two-point differentiator after low-pass filtering with a Gaussian FIR filter with a cut-off frequency of 33 Hz. The detection of either saccade or vergence was done automatically, and checked by the experimenter. The onset of saccades was determined at the point when the eye velocity of the conjugate signal exceeded 15°/s. The onset of vergence was determined at the point when the eye velocity of the disconjugate signal exceeded 5°/s. These criteria are standard (Coubard et al., 2004; Takagi et al., 1995).

(ii). The minimum amplitude of the pro- or anti-movements had to be 0.5°. Studies with normal subjects showed that movements with amplitude above this limit attest a failure in maintaining fixation either in the horizontal axis ( Pitzalis & Di Russo, 2001; Smyrnis et al., 2003) or in the vertical axis (Pitzalis & Di Russo, 2001). As commented in the Discussion, the 0.5° limit should exclude most microsaccades.

(iii). SWJ typically consist in a saccade away from the intended position of fixation followed, up to 500 ms later, by a return saccade to the FP (for a review, see Elidan et al., 1984). To exclude such movements, we did not select eye movements that were either preceded in the last 500 ms, or followed in the next 500 ms, by a movement in the opposite direction. For example, if the onset movement occurred 400 ms after distracter onset, we checked that no movement occurred in the opposite direction, backward for the period from –100 ms to +400 ms and forward for the period from +400 ms to +900 ms. Thus, we selected eye movements occurring singly for a time period of 1 s, which was centered on the onset of the movement. Our criterion of 500 ms for the inter-saccades interval of SWJ is also commented in the Discussion.

The effects of the type of movement, the direction of distracter, and the viewing distance on the frequencies of pro- and anti-movements were tested using Friedman ANOVA for the group of subjects; two-by-two comparisons were performed with Wilcoxon test.

Beside erratic pro- and anti-movements, it was interesting to search further for deviations in the position signal, which would drift toward the location of the distracter (or in the opposite direction), even though the movement failed to reach the velocity threshold and to elicit a clear eye movement.We grouped all remaining signals for the group of subjects and averaged them for both the distracter task and the no-distracter task. The group averages for both tasks were plotted together for each type of eye movement. Statistics were applied to the amplitude difference between the two tasks. For our window of interest (from –600 to +400 ms relative to the distracter onset), we divided the time into intervals of 20-ms width. Global difference between the distracter and the no-distracter tasks was tested using Friedman ANOVA. When significant, two-by-two comparisons were performed between time intervals using Wilcoxon test.

Here we report eye movements that were detected after the onset of the distracter. Using our selective criteria (see Methods), no movement was found in the no-distracter task. Thus, the movements described below only concern the distracter task. The errors were equally distributed throughout the experiment and no learning or fatigue effect was found. The following description first emphasizes the frequencies of errors (pro- and anti-movements) in the different conditions, and indicates afterward their temporal and dynamic characteristics.

Figure 2 shows the trajectories and Table 2 presents the numerical values. Distracters located either leftward or rightward yielded prosaccades in all conditions (mean = 2.64%). Spontaneous antisaccades were also observed at a similar rate (mean = 2.64%).

When pro- and antisaccades were pooled, we found a distance effect. The error rate increased significantly with decreasing viewing distance: 1.45% at 150 cm, 2.08% at 40 cm, and 4.37% at 20 cm (Chi²_8,2 = 6.9, p = .0323).

The leftward distracter at 20 cm yielded the highest error rate: 6.67% of prosaccades and 5% of antisaccades (see Figure 2A, left). At 40 cm, prosaccades occurred at similar rates for leftward (1.25%) and rightward (2.08%) distracters (Z < 1). Antisaccades to leftward distracter (i.e., rightward saccades) were more frequent than antisaccades to rightward distracter (i.e., leftward saccades): 4.58% and 0.42%, respectively (Z₈ = 2.0, p = .0431). At 150 cm, the error rate was low and similar for the left side and the right side. All together, these results suggest more errors when the distracter was leftward and at a close distance. The highest error rate was for the leftward distracter at 20 cm. At 40 cm, rightward saccades were the most frequent, but only in response to the leftward stimulus.

All subjects showed prosaccades (range = 0.56-7.78%) or antisaccades (range = 0.56-5%). Nevertheless, the absence of errors for some subjects in specific conditions did not allow a statistical analysis of their parameters (latency, amplitude, velocity; see Table 2), so we comment only qualitatively.

Almost all latencies were slow regular ones (181-400 ms; Fischer et al., 1997b). Prosaccades tended to show shorter latencies and higher amplitude than did antisaccades. However, as shown in Figure 2A, prosaccades tended to exhibit two subpopulations: A first population of saccades had a latency below 300 ms and a large amplitude; a second one had a higher latency (above 300 ms) and a small amplitude. Taken together, the amplitude of prosaccade reached at least 40% of distracter eccentricity, and it got closer to the latter with increasing viewing distances. The absolute mean was 4.3° at 20 cm, 4.15° at 40 cm, and 8.8° at 150 cm for both leftward and rightward prosaccades.

The population of antisaccades was more homogeneous than that of prosaccades. Their latency and amplitude were similar for both sides and independent of viewing distances. The latency was in the range of that of small prosaccades (above 300 ms). The amplitude did not exceed 10% of distracter eccentricity.

To summarize, the ability to suppress saccades toward lateral distracters was not perfect, and it led to pro- and antisaccades at similar rates. The error rate increased with proximal viewing distance, and was higher for leftward distracters, especially at close distance.

Trajectories are shown in Figure 3 and numerical values in Table 3. Recall that we differentiated four types of vergence: convergence from either 40 cm (see Figure 3A) or 150 cm (see Figure 3B) and divergence from either 40 cm (see Figure 3C) or 20 cm (see Figure 3D).

Distracters calling for convergence yielded the highest error rates (see Figure 3A-B). The errors were mostly directed toward the distracter (i.e., convergence), while antivergence divergence was either scarce or null. The vicinity of the FP and the distracter in depth tended to influence the error rate. Indeed, a convergence distracter (at 20 cm) with the eyes fixating at 40 cm yielded 8.75% convergence, while fixation at 150 cm yielded 5.42% convergence. However, the difference failed to reach statistical significance (Z₈ = 1.05, p = .294).

Distracters calling for divergence yielded almost insignificant error rates (see Figure 3C-D). Only 0.42% of divergence was found, whatever the initial fixation point, and little convergence was observed.

All subjects produced erratic provergence (range = 1.67-14.2%), but only four of the eight subjects exhibited antivergence (range = 0.83-1.67%). Thus, no statistical analysis was performed on timing and dynamic parameters. As shown in Table 3, the latencies of errors were in the range of fast regular movements (121-180 ms; Fischer et al., 1997b), and some scarce express movements could be observed. The amplitude of divergence (either pro- or antivergence) did not exceed 1°, whatever the initial point. In contrast, for distracters calling for convergence, unwanted convergence could reach the amplitude of the distracter depth (see Figure 3A-B). Nevertheless, in most convergence trials (for 71% and 77% of trials, for convergence from 40 cm and convergence from 150 cm, respectively), unwanted convergence stayed below an amplitude of 1°, which corresponds to 12% and 7% of distracter depth for convergence from 40 cm and convergence from 150 cm, respectively.