Free
Article  |   June 2012
Saccadic inhibition is accompanied by large and complex amplitude modulations when induced by visual backward masking
Author Affiliations
  • Alain Guillaume
    Laboratoire de Neurosciences Cognitives, Centre National de la Recherche Scientifique, Aix-Marseille Université, Marseille, France
    alain.guillaume@univ-amu.fr
Journal of Vision June 2012, Vol.12, 5. doi:10.1167/12.6.5
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Alain Guillaume; Saccadic inhibition is accompanied by large and complex amplitude modulations when induced by visual backward masking. Journal of Vision 2012;12(6):5. doi: 10.1167/12.6.5.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract
Abstract
Abstract:

Abstract  Saccadic inhibition refers to the strong temporary decrease in saccadic initiation observed when a visual distractor appears shortly after the onset of a saccadic target. Here, to gain a better understanding of this phenomenon, we assessed whether saccade amplitude changes could accompany these modulations of latency distributions. As previous studies on the saccadic system using visual backward masking—a protocol in which the mask appears shortly after the target—showed latency increases and amplitude changes, we suspected that this could be a condition in which amplitude changes would accompany saccadic inhibition. We show here that visual backward masking produces a strong saccadic inhibition. In addition, this saccadic inhibition was accompanied by large and complex amplitude changes: a first phase of gain decrease occurred before the saccadic inhibition; when saccades reappeared after the inhibition, they were accurate before rapidly entering into a second phase of gain decrease. We observed changes in saccade kinematics that were consistent with the possibility of saccades being interrupted during these two phases of gain decrease. These results show that the onset of a large stimulus shortly after a first one induces the previously reported saccadic inhibition, but also induces a complex pattern of amplitude changes resulting from a dual amplitude perturbation mechanism with fast and slow components.

Introduction
Saccadic eye movements allow us to explore our visual environment by aligning the fovea with targets of interest in the visual scene. In reaction to a stimulus appearance, a saccade will be produced to identify it. Latency of this saccade with respect to target appearance will belong to the classical reaction time distribution of saccades with a median around 180–200 ms (Becker, 1989). It has been shown that if another stimulus, a distractor, appears shortly after the first one that has been selected for the saccade, there is a strong temporary decrease in the probability of saccadic initiation around 90 ms after its onset, hence creating a dip in the classical latency distribution (Reingold & Stampe, 2002, 2004). Studying this so-called “saccadic inhibition” phenomenon may shed light on the control of saccade initiation and more generally on the organization and functioning of the human oculomotor system (for example, Bompas & Sumner, 2011). 
The saccadic inhibition phenomenon consists of modulations of latency distributions. It probably results from a competition between the target and the distractor in visuomotor maps (Reingold & Stampe, 2002, 2004). In addition to being involved in the initiation processes, these maps obviously also contain information regarding saccade metrics. The question addressed by the present study concerns potential saccadic amplitude changes that may accompany these latency distribution modulations. Previous studies concerning saccadic inhibition either reported no accompanying amplitude changes (Reingold & Stampe, 2002, 2004) or did not look for it (Buonocore & McIntosh, 2008; Bompas & Sumner, 2011). The only mention of a significant amplitude change was reported by Edelman and Xu (2009) in a task without the presence of a visual target at the saccade time, i.e., in a task requiring saccades toward remembered targets. In this task, a distractor appearing after the go-signal for a saccade toward a remembered target induced saccadic inhibition accompanied by a decrease in saccadic amplitude. The discrepancy concerning amplitude modulation in previous studies (no change vs. change) maybe be related to the fact that saccades toward remembered targets are associated with less intense neuronal activation (e.g., Edelman & Goldberg, 2001). This lower neuronal activation related to the target may bias the result of the competition in favor of the distractor (see Discussion for a proposed explanation of amplitude change given by Edelman & Xu, 2009). 
Given this hypothesis of competition among visuomotor maps, the aim of the present work is to test whether an amplitude change would accompany saccadic inhibition in a task requiring visually guided saccades (stable visual targets) and involving a powerful distractor. Compared to punctate stimuli that have been used as distractors in previous studies, one such stronger distractor could be formed by a visual mask. Visual backward masking consists of inducing a time-controlled disruption of the processing of a visual stimulus (the target) by obscuring it after a very short time interval with a second generally large stimulus (the mask) (Macknik, 2006; Ansorge et al., 2007). Hence, a second visual stimulus is presented shortly after the first one. 
Visual backward masking was used in two studies aiming to decipher whether perceptual localization and saccade programming were based on the same target representation (Aitsebaomo & Bedell, 1992; Eggert et al., 2002). The visual mask was used to interrupt visual processing in order to equate the processing time in both visual perception and saccade production tasks. In both studies, a saccadic hypometria was observed when the mask interrupted the stimulus processing for saccade production. In addition, Aitsebaomo and Bedell (1992) showed that there was a relationship between saccade gain and the target duration (see their Figures 1 and 2). Interestingly, in addition to amplitude modulations, both studies also noted latency increases. Nevertheless, no quantitative analysis of latency data and hence no characterization of latency distributions were performed. We suspected that these latency modulations could be the result of a saccadic inhibition due to the subsequent appearance of a second stimulus consisting of a mask shortly after the target onset. Indeed, the mask could have induced a dip in the latency distribution (saccadic inhibition) and hence postponed some saccades, leading to higher latency values. If this scenario is correct, we should be able to observe saccadic inhibition with accompanying amplitude changes. This will allow us to study the relationships between these changes in latency and amplitude. 
Results of the present study will show that when the target of a saccadic eye movement is quickly masked after its appearance (visual backward masking), there is a strong transient decrease in the probability of saccade initiation, a saccadic inhibition, that occurs around 90 ms after the mask appearance, similar to what has been previously reported for the appearance of a non-overlapping stimulus (Reingold & Stampe, 2002, 2004) or of a remote punctate distractor (Graupner et al., 2007; Buonocore & McIntosh, 2008; Edelman & Xu, 2009). By plotting the gain of saccades as a function of their initiation time we also observed a complex pattern of saccade gain evolution: approximately 50 ms after the mask onset, saccadic gain decreased until the beginning of the period of saccadic inhibition. When saccades began to occur again, their gain was around 1.0 and rapidly entered into a second decrease phase. Part of this work has already been published in abstract form (Guillaume, 2009). 
Materials and methods
Participants, experimental set up, and eye movements recording
Twenty volunteers from the Aix-Marseille University (6 females and 14 males; ages between 18 and 41 years) participated in the study. All had normal or corrected-to-normal vision. They were naïve as to the purpose of the study and gave their informed consent to participate. The study was performed in accordance with the ethical standards laid down in the Declaration of Helsinki (last modified, 2004). Participants were installed in a completely dark room in a chin rest at a distance of 46 cm from a 20-in. CRT Monitor (Silicon Graphics, GDM 4011P; screen of 37.4 × 28.1 cm). The resolution of the screen was 640 × 480 and the refresh rate was 150 Hz. A helmet-mounted infrared sensor (Eyelink II system, SR research, Mississauga, Ontario, Canada) allowed recording of the left eye position at 500 Hz. 
Visual stimulus
A green cross corresponded to the fixation point, and targets for saccades were red circles that subtended 0.5° of visual angle with smaller dark circles inside. The background was black. Targets were presented for short durations (13, 33, 53, 73, 93, and 113 ms) and, after this duration, were either switched off or covered by the presentation of a visual mask. The mask was a structured mask (i.e., a mask that shares many of the structural features of the target; Breitmeyer & Ogmen, 2000) that consisted of the three central rows of the screen filled with target patterns (Figure 1). 
Figure 1
 
Visual stimulus and task. Time course of the different trial types. Participants were required to fixate on a green cross for a random duration. After the fixation point offset, there was a gap of 250 ms, and then the target (a red circle with a dark circle inside) was presented for a short period of time (13, 33, 53, 73, 93, or 113 ms). After this duration, the target was either switched off (NoM condition) or covered by a pattern mask consisting of three rows of 79 target patterns each (simplified on this figure; FullM condition). Trials with a visual mask could also be of the HalfM type: the fixation point was displaced either to the left or to the right, and the mask was only covering half of the visual field. It consisted of three rows of 68 target patterns each. In the HalfM condition, the center of gravity of the mask was always located further away from the target, which was not the case for the FullM condition. This allowed testing for a potential global effect between the target and the center of gravity of the mask. See text for further details.
Figure 1
 
Visual stimulus and task. Time course of the different trial types. Participants were required to fixate on a green cross for a random duration. After the fixation point offset, there was a gap of 250 ms, and then the target (a red circle with a dark circle inside) was presented for a short period of time (13, 33, 53, 73, 93, or 113 ms). After this duration, the target was either switched off (NoM condition) or covered by a pattern mask consisting of three rows of 79 target patterns each (simplified on this figure; FullM condition). Trials with a visual mask could also be of the HalfM type: the fixation point was displaced either to the left or to the right, and the mask was only covering half of the visual field. It consisted of three rows of 68 target patterns each. In the HalfM condition, the center of gravity of the mask was always located further away from the target, which was not the case for the FullM condition. This allowed testing for a potential global effect between the target and the center of gravity of the mask. See text for further details.
As in Aitsebaomo and Bedell (1992), the fixation point was at the center of a screen in the first trial type so targets required centrifugal saccades with respect to the orbit center. The mask was a horizontal band that spanned the entire horizontal extent of the screen (FullM, Figure 1). Aitsebaomo and Bedell noted that hypometria observed in these masking experiments (see Introduction) could be due to a global effect (saccadic averaging phenomenon): when two targets are presented in spatial and temporal proximity, a saccade that lands in-between the two targets (an averaging saccade) could be elicited (Coren & Hoenig, 1972; Findlay, 1982; Ottes et al., 1984). Aitsebaomo and Bedell reasoned that hypometria could be the result of an average between the target and the center of gravity of the mask (considered here as a second target). They tested a supplementary condition with a displaced mask that put the center of gravity of the mask beyond the target. As they obtained identical results for this supplementary condition, they concluded that observed hypometria was not due to an averaging phenomenon. Nevertheless, they tested this condition of a displaced mask for only two target durations (33 and 50 ms). In the present study, we systematically tested for each target duration a condition comprising of a displaced mask with the center of gravity always beyond the target. We named this condition HalfM (Half Mask, Figure 1) because the mask stimulated only half of the visual field from the fixation point to the opposite border of the screen whereas, in FullM condition, the mask stimulated the whole visual field. Hence, two similar masks of different sizes were used. The FullM consisted of three rows of 79 target patterns while the HalfM consisted of three rows of 68 target patterns (Figure 1). 
All visual presentations were controlled at the ms scale with the Psychophysics Toolbox for Matlab (Brainard, 1997; Pelli, 1997). 
Protocol and task
There were three types of trial: with no mask (NoM), with a full mask (FullM), and with a half mask (HalfM) (Figure 1). 
In NoM and FullM, the green cross appeared at the center of the screen for a period of random duration (800-1400 ms). After it disappeared, there was a gap period of 250 ms after which a target was presented on the horizontal central row of the screen for one of the six durations. The gap paradigm was chosen to maximize the presence of short latency saccades. Hence, the effects of the mask could be studied on these short latency saccades and, globally, on a wide array of saccade latencies. There were 16 possible target locations: 4°, 6°, 8°, 10°, 12°, 14°, 16°, or 18° to the left or the right of the fixation point. After the duration of the target presentation, the target was either switched off (NoM) or immediately covered by the mask (FullM). In the latter case, the mask was set to screen width, and all possible target locations corresponded exactly to a given target pattern in the mask. The mask remained present for 1 second. Hence, the saccadic response was produced while the mask was still on the screen. A whole trial lasted 3 seconds. 
In HalfM, timing of events was the same as for FullM but there were only four possible target locations (4°, 8°, 12°, or 16°). The fixation point was shifted toward one border of the screen (16° of visual angle; randomly to the left or the right), and the targets appeared along the opposite direction. Hence, saccades were centripetal with respect to the orbit center. Under these conditions, the mask covered only half the visual field, from the fixation point to the opposite border of the screen. With this configuration, the center of gravity of the mask was always beyond the target. There was no equivalent to NoM with the displaced fixation point. 
Using many different targets was preferable to the use of only few targets on each side of the fixation point in order to avoid any possibilities of memorization of target location. As saccade gain was computed, values for all target positions in a given condition were merged. An analysis based on saccades produced in the NoM condition showed that there was a small effect of target position on saccade gain: a linear regression fit indicated a gain: decrease of 0.11 from the minimum to maximum eccentricity. 
Each session corresponded to 240 trials during which all combinations of target duration and position for the three mask conditions (NoM, FullM, and HalfM) appeared in a randomized order. In order to gain more data, each participant performed two experimental sessions on two different days. For each session the instruction to the participant was simply to fixate the green cross and to make a saccade to the red target as quickly and as precisely as possible. 
Data analysis
As all possible visual targets were on the horizontal meridian, only the horizontal component of eye movements was studied in detail. Saccade onset and offset were identified when instantaneous horizontal velocity exceeded 20°/s and dropped below 30°/s, respectively. Trials containing no saccade at all, saccades in the wrong direction, or anticipatory saccades (latency <70 ms) were discarded from further analysis. Only the primary saccades were analyzed. For each saccade, we computed its gain (actual saccade horizontal amplitude/required saccade horizontal amplitude), latency, and peak velocity. 
We conducted analyses of latency distributions and saccade gain as a function of time elapsed since the appearance of the target. These analyses were first performed separately for each Mask Condition (NoM, FullM, HalfM) and Target Duration (13–113 ms) on data from all participants (N = 20) pooled together. Similar analyses were then conducted separately for each participant but with data for all Target Durations pooled together and aligned on mask onset (see Results). 
Characterization of latency distributions
To characterize latency distributions as a function of time elapsed since the appearance of the target, we computed probability density functions with the kernel density estimation (ksdensity.m) of the Matlab Statistics Toolbox (Bandwith = 14; Van Zandt, 2000). The visual mask produced a strong transient decrease in saccadic initiation probability (see Results). A similar saccadic inhibition due to the onset of a second visual stimulus after the appearance of the main target has already been described by several studies (Reingold & Stampe, 2002, 2004; Buonocore & McIntosh, 2008; Edelman & Xu, 2009). Here, we used some of the measures defined by Reingold and Stampe (2002) (Figure 2): the time of maximum saccadic inhibition (MSI), the two time markers corresponding to 50% of maximum inhibition (50%SIbeg and 50%SIend), and the duration of the inhibition (= 50%SIend − 50%Sibeg). We added the measure of the first effect on latency distributions (FEL) which corresponds to the onset time of the first decrease in saccade density, in other words the time of the first peak in the probability density function (Figure 2). 
Figure 2
 
Measures taken to characterize the saccadic inhibition. The mask appearance always induced a strong transient decrease in probability of saccade initiation. The measures defined to characterize this inhibition are indicated on the probability density function computed from values of latency of all participants in the FullM – 113 ms condition (in grey, probability density function of NoM control values for the same target duration). FEL: first effects on latency = the onset time of the first decrease in saccade density. MSI: time of the maximum saccadic inhibition. 50%SIbeg: time corresponding to 50% of the saccadic inhibition − beginning of saccadic inhibition. 50%SIend: time corresponding to 50% of the saccadic inhibition − end of saccadic inhibition. SID: duration of the saccadic inhibition = 50%SIend − 50%SIbeg.
Figure 2
 
Measures taken to characterize the saccadic inhibition. The mask appearance always induced a strong transient decrease in probability of saccade initiation. The measures defined to characterize this inhibition are indicated on the probability density function computed from values of latency of all participants in the FullM – 113 ms condition (in grey, probability density function of NoM control values for the same target duration). FEL: first effects on latency = the onset time of the first decrease in saccade density. MSI: time of the maximum saccadic inhibition. 50%SIbeg: time corresponding to 50% of the saccadic inhibition − beginning of saccadic inhibition. 50%SIend: time corresponding to 50% of the saccadic inhibition − end of saccadic inhibition. SID: duration of the saccadic inhibition = 50%SIend − 50%SIbeg.
Characterization of gain evolution
The evolution of saccadic gain as a function of time elapsed since the appearance of the target was characterized by using a Locally Weighted Scatterplot Smoothing (lowess), a statistical method for robustly fitting smoothing curves without prior assumptions about the shape or form of the curve1 (Cleveland, 1979). The 95% confidence intervals (±1.96 * se) were added to the plots of saccadic gain as a function of latency. Finally, we used two measures to characterize these gain curves in the mask conditions: we measured the timing of the first effect on amplitude curves (FEA) and the timing of the dip in amplitude (DA) (see the description of Figure 4 in Results for details). 
Gain values were compared between the three conditions with a two-way ANOVA with factors “Mask Condition” (NoM, FullM, or HalfM) and “Bin of latency” (from 90–110 to 370–390 ms; 15 levels) (unbalanced design). Significant differences were followed up with Holm-Sidak post-hoc tests (statistical threshold of p < 0.05). This post-hoc method was selected for its control of type-I family-wise error rate that is required when numerous comparisons are performed. 
We also conducted an analysis of variability of gain evolution: for several time periods (see Results), gain variability was computed as the mean of standard error of the lowess fitting. For each target duration, a two-way ANOVA with factors “Mask Condition” (NoM, FullM, or HalfM) and “Time Period” (four levels) was then used to compare obtained variability values. Significant differences were followed up with Newman-Keuls post-hoc tests. 
After these analyses as a function of initiation time, in order to compare our data to those of Aitsebaomo and Bedell (1992), mean horizontal saccade gain was calculated (±SEM) for each target duration in each mask condition for each participant. This data was submitted to repeated measures ANOVAs with two within-subjects factors: “Mask Condition” (NoM, FullM, or HalfM) and “Target Duration” (13, 33, 53, 73, 93, or 113 ms). Newman-Keuls post-hoc tests were used to perform specific comparisons. 
Correlation and kinematics analyses
Correlation analyses were used to evaluate relationships between target duration and two parameters: the time of maximum saccadic inhibition (MSI) and the time of the dip in amplitude (DA). 
Finally, a one-way ANOVA allowed to compare differences in peak velocity between NoM and FullM saccades during four time periods (see Results). Newman-Keuls post-hoc tests were used to perform specific comparisons. 
Results
General observations
Across all tested conditions (target duration and mask presence), the only two in which participants sometimes reported (this was not systematic) that they did not perceive the target were those with the shortest target duration (13 ms) and the presence of a mask (FullM and HalfM). In these “no-perception trials,” a saccade in the target direction was nevertheless sometimes produced. As these cases represented only a minor part of the data and as the question of the perception was not central to the present study, these trials were included for analysis when a saccade was present without more special attention. 
The trial selection based upon the criteria described in the Materials and methods section was conducted and led to the rejection of 106/3840 trials (2.8%) in NoM, 616/3840 trials (16.0%) in FullM, and 580/1920 trials (30.2%) in HalfM conditions. The higher rejection percentage in the HalfM condition was due to a stronger tendency to anticipation, which was not surprising given the complete advance certainty about the upcoming saccade direction in this condition. 
The mask greatly reduced the probability of saccade initiation for a short period of time
The latency distributions revealed a strong transient saccadic inhibition following the mask appearance. The probability density functions were calculated for each of the three mask conditions (NoM, FullM, and HalfM) and for each target duration (see Materials and methods). These probability density functions contain values from all subjects for a given combination of target duration and mask condition. On each panel of Figure 3 (Panels a–f), the probability density functions of the three mask conditions have been superimposed for a given target duration. 
Figure 3
 
Inhibition of saccadic initiation produced by the mask appearance. (a–f) For each target duration, superimposition of the three probability density functions for NoM, FullM, and HalfM conditions. In each case, a strong transient decrease of saccade initiation was observed for FullM and HalfM conditions around 100 ms after mask appearance. See text for further details.
Figure 3
 
Inhibition of saccadic initiation produced by the mask appearance. (a–f) For each target duration, superimposition of the three probability density functions for NoM, FullM, and HalfM conditions. In each case, a strong transient decrease of saccade initiation was observed for FullM and HalfM conditions around 100 ms after mask appearance. See text for further details.
Figure 4
 
Saccade gain as a function of initiation time: Examples. (a) NoM condition, target duration = 113 ms. (b) FullM condition, target duration = 113 ms. In both cases, lowess regression applied to the data set (bold trace) and 95% confidence interval of the lowess regression (grey area) have been added. In addition, probability density functions corresponding to NoM − 113 ms and FullM – 113 ms have been added below. On the probability density function of FullM, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). FEA: first effects on amplitude = the time of the onset of the first gain decrease phase. DA: dip in amplitude = the time of the transition between the first gain decrease phase and the following increase phase. P1 to P4 correspond to four 40-ms periods used to assess gain variability. See text for further details.
Figure 4
 
Saccade gain as a function of initiation time: Examples. (a) NoM condition, target duration = 113 ms. (b) FullM condition, target duration = 113 ms. In both cases, lowess regression applied to the data set (bold trace) and 95% confidence interval of the lowess regression (grey area) have been added. In addition, probability density functions corresponding to NoM − 113 ms and FullM – 113 ms have been added below. On the probability density function of FullM, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). FEA: first effects on amplitude = the time of the onset of the first gain decrease phase. DA: dip in amplitude = the time of the transition between the first gain decrease phase and the following increase phase. P1 to P4 correspond to four 40-ms periods used to assess gain variability. See text for further details.
In every case, the first initiated saccades had very short latencies, around 110 ms for NoM and FullM, and slightly earlier (around 90 ms) for HalfM. These short latencies were the result of using the gap paradigm. The difference between the HalfM condition and the two others can be accounted for by the predictability in the direction of the target onset under this condition and by the fact that saccades were centripetal with respect to orbit center (Paré & Munoz, 1996; see Materials and methods). 
In the NoM condition, for each duration, we obtained a classical reaction time distribution that rose rapidly on the left and dropped off more slowly on the right (see Sumner, 2011, for review). For both FullM and HalfM conditions, plots of probability density functions showed that, irrespective of the target duration, a strong transient decrease in the probability of saccadic initiation always occurred around 100 ms after mask onset. Table 1 summarizes measures taken to characterize this saccadic inhibition (Figure 2). For example, in FullM with 113 ms target duration, the first effect of the mask appearance on the latency distribution (FEL) took place 63 ms after its appearance. The 50% level of saccadic inhibition (50%SIbeg) was reached in 87 ms, and the maximum of inhibition (MSI) occurred at 114 ms after the appearance of the mask. As the duration of the target was reduced, the number of saccades occurring before the probability dip was reduced because this dip moved closer to the bins of the first saccades (90–110 ms or 70–90 ms). This displacement of the probability dip finally produced the disappearance of the first peak in the latency distribution. This disappearance occurred for the 33 ms target duration in the FullM condition whereas it occurred only for the 13 ms target duration in the HalfM condition. This is explained by the presence of saccades with shorter latency in the HalfM condition. The disappearance of this first peak explains why some values are lacking in Table 1. Figure 6a shows the high correlation (r2 = 0.97, p < 0.001) between target duration (and hence mask appearance) and the timing of the latency dip. 
Table 1
 
Measured parameters characterizing latency distributions and gain evolution as a function of initiation time for FullM and HalfM conditions.
Table 1
 
Measured parameters characterizing latency distributions and gain evolution as a function of initiation time for FullM and HalfM conditions.
Time [with respect to mask onset] of… (in ms) Sacc inhib duration SID
… first effects on amplitude curves FEA … first effects on latency PDF FEL … the dip in amplitude DA … 50% sacc inhib (beginning) 50%SIb … max sacc inhib (dip in latency) MSI … 50% sacc inhib (end) 50%SIe
FullM – 113 56 63 86 87 114 155 68
HalfM – 113 42 49 77 83 132 158 75
FullM – 93 40 63 85 88 110 159 71
HalfM – 93 59 64 89 87 115 157 70
FullM – 73 66 88 91 109 157 66
HalfM – 73 60 88 94 123 169 75
FullM – 53 78 91 108 166 75
HalfM – 53 68 71 90 117 167 77
FullM – 33
HalfM – 33 81 99 116 156 57
FullM – 13
HalfM – 13
Mean values (± SD) 49.3 (± 9.6) 65.7 (± 9.6) 83.4 (± 6.8) 90.0 (± 4.6) 115.3 (± 7.5) 161.4 (± 5.9) 70.4 (± 6.2)
These plots of probability density functions showed that, for NoM and FullM conditions (i.e., comparable conditions in terms of target direction predictability and of direction of saccades with respect to the orbit center), the first initiated saccades occurred at the same time. Nevertheless, if mean values would be calculated, an increase of latency would be observed for FullM because of the inhibition that pushed away “remaining saccades” toward later times. This explains observations by Aitsebaomo and Bedell (1992) and Eggert et al. (2002) concerning latency increase in masking protocol. Hence, to characterize completely the effect of visual masking on saccade initiation, latency distributions have to be studied. 
Finally, the presence of the mask produced a decrease in the probability of saccadic initiation around 100 ms after its appearance, then aligning data of all target durations on mask onset should result in a probability density function with a single dip in initiation probability. This was found when data from all participants were pooled together (see Panel a of Supplementary Data, Figure S1). Importantly, we checked that the bimodality in latency distributions was not the result of different behaviors from different participants, e.g., some producing a large amount of short latency saccades and other less. The alignment on mask onset was also performed separately for each participant (Panels b–u of Supplementary Data, Figure S1). A dip in latency distribution was found in 18 participants out of 20. This result obtained with each participant considered separately validates the population analysis. 
Saccadic gain showed a complex pattern of evolution as a function of the time elapsed from the target onset
The evolution of saccade gain was studied as a function of saccadic initiation time. As this analysis was conducted separately for each target duration, it showed the evolution of the gain as a function of the duration of mask influence. The first analyses considered data from all participants pooled together. We subsequently carried out an analysis at the participant level. 
Example of gain evolution as a function of initiation time
Figure 4 shows plots of saccade gain as a function of initiation time for the conditions NoM − 113 ms (Panel a) and FullM − 113 ms (Panel b). Each circle is an individual saccade, and bold curves correspond to the lowess regressions (see Materials and methods). In the bottom part of Figure 4, we have added the probability density function of saccade initiation for these conditions (Figure 3). For the NoM condition, the mean gain of the very first initiated saccades was slightly hypometric. This has already been described by Fischer and Weber (1993). The gain then progressively increased and reached values of ∼1.0 for saccades with latencies greater than ∼155 ms and then remained around this value. 
Panel b shows what happened when a mask was presented immediately after the duration of the target (FullM). Saccades initiated very early were not disrupted by the mask, probably because their initiation occurred before the mask had been processed and had started interfering with saccade preparation. Hence, gain values for these earliest initiated saccades were similar to the NoM condition. Then gain values entered into a first decrease phase. We measured this timing of first effects on amplitude (FEA). It corresponded to the timing of transition between the gain increase observed for the first initiated saccades to this decrease phase. For the present example, it occurred 56 ms after the mask appearance (see Table 1 and Figure 5 for superimposed traces). This decrease phase lasted for around 40 ms before the lowess regression began to increase again, hence forming a “dip in amplitude.” The timing of this dip in amplitude (DA) formed a second measure used to characterize gain curves. Here, DA occurred 86 ms after mask appearance, a value that corresponded exactly to the 50% (beginning) of saccade inhibition (Table 1). When saccades reappeared after the saccadic inhibition phase (gray band on the lower part of Panel b), their mean gain had values around 1.0. Finally, a second phase of gain decrease began and lasted up to the last produced saccades, bringing the gain down to around 0.8. We also defined four time periods (P1 to P4) that will be used to assess response variability (see the paragraph concerning variability for details). 
Figure 5
 
Saccade gain as a function of initiation time. Right column contains for each target duration (Panels a–f), superimposition of the three lowess regressions for NoM, FullM, and HalfM. For the NoM condition, the 95% confidence interval (grey area) has been added. For longest target durations (Panel a and b: 113 and 93 ms) the mask presence induced a complex pattern of gain change that contained two decrease phases surrounding a return to gain values around 1.0. As target duration was decreased (from 73 ms to 13 ms, Panels c to f), first phases of this complex pattern of gain modulation was more and more cropped. On each plot, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). Tables above the x-axis of each plot contain results of post-hoc breakdown (Holm-Sidak tests) after ANOVAs (see text for design), grey box = (p < 0.05) and empty box (p > 0.05). Left column contains variability of gain values assessed through computation of mean of standard error for the four defined periods (Figure 4). Vertical grey bars symbolize saccadic inhibition. Arrows show cases for which post-hoc tests after ANOVAs revealed no significant difference between phase 2, 3, and 4. See text for further details.
Figure 5
 
Saccade gain as a function of initiation time. Right column contains for each target duration (Panels a–f), superimposition of the three lowess regressions for NoM, FullM, and HalfM. For the NoM condition, the 95% confidence interval (grey area) has been added. For longest target durations (Panel a and b: 113 and 93 ms) the mask presence induced a complex pattern of gain change that contained two decrease phases surrounding a return to gain values around 1.0. As target duration was decreased (from 73 ms to 13 ms, Panels c to f), first phases of this complex pattern of gain modulation was more and more cropped. On each plot, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). Tables above the x-axis of each plot contain results of post-hoc breakdown (Holm-Sidak tests) after ANOVAs (see text for design), grey box = (p < 0.05) and empty box (p > 0.05). Left column contains variability of gain values assessed through computation of mean of standard error for the four defined periods (Figure 4). Vertical grey bars symbolize saccadic inhibition. Arrows show cases for which post-hoc tests after ANOVAs revealed no significant difference between phase 2, 3, and 4. See text for further details.
Figure 6
 
Correlation analyses. The time of maximum saccadic inhibition (MSI, Panel a) and the time of the dip in amplitude (DA, Panel b) were highly correlated with the target duration that corresponds to the time of the mask onset. DA was also highly correlated with a parameter that characterizes latency distribution: the time of 50% of saccadic inhibition − beginning (Panel c).
Figure 6
 
Correlation analyses. The time of maximum saccadic inhibition (MSI, Panel a) and the time of the dip in amplitude (DA, Panel b) were highly correlated with the target duration that corresponds to the time of the mask onset. DA was also highly correlated with a parameter that characterizes latency distribution: the time of 50% of saccadic inhibition − beginning (Panel c).
Gain evolution for the different target durations
In the left column of Figure 5, we have superimposed the three lowess regressions corresponding to the three conditions for the six possible target durations (Panels a–f). For the NoM condition, the 95% confidence interval of the lowess fitting has been added to the plot. This confidence interval allowed us to easily evaluate the difference between the FullM or HalfM conditions and the NoM condition: a statistically significant difference was observed as soon as the FullM or HalfM curve left the NoM confidence interval. Nevertheless, we also added a more conservative comparison. Separately, for each target duration, a statistical analysis was conducted to compare gain values between the three conditions. Saccadic gain values were grouped in latency bins of 20 ms, from 90–110 ms to 370–390 ms. Then, values were entered into a two-way ANOVA with factors “Mask Condition” (NoM, FullM, or HalfM) and “Bin of latency” (15 levels) (unbalanced design). The main effect of each factor and the interaction were found to be statistically significant for all durations (p < 0.05). Results of specific comparisons through post-hoc tests (Holm-Sidak post-hoc tests, see Materials and methods) were indicated above the x-axis of each graph of the left column; significant difference (p < 0.05) for a bin was indicated by a grey box. 
For 113 ms (Panel a), we can observe that the fitting curve for the HalfM condition was very similar to the FullM one (no statistically significant difference). It also showed a first phase of gain decrease, a return to accuracy after the saccadic inhibition, and, finally, a second phase of gain decrease. In addition, at no time, was gain value for HalfM larger than that for FullM, a result that should have been found if the effect of the mask on saccade gain was fully or partly due to a global effect with the center of gravity of the mask (see Materials and methods). Table 1 contains defined measures (FEA and DA) used to characterize the gain modulation pattern. For this target duration, in the HalfM condition, the FEA and the DA occurred at 42 ms and 77 ms, respectively, after the mask appearance. In both the FullM and HalfM cases, the DA value was very similar to the value of 50% of saccade inhibition − beginning (50%SIbeg). 
For the 93 ms duration (Panel b), again results for the FullM and the HalfM conditions were similar to each other and similar to those obtained for the 113 ms duration. A first phase of gain decrease took place 40 ms (FullM) or 59 ms (HalfM) after mask appearance; a return to gain values around 1.0 was observed after the saccade inhibition and finally a second phase of gain decrease occurred. 
Panel c shows that, for the 73 ms target duration when the first saccades were initiated, the mask had already begun to influence the saccade gain. Hence, these initial saccades were already in the first decrease phase observed for longer durations, making the measurement of FEA timings impossible. Nevertheless, the two DA times (for FullM and HalfM) were easily measurable: these measurements were again very similar to the 50%SIbeg values. After the saccadic inhibition, gain was back to value of ∼1.0 under the FullM condition and even more than 1.0 (hypermetria) for the HalfM condition (contribution of global effect, see Discussion). Finally the second decrease phase brought saccade gain to the low values of 0.5 (FullM) and 0.7 (HalfM). 
For the 53 ms (Panel d) and the 33 ms (Panel e) durations, patterns of gain evolution were quite similar: the first initiated saccades seemed to form the bottom of the dip in the amplitude that we observed for longer durations. The DA value was measurable only on the smooth for 53 ms in HalfM condition. The return of gain to value around 1.0 was present at the end of saccadic inhibition. At this time, gain was slightly hypometric in every case except for the HalfM − 33 ms for which a slight hypermetria was recorded. Afterward, a phase of gain decrease again brought gain to low values. 
Finally, for the shortest duration tested (13 ms, Panel f), the saccadic inhibition occurred too early to allow saccade initiation before its beginning. At the end of saccadic inhibition, gain values were very different for the two conditions: a marked hypermetria for HalfM and a marked hypometria for FullM. From these values, in both cases, a final gain decrease phase was observed. 
Figure 6b shows the strong relationship between the target duration, and hence the time of mask appearance, and the time of the dip in amplitude smoothes (r2 = 0.93, p < 0.001). Panel c relates the effect of the mask on latency distributions to its effect on gain evolution: as already mentioned, there was a strong correlation between the time of 50% of saccadic inhibition − beginning (50%SIbeg, Figure 2) and the time of the dip in amplitude (r2 = 0.96, p < 0.001). 
Can the mask itself become a target?
Concerning the second phase of gain decrease, a question may arise: could this gain decrease be due to the fact that, because already being present for a certain duration, the mask itself became a target for saccades? If so, two observations should have been made. First, there should have been a difference during this second phase between FullM and HalfM because position of the mask was different in these two conditions. This was not the case, apart for the 13 ms target duration and for a short period of time after saccadic inhibition for the 73 ms target duration during which an hypermetria was observed for the HalfM condition. Second, given the size of the mask, if it became a target, an important increase in gain variability should have been observed during this second phase (Ploner et al., 2004). To assess this second point, the variability of responses was measured during four time periods (Panel b of Figure 4). There was approximately 40 ms between the first effect of the mask and the dip in amplitude smoothes (DA), which was also roughly similar to the 50% saccadic inhibition − beginning (50%SIbeg); this period of 40 ms formed our period 2 (P2) corresponding to the period of the first phase of gain decrease. Period 1 (P1) consisted of the 40 ms before the P2 that corresponded to a period before gain decrease. The two remaining periods were from 50%SIend to 50%SIend + 40 ms and from 50%SIend + 40 ms to 50%SIend + 80 ms. These last two periods hence characterized the phases during which saccades recovered accuracy and entered into a second phase of gain decrease, respectively. The right column of Figure 5 plots the variability (mean of standard error, see Materials and methods) of responses for these four periods of 40 ms. For each target duration, a two-way ANOVA with Phase (P1 to P4) and Masking Condition (NoM, FullM, and HalfM) as factors was used to characterize response variability. For each duration, there was a significant effect of each factor and of the interaction between them (p < 0.05). Nevertheless, importantly in the present context, in several cases post-hoc tests (Newman-Keuls) showed that variability was not significantly different between phases 2, 3, and 4. Arrows on Figure 5 (left column) show these cases. This means that the second phase of gain decrease was not inevitably associated with a significant increase in the variability of the responses. In sum, present evidences do not favor the possibility that the mask itself became a target during the second phase of gain decrease. 
Gain evolution as a function of initiation time for each participant
The complex pattern of gain evolution as a function of initiation time was also found when data were analyzed separately for each participant. Using the same method of alignment as the one used at the end of the latency analysis, we obtained for each participant a plot for the evolution of gain as a function of the initiation time with respect to mask onset (see one example on Figure 7 and graphs for all participants on Panels b–u of Supplementary Data, Figure S1). The complex pattern of gain evolution with two decrease phases separated by the saccadic inhibition was found for 15 out of 20 participants. These 15 participants were those for whom a mean of 49.6% of saccades were produced before 200 ms after target onset. Thus, for these participants, enough early initiated saccades (before 200 ms) were present to observe the first phase of gain decrease. The remaining 5 participants were those presenting only a mean of 10.8% of early initiated saccades and, hence, presented only the second phase. Note finally that to construct this plot separately for each participant, data from the FullM and the HalfM conditions have been merged. This explains why a tendency to hypermetria was sometimes observed after saccadic inhibition (e.g., Figure 7). 
Figure 7
 
Saccade density and saccade gain as a function of initiation time for a participant. For a single participant, lowess regression on data for all target durations and for the two masking conditions (FullM + HalfM) that have been aligned with respect to mask onset. Similarly as for the population analysis, gain data showed a complex pattern of modulation with a first gain decrease phase, a return to value around 1.0 (the trend to hypermetria was due to data from the HalfM condition) and, finally, a second decrease phase. P1 to P4 correspond to the four 40-ms periods considered to study peak velocity difference between FullM and NoM conditions (Figure 10). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Figure 7
 
Saccade density and saccade gain as a function of initiation time for a participant. For a single participant, lowess regression on data for all target durations and for the two masking conditions (FullM + HalfM) that have been aligned with respect to mask onset. Similarly as for the population analysis, gain data showed a complex pattern of modulation with a first gain decrease phase, a return to value around 1.0 (the trend to hypermetria was due to data from the HalfM condition) and, finally, a second decrease phase. P1 to P4 correspond to the four 40-ms periods considered to study peak velocity difference between FullM and NoM conditions (Figure 10). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Mean gain values as a function of target duration
Figure 8 shows amplitude results with the same analysis as the one performed by Aitsebaomo and Bedell (1992): the mean saccadic gain is represented as a function of the target duration for the three conditions (NoM, FullM, and HalfM) with data from all participants pooled together. The curves obtained for the NoM and the FullM conditions are very similar to figure 1 and 2 of Aitsebaomo and Bedell. Nevertheless, among the two target durations (33 ms and 50 ms), they tested in a condition similar to HalfM (their figure 3); the result of the present study for 33 ms diverge from their own result (see Discussion). Interestingly, we can note that when the mean is considered (as opposed to an analysis as a function of initiation time), we don't observe a statistically significant change of amplitude for the 93 and 113 ms stimulus durations, despite the phases of significant gain change (Figure 5). This could explain why absence of amplitude change has been reported by some previous studies on saccadic inhibition (Reingold & Stampe, 2002, 2004; see Discussion). Finally, the divergence between the FullM and the HalfM curves observed for the shortest target durations may indicate an involvement of the global effect in these cases (see Discussion). 
Figure 8
 
Mean saccade gain as a function of target duration. Error bars symbolize standard error of the mean. A hypometria dependent on target duration was obtained in the FullM condition. In the HalfM, a trend to hypermetria was found for shortest target durations, whereas a hypometria was observed for longest ones. The table above the panel shows the results of post-hoc breakdown (Newman-Keuls tests: * p < 0.05 and ns p > 0.05) after ANOVA (see text for design).
Figure 8
 
Mean saccade gain as a function of target duration. Error bars symbolize standard error of the mean. A hypometria dependent on target duration was obtained in the FullM condition. In the HalfM, a trend to hypermetria was found for shortest target durations, whereas a hypometria was observed for longest ones. The table above the panel shows the results of post-hoc breakdown (Newman-Keuls tests: * p < 0.05 and ns p > 0.05) after ANOVA (see text for design).
Saccade kinematics during the different phases of masking effect
In their distractor study, Edelman and Xu (2009) provided evidence of saccadic interruption when the distractor was presented: hypometric saccades showed similar initial rise time of the velocity trace than saccades produced without distractor, but they apparently diverged as the distractor's influence manifested (see their figure 6). Considering the four phases defined for Figure 5, we found similar results for the two phases for which there was an hypometria (P2 and P4). For these two phases, Figure 9 shows examples of mean velocity traces. On this figure, we also added mean velocity traces of a control saccade of similar amplitude as hypometric ones: these control saccades have lower peak velocity and longer duration. These observations were quantified by comparing main sequence relationships between the two conditions for the four phases (P1 to P4). Nevertheless, to take into account the interparticipant variability in saccade kinematics, these four periods here were determined separately for each participant with data aligned on mask onset (Figure 7). Panels a–d of Figure 10 show main sequence plots (peak velocity as a function of saccade amplitude) with NoM and FullM conditions superimposed for each period. Mean peak velocity was computed across all participants for bins of 2°, from 0–2° to 20–22° of saccadic amplitude. Note that HalfM saccades were not considered because initial eye position influences peak velocity of saccades: centripetal saccades have higher peak velocity than centrifugal ones (Pélisson & Prablanc, 1988). These plots show that FullM saccades tended to have higher peak velocity than NoM saccades mainly for the phases P2 and P4, i.e., during the first and the second phase of gain decrease. A statistical analysis was performed as follows: separately for each participant, peak velocity differences between NoM and FullM were computed for each amplitude bin in each period (4 saccades minimum per bin to consider the value). This resulted in 267 values of differences. Mean of all these differences per periods are plotted on Figure 10e. A one-way ANOVA revealed a significant effect of the factor period on peak velocity differences (F(3,263) = 6.31, p < 0.001). Differences for periods P1 and P3 were near zero and were not statistically different from each other but were different from the two others (P2 and P4) that were themselves not different from each other (Newman-Keuls tests). Thus, hypometric saccades in Period 2 and Period 4 similarly showed a higher peak velocity than normometric saccades of similar amplitude. 
Figure 9
 
Examples of mean velocity traces for saccades produced in the NoM and the FullM conditions. Mean velocity traces of saccades produced for a 14° target in Phase 2 (participant ASI) and Phase 4 (participant LFE) in the NoM condition (n = 5, thin line) and the FullM condition (n = 4, thick line). Hypometric saccades observed in FullM condition showed similar initial rise time of the velocity trace than accurate saccades in NoM condition. Mean velocity traces for control saccades (NoM condition, n = 5, dashed line) of similar amplitude than those of FullM condition show lower peak velocity and longer duration.
Figure 9
 
Examples of mean velocity traces for saccades produced in the NoM and the FullM conditions. Mean velocity traces of saccades produced for a 14° target in Phase 2 (participant ASI) and Phase 4 (participant LFE) in the NoM condition (n = 5, thin line) and the FullM condition (n = 4, thick line). Hypometric saccades observed in FullM condition showed similar initial rise time of the velocity trace than accurate saccades in NoM condition. Mean velocity traces for control saccades (NoM condition, n = 5, dashed line) of similar amplitude than those of FullM condition show lower peak velocity and longer duration.
Figure 10
 
Saccade kinematics during the different phases of mask effect. Panels a–d show for each of the four 40-ms periods of time considered (P1–P4, Figure 7) the superimposition of the main sequence plot for the NoM and FullM conditions. A trend to an increase in peak velocity could be seen for the periods P2 and P4. (e) Mean peak velocity difference between similar amplitude saccades in the NoM and the FullM conditions for the four periods (P1–P4). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Figure 10
 
Saccade kinematics during the different phases of mask effect. Panels a–d show for each of the four 40-ms periods of time considered (P1–P4, Figure 7) the superimposition of the main sequence plot for the NoM and FullM conditions. A trend to an increase in peak velocity could be seen for the periods P2 and P4. (e) Mean peak velocity difference between similar amplitude saccades in the NoM and the FullM conditions for the four periods (P1–P4). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Discussion
The appearance of a distractor shortly after the onset of a saccadic target induces a strong decrease in the probability of saccade initiation around 90 ms after this appearance, creating a dip in the latency distributions. This phenomenon is referred to as “Saccadic inhibition” (Reingold & Stampe, 2002). The purpose of the present experiment was to test whether amplitude modulations can accompany these changes in latency distributions. Visual backward masking was suspected to potentially induce a saccadic inhibition that would be accompanied by amplitude changes (see Introduction). We showed here that this protocol of visual backward masking indeed produced a strong saccadic inhibition that was accompanied by complex saccade amplitude changes. When the gain of saccades was plotted as a function of their initiation time, we observed a first phase of amplitude decrease beginning shortly after the mask onset that was interrupted by the saccadic inhibition; then, when initiated again, saccades were accurate before entering into a second phase of amplitude decrease. These results may be accounted for by a fast and a slow process of amplitude modulation occurring before and after the decrease of the initiation probability (saccadic inhibition). Depending on the target duration, this complex amplitude modulation was observed either in totality or only partially, its initial phases being cropped by the earlier mask influence. 
The effects of the mask on saccade initiation
Whatever the target duration, the appearance of the mask induced a strong transient decrease in the probability of saccadic initiation with a delay of approximately 90 ms (50%SIb). Reingold and Stampe (2002) observed a similar inhibitory phenomenon when a large (nonoverlapping) visual stimulus was flashed after a target onset, calling this phenomenon “saccadic inhibition.” In their study, the time at which 50% of the inhibition was reached occurred earlier (67.8 +/– 1.1 ms) than in the present study (90.0 ± 4.6 ms, Table 1). Considering the explanation of lateral inhibitory influences at the level of a visuomotor map (see Introduction), the difference could be the consequence of the fact that our mask also activated the region corresponding to the target, whereas this was not the case in the study of Reingold and Stampe. Indeed, in their case, the flashed stimulus was situated outside of the working area that contained the fixation point and the target of the saccade (top and bottom third of the display). Thus, in our study, as the site corresponding to the target was also activated by the mask, its inhibition could have taken a longer time to be achieved. 
A second difference is that our inhibition lasted longer (70.4 ± 6.2 ms vs. 36.8 ms in the Reingold and Stampe study). This result could be due to the very short presentation time of the new visual stimulus in their study (33 ms), while in ours the mask was still present after the saccade had been produced. 
Interestingly, it has also been shown that the increase in latency observed when a small distractor was presented at a nontarget location before the initiation of the saccades (the remote distractor effect, Walker et al., 1997) was due to a saccadic inhibition (Graupner et al., 2007; Buonocore & McIntosh, 2008; Edelman & Xu, 2009). Edelman and Xu (2009) showed that the strength of the saccadic inhibition due to a small distractor was related to the strength of the visuomotor activation induced by the task. Saccadic inhibition was larger when visuomotor activation was not too strong, such as in a task in which a memory-guided saccade had to be produced toward a blank space (i.e., with no visual target on the display, Edelman & Goldberg, 2001). Conversely, if subjects had to produce reflexive saccades toward visual targets, the saccadic inhibition was reduced. In the present study, we showed that the mask inhibited saccadic production even in the case in which there was a strong visuomotor activation, i.e., a task involving an unpredictable visual target with a gap paradigm. It is likely that there is a continuum between weak distractors and stronger ones, such as the one formed by the mask. 
The effects of the mask on saccade gain
For each target duration, we observed some phases of significant saccadic gain change, whereas in the study of Reingold and Stampe (2002) no amplitude changes were observed. What could explain this discrepancy? As explained in the Introduction, the relative strength ratio between the target and the distractor could be evoked (Edelman & Xu, 2009). Reingold and Stampe (2002) observed a larger amplitude change, although not statistically significant, when they tested antisaccades for which the visuomotor activation is lower (Everling et al., 1999), which also supports this point. But still, Reingold and Stampe used in their study a rather powerful distractor consisting of 2/3 of the screen flashing to white, though not overlapping the target. Hence, we believe that an explanation for their report of no amplitude change is the fact that saccadic gain was not studied as a function of initiation time as in the present study. Indeed, in the present study, when means of gain were computed (Figure 8) for longer target duration (93 ms and 113 ms), there was a small, but not statistically significant decrease of amplitude, a result identical to the one obtained by Reingold and Stampe (2002) (see their table 1). We believe that, in protocols inducing saccadic inhibition, analyzing the saccadic gain dynamically as a function of the saccade initiation time may uncover amplitude changes that otherwise would go unnoticed. 
The backward masking of the target most of the time induced a saccadic hypometria; nevertheless, a hypermetria could be observed for specific conditions and timings. 
Global effect with the center of gravity of the mask?
A first explanation of amplitude modulation when the mask appeared would be the presence of a global effect with the center of gravity of the mask. Indeed, when two targets are presented in spatial and temporal proximity, a saccade that lands in-between the two targets (an averaging saccade) could be elicited (Coren & Hoenig, 1972; Findlay, 1982; Ottes et al., 1984). In the present study, there would have been a global effect between the target and the center of gravity of the mask. If the global effect was the origin of gain modulation, we should have observed opposite effects in our two masking conditions at the same time: hypometria in FullM and hypermetria in HalfM (see Materials and methods). At the very least, differences between these two conditions should have been observed. Results obtained for target durations of 113 and 93 ms, for which there was no significant difference between FullM and HalfM (Figure 5), clearly show that the global effect cannot be the only origin of gain change2. Nevertheless, differences between FullM and HalfM, and specifically hypermetria in HalfM, were observed when the target duration was shorter: for the two shortest target durations (13 and 33 ms), at the end of the saccadic inhibition, hypometria (FullM) and hypermetria (HalfM) were observed. For the same period, there was also a hypermetria in the HalfM − 73 ms but without hypometria in the FullM. We suggest that, for target durations from 73 ms to the shortest ones, a global effect between the target and the center of gravity of the mask contributed to gain variation observed right after the saccadic inhibition. In total, the global effect could contribute in certain cases (after saccadic inhibition for the shortest target durations), but it cannot be the only explanation for all the effects on gain recorded in the present study. 
Activation of fixation neurons?
A second origin to amplitude variation could be the fact that the mask used in the present study contained target patterns at the location of the fixation point. It is known that active fixation processes are at work in the oculomotor system (Munoz et al., 2000; Munoz & Fecteau, 2002; Stuphorn & Schall, 2002). Activation of fixation neurons during the production of a saccade reduces the amplitude of the movement (Sparks & 5s, 1983; Paré & Guitton, 1994; Goldberg et al., 1986). In the present study, the appearance of the mask could induce an activation of this fixation system. Nevertheless, even if this mechanism could account for a part of the effect observed in the present study, we argue that there are at least two reasons why it cannot be the sole explanation. First, Aitsebaomo and Bedell (1992) used, in a condition designed to test the hypothesis of the global effect (see Materials and methods), a mask extending only in the peripheral visual field that was not impacting fixation zone (see their figure 3). In spite of this, they found exactly the same results than those observed with a mask equivalent to the one used in our FullM condition. Second, Edelman and Xu (2009) in their study of the influence of a small distractor also observed hypometric saccades. This hypometria was observed in spite of the fact that only a 2 × 2° distractor was presented at a location that was 10° remote from the fixation point and hence that did not cover the fixation area. Thus, for these two reasons, we believe that an activation of the fixation subsystem could not be the principal explanation for the effects recorded here3
Selection of a random portion of the mask as a target?
A third possibility would be that, given the very short target duration in some trials, a more or less random portion of the mask was selected as a target for the saccade. As already mentioned in the Results section, the absence of amplitude difference between FullM and HalfM conditions argues against this possibility. In addition, such a random selection process should also lead to a larger variability. We showed that this increase of variability, although present in some cases, was not the rule (right column of Figure 5). Finally, although not studied in detail, a complex pattern of amplitude change similar to the one reported in the present study was also apparent in figure 5 of Edelman and Xu (2009). This also argues against the possibility that the mask was taken as a target. Indeed, in the study of Edelman and Xu, this pattern with two decrease phases was obtained with a single punctate distractor. 
Saccade interruption?
A fourth possibility is that activity induced by the mask appearance could have prematurely terminated saccades triggered by the visual target. This explanation was proposed by Edelman and Xu (2009) to account for hypometric saccades they recorded in their distractor task. Saccades would be initiated based on target information, but the arrival of a new stimulus (a distractor or a mask) would interrupt the saccade execution. A consequence of such an interruption is that kinematics of saccade should change. For example, a target located at 14° would trigger a saccade with similar initial rise time of velocity trace than for a control 14° saccade. But the interruption would reduce the amplitude and hence would associate a smaller saccade with an abnormally high peak velocity (Figure 9). This would result in a change in the main sequence. Present results concerning saccade kinematics fully support this explanation of saccade interruption. Importantly, we found here a significant increase in peak velocity for similar amplitude saccades for both phases of gain decrease. This suggests a similar mechanism of gain modulation for these two phases. Obviously, this hypothesis of saccade interruption is valid only for hypometria. 
Hence, to sum up, amplitude modulation corresponding to hypometria would have been due to saccade truncation by the activity related to the mask. Additionally, a participation of the global effect could account for hypermetria when observed (i.e., after saccadic inhibition for shortest target durations). 
Potential neurophysiological substrates of the observed effects
Saccadic inhibition mechanism
Reingold and Stampe (2002, 2004) argued that “saccadic inhibition” could be the result of inhibitory processes that potentially take place at the level of the superior colliculus (SC), a fundamental structure for the initiation of saccade (Dorris et al., 1997; Munoz et al., 2000; Sparks et al., 2000; Krauzlis, 2003). They observed first effects on latency distributions around 60 ms, a value similar to the one found in the present study (65 ms). As discussed by Reingold and Stampe, such a value is consistent with the hypothesis of a collicular involvement. First, visual activities could be found in the superficial SC 35–47 ms after the onset of a target (Rizzolatti et al., 1980); then, the time to transfer this activity to the intermediate and deep layers of the SC is 5–10 ms (Lee et al., 1997). Munoz and Wurtz (1993) found a mean of 20 ms for a collicular activity to delay saccades coded by a remote collicular area. Finally, it has been shown in a monkey that evoking activity in any part of this visuomotor map entails an inhibition of remote areas of the map (Munoz & Istvan, 1998). Consequently, in both the Reingold and Stampe study (2002) and the present study, activity related to the target could have been inhibited through lateral inhibitory connections (Munoz & Istvan, 1998; Meredith & Ramoa, 1998; Dorris et al., 2007; Li et al., 2006) by activity related to the second nonoverlapping (Reingold & Stampe, 2002) or overlapping (the present study) stimulus. This competition would have produced the saccadic inhibition. In fact, the present study complements the scenario: the arrival of mask-related activity in the SC would first interrupt saccades (first effect around 50 ms, Table 1), then begin to delay saccades (from 65 ms) before strongly inhibiting saccadic initiation (115 ms). 
The complex pattern of amplitude variation
A remaining question concerns the origin of the complex pattern of gain modulation containing two decreasing phases surrounding return to gain values around 1.0. We observed this pattern either in totality (for 93 and 113 ms) or only partly for the remaining target durations because the mask onset was much earlier (see Figure 5). Interestingly, a similar complex pattern of gain modification, although not studied in detail, is also apparent in figure 5 of the Edelman and Xu (2009) study on distractor effect. What could be the origin of such a complex pattern? 
One possibility, admittedly speculative, would be to relate this finding to the organization of the saccadic system. Several pathways for the production of saccades have been isolated: a first, very direct, pathway connects the retina to the brainstem. This pathway brings information directly to the SC, where a motor command could be immediately prepared (Isa, 2002). It could be the substrate for the production of early initiated saccades (Schiller et al., 1980, 1987; Edelman & Keller, 1996; Sparks et al., 2000). A second pathway is a more indirect connection between the retina and the SC. Visual information travels from the retina to cortical visual areas where it is distributed to at least two cortical areas involved in the generation of saccades: the parietal and fontal eye fields (LIP and FEF; Wurtz et al., 2001; Büttner & Büttner-Ennever, 2006). These areas then project to the SC. 
The complex pattern of gain modulation observed in the present study could be the result of information flow—and its perturbation by the mask—into these two neural pathways of different length up to saccade production. The first initiated saccades observed in our study could be produced by the direct pathway. Rapid first perturbation due to the mask would occur (see previous—saccadic inhibition mechanism; Edelman & Xu, 2009; and the present data). This would correspond to the first gain decrease phase. This disruption would continue up to a point at which there is almost no saccade at all. This would correspond to the saccadic inhibition phenomenon. Meanwhile, activity related to the target would have travelled through the second cortical pathway. After saccadic inhibition, saccade gain returned to values around 1.0. This return to accuracy could be the result of signals arriving to the SC from the cortical areas (LIP and FEF), signals that have not yet been disrupted by the mask. Soon after that, the mask would also reach cortical areas and disrupt this processing stage: the second gain decrease phase would begin. Concerning this second gain decrease phase, contrary to the transient inhibition that has been seen in the SC after the presentation of a distractor (Dorris et al., 2007; Li et al., 2006), no such inhibition of the saccadic site in cortical areas was observed when a distractor was presented (McPeek, 2006, for FEF; Bisley & Goldberg, 2006, for LIP; but see Powell & Goldberg, 2000). Nevertheless, it is possible that the activity induced by the distractor or the mask in cortical areas inhibited the whole SC through corticotectal connections, thereby interrupting saccade in a similar manner as during the first phase of gain decrease. This last suggestion is supported by results concerning saccade kinematics showing a similar increase in peak velocity in both phases of gain decrease. 
This suggestion concerning the origin of the observed complex pattern of amplitude variation remains speculative and needs electrophysiological studies to be ascertained. But provided that it is correct, saccadic inhibition and accompanying amplitude changes could be a model to behaviorally study the dynamics of both oculomotor pathways in humans. 
Conclusion
The aim of the present study was to search for potential amplitude modifications that could go along with saccadic inhibition in certain cases. Visual backward masking induced saccadic inhibition that was accompanied by large and complex amplitude modulations. These modulations comprised two phases of gain decrease surrounding a phase during which the gain temporarily returned to values close to 1.0. We suggest that these amplitude modulations would be due to an interruption occurring through a fast and a slow mechanism. These results show that the onset of a large stimulus shortly after the first one induces the previously described saccadic inhibition and also amplitude modulations that follow a complex pattern. 
Supplementary Materials
Acknowledgments
I thank J. Blouin, G. Malhotra, H. Mokh, and D Pélisson for helpful comments on the manuscript. I also thank anonymous reviewers for insightful comments. This work was supported by the ANR grant BLAN07-1_185594 (MAPS). 
Commercial relationships: none. 
Corresponding author: Alain Guillaume. 
Email: alain.guillaume@univ-amu.fr. 
Address: Laboratoire de Neurosciences Cognitives, Aix-Marseille Université, Marseille, France. 
References
Aitsebaomo A. P. Bedell H. E . (1992). Psychophysical and saccadic information about direction for briefly presented visual targets. Vision Research, 32,1729–1737. [CrossRef] [PubMed]
Ansorge U. Francis G. Herzog M. H. Ogmen H . (2007). Visual masking and the dynamics of human perception, cognition, and consciousness A century of progress, a contemporary synthesis, and future directions. Advances in Cognitive Psychology, 3,1–8. [CrossRef]
Becker W . (1989). Metrics. In Goldberg M. E. Wurtz R. H . (Eds.), The neurobiology of saccadic eye movements. Elsevier Science Publishers.
Bisley J. W. Goldberg M. E . (2006). Neural correlates of attention and distractibility in the lateral intraparietal area. Journal of Neurophysiology, 95,1696–1717. [CrossRef] [PubMed]
Bompas A. Sumner P . (2011). Saccadic inhibition reveals the timing of automatic and voluntary signals in the human brain. Journal of Neuroscience, 31(35),12501–12512. [CrossRef] [PubMed]
Brainard D. H . (1997). The psychophysics toolbox. Spatial Vision, 10,433–436. [CrossRef] [PubMed]
Breitmeyer B. G. Ogmen H . (2000). Recent models and findings in visual backward masking: A comparison, review, and update. Perception & Psychophysics, 62,1572–1595. [CrossRef] [PubMed]
Buonocore A. McIntosh R. D . (2008). Saccadic inhibition underlies the remote distractor effect. Experimental Brain Research, 191,117–122. [CrossRef] [PubMed]
Büttner U. Büttner-Ennever J. A . (2006). Present concepts of oculomotor organization. Progress in Brain Research, 151,1–42. [PubMed]
Cleveland W. S . (1979). Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association, 74(368),829–836. [CrossRef]
Coren S. Hoenig P . (1972). Effect of non-target stimuli upon length of voluntary saccades. Perceptual & Motor Skills, 34,499–508. [CrossRef]
Dorris M. C. Pare M. Munoz D. P . (1997). Neuronal activity in monkey superior colliculus related to the initiation of saccadic eye movements. Journal of Neuroscience, 17,8566–8579. [PubMed]
Dorris M. C. Olivier E. Munoz D. P . (2007). Competitive integration of visual and preparatory signals in the superior colliculus during saccadic programming. Journal of Neuroscience, 27,5053–5062. [CrossRef] [PubMed]
Edelman J. A. Goldberg M. E . (2001). Dependence of saccade-related activity in the primate superior colliculus on visual target presence. Journal of Neurophysiology, 86(2),676–691. [PubMed]
Edelman J. A. Keller E. L . (1996). Activity of visuomotor burst neurons in the superior colliculus accompanying express saccades. Journal of Neurophysiology, 76,908–926. [PubMed]
Edelman J. A. Keller E. L . (1998). Dependence on target configuration of express saccade-related activity in the primate superior colliculus. Journal of Neurophysiology, 80,1407–1426. [PubMed]
Edelman J. A. Xu K. Z . (2009). Inhibition of voluntary saccadic eye movement commands by abrupt visual onsets. Journal of Neurophysiology, 101,1222–1234. [CrossRef] [PubMed]
Eggert T. Sailer U. Ditterich J. Straube A . (2002). Differential effect of a distractor on primary saccades and perceptual localization. Vision Research, 42,2969–2984. [CrossRef] [PubMed]
Everling S. Dorris M. C. Klein R. M. Munoz D. P . (1999). Role of primate superior colliculus in preparation and execution of anti-saccades and pro-saccades. Journal of Neuroscience, 19(7),2740–2754. [PubMed]
Findlay J. M . (1982). Global visual processing for saccadic eye movements. Vision Research, 22,1033–1045. [CrossRef] [PubMed]
Fischer B. Weber H . (1993). Express saccades and visual attention. Behavioral and Brain Sciences, 16,553–610. [CrossRef]
Goldberg M. E. Bushnell M. C. Bruce C. J . (1986). The effect of attentive fixation on eye movements evoked by electrical stimulation of the frontal eye fields. Experimental Brain Research, 61,579–584. [CrossRef] [PubMed]
Graupner S. T. Velichkovsky B. M. Pannasch S. Marx J . (2007). Surprise, surprise: Two distinct components in the visually evoked distractor effect. Psychophysiology, 44,251–261. [CrossRef] [PubMed]
Guillaume A . (2009). Saccadic eye movements towards visual targets followed by a visual mask. Progress in Motor Control VII abst. A2-26: 7 23–25, Marseille, France.
Hafed Z. M. Goffart L. Krauzlis R. J . (2009). A neural mechanism for microsaccade generation in the primate superior colliculus. Science, 323,940–943. [CrossRef] [PubMed]
Isa T . (2002). Intrinsic processing in the mammalian superior colliculus. Current Opinion in Neurobiology, 12,668–677. [CrossRef] [PubMed]
Krauzlis R. J . (2003). Neuronal activity in the rostral superior colliculus related to the initiation of pursuit and saccadic eye movements. Journal of Neuroscience, 23,4333–4344. [PubMed]
Lee P. H. Helms M. C. 8ine G. J. Hall W. C . (1997). Role of intrinsic synaptic circuitry in collicular sensorimotor integration. Proceedings of the National Academy of Sciences of the United States of America, 94,13299–13304. [CrossRef] [PubMed]
Li X. Kim B. Basso M. A . (2006). Transient pauses in delay-period activity of superior colliculus neurons. Journal of Neurophysiology, 95,2252–2264. [CrossRef] [PubMed]
Macknik S. L . (2006). Visual masking approaches to visual awareness. Progress in Brain Research, 155,177–215. [PubMed]
Meredith M. A. Ramoa A. S . (1998). Intrinsic circuitry of the superior colliculus: Pharmacophysiological identification of horizontally oriented inhibitory interneurons. Journal of Neurophysiology, 79(3),1597–1602. [PubMed]
McPeek R. M . (2006). Incomplete suppression of distractor-related activity in the frontal eye field results in curved saccades. Journal of Neurophysiology, 96,2699–2711. [CrossRef] [PubMed]
Munoz D. P. Wurtz R. H . (1993). Fixation cells in monkey superior colliculus. II. Reversible activation and deactivation. Journal of Neurophysiology, 70,576–589. [PubMed]
Munoz D. P. Istvan P. J . (1998). Lateral inhibitory interactions in the intermediate layers of the monkey superior colliculus. Journal of Neurophysiology, 79,1193–1209. [PubMed]
Munoz D. P. Dorris M. C. Pare M. Everling S . (2000). On your mark, get set: Brainstem circuitry underlying saccadic initiation. Canadian Journal of Physiology and Pharmacology, 78,934–944. [CrossRef] [PubMed]
Munoz D. P. Fecteau J. H . (2002). Vying for dominance: Dynamic interactions control visual fixation and saccadic initiation in the superior colliculus. Progress in Brain Research, 140,3–19. [PubMed]
Ottes F. P. Van Gisbergen J. A. Eggermont J. J . (1984). Metrics of saccade responses to visual double stimuli: Two different modes. Vision Research, 24,1169–1179. [CrossRef] [PubMed]
Pare M. Guitton D . (1994). The fixation area of the cat superior colliculus: Effects of electrical stimulation and direct connection with brainstem omnipause neurons. Experimental Brain Research, 101,109–122. [PubMed]
Pare M. Munoz D. P . (1996). Saccadic reaction time in the monkey: Advanced preparation of oculomotor programs is primarily responsible for express saccade occurrence. Journal of Neurophysiology, 76,3666–3681. [PubMed]
Pelli D. G . (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10,437–442. [CrossRef] [PubMed]
Pélisson D. Prablanc C . (1988). Kinematics of centrifugal and centripetal saccadic eye movements in man. Vision Research, 28,87–94. [CrossRef] [PubMed]
Ploner C. J. Ostendorf F. Dick S . (2004). Target size modulates saccadic eye movements in humans. Behavioral Neuroscience, 118,237–242. [CrossRef] [PubMed]
Powell K. D. Goldberg M. E . (2000). Response of neurons in the lateral intraparietal area to a distractor flashed during the delay period of a memory-guided saccade. Journal of Neurophysiology, 84,301–310. [PubMed]
Reingold E. M. Stampe D. M . (2002). Saccadic inhibition in voluntary and reflexive saccades. Journal of Cognitive Neuroscience, 14,371–388. [CrossRef] [PubMed]
Reingold E. M. Stampe D. M . (2004). Saccadic inhibition in reading. The Journal of Experimental Psychology: Human Perception and Performance, 30,194–211. [CrossRef]
Rizzolatti G. Buchtel H. A. Camarda R. Scandolara C . (1980). Neurons with complex visual properties in the superior colliculus of the macaque monkey. Experimental Brain Research, 38,37–42. [CrossRef] [PubMed]
Schiller P. H. True S. D. Conway J. L . (1980). Deficits in eye movements following frontal eye-field and superior colliculus ablations. Journal of Neurophysiology, 44,1175–1189. [PubMed]
Schiller P. H. Sandell J. H. Maunsell J. H . (1987). The effect of frontal eye field and superior colliculus lesions on saccadic latencies in the rhesus monkey. Journal of Neurophysiology, 57,1033–1049. [PubMed]
Sparks D. Rohrer W. H. Zhang Y . (2000). The role of the superior colliculus in saccade initiation: A study of express saccades and the gap effect. Vision Research, 40,2763–2777. [CrossRef] [PubMed]
Sparks D. L. 5s L. E . (1983). Spatial localization of saccade targets. I. Compensation for stimulation-induced perturbations in eye position. Journal of Neurophysiology, 49,45–63. [PubMed]
Stuphorn V. Schall J. D . (2002). Neuronal control and monitoring of initiation of movements. Muscle Nerve, 26,326–339. [CrossRef] [PubMed]
Sumner P . (2011). Determinants of saccade latency. In Liversedge S. P. Gilchrist I. D. Everling S . (Eds.), The Oxford handbook of eye movements (pp.413–424). Oxford: Oxford University Press.
Van Zandt T . (2000). How to fit a response time distribution. Psychonomic Bulletin & Review, 7,424–465. [CrossRef] [PubMed]
Walker R. Deubel H. Schneider W. X. Findlay J. M . (1997). Effect of remote distractors on saccade programming: Evidence for an extended fixation zone. Journal of Neurophysiology, 78,1108–1119. [PubMed]
Wurtz R. H. Sommer M. A. Pare M. Ferraina S . (2001). Signal transformations from cerebral cortex to superior colliculus for the generation of saccades. Vision Research, 41,3399–3412. [CrossRef] [PubMed]
Footnotes
1  The “loess” function of the R software was used to compute these fits. In R, the “loess” function has more features than the “lowess” one. The “loess” function gives the exact equivalent of the “lowess” one if the parameters “family,” “iterations,” and “surface” are fixed to “symmetric,” “4,” and “direct,” respectively. The order of the polynomial that was locally fitted was fixed to 1; the span (degree of smoothing) was fixed to 0.25; and 4 iterations was used for the robust fitting. The “loess” function of the R software also allows us to compute the standard error of these fits.
Footnotes
2  The hypothesis of a global effect would additionally suggest a similar peak-velocity/amplitude relationship for control and hypometric saccades (Edelman & Keller, 1998). This was not the case as shown by the kinematics analyses (Figures 8 and 9).
Footnotes
3  The fact that Hafed et al. (2009) emphasized the role of the foveal part of the SC in the control of microsaccades implies that any rostral activity would have to be integrated in the computation of the center of gravity of the mask. This scenario would hence refer to the previous paragraph entitled “Global effect with the center of gravity of the mask?”
Figure 1
 
Visual stimulus and task. Time course of the different trial types. Participants were required to fixate on a green cross for a random duration. After the fixation point offset, there was a gap of 250 ms, and then the target (a red circle with a dark circle inside) was presented for a short period of time (13, 33, 53, 73, 93, or 113 ms). After this duration, the target was either switched off (NoM condition) or covered by a pattern mask consisting of three rows of 79 target patterns each (simplified on this figure; FullM condition). Trials with a visual mask could also be of the HalfM type: the fixation point was displaced either to the left or to the right, and the mask was only covering half of the visual field. It consisted of three rows of 68 target patterns each. In the HalfM condition, the center of gravity of the mask was always located further away from the target, which was not the case for the FullM condition. This allowed testing for a potential global effect between the target and the center of gravity of the mask. See text for further details.
Figure 1
 
Visual stimulus and task. Time course of the different trial types. Participants were required to fixate on a green cross for a random duration. After the fixation point offset, there was a gap of 250 ms, and then the target (a red circle with a dark circle inside) was presented for a short period of time (13, 33, 53, 73, 93, or 113 ms). After this duration, the target was either switched off (NoM condition) or covered by a pattern mask consisting of three rows of 79 target patterns each (simplified on this figure; FullM condition). Trials with a visual mask could also be of the HalfM type: the fixation point was displaced either to the left or to the right, and the mask was only covering half of the visual field. It consisted of three rows of 68 target patterns each. In the HalfM condition, the center of gravity of the mask was always located further away from the target, which was not the case for the FullM condition. This allowed testing for a potential global effect between the target and the center of gravity of the mask. See text for further details.
Figure 2
 
Measures taken to characterize the saccadic inhibition. The mask appearance always induced a strong transient decrease in probability of saccade initiation. The measures defined to characterize this inhibition are indicated on the probability density function computed from values of latency of all participants in the FullM – 113 ms condition (in grey, probability density function of NoM control values for the same target duration). FEL: first effects on latency = the onset time of the first decrease in saccade density. MSI: time of the maximum saccadic inhibition. 50%SIbeg: time corresponding to 50% of the saccadic inhibition − beginning of saccadic inhibition. 50%SIend: time corresponding to 50% of the saccadic inhibition − end of saccadic inhibition. SID: duration of the saccadic inhibition = 50%SIend − 50%SIbeg.
Figure 2
 
Measures taken to characterize the saccadic inhibition. The mask appearance always induced a strong transient decrease in probability of saccade initiation. The measures defined to characterize this inhibition are indicated on the probability density function computed from values of latency of all participants in the FullM – 113 ms condition (in grey, probability density function of NoM control values for the same target duration). FEL: first effects on latency = the onset time of the first decrease in saccade density. MSI: time of the maximum saccadic inhibition. 50%SIbeg: time corresponding to 50% of the saccadic inhibition − beginning of saccadic inhibition. 50%SIend: time corresponding to 50% of the saccadic inhibition − end of saccadic inhibition. SID: duration of the saccadic inhibition = 50%SIend − 50%SIbeg.
Figure 3
 
Inhibition of saccadic initiation produced by the mask appearance. (a–f) For each target duration, superimposition of the three probability density functions for NoM, FullM, and HalfM conditions. In each case, a strong transient decrease of saccade initiation was observed for FullM and HalfM conditions around 100 ms after mask appearance. See text for further details.
Figure 3
 
Inhibition of saccadic initiation produced by the mask appearance. (a–f) For each target duration, superimposition of the three probability density functions for NoM, FullM, and HalfM conditions. In each case, a strong transient decrease of saccade initiation was observed for FullM and HalfM conditions around 100 ms after mask appearance. See text for further details.
Figure 4
 
Saccade gain as a function of initiation time: Examples. (a) NoM condition, target duration = 113 ms. (b) FullM condition, target duration = 113 ms. In both cases, lowess regression applied to the data set (bold trace) and 95% confidence interval of the lowess regression (grey area) have been added. In addition, probability density functions corresponding to NoM − 113 ms and FullM – 113 ms have been added below. On the probability density function of FullM, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). FEA: first effects on amplitude = the time of the onset of the first gain decrease phase. DA: dip in amplitude = the time of the transition between the first gain decrease phase and the following increase phase. P1 to P4 correspond to four 40-ms periods used to assess gain variability. See text for further details.
Figure 4
 
Saccade gain as a function of initiation time: Examples. (a) NoM condition, target duration = 113 ms. (b) FullM condition, target duration = 113 ms. In both cases, lowess regression applied to the data set (bold trace) and 95% confidence interval of the lowess regression (grey area) have been added. In addition, probability density functions corresponding to NoM − 113 ms and FullM – 113 ms have been added below. On the probability density function of FullM, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). FEA: first effects on amplitude = the time of the onset of the first gain decrease phase. DA: dip in amplitude = the time of the transition between the first gain decrease phase and the following increase phase. P1 to P4 correspond to four 40-ms periods used to assess gain variability. See text for further details.
Figure 5
 
Saccade gain as a function of initiation time. Right column contains for each target duration (Panels a–f), superimposition of the three lowess regressions for NoM, FullM, and HalfM. For the NoM condition, the 95% confidence interval (grey area) has been added. For longest target durations (Panel a and b: 113 and 93 ms) the mask presence induced a complex pattern of gain change that contained two decrease phases surrounding a return to gain values around 1.0. As target duration was decreased (from 73 ms to 13 ms, Panels c to f), first phases of this complex pattern of gain modulation was more and more cropped. On each plot, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). Tables above the x-axis of each plot contain results of post-hoc breakdown (Holm-Sidak tests) after ANOVAs (see text for design), grey box = (p < 0.05) and empty box (p > 0.05). Left column contains variability of gain values assessed through computation of mean of standard error for the four defined periods (Figure 4). Vertical grey bars symbolize saccadic inhibition. Arrows show cases for which post-hoc tests after ANOVAs revealed no significant difference between phase 2, 3, and 4. See text for further details.
Figure 5
 
Saccade gain as a function of initiation time. Right column contains for each target duration (Panels a–f), superimposition of the three lowess regressions for NoM, FullM, and HalfM. For the NoM condition, the 95% confidence interval (grey area) has been added. For longest target durations (Panel a and b: 113 and 93 ms) the mask presence induced a complex pattern of gain change that contained two decrease phases surrounding a return to gain values around 1.0. As target duration was decreased (from 73 ms to 13 ms, Panels c to f), first phases of this complex pattern of gain modulation was more and more cropped. On each plot, the grey bar shows the duration of the saccadic inhibition (time between 50%SIbeg and 50%SIend). Tables above the x-axis of each plot contain results of post-hoc breakdown (Holm-Sidak tests) after ANOVAs (see text for design), grey box = (p < 0.05) and empty box (p > 0.05). Left column contains variability of gain values assessed through computation of mean of standard error for the four defined periods (Figure 4). Vertical grey bars symbolize saccadic inhibition. Arrows show cases for which post-hoc tests after ANOVAs revealed no significant difference between phase 2, 3, and 4. See text for further details.
Figure 6
 
Correlation analyses. The time of maximum saccadic inhibition (MSI, Panel a) and the time of the dip in amplitude (DA, Panel b) were highly correlated with the target duration that corresponds to the time of the mask onset. DA was also highly correlated with a parameter that characterizes latency distribution: the time of 50% of saccadic inhibition − beginning (Panel c).
Figure 6
 
Correlation analyses. The time of maximum saccadic inhibition (MSI, Panel a) and the time of the dip in amplitude (DA, Panel b) were highly correlated with the target duration that corresponds to the time of the mask onset. DA was also highly correlated with a parameter that characterizes latency distribution: the time of 50% of saccadic inhibition − beginning (Panel c).
Figure 7
 
Saccade density and saccade gain as a function of initiation time for a participant. For a single participant, lowess regression on data for all target durations and for the two masking conditions (FullM + HalfM) that have been aligned with respect to mask onset. Similarly as for the population analysis, gain data showed a complex pattern of modulation with a first gain decrease phase, a return to value around 1.0 (the trend to hypermetria was due to data from the HalfM condition) and, finally, a second decrease phase. P1 to P4 correspond to the four 40-ms periods considered to study peak velocity difference between FullM and NoM conditions (Figure 10). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Figure 7
 
Saccade density and saccade gain as a function of initiation time for a participant. For a single participant, lowess regression on data for all target durations and for the two masking conditions (FullM + HalfM) that have been aligned with respect to mask onset. Similarly as for the population analysis, gain data showed a complex pattern of modulation with a first gain decrease phase, a return to value around 1.0 (the trend to hypermetria was due to data from the HalfM condition) and, finally, a second decrease phase. P1 to P4 correspond to the four 40-ms periods considered to study peak velocity difference between FullM and NoM conditions (Figure 10). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Figure 8
 
Mean saccade gain as a function of target duration. Error bars symbolize standard error of the mean. A hypometria dependent on target duration was obtained in the FullM condition. In the HalfM, a trend to hypermetria was found for shortest target durations, whereas a hypometria was observed for longest ones. The table above the panel shows the results of post-hoc breakdown (Newman-Keuls tests: * p < 0.05 and ns p > 0.05) after ANOVA (see text for design).
Figure 8
 
Mean saccade gain as a function of target duration. Error bars symbolize standard error of the mean. A hypometria dependent on target duration was obtained in the FullM condition. In the HalfM, a trend to hypermetria was found for shortest target durations, whereas a hypometria was observed for longest ones. The table above the panel shows the results of post-hoc breakdown (Newman-Keuls tests: * p < 0.05 and ns p > 0.05) after ANOVA (see text for design).
Figure 9
 
Examples of mean velocity traces for saccades produced in the NoM and the FullM conditions. Mean velocity traces of saccades produced for a 14° target in Phase 2 (participant ASI) and Phase 4 (participant LFE) in the NoM condition (n = 5, thin line) and the FullM condition (n = 4, thick line). Hypometric saccades observed in FullM condition showed similar initial rise time of the velocity trace than accurate saccades in NoM condition. Mean velocity traces for control saccades (NoM condition, n = 5, dashed line) of similar amplitude than those of FullM condition show lower peak velocity and longer duration.
Figure 9
 
Examples of mean velocity traces for saccades produced in the NoM and the FullM conditions. Mean velocity traces of saccades produced for a 14° target in Phase 2 (participant ASI) and Phase 4 (participant LFE) in the NoM condition (n = 5, thin line) and the FullM condition (n = 4, thick line). Hypometric saccades observed in FullM condition showed similar initial rise time of the velocity trace than accurate saccades in NoM condition. Mean velocity traces for control saccades (NoM condition, n = 5, dashed line) of similar amplitude than those of FullM condition show lower peak velocity and longer duration.
Figure 10
 
Saccade kinematics during the different phases of mask effect. Panels a–d show for each of the four 40-ms periods of time considered (P1–P4, Figure 7) the superimposition of the main sequence plot for the NoM and FullM conditions. A trend to an increase in peak velocity could be seen for the periods P2 and P4. (e) Mean peak velocity difference between similar amplitude saccades in the NoM and the FullM conditions for the four periods (P1–P4). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Figure 10
 
Saccade kinematics during the different phases of mask effect. Panels a–d show for each of the four 40-ms periods of time considered (P1–P4, Figure 7) the superimposition of the main sequence plot for the NoM and FullM conditions. A trend to an increase in peak velocity could be seen for the periods P2 and P4. (e) Mean peak velocity difference between similar amplitude saccades in the NoM and the FullM conditions for the four periods (P1–P4). Vertical grey bar symbolizes saccadic inhibition. See text for further details.
Table 1
 
Measured parameters characterizing latency distributions and gain evolution as a function of initiation time for FullM and HalfM conditions.
Table 1
 
Measured parameters characterizing latency distributions and gain evolution as a function of initiation time for FullM and HalfM conditions.
Time [with respect to mask onset] of… (in ms) Sacc inhib duration SID
… first effects on amplitude curves FEA … first effects on latency PDF FEL … the dip in amplitude DA … 50% sacc inhib (beginning) 50%SIb … max sacc inhib (dip in latency) MSI … 50% sacc inhib (end) 50%SIe
FullM – 113 56 63 86 87 114 155 68
HalfM – 113 42 49 77 83 132 158 75
FullM – 93 40 63 85 88 110 159 71
HalfM – 93 59 64 89 87 115 157 70
FullM – 73 66 88 91 109 157 66
HalfM – 73 60 88 94 123 169 75
FullM – 53 78 91 108 166 75
HalfM – 53 68 71 90 117 167 77
FullM – 33
HalfM – 33 81 99 116 156 57
FullM – 13
HalfM – 13
Mean values (± SD) 49.3 (± 9.6) 65.7 (± 9.6) 83.4 (± 6.8) 90.0 (± 4.6) 115.3 (± 7.5) 161.4 (± 5.9) 70.4 (± 6.2)
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×