Selecting a stimulus as the target for a goal-directed movement involves inhibiting other competing possible responses. Both target and distractor stimuli activate populations of neurons in topographic oculomotor maps such as the superior colliculus. Local inhibitory interconnections between these populations ensure only one saccade target is selected. Suppressing saccades to distractors may additionally involve inhibiting corresponding map regions to bias the local competition. Behavioral evidence of these inhibitory processes comes from the effects of distractors on oculomotor and manual trajectories. Individual saccades may initially deviate either toward or away from a distractor, but the source of this variability has not been investigated. Here we investigate the relation between distractor-related deviation of trajectory and saccade latency. Targets were presented with, or without, distractors, and the deviation of saccade trajectories arising from the presence of distractors was measured. A fixation gap paradigm was used to manipulate latency independently of the influence of competing distractors. Shorter-latency saccades deviated toward distractors and longer-latency saccades deviated away from distractors. The transition between deviation toward or away from distractors occurred at a saccade latency of around 200 ms. This shows that the time course of the inhibitory process involved in distractor related suppression is relatively slow.
Inhibition of inappropriate behavior is a major function of the cerebral cortex. The visual world contains multiple stimuli affording action. At any one time, we typically direct our behavior by selecting just one stimulus as a goal. Neural activity corresponding to other, competing stimuli, and the responses to them, may be inhibited (Allport 1993; Desimone and Duncan 1995; McPeek et al. 2003; Munoz and Istvan 1998). Recent oculomotor and reaching studies have provided direct evidence for inhibition, by studying how goal-directed movement trajectories are modulated when movement to an alternative, competing stimulus must be inhibited (Godijn and Theeuwes 2002, 2004; Ludwig and Gilchrist 2003; McPeek et al. 2003; McSorley et al. 2004; Theeuwes and Godijn 2004; Van der Stigchel et al. 2006; Walker et al. 2006). Thus saccade trajectories have been shown to deviate away from competing distractors. In a study by Sheliga et al. (1994) subjects attended to stimuli presented laterally, whose form specified whether an upward or downward vertical saccade should be made. The vertical saccades were found to deviate away from the lateral stimulus location. On their interpretation, attention to the lateral stimulus elicited the motor program for a saccade to that location. Active inhibition of this program caused the saccade to the target to deviate away from the attended location. Several other studies have also reported initial trajectory deviations away from distractors presented along with the saccade target (Doyle and Walker 2001, 2002; Findlay and Harris 1984; Godijn and Theeuwes 2002; McPeek and Keller 2001; McSorley et al. 2004; Sheliga et al. 1995a,b, 1997; Theeuwes and Godijn 2004; Tipper et al. 2001; Van der Stigchel et al. 2006; van Gisbergen et al. 1987; Walker et al. 2006).
By contrast, other behavioral studies have reported initial trajectory deviations toward distractors, in apparently similar experimental conditions. For example, saccades during visual search consistently deviate toward distractors (McPeek and Keller 2001; McPeek et al. 2003) as do saccades made in other saccade paradigms in which the distractor location is not known before the onset of the saccade goal (Walker et al. 2006). Deviation toward distractors has been explained in terms of partial activation of the saccade program associated with the distractor, followed by an averaging process within the motor map encoding saccade metrics (McPeek and Keller 2001; McPeek et al. 2003; Port and Wurtz 2003; Tipper et al. 2001). Importantly, these explanations involve a purely feedforward process, without any top-down inhibition of visuomotor processing.
These distractor-induced effects on saccade trajectories have been attributed to competitive interactions operating in the underlying neural map that specifies potential saccade goals (perhaps in the superior colliculus; Aizawa and Wurtz 1998; McPeek and Keller 2001; McPeek et al. 2003; Port and Wurtz 2003; Quaia et al. 1998). First, an averaging process ensures that adjacent peaks of activation in the map are merged together. Then, the initial saccade direction is assumed to be specified by the location of peak activation in the map. When the distractor-related activity is above the surrounding baseline at the time of saccade initiation, it may merge with target-related activity, resulting in a deviation of initial saccade direction toward the distractor location. In other situations an additional, external inhibitory process may be applied to nontarget regions of the map. The projection from the frontal eye fields (FEFs) to the superior colliculus (Schlag-Rey et al. 1992; also see Tehovnik 2000) may perform this function. This top-down inhibition suppresses the distractor-related activity below baseline, so the averaging process now includes a negative contribution (Tipper et al. 2001), with the result that initial saccade direction deviates away from the distractor location. This inhibitory process has been shown to be spatially coarse (McSorley et al. 2004, 2005), which is consistent with the broad inhibition applied by the FEFs onto corresponding regions of the colliculuar motor map (Schlag-Rey et al. 1992). The observed curvature of trajectories back toward the saccade goal has been attributed to a separate process that could involve the cerebellum (McSorley et al. 2004; Quaia et al. 1999).
In many situations, both the top-down inhibition and the feedforward mechanisms may operate in parallel. Feedforward drive from visual representations of distractors and top-down inhibitory drive from cortical areas may converge on the motor layers of the superior colliculus. The initial direction of a saccade trajectory will thus reflect the combined influence of both top-down and feedforward processes at the moment of saccade initiation. Indeed, close examination of several studies reveals that both deviations toward and away from distractors can be seen in individual trials, even when the average deviation shows significant deviation toward (McPeek and Keller 2001) or away from (Theeuwes and Godijn 2004; Walker et al. 2006) the distractor. We hypothesized that this variability could be explained by the different time courses of the top-down and feedforward processes. For example, top-down inhibition of a target location may be slower than feedforward activation of that target location (McPeek et al. 2003).
On this view, increasing saccade latency should therefore promote the effects of top-down inhibition, increasing deviation away from a distractor. However, the relation between saccade latency and direction of saccade deviation remains unclear. Distractors influence latency, producing both increases and decreases, depending on the temporal relationship between stimulus onset asynchrony (Walker et al. 1995) and also cause trajectory deviations. Therefore a difference in direction of deviation between conditions with and without distractors could arise from the presence of the distractors or from the difference in saccade latency between these conditions. Previous studies reported correlations between latency and deviation (McSorley et al. 2004; Theeuwes and Godijn 2004), although these cannot be interpreted as a direct effect of latency on deviation because their experimental designs did not control for the additional influence of distractors on latency. Because these studies do not show a direct effect of latency on deviation away for the distractor, they cannot be taken as unambiguous evidence for the time course of top-down inhibition of distractors. A stronger test of this hypothesis would involve an experimental manipulation of saccade latency independent of the distractor effects on trajectory. Here we used a fixation gap paradigm (Ross and Ross 1980; Saslow 1967) to vary saccade latency. This manipulation, known to be independent of distractor effects on latency (Walker et al. 1995), allowed us to measure the time course of distractor-induced saccade deviation in a systematic way.
Six subjects made 420 saccades to a target cross (a white “×” 1-deg square with a line thickness of 6′ arc) 10 deg from central fixation (0.5-deg filled white circle) situated in one of four positions on the primary obliques (45, 135, 225, 315 angular deg). Distractors (an unfilled white circle 1 deg in diameter with a line thickness of 6′ arc) could appear 10 deg from fixation in one of two unpredictable flanking locations, each set 45 angular degrees away from the target (i.e., at 0, 90, 180, or 270 deg depending on the target position) as shown in Fig. 1. There were thus eight distractor present conditions and four no distractor single target (baseline) conditions. Distractors appeared simultaneously with the onset of the saccade target. The fixation spot was removed from the display at stimulus onset asynchronies (SOAs) of: −200, −100, −50, 0, 50, 100, or 200 ms relative to target onset. Negative values indicate that the fixation spot was removed before onset and positive values indicate that fixation was removed after onset. Collapsing across target directions there were 40 distractor trials and 20 no distractor trials for each of the seven gap intervals (total 420 trials per subject). Each trial began with the appearance of the central fixation spot that was displayed initially for a random foreperiod of 800–1,300 ms. After this period the target appeared randomly at one of the four locations for a period of 1,000 ms. All trial types were randomly interleaved within a block. Subjects were instructed to make a saccade to the target onset and to ignore the distractors. The display was then blanked for an intertrial delay of 600 ms. The spatial location of targets and distractors along with three sample saccade trajectories (made to a single target or to a target with a distractor) are shown in Fig. 1. The experiment was approved by the ethics committee at Royal Holloway University of London and was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki.
Eye movements were recorded using a head-mounted video-based eye tracker (Eyelink, Sensomotoric Instruments) with a sampling frequency of 250 Hz. Before each block of trials a calibration of the subjects’ eye position relative to fixed points on the monitor was performed. Eyelink software identified saccade start- and endpoints using a 22 deg/s velocity and 8,000 deg/s2 acceleration criterion. Saccades were excluded from further analysis if: the direction of the saccade was >30 angular deg either side of the target; amplitudes were 2.5 SDs away from the mean; or blinks occurred during the saccade.
The maximum trajectory deviation of each saccade relative to the direct path between fixation and landing position was found (Ludwig and Gilchrist 2002). Because saccade trajectories are never completely straight, the trajectory deviation observed in no distractor (baseline) conditions was subtracted from that for distractor conditions. Trajectories deviating toward the distractor were assigned positive values and those deviating away from the distractor, negative values. Trajectory deviation toward or away from the distractor was calculated for each factorial combination of target position and fixation offset time.
Trials were excluded from further analysis on the basis of latency (3%), amplitude (3%), and direction (7%). The remaining trials were collapsed across target direction. Figure 2 shows mean saccade latency for each SOA, collapsed across subjects, under baseline (no-distractor) and distractor trials. A monotonic change in latency was observed in both the baseline and distractor conditions across SOA. These mean differences were examined using a general linear model ANOVA. A two-factor ANOVA (SOA × distractor) showed a significant effect of the SOA (gap condition) on latency [F(6,30) = 34.1, P < 0.01] and a significant remote distractor effect (RDE) increase in latency on distractor trials [F(1,5) = 40.2, P < 0.01]. Importantly, the magnitude of this distractor effect on latency can be seen to be comparable across gap conditions as confirmed by the nonsignificant two-way interaction effect (F < 1). Thus the influence of distractors on latency (increase) was independent of the influence of fixation offset on latency (decrease). To examine the effect of saccade latency on saccade curvature, the latency distribution across all target–distractor conditions, for each subject, was divided into eight equal groups (octiled). The mean saccade trajectory deviation was then determined for successive 12.5% portions of the saccade latency distribution (while ignoring the fixation gap intervals). The mean deviation and latency for each octile were averaged across subjects and are plotted separately for gap and overlap conditions in Fig. 3A and combined in Fig. 3B. The ordinate shows the average curvature and the abscissa shows latency octiles. It is clear that quicker saccades tended to deviate toward the distractor whereas slower saccades deviated away. When latency was quick (<200 ms) saccades deviated toward distractors and when latency was longer (>200 ms approx.) saccades deviated away from distractors. A one-way ANOVA on the data shown in Fig. 3B confirmed a significant effect of saccade latency on saccade curvature [F(7,35) = 2.95, P < 0.05].
This study investigated the relationship between the distractor modulation of saccade trajectory and saccade latency. The well-established fixation “gap paradigm” (Dorris and Munoz 1995; Ross and Ross 1980; Saslow 1967) was used to modulate saccade latency by facilitating or impairing disengagement from central fixation. The use of a gap paradigm enabled us to manipulate saccade latency independently of the influence of distractors on latency. A clear relationship was revealed between saccade latency and the effects of distractors on the direction of initial trajectory deviation. Shorter-latency saccades deviated toward distractors whereas longer-latency saccades deviated away. Because the occurrence and magnitude of the fixation gap effect were robustly independent of the influence of distractors on latency (see also Walker et al. 1995) we believe that the gap effect does not directly influence saccade trajectory deviation. Rather, the gap effect influenced the time to initiate the saccade to the target. Gap-induced delays in saccade latency in turn allowed the process of distractor inhibition to influence the neural map responsible for saccade generation. Thus it cannot be the presence or absence of the fixation stimulus per se that influences the initial direction of trajectory deviation because shorter-latency saccades in the fixation overlap conditions deviated toward distractors and longer-latency saccades in the fixation gap conditions deviated away from distractors (Fig. 3A). Thus saccade latency appears to be the underlying factor involved in modulating the initial direction of saccade trajectory.
The results presented here showed an almost linear relation between latency and distractor-induced saccade deviation. This leads first to decreasing deviation toward the distractor, followed by an overt deviation away from it. The initial decrease in deviation toward the distractor could reflect the feedforward neural processes that select a single target location for saccade programming from among competing stimuli. Selection may involve local competitive inhibition between target and distractor locations in the neural map, leading to gradual suppression of the distractor location (Port and Wurtz 2003; Walton et al. 2005). The deviation away from the distractor observed at longer saccade latencies requires a different neural mechanism. Specifically, curvature away implies suppression of the distractor location below baseline, which cannot easily be accommodated by feedforward models of either averaging (Glimcher and Sparks 1993; van Opstal and van Gisbergen 1989) or of local inhibition (Munoz and Istvan 1998). Instead, we suggest that curvature away arises from a second, top-down, inhibitory input into the neural map for saccade generation. The bias exerted by the frontal eye fields on saccade-related neurons in the intermediate layers of the superior colliculus (Schlag-Rey et al. 1992) may be one part of this inhibitory circuit. The saccade latency at which the presence of a distractor begins to cause deviation away from rather than deviation toward may reflect the time course of this top-down process. In our data, this was around 200 ms, in close agreement with estimates based on different methods by Theeuwes and Godijn (2004) and Walker et al. (2006). However, the present study, unlike previous ones, manipulates saccade latency by a parameter that does not directly affect distractor-related processes. This would correspond to the point at which inhibition has successfully reduced the level of activity associated with the distractor to a level below that of the surrounding baseline so the initial saccade direction is away from that location. Our result is consistent with the finding in antisaccade studies (Hallett 1978, 1980), and in the control of voluntary action more generally (Day and Lyon 2000), that generating a motor response to a target is fast while preventing it is relatively time consuming. Finally, it is worth noting that deviation toward distractors has been observed in studies using humans and monkeys as participants, but deviation away from distractors has been observed with human participants only (Van der Stigchel et al. 2006). As monkeys tend to have shorter latency saccades than humans (Fischer and Weber 1993) it may be that latency is again the underlying factor involved in trajectory deviation.
The work was supported by a Leverhulme Trust grant to R. Walker and P. Haggard.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2006 by the American Physiological Society