Remote distractor effects and saccadic inhibition: Spatial and temporal modulation

Abstract

Abstract

Abstract:

Abstract  The onset of a visual distractor remote from a saccade target is known to increase saccade latency (the remote distractor effect [RDE]). In addition, distractors may also selectively inhibit saccades that would be initiated about 90 ms after distractor onset (termed saccadic inhibition [SI]). Recently, it has been proposed that the transitory inhibition of saccades (SI) may underlie the increase in mean latency (RDE). In a first experiment, the distractor eccentricity was manipulated, and a robust RDE that was strongly modulated by distractor eccentricity was observed. However, the underlying latency distributions did not reveal clear evidence of SI. A second experiment manipulated distractor spatial location and the timing of the distractor onset in relation to the target. An RDE was again observed with remote distractors away from the target axis and under conditions with early-onset distractors that would be unlikely to produce SI, whereas later distractor onsets produced an RDE along with some evidence of an SI effect. A third experiment using a mixed block of target-distractor stimulus-onset asynchronies (SOAs) revealed an RDE that varied with both distractor eccentricity and SOA and changes to latency distributions consistent with the timing of SI. We argue that the notion that SI underpins the RDE is similar to the earlier argument that express saccades underlie the fixation offset (gap) effect and that changes in mean latency and to the shape of the underlying latency distributions following a visual onset may involve more than one inhibitory process.

Introduction

Saccadic eye movements are made to shift our gaze and attention onto objects of interest. The selection of a saccade goal reflects competitive interactions between competing stimuli that can be probed by presenting distractors along with the saccade target. The effects of distractors can range from facilitation to inhibition depending on the spatial and temporal relationship between the target and distractor. Distractors at remote spatial locations, presented in close temporal proximity to the target onset, increase saccade latency (the remote distractor effect [RDE]) but have little effect on landing position (Walker, Deubel, Schneider, & Findlay, 1997). By contrast, distractors in close spatial proximity to the target have small or negligible effects on latency but can modulate landing position (the “global effect”; Findlay, 1982). Studies examining the influence of remote distractors on saccades have adopted two different methodological approaches: one is to report the influence of distractors on mean saccade latency (the RDE; Ludwig, Gilchrist, & McSorley, 2005; Walker et al., 1997; Walker, Kentridge, & Findlay, 1995), whereas another approach is to examine the underlying latency distributions for evidence of a transitory inhibitory effect (termed saccadic inhibition [SI]) that knocks out a proportion of responses time locked to distractor onset. The SI effect is manifest as a decrease in saccade frequency observed shortly after distractor onset, with a maximum inhibitory influence occurring around 90 ms later (Reingold & Stampe, 1999, 2002, 2004). 

Recently, it has been proposed that both the RDE and SI reflect the same underlying phenomena and that the increase in mean latency observed in the RDE can be attributed to the inhibition of a population of saccades, about 60 to 120 ms after distractor onset, which results in an increase in the frequency of longer-latency responses (Bompas & Sumner, 2011; Buonocore & McIntosh, 2008). The effect of distractors on both mean latency and SI has been shown to be modulated by the timing of the distractor onset, but investigations of spatial effects have largely been confined to studies of the RDE. If both the RDE and SI reflect similar inhibitory processes, both effects would be expected to show similar modulation depending on the spatial location and eccentricity of the distractor. This was examined here in three experiments that manipulated distractor spatial location and timing. 

Spatial modulation of RDE

The increase in saccade latency with remote distractors was first reported by Lévy-Schoen (1969), who found that saccade latency increased when bilateral targets were presented simultaneously in both visual fields. This is in contrast with the small effect on latency of a distractor presented close to the target that modulates the saccade landing position (global effect). The RDE has since been shown to depend, in a highly systematic way, on the spatial relationship between the distractor and saccade target. Walker et al. (1997) found a large RDE when a small visual distractor (0.5°) appeared at fixation, and the effect decreased dramatically as distractor eccentricity increased. A small RDE effect can be observed with distractors as far as 10° from fixation, showing that inhibitory effects on saccade initiation are not confined to a small central fixation zone (Gandhi & Keller, 1999). It was also shown that the magnitude of the distractor effect depended most critically on distractor eccentricity, rather than the target-distractor distance. A distractor located in the same hemifield as the saccade target produced a similar RDE increase in latency as a distractor in the opposite hemifield (as long as it was located outside the critical window of 20° around the target axis). The exact relationship between the magnitude of the RDE and distractor eccentricity was best described as a function of the ratio of the distractor and target eccentricity rather than the absolute distance between the target and distractor (Walker et al., 1997). Thus, the inhibitory effect of distractors, outside a critical spatial window of ±20° around the target axis, depends on the relative eccentricity of the distractor and target from fixation rather than on the target-to-distractor distance as might be expected. It has also been found that these effects are reliably observed when target direction is predictable, but if target direction is not predictable, then additional higher-level cognitive factors (possibly related to the additional process of target discrimination) can change the nature of the observed RDE (Benson, 2008). 

Temporal modulation of RDE

Walker et al. (1995) examined the effect of manipulating the distractor-target stimulus-onset asynchrony (SOA) and observed the largest increase in latency (RDE) when distractors appeared simultaneously to, or shortly after (40 ms), the target (see also Buonocore & McIntosh, 2008). By contrast, distractors that appeared at longer intervals (160–240 ms) before the target produce a substantial reduction in latency, consistent with warning signal explanations (Ross & Ross, 1980, 1981). Thus, distractors appearing close to the time of the target onset increase latency, distractors appearing at longer intervals before the target reduce latency, and distractor onsets delayed for longer intervals (>50 ms) after the target onset have negligible effects. The shape of the underlying latency distributions is also found to be modulated by distractor-target SOA: Near-simultaneous distractor onsets produced an increase in positive skew (reflecting an increase in variance and longer-latency responses), whereas distractors presented before the target produced a leftward shift, reflecting an increase in short-latency express saccades (Fischer & Ramsperger, 1984) when compared with the baseline distributions on single-target trials. The temporal modulation effect has been shown to depend on a critical window (±20 ms around target onset) during which time the distractor-related inhibition is able to interfere with saccade target selection (Bompas & Sumner, 2009). 

Saccadic inhibition

Reingold and Stampe (1999, 2002, 2004) first noted SI in a series of studies designed to examine the effect of large task-irrelevant visual transient display changes (as brief as 33 ms) on saccades made in reading. They showed that the increase in fixation durations observed on flash trials could be related to a selective notched dip in the underlying frequency distributions that started around 70 to 90 ms after the display change was made. The decrease in saccade frequency was attributed to low-level automatic inhibitory processes rather than to high-level cognitive factors and was time locked to the distractor (flash) onset (Reingold & Stampe, 2004). SI has since been shown to generalize to less cognitive tasks, for example, in the case of stimulus-elicited (so-called “reflexive”) saccades made toward a peripheral visual target and also for anti-saccades that depend on higher-level endogenous control processes (Reingold & Stampe, 2002). The most important factor is to present the flash at the critical time period, in relation to the underlying central tendency (mean or median latency), so it can exert the greatest influence on the maximum number of saccades (Reingold & Stampe, 2002). The SI effect has been interpreted in terms of inhibitory oculomotor processes similar to those developed in studies of the RDE (Walker et al., 1995; Walker et al., 1997). 

SI and the RDE

The early studies of Reingold and Stampe (2002) used large visual display changes, although similar SI effects have since been observed with smaller distractor stimuli (0.5°) more similar to those used in studies of the RDE. In these studies, the maximum SI effects were found about 90 to 100 ms after distractor onset (Buonocore & McIntosh, 2008), supporting the idea that both effects have similar underlying causes. In contrast to studies of the RDE, the spatial modulation of distractor location on SI has received relatively little attention. One exception is a study by Edelman and Xu (2009), who reported an SI effect with both contralateral and ipsilateral distractors as long as distractors appeared outside a 20° window around the target axis. This spatial region is comparable to that over which the RDE operates (Walker et al., 1997). In this study, distractor eccentricity was fixed and the effects on mean latency (and the RDE) were not reported, preventing conclusions about the relationship between SI and the RDE to be drawn. 

The neural basis of SI and the RDE

The neural basis of saccade target selection involves multiple cortical regions including the intraparietal area, the frontal eye fields (FEFs), and the supplementary eye fields that project to brainstem structures including the superior colliculus (SC) and the brainstem oculomotor nuclei (Krauzlis, Liston, & Carello, 2004; Schall, 1995). Accounts of both SI and the RDE have invoked competitive interaction effects operating between separate populations of neurons in the intermediate layers of the SC. The onset of a peripheral stimulus increases activity in caudal buildup neurons that encode the spatial location of a planned motor response. Activity in populations of buildup neurons is thought to be mutually inhibitory if remote and excitatory if close (Meredith & Ramoa, 1998). According to lateral inhibition models of the RDE (Trappenberg, Dorris, Munoz, & Klein, 2001), neurons encoding a remote distractor inhibit the rise of target-related activity, thus delaying saccade initiation. An alternative view is that the distractor produces activity in a fixation system that serves to inhibit saccades by downstream projections to the brainstem (Walker et al., 1997). One argument against the fixation model was that fixation neurons were located in the foveal zone and would not account for distractor effects outside this region. More recent studies have shown that fixation neurons are best conceived of as saccade-related neurons that encode small amplitudes (Hafed & Krauzlis, 2012; Krauzlis, Basso, & Wurtz, 1997), and distractors outside the fixation zone may therefore be expected to have qualitatively similar inhibitory effects as those at fixation. Which of these models best accounts for the RDE is controversial (Casteau & Vitu, 2012), and inhibitory influences may not be restricted to the SC alone (Tehovnik, Sommer, Chou, Slocum, & Schiller, 2000). SI has also similarly been attributed to lateral interactions between target and distractor-related activity (Bompas & Sumner, 2011; Buonocore & McIntosh, 2008; Edelman & Xu, 2009; Reingold & Stampe, 2002). Reingold and Stampe (2002) discounted the possibility that SI reflects fixation activity because the display flash was located outside the region that would classically be regarded as containing fixation neurons (Munoz & Wurtz, 1993) and attributed SI to the inhibition of remote buildup neurons. A further possibility is that SI may be attributed to the fast-acting transitory inhibition of preparatory activity in the SC (Dorris, Olivier, & Munoz, 2007) rather than the target-related buildup activity. SI may therefore reflect a short-lasting transitory inhibitory effect in the SC, whereas the RDE could involve inhibition within the SC along with other slower-acting inhibitory influences from cortical regions such as the FEFs (Sommer & Wurtz, 2000). 

The goal of the present study was to examine SI and the RDE to see if both effects show a similar spatial modulation as would be predicted if they reflect similar inhibitory processes. We reasoned that if this were the case, both SI and the RDE should be modulated by distractor eccentricity, with the maximum effects expected with distractors at central fixation, and the effects decreasing with increasing distractor eccentricity. This was examined here in a series of experiments in which distractor eccentricity (Experiment 1) and distractor location and timing (Experiments 2 and 3) were manipulated. 

Experiment 1

Methods

Participants

Five unpaid volunteers (three male), age range 21 to 48 years (four naïve plus one of the authors), participated in Experiment 1. Each participant completed 2,880 trials in three separate testing sessions of approximately 40 min each. The experiments were approved by the Psychology department's ethical committee and conformed to the Declaration of Helsinki for testing human participants. 

Apparatus

Eye movements were recorded using the SR Research Ltd. Eyelink II video-based system. This tracker has a spatial accuracy of <0.5° and a sampling rate of 250 Hz (4 ms temporal resolution). The left eye was recorded for all participants in pupil-only mode. The online saccade detection algorithm used by the eye tracker detected saccades greater than 0.5° using an acceleration (9500°/s₂) and velocity (30°/s) threshold. Participants viewed a 21-in. CRT monitor (100 Hz). A head and chin rest was used to stabilize the participants' position, which was 57 cm from the screen with the eyes level with the screen center. The two video cameras used to track the eyes were mounted onto the head and chin rest, and the head-tracking mode was not required. 

Materials and design

Stimuli were white presented on a black background. At the start of each trial, a small (0.5°) white plus sign appeared in the center of the screen and acted as a fixation stimulus (see Figure 1). After an interval of 500 ms, the fixation stimulus disappeared, and simultaneously with this change, a small target (white triangle 0.5°) appeared at either 4° or 8° in a fixed direction along a horizontal axis aligned with the screen center. The fixed target direction alternated every 48 trials (left or right). On the majority of trials (7/8), a distractor (white circle 0.5°) appeared either at fixation (0°) or 2°, 4°, 8°, or 16° in the opposite hemifield. Distractors were presented along a horizontal axis, level with the center of the screen and appeared simultaneously with the target. The screen was then cleared for an intertrial delay of 1 s. Participants completed 10 blocks of 96 trials in a single session and completed three sessions in total on different days. Each participant completed 240 trials per distractor condition or 2,880 trials in total. 

Figure 1

 

(a) A schematic representation (not to scale) of the display sequence used for a single target (control, right) or target and distractor (left) trial in Experiment 1. A small plus sign acted as a fixation stimulus at the start of each trial for 500 ms. Targets (small white triangles) appeared (randomly) at either 4° or 8° in a fixed direction that alternated every 48 trials. Distractors (white circles) appeared, simultaneously with target onset, at eccentricities of 0°, 2°, 4°, 8°, or 16° in the opposite visual field (as shown, far right panel). The distractor and target stayed on for 1000 ms and were followed by a blank screen (not shown) for an intertrial interval (ITI) of 1000 ms. (b) Display sequence for a contralateral (45° axis) distractor trial (left panel) and an ipsilateral (0° axis) distractor trial (right panel) in Experiment 2. The target and distractor stimuli were filled rather than empty, and a trial sequence was as follows: random fixation foreperiod (1000–1200 ms, step 10 ms), target (1000 ms), blank screen for ITI (1000 ms). Targets always appeared at an eccentricity of 4° in a fixed direction that alternated every 49 trials. Distractors appeared at an eccentricity of 2° in either the ipsilateral or contralateral visual fields at one of six locations: on the horizontal target axis (at ±180°) or on an axis ±22°, ±45°, from horizontal in the upper visual field.

Procedure

A nine-point two-dimensional (2D) calibration routine was performed at the start of each block of trials and was repeated if the average spatial accuracy was >0.5°. A drift correction procedure was used to correct for small head movements at the start of a block of 48 trials, and participants were then informed of the upcoming target direction. Participants were instructed to look at the central fixation stimulus and to then make an eye movement to the target when it appeared but to avoid making anticipatory responses. A minimum of 20 practice trials was performed at the start of each testing session as required. 

Data analysis

Saccades were excluded from the analysis if they were made in the wrong direction or if they were outside ±2° of the target amplitude. They were also excluded if they had a latency of <60 ms or if they had latency greater than the mean plus 2.5 times the standard deviation for all trials for that participant. 

SI analysis

The analysis had a number of discrete steps and was based on that used by Buonocore and McIntosh (2008) and Reingold and Stampe (2002). Frequency distributions of saccade latency were created for each individual participant (bin width = 4 ms) collapsed across target eccentricity. The maximum number of data points per condition was 240, and some bins (those <100 ms) tended to contain smaller numbers of trials. To reduce noise in these distributions, each bin was replaced by an average of itself and the three preceding and three following bins (cf. Reingold & Stampe, 2002). 

Results

All trials in which eye movements were made in the wrong direction were discarded (2.6%), and all those outside ±2° of the target were rejected (8.1%); saccades with very short latency (3.3%) and those with excessively long latency (1.8%) were also excluded from further analysis. 

Remote distractor effect

The mean of the median RDE (latency on distractor trials – single target baseline mean latency) is shown in Figure 2 for each target eccentricity as a function of distractor condition. From Figure 2, it can be seen that the greatest RDE effect (+22.9 ms collapsed across target eccentricity) occurred with distractors at fixation (D0°), which decreases as distractor eccentricity increases but is still apparent (+5.6 ms) with distractors at the furthest eccentricity (D16°). A two-factor (target eccentricity [2], distractor condition [6]) analysis of variance (ANOVA) revealed no effect of target eccentricity, F(1, 4) < 1, but a highly significant effect of distractor eccentricity, F(4, 16) = 10.6, p = 0.012, and no interaction effect between target and distractor eccentricity, F(4, 16) < 1. The data were collapsed across target eccentricity, and the mean latency and RDE effect are shown in Table 1 for each distractor eccentricity. Simple contrasts revealed a significant increase in latency for all distractor eccentricities compared with the single target mean (123.45 ms): D0°, F(1, 4) = 47.93, p = 0.002; D2°, F(1, 4) = 17.54, p = 0.014; D4°, F(1, 4) = 13.36, p = 0.022; D8°, F(1, 4) = 13.20, p = 0.022; D16°, F(1, 4) = 20.4, p = 0.011. Thus, a robust RDE effect was observed, which decreased with distractor eccentricity but remained significant with distractors as far as 16° from fixation. 

Figure 2

 

Mean (of median) remote distractor effect (in ms) observed in Experiment 1 for targets at eccentricities of 4° (filled circles) or 8° (filled diamonds) as a function of distractor eccentricity. D0: Distractor at fixation. D2, D4, D8, D16: Distractors at 2°, 4°, 8°, or 16° in contralateral visual field.

Table 1

 

Mean (of median) saccade latency (standard error shown in parentheses) and remote distractor effect (RDE, in milliseconds), collapsed across target eccentricity for Experiment 1. Notes: The average value of Dip_max (the minimum value of distractor minus baseline distributions) and the time to maximum dip – Dip_T (ms) is also shown for each distractor condition. **Indicates latency is significantly greater than the single target mean at p < 0.01 or *p < 0.05.

	Mean (of median) latency	RDE (ms)	Dipmax	DipT
Single target	123.45 (2.48)	—	—	—
Distractor: 0°	146.35 (2.72)	22.90**	−0.0410	120
Distractor: 2°	135.15 (3.98)	11.70**	−0.0316	116
Distractor: 4°	131.70 (3.76)	8.25*	−0.0283	116
Distractor: 8°	129.65 (3.35)	6.20*	−0.0229	112
Distractor: 16°	129.00 (3.34)	5.55*	−0.0196	120

Saccadic inhibition

Figure 3a shows the latency distributions (collapsed across target eccentricity) for each individual participant (P1–P5) in the single-target (baseline) condition (shaded histogram), along with the distributions for each distractor eccentricity overlaid (line plots). Figure 3b shows average distributions for all of the data along with plots resulting from the bin-by-bin subtraction method (dip plots). The average values of Dip_max (the minimum value of distractor minus baseline distributions and the time to reach minimum) – Dip_T (calculated using the individual data) are shown in Table 1. The baseline histograms reveal wide variability in their underlying shape across participants. The overlaid distractor distributions show an increase in positive skew and general broadening that is modulated by distractor eccentricity. The modulation is most pronounced for the distractor at fixation (D0°) condition. The overlaid distributions for the distractor conditions do not reveal evidence of a notched dip that would be expected around 90 to 100 ms (express saccades) after distractor onset, for any of the individual participants or for the average (combined) data. Although the distributions do not show evidence of a dip, the results of the bin-by-bin subtraction (Figure 3b) do reveal an apparent dip in saccade frequency occurring at about 110 to 120 ms across all distractor eccentricities. The maximum decrease is shown for distractors at fixation (D0°), and this decrease reduces with increasing distractor eccentricity (see Table 1). Thus, the distributions themselves do not show clear evidence of SI, whereas the subtraction plots appear to show a decrease in saccade frequency that could be interpreted as an SI effect. The Dip_max measure of peak SI (Table 1) shows a similar modulation (increasing as distractors approach fixation) to the effect observed on mean latency (Figure 2). 

Figure 3

 

(a) Latency distributions for Participants 1 to 5 for each distractor location (see text for details) collapsed across target eccentricity. The histogram shows the latency distribution for the single-target baseline condition, and the superimposed line shows the distribution for the distractor condition (D0°–D16° columns from left to right). (b) Latency distributions for combined data for all participants and the subtraction dip plots (baseline subtracted from distractor distribution) for each distractor eccentricity.

The dips observed in the plots following the bin-by-bin subtraction of each latency bin (see Table 1 for maximum values of each dip [Dip_max] and time to dip max [Dip_T]) are, however, potentially misleading as the defining diagnostic criteria for SI. This is because such dips can be attributed either to the selective inhibition of a population of short-latency saccades (SI) or, alternatively, to other changes in the shape of the distractor-related distribution that may reflect both transitory and sustained inhibitory influences. This can be illustrated by the following example in which two (identical) distributions are shifted (by 4 ms) and the bin-by-bin subtraction technique is performed (see Figure 4). The subtraction plot in Figure 4 shows a pronounced dip at about 100 ms, which appears to be an SI-like effect but actually reflects the result of shifting the identical distributions. This diplike effect can be attributed to the properties of the normal (or approximately normal) distribution, whereby the curve reflects a rapidly increasing saccade frequency, which is then subtracted from the much smaller frequency located in the left tail of the original distribution. The apparent dips revealed in the subtraction plots in Figure 3 (and associated measures such as Dip_max) can therefore be attributed to the increase in positive skew of the distractor distributions rather than a transitory inhibitory effect (SI). The presence of dips in the distractor-related latency distributions would provide a less equivocal diagnostic signature of SI, although the timing of distractor onset in Experiment 1 may not have been ideal to reveal it. 

Figure 4

 

An illustration of a dip plot produced by the bin-by-bin subtraction of two identical distributions shifted (by 4 ms) with respect to each other (left). A clear dip can be seen in the subtraction plot (right), which in this case cannot be attributed to the selective inhibition of a population of short latency responses.

Discussion

Experiment 1 has shown that a distractor onset, appearing simultaneously with the target, located in the opposite visual field, produces a reliable increase in saccade latency that is strongly modulated by the distractor eccentricity (Walker et al., 1997). As expected, the maximum RDE occurred with distractors close to fixation and the effect decreased with increasing distractor eccentricity (Walker et al., 1997), although here a small effect was apparent with distractors as far as 16° from fixation. The latency distributions for the distractor conditions do not, however, show clear evidence of a notched dip that would be expected at about 60 to 110 ms if the data revealed an SI effect (Reingold & Stampe, 2002). Instead, distractors appear to produce a generalized rightward shift in the distribution arising from an increase in variance and in the frequency of longer-latency saccades. The Dip_max measure (of peak SI) produced using the bin-by-bin subtraction method showed a similar modulation with distractor eccentricity, as did the RDE. This does not, however, have to be attributed to a transitory inhibitory effect but can be accounted for by the increase in positive skew of the distractor-related distributions. Although the Dip_max measure is not a reliable diagnostic indicator of SI, it can be taken as a measure of the time at which inhibition starts to influence saccade generation. Bompas and Sumner (2011) have reported a high correlation between the timing of dips, across a range of SOAs, including those with simultaneous target-distractor onsets as used here. Thus, the increase in skew observed with simultaneous onsets may reflect SI (a short-lasting effect) or could potentially be attributed to more than one inhibitory process. 

SI has most often been revealed in studies in which the timing of the distractor onset is manipulated taking into account the underlying central tendency for the single-target condition (Bompas & Sumner, 2011; Buonocore & McIntosh, 2008; Edelman & Xu, 2009; Reingold & Stampe, 2002). A second experiment was performed to further examine the relationship between the RDE and the SI effect, manipulating target-distractor SOA. Distractors appeared either before, simultaneously with, or after the target (in separate blocks). In addition, a wider range of distractor spatial locations was examined, with distractors appearing at a range of 2D locations. The conditions included ipsilateral distractors on the same axis as the saccade target, a situation that has small or negligible effects on latency but modulates amplitude (the global effect (Findlay, 1982) as well as remote distractors (on an axis >20° from the horizontal target axis) in both the ipsilateral and contralateral fields, which should increase saccade latency without modulating amplitude. 

Experiment 2: The RDE and SI across SOA

In Experiment 2, the spatial relationship between distractor location, the RDE, and SI was examined under conditions in which the timing of distractor-target onset was manipulated. Distractors appeared 60 ms before, simultaneous with, or 60 ms after the target, in separate blocks of trials. A distractor appearing 60 ms before the target would be too early to produce an observable SI effect and may also result in a reduction in mean latency due to a generalized nonspatial warning-signal effect (Walker et al., 1995). By contrast, delaying distractor onset by 60 ms should allow the SI effect to be observed as a notched dip in the underlying latency distribution some 90 ms after distractor onset and may result in a reliable RDE (Buonocore & McIntosh, 2008). A simultaneous target-distractor onset (0 ms SOA) condition may allow both SI and the RDE to be observed when a fixed target location is used. Experiment 2 included a wider range of distractor locations with distractors presented in both the ipsilateral and contralateral visual fields at a range of 2D positions (similar to the studies of Walker et al., 1997). Remote distractors appearing in both the contralateral and ipsilateral visual fields should produce similar effects on the RDE and would therefore be predicted to have similar effects on SI if the same inhibitory processes underpin both effects. Distractors close to the target (<20° from target axis), by contrast, are known to modulate saccade landing position but not latency and so would not be expected to produce an SI effect. 

Methods

Apparatus, procedure, and analysis

The apparatus, procedure, and analysis were as described for Experiment 1. 

Participants

Five participants (four naïve undergraduate volunteers and one of the authors) participated in each of the conditions. The ages ranged from 20 to 48 years (mean = 28 years), and all had normal or corrected-to-normal vision. 

Materials and design

Stimuli were white presented on a black background (see Figure 1b). A small white plus sign (0.5°) appeared as a fixation stimulus at the start of each trial. A variable fixation interval (1000–1200 ms, random 10-ms step) was incorporated to reduce anticipatory responses. Following this period, a small target (filled white triangle 0.5°) appeared at 4°, in a fixed direction, along a horizontal axis that was level with fixation. On the majority of trials (6/7), a distractor (white filled circle 0.5°) appeared at an eccentricity of 2° from fixation. Distractor location varied (randomly) across the six positions and could appear in either the contralateral or ipsilateral visual field to the target. Distractors appeared at 2D locations on axis 22° or 45° from horizontal (see Figure 1b), and those presented off the horizontal axis appeared in the upper visual field only. The fixed target direction (left or right) alternated every 49 trials, and a drift correction procedure was used following each change of direction. Distractors appeared at SOAs of −60 ms before, simultaneously with (0 ms), or +60 ms after the target onset (separate blocks of trials). The baseline condition used an invisible distractor (Edelman & Xu, 2009) in which a black distractor was presented, thereby maintaining the same timing for single-target trials as in distractor trials. Participants first completed a practice session of 20 trials and then a total of 10 blocks of 98 trials in a single session, producing a maximum of 140 trials for each distractor position and SOA. The timing of distractor onset SOA (−60 ms, 0 ms, or +60ms) was fixed within a session (Conditions 2a–c, respectively). 

Results

During the practice sessions, it was noticeable that some participants made many anticipatory saccades around the time of the target onset. In these instances, they were reminded not to make a response until the target had appeared, and further practice trials were given. The number of trials excluded and the mean (of the medians) saccade latency for each condition are shown in Table 2. The mean baseline latency of saccades to single targets in the −60 ms SOA condition is shorter (112.0 ms) than in the 0 ms and +60 ms SOA conditions (155.0 ms and 148.1 ms, respectively). The overall reduction in latency with an early-onset distractor may reflect a generalized nonspatial alerting effect (see the Discussion section). The pattern of distractor modulation on mean latency as shown in Figure 5 can be seen to depend on both the timing (SOA) and spatial location of the distractor onset. Ipsilateral distractors appearing before (−60 ms SOA) the target reduced latency (Walker et al. 1995) but only when they appeared close to the target axis (0°–22°). By contrast, remote distractors increased latency, with the greatest RDE observed with distractors in the contralateral visual field. Distractors appearing at 0 ms and +60 ms SOAs in both the ipsilateral and contralateral visual field all produced an RDE increase in latency, which appears greatest for the simultaneous (0 ms SOA) onset condition. For the simultaneous (0 ms SOA) and delayed (+60 ms SOA) distractor onset conditions, the greatest RDE effect is observed with contralateral distractors, although a small latency increase is apparent with ipsilateral distractors remote from and on the target axis itself. 

Figure 5

 

Mean of median RDE (ms) for Experiment 2a–c, collapsed across participants for each distractor location and SOA (0 ± 60 ms). Contra: Distractor in hemifield opposite target. Ipsi: Distractor in same hemifield as target. Distractor axis is as follows: horizontal = 0°, ±22°, or ±45° away from the horizontal in upper field (only).

Table 2

 

Mean (of median) saccade latency for Experiment 2 (standard error in parentheses). Notes: The average values of Dip_max and Dip_T (the time to reach maximum decrease) time locked to the target onset are shown in parentheses. Distractors appeared at SOAs of −60 ms before, simultaneously with, or +60 ms after the target. Distractor locations were as follows: Ipsi, same hemifield as target; Contra, opposite hemifield to target; 0 = horizontal axis, 22°, 45° = angular distance from horizontal axis in upper visual field. The percentage of trials removed, based on the exclusion criteria applied, are also shown.

	Distractor −60 ms SOA	Distractor 0 ms SOA	Distractor +60 ms SOA
Single target
Latency	112.0 ms (1.3)	155.0 ms (8.8)	148.1 ms (5.6)
Amplitude	3.50°	3.49°	3.33°
Ipsi 0° axis
Latency	98.5 ms (4.2)	163.5 ms (9.4)	151.7 ms (4.4)
Amplitude	3.09°	3.12 °	3.15 °
Dipmax (DipT ms)	−0.0548 (113 ms)	−0.0203 (120 ms)	−0.0177 (132 ms)
Ipsi 22° axis
Latency	111.1 ms (5.0)	164.8 ms (9.2)	153.2 ms (5.9)
Amplitude	3.38°	3.40 °	3.25 °
Dipmax (DipT ms)	−0.0392 (111 ms)	−0.0212 (127 ms)	−0.0254 (143 ms)
Ipsi 45° axis
Latency	117.7 ms (3.6)	165.6 ms (8.6)	155.8 ms (7.7)
Amplitude	3.55°	3.52 °	3.30 °
Dipmax (DipT ms)	−0.0283 (112 ms)	−0.0238 (127 ms)	−0.0274 (150 ms)
Contra 45° axis
Latency	127.2 ms (1.6)	167.0 ms (9.9)	149.8 ms (5.6)
Amplitude	3.64°	3.56 °	3.36 °
Dipmax (DipT ms)	−0.0358 (98 ms)	−0.0256 (119 ms)	−0.0255 (144 ms)
Contra 22° axis
Latency	128.5 ms (1.6)	167.4 ms (9.2)	153.8 ms (5.8)
Amplitude	3.59°	3.54 °	3.33 °
Dipmax (DipT ms)	−0.0405 (94 ms)	−0.0226 (122 ms)	−0.0242 (145 ms)
Contra 0° axis
Latency	131.8 ms (3.6)	170.5 ms (8.4)	154.1 ms (6.2)
Amplitude	3.66°	3.53 °	3.33°
Dipmax (DipT ms)	−0.0425 (94 ms)	−0.0245 (122 ms)	−0.0239 (142 ms)
Exclusion rates (%)
Short latency	2.3	0.7	3.2
Long latency	10.1	14.1	4.4
Amplitude	2.5	1.1	0.4
Wrong direction	0.75	0.3	0.1

For the −60 ms SOA condition, ipsilateral distractors reduced latency, and remote distractors (in both the ipsilateral and contralateral hemifields) increased latency. A one-way ANOVA confirmed a significant main effect of distractor location, F(6, 24) = 22.86, p < 0.0001, ηp2 = 0.851. Paired contrasts confirmed that ipsilateral distractors on the horizontal target axis significantly reduced saccade latency, Ipsi 0°-axis F(1, 4) = 12.60, p = 0.024, whereas those at 22° and 45° had no significant effect, Ipsi 22°-axis F(1, 4) < 1, p = 0.835; Ipsi 45°-axis F(1, 4) = 4.85, p = 0.092, respectively. By contrast, contralateral distractors significantly increased saccade latency, compared with the single target, for all distractor locations, Cont 45°-axis F(1, 4) = 76.0, p = 0.001; Cont 22°-axis F(1, 4) = 77.78, p = 0.001; Cont 0°-axis F(1, 4) = 31.45, p = 0.005. 

For the simultaneous onset (0 ms) SOA distractor condition, a small increase in latency was observed with ipsilateral distractors and a larger increase with contralateral distractors. The ANOVA confirmed a significant effect of distractors, F(6, 24) = 10.28, p < 0.0001, ηp2 = 0.72, and paired contrasts showed that this increase was significant across all distractor locations, Ipsi 0°-axis F(1, 4) = 26.27, p = 0.007; Ipsi 22°-axis F(1, 4) = 19.06, p = 0.012; Ipsi 45°-axis F(1, 4) = 28.19, p = 0.006, and contralateral, Cont 45°-axis F(1, 4) = 43.63, p = 0.003; Cont 22°-axis F(1, 4) = 48.65, p = 0.002; Cont 0°-axis F(1, 4) = 57.89, p = 0.002, visual fields. For the delayed distractor onset (+60 ms) SOA condition, a small increase was observed across all distractor locations, similar to that produced with simultaneous distractor onsets but of smaller magnitude. An ANOVA showed that this small increase was not significant, F(6, 24) = 2.92, p = 0.163, ηp2 = 0.422. 

Saccadic inhibition

Figure 6 shows the combined latency distributions for the data obtained from all five participants and the distributions for two participants (P1 and P2) who participated in all three conditions (enabling direct visual comparisons to be made across all conditions). The baseline (no distractor) distributions (histogram) are shown, with the distributions for ipsilateral and contralateral distractors for each distractor axis overlaid in separate plots. 

Figure 6

 

Latency distributions obtained from data from all participants combined (top row) and two participants (P1 and P2; center and lower rows) from Experiment 2. Left panels: SOA −60 ms. Center panels: SOA 0 ms. Right panels: SOA +60 ms. Separate lines are used to indicate the distractor axis. Ipsi: Ipsilateral distractors. Contra: Contralateral distractors, presented on axis 0°, 22°, or 45° from horizontal. The target and distractor eccentricities were 4° and 2°, respectively.

−60 ms SOA

The latency distributions for the combined data (n = 5) in Figure 6 show that the early onset of an ipsilateral distractor, close to the target axis (0° and 22° axis), has facilitated the production of short-latency saccades producing a leftward shift, whereas ipsilateral distractors away from (45° axis) the horizontal axis have had little observable effect compared with the baseline distributions. Contralateral distractors, at all locations, produced a small increase in the frequency of longer latency saccades, producing an increase in skew in the underlying distribution (similar to that observed in Experiment 1). The overlaid distractor distributions do not reveal a notched dip characteristic of an SI effect. The values of Dip_max and the late timing of Dip_T ∼ 170 ms after distractor onset can here be attributed to the leftward shift in the distractor-related latency distributions with a later decrease in saccades rather than SI (similar to the apparent dip in Figure 4). The plots for the two individual participants (P1 and P2) show some variability in the underlying latency distributions as would be expected. The influence of distractors for P1 appears comparable to those in the overall combined data. Ipsilateral distractors close to the target axis have produced a leftward shift, whereas remote contralateral distractors have produced a small rightward shift. There is no evidence of notched dips for P1 with any distractor condition. The distractor distributions for P2 show a decrease in saccade frequency with near ipsilateral distractors (0° and 22°), although this appears at about 140 ms after target onset (200 ms after distractor onset), which is much later than would be expected for the SI effect. It was noticeable that this participant found it difficult to avoid making anticipatory responses in this condition, and it can be seen that she made many short latency saccades with near ipsilateral distractors (0° and 22° axis) that may account for the dip observed at 140 ms under these conditions. There is no evidence of a notched dip occurring with remote ipsilateral (45° axis) and contralateral distractors. Thus, the −60 ms SOA has not revealed evidence of an SI effect, as would be expected given the expected time course of SI, which would be predicted to peak about ∼40 ms after target onset, which is potentially too early to influence the latency distributions. 

0 ms SOA

The combined latency distributions (n = 5) with both ipsilateral and contralateral distractors appear broadly similar to those observed in Experiment 1 with a tendency for a rightward shift across all distractor locations. It should be noted here that ipsilateral distractor on the horizontal target axis is the condition expected to modulate the saccade landing position (global effect condition) and not latency, but a rightward shift is observed in the latency distribution consistent with the small RDE effect that was observed. There is no evidence of a notched dip in the overall combined data for ipsilateral or contralateral distractor conditions, but the increase in skew produces a maximum Dip_max some 120 ms after distractor onset, similar to that in Experiment 1. There is some evidence of SI in the latency distributions for participant P1, and there is some suggestion of a dip occurring about 180 ms (rather late for SI) for P2, with contralateral distractors on the horizontal (0°) axis only. Thus, and similar to the findings from Experiment 1, the simultaneous distractor condition has produced an RDE effect with changes to the shape of the distractor-related latency distributions starting about 90 to 100 ms after distractor onset. 

+60 ms SOA

The combined latency distributions (n = 5) for the delayed onset distractor (+60 ms) SOA are broadly similar to those observed in the 0 ms SOA condition. The predicted SI effect would be expected to peak at about 150 ms after target onset in this condition, and there is some evidence of a selective dip in saccade frequency with remote ipsilateral distractors on the 45° axis only at around this time. A small RDE (of 7.8 ms) was observed for this condition, although a similar small RDE was observed with contralateral distractors (see Table 2), and in this case, there is no evidence of a notched dip in the distributions. What is apparent is a generalized reduction in saccades at about 150 ms, which is also shown in the distribution for P1 with contralateral and ipsilateral distractors (the condition that also produced an RDE). The distribution for participant P1 shows some signs of a dip at about 150 ms with remote ipsilateral distractors (45° axis) only. Participant P2 also shows some evidence of dips, with ipsilateral distractors most apparent with distractors 0° and 45° from the target axis (a small RDE effect of between 5 and 10 ms was observed for these conditions). The timing of Dip_max can be seen to be about 80 to 85 ms after the onset of an ipsilateral or contralateral distractor, which is consistent with the expected time course of SI. 

Discussion

In Experiment 2, distractors appearing before, or simultaneously with, the saccade target increased saccade latency (RDE) without clear evidence of notched dips in the underlying latency distributions. The effects of SI are known to be time locked to distractor onset, starting around 60 ms and peaking around 90 ms, after distractor onset. The absence of a notched dip for the early-onset (−60 ms SOA) condition is therefore not unexpected, as SI would be expected to peak at about 30 ms after target onset, an interval too early to influence the target selection if it is a short-lasting effect. A robust RDE effect was observed, however, along with an increase in positive skew in the distractor-related latency distributions similar to that observed in Experiment 1 with simultaneous onsets. A notched dip was predicted in the delayed distractor onset (+60 ms) condition, peaking at about 150 ms after target onset, which should correspond with the peak in the baseline latency distribution for this condition. A decrease in saccade frequency can be observed (at about 140 ms), consistent with the expected time course of SI. The clearest evidence of a notched dip is apparent with ipsilateral distractors presented 45° from the target axis. This finding may appear to be consistent with that of Buonocore and McIntosh (2012), who have recently reported that SI can potentially be stronger with ipsilateral rather than contralateral distractors, although in their study, ipsilateral distractors appeared at the target location. The possibility that SI is stronger with ipsilateral distractors contrasts with the spatial modulation of the RDE. 

Experiment 2 has shown that an RDE effect can be observed, which is reflected in the underlying distractor-related latency distributions as an increase in skew, with some evidence of a short-lasting inhibitory SI effect most apparent with delayed distractor onsets in the target hemifield. The underlying single-target (baseline) latency distributions appear different in the early-onset (−60 ms SOA) condition compared with those from the simultaneous (0 ms) and delayed (+60 ms) SOA conditions (which are more similar). The distributions for the early-onset condition show a shift toward short-latency saccades and are less skewed than the simultaneous and delayed-onset conditions. This highlights how the underlying latency distributions and central tendencies can change across conditions that otherwise appear similar. The effects of presenting distractors either before, or after, the target may differ due to a range of low-level bottom-up and high-level cognitive factors that are not equivalent. For example, a distractor appearing before the target may reduce latency due to warning-signal effects (Reuter-Lorenz, Oonk, Barnes, & Hughes, 1995; Ross & Ross, 1980, 1981) as well as produce inhibitory effects that that may be masked by the warning-signal facilitation effect. Distractors appearing simultaneously or after the target may not contribute to such alerting effects, or in the case of simultaneous onsets, they may be reduced. Thus, the shape of the underlying latency distributions will reflect a range of facilitatory and inhibitory processes as well as cognitive factors influenced by the spatial and temporal predictability of targets and distractors that may all influence the decision of when and where to initiate a saccade. The bin-by-bin subtraction technique shows a late Dip_max in the −60 ms SOA condition, which is due to the leftward shift of the distractor distribution rather than a late inhibitory effect. 

There are some minor differences between this experiment and other studies reporting an SI effect. One of these is the use of briefly presented (33–100 ms) distractor flashes (Bompas & Sumner, 2009, 2011; Buonocore & McIntosh, 2008; Reingold & Stampe, 2002) rather than distractor stimuli remaining on until the end of the trial as here. Under such conditions, the offset of the distractor could potentially modulate the inhibitory effect, thus producing a more transitory SI effect, as revealed by the selective dip in saccade frequency. This seems an unlikely explanation for the differences in findings, as Edelman and Xu (2009) reported an SI effect with a distractor presented until the saccade was initiated. Studies have also reported an SI effect with small distractor onsets, presented at comparable eccentricities, comparable to those used here (Bompas & Sumner, 2011; Buonocore & McIntosh, 2008; Edelman & Xu, 2009). A further possibility is that the use of a fixed target direction in our experiments is a factor in reducing the likelihood of observing SI. This seems unlikely as Buonocore and McIntosh (2008) used a fixed target direction, and in Edelman and Xu's (2009) study, target direction was prespecified by a central arrow cue (although the effects on latency and the RDE were not reported). One other difference between our experiment and those reporting an SI effect is the use of a blocked design for the target-distractor SOAs. Therefore, in a further attempt to observe SI in an RDE experiment, we included a range of SOAs in a mixed design with distractors presented at a range of eccentricities. 

Experiment 3: The RDE and SI mixed-target distractor delays (SOAs)

Experiment 3 examined SI and the RDE using a range of SOAs selected with the optimization timing similar to Edelman and Xu's (2009) in a mixed SOA design. Edelman and Xu first assessed the baseline latency for each participant and then selected a small range of SOAs that resulted in the distractor appearing approximately 90 ms before saccade onset. To reduce the number of trials required in Experiment 3, a single-target eccentricity (4°) was used with contralateral distractors at a range of distractor eccentricities (0°, 2°, and 4°) that enabled direct comparisons with some of the conditions from Experiments 1 and 2 reported above. The present experiment again used a visual saccade target, in order to be more similar to the studies reported above, and also to Edelman and Xu's (2009) experiment 2b that reported an SI effect with visually guided saccades. Participants first performed a block of 34 single-target trials (fixed direction and eccentricity) to assess their baseline latency, although we note there are limitations with this approach, as latency will most likely be reduced over the large number of trials performed in the main experiment. 

Methods

Apparatus, procedure, and analysis

The apparatus, procedure, and analysis are as described for Experiment 1. 

Participants

Four participants (three naïve postgraduate volunteers and one of the authors) participated in each of the conditions. The ages ranged from 20 to 48 years (mean = 28 years), and all had normal or corrected-to-normal vision. 

Materials and design

The median latency for each participant made to single targets in the initial single-target block varied from ∼150 to 165 ms. From these, three distractor-delays (SOAs) of +30 ms, +60 ms, and +90 ms were chosen that would result in average distractor saccade onset delay times of ∼125, 95, and 65 ms (average = 95 ms) that should have been approximately ideal for observing an SI effect. In the main experiment, each trial started with a small white fixation cross (0.5°) presented for a variable fixation interval (1000–1200 ms, 10-ms steps) on a black background. Following this, a small target (white filled triangle 0.5°) appeared at 4°, in a fixed direction (alternated every 48 trials), along a horizontal axis that was level with fixation. On the majority of trials (3/4), a distractor (white filled circle 0.5°) appeared at an eccentricity of 0°, 2°, or 4° in the contralateral visual field on a horizontal axis. Distractors appeared after delays (SOAs) of 30, 60, or 90 ms following target onset, randomly interleaved within a block of (96) trials. On single-target trials, an invisible distractor (as background) was presented after a similar SOA period, which ensured single and distractor trials had comparable timing. Each participant completed 20 blocks of (96) trials, producing a total of 1,920 trials; of these, there were 160 trials for each distractor eccentricity and SOA and 120 single-target trials for each invisible distractor SOA. 

Results

The mean (of median) of the latency RDE effect is shown in Figure 7. It can be seen that distractors at fixation (0°), appearing +30 ms after target onset, produced a large RDE effect of some 32.9 ms. A large effect of SOA is also apparent on the magnitude of the RDE for distractors at fixation, which decreased to some 15.8 ms and 13.0 ms with longer SOAs of +60 ms and +90 ms. By contrast, SOA had a smaller impact on the magnitude of the RDE, with distractors at eccentricities of 2°, which decreased from 18.1 ms to ∼12.3 ms and 9.0 ms across SOA. Similarly, the RDE with distractors at 4° decreased only slightly with increasing SOA (from 9.0 ms to a minimum of 6.5 ms). A two-factor repeated-measures ANOVA revealed a nonsignificant effect of distractor eccentricity, F(2, 6) = 3.88, p > 0.05, ηp2 = 0.56; a significant effect of SOA, F(2, 6) = 9.15, p < 0.05, ηp2 = 0.75; and a significant two-way interaction, F(4, 12) = 5.09, p < 0.05, ηp2 = 0.63. The interaction effect can be attributed to the much greater effect of SOA with distractors at fixation than with contralateral distractors at 2° and 4°. Post hoc t tests confirmed that the RDE was significantly greater with distractors at fixation (D0°) for the +30 ms SOA than for the +90 ms SOA, t(3) = 3.86, p < 0.05, and was approaching significance for the intermediate +60 ms SOA, t(3) = 2.96, p = 0.06. The RDE with distractors at fixation in the +30 ms SOA condition was also significantly greater than with distractors at 2° and 4° across all SOAs (all p values <0.01). None of the other pairwise comparisons were significant. 

Figure 7

 

Mean of median RDE (ms) in Experiment 3, collapsed across participants for each distractor eccentricity (0°, 2°, and 4°) and SOA (filled circle +30 ms, square +60 ms, and diamond +90 ms).

Saccadic inhibition

The combined latency distributions for all four participants (collapsed across bins) for each distractor eccentricity and SOA are shown in Figure 8a (upper row). The Dip plots produced by the bin-by-bin subtraction technique are shown below. Figure 8b shows the latency distributions from each individual participant (P1–P4), which reveal a broadly consistent pattern of distractor-related modulation to that observed in the overall distributions. 

Figure 8

 

(a) Latency distributions obtained with data from all participants collapsed across bins (upper row) from Experiment 3 under conditions in which a distractor appeared at fixation (left) or in the contralateral visual field at eccentricities of 2° (center) or 4° (right). Each plot shows the single-target baseline (shaded histogram) along with the distractor-related distribution for each SOA → ← (filled circle = +30 ms, square = +60 ms, diamond = +90 ms) plotted in relation to target onset. The plots obtained from the bin-by-bin subtraction method (distractor-related distribution – baseline) plotted with respect to target onset are shown below. (b) Latency distributions from the individual participants (P1–P4), which reveal similar patterns of distractor-related modulation (same conventions as above).

Distractors at fixation (D0 left plots)

A modulation in the shape of the distractor-related distributions was produced starting at about 110 ms that was strongly influenced by SOA. For the +30 ms SOA, the decrease in saccade frequency starts at ∼90 ms, whereas the decrease in saccades in the +60 and +90 ms SOAs starts later at ∼140 ms and ∼180 ms (with respect to target onset). The proportion of saccades that can be inhibited, in relation to the baseline distribution, is therefore greatest at the +30 ms SOA. In this condition, a later rebound is apparent with an increase in longer-latency saccades (>200 ms). A similar but reduced effect is apparent at the +60 ms SOA, whereas the later +90 ms SOA has a smaller effect. 

Contralateral distractors at 2° (D2 center plots)

The decrease in saccades for the +30 ms SOA starts at ∼110 ms, resulting in an increase in positive skew for the distractor-related distribution and an increase in saccades with a latency of >180 ms. The decrease in saccade frequency for the +60 ms and +90 ms SOAs start at about 140 ms and 170 ms (with respect to target onset). 

Contralateral distractors at 4° (D4 right plots)

The decrease in saccades varies across SOA starting at ∼110 ms, 140 ms, and 180 ms, respectively (with respect to target onset). The largest modulation is again apparent for the earlier +30 ms SOA with similar but smaller effects for the +60 ms and +90 ms SOAs comparable to those observed with a distractor at 2°. 

Dip_max (Decrease in saccade frequency)

The mean measure of dip magnitude (Dip_max), as shown in Figure 9a, varied with distractor eccentricity and SOA, with the greatest decrease in saccades occurring for distractors presented at the shortest SOA. An effect of distractor eccentricity is also apparent, with the magnitude of SI increasing as distractors approach fixation. A two-factor repeated-measures ANOVA confirmed a significant effect of eccentricity, F(2, 6) = 7.57, p < 0.05, ηp2 = 0.72, and SOA, F(2, 6) = 6.35, p < 0.05, ηp2 = 0.68, along with a two-way interaction effect, F(4, 12) = 3.32, p < 0.05, ηp2 = 0.52. Post hoc t tests showed that the distractor at fixation produced a larger decrease in saccade frequency for the +30 ms SOA than +60, t(3) = −3.72, p < 0.05, and +90 ms, t(3) = −5.62, p < 0.01, SOAs. For distractors at 2° and 4°, the increase in Dip_max for the +30 ms SOA did not differ significantly from that observed at the +60 ms and +90 ms delays. Distractors at fixation, at the 30 ms SOA, produced a greater decrease in saccades compared with that observed with distractors at 4°, t(3) = −4.28, p < 0.05, but not with distractors at 2°, t(3) < 1. Thus, the decrease in saccade frequency (Dip_max) is influenced by SOA and by distractor eccentricity, but the effect is greatest with distractors at fixation when presented +30 ms after the target. Distractor eccentricity had little effect on Dip_max when presented at the longer SOAs. 

Figure 9

 

(a) Mean measures of Dip_max (change in proportion of saccades) obtained from the bin-by-bin subtraction using the data from each participant. Separate lines depict values for each SOA (circles = +30 ms, squares +60 ms, diamonds = 90 ms) for each distractor eccentricity as shown on the abscissa. (b) Mean time to reach maximum “dip” (Dip_T) in milliseconds from distractor onset (same conventions as for [a]).

Dip_T (Time to reach Dip_max)

The mean time to reach maximum inhibition (Dip_T, shown in Figure 9b) also reveals an influence of both distractor eccentricity and SOA. A two-factor repeated-measures ANOVA confirmed a significant effect of distractor eccentricity, F(2, 6) = 7.37, p < 0.05, ηp2 = 0.71, SOA, F(2, 6) = 9.63, p < 0.01, ηp2 = 0.76, and no interaction effect (F < 1). The average values of Dip_T collapsed across distractor eccentricity, in relation to distractor onset, are 123.7 ms, 104.0 ms, and 83.0 ms for the +30, +60, and +90 ms SOAs, respectively. Similarly, the average Dip_T times collapsed across SOA are D0 = 115.7 ms, D2 = 101.3 ms, and D4 = 93.3 ms. Thus, the maximum value of Dip_max can be observed approximately 90 ms after distractor onset with distractors at 4° that appear at SOAs of +60 to 90 ms after target onset. Distractors closer to fixation and those appearing at shorter SOAs produce a later maximum decrease in saccade frequency (Dip_max). 

Discussion

In Experiment 3, distractors appeared at three delays of 30, 60, or 90 ms after target onset, at three eccentricities (0°, 2°, or 4°). A fixed target eccentricity (4°) was used. The magnitude of the RDE depended on both distractor eccentricity and SOA. Distractors at fixation produced a large RDE (∼32 ms) when presented shortly after (+30 ms) target onset, but this effect dropped substantially with longer distractor delays. Contralateral distractors at 2° and 4° produced a smaller RDE (∼12.5–6.5 ms) that also showed a smaller effect of SOA. The large RDE observed with distractors at fixation is comparable with that observed from studies using simultaneous target-distractor onsets (Experiments 1 and 2 here), but the modulation with SOA has not previously been observed. This highlights the need to investigate distractor effects in terms of both the spatial and temporal domains (Bompas & Sumner, 2009, 2011). One possible account of the greater effect of foveal distractors when presented close to the time of target onset is in relation to fixation activity. The activity of fixation neurons would be expected to decrease following the offset of fixation, which occurs simultaneous with target onset; if a foveal distractor then appears a short time afterward, fixation-related activity may be swiftly enhanced. At longer delays, the level of fixation activity may have decreased further because of the effect of fixation offset and lateral inhibition from target-related activity. In this situation, a delayed foveal onset may not produce a rapid rise in fixation-related activity, resulting in a reduced inhibitory effect. A delay of 60 to 90 ms appears sufficient for fixation activity to be suppressed so that a later distractor onset exerts a smaller influence on saccade triggering. 

The latency distributions and subtraction dip plots in Figure 8a reveal a modulation in saccade frequency that is strongest for the short +30-ms distractor delay. Distractors at fixation start to exert an influence some 90 ms after distractor onset, producing the maximum dip at about 135 ms. A later rebound effect is apparent, which can be attributed to the production of longer latency (>200 ms) saccades. The decrease in saccade frequency at the later SOAs produces a much smaller dip, as there are now fewer saccades that can be inhibited by the distractor. Distractors at 2° and 4° at the short +30-ms delay show a similar modulatory effect starting at about 100 ms, resulting in an increase in skew. The influence of distractors appearing at the longer SOAs produces smaller changes to the shape of the distributions because there are fewer saccades that can be inhibited. How these distractors influence the shape of the underlying latency distribution will depend on the characteristics of the underlying baseline distributions for individual participants and the time at which the distractor appears. 

The baseline distribution for the four participants in Experiment 3 had a mean latency of 163 ms and a standard deviation of 28.8. SI is thought to start at about 60 to 70 ms after a visual onset and to continue for about 60 to 70 ms following that (Reingold & Stampe, 2002). A distractor onset 30 ms after the target can potentially influence a large proportion of saccades that would otherwise be initiated with a latency of ∼100 to 170 ms. Far fewer saccades would be influenced by the later onsets, reflecting the standard deviation of the distribution and also because they may occur during the saccade “dead time” period (estimated as being 60–70 ms before saccade initiation, during which time a visual change will not influence saccade generation; Ludwig, Mildinhall, & Gilchrist, 2007). The influence the distractor has on overall mean latency reflects the way in which the distractor-related distribution changes in relation to the characteristics of the baseline. Here, the 30-ms delay appears optimal, and inhibition decreases with increasing distractor eccentricity. We return to the relationship between SI and the RDE in the General discussion. 

General discussion

This study aimed to investigate the relationship between SI and the RDE and specifically whether there was evidence to support the idea that SI underpins the RDE across a range of distractor locations. The first experiment used conditions similar to those in the original RDE studies with simultaneous target and distractor onsets and revealed a reliable RDE that, as expected, increased as distractors approached fixation. The bin-by-bin subtraction method, which is often used to produce the measure of maximum SI (Dip_max), showed that this measure, like the RDE, varied systematically with distractor eccentricity. However, this method is not a reliable diagnostic measure of a selective short-lasting SI effect alone as it can arise due to other changes to the shape of the distractor-related distribution. An examination of the latency distributions for Experiment 1 showed no evidence of selective inhibition, as would be expected with simultaneous target-distractor onsets, but rather an increase in positive skew. If a weak SI was influencing saccades starting around 60 ms, that resolved by 110 ms after distractor onset, it would influence only a small proportion of saccades in the leftmost tail of the distributions, and unless it operates over a longer time scale (during the later recovery period; Bompas & Sumner, 2011), it would not be expected to account for all of the increase in longer-latency responses apparent in these distributions. We think that this provides support for the view that a transitory inhibitory effect (SI) does not account for all of the observed change in mean latency (RDE). 

The second experiment further examined the RDE and SI with distractors at locations in both the ipsilateral and contralateral hemifields across a range of SOAs. Presenting a remote distractor 60 ms before the target onset produced an RDE latency increase, although this early onset would be regarded as being too soon to produce an SI effect. The underlying latency distributions showed an increase in positive skew with contralateral distractors and a facilitation effect with early-onset ipsilateral distractors presented close to the saccade target (Figure 6). The onset of remote distractors simultaneous with the target tended to produce an increase in positive skew similar to that observed in Experiment 1 along with a similar RDE. In the +60 ms SOA condition, there was some evidence of a dip about 90 ms after distractor onset, consistent with SI that appeared strongest with ipsilateral distractors (Buonocore & McIntosh, 2012). These findings are suggestive that there may be more than one inhibitory process operating to produce SI and the overall change in mean latency that is labeled as the RDE. 

Experiment 3 included a range of target-distractor SOAs selected as being optimal for observing SI at an individual participant level (Edelman & Xu, 2009). As in their study, the SOAs were mixed within a block of trials, and a small range of distractor eccentricities was used. A reliable RDE that was modulated by distractor eccentricity was again observed, but this depended on SOA. A large RDE was observed with distractors at fixation when presented a short time after the target. The RDE reduced substantially for the longer distractor delays. The effect of contralateral distractors at 2° and 4° eccentricities was also greatest at the shorter SOA, although the influence of SOA was less marked for the more eccentric distractors. We suggest that the greater effect for distractors at fixation presented at the shorter SOA is consistent with the notion of increased fixation activity that inhibits saccade generation. When a distractor appears soon after fixation offset, it may enhance fixation activity, whereas at the longer delays, fixation activity may be more strongly inhibited by rising target-related activity, thus reducing the influence of a foveal onset. 

The modulation of the RDE by distractor eccentricity and location reported here is similar to our original report (Findlay & Walker, 1999) and shows that the RDE operates in both the ipsilateral and contralateral visual fields with remote distractors as far as 16° from fixation (Experiment 1 here). This supports the view that the inhibitory influences produced by a remote visual onset are not restricted to a central fixation zone of less than 10°. We suggest that SI and the RDE are partly dissociable and that each may have a separate neurophysiological basis. A study of Dorris et al. (2007) showed how saccade-related neurons in the SC were briefly inhibited by a remote distractor, a finding that appears consistent with both the RDE and SI effects. However, in the study by Dorris et al. (2007), the distractor appeared 100 ms before the target (a condition unlikely to produce an RDE), and the stimulus onset suppressed preparatory activity, related to the use of completely predictable target timing and location, rather than target-related activity. This inhibitory effect of preparatory saccade-related activity in the SC by a visual onset could potentially be related to SI (Edelman & Xu, 2009). The RDE is attributed to similar competitive interaction effects operating in the SC, but in this case, it may be a more sustained inhibitory effect. The exact nature of these lateral interactions is not fully understood, and they may operate between populations of neurons encoding the distractor and saccade target (Trappenberg et al., 2001) or between fixation-related and saccade-related activity (Casteau & Vitu, 2012; Findlay & Walker, 1999; Walker et al., 1997). There is, therefore, a possibility that separate, but interacting, inhibitory influences account for the SI and the RDE and that both of these may be observed under certain conditions, such as when a highly predictable target paradigm is used or when the target activity is weak, such as in the delayed and memory-guided paradigms used by Edelman and Xu (2009). 

The examination of changes to latency distributions and the bin-by-bin subtraction technique used to examine SI show that inhibition can start at about 60 to 70 ms after a visual onset, which is consistent with physiological evidence of the time taken for collicular neurons to respond to a visual stimulus (see Reingold & Stampe, 2002). The relationship between SI and the RDE depends on the definitions of these effects. We regard SI to be a short-lasting effect, starting at about 60 ms after a visual onset and lasting for about 60 ms. The RDE, as defined as an increase in latency, can reflect this inhibitory effect along with other competitive inhibitory processes that operate over a longer time scale that depend on both stimulus timing and eccentricity. A visual onset could potentially increase activity in brainstem omnipause neurons, resulting in a short-lasting inhibitory effect. The potential advantage of this delay is it allows time for neural activity associated with potential targets to evolve for the selection of the desired saccade target. A visual onset may also result in increased activity of rostral collicular neurons that also serve to inhibit saccade generation. The strength of this activity will depend on the eccentricity of the distractor and target (Casteau & Vitu, 2012; Walker et al., 1997) rather than on the target-distractor spatial separation (Bompas & Sumner, 2011; Trappenberg et al., 2001) and on the level of fixation activity at the time the distractor appears. The competition between the target-related and fixation-related activity is further influenced over time by sustained endogenous inputs, making this a highly nonlinear process, as noted by Bompas and Sumner (2011). However, in their model, the competition is between the target and distractor, with saccade triggering depending on the rise of target-related activity. 

Conclusions

The argument that SI (alone) underpins the RDE is reminiscent of the much earlier debate regarding the relationship between express saccades and the fixation offset (or gap) effect (Kingstone & Klein, 1993; Reuter-Lorenz et al., 1995). Express saccades, as noted in humans by Fischer and Ramsperger (1984, 1986), refer to a population of short-latency saccades occurring 80 to 120 ms observed in bimodal or multimodal latency distributions, whereas the fixation-offset (or gap) effect refers to the decrease in mean latency observed following the prior offset of central fixation. The offset of fixation about 200 ms before target onset is the ideal situation for producing short-latency express saccades and the maximum gap effect reduction in mean latency. However, not all studies observed clear evidence of bimodality (or multimodality) in the latency distributions (Kingstone & Klein, 1993; Reuter-Lorenz et al., 1995; Reuter-Lorenz, Hughes, & Fendrich, 1991; Wenban-Smith & Findlay, 1991), and the diagnostic feature of an express saccade remains unclear (Kingstone & Klein, 1993; Reuter-Lorenz et al., 1995). Similarly, the defining feature of SI is the transitory inhibitory effect that knocks out saccades that would be expected 60 to 120 ms after distractor onset, whereas the RDE is defined in terms of an increase in mean latency. Changes to central tendency will naturally be reflected in observable changes to the distractor-related latency distributions, but these do not have to be attributed to a short-lasting inhibitory effect alone. The conditions typically used to demonstrate SI (delayed distractor onsets) are not optimal for maximizing the RDE, and those used to produce the maximum RDE (near simultaneous onsets) are not ideal for observing a selective dip in the latency distributions. We agree with Buonocore and McIntosh (2012) that an examination of latency distributions along with changes to central tendency is useful as it can reveal more about the time course of these inhibitory processes. It may also prove beneficial to examine how changes to the shape and variance of latency distributions can influence the measures of SI. 

Acknowledgments

We would like to thank Casimir Ludwig and Colin Davis for insightful discussions on the studies reported here and particularly for their suggestions on the analysis techniques used to reveal SI. We thank also Frouke Hermens for providing feedback on an earlier draft of the manuscript and Hannah Harvey for her assistance in testing participants. We are also grateful for the helpful comments of two anonymous referees. 

Commercial relationships: none. 

Corresponding author: Robin Walker. 

Email: [e-mail]. 

Address: Department of Psychology, Royal Holloway, University of London, Egham, Surrey, United Kingdom.