Inhibition of saccades elicits attentional suppression

Abstract

Abstract

Abstract:

Abstract  Visuospatial attention has been shown to have a central role in planning and generation of saccades but what role, if any, it plays in inhibition of saccades remains unclear. In this study, we used an oculomotor delayed match- or nonmatch-to-sample task in which a cued location has to be encoded and memorized for one of two very different goals—to plan a saccade to it or to avoid making a saccade to it. We measured the spatial allocation of attention during the delay and found that while marking a location as a future saccade target resulted in an attentional benefit at that location, marking it as forbidden to saccades led to an attentional cost. Additionally, saccade trajectories were found to deviate away more from the “don't look” location than from a saccade-irrelevant distractor confirming greater inhibition of an actively forbidden location in oculomotor programming. Our finding that attention is suppressed at locations forbidden to saccades confirms and complements the claim of a selective and obligatory coupling between saccades and attention—saccades at the memorized location could neither be planned nor suppressed independent of a corresponding effect on attentional performance.

Introduction

Attentional processes that selectively deliver the spatial coordinates of the intended target object are one of the essential elements of goal-directedness of saccadic eye movements. A long line of psychophysical studies has shown a central role for visuospatial attention in planning and generation of saccades (Deubel & Schneider, 1996; Hoffman & Subramaniam, 1995; Jonikaitis & Deubel, 2011; Kowler, Anderson, Dosher, & Blaser, 1995; Rizzolatti, Riggio, Dascola, & Umiltá, 1987). A number of neurophysiological studies in nonhuman primates have confirmed a causal role of oculomotor circuits in the deployment of visuospatial attention (Moore & Fallah, 2001, 2004; see Awh, Armstrong & Moore, 2006 for a review). While subsequent studies have identified single neurons intertwined within these circuits that are specialized for either visual selection or oculomotor control; they also found a “visuomotor” class of neurons that had a dual role in both (Ignashchenkova, Dicke, Haarmeier, & Thier, 2004; Thompson, Biscoe, & Sato, 2005). 

However, what role, if any, visuospatial attention plays in saccade inhibition remains unclear. This question has assumed increased importance for a number of reasons. First, saccade inhibition has emerged as an important model for studying response inhibition, cognitive control, and cortical executive processes in general (Hutton, 2008; Munoz & Everling, 2004). Second, deficits in the inhibition of saccades have been shown to be associated with a whole range of neurological and psychiatric disorders including schizophrenia, attention deficit disorder, closed head injuries, and frontal lesions (Clementz, McDowell, & Zisook, 1994; Everling & Fischer, 1998; Mostofsky, Lasker, Cutting, Denckla, & Zee, 2001). While clinical studies have extensively characterized the use of errors in saccade inhibition as a diagnostic measure (Everling & Fischer, 1998), and neurophysiological work has elucidated the underlying neural processes (Munoz & Everling, 2004), the exact nature of cognitive dysfunction in patient populations that show saccade inhibition deficits has remained unclear. Especially, it remains unclear whether spatial attention, which is otherwise known to be inextricably linked to saccade planning, plays any role in saccade inhibition. 

In this study, we examined the consequences for visuospatial attention, if oculomotor programs to a location are being actively inhibited. Using the oculomotor delayed match- and nonmatch-to-sample task (Figure 1), we asked if the inverse of the attention-saccade coupling holds true—given that marking a location as a saccade target leads to attentional enhancement at that location, will a command to actively inhibit saccades to a specific location lead to attentional suppression at that location? Participants were shown a sample cue and were instructed to remember its location. After a delay, two targets appeared, one at the previously memorized and another at a random location. In the match trials, participants had to make a saccade to the memorized location and ignore the other while in the nonmatch trials they had to ignore the memorized location and make a saccade to the nonmemorized target. The memorized location in the two cases differed only by the saccadic target it was encoded for—to plan a saccade to it or to avoid making a saccade to it when the target reappeared. We tested how planning or avoiding saccades at a location would affect visual discrimination performance, a frequently used diagnostic measure of visuospatial attention (Carrasco, 2011; Deubel & Schneider, 1996). 

Figure 1

 

Schematic depiction of the sequence of stimuli and experimental conditions. See Stimuli and tasks under Methods for details.

The nonmatch task as described here presents certain unique advantages over other paradigms used to study saccade suppression. Like the antisaccade task, it involves active suppression of a specific eye movement as opposed to global suppression of all the saccades as in countermanding and go/no-go tasks. But unlike the antisaccade task, which involves ruling out a location as a saccade target and simultaneously planning the opposite saccade (Munoz & Everling, 2004), the nonmatch task only involves the ruling out of a location since the only information available through the delay is where the subject must not look and the correct saccade cannot be planned until after the delay when the two potential targets appear. Hence, the delay period in the nonmatch task uniquely allows us to measure spatial allocation of attention during sustained suppression of saccades without being confounded by the processes involved in making the oculomotor response. 

How is the nonmatch task solved and what role does attention play in it? In a remarkable neurophysiological study using this task, Hasegawa, Peterson, and Goldberg (2004) identified a group of neurons in the frontal eye fields and the caudal prefrontal cortex that show enhanced activity when a stimulus appears at a specific location in the visual field and when saccades to that location are forbidden (the memorized location in the nonmatch task). Their results show a population of neurons that represents a command—“don't look at this particular location.” Their findings suggest that the strategy used to identify the nonmatch target is by actively suppressing a response to the memorized location so that the desired saccade could be made to the nonmatching location. But whether a saccade to a location is suppressed by passively ignoring it or by actively suppressing spatial attention at that location remains uncertain. A different line of reasoning would actually predict attentional enhancement at the suppressed location. The attention-based rehearsal hypothesis (Awh, Vogel, & Oh, 2006; Smyth & Scholey, 1994) suggests that attention is enhanced at the locations held in spatial working memory and that such attention-based rehearsal plays an essential role in maintenance of information in visual working memory. In the nonmatch task, even though a saccadic response to the sample location needs to be suppressed, it remains task relevant and must be maintained in working memory through the delay. Thus the attention-based rehearsal hypothesis would predict that the nonmatch task is solved by keeping sustained covert attention at the sample location to help sustain it in working memory through the delay. 

Finally, several recent neuroimaging studies (Curtis, 2006; Curtis, Rao, & D'Esposito, 2004; Curtis, Sun, Miller, & D'Esposito, 2005) have shown that match and nonmatch tasks lead to distinct patterns of sustained activation during their respective delay periods. They argue that these distinct activations represent neural correlates of maintenance of a motor intention (in the match task) as opposed to that of sustained spatial attention (in the nonmatch task). However, this equivalency of delay period activity in match task as motor intention and in nonmatch task as enhanced attention is based on untested theoretical assumptions regarding the role of visuospatial attention in this paradigm. By measuring the spatial allocation of attention in these tasks, our study puts these assumptions to a direct test and will thus help clarify what the particular brain activation patterns seen in match versus nonmatch tasks represent. 

Methods

Participants

Ten participants (five males and five females, aged between 18 and 28, vision normal or corrected-to-normal) took part in the experiments. All except for two of the authors (SD, DJ) were naive as to the purpose of the study. The experiments were undertaken with the understanding and consent of each subject and were conducted in accordance with the Declaration of Helsinki. 

Experimental setup

Observers sat in a silent and dimly lit room. Dominant eye gaze position was recorded using an EyeLink 1000 Desktop Mount (SR Research, Osgoode, Ontario, Canada) with a spatial resolution finer than 0.25° and a sampling rate of 1000 Hz. Head movements were minimized using a chin rest. Stimuli were presented on a 22-inch Lacie Electron 22 Blue screen (spatial resolution of 1280 × 1024, 85 Hz vertical refresh rate; Lacie SA, Paris, France) at a viewing distance of 70 cm. Stimulus presentation and response collection was controlled by a Dell computer running Windows XP and implemented in Matlab (MathWorks, Natick, Massachusetts, USA) using the Psychophysics (Brainard, 1997; Pelli, 1997) and Eyelink (Cornelissen, Peters, & Palmer, 2002) toolboxes. 

Stimuli and tasks

Figure 1 provides a schematic illustration of the sequence of stimuli and the experimental conditions. Each trial comprised six phases: (1) Fixation: The trial started with the presentation of a fixation point (a red circle of radius 0.3°) in the center of the display. Participants were instructed to fixate all through the trial except when making the saccade response. Mean fixation duration was 1000 ms selected from a Gaussian distribution with a standard deviation (SD) of 300 ms. (2) Sample: After a fixation period, a spatial cue (a red circle of radius 0.8°) appeared for 1000 ms at a randomly selected location, 5° away from the fixation. Participants were asked to memorize the location of the circle. (3) Delay: After the cue disappeared, participants continued to fixate for 200, 500, 1000 or 2500 ms (randomly selected on each trial). (4) Probe presentation: After this variable delay, two rings were presented for two video frames (∼24 ms): a Landolt ring (the discrimination probe with opening either to the left or right) and a complete ring. One of the rings appeared at the memorized location and the other at a location chosen randomly from one of the six possible nonmemorized locations at 36°, 72°, or 108° (radial directions with the line between central fixation and the memorized location serving as the 0° polar axis) clockwise or counterclockwise from the memorized location. Probability of the discrimination probe appearing at the memorized or nonmemorized position was equal (50%). Subsequently, both rings were masked by a white noise mask (each pixel's gray value taken from a Gaussian distribution with mean of 0.5 (where 0 is black and 1 is white) and SD = 0.5; duration ∼24 ms). (5) Saccade response: 50 ms after the mask disappeared, two response objects (filled blue circles) appeared for 500 ms at the same locations as the two rings presented before (one at the memorized location and another at the nonmemorized location). In the match task, participants had to make a speeded saccade to the target at the memorized location, while in the nonmatch task, they had to make a saccade to the target at the nonmemorized location. A horizontal red rectangle was shown at center as feedback for incorrect eye movements (saccades not directed to the instructed target). (6) Probe response: Upon the disappearance of the test stimuli, subjects had to indicate by pressing a button, in a nonspeeded manner, whether the left or right orientation of the Landolt ring was shown. Performance on the probe response was indicated to the participants by auditory feedback—a single beep for correct and a double beep for incorrect probe response. The next trial started immediately after the button press. 

The experiment comprised five consecutive blocks of the match and five consecutive blocks of the nonmatch task, each block consisting of 50 trials. The task order was determined randomly for each subject. The experiment was preceded by a single practice block of 50 trials. 

Measuring attention

Visual attention, operationally defined as prioritized processing of certain pieces of visual information can be measured experimentally by accuracy differences in identifying visual targets at the attended locations (Deubel & Schneider, 1996; Yeshurun & Carrasco, 1999). In general, the accuracy of reporting visual attributes of objects at the attended locations is found to be higher as compared to locations that are not attended. We chose to use relative accuracy in discriminating the orientation of a Landolt ring (Yeshurun & Carrasco, 1999) presented during the delay period as a measure of spatial attention. 

Our experiment was designed only to find how the discrimination performance at a specific location, memorized either as a valid or forbidden saccade target, differs from the attentional resources allocated to the rest of the display. For this we measured the allocation of attention at iso-eccentric locations spread around the display (at 0°, 36°, 72°, and 108° on either side of the memorized location). We took attentional performance at the rest of the measured locations within each task to be the baseline by which to judge the performance at the memorized location in that task. 

We did not seek to establish a neutral baseline performance using a separate attention task performed without any saccade planning, since visuospatial attention has been shown to be influenced in various ways when the task demands change (e.g., Folk, Remington, & Johnston, 1992; Lupiáñez, Milán, Tornay, Madrid, & Tudela, 1997; Lupiáñez & Milliken, 1999). Hence, spatial allocation of attention without any saccade planning would not have been considered directly comparable to the same during saccade planning. Probe performance at the nonmemorized locations, within each task, is the only control that keeps every other factor constant (except the fact that these locations were nonmemorized) and can, thus, serve as a meaningful comparison for the performance at memorized locations to which saccades have to be made or suppressed. 

Data analysis

Eye movements were recorded during the experiment and saccades were detected offline using an algorithm based on two-dimensional eye velocity (Engbert & Kliegl, 2003) computed from subsequent samples in the eye position series. The threshold for peak velocity was 3.0 SD and the minimum saccade duration was 20 ms (Engbert & Kliegl, 2003). We accepted only trials where the saccade latency was above 100 ms and below 500 ms. We analyzed only trials in which the saccade landed within a 2.5° radius around its goal. In total we accepted 82% trials: 4% of all trials were rejected due to blinks, 7% due to inaccurate saccades, and 7% due to participants looking at the incorrect target. 

The performance in the probe discrimination task was expressed by the percentage of correct Landolt ring discriminations. Since there were two response alternatives, chance probe discrimination level was at 50%. For the analysis of probe discrimination performance, we pooled all saccade directions together. Statistical analyses included repeated-measure analyses of variance (ANOVA) and two-tailed paired t-tests. 

We also analyzed saccade trajectory deviations. We defined the distractor as an object to which participants were not supposed to look at: In the match task the distractor was presented at the saccade-irrelevant non-memorized location, whereas in the nonmatch task the distractor was the object at the saccade-relevant memorized location. 

Saccadic data points were rotated and pooled together in one direction so that all the saccades started at the center of the display and were directed vertically (90° upward) while the relationship of the distractor being presented clockwise or counterclockwise to the saccade target was preserved. We normalized saccade duration to the interval from 0 (saccade start) to 1 (saccade finish), with data points recorded during the middle of the saccade corresponding to a value of 0.5 (Figure 4a). Saccade curvature for each participant and condition was defined as median angular deviation of the saccade path relative to a straight line from the starting point of the saccade to the saccade endpoint (Belopolsky & Theeuwes, 2011, Van der Stigchel, Meeter, & Theeuwes, 2006). Condition specific saccade curvature was defined as the difference between the trajectory of saccades when the distractor was presented clockwise and when it was presented counterclockwise from the saccade target (Figure 4a). This calculation takes into account any individual biases in trajectory deviations in absence of any distractor. Negative differences indicate curvature away from the distractor, while positive differences indicate curvature towards the distractor. 

Results

Saccade performance

Average saccade latencies were shorter in the match task—224 ± 22 ms (mean ± SEM)—than in the nonmatch task—254 ± 22 ms, t(9) = −2.86, p < 0.05). This difference in latencies is not surprising, as during the nonmatch task there was spatial uncertainty about the location of the saccade target through the memory delay. Saccade latencies, in either of the tasks (match: t[9] = 0.66, p = 0.52); nonmatch: (t[9] = 1.65, p = 0.13), did not depend on whether the attentional probe was at the same location as the saccade target or not (Figure 2a). This shows that our perceptual task did not affect oculomotor performance. Similar results were obtained for saccadic accuracy with accuracy being comparable for probe at the memorized versus nonmemorized location in match and nonmatch tasks (Figure 2b). 

Figure 2

 

(a) Cumulative probability of saccade latencies for match versus nonmatch tasks with the probe at either memorized or nonmemorized location. Horizontal lines represent standard error of the mean of average saccade latencies. (b) Saccade accuracy for match versus nonmatch tasks with the probe at either memorized or nonmemorized location. Lines represent area within which two-thirds of saccades ended within a particular condition. Corresponding color markers represent average end point accuracy with respect to the saccade target location (saccade target location coded as 0,0).

Discrimination performance at the memorized location

See Figure 3. Repeated measures ANOVAs with two factors—probe discrimination at memorized versus nonmemorized location and memory delay duration were carried out for the match and the nonmatch tasks. In the match task, probe discrimination was better at the memorized location (to which a saccade was planned) than at the nonmemorized locations, F(1, 9) = 12.14, p < 0.01. This shows that spatial attention was enhanced at the memorized saccade target location. This effect did not depend on the memory delay duration, F(3, 27) = 0.78, p = 0.55, and the interaction between the two factors was not significant, F(3, 27) = 1.40, p = 0.26. In the nonmatch task, on the other hand, probe discrimination at the memorized location (to which saccades were forbidden) was worse as compared to nonmemorized locations, F(1, 9) = 11.30, p < 0.01, whereas memory delay, F(3, 27) = 0.07, p = 0.97, and the interaction between the two factors, F(3, 27) = 0.59, p = 0.62, was not significant. This shows that memorizing a location but as a marker of “where not to look” leads to a suppression of attentional performance at that location. These results clearly demonstrate that attentional deployment at the memorized location depends on the contextual condition under which it was memorized, that is, whether later a saccade has to be made (match task) or avoided (nonmatch task) at that location. Additionally, no significant effect was found when we compared the discrimination performance at the nonmemorized locations in the match task with those in the nonmatch task, F(1, 9) = 0.59, p = 0.46; Figure 3a lighter-color lines. 

Figure 3

 

(a) Perceptual performance in identifying the discrimination target as a function of the delay condition. (b) Attention benefit at memorized versus nonmemorized locations. Black and red bars show the relative discrimination performance for each delay condition, respectively for match and nonmatch trials, obtained by subtracting the performance at the memorized locations from that at the nonmemorized locations.

Figure 4

 

Effects of probe distance from the memorized location on discrimination performance. Position 0° refers to the memorized location and the other three positions to nonmemorized locations at 36°, 72°, and 108° (radial directions from the line between central fixation and the memorized location serving as the 0° polar axis).

In a more direct comparison, Figure 3b presents relative probe discrimination obtained by subtracting the performance at the memorized location from that at the nonmemorized locations, separately for both match and nonmatch tasks for each delay condition. The data show that memorizing a location as a saccade target binds attention to that location giving it local processing benefits compared to nonmemorized locations. On the other hand, memorizing a location to which saccades are forbidden leads to an attentional cost at that location compared to the nonmemorized locations that are not forbidden to saccades. 

Effect of probe distance from the memorized location

Figure 4 shows discrimination performance as a function of radial direction of the probe from the memorized location. The probe was presented at a radius of 5° of visual angle from the central fixation with a polar angle of 0°, 36°, 72°, or 108° to the line, between the central fixation and the memorized location, serving as the polar axis (0° represents probe appearing at the memorized location). A two-way repeated measures ANOVA with main factors task and probe distance showed no effect of the task, F(1, 9) = 0.43, p = 0.52, no effect of distance, F(3, 27) = 1.00, p = 0.13, but interaction between the two factors was significant, F(3, 27) = 7.90, p < 0.01. Additionally, to evaluate the spatial specificity of enhancement/suppression at the memorized location, when the probe discrimination performance at the memorized location (0°) was compared only to the location right besides it (36° away in radial direction with an eccentricity of 5° from central fixation, equivalent to a distance of 3.09° of visual angle), it was found to be somewhat better in the match task, t(9) = 2.04, p = 0.07, and worse in the nonmatch task, t(9) = −3.36, p < 0.01. This analysis shows that the increase/decrease in perceptual performance was spatially specific to the memorized location. 

Saccadic trajectories and distractor suppression

While the attentional task was used as a direct measure of how visual processing resources are distributed in the visual scene during the memory delay period, we also used an indirect measure of task effects on saccade planning. We took advantage of the known finding that saccade trajectories deviate away from a distracting stimulus in the visual field (McSorley, Haggard, & Walker, 2006; Van der Stigchel et al., 2006). In our experiment, the nonmemorized location in the match task and the memorized location in the nonmatch task act as a distractor present close to the path between the fixation and the respective saccade target but with one crucial difference between the two conditions—while the distractor in the match task was saccade-irrelevant, the same in the nonmatch task was encoded as an actively suppressed “don't look” location. We wanted to see if we will observe similar trajectory deviation in our data, and if so, whether the extent and nature of the deviation would differ between the two conditions. Figure 5a shows averaged saccade trajectories of an individual participant, which deviated away from the distracting stimulus location in both match and nonmatch tasks. A two-way repeated measures ANOVA with factors task and distractor distance showed a main effect of the task, F(1, 9) = 13.16, p < 0.01, which means that saccade trajectories in the match task deviated away from the distractor less than in the nonmatch task (Figure 5b). The main effect of the distance between the saccade target and the distractor was not significant, F(2, 18) = 0.22, p = 0.80, nor was the interaction, F(2, 18) = 0.49, p = 0.61. Additional analyses showed that delay duration also did not affect saccade trajectories, F(3, 27) = 0.62, p = 0.60. One could argue that larger deviations of saccades in the nonmatch task were due to longer saccade latencies—as the saccadic system had more time to suppress the distractor in the nonmatch task. To test this possibility, we analyzed whether there was a relationship between saccade trajectory deviations and saccade latencies. We divided the trials into three groups based on the saccade latency tercile (Figure 5c). We found no effect of saccade latencies on trajectory deviations, F(2, 18) = 1.89, p = 0.18, and no interaction between saccade latency tercile and task, F(2, 18) = 2.13, p = 0.14). Thus, while visual displays in both the conditions were identical, saccades deviated away more from the memorized “don't look” distractor in the nonmatch task than from the nonmemorized saccade-irrelevant distractor in the match task. This is in line with our aforementioned finding that spatial attention at the “don't look” location was actively suppressed. 

Figure 5

 

Saccade trajectory deviations. (a) Averaged saccade trajectories of an individual participant. Circles illustrate distractor position, irrespective of its distance from the saccade target—Light-colored circle represents the clockwise distractor while dark-colored circle represents the counterclockwise distractor. Average saccade trajectories for distractors presented clockwise (light-colored curves) and counterclockwise (dark-colored curves). Saccade trajectories deviate away from the distractor. (b) Averaged saccade curvature for all participants as a function of the distance between saccade target and distractor location. (c) Averaged saccade curvature for all participants as a function of saccade latency. Saccade latencies were divided into terciles for each participant.

Discussion

Inhibitory processes in the saccadic system are crucial for the coordination between reflexive and goal-directed responses that lets our eye movements be automatic and flexible at the same time. However, despite the behavioral and clinical importance of saccade inhibition, the role of attentional processes in inhibiting a saccade to a location remains unclear. In this study, by adding a behavioral measure of attention to the oculomotor match- and nonmatch-to-sample tasks, we investigated the role of attentional allocation in maintenance of a location as allowed or forbidden for saccades. The memorized location in the two cases differed only by the saccadic task it was encoded for—to plan a saccade to it or to avoid making a saccade to it when the target reappeared. Particularly, during the delay in the nonmatch task, no saccade could be prepared and the only information available was the memorized location where the subjects had to avoid a saccade. We found that ruling out a location as a saccade target led to an attentional cost at that location as compared to the nonmemorized location that was not ruled out. On the other hand, preselecting a location as a saccade target enhanced attention at that location giving it local processing benefits as compared to the nonmemorized location that was not a saccade target. 

Our results also provide a novel distinction between the saccade trajectory deviation observed because of saccade-irrelevant distractors and that because of a location being forbidden to saccades by top-down processes (Figure 5). Saccade trajectories are known to deviate away from irrelevant distractors present close to the path between the fixation and the target and this is interpreted as inhibition of the distractor location in action-based representations (McSorley et al., 2006; Van der Stigchel et al., 2006). In the match task, the second element in the display at the time of the saccade is a saccade-irrelevant distractor and we observed the expected deviation of the saccade trajectory away from it. But importantly, in the nonmatch task, in which the second element in the display is not a saccade-irrelevant distractor but a location memorized as forbidden to saccades, the trajectory deviation was found to be significantly larger. The higher trajectory deviation away from the “don't look” location as compared to a visually similar distractor suggests active inhibition of the “don't look” location in oculomotor programming. 

We will first discuss the functional role of suppression in this task and how this suppression differs from inhibition of return and then the implications of this finding for attention-saccade coupling, several recent neurophysiological and neuroimaging studies that used this paradigm, and the role of attentional resources in working memory function. 

A number of factors warrant suppression of the memorized location in the nonmatch task. Peripheral cues with an abrupt onset can capture attention (Yantis & Jonides, 1984) and saccades (Theeuwes, Kramer, Hahn, Irwin, & Zelinsky, 1999) in an involuntary and stimulus-driven manner. Theeuwes et al. (1999) showed that, in a task that required a goal-directed saccade to a specific target, an abrupt visual onset, which was otherwise irrelevant to the task, elicited reflexive eye movements on more than one-third of the trials. In our experiment, the nonmatch task required participants to make a saccade immediately upon the sudden presentation of two highly salient objects. To do this task successfully hence requires not only that the nonmatching location be identified as the saccade target but also that a reflexive saccade to the matching location be inhibited to lower the chances of error. Furthermore, several recent studies have shown that an object held in working memory will automatically capture attention on a subsequent presentation (Pashler & Shiu, 1999; Soto, Hodsoll, Rotshtein, & Humphreys, 2008). This further emphasizes the behavioral necessity for active suppression of the memorized location to execute the nonmatch task as otherwise when the two potential targets appear at the response time, attention could be involuntarily captured by the matching location that is being held in working memory, increasing the likelihood of an erroneous saccadic response to that location. 

The “don't look” inhibition, or attentional suppression at a location that is forbidden to saccades, is a novel example of suppression of visuospatial attention, which is markedly different in its spatial and temporal characteristics from attentional inhibition observed in inhibition of return (Klein, 2000; Posner & Cohen, 1984). Firstly, inhibition of return (IOR) is elicited in experiments where the cue is uninformative and irrelevant to the task (Klein, 2000), while in both match and nonmatch tasks, the cued location is memorized and is relevant to both the discrimination and the saccadic response. Secondly, inhibitory effects in IOR are not restricted to the originally cued location but have been shown to be also present at nearby locations with the magnitude of inhibition declining with increasing angular distance from the cue. Dorris, Taylor, Klein, and Munoz (1999) found significant inhibitory effects for stimuli at 30° and 60° (radial direction with an eccentricity of 10° from the central fixation; 60° was the maximum distance that was tested in the same hemifield) on either side of the cued location. In a more detailed examination of spatial distribution of IOR, Bennett and Pratt (2001) found that “IOR spreads beyond the cued location to affect the cued hemifield”. In contrast, the decrease in perceptual performance in the nonmatch task was restricted to the cued location and was not found for probes presented at 36°, 72°, and 108° (radial directions with an eccentricity of 5° from central fixation) from the cued location. Performance at these adjacent locations was statistically indistinguishable from each other and was found to be at the same level as the performance at the saccade-irrelevant locations in the match task (Figure 4). Thirdly, the temporal parameters required to elicit IOR vary from those used in our experiment. The onset of IOR has been shown to depend on the attentional probe that is used. When IOR was tested with a discrimination task (which is more comparable to our experiment) instead of a detection task, Klein and Taylor (1994), Lupiáñez, Milán, Tornay, Madrid, and Tudela (1997); and Lupiáñez and Milliken (1999) found facilitation instead of inhibition at cue-target intervals of 500, 400, and 400 ms, respectively. Lupiáñez et al. (1997, 1999) showed that IOR can be observed with discrimination tasks but later (at 700 ms) and that it disappears sooner (at 1300 ms). In our results, on the other hand, inhibition could already be observed at 200 ms and, and was still present at a delay of 2500 ms (Figure 3b). 

Visuospatial attentional resources have been shown to be coupled to and deployed according to the demands of saccadic eye movements by a number of psychophysical (Deubel & Schneider, 1996; Jonikaitis & Deubel, 2011; Kowler et al., 1995; Rizzolatti et al., 1987), neurophysiological (Moore & Fallah, 2001, 2004), and neuroimaging (Beauchamp, Petit, Ellmore, Ingeholm, & Haxby, 2001; Corbetta, Akbudak, Conturo, Snyder, Ollinger, Drury, & Linenweber, 1998) studies. Deubel and Schneider (1996) showed that attention remains enhanced at future saccade targets until the onset of the saccade and argued for an obligatory and selective coupling between these two mechanisms. Our results confirm and complement the claim of an obligatory coupling between saccades and attention—saccades at the memorized location could neither be planned nor suppressed independent of a corresponding effect on attentional performance. These results point to a common signal, which can either be enhanced at a given location leading to a positive saccadic response and enhanced object discrimination ability, or be suppressed at a particular location leading to suppression of both saccades and attention. 

Inhibition of saccades in the nonmatch task was also investigated in a monkey neurophysiology study by Hasegawa et al. (2004), who discovered a counterpart to the saccade planning neurons in the frontal cortex in the form of a distinct neuronal population dedicated to saccade suppression in a spatially selective manner. Using an experimental paradigm similar to ours, they identified a group of neurons (named as “don't look” neurons) in the frontal eye field and the caudal prefrontal cortex that code selectively for the forbidden stimulus in the nonmatch task. The delay period activity of these “don't look” neurons was spatially tuned for the memorized location only in the nonmatch task while they were not tuned at all for the same location in the match task. Around the same time, a series of elegant studies by Moore and colleagues had established that neural circuits involved in saccade preparation play a causal role in directing and modulating covert spatial attention. Subthreshold microstimulation of sites in frontal eye field (FEF) (Moore & Fallah, 2001, 2004) that are responsible for generating saccades to a particular location led to improved performance in psychophysical visual attention tasks involving stimuli at that location. Similar microstimulation studies have also demonstrated an attention-like enhancement in activity of V4 neurons at retinotopically corresponding locations (Moore & Armstrong, 2003). Collating these results with our data, we believe it to be plausible that the sustained frontal signal demonstrated by Hasegawa et al. (2004) in the nonmatch task leads to a concomitant suppression of both, saccades and attention. 

Recent human neuroimaging studies (Curtis, 2006; Curtis, Rao, et al., 2004; Curtis, Sun, et al., 2005) have shown that match and nonmatch tasks lead to sustained activation of different brain areas during their respective delay periods, despite the fact that the sample cue being maintained in the working memory, through the delay, is visually the same in both. The authors attribute this to the difference in the purpose for which the cue is encoded in each of the tasks. In the match task, as soon as the sample cue appears participants can make the saccade plan and simply postpone its initiation until after the delay, hence sustained activation seen during the delay period should represent the maintenance of a motor code in working memory. In the nonmatch task on the other hand, since no saccade can be planned until after the delay, the only information to be retained is the location of the sample cue, sustained activation seen during the delay period should represent the maintenance of a sensory code in working memory. Curtis and colleagues used this to argue that when a visually similar cue has to be maintained in working memory over a delay for two different purposes (as a motor plan in the match task and as a sensory memory of the location in the nonmatch task), it activates two different nodes within the frontoparietal network, which represent the neural correlates of two different subtypes of working memory representation. 

However, none of these studies (Curtis, 2006; Curtis, Rao, et al., 2004; Curtis, Sun, et al., 2005) directly investigated the role attention might have in performing this task. Our results corroborate the primary assumption made by these studies that the match and nonmatch tasks are performed using two different cognitive strategies but question the assumptions made regarding what exactly these strategies are and hence what the observed sustained activity represents. In the match task, Curtis and colleagues argue that a saccade plan/prospective motor code is generated right after the sample cue appears but do not ascribe any role to the attentional processes. Our results indicate that attention at the memorized location in the match task was enhanced throughout the delay and in fact was significantly higher than all other conditions in either task. These results confirm previous studies (Deubel & Schneider, 1996, 2003; Kowler et al., 1995) which also show that spatial attention is obligatorily maintained at the location selected as a future saccade target till the saccade is carried out. Hence the sustained activation seen in the oculomotor areas in the match task is likely to represent not only the maintenance of the motor code but also sustained attention. At the same time, in the nonmatch task, they argue that the way to make the right saccade is by sustaining attention at the memorized location throughout the delay right up to the response time and hence the sustained parietal activity seen in the nonmatch task represents sustained spatial attention. Our results show that the nonmatch task delay period involves suppression rather than maintenance of spatial attention at the cue location and hence call for refining the interpretation of the sustained neural activity seen in the posterior parietal cortex (PPC) and inferior frontal cortex during the nonmatch task delay in terms of suppression rather than maintenance of attention. Lindner, Iyer, Kagan, and Andersen (2010), in a neuroimaging study using a manual reaching version of the match and nonmatch tasks, also argued along these lines by maintaining that the nonmatch task is solved by sustained inhibition of actions towards the precued undesired locations. In indirect support of this notion, they found that the fMRI activity in the PPC increased with the number of undesired target locations but not with the number of default locations and suggested that this activity reflects suppression of movements directed towards the memorized undesirable targets. 

Remembering a location as forbidden to a certain response represents a conflict for cognitive processes such as attention that have been shown to be essential to both remembering and responding. Modulation of attentional resources has been found to be essential for not only saccadic planning but also working memory. The attention-based rehearsal hypothesis of working memory (Awh et al., 2006; Smyth & Scholey, 1994) suggests that attention is deployed at locations held in visuospatial working memory and that such attention-based rehearsal plays an essential role in maintenance of information in working memory. However, while visuospatial attention has been shown to be coupled to both saccadic planning and working memory rehearsal, it is not clear what happens in cases where these two processes make conflicting demands on attentional resources. The nonmatch task that we used represents one such unique experimental situation in which saccadic planning and working memory make opposite demands on attentional modulation—the “don't look” location has to be kept in working memory, but saccades to its location have to be suppressed. We found that in such a scenario, attentional resources were preferentially allocated in accordance with the saccadic demands, despite the conflicting needs of working memory. Our results show, for the first time, that a location can be maintained in working memory through a delay while attention at that location is suppressed instead of being enhanced. Sustained attentional suppression of a task-relevant location, by definition, requires its spatial coordinates to be maintained through the delay. Such an inhibitory trace could simultaneously prevent an error saccade and serve as a memory trace for that location. Hence, our results suggest that active attentional suppression, much like attentional enhancement, could contribute to the rehearsal of working memory representations thereby allowing working memory to interact with attentional resources in a less obligatory and more context-dependent manner. 

Lastly, saccade inhibition deficits have been found in a whole variety of clinical populations (see Everling & Fischer, 1998, for a review) including schizophrenia and first-degree relatives of schizophrenia patients (Clementz et al., 1994), Huntington's and Alzheimer's related dementia (Currie, Ramsden, McArthur, & Maruff, 1991), frontal lesions (Fukushima, Fukushima, Miyasaka, & Yamashita, 1994) and ADHD (Mostofsky et al., 2001). Despite the well-established links between attention and saccades, the possible (and very plausible) role of visuospatial attention in saccade inhibition deficits has remained neglected. The close attention-saccade coupling, along with our finding that inhibition of saccades is accompanied by attentional suppression, point to the possibility that various clinical populations that show deficits in the inhibition of saccades might also show deficits in attentional suppression. 

Acknowledgments

This work was supported by the Deutsche Forschungsgemeinschaft grants GRK 1091 and JO980-1/1. 

*SD and DJ contributed equally to this article. 

Commercial relationships: none. 

Corresponding author: Donatas Jonikaitis. 

Email: [e-mail]. 

Address: Allgemeine und Experimentelle Psychologie, Ludwig Maximilians Universität München, Munich, Germany.