Visual sensitivity of frontal eye field neurons during the preparation of saccadic eye movements

Primates make frequent saccades to inspect their surroundings and yet must maintain perceptual stability across saccades. The oculomotor system has long been posited to play a role in perceptual stability by reducing visual sensitivity around the time of saccades. We demonstrate that neurons in the frontal eye field (FEF), a prefrontal oculomotor area, exhibit reduced visual sensitivity around the time of saccades, suggesting that the FEF may be a source of perisaccadically reduced visual sensitivity in other brain areas.

Abstract

Primate vision is continuously disrupted by saccadic eye movements, and yet this disruption goes unperceived. One mechanism thought to reduce perception of this self-generated movement is saccadic suppression, a global loss of visual sensitivity just before, during, and after saccadic eye movements. The frontal eye field (FEF) is a candidate source of neural correlates of saccadic suppression previously observed in visual cortex, because it contributes to the generation of visually guided saccades and modulates visual cortical responses. However, whether the FEF exhibits a perisaccadic reduction in visual sensitivity that could be transmitted to visual cortex is unknown. To determine whether the FEF exhibits a signature of saccadic suppression, we recorded the visual responses of FEF neurons to brief, full-field visual probe stimuli presented during fixation and before onset of saccades directed away from the receptive field in rhesus macaques (Macaca mulatta). We measured visual sensitivity during both epochs and found that it declines before saccade onset. Visual sensitivity was significantly reduced in visual but not visuomotor neurons. This reduced sensitivity was also present in visual neurons with no movement-related modulation during visually guided saccades and thus occurred independently from movement-related activity. Across the population of visual neurons, sensitivity began declining ∼80 ms before saccade onset. We also observed a similar presaccadic reduction in sensitivity to isoluminant, chromatic stimuli. Our results demonstrate that the signaling of visual information by FEF neurons is reduced during saccade preparation, and thus these neurons exhibit a signature of saccadic suppression.

NEW & NOTEWORTHY

Primates make frequent saccades to inspect their surroundings and yet must maintain perceptual stability across saccades. The oculomotor system has long been posited to play a role in perceptual stability by reducing visual sensitivity around the time of saccades. We demonstrate that neurons in the frontal eye field (FEF), a prefrontal oculomotor area, exhibit reduced visual sensitivity around the time of saccades, suggesting that the FEF may be a source of perisaccadically reduced visual sensitivity in other brain areas.

all sensory systems face the problem of potential disruption created by self-generated sensory input. As animals move through the world, their movements generate sensory signals that may distract from or confound representations of the external world. This problem is especially apparent in the visual system of primates, because they make frequent saccadic eye movements to bring objects of interest onto the fovea for high-acuity examination. Surprisingly, however, these ballistic eye movements do not create a disturbing percept of motion. One mechanism that is thought to contribute to visual stability is saccadic suppression, a psychophysical phenomenon consisting of a 3- to 10-fold loss of visual sensitivity across the entire visual field that occurs just before, during, and after saccadic eye movements (Burr et al. 1982; Diamond et al. 2000; Knöll et al. 2011; Latour 1962; Shioiri and Cavanagh 1989; Zuber and Stark 1966). This loss of visual sensitivity around the time of eye movements could increase perceptual stability by reducing perception of potentially disturbing, self-generated visual motion. Previous studies have identified neural correlates of saccadic suppression in visual cortex in the form of a reduction of visual responses and contrast sensitivity for visual stimuli appearing before or during saccades directed away from the receptive field (RF) (Bremmer et al. 2009; Han et al. 2009; Ibbotson et al. 2007; Kleiser et al. 2004; Sylvester et al. 2005; Thiele et al. 2002). However, it is unclear where and how saccadic suppression signals originate.

It has long been posited that saccade generation must invoke an active, inhibitory process that suppresses perisaccadic vision in some way (Holt 1903; Sperry 1950; von Helmholtz 1925; von Holst and Mittelstaedt 1950). Consistent with this, stimuli that are imperceptible during saccades remain visible when they are passively moved over a stable eye at saccadic speeds (Burr and Ross 1982; Diamond et al. 2000). However, a neural source of active saccadic suppression has yet to be identified. Structures in the saccadic system that both contribute to saccade generation and modulate visual cortical responses are good candidates for a central source of saccadic suppression. One such area is the frontal eye field (FEF). In addition to contributing to the generation of saccades, the FEF has reciprocal, monosynaptic connections with dorsal and ventral extrastriate visual cortex (Anderson et al. 2011; Schall et al. 1995; Stanton et al. 1995) and has an established causal role in modulating signals in visual cortex (Ekstrom et al. 2009; Moore and Armstrong 2003; Noudoost et al. 2014; Noudoost and Moore 2011).

To determine whether the FEF exhibits a signature of saccadic suppression, we investigated changes in visual sensitivity within the FEF during the preparation of saccadic eye movements. FEF neurons can exhibit visual responses, motor preparatory activity, or both; neurons with these properties are, respectively, termed visual, motor, and visuomotor neurons. During saccade preparation, motor and visuomotor neurons exhibit motor preparatory activity broadly tuned for saccades directed toward the neuron's “movement field” (Bruce and Goldberg 1985; Schall 1991). Simultaneously, FEF visual activity is enhanced when saccades are directed into the visual RF (Bruce and Goldberg 1985; Wurtz and Mohler 1976). However, it is not known whether visual information in the FEF is reduced, enhanced, or unaffected when saccades are directed away from the RF. We investigated this question by measuring the visual sensitivity of FEF neurons to brief visual probes presented during fixation and before saccades directed away from the RF and observed that visual sensitivity declines in the FEF before saccade onset.

MATERIALS AND METHODS

General and surgical procedures.

Two male rhesus monkeys (Macaca mulatta, 11–14 kg) were used in these experiments. All experimental procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Society for Neuroscience Guidelines and Policies and approved by Stanford University Animal Care and Use Committee. Each animal was surgically implanted with a titanium head post and a cylindrical titanium recording chamber (20 mm diameter), allowing access to the FEF. Surgery was conducted using aseptic techniques under general anesthesia (isoflurane), and analgesics were provided during postsurgical recovery. Eye-position monitoring was performed with 1,000 Hz resolution eye-tracking cameras (EyeLink 1000; SR Research, Mississauga, Ontario, Canada). Eye monitoring, stimulus presentation, data acquisition, and behavioral monitoring were controlled using the NIMH Cortex system (National Institute of Mental Health, National Institutes of Health, Bethesda, MD). Saccades were detected in the eye-tracking data using the combination of a velocity threshold (10°/s) and a “moving boxcar” technique for detecting deflections in eye position (Armstrong et al. 2006). Visual stimuli were presented on liquid-crystal display monitors at a 60-Hz refresh rate (Samsung S23A700D). Background illumination was 34 cd/m². Onset and offset of visual stimuli were measured with a photodiode placed over a small, white bar whose onset was synchronized to stimulus onset; this photodiode stimulus was placed on the edge of the monitor contralateral to RFs and hidden from the monkey's view, such that it was not part of the visual display.

Multielectrode array recording in FEF.

Single-neuron recordings were made with varnish-coated tungsten microelectrodes (0.1–0.5 MΩ impedance measured at 1 kHz; FHC, Bowdoin, ME) or 16-channel axial array electrodes (U-probe and V-probe; Plexon, Dallas, TX), lowered using a hydraulic microdrive (Narishige, Amityville, NY) through a recording chamber over a craniotomy covering the anterior bank of the arcuate sulcus where the FEF is located. In each monkey, the FEF was first localized by using single-neuron recording and microstimulation to identify surrounding physiological and anatomical landmarks, including premotor cortex and the arcuate sulcus. Individual sites were defined as FEF if fixed-vector, short-latency (<75 ms) saccadic eye movements could be evoked with <50 μA microstimulation (Bruce et al. 1985). Microstimulation consisted of a 100-ms train of biphasic current pulses (0.25 ms, 333 Hz) delivered with a Grass S88 stimulator and two Grass stimulation isolation units (PSIU-6; Grass Instrument, West Warwick, RI). The amplitudes of evoked saccades ranged from 5 to 20 degrees of visual angle (DVA) in Monkey S and 5–18 DVA in Monkey B. Following identification of FEF sites using microstimulation, we performed FEF recordings using 16-channel axial array electrodes, placed at the coordinates of confirmed FEF sites. Extracellular waveforms were manually classified as single neurons offline (Offline Sorter; Plexon).

Visually guided saccade task with probe stimuli.

We trained monkeys to perform a delayed, visually guided saccade task [e.g., Sommer and Wurtz (1998)] in which task-irrelevant visual probes flashed briefly onscreen during fixation and just before onset of saccades directed away from the RF (see Fig. 1A). In the task, monkeys fixated on a central fixation spot subtending 0.1 DVA within a 1.5-DVA fixation error window. After the animal initially achieved fixation, the saccade target appeared in the hemifield opposite from the recorded RFs at an eccentricity of 9 DVA from the central fixation spot and a randomly chosen angle within −45° to +45° relative to the horizontal meridian. This delayed saccade task design, with the target appearing before the cue to saccade, was chosen to ensure that the target onset did not disrupt visual responses to the fixation or presaccadic probe. After a 500- to 750-ms interval following target onset, a first visual probe stimulus (hereafter called the fixation probe, described in more detail below) appeared for a single, 16-ms screen refresh. After another 500–750 ms, the monkey was cued to saccade by the disappearance of the fixation spot. A second, identical probe stimulus (the presaccadic probe) was timed to appear with high likelihood, after the cue to saccade but before saccade onset, by placing it after the cue at an interval that was shorter than the mean reaction time across the previous 20 trials by a fixed amount. Although saccadic suppression occurs for probes presented throughout the perisaccadic epoch, the second probe was presented only during the presaccadic epoch to permit measurement of neuronal responses to a stable stimulus during fixation. No-probe stimuli appeared once the eye position left the fixation error window. The monkey was rewarded with juice for correctly executing a saccade into a 5-DVA window around the target.

Fig. 1.

Probing visual sensitivity of FEF neurons during fixation and before visually guided saccades. A: visually guided saccade task. A full-field, 16-ms probe stimulus appeared during fixation and again between the cue to move and the saccade onset. The probe consisted of pseudorandomly positioned gray circles, 1 DVA in diameter, and spaced 5–8 DVA apart; the contrast of the circles with respect to the background was uniform within trials but varied across trials. The monkey was rewarded for responding to the offset of the fixation spot (white circles) by saccading to a target (yellow circles). The target was placed in the hemifield opposite recorded RFs and appeared at a random angle between −45° and +45° of the horizontal meridian at the beginning of each trial; the dashed, yellow arc in the left panel indicates the range of possible target locations. Linear electrode arrays were used to record from multiple neurons with overlapping or nonoverlapping RFs; dashed, black circles indicate a schematic example of RF locations. B: control of presaccadic stimulus-onset times. Probe stimulus onsets were measured using a photodiode (top; sample traces) and timed to fall with high probability just before the expected saccade onset (middle; sample eye traces), based on past reaction times. Trials with correctly timed presaccadic probes were selected post hoc for further analysis (bottom; example experiment; red, selected bins). C: visual responses of 2 FEF neurons recorded in a single session using a linear electrode array; shaded regions indicate means ± SE. Top: a cell with equal visual responses to probes presented during fixation (black) and presaccadically (red); the yellow trace shows presaccadic (presac) activity on trials when no probe was presented. Bottom: a cell with presaccadically suppressed visual responses.

Visual probe stimulus.

Since each recording included multiple neurons with distinct RF locations, we used a “full-field” probe stimulus covering the extent of the liquid-crystal display monitor to probe the visual responses of all neurons simultaneously. Furthermore, given that FEF RFs shift dramatically during saccades (Zirnsak et al. 2014), use of a full-field probe also eliminated the need to map the presaccadic RF. The full-field probe consisted of a field of 1 DVA-diameter gray circles with pseudorandomized positions separated by 5–8 DVA, such that the average distance between circles was no greater than the minimum RF diameter across the majority of multielectrode array channels. RF diameters were estimated online by listening to amplified activity of each isolated neuron as an oriented bar stimulus was swept across the screen through a variety of trajectories. Luminance contrast was defined as the Michelson contrast [C = (max − min)/(max + min)] between circle and background luminance. Two percent to 32% Michelson contrast probes were randomly interleaved across trials. For each contrast level, an identical library of images with pseudorandomized probe positions was used, such that nothing varied between conditions except luminance contrast. Within trials, the fixation and presaccadic probes were identical. The size and spacing of the scattered circles were appropriately optimized for both the fixation and perisaccadic RFs, because RF diameter does not decrease before saccades (M. Zirnsak and T. Moore, unpublished data).

Visual responses in the FEF adapt in response to repeated stimulation, as in other areas of the visual system (Brown et al. 1987; Motter 2006; Schiller 1968); however, FEF visual responses have been shown to recover from adaptation within 400 ms of visual stimulus onset (Mayo and Sommer 2008). To avoid any possibility of adaptation effects contributing to decreased presaccadic visual responses, we used an interval of 500–750 ms between the fixation probe and the cue to saccade. The presaccadic probe followed the cue to saccade by an interval determined by the monkey's reaction time; as a result, the interval from the fixation probe to the presaccadic probe (interprobe interval) was 738.5 ± 93.4 ms (mean ± SD).

Many FEF neurons exhibit broadly tuned movement fields that may stretch across the vertical meridian and extend into two or three quadrants of visual space or, more rarely, be untuned altogether (Bruce and Goldberg 1985). Saccades were directed into the hemifield opposite from RFs to minimize movement-related activity, because visual activity was of primary interest in this study; nonetheless, we anticipated that some neurons in our sample might exhibit movement-related increases or decreases in firing rate during the presaccadic epoch. To control for movement-related changes in neuronal activity within the presaccadic probe presentation window, we included trials in which no visible probe was presented (hereafter termed “no-probe” trials). On no-probe trials, the hidden photodiode stimulus was presented with the same statistical distribution of onset times as on probe trials to obtain a measurement of baseline spiking activity in the absence of visual stimulation. Any movement-related dynamics present during the presaccadic epoch would be observed during both no-probe and probe trials. Neuronal activity attributable solely to visual stimulation, in contrast, could be isolated by comparing probe trials with the “baseline” measured during no-probe trials.

Twenty-seven of the 37 recording sessions also included a subset of trials in which the visual probes were varied in color rather than luminance. These probe stimuli consisted of physically equiluminant, 34 cd/m², 0.3 cycles/DVA, red-green square waves within 3 DVA Gaussian envelopes, at the same pseudorandomized positions as the luminance contrast probe stimuli used on other trials.

Memory-guided saccade task.

The visually guided saccade task was followed by a memory-guided saccade (MGS) task. This allowed us to classify neurons into standard, functional categories: visual, motor, and visuomotor neurons (Bruce and Goldberg 1985; Sommer and Wurtz 1998). In this task, the monkey was trained to fixate while a saccade target (33% Michelson contrast) was presented for 300 ms. After a 500-ms delay, the disappearance of the fixation spot cued the monkey to execute a saccade to the remembered target location. On each trial, the target could appear at six possible angles (0, 45, 90, 180, 270, and 315°) from the horizontal meridian of the recorded hemisphere and up to two possible radial distances (6 or 9 DVA) from the fixation spot. The radial distance(s) were chosen each day based on the radial distance(s) of recorded neurons' RFs (estimated using the procedure described above). Functional types were identified by comparing spike-rate distributions (two-sample t-test with Holm-Bonferroni corrections for multiple comparisons) in the target or saccade epoch with their respective baselines for each target location. Specifically, activity in the target-onset epoch (50–350 ms following target onset) was compared with the target-baseline epoch (350–50 ms preceding target onset), and activity in the saccade epoch (150 ms preceding 50 ms following saccade onset) was compared with the saccade-baseline epoch (400–250 ms preceding saccade onset), similar to previous studies (Sommer and Wurtz 1998). Visual neurons were defined as those that exhibited significant target-epoch activity at one or more target positions but saccade-epoch activity at none. Visuomotor neurons were defined as exhibiting both target-epoch and saccade-epoch activity at one or more target positions.

Data analysis.

All data analyses were performed in Matlab (MathWorks, Natick, MA). Only correctly completed trials were included in analyses. Only neurons with significant visual responses to the full-field probe during fixation were included in analyses. To identify these neurons, spikes were counted in a 30- to 200-ms window following probe presentations, and a right-sided Wilcoxon rank sum test (P < 0.01) with Holm-Bonferroni corrections across contrast conditions was used to compare firing rates following the onset of the probe at each contrast level with firing rates in the same window following the onset of the photodiode patch on no-probe trials. A receiver operating characteristic (ROC) analysis was used to compute neuronal visual sensitivity, based on the degree of overlap of the distribution of spike rates in the probe-response window on probe trials and the distribution of spike rates in the same window on no-probe trials. This analysis asks how accurately an ideal observer can classify trials as having had a probe or not by setting a criterion for the spike count on each trial, above which the trial would be classified as a “probe” trial. The area under the ROC curve (AUC) gives the probability of an observer correctly classifying a trial as having a probe or not, based on the spike rates (Britten et al. 1992; Green and Swets 1966). An AUC value of 0.5 indicates a complete overlap of the probe and no-probe spike-rate distributions, whereas 1.0 indicates perfect separation of these distributions. Across the population, contrast-sensitivity functions were averaged together, and the functions for the fixation probe and presaccadic probe were compared using a two-way repeated-measures ANOVA (significance threshold, P = 0.05).

To characterize basic properties of visual and visuomotor neurons, we used a standard Poisson spike-train analysis to determine the latency and duration of visual responses to the memory-guided saccade target (Burrows and Moore 2009; Légendy and Salcman 1985; Schmolesky et al. 1998; Sheinberg and Logothetis 2001; Thompson et al. 1996). For this analysis, we included neurons identified as visual or visuomotor using the MGS task that also had significant responses to a full-field probe used in the visually guided saccade task. We calculated a “surprise index” [S = −log(P)], where P(n) = e^−rτ[∑_{i = n}^∞(rτ)ⁱ/i!] indicates the probability of observing n or more spikes in a given time interval τ if spike rates are determined by a Poisson process with mean equal to the neuron's baseline firing rate r [P(n) = (e^−λλⁿ)/n!], where λ = rτ. We summed spikes in τ = 100 ms sliding windows, centered on each time point, shifted in 1 ms steps, for a total epoch of 400 ms following target onset. The baseline firing rate was measured from the 450-ms window preceding target onset. The onset of significant visual activity on a given trial was taken to be the time when the surprise index exceeded 2 and remained >2 for at least 100 consecutive windows. After identifying the visual response onset and offset, we then eliminated the first 1–50 ms, one by one, from the period of significant responses, if doing so resulted in an increase in the surprise index for the entire period of visual activity (Légendy and Salcman 1985; Sheinberg and Logothetis 2001).

RESULTS

We recorded the activity of 170 single FEF neurons from 2 monkeys (97 in Monkey S and 73 in Monkey B) during 37 recording sessions (23 in Monkey S and 14 in Monkey B) using 16-channel linear multielectrode arrays. We probed the visual sensitivity of FEF neurons by presenting brief (16 ms) visual stimuli during fixation and also before visually guided saccades directed away from their RFs (Fig. 1A). Precise control of presaccadic probe timing was achieved by measuring probe onset with a photodiode (Fig. 1B). Only trials in which the probe onset was determined post hoc to precede saccade onset by 16–144 ms (1–9 screen refreshes) were included in analyses (Fig. 1B). Across all recordings, there were 36.1 ± 15.9 (means ± SD) and 27.4 ± 9.3 trials, on average, per contrast level for the fixation and presaccadic probes, respectively, and 55.9 ± 41.8 and 41.6 ± 26.7 trials in the fixation and presaccadic no-probe conditions, respectively. Figure 1C shows visual responses to the fixation and presaccadic probe for two neurons recorded in a single session: one with visual responses maintained in the presaccadic period (two-sample t-test, P = 0.10) and one with significantly reduced presaccadic visual responses (two-sample t-test, P = 0.0031; for both neurons, fixation: n = 173 trials; presaccadic probe: n = 122 trials; and presaccadic epoch, no probe: n = 71 trials). For all analyses, we included only neurons with significant visual responses to at least one contrast level of the full-field probe presented during fixation (right-sided Wilcoxon rank sum test, P < 0.01, with Holm-Bonferroni corrections). These exclusion criteria identified a total of 47 neurons (33 in Monkey S and 14 in Monkey B) that responded to the full-field probe. All RFs were >5 DVA in diameter and located in the hemifield opposite from the saccade target.

Visual sensitivity is suppressed before saccades.

We assessed neuronal visual sensitivity using an ROC measure that quantified the ability of an observer to discriminate the presence or absence of the visual probe based on neuronal spike rates (Britten et al. 1992; Green and Swets 1966). Figure 2 demonstrates this analysis for an example neuron. For this neuron, fixation probes were presented 705 ± 42 ms before saccade onset; presaccadic probes were presented 81 ± 36 ms before saccade onset (Fig. 2A). Visual responses to the presaccadic probe were reduced relative to visual responses to the fixation probe at 16% contrast (two-sample t-test, P = 0.0001) and 32% contrast (Fig. 2B; two-sample t-test, P = 0.021; fixation probe: 53.8 ± 16.1 trials per condition; presaccadic probe: 29.8 ± 7.9 trials per condition). Figure 2C plots the sensitivity of this neuron to the fixation and presaccadic probes across contrasts, specifically the discriminability of the probe responses from the no-probe responses.

Fig. 2.

A single FEF neuron with suppressed presaccadic visual sensitivity. A: probe time distributions for all trials of a single recording (blue, fixation probe; red, presaccadic probe). Boxed areas indicate middle 50% of data; whiskers enclose all data. B: visual responses to 2, 4, 8, 16, and 32% Michelson contrast (inset color bars; shaded regions indicate means ± SE), measured in a window 30–200 ms following the probe onset (gray regions), for a single unit. C: ROC analysis was used to measure visual sensitivity (AUC) by comparing probe responses at each contrast with responses on no-probe trials (blue, fixation probe; red, presaccadic probe).

We used the ROC analysis to compare sensitivity to fixation and presaccadic probes across the population of visually responsive neurons (Fig. 3). Fixation probes were presented, on average, 824 ± 97 ms before saccade onset, whereas all presaccadic probes were restricted to a window, 16–144 ms before saccade onset (means 87 ± 34 ms; Fig. 3A). Over the population, sensitivity to presaccadic probes was reduced relative to sensitivity to fixation probes (Fig. 3B; two-way repeated-measures ANOVA, n = 47 neurons; epoch: P < 0.001; contrast: P < 0.00001; contrast × epoch: P < 0.0001). The magnitude of this reduction of sensitivity depended on stimulus contrast. At the highest contrast level, the presaccadic decrease in ROC AUC value was 0.089 (Fig. 3B). Figure 3C plots the fixation and presaccadic sensitivities across all neurons and contrasts, with values below the line of unity indicating lower sensitivity in the presaccadic epoch than in the fixation epoch. The population distribution of presaccadic sensitivity values was consistently shifted below the line of unity across all contrast levels.

Fig. 3.

The visual sensitivity of FEF neurons during fixation and before saccades. A: probe time distributions for all trials of all recordings (blue, fixation probe; red, presaccadic probe). Boxed areas indicate middle 50% of data, whiskers enclose middle 99.3% of data, and crosses indicate any values outside of this range. B: ROC analysis quantifying the separation of distributions of neuronal responses to 2–32% contrast probes presented during fixation (blue) or before saccades (red) from distribution of responses measured in the same time windows on trials where no visual probe was presented. Inset: ROC difference (Δ) across contrast. C: sensitivity values shown in B plotted for individual neurons. Each point is 1 neuron at 1 contrast level (contrast values represented in color bar, right). The mean differences between fixation and presaccadic sensitivity values for each contrast are indicated by arrows in the top right corner. Circles below the line of unity demonstrated lower sensitivity in the presaccadic epoch than in the fixation epoch.

Visual sensitivity across functional classes of neurons.

To determine whether the reduction in visual sensitivity differed across functional classes of neurons, we classified FEF neurons during a separate MGS task in which the monkey saccaded toward a remembered target location that could be inside or outside of the RF. Of the 47 neurons that responded to the full-field visual stimulus, 12 were identified as visual and 12 as visuomotor using the MGS task. For the remaining 23 out of 47 neurons with visual responses to the full-field probe, we were unable to determine a type using the MGS task, likely because we could not simultaneously optimize MGS target locations for RFs on all 16 channels of the electrode array. For visual and visuomotor neurons, respectively, mean onset latencies of visual responses to the MGS target were 99.54 ± 33.8 (means ± SD) and 81.7 ± 45.2 ms; durations of visual responses were 198 ± 56.0 and 191 ± 79.5 ms; baseline firing rates were 8.26 ± 10.5 and 4.69 ± 5.40 Hz; and firing rates during periods of significant visual responses were 30.4 ± 23.9 and 31.1 ± 29.4 Hz compared with previously reported measurements (Bruce and Goldberg 1985; Schall 1991; Schmolesky et al. 1998; Thompson et al. 1996).

We compared sensitivity to the full-field probe during fixation and before visually guided saccades for neurons with MGS-defined visual and visuomotor activity. The relative responsiveness of MGS-defined visual, visuomotor, and motor neurons during the target and saccade epochs of the MGS task is shown in Fig. 4A. Visual neurons demonstrated significant reduction of presaccadic visual sensitivity during the visually guided saccade task (Fig. 4B; two-way ANOVA, n = 12 neurons; epoch: P = 0.013; contrast: P < 0.0001; epoch × contrast: P = 0.72). In contrast, visuomotor neurons exhibited only a trend toward reduced sensitivity (two-way ANOVA, n = 12 neurons; epoch: P = 0.12; contrast: P < 0.0001; epoch × contrast: P = 0.86). Thus presaccadic sensitivity was significantly reduced in MGS-defined visual but not visuomotor neurons.

Fig. 4.

Functional typing of FEF neurons using a memory-guided saccade task. A: population mean spike rates during the target (gray) and saccade (black) epochs for neurons classified as visual, visuomotor, and motor. For each neuron, epoch-appropriate baseline activity was subtracted, and firing rates were normalized to the epoch in which the neuron demonstrated a higher firing rate. *P < 0.05. B: sensitivity to fixation (blue) and presaccadic (red) probes presented in the visually guided saccade task, for visual (left) and visuomotor (right) neurons.

We next asked whether suppression would be present in neurons that demonstrated no movement-related activity during the visually guided saccade task. Previous work has demonstrated that FEF neurons that exhibit no premotor activity on a learned saccade task, similar to the MGS task used here, may still exhibit movement-related modulation during a visually guided saccade task (Bruce and Goldberg 1985), a pattern that is also present in our sample. To control for movement-related baseline activity preceding visually guided saccades, we restricted our measurement of visual sensitivity to neurons with no movement-related modulation of baseline activity during the visually guided saccade task (Vis_VGS neurons). To identify Vis_VGS neurons, we compared the within-neuron-normalized baseline (no-probe condition) firing rates during fixation and before saccades (Fig. 5B). Figure 5C shows mean firing rates for Vis_VGS neurons during no-probe trials and during trials with 8–32% contrast probes (both with fixation and presaccadic); note that unlike for the population (Fig. 5A), firing rates in the no-probe condition do not vary between the fixation and presaccadic epochs. Eighteen of the 47 neurons with full-field probe responses exhibited no difference in baseline activity between the fixation and presaccadic periods (two-sample t-test, P > 0.05). Vis_VGS neurons demonstrated reduced presaccadic visual sensitivity (Fig. 5D; two-way repeated-measures ANOVA, n = 18 neurons; epoch: P = 0.012; contrast: P < 10⁻⁹; epoch × contrast interaction: P < 0.01). Neurons that exhibited increased movement-related baseline modulation in the presaccadic epoch (VM_VGS neurons; n = 20 of 47 neurons with visual responses to the full-field probe) also exhibited significantly reduced sensitivity (repeated-measures ANOVA, n = 20; epoch: P = 0.0025; contrast: P < 10⁻⁵; epoch × contrast interaction: P = 0.011). Thus presaccadic reduction of sensitivity was present regardless of whether cells exhibited movement-related baseline activity before visually guided saccades.

Fig. 5.

A: population peristimulus time histogram comparing fixation (blue) and presaccadic (red) visual responses with 8–32% probes (shaded regions indicate means ± SE). Values are normalized within each cell to the highest activity during the fixation probe condition. Motor-related activity is illustrated by cyan (fixation) and orange (presaccadic) traces, demonstrating activity on trials with no visual probe. Hatched lines indicate range of saccade onset times (16–144 ms following probe onset). B: neurons with no motor preparatory activity (Vis_VGS; cyan circles, two-sample t-test, P > 0.05) were identified by comparing activity on no-probe trials during fixation (x-axis) with the presaccadic epoch (y-axis), each normalized within each neuron to the maximum firing rate during the fixation probe condition as in A [black circles, neurons with increases in baseline activity before visually guided saccades (VM_VGS neurons); open circles, remaining neurons]. C: within-neuron-normalized population peristimulus time histogram, demonstrating fixation and presaccadic visual responses for neurons with no motor preparatory activity during the presaccadic epoch (Vis_VGS neurons). Conventions are as in A. D: sensitivity to fixation (blue) and presaccadic (red) probes for Vis_VGS neurons.

Time course of sensitivity reduction.

We also examined the time course of the reduction of sensitivity in Vis_VGS neurons and the whole population of neurons with visual responses to the full-field probe. Figure 6 shows the decline in mean presaccadic sensitivity relative to fixation at the mean times of probe-saccade interval quintiles (means ± SD: −127 ± 5.6, −103 ± 10.3, −84 ± 12, −62 ± 11, and −34 ± 5.4 ms). Across the population and in Vis_VGS neurons, suppression began 84 ± 12 ms before saccade onset (n = 47, two-sample t-test comparing fixation and presaccadic sensitivity values: all neurons, P = 0.014; Vis_VGS neurons, P = 0.0323). In VM_VGS neurons, sensitivity was significantly reduced in the earliest measured time bin. The decrease of sensitivity in the earliest time bin in VM_VGS neurons is consistent with the time course of motor preparatory activity (Fig. 5A; no-probe condition). The difference in time course for Vis_VGS and VM_VGS neurons indicates that whereas the early decreases in sensitivity may derive both from motor preparation and reduced visual responses, later decreases in sensitivity solely reflect changes in visual responses.

Fig. 6.

Time course of sensitivity reduction for Vis_VGS neurons (red) and VM_VGS neurons (gray). For each neuron, trials were divided into quintiles, according to the time from probe onset to saccade onset (probe-saccade interval), and presaccadic AUC was normalized to mean fixation AUC (blue, dashed line). Asterisks indicate results of two-sample t-tests comparing presaccadic sensitivity with fixation sensitivity at each time point (*P < 0.05; **P < 0.01; ***P < 0.001).

Presaccadic sensitivity is also reduced for equiluminant color stimuli.

We hypothesized that if sensitivity reduction in the FEF is inherited from its feedforward connections from visual cortex, then sensitivity to equiluminant color stimuli would be maintained presaccadically, reflecting the pattern observed in visual area 4 (V4) (Han et al. 2009). Therefore, in a subset of recordings (27 out of 37), we interleaved luminance-contrast trials with physically equiluminant red-green contrasting probes (Fig. 7A). In the Vis_VGS neurons that we recorded during these experiments, color stimuli evoked robust visual responses during fixation. However, presaccadic color responses of Vis_VGS neurons were suppressed relative to fixation (Fig. 7B). Fixation visual responses to color stimuli were not different from fixation visual responses to 32% contrast luminance-varying stimuli (two-sample t-test, P = 0.56). Reduction of sensitivity occurred for both 32% luminance contrast and color stimuli (Fig. 7C; repeated-measures ANOVA, luminance, P = 0.0087; color, P = 0.012). The magnitude of reduction of sensitivity to luminance and color stimuli was not different (Fig. 7D; two-sample t-test, P = 0.93). Thus, the presaccadic reduction of sensitivity that we observed in the FEF is not selective for luminance contrast.

Fig. 7.

A: physically equiluminant color stimuli were presented on randomly interleaved trials instead of luminance contrast-varied stimuli. Stimuli consisted of pseudorandomly scattered red-green square-wave gratings with 0.3 cycle/degree spatial frequency in Gaussian envelopes, spaced 5–8 DVA apart. B: visual responses evoked by color stimuli. C: population mean sensitivity to color and luminance (Lum) contrast stimuli presented during fixation (blue) and before saccades (red; repeated-measures ANOVA; *P < 0.05; **P < 0.01). D: presaccadic AUC relative to fixation AUC for luminance-contrast stimuli (x-axis) vs. color-contrast stimuli (y-axis).

DISCUSSION

We compared the visual sensitivity of FEF neurons during fixation and during the preparation of saccades and observed a presaccadic reduction in visual sensitivity in FEF neurons that impaired their ability to signal information about visual stimuli. The FEF therefore demonstrates a neural correlate of saccadic suppression. This reduction in sensitivity was present in MGS-identified visual neurons but not visuomotor neurons, although the population of visuomotor neurons exhibited a trend toward reduced sensitivity. Sensitivity was also reduced in neurons with and without movement-related activity during visually guided saccades. Presaccadic reduction in visual sensitivity in the FEF thus occurs independently of movement-related modulation in individual neurons. In addition, we observed that the reduced sensitivity began ∼80 ms before saccade onset and that it affected luminance and color responses equally.

Presaccadic reduction of sensitivity occurs independent of movement-related activity.

Past studies of neural correlates of saccadic suppression in the lateral intraparietal area and in the superior colliculus (SC) have also found evidence of perisaccadic reduction of visual responses (Bremmer et al. 2009; Robinson and Wurtz 1976). However, given the concurrent dynamic changes in motor and visual activity during the perisaccadic epoch in both lateral intraparietal area and the SC, it is unclear how much of the measured reductions in activity can be attributed to changes in visual responses per se. This problem is also evident in the FEF, and previous studies of perisaccadic activity modulation in the FEF have not distinguished between visual and motor components of perisaccadic FEF activity (Bruce and Goldberg 1985; Joiner et al. 2013). We therefore took multiple approaches to discerning whether the visual component of FEF responses exhibits a signature of saccadic suppression. First, we confirmed that the presaccadic reduction in sensitivity was present among neurons classified as visual using a standard MGS task. Second, we confirmed that the reduced sensitivity was also present among a subpopulation of Vis_VGS neurons. Thus presaccadic reduction of visual sensitivity occurs in FEF neurons even when they exhibited no motor preparatory activity.

The onset of decreases in sensitivity in FEF precedes decreases in sensitivity in extrastriate visual cortex.

Sensitivity of VM_VGS neurons began to decline in the earliest bin we measured (∼130 ms before saccade onset), consistent with the measured time course of motor preparation. In contrast, sensitivity of Vis_VGS neurons, as well as that of the overall population, began to decline ∼80 ms before saccade onset. Human psychophysical contrast sensitivity begins to decline ∼50–75 ms before saccade onset (Diamond et al. 2000; Knöll et al. 2011; Krock and Moore 2014). The time course of sensitivity reduction has also been measured to begin ∼50 ms before saccade onset in macaque extrastriate visual cortical areas V4, middle temporal (MT), and medial superior temporal (Bremmer et al. 2009; Han et al. 2009; Ibbotson and Krekelberg 2011; Ibbotson et al. 2007). Therefore, the time course of sensitivity reduction that we observed in FEF is early enough to account for the time course of reductions in human psychophysical contrast sensitivity and reductions in visual sensitivity in V4, MT, and the medial superior temporal area. Microsaccades have also been shown to suppress visual bursts in the SC beginning ∼70 ms before microsaccade onset (Hafed and Krauzlis 2010), similar to the onset of suppression of visual sensitivity that we observed in the FEF. The attainment of a precise comparison of onset times, however, will require simultaneous measurements of presaccadic sensitivity in FEF, visual cortex, the SC, and any other relevant brain regions in the same experimental paradigm with the same visual stimulation and with the same measures of visual sensitivity.

Color and luminance sensitivity are equally reduced.

Psychophysical experiments have established that sensitivity to equiluminant color stimuli is maintained during saccades, unlike for luminance-varying stimuli (Burr et al. 1994). Evidence has also been found for perisaccadic maintenance of color vision in macaque (Hass and Horwitz 2011). As a result of these observations, it has been hypothesized that the primary target of saccadic suppression is the color-insensitive magnocellular pathway of the lateral geniculate nucleus but that color representations in the parvocellular and koniocellular layers are spared (Burr et al. 1994). However, neurophysiological evidence has shown that magnocellular and parvocellular lateral geniculate nucleus neurons are equally affected by saccadic suppression, and both layers' response patterns are dominated by strong postsaccadic enhancement rather than presaccadic suppression (Reppas et al. 2002; Royal et al. 2006). Likewise, responses of L − M opponent and L + M nonopponent neurons in V1/V2 are similarly affected by microsaccades (Hass and Horwitz 2011). In macaque V4, on the other hand, color sensitivity is relatively spared compared with luminance sensitivity (Han et al. 2009). Many questions remain about the mechanism of the luminance specificity of saccadic suppression.

We wondered whether the FEF would exhibit a pattern of luminance-specific sensitivity reduction similar to V4. To determine whether the presaccadic reduction of sensitivity in the FEF exhibits luminance specificity, we compared presaccadic visual sensitivity to luminance vs. color stimuli. We found that color sensitivity decreased before saccades, and the magnitude of reduction of color sensitivity did not differ from that of luminance sensitivity. Thus color sensitivity in the FEF is not maintained before saccades, in contrast with V4 (Han et al. 2009). The reduction of color sensitivity in the FEF precludes the possibility that reduced FEF sensitivity is a simple readout of reduced sensitivity inherited via feedforward connections from V4. An alternative possibility consistent with our results is that reduction of sensitivity is generated within the FEF. It is also possible that reduced sensitivity in FEF is inherited via feedforward connections from another structure in which color sensitivity is reduced before saccades, although a structure with projections to FEF that exhibits reduction of sensitivity to both chromatic and achromatic stimuli has not yet been identified.

Source(s) of saccadic suppression.

Neural correlates of saccadic suppression have been observed in visual responses throughout the visual and saccadic systems, but the source of suppression in these areas is not known (Bremmer et al. 2009; Han et al. 2009; Hass and Horwitz 2011; Krock and Moore 2014; Robinson and Wurtz 1976; Thiele et al. 2002). Psychophysical and neurophysiological studies have furnished evidence for the involvement of an active, extraretinal-suppressive mechanism coupled to eye-movement planning, implicating the saccadic system (Burr et al. 1982; Diamond et al. 2000; Han et al. 2009; Robinson and Wurtz 1976). A simple model of a circuit for saccadic suppression might consist of an inhibitory signal that originates in a saccadic structure and directly suppresses visual representations in the visual system. Yet because long-range cortical connections are overwhelmingly excitatory, it seems unlikely that a saccade command could directly inhibit neurons in visual cortex. For example, ∼90% of the projections of FEF neurons to extrastriate area V4 terminate directly onto the dendritic spines of pyramidal neurons, making it unlikely that FEF neurons directly inhibit V4 neurons (Anderson et al. 2011).

Another way that saccadic structures could contribute to implementing saccadic suppression in visual cortex, however, is via a reduction in excitatory feedback. Congruent with this type of mechanism, Berman and Wurtz (2011) observed that a correlate of saccadic suppression is present in a pathway from the SC to MT via pulvinar relay neurons. This suppression began as early as 60 ms before saccade onset, slightly preceding suppression in MT (Bremmer et al. 2009; Ibbotson et al. 2007), consistent with a model in which pulvinar relay neurons transmit suppressed visual signals from the SC to the MT. Similarly, we observed a correlate of saccadic suppression in the visual activity of FEF neurons, beginning 80 ms before saccade onset; this early loss of sensitivity could be transmitted elsewhere in the saccadic system (e.g., the SC) or in the visual system (e.g., V4) via known excitatory connections. Given that we observed reduced sensitivity in visual neurons, we hypothesize that visual neurons transmit this reduction of sensitivity to V4. Future work will need to determine which functional classes of FEF neurons project to V4 and other extrastriate areas.

If saccadic suppression is transmitted from the saccadic system to visual cortex via excitatory projections, then the inhibition necessary to generate that suppression in the first place must take place within the saccadic system. For example, the reduction in sensitivity that we observed in the FEF could be the result of increased motor activity, either within the FEF or from an input area, such as the SC. Consistent with a role for the FEF itself in global presaccadic suppression of FEF activity, Schlag et al. (1998) found that microstimulation of one FEF hemisphere suppresses activity of neurons in the contralateral FEF hemisphere when the preferred saccade vector of the neurons does not match the evoked saccade vector. Similarly, microstimulation of the FEF can inhibit neurons in the SC (Schlag-Rey et al. 1992). Lateral inhibition is one possible circuit mechanism that could mediate this suppression. However, although the FEF is known to have large, suppressive surrounds (Cavanaugh et al. 2012), it is unclear whether lateral inhibition could account for the global extent of suppression. Bruce and Goldberg (1985) identified 8.6% of FEF neurons that exhibited untuned movement activity, and Sommer and Wurtz (1998) identified 10% of corticotectal FEF neurons with untuned visual and motor activity (and these neurons had no topographies, either of distribution in the FEF or projection across the SC); neurons with these properties could contribute to global suppression of presaccadic visual signals within the FEF or SC, directly or via inhibitory interneurons.

In future experiments, it will be critical to test the causal role of saccade vector representations in areas such as the FEF and the SC, in the reduction of visual sensitivity within the FEF, in extrastriate visual cortex, and in perception. Previous evidence suggests that FEF activity, generated via microstimulation, is sufficient to suppress V4 visual responses at retinotopically distant locations, even when no saccade occurs (Moore and Armstrong 2003). It remains unknown, however, whether saccade-generating activity in either the FEF or the SC is necessary for saccadic suppression of visual cortical representations or perception. Given that a correlate of saccadic suppression is present in the reduction of presaccadic visual sensitivity of FEF neurons, it seems likely that this reduction in sensitivity would be transmitted to the visual cortical areas that the FEF is known to modulate.

GRANTS

Support for this work was provided by the National Eye Institute (EY014924; to T. Moore) and the National Science Foundation (Graduate Research Fellowship Program; to R. M. Krock).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.M.K. and T.M. conceived and designed research; R.M.K. performed experiments; R.M.K. analyzed data; R.M.K. and T.M. interpreted results of experiments; R.M.K. prepared figures; R.M.K. and T.M. drafted manuscript; R.M.K. and T.M. edited and revised manuscript; R.M.K. and T.M. approved final version of manuscript.