Enhancement of motion perception in the direction opposite to smooth pursuit eye movement

Abstract

When eyes track a moving target, a stationary background environment moves in the direction opposite to the eye movement on the observer's retina. Here, we report a novel effect in which smooth pursuit can enhance the retinal motion in the direction opposite to eye movement, under certain conditions. While performing smooth pursuit, the observers were presented with a counterphase grating on the retina. The counterphase grating consisted of two drifting component gratings: one drifting in the direction opposite to the eye movement and the other drifting in the same direction as the pursuit. Although the overall perceived motion direction should be ambiguous if only retinal information is considered, our results indicated that the stimulus almost always appeared to be moving in the direction opposite to the pursuit direction. This effect was ascribable to the perceptual dominance of the environmentally stationary component over the other. The effect was robust at suprathreshold contrasts, but it disappeared at lower overall contrasts. The effect was not associated with motion capture by a reference frame served by peripheral moving images. Our findings also indicate that the brain exploits eye-movement information not only for eye-contingent image motion suppression but also to develop an ecologically plausible interpretation of ambiguous retinal motion signals. Based on this biological assumption, we argue that visual processing has the functional consequence of reducing the apparent motion blur of a stationary background pattern during eye movements and that it does so through integration of the trajectories of pattern and color signals.

Introduction

When we perform smooth pursuit, an environmentally stationary visual scene inevitably moves on the retina in the opposite direction. Studies have shown that motion perception for such retinal slip, produced by pursuit, goes more or less unnoticed to preserve the perceptual stability of the stationary background. The velocity of the retinal image of the background is partially canceled through visual processing by taking into consideration biological signals for current eye velocity (for a review, see Goldstein, 2007; Spering & Montagnini, 2011). In addition, the contrast sensitivity for stimuli moving in the direction opposite to pursuit, and the perceptual saliency of concomitant motion streaks, are reduced during smooth pursuit compared with the same stimuli, retinally, but viewed during fixation (Bedell & Lott, 1996; Schütz et al., 2007). 

In contrast to these previous findings, a recent study suggested the possibility that motion signals for such retinal slip produced by smooth pursuit can also be useful. The chromatic critical fusion frequency during pursuit is higher when color changes belong to a motion trajectory originating from a stationary background than when they belong to one in the same direction as that of smooth pursuit (Terao, Watanabe, Yagi, & Nishida, 2010). Consistent with studies demonstrating that pattern and color signals are integrated along the motion trajectory to improve the chromatic critical fusion frequency (Burr, 1981; Burr & Ross, 1986; Nishida, 2004; Nishida, Watanabe, Kuriki, & Tokimoto, 2007), this effect shows that smooth pursuit can enhance signal integration along the motion trajectory in the direction opposite to smooth pursuit, and thus suggests that smooth pursuit may, in fact, enhance motion-related signals in the direction opposite to smooth pursuit, in useful ways. 

A counterphase grating is one of the most suitable tools for investigating the strength of motion-related signals. Mathematically, a counterphase sinusoidal grating (i.e., a grating whose luminance is modulated around the mean luminance as a sinusoidal function of space and time) is equivalent to the linear sum of two spatially sinusoidal gratings drifting opposite each other at the same speed. It is well known that, during fixation, the overall perceived direction of a counterphase grating depends on the strength ratio of the motion signals conveyed by the two components. When the strength ratio is unity, the perceived direction is ambiguous with no net motion; however, when the strength ratio is sufficiently large, the stimulus appears to move in the direction of the component with the higher strength. Note that “strength” is often quantified by the modulation amplitude (i.e., luminance contrast); however, it can also depend on the state of our visual system, such as the observer's perceptual set and priming (e.g., Gorea & Lorenceau, 1984; Qian, Andersen, & Adelson, 1994; for a review, see Nishida, 2011). If smooth pursuit biologically enhances the motion signal in the direction opposite to eye movements, we may well expect that smooth pursuit will accordingly bias motion perception for a counterphase grating rendered on the moving retina. Therefore, in this study, observers were presented with a counterphase grating in retinal coordinates while performing smooth pursuit. The stimulus consisted of two drifting components, one drifting in the opposite direction and the other drifting in the same direction as that of the smooth pursuit. When the luminance contrast was unity, the counterphase grating was almost always perceived as opposite to smooth pursuit. This directional bias was nulled by reducing the luminance contrast of the component grating drifting in the direction opposite to that of smooth pursuit. 

Experiment 1: A retinally counterphase grating appears to move in the direction opposite to that of pursuit

Methods

This section describes the methods used in Experiment 1. Specific changes made in Experiments 2 and 3 will be described later. 

Observers

The observers in this study included two of the authors (M. T. and S. N.) and two volunteers who were unaware of the purpose of the experiments. All observers had normal or corrected-to-normal vision. All observers provided written, informed consent prior to the experiments. The experiments and the consent form were approved by the NTT Communication Science Laboratories Research Ethics Committee. 

Apparatus

Visual stimuli were displayed on a cathode ray tube (CRT) monitor (GDM-F520, Sony, Inc., Tokyo, Japan) at a refresh rate of 120 Hz. The monitor was connected to a visual stimulus generator (VSG2/5, Cambridge Research Systems, Rochester, UK) installed in a workstation (Dell Precision 350, Dell, Inc., Round Rock, TX). The monitor was calibrated with a colorimeter (Color Cal, Cambridge Research Systems) to linearize the gamma relationship. The spatial resolution of the monitor was 800 × 600 pixels, with each pixel subtending 1.5 min at a viewing distance of 113 cm. The observer sat with his or her head fixed on a chin rest and viewed the display binocularly. The room had no illumination except for the stimulus presented on the CRT. A keyboard was placed in front of the observer to record responses. The movements of the dominant eye were monitored at 500 Hz with a video-based eye tracker (EyeLink II, SR Research, Ltd., Ottawa, ON, Canada); analog output data from the eye-tracker system were recorded on a disk for off-line analysis using a data-acquisition system (NR-2000; Keyence) at a 1 kHz sampling rate. To synchronize the timing of the stimulus presentation and eye data with high precision, we chose a higher re-sampling rate than the original sampling rate of the eye tracker, and in the off-line analysis the resampled eye position time series was low-pass-filtered using a Butterworth filter with a cut-off frequency of 30 Hz. 

Stimulus

The stimulus consisted of a white rectangular (6 × 60 min, 136 cd/m²) tracking target and a vertically oriented counterphase flickering grating, that were generated by superimposing two sinusoidal gratings moving in opposite directions on the retina. The grating subtended 20° in width and 1° in height, with a spatial frequency of 0.33 cpd and a temporal frequency of 4 Hz on the retina. The grating was located above the tracking target with a center-to-center distance of 1.5° (Figure 1A). The background was a uniform gray field at the mean luminance (68 cd/m²). The grating was presented within a Gaussian temporal window; the full-width at half-maximum (FWHM) of the temporal window was 500 ms. 

Figure 1

 

Experimental setup. (A) Schematic spatial configuration of the grating and the pursuit target. (B) Space–time plot of two components of the counterphase grating. The white arrow indicates the gaze position over time. During fixation, the spatiotemporal stimulus configuration on the screen and that on the retina were identical; the counterphase grating and its components under this fixation condition are shown in the upper panel. During smooth pursuit, one component was static on the screen, whereas the other component moved twice as fast as the pursuit target on the screen (bottom panel). As a result of smooth pursuit, both components under this pursuit condition drifted in mutually opposite directions at the same speed on the retina (upper panel), rendering a counterphase grating at this location.

A counterphase grating, the equivalent of the linear sum of two spatially sinusoidal gratings drifting opposite to each other at the same speed, was used to render the target stimulus on the retina under both fixation and pursuit conditions. Under the fixation condition, observers did not move their eyes, and, on the screen, one grating drifted leftward at a speed of 12.17°/s, whereas the other grating drifted rightward at the same speed (Figure 1B, upper row). Under the pursuit condition, one grating was environmentally stationary (i.e., on the screen) and the other grating was environmentally drifting in the pursuit direction at a speed twice as fast as the target speed (Figure 1B, bottom row). This type of stimulus is sometimes referred to as unikinetic plaid. As the observer performs smooth pursuit to track the horizontally moving target, this environmentally unikinetic plaid is rendered on the retina as a counterphase grating, given perfect tracking (Figure 1B, upper row). 

The physical balance in motion signals between the two gratings was manipulated by changing their luminance contrast ratio. The overall luminance contrast of the counterphase grating was held constant at 25%. The contrast ratio was randomly changed within each experimental block. 

Procedure

Three conditions were tested. Under the fixation condition, the grating appeared on the screen in the midst of a fixation period lasting for 1.6 s during which the observer maintained fixation on a central stationary target. After the fixation target disappeared, the observer judged the motion direction of the grating in response to a two-alternative, forced-response question (leftward or rightward). Under the rightward-pursuit condition, a key press by the observer initiated a horizontal movement of the pursuit target at a constant speed of 12.17°/s. Within 1.6 s, the target moved 20° from the left end to the right end of the CRT screen. The observer was asked to pursue the tracking target as precisely as possible. When the pursuit target arrived around the horizontal center of the screen, the counterphase grating was presented, with the Gaussian temporal window centered at the time at which the target reached the center of the screen. After the tracking target disappeared, the observer judged the motion direction of the grating by pressing one of two computer keys corresponding to the two alternatives, “leftward” or “rightward.” It should be noted that observers were not given any instructions regarding specific coordinates (retinal or environmental). In preliminary experiments, we confirmed that observers performing smooth pursuit reliably reported the direction of each of the two component gratings as it was presented alone. Thus, as expected, when the grating moved in the pursuit direction, it appeared to move “rightward” during rightward pursuit. Likewise, when the grating moved in the direction opposite to the pursuit direction, it appeared to move “leftward.” We assumed that the observers behaved based on a similar criterion in response to the counterphase grating in the main experiment as well. The procedure used under the leftward-pursuit condition paralleled that used under the rightward-pursuit condition, with the exception that all the stimuli were mirror reversed. The three eye-movement conditions were tested in separate blocks presented in random order separated by an interblock rest interval of at least 1 min. 

Data analysis

The proportion of seeing rightward motion was plotted as a function of the contrast ratio (the ratio of the luminance contrast of the rightward moving grating to that of the leftward moving grating based on retinal coordinates). A psychometric function was derived by fitting the cumulative Gaussian function to the data using the constrained maximum likelihood method. We determined the luminance contrast that yielded 0.5 in the ordinate as the point of subjective directional ambiguity for each observer under each condition. The 95% confidence interval was computed with the bootstrap method and 20,000 iterations using the software Psignifit toolbox for MATLAB (Wichmann & Hill, 2001a, 2001b). 

Calibration of the eye tracker was performed at the beginning of each experimental block. Eye velocity was obtained by digital differentiation of the eye position over time. Trials were excluded from further analysis if the pursuit gain deviated from 100 ± 20%. Trials were also excluded if they contained eye velocities exceeding 30°/s, which indicated catch-up saccades. We collected data until at least 12 (typically 24) trials were obtained within the above eye-velocity criteria for each sampling point along the psychometric function. The median eye velocity of the trials used for analysis and the quartile deviation were 12.47 ± 0.64°/s, respectively. 

Results

Figure 2 shows the results for each observer in separate panels. The physical balance in motion signal was quantified by calculating the logarithm of the contrast ratio (hereinafter referred to as the log contrast ratio). Under the fixation condition, the perceived direction was ambiguous when the log contrast ratio was 0 and was gradually predominated by the stronger component as the contrast ratio became smaller or larger than 0. Under the pursuit conditions, the perceived direction was predominated by the stronger component when the log contrast ratio was sufficiently smaller or larger than 0, similar to the fixation condition. However, the perceived direction at the log contrast ratio of 0 was primarily rightward and leftward under the leftward and rightward pursuit conditions, respectively, and thus nearly always opposite to the pursuit direction. Additionally, the psychometric functions for these pursuit conditions significantly shifted toward the direction in which the grating that was moving in the same direction as that of the pursuit had a higher contrast. Under the leftward and rightward pursuit conditions, the log contrast ratios at the point of subjective directional ambiguity were roughly −0.2 and +0.2, respectively. Given that the overall luminance contrast of the counterphase grating was always 25%, the point of subjective directional ambiguity was established by having the luminance contrast of the component grating that moved in the direction opposite to the pursuit direction ∼ 6% lower than that of the component grating that moved in the pursuit direction. These results were consistent across all observers. 

Figure 2

 

Results of Experiment 1. The perceived direction of motion is plotted as a function of the contrast ratio (ratio of the luminance contrast of the rightward moving grating to that of the leftward moving grating). Data were obtained with rightward smooth pursuit, leftward smooth pursuit, and fixation. Positive values of the abscissa indicate that the luminance contrast of the rightward moving grating was higher than that of the leftward moving grating. Each horizontal bar indicates a 95% confidence interval by the bootstrapping method. Individual panels correspond to each of the four observers.

Because in some situations a motion stimulus perturbs the performance of smooth pursuit (e.g., Spering & Gegenfurtner, 2007), one might question how well the counterphase grating in our experiment was rendered on the retina. We traced instantaneous eye velocity and its standard deviation for the trials that were included in the psychophysical data analysis. Instantaneous eye velocity was obtained by digital differentiation of eye position signals with respect to time. The data for leftward pursuit were flipped and merged with the data for rightward pursuit, as if the observer was always performing rightward pursuit. Figure 3A shows the averaged eye trace of this “rightward” pursuit for each observer, as well as the grand average across all four observers, plotted from 250 ms before the stimulus onset to 250 ms after the stimulus offset. Figure 3B depicts the standard deviation of eye velocity for each observer and the grand average. Data plotted here are based on the trials at the contrast ratio of 0, but essentially the same oculomotor patterns were confirmed for other contrast ratios. Instantaneous eye velocity was similar to the tracking target velocity, and there was no systematic deceleration or acceleration in response to the stimulus presentation (Figure 3A). Standard deviation during stimulus presentation was not larger than that before/after stimulus presentation (Figure 3B). Thus, smooth pursuit was not perturbed by stimulus presentation, and the counterphase grating on the retina had no systematic distortion. 

Figure 3

 

Oculomotor data around the motion-stimulus presentation, during which the eyes were supposed to be tracking the target horizontally moving at uniform velocity. (A) Instantaneous eye velocity and (B) its standard deviation are plotted as a function of time. Each colored curve represents the eye trace of each observer, and the black curve represents the grand average. The dashed line indicates the tracking target velocity. The shaded area indicates the period of stimulus presentation.

Experiment 2: The effect is eliminated at lower overall contrasts

In contrast to the above finding, a previous study found that the contrast sensitivity for a luminance grating was reduced for stimuli moving retinally opposite to the direction of smooth pursuit compared with stimuli moving retinally in the same direction (Schütz et al., 2007). Accordingly, Schütz et al. (2007) predicts that retinally ambiguous opposing motions would appear to move predominantly in the pursuit direction; however, we found the reverse. This contradiction may be attributed to the difference in stimulus contrast. Whereas suprathreshold contrast stimuli were used in the present study, Schütz et al. (2007) measured the detection threshold contrast. One possible way to reconcile our findings with those of Schütz et al. is to assume that the directional bias in perceptual judgment during smooth pursuit depends on the contrast. 

To address this issue, we conducted two experiments that examined how stimulus contrast affects motion perception during smooth pursuit. First, we determined the perceived direction of a retinally counterphase grating during pursuit across a range of grating contrasts. Next, we measured the detectability of a drifting grating in two directions, in the same direction as and in the direction opposite to that of the pursuit direction, across the same range of grating contrasts. 

Methods

In terms of the measurement of the perceived direction of the counterphase grating, observers were asked to judge its direction of motion during pursuit, following the same protocol as in Experiment 1, with the exception that we manipulated the overall contrast of the stimulus, with its contrast ratio fixed at 1. The overall contrast was randomly changed within each experimental block. Two pursuit directions (leftward and rightward) were tested in separate blocks. The same observers who took part in Experiment 1 participated. 

With regard to the detectability measurements of the drifting grating, we presented a unidirectional drifting grating that moved on the retina in either the same or in the opposite direction relative to the pursuit direction. The same stimulus parameters as in Experiment 1 were used, with the exception of the following: We presented the drifting grating at either 1.5° above or 1.5° below the tracking target, manipulated the contrast of the grating, and asked observers to make a two-alternative, forced-choice judgment about the position of the grating (above or below). The contrast and direction of the grating were randomly changed within each experimental block. Two pursuit directions (leftward and rightward) were tested in separate blocks. 

Results

Figure 4 shows the effects of the overall contrast on the perceived direction of the counterphase grating during smooth pursuit and on the detection rate for the unidirectional drifting as measured by the rate of correct response. Although we collected data for both rightward and leftward pursuit directions, for clarity we flipped the data for the rightward pursuit direction and merged them with those for the leftward pursuit direction, as if the observer were always making “leftward” pursuit while viewing either the “leftward” (i.e., same direction) or “rightward” (i.e., opposite direction) unidirectional grating. The point at 25% grating contrast was the same as the point at 0 log contrast ratio in Experiment 1 (Figure 2). In response to a reduction in the overall contrast, the perceived direction became ambiguous or even reversed, rendering it the same as the pursuit direction, among some observers. Consistent with this reduction in the effect at low contrast, the detectability of the unidirectional grating in the direction opposite to that of pursuit decreased as the contrast was reduced, becoming worse than the detectability of the unidirectional grating in the same direction, which replicates the results reported by Schütz et al. (2007). In contrast, the detectability of the unidirectional grating in the same direction changed only slightly across the tested contrast range and tended to saturate at perfect detectability. 

Figure 4

 

Effects of stimulus contrast on the perceived direction of a counterphase grating during smooth pursuit and on the contrast detectability for a unidirectional grating. (A) Proportion of seeing the “rightward” direction in the counterphase grating during “leftward” pursuit plotted as a function of overall contrast. Data for the rightward pursuit direction were flipped and merged into those for the leftward pursuit direction, as if the observer were always making “leftward” pursuit. The rightmost point is identical to the data point in Figure 2 at a log contrast ratio of 0. Each panel corresponds to each observer. (B) Detection rate of a unidirectional grating as a function of overall contrast. Data for the rightward pursuit direction were flipped and merged into the those for the leftward pursuit direction, as if the observer were always making “leftward” pursuit while viewing either the “leftward” (red symbols) or “rightward” (green symbols) unidirectional grating. Each panel corresponds to each observer.

These results indicate that smooth pursuit affects the processing of motion-related signals in two different ways, depending on the level of overall contrast; that is, it reduced the detectability of the stimulus moving in the opposite direction at low contrast (Schütz et al., 2007) and enhanced the strength of the subjective motion of the stimulus moving in the opposite direction at contrasts well above threshold. 

Experiment 3: The effect is not the result of motion capture by a larger frame of reference

In the above experiments, we presented grating stimuli in a rectangular window spanning the full width of the monitor screen (Figure 1A). When the retinal images for the fixation and pursuit conditions were compared, the window movement differed, although the pattern inside the window was the same; that is, the side boundaries of the window (and those of the gray rectangular background) were stationary under the fixation condition but moving on the retina concomitantly in the direction opposite to the eye movement under the pursuit condition. It is known that the perceived direction of ambiguous motion is sometimes biased toward nearby unambiguous image motion (e.g., Murakami & Shimojo, 1993; Ramachandran & Cavanagh, 1987). Given the existence of such an effect of motion capture, one may suspect that it was the retinal motion of the window boundaries rather than the eye movement per se that induced the motion perception in the counterphase grating during pursuit. To test this possibility, the next experiment presented a retinally counterphase grating within a smaller window that moved with the pursuit on an otherwise dark background. 

Methods

The methods were identical to those in Experiment 1, with the exception of the following. The background was completely dark (<0.2 cd/m²). We presented a gray horizontally elongated box (10° wide and 1° high) above a tracking target. Under the pursuit condition, the gray box moved horizontally with the tracking target. The target moved for a distance of 10° from left to right or from right to left. When the target came around the horizontal center of the screen, the gray box changed to a counterphase grating with the same spatiotemporal properties as those in Experiment 1. The display under the pursuit condition is schematized in Figure 5A. Under the fixation condition, we mimicked what would occur on the retina under the above condition: A counterphase grating appeared within the region that the stationary gray box had occupied while the observer was fixating on a central stationary target. The observers in this study were two of the authors (M. T. and S. N.) and a new volunteer who had not participated in Experiments 1 and 2. 

Figure 5

 

Method and results of Experiment 3. (A) Schematic spatial configuration of the display for the rightward-pursuit condition. The black arrow indicates the order of the temporal sequence in each trial. A retinally counterphase grating was presented at the middle of the sequence within the region of the gray box. Only three representative frames are illustrated; however, the stimuli were actually moving smoothly on the screen. (B) Perceived direction of motion. The conventions are identical to those in Figure 2. Individual panels are shown for each observer.

Results

The results are shown in Figure 5B. The pattern of each psychometric function is very similar to the results of Experiment 1 (Figure 2). Even though the current procedure excluded window motion in retinal coordinates, we replicated the original effect of directional bias. Therefore, our effect was not the result of motion capture by the movement of the stimulus window or of a larger background context serving as a frame of reference. 

General discussion

To clarify how smooth pursuit may enhance motion perception, we presented observers with a retinally rendered counterphase grating and asked them to report in which direction it appeared to move. Experiment 1 demonstrated that this stimulus usually appeared to move in the direction opposite to that of smooth pursuit. This effect was eliminated by reducing the relative contrast of the component grating moving opposite to the pursuit by ∼ 6%. Experiment 2 revealed that this effect required sufficiently high overall contrast and that the perceived direction became ambiguous or even reversed at near-threshold contrast. Experiment 3 confirmed that the effect could not be ascribed to motion capture by a visible reference frame served by peripheral images, such as the stimulus window and the border of the uniform background. These results indicate that smooth pursuit exerts a facilitative influence on suprathreshold motion perception in the direction opposite to that of eye movement. 

One might consider that our findings are related to Kato's (2007) finding that a dynamic random-dot pattern appears to move in the direction opposite to smooth pursuit. According to our pilot observation, Kato's illusion requires an environmentally stationary boundary. When a boundary demarcating the random-dot pattern moves in the same direction and at the same speed as a pursuit target, the illusion does not occur. This observation suggests that Kato's illusion stems from motion capture by a retinal motion of the stimulus boundary, rather than from the eye movement per se. Thus, it is most likely that our effect and Kato's illusion have different mechanisms. 

It is known that attention can affect the perceived direction of an ambiguous motion (e.g., Cavanagh 1992; Culham, Verstraten, Ashida, & Cavanagh, 2000; Verstraten, Cavanagh, & Labianca, 2000; Wertheimer, 1961). Attention is necessary for target selection, pursuit initiation, and pursuit maintenance (for a review, see Kowler, 2011; Schütz, Braun, & Gegenfurtner, 2011). In each of these phases, however, the spatial location of attention must be directed either precisely at the target location or ahead of the target trajectory. In our experiments, observers were able to keep tracking the moving target, with good accuracy and precision, even when the grating stimulus was presented on the display. Thus, at least overt attention was maintained high, at or around the gaze point, even though it moved together with the pursuit target. Given that the spatial location of attention moves together or ahead of the pursuit target and that an attended part of stimuli appears more conspicuous, any attentional effects would render the perceived direction of the retinally counterphase grating either ambiguous or biased toward the pursuit direction. However, our data demonstrated the opposite, indicating that our effect was presumably caused by smooth pursuit per se. 

The phenomenon in which the opposite direction is seen during pursuit may sound reminiscent of the classical Filehne illusion, in which an environmentally stationary pattern appears to move slightly backward during smooth pursuit (Filehne, 1922). The Filehne illusion is viewed as a product of an imperfect velocity compensation between the retinal and extraretinal velocity signals to compensate for spurious image motions originating from smooth pursuit; this computation is believed to occur at a later stage of visual motion processing that is devoted to external object velocity estimations (for a review, see Furman & Gur, 2012; Goldstein, 2007; Spering & Montagnini, 2011). 

Although the observation of motion in our experiments may involve the same velocity compensation process as that underlying the Filehne illusion, the process is unlikely to be responsible for our effect. In our experiment, one grating was stationary on the screen, and the other was drifting at a speed twice as fast as the pursuit target. Thus, if the retinal velocity of each component grating was compensated for by a biologically estimated eye velocity, the resulting subjective velocity of the one moving retinally in the opposite direction would have been much slower (ideally zero at perfect compensation) than that of the one moving in the pursuit direction, with the compensation serving as an estimate of the velocity in the external world. When faster and slower motions are represented simultaneously, it is unlikely that the slower one is always chosen as the dominant direction. In fact, we performed preliminarily tests to determine whether our observers showed any perceptual biases in their judgments of the direction of slower motion as the dominant direction when they observed a superposition of two gratings that were drifting in mutually opposite directions at different speeds; such biases were not observed. 

In our experiment, the retinal speeds of the two component gratings of the retinally counterphase grating were the same as the speed of the pursuit target on the display. One would naturally wonder whether the perceptual enhancement we observed is selective for one particular retinal motion or is more broadly tuned. Regarding this concern, we conducted a preliminarily investigation into what would happen when the speeds of the component gratings were changed, and found that the direction, opposite pursuit, was always seen, even at a doubled or halved component speed compared with the target speed. The main effect reported here, i.e., prioritizing motion in a direction opposite to pursuit, may not be strictly selective for one particular retinal speed equal to the pursuit speed, but may be broadly tuned in the speed domain. This observation does not contradict our interpretation of the main results because it is possible that the slow-motion prior is broadly tuned and that the slower one is always chosen as the dominant direction when faster and slower environmental motions are present simultaneously at the same place. We are currently conducting a systematic investigation to resolve this issue. 

The preference for the antipursuit—and in effect environmentally slower—visual motion may also serve a useful function for perceptual stability, because this preference may facilitate the default interpretation of directionally ambiguous retinal events as belonging to an untracked stationary background. Retinal motion signals are often noisy and ambiguous, particularly during eye movements, because they add extra spatiotemporal variations to the retinal image. These variations, mixed with the low temporal resolution of the visual system, would result in severe motion blur in an environmentally stationary background (e.g., Burr, 1981). However, we seldom perceive blurred images during smooth pursuit. Additionally, perceived motion smear is less severe during smooth pursuit than when comparable retinal motion is presented during fixation (e.g., Bedell & Lott, 1996). These factors indicate that the visual system must have some special way to prevent motion blur/smear from occurring perceptually during smooth pursuit. A possible way to counteract the detrimental effects of retinal motion may be the contrast sensitivity elevation for chromatic and high-spatial frequency modulations during pursuit (Schütz, Braun, & Gegenfurtner, 2009; Schütz, Braun, Kerzel, & Gegenfurtner, 2008). The other proposed mechanism for reducing motion blur perception involves the neural integration of visual signals along the trajectory of moving objects (Burr, 1981; Burr & Ross, 1986; Nishida, 2004; Nishida et al., 2007), because this integration is expected to prevent signal mixture among spatially adjacent inputs that would otherwise cause image degradation (Terao, Watanabe, Yagi, & Nishida, 2011; Watanabe & Nishida, 2007). The effective enhancement of visual signals from an environmentally stationary, though retinally moved, background may facilitate visual signal integration along the retinal motion trajectory originating from the background pattern. Given that a background pattern usually occupies most of the visual field, with this information serving as a frame of reference, we may be able to obtain true velocities of other objects and, accordingly, an impression of clear vision over the entire visual field. 

A preference for slow motion is conceptually consistent with computational theories that have been proposed in other contexts to stress that motion computation has a slow motion preference (Weiss, Simoncelli, & Adelson, 2002) and that the visual system has a stable-world assumption during self-motion (for a review, see Wexler & van Boxtel, 2005). Recently Freeman, Champion, and Warren (2010) successfully extended the above concept of slow motion preference to speed computation during smooth pursuit. These theories, together with our findings, suggest that the slow-motion prior is widely used in different aspects of motion processing, such as speed computation and motion selection. 

Conclusions

In conclusion, this study demonstrated for the first time that smooth pursuit can enhance motion perception in the direction opposite to that of eye movement. Our findings reveal a novel form of the modulation of motion perception by smooth pursuit; that is, smooth pursuit affects direction selection in such a way that is independent of the classically investigated object velocity estimate during smooth pursuit. Additionally, the results of this study suggest that motion perception is biased toward a spatiotemporal trajectory that is presumably stationary in the world. This preference may be used to help integrate visual signals along the motion trajectory of an environmentally stationary background pattern to reduce image degradation. 

Acknowledgments

This research was supported by the JSPS Funding Program for Next Generation World-Leading Researchers (LZ004) and a JSPS Grant-in-Aid for Scientific Research on Innovative Areas (25119003) to IM. A substantial proportion of this study was carried out while MT was a visiting researcher of NTT Communication Science Laboratories under the support of Japan Society for the Promotion of Science Research Fellowships for Young Scientists. We are grateful to Dr. Hsin-Ni Ho for comments on the manuscript. 

Commercial relationships: none 

Corresponding author: Masahiko Terao.