Effect of light level on the reference frames of visual motion processing

Abstract

It is empirically known that some action-related visual tasks, which may rely on the construction of spatiotopic coordinates, are not well conducted under mesopic vision. The aim of this study was to clarify the effect of light level on the reference frame, such as retinotopic and spatiotopic coordinate bases, associated with visual motion processing. For this purpose, we used a phenomenon called visual motion priming in which the perceived direction of a directionally ambiguous test stimulus is influenced by the moving direction of a priming stimulus. Previous studies have shown that negative and positive motion priming are conspicuously observed in retinotopic and spatiotopic coordinates, respectively. In the experiments, participants made a saccade after the termination of the priming stimulus and judged the perceived direction of the test stimulus presented subsequently in retinotopic or spatiotopic coordinates at different light levels. We found that in retinotopic coordinates, negative motion priming was observed at all light levels. In spatiotopic coordinates, positive motion priming was observed at photopic and scotopic light levels, whereas the strength of motion priming was greatly reduced at mesopic light levels. These results were robust to the change in the luminance contrast or the saccadic eye movement per se. Different spatiotemporal properties of cones and rods at mesopic light levels may disturb the construction of a spatiotopic representation of motion, which leads to the disappearance of visual motion priming in spatiotopic coordinates during mesopic vision.

Introduction

In natural environments, ambient light levels may change between day and night by a factor of 10¹¹ (Hood & Finkelstein, 1986; Stockman & Sharpe, 2006). However, the dynamic range of constituent neurons of the visual system is limited to approximately 10² (Wandell, 1995). The visual system deals with this by switching between two different types of photoreceptors as follows: cones, which function at higher levels of illumination, and rods, which function at lower levels of illumination. Photopic, mesopic, and scotopic regions are defined according to whether cones operate alone, rods and cones operate together, or rods operate alone, respectively. In our daily lives, scotopic and mesopic vision extends over an illumination range as wide as 10⁶. Therefore, an understanding of the characteristics of visual functioning at low light levels is very important from both scientific and practical perspectives (Hess, 1990; Hess, Sharpe, & Nordby, 1990). 

Rods preferentially project into the magnocellular lateral geniculate nucleus (LGN) layers (Purpura, Kaplan, & Shapley, 1988). Because motion-selective areas such as the middle temporal (MT) area receive dominant inputs from the magnocellular LGN layers, visual motion processing could be selectively influenced by rod-based inputs (Hadjikhani & Tootell, 2000; Maunsell, Nealey, & DePriest, 1990; Maunsell & van Essen, 1983). Human motion perception indeed changes as light intensity decreases. Velocity perception (Gegenfurtner, Mayser, & Sharpe, 2000; Hammett, Champion, Thompson, & Morland, 2007; Pritchard & Hammett, 2012; Vaziri-Pashkam & Cavanagh, 2008), velocity discrimination thresholds (Takeuchi & De Valois, 2000), short-range motion perception (Dawson & Di Lollo, 1990), complex-motion perception (Billino, Bremmer, & Gegenfurtner, 2008), biological-motion perception (Billino et al., 2008; Grossman & Blake, 1999), perception of static-motion illusions (Hisakata & Murakami, 2008), perception of interstimulus interval (ISI) reversal (Sheliga, Chen, FitzGibbon, & Miles, 2006; Takeuchi & De Valois, 1997, 2009; Takeuchi, De Valois, & Motoyoshi, 2001), perception of two-stroke motion (Challinor & Mather, 2010; Mather & Challinor, 2009), the coherent-motion threshold (Billino et al., 2008; Lankheet, van Doorn, & van de Grind, 2002; van de Grind, Koenderink, & van Doorn, 2000), moving texture segregation (Takeuchi, Yokosawa, & De Valois, 2004), and visual motion priming (Takeuchi, Tuladhar, & Yoshimoto, 2011; Yoshimoto & Takeuchi, 2013) have all been shown to vary with the light level. 

Outside the laboratory, we empirically know that motion perception could be affected by ambient light level, particularly at the mesopic light level. For example, the number of traffic accidents greatly increases at dusk both in the United States (Owens, Wood, & Owens, 2007) and in Japan (Yamadaya, 2010). According to a report from Japan's Metropolitan Police Department (Yamadaya, 2010), the deterioration in visual motion sensitivity that has been observed at dusk results from the cone–rod interaction and could be a possible explanation for the increase in the traffic accident rate at dusk. However, there is no experimental evidence for this. Another example is related to ball games such as baseball. It is known that a baseball outfielder is more likely to lose sight of a fly ball in the twilight sky and commit errors, both in actual games and in experimental settings (Oudejans, Michaels, Bakker, & Davids, 1999). A professional baseball outfielder, Ichiro Suzuki of the New York Yankees, said, “I just hope no fly ball comes in the twilight zone” (Tsugawa, 2007). These two examples are particularly interesting since most previous laboratory studies have shown an invariance of motion perception under various light levels or a deterioration of motion sensitivity at the darkest light levels examined, such as a scotopic level. Only one study by Billino et al. (2008), who estimated the detection sensitivity of biological motion, has indicated a specific loss of motion sensitivity at the mesopic light level. 

One common aspect of the two abovementioned examples under mesopic vision is their close relationship to the action or body movement. During vision-related actions such as driving or catching a ball, the eye, head, and body movements induce retinal image changes in a complex manner. In early visual areas such as V1, input information is encoded retinotopically (e.g., Wandell, Brewer, & Dougherty, 2005). However, to accomplish the abovementioned tasks successfully, a mechanism that allows a nonretinotopic representation to be encoded in spatiotopic coordinates based on the spatial layout of the external world must exist (Burr & Morrone, 2011, 2012; Cavanagh, Hunt, Afraz, & Rolfs, 2010; Wurtz, 2008). Burr and Morrone (2012) have argued that the function of the spatiotopic coordinates may be connected predominantly with action rather than perception and guide our physical interaction with the world. Even when visual information is unavailable, we can manage to perform actions such as pointing quite precisely after conducting body movements (Byrne, Becker, & Burgess, 2007). As summarized by Land (2012), previous studies have suggested that an internal model of the external world should be constructed and updated as we move through the world to make proper actions. Based on the empirical observations regarding driving or catching, we speculate that visual motion representation, particularly in spatiotopic coordinates, may be influenced by light level, presumably at the mesopic light level. Although many aspects of motion perception at low light levels have been studied, the coordinates associated with motion mechanism (i.e., the reference frames of motion processing) have not been investigated to date. The present study aimed to examine the effect of light level on the reference frames of visual motion processing. 

We examined the perception of visual motion priming on the basis of both retinotopic and spatiotopic coordinates under different ambient light levels. Visual motion priming is a phenomenon in which the perceived direction of a directionally ambiguous stimulus is influenced by the movement direction of the preceding stimulus. Visual motion priming has been used to examine how spatiotemporally separate inputs are integrated to induce a coherent motion perception in the underlying motion mechanisms (Kanai & Verstraten, 2005; Pantle, Gallogly, & Piehler, 2000; Pavan, Campana, Maniglia, & Casco, 2010; Pinkus & Pantle, 1997; Yoshimoto & Takeuchi, 2013). Pantle et al. (2000) used a moving sine-wave grating and showed that a subsequently presented directionally ambiguous test pattern, which was made of a 180°-shifted grating, was perceived to move in the same direction as the priming stimulus (positive motion priming) or in the opposite direction of the priming stimulus (negative motion priming). The two priming effects, either positive or negative motion priming, are observed depending on the stimulus parameter such as the presentation duration, velocity, luminance contrast, and presentation location of the priming stimulus, as described later (Kanai & Verstraten, 2005; Pantle et al., 2000; Takeuchi et al., 2011; Yoshimoto & Takeuchi, 2013). In our previous study (Yoshimoto, Uchida-Ota, & Takeuchi, 2014), we estimated the effect of visual motion priming in retinotopic and spatiotopic coordinates at photopic light levels. To distinguish between the two coordinates, the test stimulus was shifted in the same retinal location relative to fixation after a vertical saccade in the retinotopic condition. In contrast, the test stimulus was kept at the same location in the spatiotopic condition after a saccade. We found that when participants executed saccadic eye movements, negative priming was observed exclusively in retinotopic coordinates, whereas positive priming was observed only in spatiotopic coordinates. 

In this study, we aimed to evaluate the effects of light level on the coordinate bases associated with visual motion processing. If the deterioration of motion perception at mesopic light levels in our daily lives is related to spatiotopic representation rather than retinotopic representation, motion processing in retinotopic coordinates should not be affected by the light level. In other words, negative motion priming would be constantly observed in retinotopic coordinates at low light levels as well as photopic light levels. In contrast, if motion processing in spatiotopic coordinates is affected by the light level, the perception of positive motion priming in spatiotopic coordinates should be modulated at low light levels, presumably at a mesopic level. Two reasons have enabled us to form these proposals. First, the effect of visual motion priming, both positive and negative, reflects a function of the underlying motion mechanism that integrates spatiotemporally different visual inputs to induce coherent motion perception (Pinkus & Pantle, 1997). Second, some action-related visual tasks, which may depend on the successful integration of visual motion signals, are not conducted well at dusk, as described earlier. 

We performed three experiments. In Experiment 1, we found that in retinotopic coordinates negative motion priming was observed irrespective of light levels, whereas in spatiotopic coordinates positive motion priming was observed at photopic and scotopic light levels but not at mesopic light levels. These results were invariant with the change in the luminance contrast (Experiment 2) or the presence of saccadic eye movements (Experiment 3) per se. We concluded that the disappearance of the positive priming effect in spatiotopic coordinates at mesopic light levels is due to different spatiotemporal properties of cones and rods, which may disturb the construction of a spatiotopic representation. 

Experiment 1

Methods

Participants

Four individuals (M. I., S. Y., T. T., and Y. K.) participated in the experiments without monetary compensation. Two participants (S. Y. and T. T.) were the authors of this study. The other two participants have had experiences in various psychophysical experiments except for the visual motion priming experiments and were naïve to the purpose of the experiments reported in this study. All had normal or corrected-to-normal vision. All participants gave informed consent before their participation in the study. 

Apparatus

The stimuli were generated by MATLAB (MathWorks Inc., Natick, MA) using Psychophysics Toolbox version 3.0 extension (Brainard, 1997; Pelli, 1997) on a computer (MacPro, Apple Inc., Cupertino, CA) and were displayed on a 21-in. color monitor (Sony GDM F520, Sony Corp., Tokyo, Japan). The monitor frame rate was 120 Hz, with a spatial resolution of 1024 × 768 pixels and 12-bit gray-level resolution via a personal computer (Endeavor MT7900, Espon Direct Corp., Nagano, Japan) with a VSG 2/5 graphics card (Cambridge Research Systems Ltd., Rochester, UK). The monitor output was linearized (gamma corrected) using a ColorCAL MKII colorimeter (Cambridge Research Systems Ltd.) under software control (CRS Color Toolbox for MATLAB, Cambridge Research Systems Ltd.). For all experiments using luminance-varying stimuli, the space-averaged chromaticity (Commission internationale de l'éclairage, CIE 1931) of the display had an x-value of 0.31 and a y-value of 0.33. Participants observed the display with a headrest. Patterns were viewed monocularly, with the right eye at a viewing distance of 57 cm. The pupil diameters and the movements of the right eye of each participant were monitored using a ViewPoint EyeTracker 220 fps USB system (Arrington Research Inc., Scottsdale, AZ) that was controlled by the same computer during the whole period of the experiments. The sampling rate of this infrared video-based eye tracker was 220 Hz. Saccades were detected by applying a set of velocity and acceleration criteria using 40°/s velocity and 800°/s² acceleration thresholds (Krauzlis & Miles, 1996). 

Neutral density filters were placed in front of the monitor screen to obtain three different luminance levels (42, 0.062, and 0.00062 cd/m²). The retinal illuminances were computed from the participants' pupil diameters that were measured under these three luminance levels. The average pupil diameters for four participants were 4.64 (SD = 0.34), 7.02 (SD = 0.46), and 7.76 (SD = 0.32) mm from the highest to the lowest adaptation levels. Figure 1 presents the averaged retinal illuminances for four participants. The three retinal illuminances were approximated as 2.85 (SD = 0.063), 0.39 (SD = 0.057), and −1.53 (SD = 0.037) log photopic troland (log Td). We assumed that these retinal illuminances and luminance levels corresponded to the photopic, mesopic, and scotopic levels, respectively (Hood & Finkelstein, 1986). The room was darkened and shielded from light, with no other source of illumination present. 

Figure 1

 

The three averaged retinal illuminances (2.85, 0.39, and −1.53 log Td) obtained from the measurement of pupil diameters (N = 4). These three levels correspond to photopic, mesopic, and scotopic light levels from the highest adaptation levels. The horizontal line represents the retinal illuminance. The horizontal error bars represent ±1 standard deviation (SD). The two dashed vertical lines starting at the left are drawn at the cone threshold and at the rod saturation (based on Hood & Finkelstein, 1986).

Stimuli

Figure 2 provides a schematic description of the priming and test stimuli, trial sequence, and stimulus arrangement. To enable comparisons with our previous study, we used a stimulus that was similar to that used by Yoshimoto et al. (2014). An achromatic vertical sine-wave grating was displayed in a rectangular window that measured 10.0° (width) × 3.3° (height). The edges of the stimulus were tapered using a Gaussian function with a sigma value of 1.0°. Because high-spatial frequency grating may be scarcely detectable under low light levels (Hess et al., 1990), the spatial frequency of the stimulus was set to 0.5 c/°. The stimulus was presented on a uniform gray-colored background (CIE 1931; x = 0.31, y = 0.33) that had a luminance equal to the space-averaged luminance (42 cd/m², 0.062 cd/m², or 0.00062 cd/m² as mentioned above) of the sine-wave grating. 

Figure 2

 

Schematic description of the trial sequence and the stimulus arrangement. A small black dot (radius: 0.25°) was displayed to assist participants in maintaining fixation. The priming stimulus was presented for 167 or 1000 ms. (A) Under the retinotopic and spatiotopic conditions, the fixation point jumped by 6.7° to the upper region of the screen after the termination of the priming stimulus, and participants made a saccade to the new fixation point. After a variable ISI (400–3000 ms), the test stimulus was presented above (retinotopic condition) or below (spatiotopic condition) the fixation point. (B) Under the full and unmatched conditions, the position of the fixation point was not changed. Thus, saccades were not required. After the ISI, the test stimulus was presented above (full condition) or below (unmatched condition) the fixation point. Under the full condition, the priming and test stimuli occupied the same position on the display. Under the unmatched condition, the test stimulus was presented at the position that matched neither the spatiotopic nor retinotopic location of the priming stimulus. Participants were asked to judge whether the perceived direction of the test stimulus was leftward or rightward.

In Experiment 1, we manipulated the presentation duration and velocity of the priming stimulus. Previous studies have shown that visual motion priming depends on various stimulus parameters (Kanai & Verstraten, 2005; Pantle et al., 2000; Pavan & Skujevskis, 2013; Pinkus & Pantle, 1997; Takeuchi et al., 2011; Yoshimoto & Takeuchi, 2013). Takeuchi et al. (2011) and Yoshimoto and Takeuchi (2013) found that when the stimuli with a high luminance contrast (e.g., as high as 10 times the direction discrimination threshold) are presented in the parafovea (Figure 2), the presentation duration and velocity of the priming stimulus determine the effects of motion priming. At a velocity of around 3 Hz, positive motion priming was observed when the duration of the priming stimulus was less than approximately 300 ms. In contrast, negative motion priming was observed when the duration of the priming stimulus was longer. However, at a velocity of lower than 3 Hz, positive priming was dominant even when the priming stimulus was as long as 2000 ms. On the other hand, at a velocity higher than 3 Hz, negative priming was dominant regardless of the duration of the priming stimulus. Later, we discuss the stimulus dependency of visual motion priming. For now, we point out that by properly choosing the stimulus parameters, the specific effect of motion priming—negative or positive—can be robustly induced. We utilized this characteristic of priming effect in this study. 

Based on previous studies (Takeuchi et al., 2011; Yoshimoto & Takeuchi, 2013), we examined certain combinations of the duration and velocity of the priming stimulus to induce positive or negative priming. The presentation duration of the priming stimulus was set to 167 or 1000 ms. The velocity of the priming stimulus varied from 2 to 4 Hz. The combinations of the duration and velocity of the priming stimulus used in the experiment are summarized in Table 1. The reason why we examined several combinations is that a single parameter does not determine the effect of visual motion priming. Predictions regarding the perceived direction of motion priming under different viewing conditions at photopic light levels are also included. These predictions are based on previous studies that show the effects of duration, velocity, and luminance contrast of the priming stimulus and the study that examined the priming effect under retinotopic and spatiotopic conditions (Yoshimoto et al., 2014). 

Table 1

 

The stimulus parameter examined in Experiment 1 coupled with the predicted priming effects at the photopic light level with saccade. The luminance contrast was assumed to be 100%.

Primer duration (ms)	Velocity (Hz)	Predicted priming effects at the photopic light level
167	3	Positive under the spatiotopic and full conditions
		No priming under the retinotopic and unmatched conditions
167	4	Negative under the retinotopic and full conditions
		No priming under the spatiotopic and unmatched conditions
1000	2	Positive under the spatiotopic and full conditions
		No priming under the retinotopic and unmatched conditions
1000	3	Negative under the retinotopic and full conditions
		No priming under the spatiotopic and unmatched conditions

The Michelson contrast of the priming stimulus was set to 100%, which was sufficiently above the direction discrimination threshold at both mesopic (average = 4.5%, SD = 0.63%) and scotopic (average = 16.9%, SD = 3.8%) light levels measured before running the main experiment for all participants. The details of threshold measurement were described in our previous study (Yoshimoto & Takeuchi, 2013). We examined different luminance contrasts in Experiment 2. 

The direction of motion of the priming stimulus was either rightward or leftward. Similar to previous studies (Kanai & Verstraten, 2005; Pinkus & Pantle, 1997), an ambiguous test stimulus was generated by shifting the phase of the grating by 180°. The spatial frequency was set to 0.5 c/°, which was the same as that of the priming stimulus. To equate the velocities of the priming and test stimuli, the duration of one frame of the test stimulus was determined based on the velocity of the priming stimulus. It was set to a duration that was equal to that required for the priming stimulus to shift 180°. A total of four frames were presented for the test stimulus. The luminance contrast of the test stimulus was 100%, which was the same as that of the priming stimulus. Tables A1 and A2 in Appendix A show the frame duration, phase shift, and total frame number for the priming and test stimuli at each velocity condition. 

The priming stimulus was presented in the center of the screen. Similar to our previous study (Yoshimoto et al., 2014), we ran four experimental conditions (Figure 2). In the retinotopic condition, the fixation point jumped to the upper region of the screen soon after the offset of the priming stimulus, and the test stimulus was presented with a variable ISI in the same retinotopic position relative to fixation as the priming stimulus. The spatiotopic condition was similar to the retinotopic condition, with the exception that the test stimulus was presented in the same screen position as the priming stimulus. In the full condition, the test stimulus was presented with a variable ISI at the same location as the priming stimulus but with no shift in the fixation point (i.e., this condition represented both retinotopic and spatiotopic coordinates). Finally, in the unmatched condition, the test stimulus was presented with a variable ISI at a position that matched neither the spatiotopic nor retinotopic location of the priming stimulus, with no shift in the fixation point. In all conditions, the spatial distance between the center of the priming stimulus and the fixation point was set to 3.3°. A black dot with a radius of 0.25° was displayed to assist participants in maintaining fixation while the grating was presented in the parafovea. In preliminary observations, we collected data when the priming stimulus was presented in the upper peripheral retina (above the fixation point) and in the lower peripheral retina (below the fixation point). We did not find any systematic difference between data collected from lower and upper retinas. Therefore, the priming stimulus was presented only in the upper peripheral retina in the main experiment (Figure 2). Participants made 6.7° upward saccades with the shift of the fixation point under the retinotopic and spatiotopic conditions (Figure 2A) but not under the full and unmatched conditions (Figure 2B). The unmatched condition was adopted to examine the possibility that the effects of motion priming could result from motion integration over a large spatial region under the spatiotopic condition when the priming and test stimuli were separated retinotopically. 

Recent studies have shown that the spatiotopic representation was not available instantaneously (Burr & Morrone, 2011, 2012; Golomb, Marino, Chun, & Mazer, 2011; Golomb, Pulido, Albrecht, Chun, & Mazer, 2010; Morrone, Cicchini, & Burr, 2010; Wurtz, 2008; Zimmermann, Morrone, Fink, & Burr, 2013). Similarly, Yoshimoto et al. (2014) found that prominent positive priming in spatiotopic coordinates was observed only when the interval between the priming and test stimuli was longer than approximately 400 ms. To investigate the temporal properties of the representation process in retinotopic and spatiotopic coordinates at different light levels, the time from the offset of the priming stimulus to the onset of the test stimulus (i.e., the ISI between the priming and test stimuli) was changed for 400, 600, 1000, 1600, or 3000 ms. The ISI was also changed under the full and unmatched conditions, in which no saccade was required, in the same manner as that used for the retinotopic and spatiotopic conditions. In Experiment 3, we examined the condition in which two saccades were required, similar to the study that examined the spatiotopic representation of visual attributes (Knapen, Rolfs, & Cavanagh, 2009). The reason why we did not adopt the two saccade paradigm in Experiments 1 and 2 is because we wanted to examine the effect of ISI, which is now considered an important parameter for understanding a spatiotopic representation (Burr & Morrone, 2012; Zimmermann et al., 2013). 

Procedure

Each trial began with the presentation of a small dot, which appeared as the fixation point for 1.5 s, followed by the priming sine-wave stimulus. Under the retinotopic and spatiotopic conditions, the fixation point immediately jumped by 6.7° to the upper region of the screen after the termination of the priming stimulus, and participants executed a saccade to the new fixation point. After a variable ISI, during which the display was blank (but contained the fixation point), a directionally ambiguous test stimulus was presented. The participants' task was to judge whether the perceived direction of the test stimulus was leftward or rightward by pressing the appropriate arrow key. Participants were instructed to view the fixation point continuously throughout the trial. After a button was pressed, a 1-s intertrial interval (in which a uniform field with the space-averaged luminance was displayed) was introduced to reduce the effect of the former trial. The saccade conditions (retinotopic and spatiotopic) and the no-saccade conditions (full and unmatched) were conducted in separate sessions. Each session consisted of 160 trials. Eight trials were performed for each of the two types of presentation of the test stimulus (the retinotopic and spatiotopic conditions in the saccade session; the full and unmatched conditions in the no-saccade session), for each of the five ISIs between the priming and test stimuli, and for the two directions of the priming stimulus (rightward or leftward). The trials were presented in a random order. In each session, the duration of the priming stimulus, the velocity, and light level were fixed. The four stimulus parameter combinations shown in Table 1 (167-ms primer duration and 3-Hz velocity, 167-ms primer duration and 4-Hz velocity, 1000-ms primer duration and 2-Hz velocity, and 1000-ms primer duration and 3-Hz velocity) were examined at each of the three light levels (photopic, mesopic, and scotopic). Each participant completed two saccade sessions and two no-saccade sessions for each of the four stimulus parameter combinations at each of the three light levels (48 sessions in total) in a random order. Therefore, each participant conducted 32 trials per condition. Participants were initially dark-adapted for 30 min prior to the task. Participants underwent at least 20 practice trials under each condition prior to actual data acquisition. 

Results

Eye position was monitored during the experiment. Under the retinotopic and spatiotopic conditions, we confirmed whether all participants made saccades in accordance with the shifts of fixation point. For all light levels and trials, the duration between the offset of the priming stimulus and the end of the saccade ranged from 136 to 315 ms. Thus, ISI of 400 ms was sufficiently long to make a 6.7° saccade. In a small number of the trials, participants looked more than 1.5° away from the fixation point (both pre- and postsaccade) during a single trial (5.1% of trials across participants). We excluded those trials from subsequent analyses. 

Conditions in which negative motion priming was predicted

Figures 3 and 4 show the results for each participant for each experimental condition. The primer duration was 167 ms and velocity was 4 Hz in Figure 3; the values were 1000 ms and 3 Hz, respectively, in Figure 4. In both conditions, we predicted that negative motion priming would be observed in the retinotopic and full conditions, whereas no priming effect would be present in the spatiotopic and unmatched conditions at photopic light levels (Table 1). The percentage response to positive motion priming was plotted as a function of the ISI between the priming and test stimuli. Thus, when more than 50% of the responses represented positive motion priming, participants reported that the perceived direction of the test stimulus was in the same direction as the priming stimulus in the majority of the trials. When fewer than 50% of the responses were scored as positive priming, participants reported that the motion of the test stimulus was in the direction opposite to the priming stimulus (negative motion priming) in the majority of the trials. 

Figure 3

 

Results of Experiment 1 for the four participants with the primer duration of 167 ms and velocity of 4 Hz. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). The luminance contrast was set to 100%. Error bars represent ±1 standard error of the mean (SEM). Each curve represents the data that were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Figure 4

 

Results of Experiment 1 for the four participants with the primer duration of 1000 ms and velocity of 3 Hz. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). The luminance contrast was set to 100%. Error bars represent ±1 SEM. Each curve represents the data that were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Although there was some interparticipant variability, the results were essentially consistent across participants. Thus, the averaged data were used in our analyses. Figure 5 presents the averaged data for the four participants. The primer duration was 167 ms and velocity was 4 Hz in Figure 5A; the values were 1000 ms and 3 Hz, respectively, in Figure 5B. In both conditions, negative motion priming was dominant at both the retinotopic and full conditions during photopic vision, as predicted (Table 1). Similarly, negative priming was observed in most trials at mesopic and scotopic light levels. At all three light levels, the frequency of the perception of negative priming decayed as ISI between the priming and test stimuli became longer. Thus, the percentage response converged to 50%. 

Figure 5

 

Averaged data for four participants in Experiment 1 (Figures 3 and 4). Each curve represents data collected at different light levels (photopic, mesopic, and scotopic). Error bars represent ±1 SEM. The luminance contrast was set to 100%. The primer duration and velocity were set to (A) 167 ms and 4 Hz and (B) 1000 ms and 3 Hz, respectively.

To test whether the light level or ISI alter the negative priming in the retinotopic condition, which is our main interest in Figure 5, we conducted a within-participant two-way analysis of variance (ANOVA). The percentage data were based on counts having a binomial distribution and therefore an angular transformation was applied to normalize the distribution (Fernandez, 1992). The transformed data sets were submitted to Bartlett's test to check the homogeneity of variance before conducting the ANOVA. Generalized eta-squared values, as suggested for within-participant designs (Bakeman, 2005; Olejnik & Algina, 2003), were computed for all effects to determine the effect size. These were interpreted according to Cohen's recommendation of 0.02, 0.13, and 0.26 for small, medium, and large effects for the generalized eta squared, respectively (Cohen, 1988). 

At the primer duration of 167 ms and velocity of 4 Hz (Figure 5A), the main effect of the light level was not significant, F(2, 6) = 3.10, ns, while that of ISI was significant, F(4, 12) = 44.39, p < 0.0001, η²_G = 0.63. The interaction between the light level and ISI was not significant, F(8, 24) = 0.63, ns. At the primer duration of 1000 ms and velocity of 3 Hz of the retinotopic condition (Figure 5B), the main effect of the light level was not significant, F(2, 6) = 2.52, ns, while that of ISI was significant, F(4, 12) = 37.72, p < 0.0001, η²_G = 0.69. The interaction between the light level and ISI was not significant, F(8, 24) = 1.46, ns. Thus, the results indicate that the motion processing in the retinotopic coordinates is invariant at different light levels. In the spatiotopic and the unmatched conditions, it is confirmed that neither positive nor negative priming are prominent at all light levels, irrespective of both stimulus parameter combinations. These results are consistent with the prediction in Table 1. 

Conditions in which positive motion priming was predicted

Figures 6, 7, and 8 show the individual (Figures 6 and 7) and group-averaged (Figure 8) results for each experimental condition. The primer duration was 167 ms and velocity was 3 Hz in Figures 6 and 8A; the values were 1000 ms and 2 Hz, respectively, in Figures 7 and 8B. In both conditions, we predicted that positive motion priming would be observed in the spatiotopic and full conditions, whereas no priming effect would be present in the retinotopic and unmatched conditions at photopic light levels (Table 1). The averaged data shown in Figure 8 were used in subsequent analyses because the results were consistent across participants. 

Figure 6

 

Results of Experiment 1 for the four participants with the primer duration of 167 ms and the velocity of 3 Hz. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). The luminance contrast was set to 100%. Error bars represent ±1 SEM. Each curve represents the data that were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Figure 7

 

Results of Experiment 1 for the four participants with the primer duration of 1000 ms and the velocity of 2 Hz. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). The luminance contrast was set to 100%. Error bars represent ±1 SEM. Each curve represents the data that were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Figure 8

 

Averaged data for four participants in Experiment 1 (Figures 6 and 7). Each curve represents data collected at different light levels (photopic, mesopic, and scotopic). Error bars represent ±1 SEM. The luminance contrast was set to 100%. The primer duration and velocity were set to (A) 167 ms and 3 Hz and (B) 1000 ms and 2 Hz, respectively.

As we predicted (Table 1), positive priming was prominent in the spatiotopic condition, whereas it disappeared in the retinotopic condition at the photopic light level. The positive priming in the spatiotopic condition depended on both the light level and ISI. At the primer duration of 167 ms and velocity of 3 Hz in the spatiotopic condition (Figure 8A), within-participant two-way ANOVA showed that the main effects of the light level and ISI were significant; F(2, 6) = 38.00, p < 0.001, η²_G = 0.61 for light level; F(4, 12) = 19.66, p < 0.0001, η²_G = 0.51 for ISI. The interaction between the light level and ISI was also significant, F(8, 24) = 4.49, p < 0.01, η²_G = 0.43. At the primer duration of 1000 ms and velocity of 2 Hz of the spatiotopic condition (Figure 8B), the main effects of the light level and ISI were significant; F(2, 6) = 39.52, p < 0.001, η²_G = 0.47 for light level; F(4, 12) = 15.44, p < 0.001, η²_G = 0.52 for ISI. The interaction between the light level and ISI was also significant, F(8, 24) = 7.18, p < 0.0001, η²_G = 0.30. Based on the Cohen's criteria (Cohen, 1988), the effect sizes are large in all these cases. 

These statistical results are also confirmed visually. The function for positive priming in the spatiotopic condition at the photopic condition was a flattened inverted-V shape. The positive priming became prominent only at ISI of approximately 600 to 1600 ms. A similar function was obtained in the spatiotopic condition at the scotopic light level. However, at mesopic light levels, the strength of positive priming in the spatiotopic condition was greatly reduced at both stimulus parameter conditions. The percentage of positive priming was close to 50%, irrespective of ISI. 

In the full condition, as predicted, positive priming was prominent at the photopic light level (Table 1). Positive priming was also observed in most trials at mesopic and scotopic light levels. At all three light levels, the strength of positive priming was reduced as ISI between the priming and test stimuli became longer. In contrast to the results in the spatiotopic condition, positive priming was prominent when ISI was as short as 400 ms. Furthermore, neither positive nor negative priming were observed in the unmatched condition at all light levels. 

Discussion

We replicated our previous findings (Yoshimoto et al., 2014) under the photopic condition. When participants made saccades, negative motion priming was observed only under the retinotopic condition, whereas positive motion priming was observed under the spatiotopic condition. When the saccadic eye movement was not required, negative priming was observed under the full condition in which the retinotopic component was included (Figure 5), whereas positive priming was observed under the full condition in which the spatiotopic component was also included (Figure 8). 

Our main interest was to see whether similar results are obtained at different light levels. We found that negative priming in the retinotopic condition was observed irrespective of the light levels. This is consistent with the conclusion from previous studies in which the sensitivity to moving patterns, whose detectability equated on the basis of the light levels, was estimated under different light levels (Lankheet, van Doorn, & van de Grind, 2002; Takeuchi & De Valois, 2000; van de Grind et al., 2000). Those studies concluded that the same visual motion mechanism is functioning at different light levels (see also Duffy & Hubel, 2007). 

Positive priming in the spatiotopic condition was observed at photopic and scotopic levels but not at mesopic levels. Consistent with previous studies, the flattened, inverted-V shape in the photopic and scotopic conditions indicates that the perception of positive priming would depend on the slowly developing spatiotopic representation (Burr & Morrone, 2012; Golomb et al., 2010, 2011; Morrone et al., 2010; Zimmermann et al., 2013). We speculate that this slow-acting spatiotopic representation would not be well constructed during mesopic vision. Consequently, temporally separated visual inputs would not be well integrated and lead to the disappearance of visual motion priming. As we discuss further in the General discussion, we consider that the incomplete integration of cone-mediated signals and rod-mediated signals could be a possible candidate for the absence of the slow-acting spatiotopic representation during mesopic vision. 

Under the full condition in Figure 5, negative priming was observed at all light levels. The full condition includes both retinotopic and spatiotopic components. This result indicates that motion processing in the retinotopic coordinate is reflected in the full condition because no priming effect was observed in the spatiotopic condition. Under the full condition in Figure 8, in contrast to the results in the spatiotopic condition, positive priming was observed irrespective of the light levels and a flattened, inverted-V shape for the effect of ISI was not observed. Because no priming effect was evident in the retinotopic condition, the observed priming effect in the full condition could be related to the spatiotopic coordinate. The difference in dependency on ISI between the full condition and the spatiotopic condition may suggest the lack of a slow-acting spatiotopic representation when no saccade is required. Thus, our results indicate that when the slow-acting spatiotopic representation is not constructed, the motion priming under mesopic vision disappears. This supports our earlier speculation that when the slow-acting spatiotopic representation has to be constructed, because of the saccadic eye movement, mesopic visual priming disappears. We return to this point in Experiment 3. 

The range of visual space over which motion integration occurs is estimated to be larger at dim light levels than at photopic levels (Hess et al., 1990; Zuidema, Verschuure, Bouman, & Koenderink, 1981). However, no priming was observed in the unmatched condition in our study. Thus, positive priming observed in the spatiotopic condition at the scotopic light level cannot be explained by the hypothesis that spatially separated priming and test stimuli are integrated by a spatially large receptive field of a motion mechanism to induce a priming effect. 

Even observers whose visual functions are normal can show a clear difference in pupil diameter between the right and left eyes (Lam, Thompson, & Corbett, 1987). Because our main concern in this study is to precisely determine the retinal illuminance based on the pupil diameter, we examined only the monocular condition in this study. Thus, the participants were adapted monocularly in terms of viewing the stimuli and judging the perceived direction of the test stimulus. The results obtained here would be essentially unchanged when the stimuli were binocularly viewed because we did not find any qualitative differences between the effect of binocular visual priming previously (Yoshimoto et al., 2014) or during monocular visual priming in this study. 

Experiment 2

To confirm that the positive priming disappears in the spatiotopic condition during mesopic vision, we examined the effect of luminance contrast and saccades in Experiments 2 and 3, respectively. Yoshimoto and Takeuchi (2013) showed that if the test stimulus is presented in the parafovea, where no saccades are required and the stimulus contrast is eight times the direction discrimination threshold, both positive and negative priming are observed depending on the duration and velocity of the priming stimulus. Conversely, when the contrast is as low as two times the threshold, negative priming is dominant regardless of the other stimulus parameters. We then examined whether setting the luminance contrast of the stimulus lower would affect the priming effect (most likely the negative priming effect) in the spatiotopic condition during mesopic vision. We addressed this possibility in Experiment 2. If the priming effects were observed in the spatiotopic condition, our speculation regarding the incomplete construction of spatiotopic representation based on the results of Experiment 1 would not be supported. 

Methods

As in Experiment 1, we examined the manner in which the preceding moving grating modulated the perceived direction of a 180° phase-shifted test grating. Participants judged the perceived directions of the test stimulus presented after a variable ISI in the four experimental conditions (Figure 2). The stimulus parameter combinations of the duration and velocity of the priming stimulus, at which neither positive nor negative priming were observed in the spatiotopic condition at the mesopic light level in Experiment 1, were examined: 167-ms duration and 3-Hz velocity, and 1000-ms duration and 2-Hz velocity (Figure 8). The luminance contrast of the stimuli was set to 50% or 15%, which were approximately 11.30 (SD = 1.42) and 3.39 (SD = 0.43) times, on average, the direction discrimination threshold at the mesopic light level, respectively (average = 4.5%, SD = 0.63%). These multiple values were similar to those used in the study by Yoshimoto and Takeuchi (2013), in which the contrast was as high as eight times the threshold and induced both positive and negative priming. However, if the contrast was as low as two times the threshold, only negative priming was observed. 

We replicated the procedures in Experiment 1, except we varied the luminance contrast. In each session, the primer duration, velocity, contrast, and light level were fixed. The two stimulus parameter combinations (167-ms duration and 3-Hz velocity; 1000-ms duration and 2-Hz velocity) were examined for each of the two contrasts (50% and 15%) at mesopic light levels. Each participant completed two saccade sessions and two no-saccade sessions for each of the two stimulus parameter combinations and for each of the two contrasts (16 sessions in total), in a random order. Therefore, each participant completed 32 trials per condition. Participants were initially dark-adapted for 30 min prior to the task. The same participants that performed Experiment 1 participated in Experiment 2. 

Results

Figures 9, 10, and 11 show the individual (Figures 9 and 10) and group-averaged (Figure 11) results for each participant for each experimental condition. Data at 100% contrast (Figures 6, 7, and 8) were replotted. The primer duration was 167 ms and the velocity was 3 Hz in Figures 9 and 11A; these values were 1000 ms and 2 Hz, respectively, in Figures 10 and 11B. The averaged data shown in Figure 11 were used in subsequent analyses because the results were consistent across participants. 

Figure 9

 

Results of Experiment 2 for the four participants with the primer duration of 167 ms and velocity of 3 Hz in the mesopic light level condition. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). Each curve represents the data that were collected at different luminance contrasts (100%, 50%, and 15%). Error bars represent ±1 SEM. The dark green solid curves (100% contrast) are replots of the data from the mesopic condition in Figure 6. The data were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Figure 10

 

Results of Experiment 2 for the four participants with the primer duration of 1000 ms and velocity of 2 Hz in the mesopic light level condition. In each graph, the percentage response of positive priming (motion of test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). Each curve represents the data that were collected at different luminance contrasts (100%, 50%, and 15%). Error bars represent ±1 SEM. The dark green solid curves (100% contrast) are replots of the data from the mesopic condition in Figure 7. The data were collected in the (A) retinotopic, (B) spatiotopic, (C) full, and (D) unmatched conditions.

Figure 11

 

Averaged data for four participants in Experiment 2 (Figures 9 and 10). Data were obtained during the mesopic condition. Each curve represents data collected at different luminance contrasts (100%, 50%, and 15%). Error bars represent ±1 SEM. The dark green solid curves (100% contrast) are replots of data from the mesopic light level condition in Figure 8. The primer duration and velocity were set to (A) 167 ms and 3 Hz and (B) 1000 ms and 2 Hz, respectively.

In the retinotopic condition in Figure 11, negative priming was observed at 15% contrast, whereas no effect of motion priming was observed at the other contrast level. The strength of negative priming was reduced as ISI between the priming and test stimuli became longer. 

In the spatiotopic condition, regardless of the contrast of the stimuli, neither positive nor negative priming were dominant at all ISIs. At the primer duration of 167 ms and velocity of 3 Hz in the spatiotopic condition (Figure 11A), within-participant two-way ANOVA revealed that the main effects of the contrast and ISI were not significant; F(2, 6) = 2.62, ns for contrast; F(4, 12) = 0.68, ns for ISI. The interaction between the contrast and ISI was not significant, F(8, 24) = 0.60, ns. At the primer duration of 1000 ms and velocity of 2 Hz (Figure 11B), the main effects of the contrast and ISI were not significant; F(2, 6) = 2.05, ns for contrast; F(4, 12) = 0.95, ns for ISI. The interaction between the contrast and ISI was not significant, F(8, 24) = 0.81, ns. 

In the full condition, positive priming was prominent at 50% contrast, similar to the results of the 100% contrast condition in Experiment 1. As in the retinotopic condition, negative priming was observed in most trials at 15% contrast. In the unmatched condition, no priming effect was present. 

Discussion

Yoshimoto and Takeuchi (2013) have shown a dominance of negative motion priming in the full condition during photopic vision when the stimulus contrast is low. Thus, we were interested in observing whether the priming effect (most likely negative) would be observed in the spatiotopic condition during mesopic vision using the low-contrast stimuli. Negative priming was observed in the retinotopic and full conditions with a 15% contrast stimulus; however, no priming effect emerged under the spatiotopic condition. Thus, we conclude that the disappearance of motion priming under mesopic and spatiotopic conditions is robust against a change in the luminance contrast. On the other hand, under the retinotopic or full condition, a change in the contrast changes the effect of motion priming. We discuss this contrast dependency of motion priming in the General discussion. 

Experiment 3

In Experiment 3, we examined whether the saccadic eye movements exert a detrimental effect on positive motion priming in the spatiotopic condition during mesopic vision. At the mesopic light level, positive priming was prominent in the full condition but not in the spatiotopic condition (Figure 8). One obvious difference between the full and spatiotopic conditions was the presence of saccadic eye movements. As described above, if the slow-acting spatiotopic representation is constructed when the saccadic eye movement occurs, no priming effect would be observed when the saccade is executed in the full condition during mesopic vision. Because we have shown that positive motion priming was unaffected by saccades in the full condition during photopic vision (Yoshimoto et al., 2014), only mesopic and scotopic light levels were tested in Experiment 3. 

Methods

Stimuli

Figure 12 presents a schematic illustration of the stimuli in a single trial in Experiment 3. As in Experiment 1, we examined the manner by which a moving grating modulated the perceived direction of a subsequently presented 180° phase-shifted test grating. Stimulus configuration was similar to the full condition (Figure 2B), with the exception that the fixation point shifted (Figure 12). The fixation point jumped to the upper region of the screen after the offset of the priming stimulus and then returned to the original position. Thus, participants executed two saccades in Experiment 3. ISI between the offset of the priming and the onset of the test stimuli was changed for 400, 600, 1000, 1600, or 3000 ms. However, we found that an ISI of 400 ms was not long enough to make two saccades (as we describe later). We therefore represent the range of ISI from 600 to 3000 ms in Figure 12. The stimulus parameter combinations of the duration and velocity of the priming stimulus, at which neither positive nor negative priming were observed in the spatiotopic condition at the mesopic light level in Experiment 1, were examined: 167-ms duration and 3-Hz velocity, and 1000-ms duration and 2-Hz velocity. The luminance contrast of the stimuli was set to 100%. 

Figure 12

 

Schematic description of the trial sequence and the stimulus arrangement. A small black dot (radius: 0.25°) was displayed to assist participants in maintaining fixation. The priming stimulus was presented for 167 or 1000 ms. The fixation point jumped by 6.7° to the upper region of the screen after the termination of the primer, stayed on for 200 ms, and then returned to the original position. Thus, participants made two saccades to track the shift in the location of the fixation point. After a variable ISI (600–3000 ms), the test stimulus was presented in the same screen position as the priming stimulus. Participants were asked to judge whether the perceived direction of the test stimulus was leftward or rightward.

Procedure

As in Experiment 1, each trial began with the presentation of a small dot, which appeared as the fixation point for 1.5 s, followed by the priming sine-wave stimulus. The fixation point immediately jumped by 6.7° to the upper region of the screen after termination of the priming stimulus, stayed on for 200 ms, and then returned to the original fixation position (Figure 12). Thus, participants made two saccades to track the position of the fixation point. It should be noted that we did not expect that the saccades would finish within 200 ms. This duration was arbitrarily determined to assist the participants conducting saccades. As described below, in most trials, participants performed two saccades before the onset of the test stimulus. 

After a variable ISI between the offset of the priming stimulus and the onset of the test stimulus, a directionally ambiguous test stimulus was presented in the same screen position as the priming stimulus. Participants were asked to judge whether the perceived direction of the test stimulus was leftward or rightward. Each session consisted of 80 trials: eight trials for each of the five ISIs between the priming and test stimuli and for the two directions of the priming stimulus (rightward or leftward). The trials were presented in a random order. In each session, the duration of the priming stimulus, velocity, and light level were fixed. The two stimulus parameter combinations (167-ms duration and 3-Hz velocity; 1000-ms duration and 2-Hz velocity) were examined at each of the two light levels (mesopic and scotopic). Each participant completed two sessions for each of the two stimulus parameter combinations at each of the two light levels (eight sessions in total), in a random order. Therefore, each participant completed 32 trials per condition. Participants were initially dark-adapted for 30 min prior to the task. Participants who participated were the same as those in Experiments 1 and 2. 

Results

As in the previous experiments, eye position was monitored during the experiment. In each trial, we confirmed whether participants made two saccades in accordance with the shifts of fixation point. The total duration between the offset of the priming stimulus and the end of the second saccade at the mesopic light level ranged from 287 to 523 ms, and the total duration for the scotopic light level ranged from 415 to 673 ms. As shown by Doma and Hallett (1988), the latency for saccades was lengthened as light intensity decreased. At both mesopic and scotopic levels, the second saccade was not finished within 400 ms in the 400-ms ISI condition. We therefore excluded data for this condition from subsequent analyses. In some number of the trials under the scotopic condition, the time needed to finish the second saccade was greater than 600 ms in the 600-ms ISI condition (18.3% of trials across participants). In addition, participants looked more than 1.5° away from the fixation point on average in 2.5% of the trials across participants. We also excluded these trials (261 trials in total) from subsequent analyses. 

Figures 13 and 14 show the individual and group-averaged results for each participant and experimental condition, respectively. The averaged data shown in Figure 14 were used in subsequent analyses because the results were consistent across participants. At both mesopic (Figure 14A, B) and scotopic (Figure 14C, D) light levels, the results were considerably similar to those obtained for the full condition in Experiment 1 where no saccade was required. We compared the results of Experiment 3 with those obtained for the full condition of Experiment 1 using data for ISIs examined (600, 1000, 1600, and 3000). Under the mesopic condition, at the primer duration of 167 ms and velocity of 3 Hz (Figure 14A), within-participant two-way ANOVA revealed that the main effect of the saccade presence was not significant, F(1, 3) = 0.24, ns, while that of ISI was significant, F(3, 9) = 73.40, p < 0.0001, η²_G = 0.75. The interaction between saccade presence and ISI was not significant, F(3, 9) = 0.29, ns. At the primer duration of 1000 ms and velocity of 2 Hz (Figure 14B), the main effect of saccade presence was not significant, F(1, 3) = 0.78, ns, while that of ISI was significant, F(3, 9) = 36.13, p < 0.0001, η²_G = 0.72. The interaction between saccade presence and ISI was not significant, F(3, 9) = 1.03, ns. Under the scotopic condition, at the primer duration of 167 ms and velocity of 3 Hz (Figure 14C), the main effect of saccade presence was not significant, F(1, 3) = 1.63, ns, while that of ISI was significant, F(3, 9) = 29.24, p < 0.0001, η²_G = 0.84. The interaction between saccade presence and ISI was not significant, F(3, 9) = 0.59, ns. At the primer duration of 1000 ms and velocity of 2 Hz (Figure 14D), the main effect of saccade presence was not significant, F(1, 3) = 0.87, ns, while that of ISI was significant, F(3, 9) = 20.21, p < 0.001, η²_G = 0.70. The interaction between saccade presence and ISI was not significant, F(3, 9) = 0.30, ns. 

Figure 13

 

Results of Experiment 3 for the four participants. In each graph, the percentage response of positive priming (motion of the test stimulus in the same direction as the primer) is plotted as a function of ISI between the primer and test stimulus (ms). Each curve represents the data that were collected in different experimental conditions (full with saccade and full without saccade). Error bars represent ±1 SEM. The pink dashed curves (full without saccade condition) are replots of the data from the mesopic and scotopic conditions in Figures 6 and 7. The luminance contrast was set to 100%. (A, B) Data obtained in the mesopic condition. (C, D) Data obtained in the scotopic condition. The primer duration and velocity were set to (A, C) 167 ms and 3 Hz and (B, D) 1000 ms and 2 Hz, respectively.

Figure 14

 

Averaged data for four participants in Experiment 3 (Figure 13). Each curve represents data collected from different experimental conditions (full with saccade and full without saccade). Error bars represent ±1 SEM. The pink dashed curves (full without saccade condition) are replots of data from the mesopic and scotopic level conditions in Figure 8. The luminance contrast was set to 100%. (A, B) Data obtained under the mesopic condition. (C, D) Data obtained under the scotopic condition. The primer duration and velocity were set to (A, C) 167 ms and 3 Hz and (B, D) 1000 ms and 2 Hz, respectively.

Discussion

Irrespective of eye movements, when ISI was 600 to 3000 ms, positive priming was prominent in all stimulus parameter combinations at the mesopic and scotopic light levels. The strength of positive priming was reduced when ISI was 3000 ms. These results suggest no detrimental effect of saccadic eye movements on the disappearance of positive priming in spatiotopic coordinates at mesopic light levels. 

In the discussion of Experiment 1, we proposed that, in the full condition, a slow-acting spatiotopic representation may not been constructed because no saccade was required. We showed that even when the saccadic eye movement is required, the existence of the slow-acting spatiotopic representation, which would emerge with the effect of ISI, is not confirmed. Thus, we need to reassess our conjecture concerning the prerequisite for the emergence of slow-acting representation. One possibility is that when the eye returns to its original position, as in Experiment 3, the slow-acting spatiotopic representation is not constructed because it is not required. Future studies are required to clarify this. 

General discussion

Summary of the results

To characterize the effects of light level on the reference frames of motion processing, we estimated the effect of a priming stimulus on a test stimulus presented after saccades in the retinotopic or spatiotopic coordinates at different light levels. We found that in the retinotopic condition, negative motion priming was dominant at all three light levels (Figure 5). In the spatiotopic condition, the strength of positive motion priming was greatly reduced at the mesopic light level, whereas positive motion priming was prominent at photopic and scotopic light levels (Figure 8). The weakening of motion priming in the mesopic condition was not due to a particular stimulus parameter, such as the duration, velocity, and luminance contrast of the priming stimulus (Figure 11), or saccadic eye movements per se (Figure 14). 

Disappearance of motion priming in spatiotopic coordinates at mesopic light levels

Why is the strength of positive priming in spatiotopic coordinates greatly diminished at mesopic light levels? It has been argued that retinotopic information is remapped with each saccade in some way to enable a stable perception of our world (Burr & Morrone, 2011, 2012; Cavanagh et al., 2010; Cicchini, Binda, Burr, & Morrone, 2013; Duhamel, Colby, & Goldberg, 1992; Melcher & Colby, 2008; Sommer & Wurtz, 2004; Wurtz, 2008). Burr and Morrone (2012) proposed a receptive field oriented in space and time to enable transient spatiotopy in neurons that shift their receptive fields in anticipation of upcoming saccades (Duhamel et al., 1992). However, they did not consider that this rapid-acting system is the only one for visual stability. They also proposed a slowly developing system of spatiotopic coordinates, in which a stable spatiotopic representation builds up. As mentioned earlier, various studies have supported this proposal (e.g., Zimmermann et al., 2013). 

We speculate that the loss of motion priming in the spatiotopic condition during mesopic vision may result from incompleteness in the construction of this slowly developing spatiotopic representation after the eye movement under mesopic vision. This incomplete spatiotopic representation should be a slowly developing type because we found that positive priming in spatiotopic coordinates during photopic and scotopic vision became conspicuous when ISI was longer than approximately 600 ms. Because visual motion priming has been considered to reflect a function of the underlying motion mechanism that integrates temporally separated motion information (Pinkus & Pantle, 1997), an integration process required for the perception of visual priming may not function with the incomplete spatiotopic representation. 

The cause of a disturbance in the construction of the spatiotopic representation remains unclear. As has been shown in several attempts to determine the mesopic luminous efficiency function as a sum of the weighted cone and rod components (e.g., Goodman et al., 2007; Ikeda & Shimozono, 1981; Palmer, 1966, 1967, 1968; Sagawa & Takeichi, 1992; Trezona, 1991), visual perceptions under mesopic vision appear to be extracted by the integration of input signals processed through cone and rod pathways. Evidence exists that the simultaneous activation of cone- and rod-mediated systems deteriorates motion perception, such as biological motion (Billino et al., 2008). We speculate that the incomplete integration of cone- and rod-mediated signals disturbs not only the perception of biological motion but also the construction of the spatiotopic representation during mesopic vision. This may lead to the disappearance of positive motion priming. The emergence of positive motion priming under the spatiotopic condition during both photopic and scotopic vision where only the cones or rods are functioning supports our view. 

Hadjikhani and Tootell (2000) showed using functional magnetic resonance imaging (fMRI) that at least at the level of V1, cone- and rod-mediated pathways are separated. However, at the level of MT, the separation was not observed. Other functional magnetic resonance imaging (fMRI) studies have shown that responses in MT are tuned to spatiotopic coordinates, while those in V1 are tuned to retinotopic coordinates (Crespi et al., 2011; d'Avossa et al., 2007; see also Gardner, Merriam, Movshon, & Heeger, 2008). This indicates that the spatiotopic representation may be constructed somewhere between V1 and MT. During mesopic vision, this pathway may not function well because of a segregation of cone- and rod-mediated signals, and this could lead to the disappearance of positive motion priming. 

Characteristics of visual motion priming and its relation to reference frames

The primary premise of this study depended on our previous finding that negative and positive motion priming are conspicuously observed in retinotopic and spatiotopic coordinates, respectively (Yoshimoto & Takeuchi, 2013). A similar line of study has been previously published. Turi and Burr (2012) showed that two forms of motion aftereffects (MAEs), which probably act at different neural levels of processing, are observed in different reference frames. In other words, when saccades are involved between adaptation and test stimuli, lower-level adaptation such as the classic MAE, in which prolonged exposure to a moving stimulus makes a stationary stimulus viewed subsequently appear to move in the opposite direction, is strictly encoded in retinotopic coordinates that shift with each eye movement (Biber & Ilg, 2011; Boi, Ögmen, & Herzog, 2011; Cavanagh et al., 2010; Knapen et al., 2009; Wenderoth & Wiese, 2008). In contrast, higher-level adaptation such as the positional MAE, in which the apparent position changes by adaptation to motion, is encoded in spatiotopic screen-based coordinates. Turi and Burr (2012) concluded that low and high levels of visual analyses have retinotopic and spatiotopic coordinate bases, respectively. 

The assumption that the positional MAE occurs at a higher level than the classic MAE induced by translational motion, as used in Turi and Burr (2012), has been supported by other studies. Meng, Mazzoni, and Qian (2006) found that the significant cross-fixation transfer of MAE indicating that the involvement of dorsal medial superior temporal (MSTd) area (or higher areas) was generated by the expansion motion but not by the translational motion. By applying transcranial magnetic stimulation, McGraw, Walsh, and Barrett (2004) showed the involvement of V5/MT, but not V1/V2, activations in generating positional MAE. 

The relationship between our study and the theory of Turi and Burr (2012) needs to be addressed. We argued in our previous study (Takeuchi et al., 2011; Yoshimoto & Takeuchi, 2013) that negative motion priming is induced in low-level motion mechanisms, whereas positive motion priming is induced in high-level motion mechanisms. This conclusion is derived from the observation that negative priming is prominent in higher-velocity or lower-contrast conditions, whereas positive priming is prominent in lower-velocity and higher-contrast conditions. These observations are replicated in this study. Different motion mechanisms have different sensitivities to the velocity or temporal frequency and luminance contrast of moving patterns. The contrast sensitivity of an energy-based first-order motion system is relatively higher than that of other motion mechanisms. A plausible candidate for a directionally selective motion detector, which is sensitive to high velocity, would be energy-based. It should be noted that this idea does not exclude the possibility that the energy-based sensor is also sensitive to a lower velocity. In contrast, higher-order motion mechanisms such as an attentional tracking mechanism are insensitive to higher velocity and lower contrast (Adelson & Bergen, 1985; Burr, Ross, & Morrone, 1986; Burr & Thompson, 2011; Cavanagh, 1992, 1994; Cavanagh & Alvarez, 2005; Del Viva & Morrone, 1998; Derrington, Allen, & Delicato, 2004; Lu & Sperling, 1995, 2001; Nishida, 2011; Smith, Hess, & Baker, 1994; Takeuchi & De Valois, 1997, 2009; Verstraten, Cavanagh, & Labianca, 2000). Based on these previous studies, we assumed that a low-level motion mechanism such as the first-order motion mechanism would be responsible for the induction of negative priming. This assumption is similar to the proposal of Pantle et al. (2000), who indicated that negative priming reflects lower levels of motion processing than positive priming. On the other hand, a high-level motion mechanism such as an attentional tracking mechanism may be responsible for positive priming. If our speculation is valid, it is said that our conjecture concerning the relationship between the level of motion processing and the reference frames is consistent with the theory put forward by Turi and Burr (2012). 

We should note that the mechanism may be more complex than the abovementioned processes. The neural site responsible for visual motion priming is still debatable (Campana, Maniglia, & Pavan, 2013; Campana, Pavan, Maniglia, & Casco, 2011; Glasser, Tsui, Pack, & Tadin, 2011; Jiang, Luo, & Parasuraman, 2002; Kanai & Verstraten, 2005; Pantle et al., 2000; Pavan et al., 2009, 2010). Neurophysiological studies have shown that a moving stimulus with a very short duration such as several hundred milliseconds is sufficient to change the adaptation status of directionally selective neurons located early at V1 (Lisberger & Movshon, 1999; Priebe, Churchland, & Lisberger, 2002). This type of rapid adaptation of directionally selective neurons could be an underlying neural mechanism of negative motion priming (Glasser et al., 2011; Kanai & Verstraten, 2005; Pantle et al., 2000; Pavan et al., 2009). By applying repetitive transcranial magnetic stimulation, Campana et al. (2011) found a greater involvement of visual areas V1/V2 than V5/MT when negative motion priming was induced. Furthermore, MAEs involving rapid MAE (negative motion priming in our terminology) are shown to be very complex phenomena that can take place at different levels of visual motion processing (Mather, Pavan, Campana, & Casco, 2008). Campana et al. (2011, 2013) showed that rapid MAE, static MAE, and dynamic MAE have different neural loci and tap both low and high levels of motion processing depending on the spatiotemporal properties of the stimulus with repetitive transcranial magnetic stimulation. With regard to positive priming, Jiang et al. (2002) measured event-related potentials and fMRI data from human participants as they observed positive priming and showed that positive priming involves the modulation of neural responses in a later visual area such as MT (Movshon, Adelson, Gizzi, & Newsome, 1986; Rodman & Albright, 1989). Thus, mechanisms underlying positive and negative priming may not be clearly divided, as suggested above. In addition, the neural site responsible for the spatiotopic representation is debatable (Crespi et al., 2011; Gardner et al., 2008). Further studies are required to fully elucidate the processing level(s) for both visual motion priming and the construction of the reference frames of visual motion. 

Future studies

Based on this study, we speculate that the two examples described in the Introduction—driving and catching—may be partly due to the incompleteness of the construction of spatiotopic representation at dusk. This may be caused by the incomplete integration of cone- and rod-mediated signals. To further justify our hypothesis, manipulation of the cone and rod contributions to the visual motion system is required. At mesopic light levels, the cones alone define vision in the central retina, and as retinal eccentricity increases, the relative rod contribution in the peripheral retina becomes larger (Hwang, Lee, Park, & Park, 2013; Raphael & MacLeod, 2011). One way to examine our hypothesis is to have both the priming and test stimuli viewed in the rod-dominant peripheral retina under the spatiotopic condition. According to Raphael and MacLeod (2011), the ratio of cone contribution to rod contribution, at the eccentricity of 3.3° where both the priming and test stimuli were presented in our experiments, is estimated to be approximately unity at the mesopic light level. If visual inputs separately given to the cone and rod systems disturb the buildup of the spatiotopic representation at mesopic light levels, positive motion priming would become conspicuous when all the stimuli are presented in the rod-dominant peripheral visual fields. This could be examined in the future to confirm the validity of our hypothesis. 

Acknowledgments

S. Y. is supported by the Japan Society for the Promotion of Science. 

Commercial relationships: none. 

Corresponding author: Sanae Yoshimoto. 

Email: [email protected]. 

Address: Department of Psychology, Japan Women's University, Kanagawa, Japan, and Japan Society for the Promotion of Science, Tokyo, Japan. 

Appendix A

Table A1

 

Frame duration, phase shift, and total frame numbers for the smoothly drifting priming stimulus. Spatial frequency was 0.5 c/°.

Primer velocity (Hz)	Frame duration of primer (ms)	Phase shift (°)	Total frame number for 167-ms primer	Total frame number for 1000-ms primer
2 (= 4°/s)	8.33	0.033	20	120
3 (= 6°/s)	8.33	0.050	20	120
4 (= 8°/s)	8.33	0.067	20	120

Table A2

 

Frame duration and phase shift for the 180° phase-shifted test stimulus. Spatial frequency was 0.5 c/°.

Test velocity (Hz)	Frame duration of test (ms)	Phase shift (°)
2 (= 4°/s)	250.0	1.0
3 (= 6°/s)	166.7	1.0
4 (= 8°/s)	125.0	1.0