Motion perception under mesopic vision

Abstract

Mesopic and scotopic vision extend over an illuminance range of 10⁶. The goal of the present study was to determine the effect of decreasing light level on the underlying motion mechanism that integrates spatiotemporally separated motion signals. To accomplish this, we took advantage of the phenomenon of visual motion priming, in which the perceived direction of a directionally ambiguous test stimulus is influenced by the directional movement of a preceding priming stimulus. After terminating a drifting priming stimulus, a 180° phase-shifted grating was presented as a test stimulus. The priming and test stimuli were separately presented to the central and peripheral retinas, respectively. The participants judged the perceived direction of this test stimulus at various light levels from photopic to scotopic levels. We found that the effects of motion priming disappeared over 1 log unit of mesopic light levels. When the test stimulus was presented before the offset of the priming stimulus to compensate for the temporal delay in the rod pathway or when both stimuli were presented at the same location in the periphery, a motion-priming effect appeared at mesopic light levels. These results suggest that different temporal characteristics between the cone pathway and rod pathway disturb the function of the putative motion mechanism responsible for the spatiotemporal integration of motion signals, which leads to specific modulation of motion perception over a wide range of mesopic vision.

Introduction

Ambient light levels may vary by a factor of up to 10¹¹ in natural environments (Hood & Finkelstein, 1986; Stockman & Sharpe, 2006). The visual system deals with this broad dynamic range by switching between two different types of photoreceptors: cones, which function at higher levels of illumination, and rods, which function at lower levels. Photopic, mesopic, and scotopic regions are defined according to whether cones operate alone, cones and rods operate together, or rods operate alone, respectively. In our daily lives, mesopic and scotopic vision extends over an illumination range as wide as 10⁶. Therefore, understanding the effect of light levels on visual perception is scientifically and practically crucial (Hess, 1990; Hess, Sharpe, & Nordby, 1990). 

Although rods project into all retinogeniculate pathways, they primarily project into the magnocellular lateral geniculate nucleus (LGN) layers (Purpura, Kaplan, & Shapley, 1988; Zele & Cao, 2015). Because motion-selective areas, such as the middle temporal 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). In fact, various aspects of human motion perception are known to change 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; Yoshimoto, Uchida-Ota, & Takeuchi, 2014b) have all been shown to vary with the light level. 

Most of these studies have shown that motion sensitivity decreases as light levels are reduced, suggesting that changes in underlying temporal mechanisms under low light levels affect motion perception. However, Billino et al. (2008) measured the thresholds for the detection of biological motion under three conditions of luminance corresponding to photopic, mesopic, and scotopic light levels and found that threshold was exclusively increased in the mesopic condition, whereas the threshold under the scotopic condition was identical to that under the photopic condition. The authors argued that in the mesopic condition, the mismatch of cone- and rod-mediated velocity information led to impaired integration of spatiotemporally separated motion signals, producing the largest threshold increase. Inspired by this finding, we employed visual motion priming to examine a motion mechanism under low light levels (Yoshimoto & Takeuchi, 2013). 

Visual motion priming is a phenomenon in which the perceived direction of a directionally ambiguous test stimulus is influenced by the direction of movement of the preceding priming stimulus. Examination of the effect of visual motion priming could reveal a mechanism that integrates temporally separate motion signals (Kanai & Verstraten, 2005; Pantle, Gallogly, & Piehler, 2000; Pavan, Campana, Maniglia, & Casco, 2010; Pinkus & Pantle, 1997). In these previous studies, both priming and test stimuli were presented at the same location in the central visual field. Our previous work (Yoshimoto & Takeuchi, 2013) differed from these studies by separately presenting the priming and test stimuli in the central and peripheral visual fields under different light levels. The rationale for this manipulation was to create a situation in which the induction of visual motion priming reflects a function of the underlying motion mechanism that integrates not only temporally, but also spatially, separate visual inputs to induce motion perception. Our data showed that when the spatial distance between the priming and test stimuli was longer than 4°, the test stimulus was perceived as moving in the opposite direction of the priming stimulus (negative motion priming) under a high light level, presumably corresponding to a photopic level (48 cd/m²). We proposed that negative priming could be induced by a mechanism similar to the spatially center-surround antagonistic motion-contrast sensing system (Allman, Miezin, & McGuinness, 1985; Born & Tootell, 1992; Eifuku & Wurtz, 1998; Murakami & Shimojo, 1993; Tadin, Lappin, Gilroy, & Blake, 2003). Antagonistic interactions between the center and surround would induce a biased perception of motion direction for the directionally ambiguous test stimulus in the peripheral visual field when the unidirectionally drifting stimulus is presented in the central visual field. Since the priming and test stimuli were temporally separated in the visual motion-priming display, the hypothesized mechanism should be able to integrate motion signals separated by several hundreds of milliseconds. 

Then, we measured the effects of motion priming under two lower light levels, presumably corresponding to mesopic (0.048 cd/m²) and scotopic levels (0.0048 cd/m²). We found that under the mesopic level, the effects of motion priming completely disappeared. Under the scotopic level, however, negative motion priming was observed in a similar degree to that observed under the photopic level (Yoshimoto & Takeuchi, 2013). As the density of cones is higher in the central retina whereas the density of rods is higher in the periphery (Curcio, Sloan, Packer, Hendrickson, & Kalina, 1987; Osterberg, 1935), our finding suggests that such a hypothesized motion-contrast mechanism could not integrate signals originated from cones at the center and rods at the periphery under mesopic vision. 

We examined only one light level from the broad mesopic range extending over an illuminance range of 10³–10⁴ in our previous study (Yoshimoto & Takeuchi, 2013). However, motion information processing might be different even under mesopic vision because of the complex nature of cone-rod interaction and the different amount of rod contribution across light levels (Bloomfield & Dacheux, 2001; Cao, Pokorny, Smith, & Zele, 2008; Stockman & Sharpe, 2006; Zele & Cao, 2015; Zele, Maynard, Joyce, & Cao, 2014). Thus, in Experiment 1 of the present study, we measured the strength of negative motion priming at various mesopic light levels following our previous experimental method in order to examine the effect of light levels on the integration of spatiotemporally separated visual inputs. 

In our previous study (Yoshimoto & Takeuchi, 2013), we also found that negative priming reappeared at the mesopic light level when the test stimulus was presented before the offset of the priming stimulus, causing a temporary overlap between the priming and test stimuli. This indicates that the temporal delay in the rod pathway at the periphery leads to the disappearance of visual motion priming under mesopic vision. The temporal delay of the rod pathway relative to the cone pathway has been estimated to be 70 ms at maximum (Sharpe & Stockman, 1999; Sharpe, Stockman, & MacLeod, 1989; Stockman & Sharpe, 2006). The temporal delay is reduced to 8–20 ms when the state of adaptation is similar for the cones and rods (Cao, Zele, & Pokorny, 2007; Sun, Pokorny, & Smith, 2001; Zele & Cao, 2015). In Experiment 1, we manipulated the overlap time across a wide range of mesopic light levels to verify our previous observations. 

Experiment 1

Methods

Participants

Four participants (EA, SY, TT, and YI) with normal or corrected-to-normal vision participated in the experiments. SY and TT were coauthors of this study, and EA and YI were experienced in psychophysical experiments but naïve to the experimental purpose. This study was approved by the Research Ethics Committee of Japan Women's University (Tokyo, Japan) and was conducted according to the Declaration of Helsinki. All participants provided written informed consent before the study began. 

Apparatus

All stimuli were generated using Matlab (MathWorks, Inc., Natick, MA) with the Psychophysics Toolbox version 3.0 extension for PCs (Brainard, 1997; Pelli, 1997) and were displayed on a 21-in color monitor (GDM F520; Sony Corp., Tokyo, Japan) via VSG 2/5 (Cambridge Research Systems Ltd., Rochester, UK) graphics system. The monitor temporal resolution was 120 Hz with a spatial resolution of 1024 × 768 pixels and 12-bit gray-level resolution. The monitor output was gamma-corrected with a ColorCAL MKII colorimeter (Cambridge Research Systems Ltd.). Neutral density filters (Kodak Wratten 2, Edmund Optics Inc., Barrington, NJ) were placed in front of the monitor screen to obtain nine different photopic luminances (42, 3.0, 0.78, 0.21, 0.062, 0.022, 0.0065, 0.0024, and 0.00062 cd/m²); the luminances were checked by the same colorimeter. To convert photopic troland (phot Td) to scotopic troland (scot Td), the spectrum of the monitor phosphors was measured with a SpectroCAL MKII spectroradiometer (Cambridge Research Systems Ltd.). The participants observed the display with the aid of a headrest. The patterns were monocularly viewed with the right eye at a distance of 57 cm. The room was darkened and shielded against external light. The fixation of the right eye for each participant was monitored using a ViewPoint EyeTracker 220 fps USB system (Arrington Research, Inc., Scottsdale, AZ) controlled by the same PC during the entire experimental period. The sampling rate of this infrared video-based eye tracker was 220 Hz. The same eye-tracking device was used to measure the pupil diameter of each participant. 

Stimuli

Figure 1 illustrates a schematic description of the stimuli in a single trial. To allow comparisons with our previous study (Yoshimoto & Takeuchi, 2013), we used a similar stimulus. As the priming stimulus, a vertical drifting sine-wave grating was displayed in a rectangular window that measured 10.0° (width) × 3.3° (height). The edges of the stimulus were tapered by a Gaussian function with σ = 1.0°. The spatial frequency of the stimulus was set to 0.5 c/°, which is well below the cut-off spatial frequency of about 6 c/° under low light levels (Hess et al., 1990). The stimulus was presented on a uniform gray-colored background (CIE1931; x = 0.31, y = 0.33) that had a similar luminance to the space-averaged luminance of the sine-wave grating. The drift direction of the priming stimulus was either rightward or leftward. Based on our previous study (Yoshimoto & Takeuchi, 2013), the presentation duration and the velocity of the priming stimulus were set to 167 ms and 6 °/s, respectively. The luminance contrast was set to two times that of the direction discrimination threshold, as described in the experimental procedure. 

Figure 1

 

Schematic descriptions of stimuli in a single trial. The priming and test stimuli were presented to central and peripheral visual fields (VF), respectively. The spatial distance between the center of the priming stimulus and that of the test stimulus was set to 10°. The black central cross (1.0° × 1.0°) was displayed to assist participants in maintaining fixation, while the grating was presented in the periphery. The participants were requested to judge whether the perceived direction of the test stimulus was rightward or leftward.

An ambiguous test stimulus was generated by shifting the phase of the grating by 180°, as in the previous studies (Kanai & Verstraten, 2005; Pinkus & Pantle, 1997). The spatial frequency of the test stimulus was the same as that of the priming stimulus. To equate the velocities of the priming and test stimuli, the phase of the test stimulus was shifted every 167 ms, which corresponded to the time taken to shift the priming stimulus by 180°. The total duration of the test stimulus was 667 ms. The priming and test stimuli were presented to the central (0° eccentricity) and peripheral (10° eccentricity) upper visual fields, respectively (Figure 1). A black fixation cross (1.0° × 1.0°) was displayed to assist the participants in maintaining fixation while the test stimulus was presented in the periphery. 

In Experiment 1, the overlap time between the priming and test stimuli was varied in seven steps (0.0, 8.3, 16.7, 25.0, 33.3, 41.7, and 83.3 ms). Figure 2 shows a schematic representation of the temporal relationship between the priming and test stimuli. When the overlap time was 0.0 ms, the test stimulus was presented immediately after the offset of the priming stimulus (Figure 2A). 

Figure 2

 

Schematic representation of the timing relationship between the priming and test stimuli in Experiment 1. The horizontal and vertical axes represent time and luminance contrast, respectively. (A) There was no overlap time between the priming and test stimuli. (B) The overlap time between the priming and test stimuli varied from 8.3–83.3 ms. Both priming and test stimuli were presented during the overlap period, as illustrated.

Experimental procedure

For each of three measurements described below, the participants were dark-adapted for 30 min prior to the tasks. Each measurement started from the darkest adapting levels. 

Pupil measurement:

The retinal illuminances were computed from the participant pupil diameters, which were measured under the nine luminances. A uniform space-averaged luminance blank field was presented for 5 s while recording the pupil diameter. The participants were asked to focus on the center of the screen without blinking. Figure 3 presents both the individual and averaged photopic and scotopic retinal illuminances for the four participants. The averaged 5-s recording of the pupil diameter range for the four participants was 3.9–8.0 mm. The averaged nine retinal illuminances were approximated from 2.8 to −1.6 log phot Td. The conversion from phot Td to scot Td was calculated from the spectral power distribution of the monitor phosphors over a range of 380–780 nm. For the computation, the CIE photopic V(λ) modified by Judd (1951) and Vos (1978) and the CIE (1951) scotopic V'(λ) were used (Wyszecki & Stiles, 2000). The conversion factor for our monitor was 2.88. The calculated log scot Tds are shown in Figure 3. 

Figure 3

 

Individual and averaged photopic and scotopic retinal illuminances for the four participants obtained from the pupil diameter measurements. Scotopic retinal illuminance (log scot Td) is plotted as a function of the photopic retinal illuminance (log phot Td). Both vertical and horizontal error bars in the graph of averaged data represent ±1 standard deviation (SD).

According to Hood and Finkelstein (1986) and Stockman and Sharpe (2006), the cone threshold and rod saturation are approximately −1.4 and 1.9 log phot Td (or −1.0 and 2.3 log scot Td), respectively. Based on this estimation, we assumed that the measured photopic retinal illuminances were categorized as one photopic (2.8 log phot Td), seven mesopic (1.9, 1.3, 0.81, 0.33, −0.07, −0.57, and −0.99 log phot Td), and one scotopic (−1.6 log phot Td) light levels. Although we used these categorizations as a reference in our Discussion, it is important to note that these represent a rough estimation, as we did not directly measure the amount of cone/rod activation under different light levels. 

Contrast sensitivity measurement:

For the main experiment (Figure 1), we equated the effective luminance contrast under different light levels and retinal eccentricities by setting the Michelson luminance contrast of the sine-wave gratings to two times that of the direction discrimination threshold. For that, we measured the luminance contrast sensitivity for direction discrimination of the drifting sine-wave gratings at each of the nine light levels at 0° and 10° eccentricities. The presentation duration of the drifting stimulus was 167 ms, which was identical to that of the priming stimulus used in the main experiment. The participants judged the perceived drift direction of the stimulus with a forced choice of two alternatives (rightward or leftward) with no feedback. We applied a standard staircase algorithm that was designed to converge at a 79% correct level (Levitt, 1971) to estimate the threshold contrast. Further details of the experiment were described in Yoshimoto & Takeuchi (2013). When the stimulus was presented in the periphery, the participants were instructed to maintain their focus on the cross pattern displayed. 

Direction judgment for estimating motion-priming effects:

The priming stimulus was presented for 167 ms, 500 ms after a beep (Figure 1). After terminating the priming stimulus, the directionally ambiguous test stimulus was presented. The participants were asked to judge the perceived direction of the test stimulus by pressing the appropriate arrow key. After a response, a blank field was presented for 1 s during the intertrial interval to reduce any effects from the previous trial. The participants were instructed to continuously view the fixation point throughout the trial. The participants' fixation was checked with an eye-tracking device during each trial. 

The direction judgment was conducted at an overlap time of 0.0 ms and at an overlap time of 8.3–83.3 ms during the separate sessions. For the overlap time of 0.0 ms, each session consisted of 32 trials: 16 trials for each of the two priming stimulus directions which were presented randomly. Each participant completed a session at each of the nine luminances. For the overlap time of 8.3–83.3 ms, each session consisted of 96 trials: eight trials for each of the six overlap times for the two directions of the priming stimulus. Each participant completed two sessions at each of the nine luminances. The participants underwent at least 20 practice trials at each light level prior to data acquisition. 

Results

Contrast sensitivity measurement

Figure 4 shows the individual and averaged contrast thresholds for the direction discrimination of the moving sine-wave grating for the four participants at two eccentricities. The percentage-contrast threshold was plotted as a function of the photopic retinal illuminance. As shown in Figure 4, the contrast thresholds depended on both the retinal illuminance and the eccentricity. The contrast threshold increased as the retinal illuminance decreased. At high light levels, the contrast thresholds at the center (0° eccentricity) were lower than those at the periphery (10° eccentricity), whereas this tendency was reversed at low light levels. This result agrees with previous studies demonstrating that the motion sensitivity is higher at the periphery compared to the center under scotopic vision (e.g., Hess et al., 1990). The transition of the contrast threshold between the center and peripheral fields was observed at 0.81 log phot Td under mesopic vision using averaged data. Similar results were observed for all four participants. 

Figure 4

 

Individual and averaged data of the contrast thresholds obtained from direction discrimination measurements of the four participants. The percentage-contrast threshold is plotted as a function of the retinal illuminance (in log phot Td). Each curve represents data for the retinal eccentricity. The error bars in the graph of averaged data represent ±1 SD.

These interpretations were supported by statistical analyses. We conducted a within-participant, two-way analysis of variance (ANOVA) for the data. The generalized η²_G, which is suggested for analysis of within-participant designs (Bakeman, 2005; Olejnik & Algina, 2003), was used to estimate the effect size and interpreted according to Cohen's recommendation of 0.02 for a small effect, 0.13 for a medium effect, and 0.26 for a large effect (Cohen, 1988). The main effects of the retinal illuminance and eccentricity were significant, F(8, 24) = 113.2, p < 0.0001, η²_G = 0.93 for retinal illuminance; F(1, 3) = 43.19, p < 0.01, η²_G = 0.44 for eccentricity. The interaction between the light level and eccentricity was also significant, F(8, 24) = 9.56, p < 0.0001, η²_G = 0.52. 

In the following experiments, we set the luminance contrast of the gratings to be twice the measured threshold for each participant. 

Direction judgment for estimating motion-priming effects

The number of trials in which participants looked more than 1.5° away from the fixation point throughout a single trial was less than 1%. Thus, excluding data from these trials did not change our results and conclusions. Therefore, we used the data from all trials for the subsequent analysis. Figure 5 shows the individual and averaged data for the four participants when there was no overlap time between the priming and test stimuli (Figure 2A). The percentage response to the negative motion priming is plotted as a function of the retinal illuminance (in log phot Td). When more than 50% of the responses represented negative motion priming, the participants reported that the perceived direction of the test stimulus was opposite to that of the priming stimulus in the majority of the trials. When fewer than 50% of the responses were scored as negative priming, the participants reported that the motion was in the same direction as the priming stimulus (the so-called positive motion priming) in the majority of trials. 

Figure 5

 

Individual and averaged data for the four participants obtained from the direction judgment for estimating motion-priming effects. There was no overlap time between the priming and test stimuli. The percentage response to negative priming is plotted as a function of photopic retinal illuminance (in log phot Td). Error bars in the graph of averaged data represent 95% confidence intervals (CI).

Although there was some interparticipant variability, these results were essentially consistent across the participants. Thus, the averaged data were used in our analyses. As shown in Figure 5, we found that the effect of motion priming greatly depended on the retinal illuminance. A within-participant, one-way ANOVA revealed that the effect of retinal illuminance was significant, F(8, 24) = 40.63, p < 0.001, η²_G = 0.91. Under photopic conditions (2.8 log phot Td), negative priming was observed in the majority of trials, which replicated our previous results (Yoshimoto and Takeuchi, 2013). At higher mesopic light levels (1.9 and 1.3 log phot Td), negative priming was reported in more than 90% of the trials. However, at 0.81 log phot Td, the percentage of negative motion priming was approximately 50% (95% CI of [43.35%, 61.4%]), which indicates that the priming effect totally disappeared and did not recover for retinal illuminances of 0.33 (95% CI of [45.7%, 63.7%]) and −0.07 log phot Td (95% CI of [44.9%, 62.9%]). Thus, neither positive nor negative priming was dominant for an approximate 1 log unit range under mesopic conditions. The priming effect gradually reappeared from mesopic (−0.57 log phot Td) to scotopic (−1.6 log phot Td) levels, where negative priming became dominant. A reversal of the contrast threshold for the central and peripheral retinas occurred at 0.81 log phot Td (Figure 4), which was coincident with the retinal illuminance where the priming effect disappeared in Figure 5. 

Figure 6 shows the individual and averaged data for the four participants when the priming and test stimuli temporally overlapped. The data at the 0.0-ms overlap time (pale blue curves) were replotted from Figure 5. Although there was some interparticipant variability, these results were essentially consistent across the participants. Thus, the averaged data were used in our analyses. 

Figure 6

 

Individual and averaged data for the four participants. The percentage response to negative priming is plotted as a function of photopic retinal illuminance (in log phot Td). Each curve represents data for the overlap times between the priming and test stimuli (0.0–83.3 ms). The pale blue curves (overlap time of 0.0 ms) are replots of data from Figure 5. The red arrows in the averaged data graph represent mesopic conditions at which no priming effect was observed at the overlap time of 0.0 ms (0.81, 0.33, and −0.07 log phot Td). Error bars in the graph of averaged data represent ±1 standard error of the mean.

As the overlap time between the priming and test stimuli increased, the function changed from a U- to a flat-shape (Figure 6). In the mesopic conditions indicated by red arrows (0.81, 0.33, and −0.07 log phot Td), the effect of increasing the overlap time was clear. In addition, negative priming was induced by increasing the overlap time at lower mesopic levels (−0.57 and −0.99 log phot Td). In contrast, at the two highest mesopic levels (1.9 and 1.3 log phot Td) as well as under photopic (2.8 log phot Td) and scotopic (−1.6 log phot Td) conditions, negative priming was dominant irrespective of the overlap time. At overlap times longer than 25.0 ms, negative priming was observed in more than 70% of trials irrespective of retinal illuminance. 

Statistical analyses support these interpretations. A within-participant, two-way ANOVA was conducted on the average data depicted in Figure 6. The main effects of the retinal illuminance and overlap time were significant, F(8, 24) = 17.05, p < 0.0001, η²_G = 0.64 for retinal illuminance; F(6, 18) = 34.80, p < 0.0001, η²_G = 0.50 for overlap time. The interaction between retinal illuminance and overlap time was also significant, F(48, 144) = 6.50, p < 0.0001, η²_G = 0.55. 

A close examination of Figure 6 revealed that the shapes of the functions for the reappearance of negative priming which depended on the overlap time were different between 0.81 and −0.07 log phot Td (indicated by the rightmost and leftmost red arrows, respectively). Thus, we selected these data (0.81, 0.33, and −0.07 log phot Td) and replotted the percentage of negative priming as a function of the overlap time in Figure 7. The overlap time required to induce the conspicuous negative priming at 0.81 log phot Td was shorter than the overlap time required at 0.33 or −0.07 log phot Td. At the overlap time of 8.3 ms, negative priming was observed in more than 80% of trials at 0.81 log phot Td, whereas the priming effect was not observed at 0.33 (95% CI of [48.06%, 65.99%]) and −0.07 log phot Td, (95% CI of [47.27%, 65.23%]). 

Figure 7

 

The percentage response to negative priming was plotted as a function of the overlap time between the priming and test stimuli. Each curve represents data for different retinal illuminances, which are indicated by red arrows in Figure 6 (0.81, 0.33, and −0.07 log phot Td). Error bars represent 95% CI.

A within-participant, two-way ANOVA followed by Tukey's posthoc test was conducted for multiple comparisons. The main effects of the retinal illuminance and overlap time were significant, F(2, 6) = 5.27, p < 0.05, η²_G = 0.30 for retinal illuminance; F(6, 18) = 38.76, p < 0.0001, η²_G = 0.81 for overlap time. The interaction between the retinal illuminance and the overlap time was also significant, F(12, 36) = 4.02, p < 0.001, η²_G = 0.31. The Tukey's test revealed significant differences in the percentage of negative priming between 0.81 and −0.07 log phot Td at overlap times of 8.3 (q = 7.70, p < 0.0001), 16.7 (q = 6.49, p < 0.001), 25.0 (q = 5.75, p < 0.001), and 33.3 ms (q = 3.78, p < 0.05). The Tukey's test also showed significant differences between 0.81 and 0.33 log phot Td at overlap times of 8.3 (q = 7.50, p < 0.0001) and 16.7 ms (q = 6.49, p < 0.001). No significant difference was found between 0.33 and −0.07 log phot Td. 

Discussion

The changes of contrast thresholds and motion priming along with decrements in retinal illuminance were quite different (Figures 4 and 5). The contrast thresholds gradually decreased for both the central and peripheral fields, whereas negative motion priming showed a sudden disappearance and gradual recovery as the retinal illuminance decreased on a log scale. This difference suggests that the underlying motion mechanism responsible for the direction discrimination of the moving stimulus is different from that for the induction of negative motion priming. In the direction discrimination task, a motion detector which is mostly tuned to a given stimulus would determine the luminance contrast threshold (e.g., Watson, 1986). On the other hand, negative motion priming observed at a suprathreshold luminance contrast would reflect the integration of the outputs from various motion detectors that detect spatiotemporally separated priming and test stimuli (Pinkus & Pantle, 1997; Yoshimoto & Takeuchi, 2013). 

In Experiment 1, we found that the motion-priming effect disappeared around 1 log unit of the mesopic range, approximately from 0.8 to −0.1 log phot Td (Figure 5). In light of the distinct distributions of cones and rods in the retina (Curcio et al., 1987; Osterberg, 1935), we hypothesize that at this mesopic range, the motion signals from both the cone and rod systems in the center and the motion signals mainly from the rod system in the periphery could not integrate well. As described in the Introduction, we speculate that a mechanism similar to the spatially center-surround antagonistic motion-contrast sensing system is responsible for the motion-priming effect. Based on this, one possible explanation for the disappearance of negative priming at some mesopic light levels is that the temporal delay in the rod pathway fails to provide visual inputs within the given time frame that would allow such a mechanism to operate. The assumed temporal delay would be shorter at higher light levels (Figure 7), which may correspond to the known temporal characteristics of the rod pathway (Sharpe et al., 1989; Zele & Cao, 2015). 

The negative priming was observed at the two highest mesopic light levels (1.3 and 1.9 log phot Td) and at mesopic light levels lower than −0.57 log phot Td. This may be because that the primarily operating system is the cone system at the higher mesopic levels and it is the rod system at the lower mesopic levels across the retina, which did not induce temporal conflict between the cone and rod pathways. We examined this conjecture in the next experiment. 

Experiment 2

In Experiment 2, we examined whether motion priming would occur under mesopic conditions by presenting both the priming and test stimuli at the same peripheral location. Because no temporal conflict would arise between the cone and rod pathways with this manipulation, we predicted that the motion-priming effect would be observed regardless of retinal illuminance. 

Methods

Here, we presented the test stimulus at a 10° eccentricity, similar to Experiment 1. However, we also presented the priming stimulus at the same peripheral location as that of the test stimulus, not at the central retina, as in Experiment 1. To examine the effect of the rod system on the motion-priming effect, the luminance level was varied from mesopic (3.0 cd/m² or 1.9 log phot Td) to scotopic light levels (0.00062 cd/m² or −1.6 log phot Td) in eight steps. There was no overlap time between the priming and test stimuli. All other parameters were the same as in Experiment 1. 

Each session consisted of 32 trials: 16 trials for each of the two directions of the priming stimulus, which were presented randomly. Each participant completed one session at each of the eight retinal illuminances (Figure 3), starting from the darkest adaptation level. In each session, the light level was fixed. The same participants that performed Experiment 1 participated in Experiment 2. 

Results and discussion

Figure 8 shows the individual and averaged data for the four participants. The Experiment 1 data for the priming and test stimuli, which were presented separately to the central and peripheral fields, was replotted from Figure 5. Although there was some interparticipant variability, the results were relatively consistent across participants. Thus, the averaged data were used in our analyses. In contrast to Experiment 1, we found that negative priming was observed in more than 90% of trials, irrespective of retinal illuminance. A within-participant, one-way ANOVA for the averaged data revealed that there was no significant effect of retinal illuminance, F(7, 21) = 1.77, n.s. This result supports our hypothesis that the temporal delay in the rod pathway disturbs the spatiotemporal integration of the motion signals during mesopic vision. 

Figure 8

 

Individual and averaged data for the four participants in Experiment 2. The percentage response to negative priming is plotted as a function of photopic retinal illuminance (in log phot Td). Each curve represents data for the eccentricities of the priming and test stimuli. “10°–10°” denotes that both stimuli were presented in the periphery (10° eccentricity). The pale blue curves (0°–10°) are replots of data from Figure 5 in which the priming stimulus was presented to the central retina (0° eccentricity), and the test stimulus was presented to the periphery (10° eccentricity). Error bars in the graph of averaged data represent 95% CI.

General discussion

Summary and conclusion

We summarize the findings of our study below: 

1.  
   The effect of motion priming disappeared at the approximately 1 log unit range of mesopic light levels (0.8 to −0.1 log phot Td). Negative motion priming was observed at the higher and lower mesopic light levels as well as at the photopic and scotopic levels (Figure 5).
2.  
   In the mesopic range where the priming effect disappeared, the frequency of reporting negative priming gradually increased with an increase in the overlap time between the priming and test stimuli (Figure 6). The duration of the overlap time that induced negative priming was shorter at higher mesopic levels (Figure 7).
3.  
   When the priming and test stimuli were presented at the same peripheral location, negative priming was dominant irrespective of retinal illuminance (Figure 8).

The relative amount of rod contribution should have been different across the retina at the mesopic light levels, since both the cones and rods are active in the central retina, while the rods are mainly active in the peripheral retina (Raphael & MacLeod, 2011). Our results indicate that the spatiotemporally separate motion signals were not well integrated when the motion signals were processed via the central and peripheral retinas. This conclusion is in accordance with the findings reported by Billino et al. (2008), which demonstrated that the simultaneous activity of the cones and rods may exert a detrimental effect on motion integration, resulting in a mesopic-specific threshold elevation. The mesopic light level of 0.285 cd/m² (or about 0.9 log phot Td) used in their study is similar to our highest mesopic light range, where we observed a disappearance of motion priming (from approximately 0.8 to −0.1 log phot Td). 

We also demonstrated that the motion-priming effect reappeared when the overlap time of the priming and test stimuli was increased (Figure 6) or when the temporal conflict between the cone and rod pathways was removed (Figure 8). These results indicate that the incomplete spatiotemporal integration of motion information may lead to the disappearance of motion priming. This is presumably due to a maximum delay of approximately 20 ms in the rod pathway from the cone pathway in the peripheral retina (Figure 7). Although further studies are needed to confirm this, the estimation of the rod delay obtained here is consistent with the current estimations of the cone-rod latency difference in similar states of light adaptation (Cao et al., 2007; Sun et al., 2001; Zele & Cao, 2015). 

Underlying mechanism for negative motion priming

We hypothesize that a mechanism similar to the spatially center-surround antagonistic motion-contrast sensing system is a candidate for the underlying mechanism for negative motion priming. In this section, we suggest alternative explanations. One is based on the studies examining a related phenomenon called the phantom or remote motion aftereffect—MAE (von Grünau & Dubé, 1992; Snowden & Milne, 1997). In this stimulus configuration, the test stimulus was presented at a different spatial location after the adapting moving stimulus was presented for 30–60 s. The test stimulus was then perceived to move in the opposite direction of the adapting stimulus. Our data indicate that a similar phenomenon could be induced by a short-term adaptation (167 ms) to the moving stimulus, if a dynamic (i.e., temporally varying) test stimulus is used. Snowden and Milne (1997) demonstrated that the phantom MAE was observed in an area with a 5° diameter and concluded that motion detectors tuned to wide-field motions induced the phantom MAE. This model is a feasible candidate for the underlying mechanism of the negative priming effect during photopic vision reported in this study. Another candidate is a mechanism that assumes a long-range suppression between spatiotemporally separated locations. Such a mechanism has been proposed for the spatial domain (Polat & Sagi, 1993), but recent studies have demonstrated a long-range interaction in the spatiotemporal domain (Yeshurun, Rashal, & Tkacz-Domb, 2015). 

Although further studies are needed to identify the underlying mechanism of visual motion priming, our results indicate that the spatiotemporal characteristics in the early visual pathway, presumably at the retinal level, critically influence the later stages of visual motion processing. Using functional magnetic resonance imaging, Hadjikhani and Tootell (2000) demonstrated that the signals from the cone and rod pathways are combined at the level of V1. By applying repetitive transcranial magnetic stimulation, Campana et al. (2011) identified the involvement of V1 when negative motion priming was induced. Taken together, our findings indicate that under mesopic vision, the failure to provide visual input to the motion mechanism in V1 at the appropriate timing results in incomplete motion integration. 

Effect of presentation duration on visual motion priming

Previous studies have shown that when both priming and test stimuli are presented at the same location in the fovea, a priming stimulus as short as 300 ms induces positive priming in which the test stimulus is perceived to drift in the same direction as the priming stimulus (Kanai & Verstraten, 2005; Pantle et al., 2000). In contrast, we found that when both priming and test stimuli are presented at the same peripheral retinal location, negative priming is induced by the 167 ms priming stimulus (Figure 8). This finding suggests that the effect of the priming stimulus presentation duration depends on the retinal eccentricity of moving patterns: The duration that switches from positive to negative priming effect may decrease as the retinal eccentricity increases. 

We speculate that the multisystem for motion perception is responsible for this dependency on retinal eccentricity. Our previous studies (Takeuchi et al., 2011; Yoshimoto & Takeuchi, 2013; Yoshimoto, Uchida-Ota, & Takeuchi, 2014a) suggested that negative priming is induced by an energy-based, directionally selective, first-order motion mechanism (Adelson & Bergen, 1985; van Santen & Sperling, 1985; Watson & Ahumada, 1985), which is shown to be sensitive to moving stimuli with low-contrast and presented in the periphery (Ashida, Seiffert, & Osaka, 2001; Chubb & Sperling, 1989; Dosher, Landy, & Sperling, 1989; Edwards & Nishida, 2004; Lorenceau & Shiffrar, 1992; Lorenceau, Shiffrar, Wells, & Castet, 1993; Lu & Sperling, 1995, 2001; Smith, Hess, & Baker, 1994; Solomon & Sperling, 1994, Sperling, 1989; Sun, Chubb, & Sperling, 2014, 2015; Takeuchi & De Valois, 1997, 2009; Weiss, Simoncelli, & Adelson, 2002; Yo & Wilson, 1992). In the current study, the prominent negative priming disappeared when the test stimulus was presented in the periphery at two times the threshold contrast (i.e., low contrast) during mesopic vision (Figure 5). This suggests that the function of the first-order motion mechanism can be influenced under mesopic vision. Further studies are required to clarify the relationship between the light level-dependency of motion perception and its underlying motion mechanisms. 

Acknowledgments

SY is supported by the Japan Society for the Promotion of Science. This study was supported by JSPS KAKENHI Grant Number 24650058 (Japan). Part of this study was presented at Vision Sciences Society 2014 (St. Pete Beach, FL, USA). 

Commercial relationships: none. 

Corresponding author: Tatsuto Takeuchi. 

Email: [email protected]. 

Address: Department of Psychology, Japan Women's University, Kanagawa, Japan.