Journal of Vision (2012) 12(6):19, 1–14 http://www.journalofvision.org/content/12/6/19 1

The role of motion streaks in the perception of the kinetic Zollner illusion

The School of Optometry and Vision Science, The University of New South Wales, Sieu K. Khuu Sydney, New South Wales $

In classic geometric illusions such as the Zollner illusion, vertical lines superimposed on oriented background lines appear tilted in the direction opposite to the background. In kinetic forms of this illusion, an object moving over oriented background lines appears to follow a titled path, again in the direction opposite to the background. Existing literature does not proffer a complete explanation of the effect. Here, it is suggested that motion streaks underpin the illusion; that the effect is a consequence of interactions between detectors tuned to the orientation of background lines and those sensing the motion streaks that arise from fast object motion. This account was examined in the present study by measuring motion-tilt induction under different conditions in which the strength or salience of motion streaks was attenuated: by varying object speed (Experiment 1), contrast (Experiment 2), and trajectory/length by changing the element life-time within the stimulus (Experiment 3). It was predicted that, as motion streaks become less available, background lines would less affect the perceived direction of motion. Consistent with this prediction, the results indicated that, with a reduction in object speed below that required to generate motion streaks (, 1.128/s), Weber contrast (, 0.125) and motion streak length (two frames) reduced or extinguished the motion-tilt-induction effect. The findings of the present study are consistent with previous reports and computational models that directly combine form and motion information to provide an effective determinant of motion direction. Keywords: geometric illusions, kinetic Zollner illusion, motion streaks, motion and form interaction, spatial vision Khuu, S. K. (2012). The role of motion streaks in the perception of the kinetic Zollner illusion. Journal of Vision, 12(6):19, 1– 14, http://www.journalofvision.org/content/12/6/19, doi:10.1167/12.6.19

see Fineman & Melingonis, 1977;Wenderoth& Introduction Johnson, 1983; c.f., Watamaniuk, 2005). Despite the fact that geometric illusions are simplis- In classic examples of optical geometric illusions such tic in construction and form, their investigation as the Zollner and Hering illusions (Burmester, 1896; provides valuable insight into the functional nature of Coren & Girgus, 1978; Hering, 1861; Orbison, 1939; mechanisms that code orientation and, more generally, Zollner, 1860), straight target lines appear tilted in the the way in which the represents space (see direction opposite to the oriented background lines Gillam, 1998; Westheimer, 2008). For example, it is the upon which they are superimposed. Geometric distor- consensus view that tilt-induction effects observed in tions of space are not limited to static stimuli, also static geometric illusions reflect lateral/mutual inhibi- occurring when the path of a moving object traverses tion between orientation-tuned detectors with overlap- tilted background lines. This so-called ‘‘motion-tilt ping orientation-tuning profiles (see Blakemore, induction’’ effect has been reported for kinetic versions Carpenter, & Georgeson, 1970; Carpenter & Blake- of both the Zollner and Hering illusions (Swanston, more, 1973; Day, 1973; Eagleman, 2001; Westheimer, 1984); analogous to static forms of these illusions, tilted 2008; Wallace, 1975; note that, though this explanation background lines deflect the path of motion in the is by no means exhaustive, see Gillam, 1998; Morgan & opposite direction. A similar effect was reported by Casco, 1990). Line-repulsion effects arise because Cesaro and Agostini (1998): an object traversing a orientation-tuned detectors coding both the target straight horizontal path across a background of lines and background lines (that differ slightly in orienta- that alternate in tilt direction appears to take on a tion) are mutually inhibiting, shifting the peak of their sinusoidal ‘‘slalom’’ trajectory. Likewise, motion-tilt tuning profile in opposite directions. Indeed, previous induction has been noted for a kinetic version of the studies have shown that lateral inhibition accurately Poggendorff illusion. In this case, the motion path of an predicts the degree and direction of tilt induction as the obliquely translating element appears misaligned when angular separation between target and background passing behind a static occluder (Nihei, 1973, 1975; also lines increases. However, while both physiological

doi: 10.1167/12.6.19 Received September 18, 2011; published June 12, 2012 ISSN 1534-7362 Ó 2012 ARVO

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 2

(Blakemore & Tobin, 1972) and computational models motion-selective cell with that of an orientation- proposing lateral inhibition (e.g., Morikawa, 1987) are selective cell tuned for extracting the orientation of well developed to account for tilt induction in static the motion streak. The advantage of this neural geometric illusions, it remains unclear whether analo- circuitry is that, because motion streaks have compar- gous operations apply to kinetic versions of these atively narrower bandwidths than motion-selective illusions. units, orientation-tuned mechanisms provide greater Swanston (1984) speculated that distortion to the precision in signaling direction, aiding the computation motion path induced by oriented background lines in of motion. Indeed, if motion streaks are effectively kinetic forms of the illusion reflects reciprocal interac- masked or degraded, it has been shown that luminance tions between separate mechanisms sensitive to spatial thresholds for detecting a moving dot (Geisler, 1999) orientation and motion direction. It is indeed possible and discriminating the direction of motion are com- that mutual inhibition between the tuning profiles of paratively elevated (Burr & Ross, 2002). orientation-tuned detectors coding the background The reliance of the visual system on the orientation lines and direction selective cells sensing the direction of motion streaks in signaling direction of motion of object motion can account for the reported motion- presents a new hypothesis for the mechanisms under- tilt-induction effects. However, while it has been well pinning kinetic geometric illusions. In the present demonstrated that inhibition occurs between detectors study, it is proposed that kinetic geometric illusions both selective for either orientation or motion (and reflect a reciprocal interaction between orientation- accounting for repulsion effects in each stimulus tuned detectors sensing the motion streak produced by domain), it is unclear whether such a process occurs the movement of the object and the orientation of between cells separately selective for orientation and background lines. Background lines may distort the motion. Importantly, previous studies have instead orientation of the motion streak, and when combined established that cells in the are sensitive to with the output of a motion detector (as proposed by both orientation and motion information (e.g., see Geisler, 1999, and confirmed by Or et al., 2007), the Albright, 1984; Krekelberg, Dannenberg, Hoffmann, perceived direction of motion is distorted. The possi- Bremmer, & Ross, 2003; Lennie, 1998), and the outputs bility that motion streaks are important to the of such cells are likely to represent a combination of explanation of kinetic geometric illusions is supported orientation and motion information rather than mutual by a recent investigation by Apthorp and Alais (2009), inhibition. Indeed, evidence that the visual system who showed that motion streaks produced by a combines form and motion information is well background of moving dots tilt the perceived orienta- supported by recent behavioral studies demonstrating tion of a centrally presented grating. The effect is that the perceived direction of moving oriented analogous to static forms of the tilt illusion. Tilt elements is influenced by its orientation (e.g., Krekel- induction only arose when moving dots produced berg et al., 2003; Or, Khuu, & Hayes, 2010). These strong motion streaks at fast speeds or followed observations therefore cast doubt on existing assump- extended trajectories. While Apthorp and Alais showed tions about the mechanisms underpinning kinetic that motion streaks affect the perceived orientation of geometric illusions. The goal of the present study is static lines, it is still unclear whether the motion-tilt to revisit geometric illusions, seeking a more inclusive induction observed in kinetic geometric illusions arises and comprehensive explanation; one that, by necessity, because background lines affect orientation-tuned units considers the interaction between form and motion to extracting motion streaks. The present study investi- account for the effect. gates this possibility. Recently, Geisler (1999) proposed that the visual If motion-tilt induction does indeed arise from system directly considers the outputs of orientation- lateral inhibition (or an analogous operation) between tuned neurons (e.g., simple cells) when determining the background orientation and the motion streak direction of motion. Because such neurons integrate produced by object motion, it would be predicted that information over brief periods (50–100 ms; see Barlow, reduction in the availability of motion streaks would 1958; Burr, 1980; Peterson, Ohzawa, & Freeman, 2001; result in a veridical percept of the object’s trajectory. Snowden & Braddick, 1991), the image of a rapidly Previous studies have well established that reducing translating object will be smeared (along the cell’s object speed, stimulus contrast, and motion length are receptive field), producing a ‘‘motion streak’’ that effective means of attenuating the salience of motion extends away from the object (Badcock & Dickinson, streaks (see Edwards & Crane, 2007; Li, Khuu, & 2009; Geisler, 1999; Or, Khuu, & Hayes, 2007; Ross, Hayes, 2008). The present study adopted these stimulus 2004). According to Geisler, the visual system is manipulations and investigated whether attenuating the sensitive to this orientation signal and computes salience of motion streaks by reducing image speed direction of motion by combining the output of a (Experiment 1), object contrast (Experiment 2), and

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 3

streak length (Experiment 3) impacts the perception of adopted by Cesaro and Agostini are sufficient to the kinetic Zollner illusion, which typifies motion-tilt- generate a motion streak, it can be questioned whether induction effects observed in kinetic geometric illu- a motion streak is perceptible given the very small size sions. The present study reports that reducing dot of the stimulus, and the fact that motion streaks are speed, dot contrast, and streak length attenuated the likely to be low contrast given the fact that they arise perception of motion-tilt induction. These findings are from temporal integration (Alais, Apthorp, Karmann, discussed and compared against the expected outcomes & Cass, 2011; Apthorp, Cass, & Alais, 2010; Kelly, of both a motion streaks and lateral inhibition 1961). The fact that increasing speed produces a hypothesis. reduction in the ‘‘slalom’’ effect is consistent with a reduction in the visibility of the object at faster speeds. Given these inherent differences, it is highly likely that the ‘‘slalom illusion’’ might altogether reflect a different Experiment 1: the effect of object percept (given that there is no static equivalent), but speed on the perceived path of one that is perceptually analogous to the kinetic motion Zollner illusion. The conflicting reports of Nihei (1973) and Swanston (1984) might be accounted for by the presence or According to Geisler (1999) motion streaks effec- absence of speed-related motion streaks. Experiment 1 tively cue direction of motion for object speeds greater sought to elucidate this issue by using a broader range than approximately 1-element width per 100 ms. of speeds. As mentioned, the present study employed a However, integration times have been shown to be variant of the kinetic Zollner illusion. The extent of dependent on luminance contrast with shorter integra- motion-tilt induction was measured as a function of tion periods at higher contrasts (e.g., see Kelly, 1961). object speed over a range of 0.56 to 188/s in which It follows that, if motion streaks do indeed account for motion streaks are not or are generated (see following). induced motion tilt in kinetic geometric illusions, tilt If motion streaks account for tilt induction in the will be most noticeable at comparatively fast object kinetic Zollner illusion, it would be predicted that speeds at which streaks provide a reliable cue for greater motion-path distortion would occur at fast perception. Conversely, object movement ought to be object speeds for which a motion streak is generated unaffected by the background at slower speeds for and less at slower speeds for which motion streaks are which the motion streak is absent. absent. However, if motion streaks are not involved in Studies relating object speed to perceived motion- the perception of kinetic geometrical illusions, the path distortion in kinetic geometric illusions have extent of tilt induction ought not to be speed produced unclear results. Both Nihei (1973)and dependent. Fineman and Melingonis (1977) reported a speed- dependent effect for the kinetic Poggendorff illusion over a very broad range of speeds of 2–438/s. However, Swanston (1984) reported no dependence on object Methods speed for the kinetic Zollner Illusion, and Cesaro and Observers Agostini (1998) reported an inverse relationship be- Six experienced observers (aged 21–35 years) partic- tween object speed and motion-tilt induction for their ipated. One was the author while the others were naıve¨ ‘‘slalom’’ illusion. These observations may be account- to the goals of the study. All had normal or corrected- ed for by the fact that only a few slow speeds were used to-normal visual acuity. in the studies by Swanston and Cesaro and Agostini, which might not be of sufficient magnitudes to produce streaks capable of reliably influencing perception. This Stimuli finding contrasts with those used by Nihei and Fine- The stimulus was a 40-frame movie sequence (each man and Melingonis who employed comparatively frame was shown for 25 ms in rapid succession; entire faster speeds over a broader range capable of generat- duration of the stimulus was 1 s) in which a circular ing motion streaks. Indeed, Swanston measured the cloud (radius: 0.7508) of 20 circular antialiased dots kinetic Zollner illusion with speeds of 2 and 48/s that— (radius: 0.0428, luminance: 80 cd/m2, dot density of the given the stimulus size employed, which was 0.28 in dot cluster: 11.318 dots/deg2) moved vertically from visual angle—is near the critical value required for bottom to top, over light increment tilted lines (width motion streaks to aid in the perception of motion. On 0.0848, luminance, 80 cd/m2). Both the dot cloud and the other hand, Cesaro and Agostini examined motion- lines were overlaid on a midgray background set to a path distortion for speeds of 0.78, 1.55, and 38/s with a luminance of 40 cd/m2, making the Weber contrast of single dot of a diameter of 0.0228. While the speeds the dots and lines 1. Tilted lines were separated by a

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 4

Figure 1. Schematic representation of the stimulus presentation sequence employed in the present study. Initially, a cloud of moving dots is presented on a background consisting of tilted lines for 1 second. After vertical lines appear above and below the stimulus area, the task of the observer was to judge whether the dot cloud moved left or right from vertical. After judgment, 10 Hz dynamic white noise was shown for 1 second.

distance of 0.3368 (separation equal to the width of four Procedures dots) and windowed within a black circular aperture (luminance: 6 cd/m2, line width: 0.18) with a radius of A schematic of the testing procedure is shown in 108. A circular aperture was used to prevent bias in Figure 1. Observers were instructed to maintain central motion-direction judgments by using the edge of the fixation, and the aforementioned stimulus was present- stimulus as a point of reference. At the beginning of a ed to a location so the dot cloud was 38 either to the presentation, the cloud of dots was positioned at the ‘‘6 right or to the left (randomly chosen from trial-to-trial) o’clock’’ position resting near the lower edge of the of central fixation for 1 s. After the stimulus circular aperture. On each and subsequent frames of presentation, the screen was blank (set to background the movie sequence, the cloud of dots subsequently luminance), and two vertical black lines (width: 0.258, moved upward (at a fixed spatial step-size) and length: 28,luminance:4cd/m2), which acted as appeared throughout the presentation. All dots were reference lines, were displayed at the 6 and 12 o’clock generated asynchronously with a limited lifetime of 10 positions on the circumference of the aperture. The task frames (250 ms), after which they disappeared and were of the observer was to indicate whether the path of replotted to a random position in the stimulus area. motion of the dot cloud was to the left or right of the This procedure prevented the tracking of individual reference lines by pressing left or right keys on the dots, requiring the visual system to rely on the global keyboard. This key press also initiated the next direction of dots (see Khuu & Badcock, 2002). Previous stimulus presentation, which began after a display of research has shown that the visual system is well able to 1 second of 10 Hz dynamic pixilated noise was discriminate the motion direction of this stimulus with presented within the circular aperture to prevent any discrimination thresholds of approximately 0.258 of image aftereffects. visual angle over a broad range of speeds (Westheimer Swanston (1984) quantified the extent of path & Wehrhahn, 1994). Observers viewed this stimulus in displacement in the kinetic Zollner illusion by requiring a dark room at a viewing distance of 80 cm. Stimuli observers to rate the extent of tilt with reference to the were generated using MATLAB version 7 and dis- height of the moving object. The present study instead played on a linearized 24-inch Mitsubishi Diamond Pro used a ‘‘nulling’’ procedure; illusory motion tilt was monitor driven at a frame rate of 120 Hz. cancelled by physically changing the motion direction

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 5

of the dot cloud (in the opposite direction to the perceived motion tilt) until it appeared to move vertically. This procedure is efficient because it simultaneously quantifies the extent and direction of motion-tilt induction (previously used by Prinzmetal & Beck, 2001, to quantify tilt in the static Zollner illusion). A staircase procedure (converging on the 79% performance level of the psychometric function) was used to modify the motion direction of the cloud of dots until it appeared vertical. The staircase began with a motion direction randomly chosen from a range of 10 to 108 (negative and positive values signify leftward and rightward movement respectively) of visual angle (from vertical). The initial absolute step-size was 0.58. Subsequently, the step-size was halved. After the third reversal, the step-size was 0.1258 and remained at this value until the end of the staircase run. The staircase lasted for eight reversals, and the average of the last four reversals was used as an estimate of the path of motion judged to be vertical. The staircase procedure was repeated for six different speed levels: 0.56, 1.12, 2.25, 4.5, 9, and 188/s produced by displacing dots at the following step-sizes: 0.014, 0.028, 0.056, 0.113, 0.225, and 0.458/frame. Note that, Figure 2. The judged angle of motion (in8) required to perceive for the two slowest speeds of 0.56 and 1.128/s, the vertical motion plotted against the speed of dots (in log scale) for stimulus was presented for longer at 3 s (by increasing different background line orientations (908 gray circles, 158 black the number of movie frames) rather than for 1 s for squares, 158 light-gray diamond). Error bars signify 95% faster speeds. A longer stimulus presentation time for confidence intervals. slower speed conditions was adopted to ensure that the stimulus traversed a sufficiently long distance for which plots the physical angular path of motion observers to reliably detect its motion direction. In addition to the speed condition, this study included a required for observers to judge that the object is condition in which background lines were tilted 158 to moving vertically (i.e., the angle required to nullify the the left and right from vertical. As Swanston (1984) motion-tilt illusion) against the dot speed on a log noted (consistent with the line-repulsion effect observed scale. The results for background lines tilted at 158 by Carpenter & Blakemore, 1973), this orientation (black squares), 158 (light-gray diamonds), and 908 difference (relative to the trajectory of motion) produces (gray circles) are plotted as different symbols. Positive maximum path displacement. Additionally, a control values on the y-axis signify physical rightward move- condition in which background lines were horizontal ment, negative values physical leftward movement, and (i.e., 908) was included. Previous studies (e.g., Swanston, 08 indicates vertical movement. There were a number of 1984) have demonstrated that, when lines are physically noteworthy findings. horizontal, no motion-tilt induction is observed. There- First, for the baseline condition comprising a fore, this condition provides a good baseline comparison background of horizontal lines (gray circles), observers for conditions in which lines are oriented obliquely. accurately judged the perceived direction of motion of A block comprised 18 staircase runs: background the dot cloud regardless of dot speed. The adjusted lines orientated 158,158, and 908 from vertical for six angle of motion was approximately 08 across the speed different speeds. Observers each completed five blocks, range in this condition. Observers were therefore and results were averaged over the five thresholds for cognizant of the task and consistent in their perfor- each condition. The order of stimulus presentation was mance regardless of speed. randomized within and between blocks. No feedback Second, where background lines were tilted, the was given to indicate the correctness of response. perceived angle of motion required to nullify the illusory tilt deviated from 08. For leftward-tilted background lines (gray diamonds), observers judged Results and discussion that the dot cloud needed to follow a path physically tilted to the left (negative numbers on the y-axis) to be The average results (error bars signify 95% confi- perceived as travelling vertically. Conversely, for dence intervals) of Experiment 1 are given in Figure 2, rightward-tilted background lines (black squares),

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 6

observers judged that the dot cloud needed to follow a speeds, perhaps because of a reduction in sensitivity path physically tilted to the right (positive numbers on to motion at fast speeds (e.g., De Bruyn & Orban, the y-axis) to be perceived as travelling vertically. 1988). Alternatively, at faster speeds, the length of To assess the absolute effect of dot speed and motion streaks would mean that they intersect with a background lines on the judged direction of motion, number of background lines, producing a more results for the two oriented background conditions of compelling tilt effect. This is consistent with previous 15 and 158 were combined as they are mirror studies showing that, for static versions of the Zollner opposites, produced similar absolute effects (see Figure illusion, increasing line length (Wallace, 1969) and line 2), and compared to the baseline condition (908)ina density (e.g., Wallace & Crampin, 1969) increases the two-way ANOVA. This analysis revealed significant extent of illusory tilt. main effects for background orientation (F[1, 96] ¼ The speed-dependent effect with tilted background 165.55, p , 0.0001) and dot speed (F[5, 96] ¼ 15.95, p , lines shown in Figure 2 is consistent with an 0.0001), indicating that both factors affected the explanation of the motion-tilt-induction effect in terms perceived direction of motion (see following for of motion streaks. At slow speeds, motion streaks are discussion). Additionally, their interaction was signif- not generated, and observers are able to perceive the icant (F[5, 96] ¼ 15.67, p , 0.0001), indicating that the true direction of motion as given by the output of judged direction of motion of the stimulus was very direction-selective cells. However, motion-tilt induction much dependent on both the background orientation is strongest at the faster speeds that are sufficient to and the speed level. generate motion streaks. These results are consistent These findings are consistent with previous descrip- with the present hypothesis that the orientation of the tions of both static and kinetic versions of the Zollner motion streak generated by rapid motion is affected by illusion in which the angular separation between background lines and creates illusory tilt in the straight lines/motion paths and background lines are perceived direction of motion (see General discussion). exaggerated. Thus, in Experiment 1, to null the motion- It could be argued that the results shown in Figure 2 tilt-induction effect and perceive vertical movement, are accounted for by lateral inhibition between an the motion path of the dot cloud had to be physically orientation-tuned and motion-selective cell, particular- displaced in the direction of the background lines. ly if it is assumed that slower speeds result in reduced Third, the judged angle of motion tilt is dependent activation/response of motion-selective units. In this on object speed. For the slowest dot speed of 0.568/s, situation, the extent of lateral inhibition changes there was no apparent tilt; the judged angle of motion because the activation of direction-selective units is was approximately 08 regardless of the background line weaker. However, this account is improbable simply orientation, indicating veridical vertical movement. because the visual system does not code image speed in Post-hoc Bonferoni multiple comparisons (see Neter, this way. Existing research strongly indicates that Wasserman, & Kutner, 1990) revealed no significant image speed is coded in a few speed-tuned channels differences between the baseline condition (in which the (low and high) (see Edwards, Badcock, & Smith 1998; line orientation was 908) and when the background line Khuu & Badcock, 2002) with each mechanism equally was oriented (mean difference: 0.006, p ¼ 0.87). Given sensitive to speeds within a given range. The suggestion that the radius of individual dots was 0.0428, according that weaker motion-tilt induction occurs at slower to Geisler (1999), the critical speed at which motion speeds because motion detectors are not effectively streaks should be produced and contribute to percep- activated is therefore at odds with this evidence. tion is approximately 0.848/s. The lack of an effect Indeed, evidence exists showing that cells in cortical observed at this slow speed is consistent with the view areas such as in V1 and MT are selective for different that it is not sufficiently fast to generate a motion speeds (see Priebe, Cassanello, & Lisberger, 2003; streak, and thus, motion judgments are veridical. Priebe, Lisberger, & Movshon, 2006). However, as dot speed is increased, the judged motion direction deviated from 08 and was significantly different (from the baseline condition) by a dot speed of 1.128/s (mean difference: 0.948, p , 0.05). Thus, the Experiment 2: the effect of speed at which motion-tilt induction is observed in the stimulus contrast on the present study is close to the critical speed value at which motion streaks are thought to contribute to perceived path of motion perception. It was additionally observed that, for greater dot speeds (e.g., 98/s and 188/s), there is an Experiment 1 indicated that motion-tilt induction is increase in the extent of illusory tilt. This increase in strongest at speeds sufficient to generate a strong illusory tilt suggests that motion streaks provide a more motion streak. In Experiment 2, the possibility that effective cue in signaling motion direction at fast motion streaks underpin this illusion is further inves-

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 7

orientations produced only either strong motion-tilt induction or none. In different conditions, the Weber contrast of the dots forming the cloud stimulus was changed to the following levels: 2, 1, 0.5, 0.25, 0.125, and 0.0625. The lines forming the background were fixed to a contrast of 1.

Results and discussion

Data were similar across observers and were therefore averaged. The average results are shown in Figure 3 (error bars signify 95% confidence intervals). The judged angle of motion required to perceive the stimulus as moving vertically is plotted as a function of the Weber contrast of the dot cloud in log scale for backgrounds tilted at 158 and 908 (from vertical). A two-way repeated measures ANOVA performed on Figure 3. The judged angle of motion of a cloud of dots moving these data revealed that both the orientation of over tilted lines is plotted as a function of the Weber contrast of background lines (F[1, 60] ¼ 237, p , 0.0001) and dot dots (in log scale). Black circles indicate judgments with contrast (F[5, 60] ¼ 16.45, p , 0.0001) significantly background lines tilted 158, while gray squares depict the results affected the illusory motion-tilt effect. Additionally, a for conditions in which background lines were tilted 908 from significant interaction effect was found indicating that vertical. the effect of dot contrast at different levels is different for the two background line orientations (F[5, 60] ¼ tigated by varying stimulus contrast. Edwards and 15.63, p , 0.0001). Particularly, as shown in Figure 3, Crane (2007) showed that reducing the luminance when background lines are oriented 908, the perceived contrast, and therefore eliminating the availability of direction of the cloud stimulus was unaffected by motion streaks, reduced the detectability of global background lines with the judged angle of motion 08 direction of motion at high speeds. The logic of irrespective of contrast (gray squares). However, where Edwards and Crane can be applied to the kinetic background lines were tilted at 158, the judged angle of Zollner illusion; if the availability of motion streaks is motion (required to cancel the illusion) was dependent decreased or eliminated via a reduction in the on stimulus contrast with the extent of tilt evident for luminance contrast of dots, a corresponding reduction high, but not at low contrasts. Post-hoc tests (Bonfer- in their contribution to determining motion direction roni multiple comparisons) between the two back- should be expected. Accordingly, at low contrasts, the ground orientations of 158 and 908 for each contrast level indicated that the judged angle of motion was not perceived direction of motion of a fast-moving object significantly different (p . 0.05) for the two low ought to be unaffected by background tilted lines. Both contrasts 0.0625 (mean difference: 0.768, p ¼ 0.42) and Wallace (1975) and Li and Guo (1995) reported that 0.125 (mean difference: 0.288, p ¼ 0.63), but were the static Zollner illusion is dependent on luminance significantly different for higher contrasts of 0.25 (mean contrast with the strength of the illusion much weaker difference: 2.338, p , 0.0001), 0.5 (mean difference: at low contrasts. In further considering the role of 2.888, p , 0.0001), 1 (mean difference: 3.048, p , motion streaks in the kinetic Zollner illusion, Experi- 0.0001), and 2 (mean difference: 3.7378, p , 0.0001). ment 2 examined the effect of contrast change on the The results of this experiment are consistent with extent of motion-tilt induction. those of Wallace (1975) who noted that the extent of illusory tilt in the static Zollner illusion is dependent on the contrast difference between background and target Methods lines (also see Westheimer, Brincat, & Wehrhahn, 1999, for comparable effects with the static tilt illusion and Observers were the same as those in Experiment 1. Poggendorff illusions). Particularly, large contrast The stimuli and procedures were similar to those used differences between the target and background pro- in Experiment 1, except only one dot speed of 188/s was duces a larger tilt effect. According to Wallace, this used, and the background lines were either 158 effect can be accounted for by the degree of inhibition (rightward)-tilted lines or lines tilted at 908 (horizontal). between two orientation-tuned detectors responding As noted in Experiment 1, these background line differently to target and background lines of different

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 8

contrasts. In Experiment 2, it is presumed that, at very in Experiment 2) contrast dot clouds were presented low contrasts, motion streaks are not present, and (for 1 s) in sequential order (separated by a blank motion direction is derived solely from the output of interval of 0.5 s). After the presentation of both stimuli, motion detectors unaffected by background lines. This observers were required to reduce the speed of the high- observation with low-contrast stimuli is consistent with contrast stimulus at steps of 0.568/s until it perceptually those of Edwards and Crane (2007). However, at higher matched the speed of the lower-contrast cloud (which contrasts, motion streaks present and are affected by always moved at a speed of 188/s). This procedure was background lines leading to an observable change in repeated five times and the results averaged. Results the perceived direction of motion. This effect with high- indicated that observers perceived the low-contrast contrast stimuli is in agreement with the observations stimulus as moving much slower than the higher- of Apthorp et al. (2010) who demonstrated that motion contrast object. Observers on average had to reduce the streaks produced by fast-moving elements can mask the speed of the high-contrast object by approximately detection of a static grating provided that moving 30% for it to appear to move at the same perceptual elements are high contrast. speed as the lower-contrast object. While speed An alternative account of the results of Experiment 2 reduction was observed, the perceived speed (approx- might be that reducing the contrast of the object imately 12.68/s) is still sufficiently fast to generate a directly produces a change in lateral inhibition between motion streak. On the basis of this finding, it would a motion- and orientation-tuned cell because the appear that that a reduction in speed at low contrasts activation of motion-direction-selective cells to a low- cannot account for the findings of Experiment 2. contrast moving object might be comparatively weaker. In this case, a reduction in contrast will decrease the neural response of direction-selective units leading to a change in the extent of motion-tilt induction. However, Experiment 3: the effect of streak this account is improbable. A lateral inhibition length on the perceived path of mechanism would predict an enlargement of the tilt motion effect at low contrasts as higher-contrast background lines exert greater influence on the perceived motion direction. This is the opposite of the findings of the Previous research has shown that changing the present study. Additionally, previous findings have spatial length/trajectory of motion modulates motion- shown that the response of direction-selective units is streak perceptibility (Apthorp & Alais, 2009; Edwards largely unaffected by changes in luminance contrast, & Crane, 2007). For the present stimulus, this can be suggesting that contrast gain control is implemented to achieved by changing the lifetime of dots within the attenuate neuronal response relative to the contrast cloud before they are replotted to a random location in level (see Albrecht & Geisler, 1991). This physiological the stimulus. Long streaks are produced by ensuring evidence is well supported by behavioral data showing dots move along a fixed direction and extended that motion detection is largely independent of contrast duration (i.e., a long lifetime). By consequence, the above a critical value estimated to be as low as 5% arising motion streak extends over a large spatial (e.g., McKee, Silverman, & Nakayama, 1986). The distance. Conversely, short streaks are produced by contrast range employed in the present study is above assigning dots a short lifetime of one or two motion this level, making it unlikely that the reported reduction frames. Because such dots are present at a particular in the extent of motion tilt at very low contrasts is due spatial location for only a brief period of time, the to a reduction in activation of motion detectors leading traversed spatial distance is small, likewise the motion to a change in lateral inhibition. streak. Consistent with Edwards and Crane, if motion It is important to note that reducing the contrast of a streaks account for induced motion tilt, it would be moving object also reduces its apparent speed (see expected that short streaks would produce less motion Stone & Thompson, 1992), and therefore the noted tilt compared with long streaks. Alternatively, if reduction in motion path distortion reported in Figure motion-tilt induction is due to inhibition between form- 3 may reflect an elimination of motion streaks due to a and motion-selective units, and not motion streaks, the reduction in speed and not apparent contrast. To extent of the distortion might be the same for both ensure that this was not the case, a supplementary types of stimuli because motion produced with these speed-matching experiment was performed using Meth- spatial step-sizes will effectively drive local direction- od of Adjustment. The same observers (who partici- selective cells (provided that they fall within a cell’s pated in Experiment 2) had to match the speed of two receptive field; see Anderson & Burr, 1987; Morgan, dot clouds that differed in contrast. In this supplemen- 1992;Rudolph,Ferrera,&Pasternak,1994). In tary experiment, high- (Weber contrast: 1) and low- Experiment 3, the impact of short motion streaks was (Weber contrast: 0.0625, the lowest contrast value used examined by repeating the conditions of Experiment 1,

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 9

but with a background line spacing of 0.6728 (double which was significantly different, see following for that of Experiment 1) and with an object speed of 98/s, details). which corresponds to a spatial step-size of 0.2258/ While shortening the lifetime of dots resulted in a frame. These conditions are sufficient to optimize any reduction in illusory motion tilt, it did not completely differences in motion-tilt induction elicited by long and eliminate the effect. A likely explanation for this short motion streaks should such a difference exist. outcome is that, though streaks were made shorter, the spatial step-size (0.2558/frame) of the dots used in Experiment 3 might have been sufficiently large to Methods generate motion streaks capable of affecting motion perception. Comparatively, shorter motion streaks can The same observers who participated in the previous be generated by making dots jump over much smaller experiments were observers in Experiment 3. Stimuli distances, but at much higher refresh rates. It would be and procedures were similar to Experiment 1 except the expected that, if motion was generated in this manner, dot speed was 98/s and the background lines were tilted motion streaks will be considerably shorter, and under 908 and 158 from vertical. As shown in the previous these stimulus conditions may not affect perception. To experiments, tilted background of 158 produced com- verify this possible outcome, a control experiment in pelling distortion in the path of object motion, but not which Experiment 3 was repeated (only for a back- at 908. In different conditions, dots moved along a fixed ground line orientation of 158 and with the same trajectory for six frames (as in Experiments 1 and 2) to observers as in Experiment 3) with a higher refresh rate generate long motion streaks or were randomly of 120 Hz (such that each frame presented for 8.33 replotted after one frame-transition to generate short msec) and the spatial step-size of each dot of 0.0568/ motion streaks. As in the previous experiments, frame to produce a speed of 98/s. Thus, the spatial step- observers were required to judge whether the cloud of size used in the control experiment was four times dots was moving to the left or to the right of vertical, smaller than that used in the main experiment (which and the same staircase procedure was employed to null was 0.2258/frame). the motion-tilt-induction effect. The results of this control condition are shown in Figure 4 (black bar). A repeated measures ANOVA conducted on the results of Experiment 3 (for Results and discussion conditions in which background lines were 158) including the control data revealed a significant effect Figure 4 plots the averaged results of Experiment 3. of streak length (F[2, 4] ¼ 30.65, p , 0.0001) on the The judged motion direction of the cloud stimulus is judged motion direction. As evident in Figure 4, dot plotted for short (white bars) and long (gray bars) motion producing shorter motion streaks (gray and streak conditions for background orientations of 908 black bars) resulted in a comparatively smaller illusory and 158. Error bars signify 95% confidence intervals. A motion-tilt effect than when motion streaks were long repeated measures two-way ANOVA confirmed that (white bar). Indeed, post-hoc mean comparisons the judged motion direction were dependent on streak (Tukey, Honestly Significant Difference test) between length (F[1, 16] ¼ 5.50, p , 0.01) and background the long motion streaks condition and the two short orientation (F[1, 16] ¼ 116.62, p , 0.0001) as well their motion streaks conditions indicated that were signifi- interaction (F[1, 16] ¼ 11.64, p , 0.001) indicating that cantly different: long and shorter motion streaks in the effect of streak length was different for the two which the spatial step-size was 0.2258/frame (mean types of background-line orientations. Particularly as difference 2.297, p , 0.01); long and shortest motion shown in Figure 4, where background lines were tilted streaks in which the spatial step-size was 0.0568/frame 908 from vertical, observers were able to accurately (mean difference 3.663, p , 0.001). Additionally, the detect the veridical motion of the stimulus with the two short motion streaks were also significantly judged angle of motion approximately 08. Moreover, different such that the illusory motion tilt observed with background lines at 908, there was no difference with motion streaks generated with a spatial step-size between short and long streak conditions. However, of 0.2258/frame was higher than when the spatial step- when background lines were tilted, the perceived size was 0.0568/frame (mean difference 1.365, p , direction of motion indicated an induced motion tilt. 0.05). The results of this control experiment demon- To nullify this illusory effect, the judged angle of strates that motion streaks underlie the perception of motion was tilted in the direction of tilted lines. the kinetic Zollner illusion such that reducing the Importantly, for these background types, the extent length of motion streaks, by shortening the lifetime of of the distortion is dependent on streak length. Long dots, decreases the extent of illusory motion-tilt streaks resulted in a larger motion displacement induction. Particularly, for the shortest motion streaks compared with short streaks (mean difference: 2.308, condition (black bar), there appears to be little or no

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 10

that the motion-tilt-induction effect was significantly reduced at comparatively low dot contrasts. Finally, Experiment 3 varied motion-streak length by changing the lifetime of dots. It was found that short streaks produced a weaker tilt-induction effect than long streaks. Together, these findings demonstrate that, while motion streaks may not be perceptible (due to motion deblurring rendering clear vision: Burr, 1980; Burr & Morgan, 1997), the visual system is sensitive to this form cue and that they contribute directly to the perception of motion. The present findings add to a growing body of literature highlighting the importance of motion streaks to the perception of motion, and more specifically, add to the literature indicating that the processing of form and motion is interactive at early stages of processing. Kinetic illusions represent a situation in which back- ground form information affects the perceived path of motion. Clearly then, the prevailing assumption of Figure 4. The judged angle of motion is plotted for long (white separate form and motion analysis in the early stages of bars) and short (gray bars) streaks conditions for background line (see De Yoe & Van Essen, 1988; orientations of 908 and 158 denoted along the x-axis. The black Livingstone & Hubel, 1987) does not account well for bar for background line orientation of 158 represents the judged these percepts. Separate early form and motion analysis angle of motion for short motion streaks in which the spatial step- would predict that the perception of motion is size was much smaller at 0.0568/frame. unaffected by spatial orientation. The influence of form information on direction of motion in these illusory effect (mean: 0.5168, 95% confidence limits: 6 kinetic illusions would instead suggest interaction 0.8668) as the judged motion direction for this between form and motion processing in the visual condition is close to zero indicating veridical motion system (see Lorenceau & Alais, 2001; Or et al., 2007; perception. Ross, 2004; Ross, Badcock, & Hayes, 2000). The cumulative evidence from the three experiments pre- sented here suggests that this interaction is instead likely to arise from the influence of motion streaks on General discussion the perceived direction of motion. As previously mentioned, Geisler (1999) describes a The goal of this study was to determine whether the computational model in which perceived direction of induced motion-tilt effect in the kinetic Zollner illusion motion can be viewed as a product of the output of a could be understood in terms of motion streaks. It was motion-tuned cell and a line-sensitive cell tuned to the reasoned that, if motion streaks were responsible for tilt orthogonal orientation. The findings of the present induction, attenuating the availability or salience of study are best explained by Geisler’s model. Under streaks would reduce or eliminate the illusory effect. In conditions in which strong motion streaks are produced three experiments, the relative contribution of motion (i.e., fast dot speeds, high contrast, and extended streaks to the perception of motion was examined by lifetimes), tilt induction is observed because back- changing dot speed, dot contrast, and dot lifetime to ground lines distort the streak (arising from lateral generate shortened streaks. Previous investigations inhibition between cells coding orientation). When have shown that such manipulations impact on the combined with the output of direction-selective cells, availability of motion streaks in the computation of the perceived path of motion is itself distorted. motion direction (Apthorp & Alais, 2009; Apthorp However, when motion-streak information is reduced, et al. 2010; Edwards & Crane, 2007; Li et al., 2008). this process does not occur. Under these circumstances, Experiment 1 examined the effect of dot speed on the the visual system relies solely on the output of motion- extent of tilt induction. It was found that increasing dot selective cells to determine motion direction, which speed resulted in a large motion-tilt effect, particularly remains unbiased by background lines. This account of at speeds sufficient to generate a motion streak. the kinetic Zollner illusion differs from that offered by Experiment 2 examined the salience of motion streaks Swanston (1984) who suggests that motion-tilt induc- by altering the luminance contrast of dots. It was found tion arises from lateral inhibition directly between

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 11

motion-selective and orientation-tuned mechanisms. tuned mechanisms processing the motion streak pro- Importantly, as mentioned, this solution does not duced by object motion and background lines. Motion provide an adequate account of the changes in streaks are used by the visual system to signal the motion-tilt induction accompanying a reduction in direction of motion, leading to tilt in the perceived path object speed, contrast, and motion-streak length that of object motion. This was confirmed in the present are reported in the present study. study by showing that attenuation of the availability of The findings of the present study are consistent with motion streaks through a reduction of speed, contrast, the observations of Krekelberg et al. (2003), Ross and streak length reduced the extent of motion-tilt (2004), and Or et al. (2007, 2010) who demonstrated induction. that perceived direction of motion is a combination of the spatial orientation of an object and its motion direction. In those studies, Glass patterns were used to show that dipole orientation is treated as a motion Acknowledgments streak that affects the perceived direction of motion. It is likely that an analogous operation occurs for motion- I thank the observers who participated in the study tilt induction; static background lines induce tilt in the and Joanna Kidd for her helpful comments. I also orientation of motion streaks. thank the two anonymous reviewers for their helpful The findings of the present study indicate that comments. This research was supported by an Austra- motion streaks can influence the direction of motion lian Research Council (ARC) Discovery Project Grant even though they are imperceptible. It is known from (DP110104713) to S. Khuu. studies of visible persistence (e.g., Coltheart, 1980;Di Commercial relationships: none. Lollo & Hogben, 1985) that the percept of a briefly Corresponding author: Sieu K. Khuu. presented visual object persists for approximately 120 Email: [email protected]. ms. One would predict, then, that the percept of a fast- Address: The School of Optometry and Vision Science, moving object ought to be highly blurred. However, The University of New South Wales, Sydney, New this does not occur. Moving objects are typically sharp South Wales. in appearance, and their apparent position is clear (Ogmen,¨ 2007). Previous studies have attributed clear perception of the form of a moving object to the active processes of motion ‘‘deblurring’’ (Burr, 1980, 1981). References While the nature of the mechanisms underlying motion deblurring is still a matter of debate (see Burr & Albrecht, D. G., & Geisler, W. S. (1991). Motion Morgan, 1997; Ogmen,¨ 2007; Bex, Edgar, & Smith, selectivity and the contrast response function of 1995), the process obviously removes the presence of simple cells in the visual cortex. Visual Neurosci- motion streaks from conscious perception (under ence, 7, 531–546. normal viewing conditions), especially when the stim- ulus is high contrast and at fast speeds (see Bex et al., Albright, T. D. (1984). Direction and orientation 1995). Despite this, motion streaks are unequivocally selectivity of neurons in visual area MT of the Journal of Neurophysiology, 52 used by the visual system to compute direction of macaque. , 1006– motion before motion deblurring occurs. As Geisler 1130. (1999) notes, this arrangement exists because form Alais, D., Apthorp, D., Karmann, A., & Cass, J. mechanisms offer greater precision to the computation (2011). Temporal Integration of Movement: The of motion direction. However, as the present study Time-Course of Motion Streaks Revealed by suggests, the integration form and motion information Masking. PLoS ONE, 6(12), e28675. doi: 10.1371/ does not always lead to accurate estimation of motion journal.pone.0028675. direction, but can cause measurable misperception in Anderson, S. J., & Burr, D. C. (1987). Receptive field motion direction. size of human motion detection units. Vision Research, 27, 621–635. Apthorp, D., & Alais, D. (2009). Tilt aftereffects and Conclusion tilt illusions induced by fast translational motion: Evidence for motion streaks. Journal of Vision, 9(1): In conclusion, the present study reports that the 27, 1–11. http://www.journalofvision.org/content/ motion-tilt induction due to oriented background lines 9/1/27, doi: 10.1167/9.1.27. [PubMed][Article]. can be explained by interactions between orientation- Apthorp, D., Cass, J., & Alais, D. (2010). Orientation

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 12

tuning of contrast masking caused by motion visible persistence. Journal of Experimental Psy- streaks. Journal of Vision, 10(10):11, 1–13. http:// chology: Human Perception and Performance, 11, www.journalofvision.org/content/10/10/11, doi: 10. 304–316. 1167/10.10.11. [PubMed][Article]. De Bruyn, B., & Orban, G. A. (1988). Human velocity Badcock, D. R., & Dickinson, E. J. (2009). Second- and direction discrimination measured with ran- order orientation cues to the axis of motion. Vision dom dot patterns. Vision Research, 38, 1323–1335. Research, 49, 407–415. De Yoe, E. A., & Van Essen, D. C. (1988). Concurrent Barlow, H. B. (1958). Temporal and spatial summation processing streams in monkey visual cortex. Trends in human vision at different background intensities. in Neuroscience, 11, 219–226. Journal of Physiology, 141, 337–350. Eagleman, D. M. (2001). Visual illusions and neurobi- Bex, P. J., Edgar, G. K., & Smith, A. T. (1995). ology. Nature Neuroscience, 2, 920–926. Sharpening of blurred drifting images. Vision Edwards, M., Badcock, D. R., & Smith, A. T. (1998). Research, 35, 2539–2546. Independent global motion systems. Vision Re- Blakemore, C., Carpenter, R. H. S., & Georgeson, M. search, 38, 573–580. A. (1970). Lateral inhibition between orientation Edwards, M., & Crane, M. (2007). Motion streaks detectors in the human visual system. Nature, 228, improve motion detection. Vision Research, 47, 37–39. 828–833. Blakemore, C., & Tobin, E. (1972). Lateral inhibition Fineman, M. B., & Melingonis, M. P. (1977). The effect between orientation detectors in the cat’s visual of a moving dot transversal on the Poggendorff cortex. Experimental Brain Research, 15, 439–440. illusion. Perception & Psychophysics, 21, 151–156. Burmester, E. (1896). Contributions to experimental Geisler, W. S. (1999). Motion streaks provide a spatial determination of the optical-geometical illusions. code for motion direction. Nature, 400, 65–69. Zeitschrift fur Psychologie, 12, 355–394. Gillam, B. J. (1998). Illusions at Century’s End. In Burr, D. (1980). Motion smear. Nature, 284, 164–165. Perception and Cognition at Century’s End in series Burr, D. (1981). Temporal summation of moving The Handbook of Perception and Cognition. Hoch- images by the human visual system. Proceedings berg, J., Series eds. Carterette, E. & Friedman, M of the Royal Society of London B: Biological (eds.). NY: Academic Press. Sciences, 211, 321–339. Hering, E. (1861). Contributions to physiology, volume Burr, D. C., & Morgan, M. J. (1997). Motion 1. Leipzig: Engelman. deblurring in human vision. Proceedings of the Kelly, D. H. (1961). Visual responses to time-depen- Royal Society B, 264, 431–436. dent stimuli. I. amplitude sensitivity measurements. Burr, D., & Ross, J. (2002). Direct evidence that Journal of the Optical Society of America, 51, 422– ‘speedlines’ influence motion mechanisms. Journal 429. of Neuroscience, 22, 8661–8664. Khuu, S. K., & Badcock, D. R. (2002). Global speed Carpenter, R. H. S., & Blakemore, C. (1973). Interac- processing: evidence for local averaging within, but tions between orientations in human vision. Exper- not across two speed ranges. Vision Research, 42, imental Brain Research, 18, 287–303. 3031–3042. Cesaro, A. L., & Agostini, T. (1998). The trajectory of a Krekelberg, B., Dannenberg, S., Hoffmann, K.-P., dot crossing a pattern of tilted lines is misperceived. Bremmer, F., & Ross, J. (2003). Neural correlates Perception and Psychophysics, 60, 518–523. of implied motion. Nature, 424, 674–677. Coltheart, M. (1980). Iconic memory and visible Lennie, P. (1998). Single units and visual cortical persistence. Perception & Psychophysics, 27, 183– organization. Perception, 27, 889–935. 228. Li, C. Y., & Guo, K. (1995). Measurements of Coren, S., & Girgus, J. S. (1978). Seeing is deceiving: geometric illusions, illusory contours and stereo- The psychology of visual illusions. Hillsdale, NJ: depth at luminance and colour contrast. Vision Erlbaum. Research, 35, 1713–1720. Day, R. H. (1973). The oblique line illusion: The Li, W. O., Khuu, S. K., & Hayes, A. (2008). Interaction Poggendorff effect without parallels. Quarterly between complex motion patterns in the perception Journal of Experimental Psychology, 25, 535–541. of shape. Vision Research, 48, 167–178. Di Lollo, V., & Hogben, J. H. (1985). Suppression of Livingstone, M. S., & Hubel, D. H. (1987). Psycho-

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 13

physical evidence for separate channels for the macaque area MT/V5. Journal of Neuroscience, perception of form, color, movement, and depth. 23, 5650–5661. Journal of Neuroscience, 7, 3416–3468. Priebe, N. J., Lisberger, S. G., & Movshon, J. A. Lorenceau, J., & Alais, D. (2001). Form constraints in (2006). Tuning for spatiotemporal frequency and motion binding. Nature Neuroscience, 4, 745–751. speed in directionally, selective neurons of macaque McKee, S. P., Silverman, G., & Nakayama, K. (1986). striate cortex. Journal of Neuroscience, 26, 2941– Precise velocity discrimination despite random 2950. variations in temporal frequency and contrast. Prinzmetal, W., & Beck, D. M. (2001). The Tilt- Vision Research, 26, 609–619. Constancy Theory of Visual Illusions. Journal of Morikawa, K. (1987). Effects of orientation-selective Experimental Psychology, 27, 206–217. adaptation on the Zollner illusion. Perception, 16, Ross, J. (2004). The perceived direction and speed of 473–483. global motion in Glass pattern sequences. Vision Morgan, M. J. (1992). Spatial filtering precedes motion Research, 44, 441–448. detection. Nature, 355, 344–346. Ross, J., Badcock, D. R., & Hayes, A. (2000). Coherent Morgan, M. J., & Casco, C. (1990). Spatial filtering global motion in the absence of coherent velocity signals. Current Biology, 10, 679–682. and spatial primitives in early vision: an explana- tion of the Zollner-Judd¨ class of geometrical Rudolph, K. K., Ferrera, V. P., & Pasternak, T. (1994). illusion. Proceedings of the Royal Society B, 242, A reduction in the number of directionally selective 1–10. neurons extends the spatial limit of global motion perception. Vision Research, 34, 3241–3251. Neter, J., Wasserman, W., & Kutner, M. H. (1990). Applied Linear Statistical Models.3rdedition, Snowden, R. J., & Braddick, O. J. (1991). The temporal Irwin. integration and resolution of velocity signals. Vision Research, 31, 907–914. Nihei, Y. (1973). A preliminary study on the geomet- rical illusion of motion path: The kinetic illusion. Stone, L. S., & Thompson, P. (1992). Human speed Tohoku Psychologica Folia, 32, 108–114. perception is contrast dependent. Vision Research, 32, 1535–1549. Nihei, Y. (1975). The effect of direction motion on the magnitude of the geometrical illusion of motion Swanston, M. T. (1984). Displacement of the path of path: The kinetic illusion (II). Tohoku Psychologica perceive movement by intersection with static Folia, 34, 88–94. contours. Perception and Psychophysics, 36, 324– 328. Ogmen,¨ H. (2007). A theory of moving form percep- tion: Synergy between masking, perceptual group- Wallace, G. K. (1969). The critical distance of ing, and motion computation in retinotopic and interaction in the Zollner illusion. Perception and non-retinotopic representations. Advances in Cog- Psychophysics, 5, 261–264. nitive Psychology, 3, 67–84. Wallace, G. K. (1975). The effect of contrast on the Or, C. C.-F., Khuu, S. K., & Hayes, A. (2007). The role Zollner Illusion. Vision Research, 15, 963–966. of luminance contrast in the detection of global Wallace, G. K., & Crampin, D. J. (1969). The effect of structure in static and dynamic, same- and oppo- background density on the Zollner Illusion. Vision site-polarity, Glass patterns. Vision Research, 47, Research, 9, 167–177. 253–259. Watamaniuk, S. N. J. (2005). The predictive power of Or, C. C.-F., Khuu, S. K., & Hayes, A. (2010). Moving trajectory motion. Vision Research, 45, 2993–3003. Glass patterns: asymmetric interactions between Wenderoth, P., & Johnson, M. (1983). Relationships motion and form. Perception, 39, 447–463. between the kinetic, alternating-line, and Poggen- Orbison, W. D. (1939). Shape as a function of the dorff illusions: The effect of inter-stimulus interval, vector field. American Journal of Psychology, 52, inducing parallels, and fixation. Perception and 31–45. Psychophysics, 34, 273–279. Peterson, M., Ohzawa, I., & Freeman, R. (2001). Westheimer, G. (2008). Illusions in the spatial sense of Neural and perceptual adjustments to dim light. the eye: Geometrical-optical illusions and the Visual Neuroscience, 18, 203–208. neural representation of space. Vision Research, Priebe, N. J., Cassanello, C. R., & Lisberger, S. G. 48:2128–2142. (2003). The neural representation of speed in Westheimer, G., Brincat, S., & Wehrhahn, C. (1999).

Downloaded from jov.arvojournals.org on 09/27/2021 Journal of Vision (2012) 12(6):19, 1–14 Khuu 14

Contrast dependency of foveal spatial functions: tion of direction of motion in human vision. orientation, vernier, separation, blur and displace- Journal of Neurophysiology, 71, 33–37. ment discrimination and the tilt and Poggendorff Zollner, F. (1860). On a new kind of pseudoscopic illusions. Vision Research, 39, 1631–1639. viewing and its relationship with the motion- phenomenon described by Plateau and Oppel. Westheimer, G., & Wehrhahn, C. (1994). Discrimina- Annalen der Physik, 110, 500–525.

Downloaded from jov.arvojournals.org on 09/27/2021