Based on these findings, we conclude that our visual system does not contain low-level motion receptors encoded for rotational motion, and thus rotation detection is not subject to hysteresis. Braddick first proposed that our visual system consists of two distinct systems consisting of short-range and long-range processes. The long-range system operates later in our visual system and is responsible for detecting larger displacement movements.
Our visual system is also able to perceive motion when no actual motion is taking place, a phenomenon called apparent motion. Braddick discovered that when objects are moved over a relatively large distance, our visual system uses the 'long-distance system'. For example, if vertical motion is observed in the figure above, our visual system will hold its perception before switching to sensing.
Hysteresis can help with motion perception because it allows our visual system to stabilize an ambiguous stimulus presentation. If no hysteresis is experienced, we can argue that our visual system does not contain low-level rotation receptors.
Methods
An example of a static stimuli would be two tokens that rotate by 18 degrees between each frame, and an example of a dynamic stimuli would be two tokens that initially rotate by 18 degrees, then 20 degrees, then 24 degrees, then 30 degrees ... etc. if the pitch angle was 2 degrees). For the static stimulus trials, the two tokens were rotated at a constant angle, leading to the perception of either clockwise, counterclockwise, or chaotic motion. The task for the participants was to press a corresponding key depending on the perceived movement type.
Thus, with 25 angles x 2 frame durations x 5 repetitions, we had a total of 250 trials, and the trials were presented in random order. Dynamic stimuli consisted of cues that had an initial rotation angle of 18 degrees and a final rotation angle of 162 degrees, and the rotation angle increased in increments of either 2 or 6 degrees between frames. The initial rotation angle remained at 18 degrees until the participant made a response indicating the type of motion they perceived.
The first rotation angle was either to the left or to the right (these were randomly assigned but not analyzed separately, as no effect was expected); so the signs initially appear to move in a clockwise or counter-clockwise direction. As the angle of rotation increases, the signs eventually seem to move chaotically, and then they eventually seem to move in the opposite direction from the initial clockwise or counter-clockwise direction. When neither direction of rotary motion is observed, the sign's motion is as explained in Figure 1, but it is perceived as chaotic because the square is constantly changing its orientation.
The dynamic stimuli had two different frame durations, two possible gain angles, and 10 repetitions for each condition, giving us a total of four experimental conditions and 40 trials. We tested 7 participants who each completed 8 testing sessions, which consisted of 250 presentations of static stimuli and 40 presentations of dynamic stimuli. All stimuli were presented on a Mistubishi DiamondPro monitor, powered by a VSG2/5 Cambridge Research Systems graphics card, and the refresh rate was 120 Hz.
The radius of the imaginary circle on which the cues were presented was 1.5 degrees of visual angle.
Results
The purpose of the static stimulus was to establish a baseline for which angle of rotation produced which direction of perceived motion (clockwise, counterclockwise, or chaos) without the effect of hysteresis. To find a single numerical value to use as a baseline, we collapsed the static stimulus data before fitting the function. Because we are more interested in determining whether the participant perceived rotation versus chaos, direction of rotation was not a significant factor for this analysis.
For example, a participant will perceive clockwise motion at 18 degrees of rotation almost 100% of the time as he or she will almost 100% of the time. Data represent the mean angular rotation thresholds for the collapsed static stimuli for both frame durations. For the dynamic stimulus, participants completed 40 trials with a total of eight different experimental conditions.
These thresholds represent the rotation angle at which the switch from detected rotation to 167 ms 250 ms. The graph is limited to 90 degrees along the x-axis because all data points have been collapsed, with each line in the graph representing a different function. The black line represents the static condition, and the data values for this function show the proportion of trials in which the participant detected clockwise rotation for one angle or counterclockwise rotation for the mirror image of that angle.
For example, the black data point at 54 degrees represents the proportion of clockwise rotation perceived at 54 degrees or the proportion of perceived. The red curve represents the proportion of trials where rotation was perceived in the dynamic condition when the angle of rotation was increasing (ie, moving from left to right in the graph). The green curve also represents the dynamic state, but when the rotation angle was decreasing (ie going from right to left in the graph), it represents going from chaos to rotation.
To examine the effects of hysteresis, a single numerical value representing the shift between the red curve and the black curve is compared to the value representing the shift between the green curve and the black curve.
Hypothetical Graph Displaying Effects of Hysteresis
Threshold Differences (“Shift Values”)
A 2x2x2 ANOVA revealed that direction (either coming from chaos to rotation or vice versa) had a significant effect on the angle at which the change in direction occurred. This significant effect is consistent with our hypothesis that hysteresis does not occur in the RC direction (rotation in chaos). This is likely because the hysteresis duration is not only a function of angle but also of time.
When the frame duration is short, a larger "offset value" is achieved in the same period of time compared to a longer frame duration, which is consistent with the idea that simply passing time would break the hysteresis. Similar to the frame duration effect, this is likely because a shorter elevation angle yields a larger "shift value" for the same time period compared to a larger elevation angle. Average "displacement values" were plotted and analyzed in terms of angle; however, they could also have been mapped and analyzed in terms of time.
When analyzed as a function of time, the 'shift values' for the two frame durations are actually quite similar. For example, the average shift value for direction CR and frame duration of 167 ms is 20 degrees, and the increment angle was 2 degrees (20/2), a total of 10 frames. Therefore, if the average “shift values” were plotted in terms of time, the points on the graph would coincide.
No significant effect was found for the interaction of direction and angle of growth with F(1,5) < 1, p = .425. The interaction between frame duration and rise angle also did not produce a significant effect with F p = .262.
Frame Duration v. Direction
Discussion
To determine the effect of hysteresis on the perception of rotational motion, participants viewed sham motion stimuli that consisted of tokens that appeared to move in a clockwise, counterclockwise, or chaotic direction. We hypothesized that if our visual system lacks rotation-encoded receptors early in the motion detection process, hysteresis will not affect rotational motion detection. To provide support for this hypothesis, hysteresis in the sense of this experiment should have an effect in detecting the switch from rotational motion to chaos; however, our data do not support this claim.
Our results suggest that hysteresis has no effect in the transition from rotation to chaos, but instead has an effect in perceiving the transition from chaos to rotational motion. Our visual system consists of selective neurons that become excited by the perception of a series of similar directions. For example, a right-diagonal motion direction would excite many neurons coded for similar directions, not just that specific orientation, while also suppressing neurons selectively coded for other directions.
For example, neurons coded for motion by say 10 degrees will also respond well to motion by 20 or 0 degrees. When they perceive the chaotic state, these types of neurons are not aware of the overall perception of chaos, but instead respond to the basic movement of the stimulus, which makes them subject to hysteresis. For example, in the condition where the elevation angle is 2 degrees and the cues start with, say, an initial angle of 50 degrees, neurons coded for 50 degrees and 140 degrees (perpendicular to 50 degrees) will show dominance start competing. populations of neurons become excited, and the next frame exhibits a rotation of 52 and 142 degrees, which is still well within the tuning range of those neurons.
Because the difference between the angles is only 2 degrees and the range of tension too. direction-selective neurons are relatively large, there will be very little difference in the amount of neural excitation. They function so early in the visual system that they cannot tell the difference between the two stimuli. Our results support this hypothesis because the effects of hysteresis were only observed to move from the chaotic direction to rotation. only during this chaotic stage, strongly suggesting that we do not have early motion detectors coded for rotational motion, thus confirming that a rotational percept is not subject to hysteresis.
Evidence for three-dimensional spatial representation. for the detection of apparent motion of a single element: Bistability from local cooperativity.
