Session Detail

The middle temporal area (MT) computes motion direction based on the inputs it receives from direction-selective neurons in primary visual cortex (V1). While V1 neurons signal the direction of ‘local’ moving edges, MT neurons compute a ‘global’ direction that is consistent for many types of spatial patterns. We do not know how information from populations of V1 neurons is combined to produce the selectivity observed in MT. To examine how visual information is successively represented by V1 and MT, we used separate multi-electrode arrays in each area to measure neural responses simultaneously from dozens of neurons in anaesthetised marmosets. We recorded neural activity while presenting motion with different kinds of visual patterns: fields of dots, sine waves, and square waves. All of these patterns evoke strong motion signals in MT, but recruit distinct V1 populations. To examine how the V1 representation related to the representation produced in MT, we measured tuning of individual neurons, receptive field overlap, and the depth of the MT neurons. We then related these quantities to noise correlations in the populations within and between areas for different visual patterns. These results will tell us more about how information is transformed between brain areas to produce perception.

The spinner illusion is an academic form of ‘The Coyote Illusion’ (Ho & Anstis, 2013 Best Illusion of the Year contest), demonstrating that moving stimuli appear faster with higher spatial frequencies when the physical speed is constant (Ashida, Ho, Kitaoka, and Anstis, 2017, i-Perception). One problem is that the illusion persists up to 16 elements per revolution with the original ring of spots while the effect almost saturates at around 8 cycles per revolution (c/rev) with sinusoidal gratings. The reason for this discrepancy could be that the spot stimulus has sharp edges that introduce harmonic components of higher spatial and temporal frequencies. We therefore measured the spinner illusion in radial gratings of 8 c/rev vs 16 c/rev, with either sine-wave (no harmonics) or square-wave (having odd spatial harmonics) modulation of luminance. We found that the square-wave stimuli yielded a larger effect of speed overestimation for 16 c/rev than the sine-wave stimuli. This difference could be mainly attributed to the observation that 8-c/rev square-wave stimulus was perceived slower than the 8-c/rec sine-wave stimulus. These results, which are qualitatively consistent with Brooks, Morris, and Thompson (2011, J. Vis.), demonstrate a crucial role of harmonic components in speed perception.

Visual transduction at photoreceptor level in the retina is characterized by temporal dynamics that effectively implies spatio-temporal filtering at the front-end of vision, particularly enhanced for moving visual stimuli. A range of patterns that induce apparent motion, including the 'rotating snakes' and 'Ouchi' illusions, are known to elicit strong effects under at least minimal eye movements and with a constellation of visual features that act as motion 'triggers'. Here, we revisit the Ouchi illusion to show how a computational model of visual transduction in a generic cone responds to different motion trigger features. Motion blur patterns generated by the model suggests that moving orientation, contrast and spatial frequency gradients result in motion blur that could potentially cause a subsequent edge detection model to induce a 'false' relative spatial displacement between elements across a feature gradient - corresponding to the experienced apparent motion experienced by a human subject. As a windfall, the model also generates output intensity maps that partially correspond to transitory contrast - or scintillating 'illusory contrast' - patterns experienced by human subjects viewing stimulus patterns, such as the Ouchi pattern.

Parkinson’s disease with dementia (PDD) is a progressive degenerative brain disease leading to severe deficits in motor and mental capabilities. Perceiving and interpreting body movements comes easily for humans, even when depicted by point-light-display (PLD). Our main question is, if patients with PDD are incapable of executing movement, are they also unable to comprehend movements? We tested 8 patients with PDD (4 females; mean age: 76.12) and 5 healthy controls (3 females; mean age: 63.6), with biological-motion-recognition and goal-intention-task. The biological-motion-task contained 12-PLD clips depicting human motions (i.e. jumping-jack) and each was displayed 3 times. The goal-intention-task contained 4 films; each has a familiarization (participants were to answer which stuffed-animal is grabbed) and a test (with the model’s hand raised in the center and paused). Participants were asked “What will she do next?” and “Which will she grab?” In the biological-motion-task, the healthy controls had an average of 10.2 (SD=1.30) correct answers out of 12, while the PDD group performed poorly with only 2.25 (SD=2.25). In the goal-intention-task, both groups showed consistent answers in the familiarization, but the PDD patients responded unpredictably in the test. In sum, our preliminary finding suggests that patients with PDD have difficulty in comprehending movement.

Visual perception is serially dependent, influenced by immediate past experience. Here, we examined whether time perception was susceptible to the recent history of temporal information. Participants were required to reproduce time intervals (810 - 1200 ms) in either a unisensory or a multisensory context. In unisensory tasks, sample intervals were presented in a single visual or auditory modality; in multisensory tasks, half the intervals (those of shorter length) and the other half (intervals of longer length) were presented in different modalities, visual and auditory. We found that reproduced times were biased towards the mean of the distribution of time intervals in both unisensory and multisensory tasks, suggesting that participants’ timing was influenced by the prior distribution of time intervals. In addition to this classic central tendency effect, we also found positive serial dependencies in both unisensory and multisensory tasks. However, for the multisensory tasks, further analysis showed that the positive serial dependencies only appeared if the previous trial and the current trial came from the same modality. Our findings suggest that two types of past experience influence current timing: a long-term prior which is represented in a supramodal manner, and a short-term serial dependence bias which is exclusively unimodal.

Light plays an important role in modern society as it affects both image-forming and non-imaging-forming function. Here we investigated the effect of blue light on time perception, one of the most important cognitive functions maintaining essential social interactions. Previous studies on this issue rendered inconsistent results; here we re-examine it using an oddball paradigm which is more sensitive than production and reproduction paradigms with a systematic manipulation of lights. In the oddball paradigm, participants were asked to judge the duration of the target, compared to that of the standard. With the condition of either blue or red background light in Experiment 1, participant’s time perception was lengthened with blue light. Experiment 2 further clarified the contribution of a recently discovered type of retinal ganglion cells that are intrinsically photosensitive (ipRGCs) which is especially sensitive to blue light, with a multi-primary stimulation system that can increase the stimulation of ipRGCs with a metameric background. Results showed that increased stimulation of ipRGC lengthened time perception. These results suggest that blue light expands subjective duration mainly through the contribution of ipRGCs. These results shed lights on further investigations of how ipRGCs affect the timing mechanism and future applications in media and lighting designs.

The visual system constructs representations of the world through repeated observations as suggested by contextual cueing effect (CCE). CCE is the learning effect of spatial layouts revealed by reaction time shortening in visual search due to repeating presentations of the same layouts. We built a model that predicts reaction time shortening by CCE, using reinforcement learning of relationships between the target location and global features of layouts. The relationship is expressed by mapping probability of the target location on each layout. The probability map is used to weight the saliency map obtained based on visual features in order to predict where to attend for searching a target. With successful learning, the probability map is expected to show the largest probability at the actual target location. The learning speed depends on a parameter of reinforcement and we obtained the parameter for the best prediction of psychophysical experiments. The CCE was measured either with typical letter stimuli or with natural scenes. The model simulation showed that the larger reinforcement effect is required with natural scenes than with letter stimuli. This may suggest that the visual system is designed to learn natural scenes.