Shimojo, K. | Search bioRxiv

bioRxiv2024

"Magnetic Sand" : Illusions of Interactivity

We found a series of new illusions, in which actions performed near a random white-noise display lead to the perception that the display is altered interactively with the observers actions. The perceptions resemble interactions with a box of magnetic sand, where the hand can leave traces, or attract and repulse grains in its vicinity. 1) the observer puts a finger very close to a dynamic noise display, slowly moving as though drawing a letter or a shape. A trace appears left in the fingers path, decaying within 500 ms or so. 2) When the observer moves their palm toward and away from the display, opening and closing their fingers as if grabbing and releasing grains of sand, the random dots appear as though they were magnetically attracted to or repelled by the fingers. 3) When an open hand close to the display is slowly moved back and forth laterally, the nearby dots appear to get attracted to or captured by the fingers and thus appear to move with them. 4) The same kind of action capture occurs even when the hand is not visible, moving behind the display. These illusions are robust across a wide range of parameters, including frame rate, luminance contrast, dot size (spatial frequency), and finger movement factors. Inter-subject variability is not correlated across illusion types, and the illusions also diverge in behavior across dynamic and static noise conditions. This indicates that multiple mechanisms are involved to different extents across illusions. Several known visual motion detectors and other low-level mechanisms may be involved in seeding the perceptual phenomena. However, a complete explanation would require mechanisms of action capture, whereby the internal model of the persons actions and their predicted consequences organizes visual attention and processing of the random stimulus components.

neuroscience↗

bioRxiv2022

Preservation of Conditioned Behavior Based on UV Light Sensitivity in Dissected Tail Halves of Planarians- a Proof by DNN

Planarians are aquatic worms with powerful regenerative and memory retention abilities. This paper examines whether a dissected tail half of a Planarian (Dugesia Dorotocephala) can retain and exhibit a previously-conditioned response, possibly before the regeneration of the head and the ganglia. We conditioned intact Planarians in a Pavlovian procedure with an electric shock (ES) as the unconditioned stimulus and weak ultraviolet (UV) light as the conditioned stimulus. Then, we dissected their bodies into halves, keeping the dissected tail halves. Starting from the 2nd day after dissection, we presented the same UV light 3 times daily while video-recording the responses. The recorded responses were then classified by a DNN: a VGG16 model was pre-trained by ImageNet for extracting features from images and additionally trained with 211 responses to ES and 118 to UV light before conditioning/dissection to categorize planarians reactions into "UV-induced" or "ES-induced" reactions. The cross-validated accuracy in categorization was 83.6%. We then let the DNN analyze 99 recorded responses to UV from 20 individual conditioned tail halves. 96.8 % of their reactions were classified as "ES-induced" (against 22.0% wrongly classified as "ES-induced" for unconditioned samples under UV), indicating they have shown the "Conditioned Response" (p<3.06E-30). This provides evidence that planarians can conserve and reveal a learned response even without the head/ganglia, as it takes approximately 7 days for the head/ganglia to regenerate versus the given 2-3 days. Although similar findings have been reported repeatedly in the literature, this is the first positive evidence with automated procedures and DNN classification. The result implies the presence of a decentralized nervous structure outside of its head/ganglia that allows a tail half to retain memory and execute motion accordingly, despite their cephalization.

animal behavior and cognition↗