Forward and backward masking of naturalistic texture
Poster Presentation: Tuesday, May 20, 2025, 8:30 am – 12:30 pm, Pavilion
Session: Temporal Processing: Neural mechanisms, models
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Laura Palmieri1 (), Timothy D. Oleskiw1,2, J. Anthony Movshon1; 1New York University, 2University of Regina
The detection and discrimination of spatial stimuli is affected by overlaid masks presented before or after the target stimulus. When the masking involves contrast detection or simple discrimination, masking is thought to depend on contrast signals like those in primary visual cortex. We have now extended this approach to signals that rely on downstream processing. For this we used the discrimination of naturalistic texture statistics, which is known from neuronal recordings in macaque to depend on signals in areas downstream from V1, notably V2 and V4. Subjects discriminated “naturalistic” textures created from the Portilla-Simoncelli model from phase-randomized “noise” textures. We interpolated the model parameters between matched noise and naturalistic textures to vary “texture coherence” – 0% for noise textures and 100% for textures with all the P-S statistics. We used a 3-choice oddity task to measure performance. Subjects discriminated textures of variable coherence, set using a staircase procedure, to define coherence threshold for brief (40 ms) targets. These were presented with or without a zero-coherence 100 ms mask that either preceded or followed the target with a gap between 0 and 50 ms. When the target followed the mask (“forward masking”), coherence thresholds were modestly elevated (1.2–2x); masking was strongest for the 0-gap condition. When the target preceded the mask (“backward masking”), threshold elevation was greater (2.5-4x), and was maximal for a gap of 25 ms. Performance was similar for a simultaneous 3-choice task using near-peripheral viewing, and a sequential 3-choice task using central viewing. Our results can be accounted for using a delayed normalization model to describe neural response dynamics, combined with temporal crowding to capture backward masking. They are also compatible with data we have obtained using a similar paradigm from single- and multi-unit neuronal recordings from area V4 of awake, fixating macaque monkeys.