Biology subjects

Pelli, D. G.

Publications and source records attributed to Pelli, D. G..

6 recordsLinked to original sources

bioRxiv2026

Object recognition along the ventral stream: effects of visual noise

How does the visual system recognize objects in natural environments? Here, we investigate the neural computations underlying robust object recognition along the ventral stream. We adapted psychophysical critical-band masking to fMRI, measuring BOLD responses along the ventral stream to natural images perturbed by bandpass noise. We found that, along this stream, the overall BOLD response to noise alone is sensitive to an increasingly wide range of stimulus spatial frequency, from 2 octaves in V1 to 5 octaves in ventral temporal cortex (VTC). However, when we assess the effect of the same noise on the accuracy of decoding scene images from the BOLD response, we obtain a different result: Recognition bandwidth is conserved along the ventral stream at about 2 octaves, close to the 1.5-octave behavioral channel. Though the recognition band is conserved, its noise tolerance increases steadily along the ventral stream, approaching behavioral levels in VTC. These findings suggest that V1 sets the bandwidth of the object-recognition channel, while downstream areas progressively denoise the signal, establishing the channels noise tolerance. This biological architecture--early channel isolation followed by successive denoising--may explain why human vision remains more robust than current machine vision.

neuroscience↗

bioRxiv2025

EasyEyes: Crowded Dynamic Fixation for Online Psychophysics

Online vision testing enables efficient data collection from diverse participants but often requires accurate fixation. When needed, fixation accuracy is traditionally ensured by using a camera to track gaze. That works well in the lab, but tracking during online testing with a built-in webcam is not yet sufficiently precise. Kurzawski, Pombo, et al. (2023) introduced a fixation task that improves fixation through hand-eye coordination, requiring participants to track a moving crosshair with a mouse-controlled cursor. This dynamic fixation task greatly reduces peeking at peripheral targets relative to a stationary fixation task, but does not eliminate it. Here, we introduce a crowded dynamic fixation task that further enhances fixation by adding clutter around the fixation mark to leverage crowding. We assessed fixation accuracy during peripheral threshold measurement. Relative to the RMS gaze error during the stationary fixation task, dynamic fixation error was 61%, while crowded dynamic fixation error was only 47%. With a 1.5{degrees} tolerance, peeking occurred on 9% of trials with stationary fixation, 4% with dynamic fixation, and 0% with crowded dynamic fixation. This improvement eliminated implausibly low peripheral thresholds, likely by preventing peeking. We conclude that crowded dynamic fixation provides accurate gaze control for online testing.

neuroscience↗

bioRxiv2021

Informational masking vs. crowding - A mid-level trade-off between auditory and visual processing

In noisy or cluttered environments, sensory cortical mechanisms help combine auditory or visual features into perceived objects. Knowing that individuals vary greatly in their ability to suppress unwanted sensory information, and knowing that the sizes of auditory and visual cortical regions are correlated, we wondered whether there might be a corresponding relation between an individuals ability to suppress auditory vs. visual interference. In auditory masking, background sound makes spoken words unrecognizable. When masking arises due to interference at central auditory processing stages, beyond the cochlea, it is called informational masking (IM). A strikingly similar phenomenon in vision, called visual crowding, occurs when nearby clutter makes a target object unrecognizable, despite being resolved at the retina. We here compare susceptibilities to auditory IM and visual crowding in the same participants. Surprisingly, across participants, we find a negative correlation (R = -0.7) between IM susceptibility and crowding susceptibility: Participants who have low susceptibility to IM tend to have high susceptibility to crowding, and vice versa. This reveals a mid-level trade-off between auditory and visual processing.

neuroscience↗

bioRxiv2021

Cross-dataset reproducibility of population receptive field (pRF) estimates and cortical magnification asymmetries

Population receptive field (pRF) models fit to fMRI data are used to non-invasively measure retinotopic maps in human visual cortex, and these maps are a fundamental component of visual neuroscience experiments. Here, we examined the reproducibility of retinotopic maps across two datasets: a newly acquired retinotopy dataset from New York University (NYU) (n=44) and a public dataset from the Human Connectome Project (HCP) (n=181). Our goal was to assess the degree to which pRF properties are similar across datasets, despite substantial differences in their experimental protocols. The two datasets simultaneously differ in their stimulus apertures, participant pool, fMRI protocol, MRI field strength, and preprocessing pipeline. We assessed the cross-dataset reproducibility of the two datasets in terms of the similarity of vertex-wise pRF estimates and in terms of large-scale polar angle asymmetries in cortical magnification. Within V1, V2, V3, and hV4, the group-median NYU and HCP vertex-wise polar angle estimates were nearly identical. Both eccentricity and pRF size estimates were also strongly correlated between the two datasets, but with a slope different from 1; the eccentricity and pRF size estimates were systematically greater in the NYU data. Next, to compare large-scale map properties, we quantified two polar angle asymmetries in V1 cortical magnification previously identified in the HCP data. The NYU dataset confirms earlier reports that more cortical surface area represents horizontal than vertical visual field meridian, and lower than upper vertical visual field meridian. Together, our findings show that the retinotopic properties of V1, V2, V3, and hV4 can be reliably measured across two datasets, despite numerous differences in their experimental design. fMRI-derived retinotopic maps are reproducible because they rely on an explicit computational model of the fMRI response. In the case of pRF mapping, the model is grounded in physiological evidence of how visual receptive fields are organized, allowing one to quantitatively characterize the BOLD signal in terms of stimulus properties (i.e., location and size). The new NYU Retinotopy Dataset will serve as a useful benchmark for testing hypotheses about the organization of visual areas and for comparison to the HCP 7T Retinotopy Dataset.

neuroscience↗

bioRxiv2021

An enhanced Bouma model fits a hundred people's visual crowding

AbstractCrowding is the failure to recognize an object due to surrounding clutter. Our visual crowding survey measured 13 crowding distances (or "critical spacings") twice in each of 50 observers. The survey included three eccentricities (0, 5, and 10 deg), four cardinal meridians, two orientations (radial and tangential), and two fonts (Sloan and Pelli). The survey also tested foveal acuity, twice. Remarkably, fitting a two-parameter model, the well- known Bouma law -- crowding distance grows linearly with eccentricity -- explains 82% of the variance for all 13 x 50 measured log crowding distances, cross-validated. An enhanced Bouma law, with factors for meridian, crowding orientation, target kind, and observer, explains 94% of the variance, again cross-validated. These additional factors reveal several asymmetries, consistent with previous reports, which can be expressed as crowding- distance ratios: 0.62 horizontal:vertical, 0.79 lower:upper, 0.78 right:left, 0.55 tangential:radial, and 0.78 Sloan font:Pelli font. Across our observers, peripheral crowding is independent of foveal crowding and acuity. Evaluation of the Bouma factor b (the slope of the Bouma law) as a biomarker of visual health would be easier if there were a way to compare results across crowding studies that use different methods. We define a standardized Bouma factor b that corrects for differences from Boumas 25 choice alternatives, 75% threshold criterion, and linearly symmetric flanker placement. For radial crowding on the right meridian, the standardized Bouma factor b is 0.24 for this study, 0.35 for Bouma (1970), and 0.30 for the geometric mean across five representative modern studies, including this one, showing good agreement across labs, including Boumas. We found that guaranteeing fixation by gaze-contingent display halved the standard deviation across observers of the estimated log b. The reduction in standard deviation is explained by a "peeking" model in which the observer looked near an anticipated target location in 50% of unmonitored-fixation trials. Individual differences are robust, as evidenced by the much larger 0.08 SD of log b across observers than the 0.03 SD of test-retest within observers. Crowdings ease of measurement enhances its promise as a biomarker for dyslexia and visual health.

neuroscience↗

bioRxiv2017

Deep learning: Using machine learning to study biological vision

Today many vision-science presentations employ machine learning and especially \"deep learning\", one of the more recent and successful variants. Many neuroscientists use machine learning to decode neural responses. Many perception scientists try to understand how living organisms recognize objects. To them, deep neural networks offer several benchmark accuracies for recognition of learned stimuli. Originally machine learning was inspired by the brain. Today, machine learning is used as a statistical tool to decode brain activity. Tomorrow, deep neural networks might become our best model of brain function. This brief overview of the use of machine learning in biological vision touches on its strengths, weaknesses, milestones, controversies, and current directions. Here, we hope to help vision scientists assess what role machine learning should play in their research.

neuroscience↗