Conscious perception of natural images is constrained by category-related visual features

Lindh, Daniel; Sligte, Ilja G.; Assecondi, Sara; Shapiro, Kimron; Charest, Ian

doi:10.1101/509927

Cited by 7 publications

(12 citation statements)

References 57 publications

(95 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If this proves to be true, it would suggest that no matter the attentional state or task demands placed on an observer, perceptual awareness is limited by higher-level perceptual features. This prediction is somewhat supported by a recent study that used the attentional blink and also found that higher-level features best predict the magnitude of the attentional blink (Lindh et al, 2019). If instead, however, there were certain cases in which the exact same set of stimuli were used, but a different primary task made it such that earlier layers/features predicted behavior, it would suggest that the limits of perceptual awareness are flexibly dictated by the task an observer is performing at any given moment.…”

Section: Remaining Questions and Limitationssupporting

confidence: 65%

“…This prediction comes from the fact that many prior studies have found that models of early vision can predict performance on a variety of visual tasks such as crowding (Balas et al, 2009;Freeman and Simoncelli, 2011), visual search (Itti & Koch, 2000;Zhang et al, 2015), scene perception (Oliva & Torralba, 2001;, and even change blindness (Rosenholtz, 2020). Alternatively, higher-level features may best predict behavior, with several prior studies showing a direct relationship between later layers of neural networks and a variety of visual behaviors such as similarity judgements (Kubilius et al, 2016;Jozwik et al, 2013;Cichy et al, 2019), object recognition (Rajalingham et al, 2018), the attentional blink (Lindh et al, 2019), and face perception (Farzmahdi et al, 2016;Jacob et al, 2021). Finally, it is also possible that a combination of both higher-and lower-level features will best predict behavior, with observers using a combination of both higher-or lower-level cues to detect alterations to the periphery depending on the stimuli.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modeling the bandwidth of perceptual experience using deep convolutional neural networks

Cohen¹,

Lydic

Murty³

2022

Preprint

View full text Add to dashboard Cite

When observers glance upon a natural scene, which aspects of that scene ultimately reach perceptual awareness? To answer this question, we showed observers images of scenes that had been altered in numerous ways in the periphery (e.g., scrambling, rotating, filtering, etc.) and measured how often these different alterations were noticed in an inattentional blindness paradigm. Then, we screened a wide range of deep convolutional neural network architectures and asked which layers and features best predict the rates at which observers noticed these alterations. We found that higher-level layers/features predicted how often observers noticed different alterations with extremely high accuracy (i.e., at the noise ceiling). This suggests that the extent to which observers noticed alterations to a scene depend on the extent to which higher-level features of a scene were preserved or changed. These findings suggest that perceptual awareness is limited by higher-level features of the human visual system.

show abstract

Section: Remaining Questions and Limitationssupporting

confidence: 65%

Section: Introductionmentioning

confidence: 99%

Modeling the bandwidth of perceptual experience using deep convolutional neural networks

Cohen¹,

Lydic

Murty³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Besides, animate objects are more often consciously perceived than inanimate objects, thus show lower level of AB effect in previous studies [8,9]. For narrower classification, more specifically, difference of some sub-classes like fruits and vegetables, processed foods, objects, scenes, animal bodies, animal faces, human bodies and human faces have been explored in [10].…”

Section: Introductionmentioning

confidence: 94%

“…First of all, inspired by [10] and [11], we propose a two-stage model for predicting ABM values with fMRI data for every particular image. In the first stage, fMRI data collected while subjects were viewing categorical images are correlated with visual image features, which are extracted by a typical convolutional neural network, Alexnet.…”

Section: Model Structure and Trainingmentioning

confidence: 99%

Categorical Difference and Related Brain Regions of the Attentional Blink Effect

Gui¹,

Ji²

2021

Preprint

View full text Add to dashboard Cite

Attentional blink (AB) is a biological effect, showing that for 200 to 500ms after paying attention to one visual target, it is difficult to notice another target that appears next, and attentional blink magnitude (ABM) is a indicating parameter to measure the degree of this effect. Researchers have shown that different categories of images can access the consciousness of human mind differently, and produce different ranges of ABM values. So in this paper, we compare two different types of images, categorized as animal and object, by predicting ABM values directly from image features extracted from convolutional neural network (CNN), and indirectly from functional magnetic resonance imaging (fMRI) data. First, for two sets of images, we separately extract their average features from layers of Alexnet, a classic model of CNN, then input the features into a trained linear regression model to predict ABM values, and we find higher-level instead of lower-level image features determine the categorical difference in AB effect, and midlevel image features predict ABM values more correctly than low-level and high-level image features. Then we employ fMRI data from different brain regions collected when the subjects viewed 50 test images to predict ABM values, and conclude that brain regions covering relatively broader areas, like LVC, HVC and VC, perform better than other smaller brain regions, which means AB effect is more related to synthetic impact of several visual brain regions than only one particular visual regions.

show abstract

“…One of the major categorical distinctions between objects is animacy. In vision, animate objects offer substantial processing and perceptual advantages over inanimate objects, including being categorized faster, more consciously perceived, and found faster in search tasks (New et al, 2007;Jackson and Calvillo, 2013;Carlson et al, 2014;Ritchie et al, 2015;Lindh et al, 2019). Auditory studies have similarly found faster categorization times for animate objects (Yuval-Greenberg and Deouell, 2009;Vogler and Titchener, 2011).…”

Section: Introductionmentioning

confidence: 99%

Selective Enhancement of Object Representations through Multisensory Integration

2020

View full text Add to dashboard Cite

Objects are the fundamental building blocks of how we create a representation of the external world. One major distinction among objects is between those that are animate versus those that are inanimate. In addition, many objects are specified by more than a single sense, yet the nature by which multisensory objects are represented by the brain remains poorly understood. Using representational similarity analysis of male and female human EEG signals, we show enhanced encoding of audiovisual objects when compared with their corresponding visual and auditory objects. Surprisingly, we discovered that the often-found processing advantages for animate objects were not evident under multisensory conditions. This was due to a greater neural enhancement of inanimate objects-which are more weakly encoded under unisensory conditions. Further analysis showed that the selective enhancement of inanimate audiovisual objects corresponded with an increase in shared representations across brain areas, suggesting that the enhancement was mediated by multisensory integration. Moreover, a distance-to-bound analysis provided critical links between neural findings and behavior. Improvements in neural decoding at the individual exemplar level for audiovisual inanimate objects predicted reaction time differences between multisensory and unisensory presentations during a Go/No-Go animate categorization task. Links between neural activity and behavioral measures were most evident at intervals of 100-200 ms and 350-500 ms after stimulus presentation, corresponding to time periods associated with sensory evidence accumulation and decision-making, respectively. Collectively, these findings provide key insights into a fundamental process the brain uses to maximize the information it captures across sensory systems to perform object recognition.

show abstract

Conscious perception of natural images is constrained by category-related visual features

Cited by 7 publications

References 57 publications

Modeling the bandwidth of perceptual experience using deep convolutional neural networks

Modeling the bandwidth of perceptual experience using deep convolutional neural networks

Categorical Difference and Related Brain Regions of the Attentional Blink Effect

Selective Enhancement of Object Representations through Multisensory Integration

Contact Info

Product

Resources

About