Ensemble feature distribution representations support visual stability
Abstract
Visual stability refers to people’s subjective experience of a stable visual environment despite incomplete or disruptive retinal input, e.g., due to blinks and frequent saccadic eye movements. Known stability mechanisms involve updating individual objects’ positions before and after saccades. This is infeasible in real-world scenes, which contain dozens of important items. Here we present evidence supporting an Ensemble Feature Distribution hypothesis: that visual stability relies strongly on ensemble feature representations of groups of objects. In three preregistered experiments (N=51, university students), observers were sequentially shown two sets of colored disks and reported the number of disks that changed their colors. We found that if the ensemble color distribution was altered between the two displays, observers easily noticed changes, even when they occurred during saccades. Yet observers perceived the environment as stable and almost completely missed equivalent changes when the ensemble feature distribution stayed the same but individual colors swapped positions.
Related
- StackHPpred: A Stacking-based ensemble learning framework for the identification of peptide hormones using multi-view feature representations
- Ensemble-based support vector machine classifiers as an efficient tool for quality assessment of beef fillets from electronic nose data
- DAF-EfficientNetV2: distribution-aware feature enhancement for plant leaf disease recognition in complex environments
- Predicting the Potential Suitable Distribution of Tomato Brown Rugose Fruit Virus in China Using a BIOMOD2 Ensemble Model.
- Medial septum glutamatergic neurons support speed coding and grid cell spatial stability in the medial entorhinal cortex
BibTeX
@misc{khvostov2026ensemble,
title = {{Ensemble feature distribution representations support visual stability}},
author = {Vladislav Khvostov and Julie Golomb},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/3qv95_v1},
url = {https://osf.io/preprints/psyarxiv/3qv95_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Ensemble feature distribution representations support visual stability
AU - Vladislav Khvostov
AU - Julie Golomb
PY - 2026
DA - 2026/09/03
PB - PsyArXiv
DO - 10.31234/osf.io/3qv95_v1
UR - https://osf.io/preprints/psyarxiv/3qv95_v1/
AB - Visual stability refers to people’s subjective experience of a stable visual environment despite incomplete or disruptive retinal input, e.g., due to blinks and frequent saccadic eye movements. Known stability mechanisms involve updating individual objects’ positions before and after saccades. This is infeasible in real-world scenes, which contain dozens of important items. Here we present evidence supporting an Ensemble Feature Distribution hypothesis: that visual stability relies strongly on ensemble feature representations of groups of objects. In three preregistered experiments (N=51, university students), observers were sequentially shown two sets of colored disks and reported the number of disks that changed their colors. We found that if the ensemble color distribution was altered between the two displays, observers easily noticed changes, even when they occurred during saccades. Yet observers perceived the environment as stable and almost completely missed equivalent changes when the ensemble feature distribution stayed the same but individual colors swapped positions.
ER -