Feature Binding without Attentional Glue: An In Silico Lesion Study on Asymmetric Attribute Amnesia in Bimodal Systems

Read full text Summarise with VIRAT
Authors: Zilin Li, Tianle Jiang
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Feature Integration Theory (FIT) posits that focused attention acts as the "glue" binding distinct physical attributes into coherent object representations. However, behavioral paradigms investigating Attribute Amnesia (AA) reveal a puzzling asymmetric memory decay: when attentional resources are exhausted or redirected, the conjunctive object representation collapses, yet individual constituent features (e.g., race, gender) often remain accessible. The exact representational geometry underlying this asymmetric unbinding remains opaque due to the causal limitations of in vivo neuroimaging. Here, we adopt an in silico neuropsychological approach, utilizing a bimodal neural network as a functional proxy for the human visual-semantic system. By applying targeted digital lesions to the attentional control hubs governing cross-modal integration, we simulate the failure of the hypothesized "attentional glue." Representational Similarity Analysis (RSA) and manifold dimensionality measurements (Deff) reveal a distinct causal sequence: progressive attentional ablation induces a non-linear structural collapse of the high-dimensional conjunctive space, while lower-dimensional independent feature manifolds maintain structural integrity. This geometric decoupling provides formal computational evidence for the "floating features" hypothesis, suggesting that attention operates not as a structural container for features, but as a dynamic gain controller that sustains temporary non-linear topological bindings.

Fields of Study & Subjects

Visual perception and processing mechanisms Multisensory perception and integration Visual Attention and Saliency Detection

Publication details

Journal / source
PsyArXiv (OSF Preprints)
Publication date
22 March 2026
Language
EN
Document type
Preprint