Enhanced generalization and specialization of brain representations of semantic knowledge in healthy aging

Pedro Margolles,David Soto

DOI: https://doi.org/10.1101/2023.08.08.552519

2024-09-09

Abstract:Aging is often associated with a decrease in cognitive capacities. However, semantic memory appears relatively well preserved in healthy aging. Both behavioral and neuroimaging studies support the view that changes in brain networks contribute to this preservation of semantic cognition. However, little is known about the role of healthy aging in the brain representation of semantic categories. Here we used pattern classification analyses and computational models to examine the neural representations of living and non-living word concepts. The results demonstrate that brain representations of animacy in healthy aging exhibit increased similarity across categories, even across different task contexts. This pattern of results aligns with the neural dedifferentiation hypothesis that proposes that aging is associated with decreased specificity in brain activity patterns and less efficient neural resource allocation. However, the loss in neural specificity for different categories was accompanied by increased dissimilarity of item-based conceptual representations within each category. Taken together, the age-related patterns of increased generalization and specialization in the brain representations of semantic knowledge may reflect a compensatory mechanism that enables a more efficient coding scheme characterized by both compression and sparsity, thereby helping to optimize the limited neural resources and maintain semantic processing in the healthy aging brain.

Neuroscience

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

This paper aims to explore how healthy aging affects the brain's representation of semantic knowledge. Specifically, the study examines the neural representation of biological and non-biological concepts in healthy older adults through pattern classification analysis and computational models.

#### Main Objectives
1. **Changes in Semantic Category Distinction**: The study investigates how the distinction of semantic categories in brain activity patterns within the bilateral semantic network changes with age using multivariate pattern analysis (MVPA) decoding methods. Researchers use machine learning classifiers to predict the semantic categories of previously unseen biological and non-biological words and conduct decoding analysis under different processing depth conditions.

2. **Impact of Deep Processing**: The study also explores the differences between deep processing conditions (e.g., imagining specific attributes of concepts) and shallow processing conditions (e.g., silently reading words). By comparing decoding accuracy under the two conditions, researchers aim to understand how task demands affect the brain's semantic representation in older adults.

3. **Neural Mechanisms of Semantic Representation**: Through searchlight representational similarity analysis (searchlight RSA), researchers compare the neural representations of participants of different ages with language models based on word embeddings and computer vision models based on image processing to identify similarities and differences.

#### Research Hypotheses
- Under both shallow and deep processing conditions, older adults will have lower brain semantic category decoding accuracy compared to younger adults, possibly due to reduced neural differentiation of semantic representations.
- In cross-condition analysis, if aging affects the generalization of brain representations across different tasks, older adults may perform better than younger adults.
- Older adults may allocate more cognitive resources and sensory-motor capabilities under deep processing conditions, potentially leading to decreased decoding accuracy in the left hemisphere and increased accuracy in the right hemisphere.

Through these analyses, researchers hope to reveal changes in the representation of semantic knowledge in the context of healthy aging and its potential compensatory mechanisms.