Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Individual Differences in Semantic Network Structure: The Role of Cumulative Experience and Implications for Cognitive Performance
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Julieta Laurino1, Dirk U. Wulff1,2; 1Max Planck Institute for Human Development, Berlin, Germany, 2University of Basel, Basel, Switzerland
Semantic representations are a critical component of cognition, supporting processes such as memory, reasoning, and language. These representations are acquired through cumulative experience, including linguistic input and everyday interactions. Because such experiences vary across individuals, semantic representations are also likely to differ between individuals. Yet most current approaches estimate semantic representations using aggregated data (e.g., word embeddings derived from large text corpora), implicitly assuming a common cognitive system and obscuring individual-level variation. As a result, we still lack a clear understanding of the extent to which semantic representations differ across individuals, how such differences relate to cumulative experience, and whether they shape cognitive performance. Recent work, with many efforts coming from the science of aging, has begun addressing this gap by estimating semantic representations at the individual level using behavioral paradigms such as free association and relatedness judgments (Dubossarsky et al., 2017; Wulff et al., 2019; Wulff et al., 2022). However, these approaches remain limited in scope, conflate representational and retrieval control processes, and provide limited assessments of cognitive performance. Addressing this gap requires large-scale, individual-level data that simultaneously capture semantic structure, cumulative experience, and cognitive performance. This study leverages the Semantic Aging Study, one of the most comprehensive datasets to date for investigating individual semantic representations. It includes free association and relatedness judgments for 100 socially relevant words across 10 categories (e.g., health, food, emotions), collected from 360 participants spanning a wide age range (20–78 years), balanced in gender and education. The dataset also contains a broad set of tasks assessing cognitive abilities across multiple domains, including memory, language, reasoning, and exploratory behavior. Individual-level semantic representations were estimated by constructing semantic networks separately for each participant from both elicitation tasks (i.e., free associations and relatedness judgments). Each network was characterized by three macroscopic measures: connectivity (average degree), structuredness (clustering coefficient), and efficiency (shortest path length); measures previously linked to word-level performance on tasks such as verbal fluency (Cosgrove et al., 2023). First, we assessed the heterogeneity of network structures by examining how metrics varied across individuals, evaluating their convergence across the two elicitation tasks, and comparing structures across semantic categories. Second, we examined whether network structure was predicted by cumulative experience, indexed by age, gender, and education, as these factors are associated with different linguistic and non-linguistic exposure. Third, we tested whether individual differences in semantic network structure related to cognitive performance across multiple domains, and whether such associations were shared or domain-specific. Results showed substantial variation in semantic network structure across participants, demonstrating that individuals differ meaningfully in how they represent semantic knowledge. Partial convergence across elicitation tasks suggested that semantic network metrics capture both genuine representational differences and method-specific variance. Furthermore, individual differences in network structure were both shaped by demographic markers of cumulative experience and predictive of performance across cognitive domains. Overall, this study establishes individual semantic structure as a meaningful, yet largely overlooked, source of variation in how people learn, think, remember, and age.
Topic Areas: Meaning: Lexical Semantics, Computational Approaches