Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Replicability of semantic network analysis for younger and older adults
Poster Session A, Friday, September 12, 11:00 am - 12:30 pm, Field House
Abigail Cosgrove1, Megan Nakamura1, Milan Dukes1, Larisa Bryan1, Michele Diaz1; 1The Pennsylvania State University
Older adults have a continually growing vocabulary (Burke and Peters, 1986; Park et al., 2002; Salthouse, 2019) and experience more contexts and environments for semantic concepts (Federmeier and Kutas, 2005; Hoffman, 2018). However, with this greater semantic richness comes a larger set of words to choose from which can influence age-related processing deficits (Hoffman, 2019; Hills et al., 2013). Network science provides an objective and analytic approach to study semantic memory (Wulff et al., 2019; Siew et al, 2019). Semantic memory network analyses have the capability to measure the interaction between the information stored in semantic memory and the ability to navigate and retrieve relevant information (Kenett et al., 2025). Many semantic tasks can be utilized to construct networks including semantic relatedness judgements (Cosgrove et al., 2023), free association responses (Dubossarsky et al., 2017), and verbal fluency data (Cosgrove et al., 2021). Across these tasks, there have been relatively consistent age effects where older adults demonstrate less efficient, less connected, and more modular semantic memory networks (Cosgrove et al., 2021, 2023; Dubossarsky et al., 2017; Wulff et al., 2019). While several studies have shown consistent findings regarding age-related differences in semantic memory networks, there remain several ongoing questions that will push the aging mental lexicon field forward. One of the next steps in semantic network research is to determine the replicability of results. This methodological question remains especially pertinent for aging populations where differences in environment, life experiences, and cognitive abilities are more variable across the cohort (Wulff et al., 2019). In this current study we have several datasets of verbal fluency responses to multiple semantic categories including Animals, Supermarket, Courthouse, and Fruits/Vegetables. Each verbal fluency dataset contains a balanced younger and older adult sample who are cognitively healthy, English-speaking monolingual adults. Replicating the same pipeline as previous work (see Cosgrove et al., 2021), our findings show consistent results across most datasets. Specifically, both Animals and Supermarket categories revealed significant differences in network properties between older and younger adults, with younger adults exhibiting higher clustering, shorter average path lengths, and less modular network structures. However, there were some differences between categories across datasets, which may be attributed to variations in data collection modality or the type of semantic information selected. Overall, these findings suggest, that with increased age, semantic memory becomes less efficient, less organized, and more sparsely connected (consistent with prior work Cosgrove et al., 2021, 2023; Dubossarsky et al., 2017; Wulff et al., 2019, 2022). Moreover, through percolation analyses, which determine semantic network stability under simulated targeted attacks, our results reliably showed that older adults semantic memory networks break down faster than younger adults. These analyses shed light on the replicability of network science approaches to study the aging mental lexicon. Critically, this objective analytic approach could consistently capture a snapshot of a cognitive system as adaptable as semantic memory.
Topic Areas: Meaning: Lexical Semantics, Computational Approaches