Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Semantic Integration After Fast Mapping: An Arena-Based RSA Approach

Poster Session F - Tuesday, March 10, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Patric Meyer^1,2,3 (), Florian Haaf¹, Ann-Kathrin Zaiser¹, Emma Delhaye⁴, Gabriel Besson⁵; ¹School of Psychology, SRH University of Applied Sciences Heidelberg, Heidelberg, Germany, ²Department for General and Applied Linguistics, Ruprecht-Karls-University Heidelberg, Heidelberg, Germany, ³Network Aging Research (NAR), Ruprecht-Karls-University Heidelberg, Heidelberg, Germany, ⁴CICPSI, Faculty of Psychology, University of Lisbon, Portugal, ⁵CINEICC, University of Coimbra, Portugal

Fast mapping (FM) enables incidental conceptual learning and may support rapid cortical integration detectable on implicit measures even when explicit item memory is weak. We introduce an arena-based representational similarity analysis (RSA) paradigm to read out the granularity of the visuo-semantic integration of labels immediately after FM. Our structure-sensitive readout indexes visuo-semantic embedding continuously, without requiring categorical decisions or overt episodic retrieval. Participants incidentally learned pseudoword–object pairs with high (FMHO) or low (FMLO) feature overlap through FM, then arranged newly-learnt labels among known visual objects displayed according to their visuo-semantic similarity in an arena, yielding distance-based behavioral representational dissimilarity matrices (RDMs). RSA related these distance RDMs to three RDMs increasingly modeling visuo-semantic granularity: domain (animal/plant), category (e.g., birds), and matched-referent. We hypothesized that FMHO would produce finer-grained arrangements—higher RSA alignment with finer granularity—than FMLO. Participants who anticipated a memory test and items endorsed as recollected were excluded (final Ns: FMHO=12; FMLO=18). No omnibus FMHO–FMLO differences emerged. Critically, our approach revealed integration missed by explicit tests: both groups significantly placed FM-trained labels within the correct domain and category, whereas 3-alternative item recognition was at chance. RSA showed robust domain-level correlations in both groups, but category-level correlations reached significance only in FMHO. Importantly, the matched-referent model showed an emerging signal in FMHO, absent in FMLO, consistent with the expectation of greater similarity after high overlap encoding. Thus, arena-based RSA yields a sensitive readout of early, coarse→intermediate visuo-semantic integration after FM, with incipient nearest-neighbor convergence when overlap is high.

Topic Area: LONG-TERM MEMORY: Semantic