Schedule of Events | Symposia

Sketchpad Series

Investigating implicitly formed mental representations of abstract knowledge

Poster Session B - Sunday, March 30, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Irina Barnaveli¹ (barnaveli@cbs.mpg.de), Simone Viganò¹, Patrick Haggard², Christian F. Doeller^1,3,4,5; ¹Max Planck Institute for Human Cognitive and Brain Sciences, ²Institute of Cognitive Neuroscience, University College London, UK, ³Kavli Institute for Systems Neuroscience, NTNU, Trondheim, Norway, ⁴Wilhelm Wundt Institute of Psychology, Leipzig, Germany, ⁵Department of Psychology, Technische Universität Dresden, Dresden, Germany

Understanding how humans relate abstract information is a key challenge in neuroscience. One influential framework suggests that humans create low-dimensional map-like Euclidean models for comparing and selecting among multiple items. It is unclear whether these representations emerge naturally, or result from biases in experimental designs that typically teach participants to associate items with experimentally relevant arbitrary dimensions. In contrast, in natural environments, most items consist of multi-modal associations, requiring a circumstantial learning and updating. Humans may instead systematically tend to categorize these associations into non-dimensional, hierarchically organized clusters, potentially facilitating the search for relevant knowledge. To explore this, we developed a novel paradigm where participants are taught the categorical and dimensional features of different monster stimuli. After training, participants perform similarity judgments across triplets of monsters. Next, participants complete two tasks that modulate the relevance of either categorical or dimensional features, again combined with triplet similarity judgments. With this paradigm, we aim to identify how humans prioritize dimensional vs. non-dimensional mental representations by exploring how participants implicitly structure abstract knowledge and adapt this structure based on task demands. We will apply representational similarity analysis to behavioral and eye-tracking data to investigate how the monster stimuli are implicitly organized. We expect participants to form either (i) dimensional representations, (ii) categorical hierarchical clusters, or (iii) a weighted combination of the two. To understand the fine-grained details of these representations, we will further reconstruct them from triplet similarity scores using multidimensional scaling and stochastic triplet embedding algorithms.

Topic Area: LONG-TERM MEMORY: Other