Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Understanding the Neuro-Cognitive Mechanisms of Orthographic Learning in Humans and Baboons: A Comparative Study Using Domain-Specific Mechanistic and Domain-General Connectionist Models

Janos Pauli¹, Benjamin Gagl¹; ¹Universität Köln

Presenter: Janos Pauli

Learning to read is essential for social participation. Here, we investigate how humans and baboons learn orthographic information. We use a neuro-cognitive mechanistic model—the Speechless Reader (SLR) and two connectionist models (CORnet-Z and ResNet-18) to investigate a human and a baboon dataset. The connectionist models employ neuronally plausible CNN architectures, while the SLR provides transparent implementations of orthographic decision behavior using pixel, letter, and letter sequence level prediction errors as representations. To align models and data, we train the models using identical trial sequences for each human and baboon. The SLR outperforms the CNNs across both species, especially on trial-wise metrics. While CNN responses diverge from individual behavioral patterns, the SLR's interpretable errors reveal that the complexity of orthographic representations increases with training. This finding suggests that domain-specific mechanistic models offer valuable insight into learned visual behavior across species.

Topic Area: Language & Communication

Extended Abstract: Full Text PDF