Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The Influence of Knowledge on Perception: Model-Based Study of How Lexical Knowledge Optimizes Visual Word Representations
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
Benjamin Gagl1; 1University of Cologne
When we read, our brain extracts meaning from words, but the extent to which our lexical knowledge aids this process is unclear. Here, we use a transparent computational model that simulates human orthographic decision behavior to examine how word knowledge influences orthographic processing. The model incorporates pixel, letter, and letter-sequence representations based on the neuronally plausible predictive coding assumption, offering a framework for investigating the integration of sensory input with knowledge (i.e., as a top-down prediction). Here, we test different lexicon structures from which the model derives the top-down predictions. For model evaluation, we use a pseudoword learning dataset that allowed us to know which items participants learned. We compare three lexicon assumptions: (i) including only learned items, (ii) all words that participants should have learned, and (iii) a lexicon only including the foils. We analyzed response times and errors and found the highest model fit for the representations that integrated the lexicon with only the learned words, including all three representations. Thus, with this model-based approach, we find substantial support for a direct link between lexical knowledge and downstream orthographic processing, indicating that representations of visual words are optimized by our lexical knowledge to enable efficient reading.
Topic Areas: Computational Approaches, Reading