Vertex effect in words

Orthographic processing builds upon visual object recognition, inheriting its properties. Computational proximity between these cognitive systems might explain the observation of a vertex effect in words: visual word recognition is less disrupted by contour deletion when vertices (i.e., line junctions) are preserved rather than removed (i.e., midsegments preserved). In this study we investigated the cognitive locus of this effect across four experiments. Experiments 1 and 2 manipulated contour (100, 55, and 35%), preserved segment (vertex vs. midsegment), and lexicality (word vs. pseudoword). A robust vertex effect was found, larger for words than pseudowords, in both naming (Experiment 1) and lexical decision (Experiment 2). Experiment 3 combined masked priming with a lexical decision task, showing that identity priming was unaffected by vertex information, as confirmed by Bayesian analyses. Experiment 4 used a time-limited lexical decision task and confirmed a larger vertex effect for words than pseudowords, with similar magnitude for low- and high-frequency words. Together, these results indicate that vertex information shapes later stages of orthographic processing, likely during lexical access and before lexical dynamics such as those driven by word frequency emerge. Importantly, rather than a vestigial trait inherited from object recognition, vertex information seems to serve a repurposed role during orthographic processing.