The BiLex computational model of bilingual lexical access was extended with lateral excitation-inhibition dynamics to simulate naming in English and Spanish. English and Spanish phonetic self-organizing maps (SOMs), linked to a shared semantic map via bidirectional connections, used short-range excitation and long-range inhibition to focus activation over settling steps, providing a novel response time measure. Further, the BiLex model was trained for semantic effects such as familiarity, typicality, and specificity. The model achieved 73.6 percent accuracy in English and 70.5 percent in Spanish. Word frequency correlated positively with accuracy and negatively with response time in both languages (all p less than 0.001). A positive typicality-frequency correlation (p = .003) suggests atypicality contributes to naming difficulty. These findings validate BiLex as a biologically plausible bilingual naming model, with future work targeting lexical-semantic impairments and treatments.

PDF

In Proceedings of the 35th Annual Computational Neuroscience Meeting (CNS*2026), September 2026.

@inproceedings{prabhakaran:cns2026, title={Simulating bilingual naming in laterally-connected self-organizing maps}, author={Divya Prabhakaran and Uli Grasemann and Swathi Kiran and Risto Miikkulainen}, booktitle={Proceedings of the 35th Annual Computational Neuroscience Meeting (CNS*2026)}, month={September}, url="http://www.cs.utexas.edu/users/ai-lab?prabhakaran:cns2026", year={2026} }

Poster