Related Experiment Videos
The connectionist simulation of aphasic naming
1Department of Mathematical and Computing Sciences, University of Surrey, Guildford, United Kingdom. J.Wright@surrey.ac.uk
Brain and Language
|September 23, 1997
Summary
This study simulates language disorders like aphasic naming using connectionist networks. Lesioning connections in a modular network accurately reproduced a patient's naming errors.
Area of Science:
- Computational neuroscience
- Psycholinguistics
- Artificial intelligence
Background:
- Interactive activation (IA) networks and connectionist systems are used to simulate language disorders.
- Existing IA models of aphasic naming involve varying network parameters like decay rate and connection strength to match patient errors.
Observation:
- Modifying additional parameters, such as the shared weight increase factor, can also yield similar results in IA models.
- This suggests the possibility of simulating aphasic naming without explicit parameter variation.
Findings:
- A modular connectionist architecture is proposed, utilizing self-organizing Kohonen maps for semantic-lexical and phonological knowledge.
- Hebbian networks implement connections between these maps, and a linear connectionist network (Madaline) simulates nonword repetition.
- Lesioning Hebbian connections effectively reproduced the observed naming errors in the aphasic patient.
Implications:
- This modular approach offers a more biologically plausible method for simulating language disorders.
- It provides a framework for understanding the neural basis of naming deficits in aphasia.
- The model can be further developed to explore other aspects of language processing and disorders.