Spreading activation in an attractor network with latching dynamics: automatic semantic priming revisited.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psycholinguistics

Background:

• Semantic priming is crucial for understanding word recognition and semantic memory.
• Existing models, like localist spreading activation (SA) and distributed representation models, offer differing explanations.
• Reconciling these approaches is key to advancing the field.

Purpose of the Study:

• To develop a unified computational framework for semantic priming.
• To implement SA within an attractor neural network using distributed representations.
• To explain major characteristics of automatic semantic priming.

Main Methods:

• Developed an attractor neural network model incorporating synaptic depression and distributed representations.
• Simulated the model's dynamics to analyze semantic priming phenomena.
• Investigated the model's ability to account for mediated, asymmetric, and backward priming.

Main Results:

• The unified model successfully replicates major characteristics of automatic semantic priming.
• Mediated and asymmetric priming effects are natural outcomes of the model's 'latching dynamics'.
• Backward priming effects and differences between semantic and associative relatedness are explained.

Conclusions:

• The proposed attractor neural network model provides a unified framework for SA and distributed representation approaches.
• The model's synaptic depression mechanism offers a robust explanation for various semantic priming phenomena.
• This work advances computational models of semantic memory and word recognition.