Prolegomena to a neurocomputational architecture for human grammatical encoding and decoding.

Area of Science:

• Neuroscience
• Computational Linguistics
• Cognitive Science

Background:

• Syntactic structure formation in language production and comprehension remains a complex challenge for neurocomputational models.
• Existing models often struggle to explain how complex grammars are processed within fixed neural architectures.
• Evidence suggests overlap between neural mechanisms for grammatical encoding and decoding, hinting at a unified system.

Purpose of the Study:

• To develop a neurocomputational architecture for grammatical processing in both language production (encoding) and comprehension (decoding).
• To investigate how complex syntactic structures can be formed in a fixed neural network without online connection changes.
• To determine if a single neural infrastructure can support both grammatical encoding and decoding.

Main Methods:

• Development of a neurocomputational architecture based on the "Unification Space" model.
• The architecture relies on prewired circuitry, with referential processing utilizing the hippocampal complex.
• Inclusion of a mechanism to handle grammatical movement phenomena.

Main Results:

• The proposed architecture enables online syntactic structure formation within a fixed neural network.
• The model successfully integrates grammatical encoding and decoding onto a shared neural infrastructure.
• A mechanism for processing grammatical movement phenomena is incorporated.

Conclusions:

• A unified neurocomputational architecture can account for syntactic structure formation in both language production and comprehension.
• The model demonstrates the feasibility of complex grammatical processing using prewired neural circuitry.
• This framework supports the hypothesis of a single "grammatical coder" underlying language processing.