Double dissociation without modularity: evidence from connectionist neuropsychology

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psycholinguistics

Background:

• Dominant theories propose encapsulated modules for specific cognitive functions.
• Brain damage studies, particularly double dissociations, are often interpreted as evidence for modularity.
• This view suggests selective impairments result from the loss of distinct processing components.

Purpose of the Study:

• To investigate a double dissociation between concrete and abstract word reading in a connectionist network.
• To challenge the assumption that functional specialization necessitates separable processing components.
• To examine the implications of connectionist models for interpreting neuropsychological data.

Main Methods:

• Detailed examination of a connectionist network model that pronounces words via meaning without separable components.
• Analysis of a double dissociation between concrete and abstract word reading following simulated damage.
• Consideration of lesion distribution effects in quantitatively equivalent individual lesions.

Main Results:

• A double dissociation between concrete and abstract word reading was observed in the connectionist network.
• The functional specialization underlying this dissociation was not transparently linked to the network's structure.
• Lesion distribution analysis raised concerns regarding the interpretation of single-case neuropsychological studies.

Conclusions:

• Findings challenge modular theories by showing functional specialization without separable components.
• Neuropsychological data interpretation requires explicit computational assumptions about cognitive systems.
• Emphasizes the need to integrate computational modeling with cognitive theories for a comprehensive understanding.