A Roadmap for Understanding Memory: Decomposing Cognitive Processes into Operations and Representations.

Area of Science:

• Neuroscience
• Cognitive Psychology
• Computational Neuroscience

Background:

• Historically, frontal lobes were linked to behavioral control and medial temporal lobes (MTL) to declarative memory.
• Cognitive functions have evolved as labels for brain networks, encompassing broad functions and specific mental processes like recollection.
• Current approaches label brain networks, moving beyond focal regions.

Purpose of the Study:

• To evaluate if "process" labels, like recollection, are the most effective for dissecting memory.
• To determine if current functional labels adequately capture neural mechanisms of memory.
• To propose a more productive framework for understanding memory organization.

Main Methods:

• Conceptual analysis of memory labels and their neurocomputational underpinnings.
• Examining the conflation of operations and representational content within process labels like recollection.
• Reviewing evidence for pattern completion across the ventral visual stream and MTL.

Main Results:

• Process labels like recollection inadequately characterize neural mechanisms by conflating distinct components.
• Recollection combines a neurocomputational operation (pattern completion) with representational content (associative memories).
• Pattern completion is widely observed in the ventral visual stream and MTL, with functional divisions explained by representational content.

Conclusions:

• A comprehensive theory of memory necessitates separating and specifying the interaction of operations and representations.
• Dissecting memory by isolating component operations and representations offers a more productive approach.
• The organization of memory relies on pattern completion, with representational content explaining site-specific divisions of labor.