Neurocognitive networks and selectively distributed processing

M. Mesulam*

*Corresponding author for this work

Research output: Contribution to journalShort survey

67 Citations (Scopus)

Abstract

The association cortex of the human brain can be divided into unimodal and transmodal components. Unimodal (modality-specific) cortical areas are subdivided into upstream regions specialized for encoding unitary features of experience and downstream regions which are specialized for encoding composite features. Modality-specific features lead to multimodal knowledge through the mediation of transmodal areas in the brain. These transmodal areas include cortical regions that are conventionally designated as heteromodal, paralimbic and limbic cortex. Contrary to earlier formulations, it is no longer thought that these transmodal areas contain a convergent residue of knowledge. Instead, it appears that the role of these transmodal areas is to contain a road map for the multifocal binding and calling up of distributed information in multiple modalities. Knowledge can thus be encoded in a flexible distributed rather than rigid convergent form. Observations on patients with focal neurological lesions indicate that transmodal areas act like neural hubs (or gateways) for accessing critioal domains of knowledge rather than as dedicated centers for specific cognitive functions. In the processes related to memory, a limbic structure such as the hippocampus does not act as a bank for specific memories but as a critical node for accessing distributed information related to recently acquired experience. Damage to a sufficient volume of the limbic system interferes with the coherence of recall and storage even though the constituent fragments of the corresponding experiences may remain stored quite well in other parts of the brain. Additional observations based on the phenomenon of hemispatial neglect lead to the conclusion that transmodal areas and unimodal areas are interconnected with each other to form large scale neural networks that can sustain complex computational architectures including those that rely on parallel distributed processing. Networks organized in this fashion can rapidly access a vast informational landscape while simultaneously considering many goals and constraints. The final compromise into which the network settles is identified as the solution to the cognitive problem. In this neurological model of cognition, the unimodal areas of cortex provide the most veridical building blocks of experience. Transmodal nodes bind this information in a way that introduces temporal and contextual coherence. The formation of specific templates belonging to objects and memories oceurs in distributed form but with considerable regional specialization. This arrangement leads to a highly flexible and powerful computational system which could be described as Selectively Distributed Processing.

Original languageEnglish (US)
Pages (from-to)564-569
Number of pages6
JournalRevue Neurologique
Volume150
Issue number8-9
StatePublished - Jan 1 1994

Fingerprint

Cognition
Brain
Neurological Models
Perceptual Disorders
Limbic System
Hippocampus

ASJC Scopus subject areas

  • Clinical Neurology
  • Neuroscience(all)

Cite this

@article{0e09e0c1870f4f9da9b8892ae089eab2,
title = "Neurocognitive networks and selectively distributed processing",
abstract = "The association cortex of the human brain can be divided into unimodal and transmodal components. Unimodal (modality-specific) cortical areas are subdivided into upstream regions specialized for encoding unitary features of experience and downstream regions which are specialized for encoding composite features. Modality-specific features lead to multimodal knowledge through the mediation of transmodal areas in the brain. These transmodal areas include cortical regions that are conventionally designated as heteromodal, paralimbic and limbic cortex. Contrary to earlier formulations, it is no longer thought that these transmodal areas contain a convergent residue of knowledge. Instead, it appears that the role of these transmodal areas is to contain a road map for the multifocal binding and calling up of distributed information in multiple modalities. Knowledge can thus be encoded in a flexible distributed rather than rigid convergent form. Observations on patients with focal neurological lesions indicate that transmodal areas act like neural hubs (or gateways) for accessing critioal domains of knowledge rather than as dedicated centers for specific cognitive functions. In the processes related to memory, a limbic structure such as the hippocampus does not act as a bank for specific memories but as a critical node for accessing distributed information related to recently acquired experience. Damage to a sufficient volume of the limbic system interferes with the coherence of recall and storage even though the constituent fragments of the corresponding experiences may remain stored quite well in other parts of the brain. Additional observations based on the phenomenon of hemispatial neglect lead to the conclusion that transmodal areas and unimodal areas are interconnected with each other to form large scale neural networks that can sustain complex computational architectures including those that rely on parallel distributed processing. Networks organized in this fashion can rapidly access a vast informational landscape while simultaneously considering many goals and constraints. The final compromise into which the network settles is identified as the solution to the cognitive problem. In this neurological model of cognition, the unimodal areas of cortex provide the most veridical building blocks of experience. Transmodal nodes bind this information in a way that introduces temporal and contextual coherence. The formation of specific templates belonging to objects and memories oceurs in distributed form but with considerable regional specialization. This arrangement leads to a highly flexible and powerful computational system which could be described as Selectively Distributed Processing.",
author = "M. Mesulam",
year = "1994",
month = "1",
day = "1",
language = "English (US)",
volume = "150",
pages = "564--569",
journal = "Revue Neurologique",
issn = "0035-3787",
publisher = "Elsevier Masson",
number = "8-9",

}

Neurocognitive networks and selectively distributed processing. / Mesulam, M.

In: Revue Neurologique, Vol. 150, No. 8-9, 01.01.1994, p. 564-569.

Research output: Contribution to journalShort survey

TY - JOUR

T1 - Neurocognitive networks and selectively distributed processing

AU - Mesulam, M.

PY - 1994/1/1

Y1 - 1994/1/1

N2 - The association cortex of the human brain can be divided into unimodal and transmodal components. Unimodal (modality-specific) cortical areas are subdivided into upstream regions specialized for encoding unitary features of experience and downstream regions which are specialized for encoding composite features. Modality-specific features lead to multimodal knowledge through the mediation of transmodal areas in the brain. These transmodal areas include cortical regions that are conventionally designated as heteromodal, paralimbic and limbic cortex. Contrary to earlier formulations, it is no longer thought that these transmodal areas contain a convergent residue of knowledge. Instead, it appears that the role of these transmodal areas is to contain a road map for the multifocal binding and calling up of distributed information in multiple modalities. Knowledge can thus be encoded in a flexible distributed rather than rigid convergent form. Observations on patients with focal neurological lesions indicate that transmodal areas act like neural hubs (or gateways) for accessing critioal domains of knowledge rather than as dedicated centers for specific cognitive functions. In the processes related to memory, a limbic structure such as the hippocampus does not act as a bank for specific memories but as a critical node for accessing distributed information related to recently acquired experience. Damage to a sufficient volume of the limbic system interferes with the coherence of recall and storage even though the constituent fragments of the corresponding experiences may remain stored quite well in other parts of the brain. Additional observations based on the phenomenon of hemispatial neglect lead to the conclusion that transmodal areas and unimodal areas are interconnected with each other to form large scale neural networks that can sustain complex computational architectures including those that rely on parallel distributed processing. Networks organized in this fashion can rapidly access a vast informational landscape while simultaneously considering many goals and constraints. The final compromise into which the network settles is identified as the solution to the cognitive problem. In this neurological model of cognition, the unimodal areas of cortex provide the most veridical building blocks of experience. Transmodal nodes bind this information in a way that introduces temporal and contextual coherence. The formation of specific templates belonging to objects and memories oceurs in distributed form but with considerable regional specialization. This arrangement leads to a highly flexible and powerful computational system which could be described as Selectively Distributed Processing.

AB - The association cortex of the human brain can be divided into unimodal and transmodal components. Unimodal (modality-specific) cortical areas are subdivided into upstream regions specialized for encoding unitary features of experience and downstream regions which are specialized for encoding composite features. Modality-specific features lead to multimodal knowledge through the mediation of transmodal areas in the brain. These transmodal areas include cortical regions that are conventionally designated as heteromodal, paralimbic and limbic cortex. Contrary to earlier formulations, it is no longer thought that these transmodal areas contain a convergent residue of knowledge. Instead, it appears that the role of these transmodal areas is to contain a road map for the multifocal binding and calling up of distributed information in multiple modalities. Knowledge can thus be encoded in a flexible distributed rather than rigid convergent form. Observations on patients with focal neurological lesions indicate that transmodal areas act like neural hubs (or gateways) for accessing critioal domains of knowledge rather than as dedicated centers for specific cognitive functions. In the processes related to memory, a limbic structure such as the hippocampus does not act as a bank for specific memories but as a critical node for accessing distributed information related to recently acquired experience. Damage to a sufficient volume of the limbic system interferes with the coherence of recall and storage even though the constituent fragments of the corresponding experiences may remain stored quite well in other parts of the brain. Additional observations based on the phenomenon of hemispatial neglect lead to the conclusion that transmodal areas and unimodal areas are interconnected with each other to form large scale neural networks that can sustain complex computational architectures including those that rely on parallel distributed processing. Networks organized in this fashion can rapidly access a vast informational landscape while simultaneously considering many goals and constraints. The final compromise into which the network settles is identified as the solution to the cognitive problem. In this neurological model of cognition, the unimodal areas of cortex provide the most veridical building blocks of experience. Transmodal nodes bind this information in a way that introduces temporal and contextual coherence. The formation of specific templates belonging to objects and memories oceurs in distributed form but with considerable regional specialization. This arrangement leads to a highly flexible and powerful computational system which could be described as Selectively Distributed Processing.

UR - http://www.scopus.com/inward/record.url?scp=0028033849&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0028033849&partnerID=8YFLogxK

M3 - Short survey

C2 - 7754292

AN - SCOPUS:0028033849

VL - 150

SP - 564

EP - 569

JO - Revue Neurologique

JF - Revue Neurologique

SN - 0035-3787

IS - 8-9

ER -