../langbrain banner
 
 

Nections and Cortical Columns 

For the relational network model of the linguistic system to be considered neurologically plausible, it has to include a reasonable hypothesis of how the elements of the model (known as NECTIONS) might be implemented in neural structures, and the various properties of nections and their interconnections have to be present in those neural structures. It has been proposed that cortical columns (a.k.a. minicolumns) may provide the physical basis for the nections of the model, the basic elements of which the networks are formed. To test the hypothesis, we may consider the properties of cortical columns, descriptions of cortical cells, and how the structure and connectivity of columns correspond (or fail to correspond) to the requirements of the model. (The information given here is based on Pathways of the Brain, chapters 16-17 and Vernon Mountcastle Perceptual Neuroscience: The Cerebral Cortex, Harvard University Press, 1998. See also Yves Burnod, An Adaptive Neural Nework: The Cerebral Cortex (1990).)

[Cortical Column Properties]
[Cortical Layers]
[Types of Cells]
[How Nection Hypothesis Properties are Met]

 

CORTICAL COLUMN PROPERTIES


CORTICAL LAYERS

Layers of neurons as shown in Nissl stain


Photo from http://webvision.med.utah.edu/VisualCortex.html
Supragranular Layers: Layers I - III
- Layer III receives input from other cortical columns
- Cells of layers II and III project to other parts of the cortex

Granular Layer: Layer IV
- Receives inputs from the thalamus
- Sends signals to the rest of the column (primarily up to layers II and III)

Infragranular Layers: Layers V - VI
- Receive input from the supragranular cells of adjacent columns
- Send signals mainly to extracortical structures (thalamus, other subcortical structures)

TYPES OF CELLS

Pyramidal Cells
  • Dendrites branch profusely in layer I

  • Output connections are excitatory

  • Those of layers II and III project to other cortical areas

  • Those of layers V and VI project to the thalamus and other subcortical structures

  • Axon fibers branch off from pyramidal cell axons and connect to other cells in the column

Spiny Stellate Cells

Basket Cells
  • Output connections are inhibitory
  • Axons run horizontally through the grey matter to other columns (up to a distance of 1 mm or more)

Chandelier Cells

Smooth Stellate Cells


HOW NECTION HYPOTHESIS PROPERTIES ARE MET

1. Varying degrees of strength of connections

Synapses vary in their strength; larger presynaptic terminals emit greater amounts of neurotransmitter. More important, the number of synapses connecting one neuron to another (hence from one column to another, for neurons in different columns) can vary greatly. A strong connection from one nection to another is implemented, according to the hypothesis, as many synapses connecting one column to another.
2. Varying degrees of activation
Excitatory inputs are summed and inhibitory ones subtracted and the result determines the activation emitted. Based on the amount of activation received, a neuron will send out varying frequencies of action potentials. A presynaptic terminal releases varying amounts of neurotransmitter as a function of the frequency of action potentials reaching it.
3. Strengthening of connections as a learning process
Several neural mechanisms have been suggested for how a connection can be strengthened. These mechanisms are neurologically plausible, but the details are not yet fully understood:
  • facilitating the relase of neurotransmitter so that an increased amount will be released, or that it will be more easily released
  • increasing the receptivity of the post-synaptic membrane so that it more easily receives neurotransmitter or responds more readily
  • growth of new terminal buttons and/or growth of new fibers of axons or dendrites or both

4. Local and distant connections
The cortex implements local connections by horizontal axons within the grey matter and distant ones by axons which go from one cortical area to another through the white matter. These long-distance connections are myelinated (see figure). The myelination allows the action potential to travel up to one hundred times faster than in unmyelinated axons. (It is the myelin that makes the white matter white). Thus the activation on a myelinated connection of 100 mm (10 cm) can reach its target as fast as that of a local connection only 1 mm away.

5. Excitatory and Inhibitory Connections

The model requires excitatory connections to be both local and long-distance, but has only local inhibitory connections. This requirement is confirmed neurologically in that the long distance connections are almost entirely from pyramidal, cells which form excitatory synapses with other neurons. As for inhibitory connections, we find basket cells that send connections horizontally to other columns up to one millimeter away. Additionally, there are smooth stellate cells which can turn off activity in neighboring columns.
6. Wait element (needed for the 'ordered and' node)
Cortical columns can also account for the wait element used for sequencing, needed in the internal structure of the 'ordered and' node. The vertical connections of pyramidal and spiny stellate cells can activate other cells in the column, and reciprocal vertical connections between upper and lower layers can keep the activation alive while awaiting further input. This wait device can be turned off by blocking elements such as the chandelier cell, whose vertical axon terminates with inhibitory synapses on axons of pyramidal cells within the same column.
7. Blocking element
In the model, a blocking element blocks activation on a line, for example, a line leading to a default realization. In the neural implementation, blocking of such a line corresponds to one or more axo-axonal connections; that is, an inhibitory connection attaching to the axon of another neuron rather than to a dendrite or the cell body. Such connections do indeed exist in the cortex. For example, axons of the chandelier cell, described above, as well as the horizontal axon fibers of large basket cells.
8. Bidirectional Connections
The property of bidirectional processing is implemented in the cortex by the connections of the pyramidal cells of a column, which provide feed-forward and feed-backward connections to higher level and lower level structures. The axons of these neurons extend to the white matter to other cortical areas as well as locally via horizontal connections to neighboring columns.
9. Abundance of Latent Connections (Needed for Learning Hypothesis)
The typical neuron in the cortex has many thousands of connections to other neurons. It has been hypothesized, for example by Gerald Edelman (see information sources), that most of these connections are very weak (latent) and non-functioning, but available for strengthening if needed.


Go to Menu (for non-frame version.)


This page was last modified on 15 February 2002.

© 2000 Rice University. This document, or any portion hereof, may be used for non-commercial informational purposes only. Any copy of this document, or portion hereof, must include the copyright notice (http://www.rice.edu/about/cr-notice.html) in its entirety.