Draft of a review article section


Edit: What I argued in this was wrong. I might be wrong again. I’ll try to summarize it before the wall of text.

  1. Barrel cortex is functionally two regions, which are the barrels and the matrix of septa around them.

  2. To a degree, those regions share some layers, so those layers don’t respect hierarchy.
    This supports HTM theory because hierarchy isn’t as strict as usually thought for some layers.

  3. All signals from L6 to thalamus are down the thalamus/cortex hierarchy, unlike what I was convinced of.

  4. It seems like the cortex can process smaller scales in one region and then use that to help understand direct sensory inputs which represent larger scales or more complex direct sensory inputs. The direct sensory input and the cortical signal can converge in the thalamus, so it’s like adding labels to objects in a blurred or confusing image.

  5. Hierarchy is better defined according to increasingly complex or broad sensory inputs than abstracted representations. L5 converts from a cortical to a subcortical representation, so its role mirrors like the increasingly complex direct sensory inputs to higher and higher levels of the hierarchy. This supports HTM theory because every region can receive representations of the feature, which are presumably relevant only to the sensory input in current moment, even if that feature is just the subcortical representation sent by L5.

Wall of text and sources

The head and tail of VPM are probably actually higher order parts of the thalamus, meaning they receive input from L5 [1]. They also receive direct sensory input (which is normal although uncommon) with larger receptive fields than the core of VPM receives. They both target the septa of barrel cortex [2, 11].

The head of VPM drives the septa [2], which means the septal columns form a higher order cortical region because they are driven by higher order thalamus. One article argued that the septa is higher order, but that was based on its input from POm, which was later found not to drive the septa [2, 4]. That claim was actually correct, except the driver isn’t POm.

I called one type of L6 corticothalamic cell in barrel cortex dual-projecting. That’s not the right description. Instead, they target two higher order nuclei. One of those drives the septa (head of VPM) and the other drives S2 and possibly also M1 (POm). There’s a third higher order subnucleus I’m ignoring (tail of VPM), which drives S2 mainly but also the septa in S1.

Side note:
“Drives” is a bit misleading because there are a few driver inputs to S2 and M1 (meaning they mainly target L4 like the main sensory input to primary cortex. I’m not sure they actually have the driver properties described by Sherman and Guillery)." S2 receives the VPM tail in the same way as its other inputs to L4 of S2, which are POm and possibly L2/3 of S1, and M1 receives a bunch of driver inputs. There are multiple driver inputs, but I don’t think they’re much different because I think L5 converts cortical representations into a form suitable for the subcortex, including whichever cortical sensory representations aid subcortical sensory processing, and the thalamus converts back to cortical representations. Otherwise, it wouldn’t make much sense to combine signals from L5 with direct sensory inputs in the tail and head of VPM and part of POm.

Since septa is higher order, all signals from L6 to the thalamus are probably feedback. What I called dual-projecting L6 CT cells target POm and VPM, probably just the head (the border with POm) [7], which are both higher order. Those L6 cells are probably all in septal columns [8, 9], so they send feedback to the thalamus. Also, POm and the head of VPM probably receive the same coding of the sensory input or submodality, different from what the primary thalamus part of VPM receives [5].

POm drives S2 [4], whereas the head of VPM drives the septa of S1, but both are higher order. I think there are two higher order nuclei because of the what and where pathways, since barrel columns project to S2 whereas septal columns project to M1 [10], although I haven’t checked how much that exaggerates.

Some layers besides L4 have less distinction between septal columns and barrel columns. I haven’t researched this much. It’s possible that rodents re-use some layers for two cortical regions, so those layers do not require hierarchical organization. There seem to be dual-projecting L6 CT cells in rat V1 and tree shrew V1, so the same could apply to those regions.

Point 5:
I wrote this above this spoiler but it was too long.
There is a lot of convergence of L5 and sensory inputs in the thalamus, which supports the idea in HTM theory that L5 represents displacements or at least something with sensory meaning. Direct transthalamic sensory inputs makes many regions primary cortex in a way. This implies that L5 sends signals to the thalamus which is on the same hierarchical level as the subcortex.
Layer 5 could convert from cortical representations to subcortical ones (e.g. more temporally precise and integrated over less time or otherwise less abstracted), and the thalamus could convert back to cortical representations.
This would make every cortical region only one hierarchical level above the subcortex. They all receive a subcortical representation which hierarchically lower L5 produced, so they all receive input from the lowest level of the hierarchy, the subcortex or subcortex-like signals. Direct corticocortical connections are still hierarchical, just not the CTC pathway in a sense. Rather, the CTC pathway’s version of hierarchy follows the increasingly complex sensory inputs to each cortical region, or substitutes for it.

[1] Distribution of Large Terminal Inputs From the Primary and Secondary Somatosensory Cortices to the Dorsal Thalamus in the Rodent (Chia-Chi Liao, Ruei-Feng Chen, Wen-Sung Lai, Rick C. S. Lin, and Chen-Tung Yen, 2010)

[2] Septal Neurons in Barrel Cortex Derive Their Receptive Field Input from the Lemniscal Pathway (Takahiro Furuta, Takeshi Kaneko, and Martin Deschênes, 2009)

[3] Corticothalamic Projections from the Rat Primary Somatosensory Cortex (Herbert P. Killackey and S. Murray Sherman, 2003)

[4] Properties of the thalamic projection from the posterior medial nucleus to primary and secondary somatosensory cortices in the mouse (Angela N. Viaene, Iraklis Petrof, and S. Murray Sherman, 2011)

[5] Extracting functional components of neural dynamics with Independent Component Analysis and inverse Current Source Density (Szymon Łęski, Ewa Kublik, Daniel A. Świejkowski, Andrzej Wróbel, and Daniel K. Wójcik, 2010)

[6] Cell Type–Specific Three-Dimensional Structure of Thalamocortical Circuits in a Column of Rat Vibrissal Cortex (Oberlaender et al., 2011)
See the reconstructions in figure 5.

[7] Corticothalamic Projections from the Cortical Barrel Field to the Somatosensory Thalamus in Rats: A Single-fibre Study Using Biocytin as an Anterograde Tracer (Jacques Bourassa, Didier Pinault, and Martin Deschênes, 1995)
The VPM head borders POm. Figures 5 and 6 show the projections from lower L6 are to POm and the VPM head. The VPM head isn’t actually part of POm because the L6 CT cells form two terminal arbors near the VPM/POm border.

[8] Corticothalamic Projections from the Rat Primary Somatosensory Cortex (Herbert P. Killackey and S. Murray Sherman, 2003)
Figure 1 shows that VPM-projecting cells are not in L6b, which here means the developmental remnant of the subplate. It also shows that projections to POm are not in septal columns, except in L6b.

[9] Intracortical Axonal Projections of Lamina VI Cells of the Primary Somatosensory Cortex in the Rat: A Single-Cell Labeling Study (Zhong-Wei Zhang and Martin Deschênes, 1997)
This shows that cells projecting to VPM and POm are only in the lower half of L6, although it excluded the subplate remnant supposedly. It also found a cell projecting only to POm just slightly above half the depth of L6. Since [8] shows the whole depth of L6 projects to POm, cells only projecting to POm are in septal columns in upper L6. Too little is known about the subplate remnant to say whether it is part of the septal columns functionally, as the voids of cells projecting to POm in [8] are rounded.

[10] Deconstructing the Cortical Column in the Barrel Cortex (Kevin Fox, 2018)

[11] Parallel Streams for the Relay of Vibrissal Information through Thalamic Barreloids (Tiphaine Pierret, Philippe Lavallée, and Martin Deschênes, 2000)

Also see Drivers of the Primate Thalamus (Zita Rovó, István Ulbert, and László Acsády, 2012)
Some thalamic subnuclei in primates receive subcortical and cortical driver inputs.