In terms of evolution, the major changes in the cortex have affected its tangential rather than radial expansion. This coincides not with a marked change in basic cortical circuitry, but rather an increasing arealisation, i.e. cytoarchitecturally-distinct cortical regions, which increase the complexity of function (and one supposes adaptation) (Welker, 1990). I would thus suggest that there may be a basic conservation of thickness to subserve a balance between feedforward / feedback connections across functional hierarchies (Barbas), but that the more fundamental evolutionary adaptation is in the tangential expansion of the cortex and its increasing arealisation.
It's a fascinating observation, and speaks volumes about the importance of the laminar organization of the cortex. There is a recent review paper about the conservation of brain oscillations across species (http://www.ncbi.nlm.nih.gov/pubmed/24183025). It's not directly addressing your question, but is quite relevant.
Very interesting; many cytoarchitectural organizations appear conserved (i.e., the number of layers and the relative configuration of cells within Brodmann's areas) in at least comparative studies in primates and rats, so I'd be curious how far across mammalian species specific Brodmann's designations are conserved. Specific neural configurations that have survived evolution to support important discrete and oscillatory functions may have resulted in no "need" for other organizations to have arisen.There are probably many principles of biophysical and functional efficiency implicit in this arrangement.
Very interesting question! I suspect there will be multiple factors, any one of which begs the question somewhat. A major determinant will likely be the developmental timing and patterning. Given that the number and density of cells differs across species (and regions), as do inter-species differences in the period of cortical lamination, if cortical thickness is constrained within some constant range, it would stand to reason that the events necessary for normal proliferation and migration (too many to list here) down-regulate when certain spatial constraints (e.g., total thickness of layers) near some thresholds. Agree with Mikhail that blood supply (pial-glial especially) might be an important factor. If that is the case, the -- wild speculation -- fish could be the vertebrates that show the most deviation in cortical thickness, given the substantial differences in circulatory constraints compared to mammals. More generally, the thickness constraint might involve physiological limits on astrocyte-neuron interaction, particularly with respect to energy-demanding pyramidal cells. It would be very interesting to look across species at numbers, densities, and layer distribution of not only different kinds of cortical neurons (pyramidal and interneurons), but also astrocytes, and throw in the duration of lamination during cortical development as a factor. If anyone has done such a comparison across species, I'd love to see it!
Another question: how much do pyramidal cells differ in size across species? This is an ill-formed thought, but I'm thinking of spatial summation in pyramidal soma (esp. Betz cells). If layer 5 pyramidal cell morphology does not vary *too* much across species, then there will likely be hard constraints on how much afferent dendritic branching you need to support normal descending motor cell activity. Perhaps having, for example, too thick a layer 5 would makes afferent fibers too widely distributed to support normal summation and descending motor neuron firing. This is wild speculation, and I'm far from confident about it because I know, for example, that different cetacean brains have very different relative proportional thickness of different layers. One would need to consider each layer and the types of cells in it, and their input patterns, for this story to make sense. But I'd love to hear from someone who is expert in these variables across species!
So far, I'm not convinced that it has to do with geometry of blood supply. Couldn't the arterioles just branch in a different pattern if the cortex was thicker? Perhaps there is a cytoarchitectural "module" (with interneuronal feedback, etc.) that is hard to improve on functionally that has an optimal thickness of about 2 mm?
While we are at it, why are all of the pyramidal cell apical dendrites oriented in exactly the same direction?
Michael Cohen - Perhaps the 1.5 to 2.0 mm thickness has to do with preserving cortical rhythms in their heirarchical and stereotyped manner in animals with different brain sizes? Gamma and theta rhythms can develop in isolated neocortex tissue slices.
In terms of evolution, the major changes in the cortex have affected its tangential rather than radial expansion. This coincides not with a marked change in basic cortical circuitry, but rather an increasing arealisation, i.e. cytoarchitecturally-distinct cortical regions, which increase the complexity of function (and one supposes adaptation) (Welker, 1990). I would thus suggest that there may be a basic conservation of thickness to subserve a balance between feedforward / feedback connections across functional hierarchies (Barbas), but that the more fundamental evolutionary adaptation is in the tangential expansion of the cortex and its increasing arealisation.
"Couldn't the arterioles just branch in a different pattern if the cortex was thicker?"
To the best of my knowledge, cortex is receives its blood supply from the arteries located on the cortical surface. This is by the way the reason why stroke affects the cortex the way it does.
From hydraulic/geometrical considerations it might be more effective to irrigate with blood relatively thin layers of grey matter.
This might be one of the reasons why cortex expands by producing large, convoluted surfaces (again, covered by a net of blood vessels which start on the surface and penetrate deep) rather than increasing the thickness.
You might want to check comparative anatomy of other organs that have "cortcices" (kidney, lung?) to see if how they scale with animal size.
I'm no expert in brain (or any other organ) vascular, but in general, the cortex is not covered by a homogenous blanket of blood vessels. Blood needs to travel from the main vessels farther laterally than the thickness of the cortex. It's possible that blood supply is related to cortical thickness constraints, but I'd guess that's not the main reason. Anyway, if easy blood supply were so important, the brain might be located just under the heart (although then we'd either have awkwardly small heads, or eyes on our chest ;) ).
I think important insight into this problem comes from the remarkable conservation of oscillations and other biophysical properties, as they relate to neural function, across species. Consider the large layer-5 pyramidal cell (I'm furthering the argument above by Gedeon Deák): One of their main functions is to integrate inputs from superficial and deep layers. Primate L5 pyramidals can be >700 microns long, which is actually quite long, relatively. So long in fact that there are phase delays of up to several ms along the vertical axis. To compensate for these phase lags, there are active phase-advance mechanisms along the dendrites. Imagine that the cortex were 10 mm thick; the phase lags along one neuron would be huge and would likely prevent even the simplest neuronal computations (e.g., coincidence detection) from happening. Thus, it might simply be too "expensive" to have a thick cortex while maintaining computational speeds on the order of milliseconds. Pretty much all of neuroscience points to *time* as being incredibly important in neural computation, and it's possible that ~2-4 mm is about as thick as the cortex can get without sacrificing the temporal precision of information processing.
That's a lot of hand-waving, of course. The experiment to test this (I have no idea whether it or anything like it has been done) would be to compare the efficacy of sub-neuronal computations as a function of neuron length as a function of energy consumed by the neuron. My hunch is that there is some kind of "sweet spot" balance between efficiency and energy, and beyond a certain point, it takes a lot of energy to maintain only a little bit of computational power.
+1 Michael's answer (need for phase-advance mechanisms to maintain synchrony); it makes sense.
Charles: descending fibers tracts are established pretty early in embryological & fetal development and (as far as I know) the process of axonal pathfinding is fairly conserved, at least in mammals (not sure about, e.g., fish or birds). For this and other reasons I assume that thickness is constrained by (in humans) middle of third trimester (when ventricular proliferation has largely stopped and layer V pyramidal apical dendrites have started pathfinding). But there are still big scaling factors in radial distance those fibers must travel if you consider differences in mammalian brain size (not to mention overall spinal tract lengths). One way of posing the original question would be, why do these scaling factors "stop" at cortex?
re: blood flow: I'm not an expert but I don't think irrigating the surface can do it. Cells can't get enough O2 unless it diffuses from hemoglobin no more than 1-2 cell widths away; neurons are O2 greedy so there's no way that simple arterial profusion at the surface of cortex can do the job no matter the size or gyrification of cortex; you still need extensive capillary system within (plus the astrocyte network), even at the thinnest parts of cortex.
I am not really an expert on any of the topics that have been discussed here so far, but I think that Michael Cohen's answer probably holds some very important clues...It is likely that the basic style of information processing that is accompanied by cortical rhythms (alpha, gamma, etc) works well with the geometry that exists in a 2 mm thick cortex, and this might all fall apart with thicker cortex, regardless of how the geometry was expanded. This is an idea that is worth examining further!
I would suggest that there are information processing reasons for the cortex thickness. These reasons drive the need for specific numbers of layers of pyramidal neurons, then the number of layers drives thickness.
There is a tendency towards connectivity vertical to the cortex sheet, and there are often four major layers of pyramidal neurons. The result is that the receptive fields detected by later layers are slightly more complex than earlier layers but within the same input information (very roughly, neurons in later layers detect the presence of combinations of receptive fields detected by earlier layers). Thus layer IV neurons detect receptive fields that are combinations of inputs to the local region (perhaps column) of the cortex, layer II/III neurons detect fields that are combinations of layer IV fields, layers V and VI detect receptive fields that are combinations of IV and II/III fields, with V and VI fields being at different levels of complexity.
The number of layers is then dictated by the number of different levels of receptive field complexity that are required to generate the information needed to manage cortical functioning. For example, four needed functions could be 1. adding appropriate inputs from the senses or other regions to fields detected in a local area; 2. determining whether or not any changes should be made to receptive fields in a local area; 3. providing appropriate receptive field detections to other cortical areas; and 4. providing receptive field detections outside the cortex to influence behaviours. Thus layer IV receives external inputs, layers II/III mainly go to the hippocampal system, layer V to other cortical areas, and layer VI to the striatum, thalamus, cerebellum and spinal cord.
More layers would allow better discrimination in the outputs provided to other structures. But more layers will have a higher resource cost. So the number of layers is a compromise imposed by natural selection. With this compromise, some receptive fields in one layer may have complexities appropriate for more than one function. Furthermore, best compromise is different for different cortical areas, resulting in slightly different layering and different thicknesses.
Thanks to Andrew Howard for a very thoughtful answer. This sounds to me intuitively likely to be an evolutionary reason for the thickness of cortex that (together with info about integration by lengthy Layer 5 pyramidal neurons) makes a lot of sense.
I think considering neural microcircuits in all layers of cortex and also specific patterns of spatial distribution of neurons and interneurons in layers 1-6, cortex with this limited thickness have enough complexity for processing sensory or motor or ... information, in evolutionary point of view I believe increasing tangential expansion have been more necessary.
A previous literature suggests that the thickness of the human cortex ranges between 1.5 and 4.5 mm in different cortical regions (Parent and Carpenter 1995).
Parent A, Carpenter MB. 1995. Human neuroanatomy. Baltimore (MD):
This is a very interesting question and the right answer(s) needs to be read here and there. It is unlikely that vascularization has to do with the thickness of the mammalian cortex. Rather, this has to deal with its evolution in vertebrates in function of neuronal size and bioelectric properties. The cortex has a six-layer structure and mammalian neurons have a length constraint related to their computational ability. Consequently, regardless of the size of the animal, its neurons need to respect physical rules for efficient information processing, so that their size across different species is similar. In my opinion, this is what determines the architecture of the cortex, and therefore its thickness. This is therefore maintained along with evolution. On the other hand, increasing the surface area is the solution of choice to allow the increase in functional complexity of the mammalian and particularly the human brain.
Thanks, Arturo - There seems to be a consensus that one of the things driving the conserved thickness of cortex is the length constant of long pyramidal neurons that span all the layers.
Just a quick addition to the debate - in humans, grey matter thickness extends beyond 2mm, with primary visual and motor/somatosensory cortices having the thinnest GM overall.