Isn't very modular in what context, brain tissue doesn't form into tiny little squares?
No.
The modularity of the brain is evident in the various lobes, cortical specializations and the fact that those lobes interconnect with fiber tissues that act like very dense ribbon cables.
Cortical specializations such as...? Broca's area and speech? Except that even such an oft cited example of specialization is an over-simplification, as Broca's area is involved in more than speech (see e.g.,
Maess, B., Koelsch, S., Gunter, T, & Friederici, A. D. (2001). Musical Syntax is Processed in Broca's Area: An MEG Study. Nature Neuroscience, 4, 540-545. ).
There is a reason why embodied cognition is such a big debate (with probably a majority now backing the theory): functional imaging and behavioral experiments suggest that conceptual processing is at its heart multimodal, so much so that even the most abstract concepts are represented across modalities such that there is not even an amodal, purely symbolic "core".
Trans-cranial stimulation capitalizes on modularity of the brain to enhance learning.
Right. And the support for this comes from...?
So the neural architecture is modular right down to the neuron as the basic processing unit.
The neuron likely isn't "the basic processing unit" most of the time, perhaps not at all.
First, not all neurons fire spikes, some fire bursts. Second, "The discussion to this point has focused on information carried by single neurons, but information is typically encoded by neuronal populations...Synchronous firing of two or more neurons is one mechanism for conveying information" (Theoretical Neuroscience, 2001, MIT press). Third, oscillating thesholds, spike latency, switching from in integrator to a resonator, and a lot of other things constitute the mechanisms of data transference. However, much more simplistically, you can think of the data not as a spike or burst itself, but the frequency of spikes. Very simplistically, one bit might be a spike train with a particular frequency, and another a different frequency. A lot of other things come into play, but the point is the single spike conveys on its own no information. Other neurons which receive the information respond to to the spikes but to things like the frequency of spikes in a spike train. That's (again, simplistically) the "bit."
What most people don't get is the interconnectivity of neurons.
Nobody gets it. It's an open question:
I guess the best way to answer the question about whether we know how a neuron works is to quote a monograph on the subject: "In every small volume of the cortex, thousands of spikes are emitted each millisecond...What is the information contained in such temporal pattern of pulses? What code is used by the neurons to transmit that information? How might other neurons decode the signal?... The above questions point to the problem of neuronal coding, one of the fundamental isssues in neuroscience. At present, a definite answer to these questions is not known." from Gerstner & Kistler's Spiking Neuron Models: Single Neurons, Populations, Plasticity (Cambridge University Press, 2002).
Put to you this way; digital computers interconnect bits through a bus.
And you need go no further. Because neurons, neural networks, and neural signals operate using fundamentally different principles than digital processors. Almost nothing about processing speed, online or long-term memory, or really anything which has to do with the functional limits (and the reasons behind them) of computers is related to similar issues in the brain. Hence the following (again):
From Manrubia, Susanna C.; Mikhailov, Alexander S.; Zanette, Damian H..(2004). Emergence of Dynamical Order : Synchronization Phenomena in Complex Systems. World Scientific Publishing Co., p 312:
"The biological hardware on which the brain is based is extremely slow. A typical interval between the spikes of an individual neuron is about 50 ms and the time needed to propagate a signal from one neuron to another is not much shorter than such an interval. This corresponds to a characteristic frequency of merely 100 Hz. Recalling that modern digital computers should operate at a frequency of 10^9 Hz and yet are not able to reproduce its main functions, we are lead to conclude that the brain should work in a way fundamentally different from digital information processing.
Simple estimates indicate that spiking in populations of neurons must be synchronized in order to yield the known brain operations. Humans can recognize and classify complex (visual) scenes within 400-500 ms. In a simple reaction time experiment, responses are given by pressing or releasing a button. Since movement of the finger alone takes about 200-300 ms, this leaves less than 200 ms to make the decision and classify the visual scene [Gerstner (2001)l. This means that, within the time during which the decision has been made, a single neuron could have fired only 4 or 5 times! The perception of a visual scene involves a concerted action of a population of neurons. We see that exchange of information between them should take place within such a short time that only a few spikes are generated by each neuron. Therefore, information cannot be encoded only in the rates of firing and the phases (that is, the precise moments of firing) are important. In other words, phase relationships in the spikes of individual neurons in a population are essential and the firing moments of neurons should be correlated."
If you read the above carefully, you'll realize that using bits and digital systems to understand "bottlenecks" in biochemical processing systems (particularly neural processors, like brains) is so worthless.
The critical component of any parallel processing is the time necessary to communicate the results of the information that each processing unit performs and the energy consumption for each processing unit, the reason for the U shape graph.
Actually, what's really important is understanding
1) what the neural "code" is
2) How often (if ever) a single neuron's spike train constitutes a "basic unit" (i.e., do the action potentials of a single neuron ever act like a "bit" and if so, when and why as well as why not)?
3) How is the brain so slow and yet so unbelievably fast?
The edited series
Computational Neuroscience put out an edited volume in 2006 entitle
23 Problems in Systems Neuroscience. Each paper included was meant to address a specific, open, and fundamental question concerning neuroscience. All three of the above questions have at least one parallel in the volume. There is, for example, a paper entitled "How can the brain be so fast?", and several papers on different aspects of neural encoding/decoding (including "What is the neural code?").
Now Compare this to a synapse that has up to ten thousand connections. This would be like allowing each CPU register to directly connect to ten thousand other processing units!
It wouldn't, and isn't.
This is why a brain the size of an insect's can run circles around any CPU of today.
It can't.
With larger brains, as mentioned earlier, we find fiber tissues interconnecting cortical modules to link themselves together.
"fiber tissues"? A single neuron can have a dendritic tree connecting it directly to over 100,000 other neurons. In other words, one neuron can
directly connect to over 100,000 other neurons. Likewise, their axons can run across the entire brain (or body) in some cases.
These fibers are in the order of billions of transmission lines communicating information across each lobe. Nature has mastered the ability to modularize information processing using highly paralleled processing units by employing massive interconnectivity.
Where is the modularity in all of this?
