Outline continued 3

So what? - most neuroscientists say: the learning process at individual synapses is now fairly well understood, and seems to be quite accurate – presumably as accurate as needed! We disagree. We think that the other properties that synapses must have – particularly that they must actually take part in the ongoing activity of the brain – means that they cannot show extraordinary accuracy. This problem is compounded by the fact that synapses are extremely close together – far closer than the pixels on your computer screen, or in your camera. The problem is actually exactly the same as in DNA – the chemical pairing of individual bases that underlies copying is not extraordinarily specific (and if it were the double helix could never peel apart, as required for cell function and division). It has to be boosted by additional biochemical processes – in particular "proofreading". The idea behind proofreading is again very familiar and simple. If you write a text, you will make mistakes, ranging from typos to topics. The best way to correct any manuscript is to give it to an independent reader, who will quickly pick up at least the easy errors. DNA is copied the same way. The enzyme does a first draft, and then backs up, reads it, clips out mistakes, and tries again. Of course the proofreader can make mistakes too (often more than the author) but the chances of a double mistake (one made by both author and the independent proofreader) are very low.

So finally we get to our central idea, the thing we think explains understanding: the part of the brain that is thought to be the seat of the mind, the cortex, does proofreading! No cortex – no proofreading – no understanding – no mind! In the final paragraph we try to explain how the cortex does proofreading. It’s the most technical paragraph, and not essential for the crux of our argument. But we hope you are now interested enough to go this last step, and then explore the rest of the website.

The adjustment of individual synapses, each much smaller than a camera pixel, that underlies learning is thought to occur in the following way. The synapse has an input "knob" sitting on an output "knob". When both knobs are electrically active at the same time (a neural "coincidence"), because of the ongoing activity of the 2 nerve cells of which they form part, the synapse strengthens, as a result of the release of an internal chemical signal. This makes it more likely that the ongoing pattern of activity will occur again, leading to "learning", because whenever the input knob is active again, it is more likely to make the nerve cell to which the output knob belongs more active. Notice that the synapse needs to do 2 contradictory things: the chemical message must be confined to the output knob (if it spreads to other synapses they will be erroneously strengthened) but the output knob must "tell" the rest of the cell that it has been activated by the input knob. It is because these 2 functions are contradictory but unavoidable that learning cannot be 100% accurate. We think the cortex contains, in its deepest layer, a set of specialized "proofreading" nerve cells, each one corresponding to a different connection between the upper layers. These proofreading cells get electrical signals from both the input and output sides of synapses, so they can make an additional independent measurement of neural coincidence; this is then fed back, by an indirect route, to the relevant synapse, where it "approves" the draft change that occurred in the synapse. The circuitry underlying this proofreading operation is quite complicated, and probably forms the major part of all cortical wiring. Furthermore, as learning proceeds, the circuitry must itself be updated, by an offline process we think corresponds to sleep. It’s quite difficult for most people to see the connection between this intricate and unfamiliar set of processes and what we immediately recognize as "mind" – but then it’s also quite difficult to see why accurate DNA replication, and the associated intricate machinery, is the basis for "life". But it is.