UK scientists who tested rats’ brains found that brain circuits are naturally “noisy” and susceptible to the celebrated “butterfly effect”, and to overcome this natural byproduct of complexity they probably use a system called “rate” code to cut out background noise, in a similar way to computer circuits.

You can read how researchers at University College London (UCL) came to these conclusions in a paper published online in Nature on 1 July.

The authors write that we already know that neural circuits show variability: for example the same sensory stimuli (such as the chemical reaction on the tongue’s taste buds when we suck a lemon) can result in different patterns of electrical responses to the brain. But we don’t whether this variability means anything: does it carry extra information, for example about the internal state of the organism?

Lead author Dr Mickey London, and fellow experimentalists at UCL’s Wolfson Institute for Biomedical Research and Peter Latham, a theorist at UCL’s Gatsby Computational Neuroscience Unit, decided to explore this variability phenomenon “bottom up”, and took inspiration from the celebrated butterfly effect.

The butterfly effect is the idea that small air pertubations caused by a butterfly fluttering its wings on one continent grow, through a series of coincidental reinforcing events, to become a tornado on another continent.

They introduced a small pertubation, effectively a single extra “spike” into a single neuron, in the brain of a rat and observed whether it would have a knock on effect: would the cortical networks amplify the signal, or would it die out straight away?

They found it had a huge knock on effect: the single spike caused about 28 additional spikes in other neurons, which then had another knock on effect of a similar size, and so on.

This may not seem like very much, given that the brain produces millions of spikes every second, but the researchers estimated that one extra spike eventually affected millions of neurons in the brain, and showed, “using simultaneous intra- and extra-cellular recordings, that a single spike in a neuron produces a detectable increase in firing rate in the local network”.

With further analysis they estimated that this amplification of a single spike leads to “intrinsic, stimulus-independent variations in membrane potential” of the order of plus or minus 2.2-4.5 mV, which suggests the variations are “pure noise, and so carry no information at all,” they wrote.

The authors considered two current theories that attempt to explain how the brain distiguishes between noise and useful information: rate code and synfire chains.

Rate code theory proposes that the different parts of the brain and central nervous system communicate with each other using signal patterns that conform to a certain rate of firing of neurons, and ignore signals occuring at different rates.

Synfire theory proposes that pools of neurons behave like feedforward networks or chains, each able to transmit packets of synchronized spikes during very fast surges of highly reliable impulses and the computational functions of the brain occur during these chains.

However, they suggest rate code was the most likely explanation because in their in vivo recordings they found “large, fast depolarizing events, such as those proposed by the theory of synfire chains” were very rare. So they concluded:

“Our findings are thus consistent with the idea that cortex is likely to use primarily a rate code.”

London told the press that:

“This result indicates that the variability we see in the brain may actually be due to noise, and represents a fundamental feature of normal brain function.”

This study shows that the brain is much noisier than computers, and suggests how it might deal with it, but we still don’t know why.

The UCL researchers said it could be the price the brain pays for having high connectivity in neural networks.

In the human brain each neuron links to about 10,000 others, totalling some 8 million kilometers of “wiring”, so perhaps the noise is not so much a useful function as a byproduct of having a useful function.

“Sensitivity to perturbations in vivo implies high noise and suggests rate coding in cortex.”
Michael London, Arnd Roth, Lisa Beeren, Michael Häusser, Peter E. Latham.
Nature 466, 123-127 (1 July 2010)
DOI:10.1038/nature09086

Additional source: UCL.

Written by: Catharine Paddock, PhD