Brain connections cause rethink over human memory

Brain connections cause rethink over human memory
From issue 2621 of New Scientist magazine, 18 September 2007, page 23

How do we store so many memories? It is a question that has puzzled neuroscientists for decades - and now it seems that our concept of how memory works may have been wrong all along.

It was originally assumed that the number of memories was proportional to the number of neurons in a network. Given that even 1 cubic centimetre of the brain's cortex contains about 50 million neurons, it seemed that the brain could indeed store masses of information. However, this model relied on the notion that each neuron is connected to every other neuron, whereas a neuron is actually connected to between 5000 and 10,000 others.

Neuroscientists then proposed that the number of memories was proportional to the number of connections per neuron. Now Yasser Roudi and Peter Latham at University College London have found a problem with this model too. They calculated that even with 10,000 connections per neuron, a network could only store about 100 memories - regardless of how many neurons were in the network (PLoS Computational Biology, DOI: 10.1371/journal.pcbi.0030141).

This implies that to store a large amount of information, the brain would have to use multiple networks. This may be problematic for something like vocabulary, Latham says. "You wouldn't want to store 100 words in each of [many] different networks; you probably want to store them more or less in one place. Now we don't know how [the brain] does this."