Our knowledge networks are built up slowly and gradually over the years to provide us with ways of acting in the world, methods of distinguishing one thing from another, memories of what happened a long time ago and information about what has just occurred. What happens when these networks are damaged? When someone has a stroke--when an artery in the brain is blocked by a blood clot and the surrounding area no longer receives blood to nourish its neurons, so that many of them die all at once? Or when someone has a brain hemorrhage--when an artery in the brain bursts and floods the surrounding area with blood, again killing many neurons at once?
int here is that it is the event of many neurons in a single area all dying at the same time that causes the damage. The death of a few neurons in any one area is actually quite commonplace and totally unnoticeable. Neurons in our brains die every day, and we are none the worse for it, even though not many new neurons grow after infancy. The reason our brains are so resilient, and can cope so easily with the daily loss of a few neurons here and there, is precisely the fact that the information in our networks is distributed over so many neurons. Since our memories and knowledge are not contained in individual neurons or small groups of them, but are distributed over whole networks, each individual connection between two neurons contains only a tiny part of the information, and the loss of a few such connections leaves the whole memory or piece of knowledge reasonably intact. A stroke, however, whether caused by a clot or a hemorrhage, may end up killing quite a large number of neurons in the same network. In this case the damage will certainly be noticeable, but it will still be distributed. If the damage is extensive, one or more entire networks, such as the animal-name or the food-name network, may be totally destroyed, in which case the patient will not be able to say the name of any animals or any foods at all. If the damage is less severe, it will not totally destroy any particular piece of information in the network. The reason for this is that each piece of knowledge is stored over the whole network, and part of the network is still intact. What happens instead is that all the information in the network will be damaged to some extent, depending on the degree of damage to the network as a whole, rather than some bits being totally lost and some bits being totally retained.
Recall and language
For example, let's say Matilda's damaged network is the animal-name network, and let's say that about a fifth of this network is affected. If the animal names were stored at random in individual neurons or groups of neurons, then we would expect Matilda to be unable to name about a fifth of the animal pictures she is shown on a test. If names of animals that are closely associated, such as cat and dog or cat and tiger, were stored together in a physically close set of neurons, we would expect Matilda to be unable to name all the animals in some particular group, say all the rodents or all the members of the cat family. But such outcomes of strokes have not been observed. The usual case is actually quite different. Assuming that Matilda's case is typical, she will probably have some trouble naming all the animals she knows. When she is shown pictures of various animals, it will take her longer to name all of them than it would have taken her before. Moreover, she will probably have the least trouble naming familiar, well-known animals, such as cats or mice, while she may be totally unable to recall the names of unfamiliar animals, such as opossums or ocelots.
This occurs because a large number of the connections between the neurons in this network have been damaged, but a large number also remain intact. In the case of the very familiar names, the connections were all quite strong to begin with, so the ones that remain intact are sufficient to enable Matilda to recall the names. In the case of the unfamiliar names, the connections were never very strong, and so all of them working together were needed to allow the name to be activated when the picture is seen. Thus, when a large number of these connections are damaged, the remaining ones are not strong enough to enable Matilda to retrieve the names when she sees the pictures.
But what happens in intermediate cases, where the animals are moderately familiar, such as kangaroos or giraffes? Each network is connected to many others, and a particular pattern in one network can be activated by patterns in several other networks that are active at the same time. For an unimpaired person who is unable to recall the name of a big cat in a picture, the first letter 'o' might help the person recall the name 'ocelot.'
What happens to stroke patients when they try to recall moderately familiar words is often quite similar. In our example, Matilda may be unable to recall the word 'donkey' when presented with the picture of a donkey. She may even say 'horse,' just like a child learning to speak, because only the well-used connections associated with 'horse' are strong enough to activate a word. In such a case the first letter 'd' may help Matilda recall the name 'donkey.'
Indeed, this particular type of cuing has been used for a long time by neuropsychologists examining the effects of brain damage on mental functions. A stroke patient who can recall 'donkey' when presented with a picture of a donkey alone is considered in better shape than one who needs the first letter 'd' as well as the picture, while a person who can recall 'donkey' only with the aid of both the picture and the letter is still not as severely injured as one who cannot recall the word even with both cues. The theory that input from several networks provides more activation than input from only one thus helps us understand this well-known clinical finding.
