AS YOU read this magazine, think for a moment about what your brain manages to do. Within a fraction of a second, you can recognise the words and retrieve their meanings. It doesn’t matter whether you see a word in CAPITALS, in lower case or even in MiXeD cAsE, in a different font, or in an unusual position or size – you still manage to recognise it in all these varied forms. Yet you are also exquisitely sensitive to tiny subtleties that may make all the difference to a word, distinguishing “eight” and “sight”, for instance. Our visual system seems perfectly adapted to the job. But how can that be?
Reading presents a real paradox to neurobiologists. It was only invented a few thousand years ago, so there really has not been enough time for our brain to evolve specialised ways to do it. How do brain circuits produced by millions of years of evolution in a world without written words adapt to the specific challenges of reading? We know we have to learn the skill – each language or script comes with its own unique patterns and rules – but how does our brain learn to read?
In the social sciences, the majority of researchers do not see a problem. There is a widespread view that the brain is a completely plastic organ, capable of absorbing any form of culture. Yet recent findings from brain imaging studies and neuropsychology throw new light on the cerebral organisation of the reading circuits. The findings refute this simplistic model of a brain that merely absorbs everything from its cultural environment. And they suggest that the architecture of our brain is limited by strong genetic constraints, though it retains a fringe of flexibility.
Advertisement
My suggestion, then, is that learning to read, and other forms of cultural learning, are only possible if this built-in flexibility can be used to divert brain circuits from their previous uses. The brain is predisposed to develop only in certain ways. In effect, we are able to learn to read because the primate visual system evolved to do a different job that was sufficiently similar to allow it to be “recycled” into a reading machine.
Functional magnetic resonance imaging of the brain reveals a vast network of cortical areas underpinning the different stages of reading. There are around a dozen regions involved, which are spread all over the brain (see Diagram) and are only beginning to be charted. Together they seem to mediate visual recognition, extract meaning, integrate each word into a phrase and allow us to pronounce them.
For now, however, I will concentrate on just one tiny region, responsible for the earliest stages of reading. We call it the visual word form region. It lies on the left side of the brain within a strip of cortex that takes part in our object-recognition pathway.
We know that this region plays a particular role in the visual stages of reading. Firstly, it only responds to written words, not to spoken ones. Also it produces the same amount of activity whether we see real words or “pseudowords” – words such as “gub” that are pronounceable, follow the phonetic rules of the language, but are not found in the dictionary. In short, the region is interested only in visual form and not meaning.
Indeed, damage to this region produces a peculiar syndrome, known as “pure alexia” (derived from Greek, and meaning “inability to read”). People with this condition are unable to read words at normal speed, though they may be able, with some difficulty, to decipher the word letter by letter, often having to trace them with a finger. Why is this syndrome called a “pure” inability to read? There are several reasons: patients remain able to write words, yet unable to read them back; they have no particular difficulty in understanding or repeating spoken words; and other forms of visual recognition, such as identifying faces or objects, are often largely preserved. Clearly, this brain region plays a rather specialised role in the visual identification of words.
When my colleague Laurent Cohen and I started to image this area in different people, we were amazed at how consistently it is located within the brain. We found it in the same position – to within 5 millimetres or so – in all individuals. Whatever script a person had learned to use, the same region would respond most strongly to chains of characters forming a real or a plausible word.
On a strictly visual level, there is no simple rule you can use to define a word. It is only the cultural history of the individual that determines what should be seen as words and what should not. For French speakers, French words work best. For Japanese subjects, it is the characters of kanji and kana that produce maximal activation. So the reproducibility of the visual word form region across cultures tells us something very interesting about reading. Even though we can’t possibly be born with a specialised reading circuit, we all end up using exactly the same part of the brain.
Of course, this specialisation has to develop. Although we do not yet know what the region does before we learn to read, experiments carried out by Bennett and Sally Shaywitz from Yale University show that in children, the activation of the visual word form region increases progressively with expertise in reading. It is not enough simply to know how to read. Rather, the activation of this region seems to betray a special form of expert reading, in which the whole group of letters in a word is perceived at once, meaning longer words do not take longer to read than shorter words. In adults with dyslexia, who never attain great ease in recognising the visual form of words, the region’s activity never achieves the usual adult levels. This reduced response is most likely the consequence of difficulties in learning to read, rather than the cause.
But exactly how does this region work? We presume that the region is capable of analysing the series of letters that make up words, and then supplying other cortical regions with some sort of representation of their identity and order. Our experiments show that it extracts what we call an invariant visual representation of a word. It somehow picks out and encodes the identity or essence of a word, and throws away all the irrelevant visual detail.
One important form of invariance is spatial. The visual word form region is the first region in the visual processing stream that is not interested in position. It responds identically to words presented on the left or the right of the visual field, which implies some clever wiring. Words presented to our left are analysed initially by the right visual hemisphere, and words on the right by the left side of our brain. The visual word form area, which is on the left side, must collect visual signals from both hemispheres. And it all happens in less than a fifth of a second. By measuring the electrical activity in the brain by EEG, we have been able to see this happening. Around 150 milliseconds after presenting a word to the right or left of a screen, electrical activity appears on the scalp on the opposite side. At around 200 milliseconds, however, the electrical activity converges on the left, regardless of where the word appeared.
Other aspects of handling invariance concern the font of the letters and the case in which the letters are printed. We can recognise the same word written in lower case or capitals, and in many different fonts. The visual word form region seems to be responsible for extracting the word’s invariant information content, whatever the word looks like. We know this because the region shows reduced activation when a word appears for the second time in succession, even when the case is changed (“RAGE” followed by “rage”, say). It is capable of recognising the form of the word, in spite of the superficial change in letter shapes. This ability must be the result of cultural learning, because there really is no greater visual similarity between “A” and “a” than between “A” and “e”. The knack of detecting the sameness of “a”, “a” and “A” cannot be innate.
But if the region didn’t evolve for reading, what was it born to do? In primates, the homologue of this region seems entirely dedicated to visual recognition. It makes up part of the ventral visual pathway, which determines “what” something is, as opposed to the dorsal pathway, which determines “where”. In humans too, the ventral visual pathway responds to all types of visual stimuli – faces, objects and places as well as words. Even in the region of visual cortex that responds most strongly to words, there is also a strong response to pictures or drawings of faces and objects. So it seems clear that the visual word form region has developed from within cortical tissue that has a more general job of recognising visual shapes.
In primates, neurons from the homologue of the visual word form region respond to elaborate forms. Each neuron is specialised, and responds selectively to certain objects. Crucially, its responses are often invariant across a large range of object sizes and positions. One neuron might be able, for example, to respond to the sight of the head of a cat, whether near or far away, turned to the right or left. Certain neurons respond similarly to very different views of the same object – the profile and front view of a face, for example.
The circuits from which the visual word form region has developed therefore appear particularly well pre-adapted for the visual identification of objects, whatever their position, size or viewpoint. This adaptation did not evolve expressly for reading, of course, but its operation is quite close to what is needed for identifying words regardless of their position, size and case.
Experiments on primates by Keiji Sakata and Manabu Tanifuji from the Riken Centre in Tokyo tell us a bit more about the region. They asked what were the most basic components of an image that would still make a neuron respond. When it responded to the image of a cat’s head, say, they progressively simplified that image until they had the minimal features needed to evoke a response. For one neuron it might be the shape of the head and the ears, and for another the presence of two white discs on a black background. The whole area seems to be a mosaic of such “element detectors”.
Most interestingly, some of these minimal elements that the primates responded to resemble our letters. Some neurons, for instance, respond optimally to two discs one above the other, rather like a number 8, others to the shape of a T, others to an asterisk. These elementary forms make up a repertoire of shapes that can be combined to form the immense variety of object shapes found naturally in our visual world. The way this visual region responds to written characters seems to be no accident. Not only can neurons in this region learn to respond to new shapes, such as those of letters, but some of those primitive shapes are already a fundamental part of our visual processing system and were simply discovered and reused as our ancestors invented the alphabet.
A final mystery remains. Why is the position of the visual word form area in the brain so consistent? Why do words always trigger the same small subregion of the visual system? Brain imaging has shown that this consistency is also true of other categories of objects: the regions of the ventral visual cortex furthest from the centre of the brain respond preferentially to objects and words, whereas regions successively closer to the mid-line have a marked preference for faces, then buildings and finally outdoor scenes.
Recent work by Rafi Malach and his collaborators at the Weizmann Institute in Rehovot, Israel, showed that it comes down to a question of scale. The outer regions respond preferentially to the small details of the image that appear in the centre of our field of view. The mid-line regions, on the other hand, respond to the peripheral visual field – the big picture. This pattern is a major feature of brain organisation that cuts across many brain areas and may well be genetically pre-programmed. It explains why words are analysed within the outer portion of the ventral visual pathway – the part that concentrates on detailed images.
The speculative conclusion that I would like to draw is that when we learn to read, we convert a network of neurons whose initial role was object recognition into a specialised word-recognition system. The brain had neither the ability nor the need to create such a region from scratch. Our brains did not evolve for reading. On the contrary, I suggest that writing systems themselves were subjected to selective pressure and had to evolve within constraints fixed by our primate visual system.
There are signs of this in all cultures, where the development of reading began with pictorial representations. The word “bull” for example, was represented by the outline of a bull’s head. These pictorial figures are immediately recognisable by any primate. Gradually, the characters were refined into more minimalist designs, easier to write but still recognisable by our visual system. We also progressively simplified the system so that shapes came to denote single sounds rather than entire words. Our letter A, for instance, derives from the letter alpha, which itself, after rotation, derives from the bull’s head whose name in ancient Semitic was pronounced “alf” (see Diagram). Such modifications created a system of letters that may seem arbitrary, but was in fact carefully crafted to remain easily learnable.
Other human cultural inventions, such as Arabic numerals, were probably subject to similar neurological constraints, and were possible only because they could invade cerebral regions that were initially performing sufficiently similar functions. This is, in effect, “evolution as tinkering” proposed by François Jacob, or the “exaptation” dear to Stephen Jay Gould – the reuse, over the course of phylogenetic history, of ancient mechanisms for new functions. In the case of cultural development, however, this process can take place over a few weeks or years, because it relies on the development and flexibility of networks of neurons, rather than the slow evolution of the genome. I call this the hypothesis of neural recycling. Each cultural development must find its “ecological niche” in the brain – a circuit or collection of circuits with an appropriate initial role and sufficient flexibility to adapt to the new function.
This theory produces two predictions. First, our genes define a cerebral architecture that limits the range of cultural expressions. Thus, the cultural variations that our species can achieve should not be unlimited. For instance, it should be possible to identify universal patterns common to all writing systems. Second, we should be able to predict how difficult it is to learn a new concept or technique, based on the extent to which the original neuronal function has to change. Perhaps this can begin to explain why some aspects of learning to read are so hard for children. All children, for instance, go through a stage where they fail to distinguish the letters p, q, d and b. This might be explained by the fact that our visual system automatically recognises objects as the same even if they have been rotated, which makes it easy for us to recognise an object from any angle, but which is a disadvantage for reading.
Understanding how the brain adapts to novel cultural inventions may ultimately have important consequences for education. Many neuroscientists share the hope that the new understanding of brain organisation made possible by brain-imaging techniques will throw light on children’s educational difficulties, in maths as well as reading. Ultimately, this may lead to new teaching strategies that are better adapted to the structure of our primate brain.