Do speakers of different languages see the world in different ways? Mark Pagel reports on whether language structures the brain.

Modern humans are in many ways very mediocre apes. We are not so big, nor particularly strong; we are not unusually fleet of foot, nor very agile. We are fussy about what we eat and how we sleep. Just go to a zoo and ask yourself whether you would like to mix it up with one of the gorillas, chimpanzees or orang-utans. You probably should not: even the rather small baboons (a mere monkey) are likely to be several times more powerful than you are, much quicker and embarrassingly more agile.

If ever embarrassed by a physically superior ape, however, confront it with your language. Language is what sets humans apart, it is truly our quintessence. No other species' attempts at communication even come close to our language abilities. And our facility with language is acquired almost effortlessly, so much so that it has been dubbed an instinct. A typical English-speaking adolescent knows perhaps 60,000 words. These include the conjugations of 180 or so irregular verbs that cannot be obtained from a simple rule. All told, a child learns more than ten words a day from birth, or one word every two hours.

Most of the learning happens without explicit training (although I am working hard to get my one-year-old to reserve the word "daddy" just for me). Indeed, preventing humans from using language is difficult: linguists have even observed groups of young children spontaneously create a fully fledged language. Contrast this with our evolutionary brothers the chimpanzees, who only after almost continual Skinnerian harassment will agonisingly gain some facility with a few words.

³ÉÈËVRÊÓƵ

Every known human cultural group has a language and our capacity for producing languages seems boundless. There may be as many as 7,000 different languages spoken on earth. Probably many more languages co-existed in the recent past. Possibly as many as 12-15,000 different languages were spoken at the peak of linguistic diversity. This may have occurred about 10,000 years ago, a time that corresponds to the development of agriculture. The rapid success of agriculturists in peopling the globe since has homogenised language to some extent.

Even the figure of 12-15,000 seems tongue-tied compared with estimates of the numbers of languages ever spoken since humans began speaking. Many linguists and anthropologists believe two sister languages that derive from a common ancestor acquire differences at a rate that makes them mutually unintelligible after about 500 years. Using simple mathematics, the figure of 12-15,000 languages at the peak of diversity, and the assumption about how rapidly languages diverge, I have estimated the number of languages ever spoken. These range from as low as around 40,000 to as high as about 600,000. A figure of between 140,000 and 280,000 different languages having ever been heard on earth represents a best guess if we believe - as do many anthropologists and palaeontologists - that humans began talking perhaps 100,000 years ago.

³ÉÈËVRÊÓƵ

Now this remarkable Babel of diversity is crumbling. Linguists estimate that perhaps 3,000 of the languages currently spoken will not survive the next century. The reasons are obvious. A few languages, some owing to trade, others to colonialism and historical accident, are rapidly achieving linguistic hegemony. Where English is spoken, for example, typically between 80 and 90 per cent of the native languages have been lost. In Russia some 70 per cent of the indigenous languages are moribund. As a result, 30 or so languages per year are seeing their last speakers die, or are no longer being taught to the young. Some linguists estimate that as few as 500 languages will survive the 21st century. The linguistic landscape is rapidly coming to resemble a desert.

Is this a crisis? Should attempts be made to preserve linguistic diversity? The issue is fraught with social implications, but two sorts of answer might be offered. One is that linguistic diversity ought to be preserved because it is useful. We can draw upon words from other languages to enrich our poetry, literature and speech. It also has serious scientific uses. It can be used alongside genetic information to trace human migrations and to establish the likely genealogies of cultures. Differences among pairs of languages in common vocabulary terms can be used to estimate when they diverged. And linguistic differences can explain why groups of people who live close together none the less are genetically quite distinct - humans, it seems, procreate with those to whom they can speak. Valuable as these uses of linguistic diversity are, they do not require the speakers of the languages be preserved. We do not need native Latin speakers in our midst for us to sprinkle our prose with such overused utterances as ceteris paribus or mutatis mutandis. It is sufficient that the language is preserved in texts. So, to ask the second sort of question of linguistic diversity, ought we attempt to preserve not only the languages but preserve them as living languages with real speakers?

One perspective on this question was given earlier this century by two American linguists, Benjamin Whorf and Edward Sapir, who advanced the idea that "language structures the mind". Sapir said that "we see and otherwise experience very largely as we do because the language habits of our community predispose certain choices of interpretation". This has since become known as the Sapir-Whorf hypothesis of linguistic relativity. In its strong form it equates an individual's language with his or her thought and denies the possibility of translation between languages. Among the more intriguing examples are the American Hopi Indians who, according to Whorf, do not have a concept of time anything like that of a European and their language has no grammatical forms for the past, present or future. All of us have also heard the claims made of tribal societies that lack colour words and are assumed not to see those colours, or of the Inuit who with their 30 or so words for "snow" see in it things the rest of us do not. The hypothesis of linguistic relativity has since its inception been dogged by careful anthropological and psychological research. The conclusions from this research are that Sapir and Whorf got their facts wrong. The Hopi do have a concept of time like Europeans, tribes lacking the same colour words as ours still see those colours but simply label them differently, and the Eskimos apparently have two words for "snow". The flaw in Sapir and Whorf's thinking was in equating language with thought, when in fact an individual's language may only be the particular voice in which his or her thoughts are expressed. Speakers of languages lacking certain colour terms can none the less learn to make discriminations among those colours. The way we see the world (with our eyes and brains) determines what discriminations we can make; the language in which we express those discriminations does not. A real world exists independently of language.

But is there a more subtle sense in which Sapir and Whorf may have identified -even if unwittingly - a link between human consciousness and language? Some linguists believe that human infants spontaneously produce all of the sounds found in all of the world's languages. It may not be that every baby produces every sound, but that it is capable of doing so. This should not be surprising. The world's languages necessarily spring from the set of sounds that humans can produce, and words are with few exceptions arbitrary associations of sounds with meanings.

The process of learning a language reinforces some kinds of sound production but not others. Unreinforced sounds disappear over time. As adults we lose the capacity to produce some kinds of sound and, perhaps more pertinent, we lose the ability to discriminate between them acoustically. Japanese infants can discriminate between "la" and "ra" as well as any other infant, but Japanese adults cannot. The loss of the ability to discriminate certain sounds may correspond to impoverishing certain sets of synaptic connections in the brain or to building up other inhibitory synaptic connections, or both. The upshot of this is that brains of Japanese-speaking adults differ from those of non-Japanese-speaking adults, and do so at a physiological level. Language differences predispose micro-anatomical differences among their speakers.

³ÉÈËVRÊÓƵ

How far does this notion extend? Might it also be true that the discriminations one is taught to make about one's world (independently of the language in which they are made) also influence the structure of the brain? Neural nets provide a convenient way to investigate this. A neural net is an artificial perceptual device implemented as a set of instructions in a computer. In a very loose sense neural nets are analogues to our own brains, at least in the following ways: they have the capacity to learn by repeated training, they can discriminate between objects and they can make judgements or decisions. They also have the capacity to generalise. One could train a neural net to recognise cats, and then present it with a new stimulus and ask it what it sees. Neural nets can be queried to reveal their conceptual categories.

All humans very likely share a set of innate concepts about reality, concepts such as "space", "time", "object" and "causation". This makes good evolutionary sense because these elements will be present in everyone's world and acting on them correctly will probably be important. But it is unlikely we have innate concepts for everything. Much more sensible would be to have an adaptable brain. Humans probably do not need innate concepts for "baseball cap" or "bangers and mash". One philosophical view of concept learning is that non-innate concepts such as "cat", "table", "toy" and so on are acquired by normative instruction. The rules that define the concepts are shared by groups in societies. One's concept of cat is, under this view, shaped and perhaps in a sense not too different from that of the neural nets, created by the repeated trials of learning, reinforcement and discrimination that other members of one's normative group have come to share.

To investigate how concept learning might influence perceptions, a colleague and I trained two neural nets to make discriminations about the same underlying world. One was trained to see the world as falling into only large and small categories. The second was trained to sort the world into six size categories. What we had in mind with this research was a rough analogy with the differences among language communities in the ways they divide the world into categories. One net corresponds to a language community impoverished in this particular dimension, the other less so. Once we had trained our nets to see the world reliably in terms of its set of categories, we asked them to respond to a new set of stimuli neither had seen before. We found they "saw" the new stimuli differently, with the impoverished net seeing fewer dimensions. They had acquired different perceptions of the same world owing to their previous normative training.

³ÉÈËVRÊÓƵ

Our neural nets are a long way from real human brains but suggest, along with the physiological evidence, the following possibility: to the extent that different languages are associated with different normative rules in their respective cultures, the speakers of those languages have brains that differ and see their worlds in subtly different ways. Note that it is not the language itself that influences thought but the habits of the community in which that language is found. Even though we all ultimately think in the same way, the habits of our particular linguistic community will have taught each of us to define the myriad non-innate concepts and their interconnections in ways that differ even if slightly from those of other groups.

This view does not rule out the possibility that holders of different concepts can compare those concepts and even reach agreements about where they differ. But, just as a Japanese adult cannot hear the difference between "ra" and "la", holders of different normative concepts may never quite appreciate the mind of their interlocutor even if they appreciate that differences exist. Thus, we have good reason to believe language does structure the brain and mind, and may do so at very deep levels. But this does not lead us inexorably down the path of relativism.

What then of the question of preserving linguistic diversity? If French speakers' minds differ from those of German speakers which differ from those of speakers of North Fresian then we may wish to preserve these ways of mind. They are living examples that demonstrate to the sciences of the linguistics and the mind the adaptability of the human brain. This also gives us a firm footing for believing what many have long known instinctively: that a culture's literature, poetry, and song are deeply interwoven into the minds, indeed even into the brains, of its inhabitants.

Mark Pagel is in the zoology department, Oxford University.

³ÉÈËVRÊÓƵ