The Taos
Discussion
Monday, September 12, 2005
http://www.childrenofthecode.org/interviews/deacon.htm
The Co-Evolution of Language and the Brain
Remarks by Terrence Deacon from the interview cited above:
Most surprisingly is that brains are not, it's hard to say this but, brains are *not* designed the way we would design any machine. They are not built the way we would build a machine. We don’t take the parts and put them together to build a whole. In fact, what happens is just the other way around. The whole starts out, it’s just undifferentiated. The parts aren’t distinguished from each other and they become more and more different from each other. The system becomes more and more complicated. The problem is the process is very indirect.
It is not like building something from a plan. It’s a very indirect process. It became very clear that this process is very much what we would call today, self organization. A lot of the information that goes into building brains is not actually there in the genes. It’s sort of cooked up or whipped up on the fly as brains develop. And so if one is to explain how a very complicated organ like the brain actually evolved, changed it’s function to be able to do something like language, one has to understand it through this very complicated prism of self organization and a kind of mini-evolution process that goes on as brains develop in which cells essentially compete with each other for nutrients. Some of them persist and some of them don’t. Some lineages go on to produce vast structures in the brain. Other lineages get eliminated as we develop, in some ways just like a selection process in evolution.
Like evolution, it’s a way of creating information on the fly - a sort of sampling the world around you, in this case the body as well as the external world and adjusting yourself to it. In some ways the brain has that character, too. The complication of all of this has made me realize that the problem of language is a problem that’s not going to be ultimately solved by a kind of engineering model, by a kind of reverse engineering. We are going to have to think, in a sense, like biology thinks, like embryology works in this kind of self organizing, self evolving-like logic to get back at what actually causes things to occur and develop in complex ways.
I think this is the case with language itself. As it’s passed from generation to generation it has a kind of self organizing and evolution-like character to it. That is also part of where the structure comes from, where it picks it up on the fly, so to speak. We have to work that into our stories. That is really the impetus for the book I wrote. The Symbolic Species captures the essence of that idea in its subtitle, The Co-evolution of Language in the Brain. Ultimately, my argument was that language itself was part of the process that was responsible for the evolution of the brain. I mean that in the following sense.
Imagine the evolution of beavers. Beavers are aquatic animals today but they are aquatic because of what beavers in the past have done. That is, beavers have created their own world to some extent. They’ve created an aquatic world by building dams and blocking up streams and turning them into small lakes. Beavers’ bodies have evolved in adaptation to the world that beavers created. It’s a kind of complex ratcheting effect in which what you do changes the environment that produces the selection on your body.
I think language is, in a sense, our beaver dam. Language has changed the environments in which brains have evolved. That changes the picture radically because now one can look at the brain, so to speak, with an inside out perspective of the problem and ask the question, “What’s different about human brains and how might that difference tell us something about the forces that shaped it?” If those forces include language then the brain itself is a wonderful signature, a wonderful trace for the forces that helped it evolve in this complicated interaction. The title, The Symbolic Species, captures this notion that we are a species that in part has been shaped by symbols, in part shaped by what we do. Therefore our brain is going to be very different in some regards than other species' brains in ways that are uniquely human...
The truth is we don’t know much more now about intelligence and what its relationship is to whole brains than we did a century ago. We don’t really know how whole brains work yet. We don’t have a sort of general theory that everybody agrees upon, or even is close to agreeing upon....
On the other hand, this major change, if it’s not just blowing the thing up larger and adding a whole lot more to everything, if it is somehow distributed in an interesting way, that distribution of how the same parts got changed and enlarged with respect to each other might tell us something about what that difference is....
I began to look at quantitative issues and with a developmental perspective, how changes in size might change the circuits, that is, how circuits might respond to this. I think for me that has been the biggest source of insights, the biggest source of clues. They’re not answers, they’re clues. That is, I think that the connections are different because the brain is bigger and because not all parts expanded at the same rate. I think that’s the first inside out clue to what is important about this language difference. What was the change in the hardware that supported this new kind of communication and cognition?...
I think clearly there was a co-evolutionary process in which language affected brains. The issue is that language changes at a much faster rate than brains. Clearly the language we have today and the kinds of languages that are all over the world were not always the way they are. We undoubtedly passed through not maybe one stage of what you might call a proto language but probably many proto languages, many forms of this linguistic symbolic communication system over the course of our evolution, all of them leaving somewhat of a trace....
I want to say one other thing about this correspondence between language and the brain. An engineer might look for something that maps to verbs, maps to nouns, maps to the past tense, all of the various features that linguists tell us that language is broken up into. But language, of course, is broken up according to a logic that has to do with communication, has to do with symbols, has to do with the constraints we have on interacting, perhaps, with speech sounds, perhaps with gesture. The logic of the brain is a very old logic and a very conserved logic. It’s the logic of embryology. It’s the logic of self organization. In fact, it’s the logic that has been shared with a common ancestor that goes back well before vertebrates....
There’s unlikely to be a nice, neat direct map between what we see in the external world of language and what we see inside brains. In fact, the map may be very, very confused and very, very different inside the brain, that is, how the brain does what we see externally in language....
So, in one sense, we should expect the brain to reflect features of language, the features that it’s created. For example, in the metaphor that I used about brains and beaver dams: beaver bodies don’t necessarily reflect beaver dams, they reflect the environment that was created and the demands of that environment. Those demands are not necessarily directly represented in the environment. Beavers are aquatic. Beavers have all kinds of special adaptations for living in water. It’s not easy to predict from what’s produced in the world what’s going to happen inside, so to speak, in the body. I think we can say the same about this troubling and complex dynamic in which language evolution and brain evolution are working in tandem interfering with each other, confusing each other, shaping each other. That dynamic is probably one that is not going to be easy to predict....
One can say that there’s going to be some general things that have to go along with linguistic processing. The symbol processing problem, the automatization, (that is the speeding up of the automatic running of syntax and of analysis), the mnemonic problems, the short term memory problems associated with it - these are things that are going to be generally there. And of course, typically the constraints of producing and hearing sound, or producing it visually and manually and interpreting it visually. All of those are things that we could predict so in one sense we can use general characters of language to make predictions. It's probably not going to be very successful if we try to use specific characters of language to make our predictions about brains....
Language clearly forces us to do something that for other species is unnatural and it is that unnaturalness that’s probably the key story. One might want to ask, “What is so different about language? What are the aspects of language that are so different from what other species do?” To answer the question, what should that imply about how brains have adapted to deal with that?...
I do think that the development of language is the single most distinguishing attribute. Now that doesn’t mean speech, necessarily. It means the support system that’s around language - the symbolic system. Of course our communication is much more than linguistic. Some of it is very old also. But a lot of culture is also linguistic-like in a variety of ways and clearly that has been around a long time.
The ritual, the mythology…simply ways of doing things that are organized conventionally, symbolically - this is a hallmark of our species. We have, in a sense, transformed and even reinterpreted much of our biology through this system. So much of what we do, whether it’s marriage, warfare, or whatever, has been transformed by this tool that has, in a sense, taken over and biased all of our interactions with the world....
I think that there are probably many thresholds that we are talking about and early language-like behavior might look very different. I think clearly early language-like behavior had to involve much less vocalization because the brains that preceded us, mammal brains, are not well suited to organizing sound in precise, discrete and rapidly produced learned sequences.
This is something that mammal brains, in effect, are poorly designed to do precisely because the system we use to produce sound is a system that normally should be running on autopilot so that we can breathe appropriately, so that we don’t choke, and so on and so forth. There’s literally been a change in that circuitry to override those systems in order for language to be possible.
What was language in the past, or what corresponds to language homologically, as they say in biology, the homologues to language might have looked very different. Not a signed language, necessarily, but some very complicated combination of modalities. This makes it very hard to predict what the effects might have been and how those effects were layered upon layer to produce new kinds of languages, new kinds of brains....
For me, one of the things I think is really exciting about languages is this aspect of how it reflexively changes the way we think. I think that’s one of the most amazing things about being a human being.
When people talk about memory they usually talk about two kinds of memory. Episodic memory, remembering a particular thing that happened, a sort of one off event in your past; and then something called procedural memory, the kind of memory that’s involved in skill learning.
The problem with one-off memory is that there’s so many one-off events, how do you find them again in your memory? The problem with procedural memory is you’ve got to be in the same situation. I know how to ride a bike when you put me on a bike, but if you ask me to tell you exactly which leg I put on first, how do I start out, how do I stay up, I couldn’t tell you because it’s, in a sense, embedded in its context.
Language is unique in the following sense: that it uses a procedural memory system. Most of what I say is a skill. Most of my production of the sounds, the processing of the syntax of it, the construction of the sentence, is a skill that I don’t even have to think of. It’s like riding a bicycle. I don’t even have a clue of how I do it.
On the other hand I can use this procedural memory system because of the symbols that it contains, the meanings and the web of meanings that it has access to, that are also relatively automatic, to access this huge history of my episodes of life so that in one sense it’s using one kind of memory to organize the other kind of memory in a way that other species won’t have access to without this.
The result is we can construct narratives in which we link together these millions and millions of episodes in our life in which you can ask me what happened last month on a particular day and if I can think through the days of the week and the things I was doing when, I can slowly zero in on exactly what that episodic memory is and maybe even relive it in some sense....
If language is new, if language is only a hundred thousand year old, or even less, a fifty or sixty thousand year old kind of process, then we should expect that it has had little effect on human brains - that whatever tweaks were used were, in a sense, clumsy kludges to make the thing work. We shouldn’t expect that it’s easy, that it’s fluid and runs without difficulty.
On the other hand, if language has been around for a good deal of our evolutionary past, say a few million years, or even a million years, that’s adequate time for it to have structured and reshaped the brain to be better satisficed to the problem of processing and using language in real time.
Similarly, language will have adapted. We will have adapted this language process to be better fit to our own constraints as we go along. The two will, in a sense, be in tandem, converging towards each other.
The question one has to ask from the present time looking back is: Are we well adapted to this? Are we really a symbolic species or is this just a kludge added on top of a primate brain not well designed for this? I think the evidence for the age of this probably can be best picked out by comparing to something quite relevant to what we are discussing, and that is writing and reading.
Writing and reading occurred recently. Therefore, whatever we use to do this was a kludge, that is, an engineering patchwork that just sort of barely gets by. We are not well designed to do so and as a result a lot of people have difficulty acquiring reading and writing. There are many kinds of difficulty and many people will never be able to read or write because their brain hasn’t had a chance, so to speak, to keep up with this process.
If language itself were like that we should expect to find those kinds of problems with our ability to acquire language. In fact, I think the surprising feature is that not only is language incredibly complex, it is acquired rapidly in our development, even when children are quite young. Reading, of course, takes quite a while to acquire; you have to wait until your brain has matured a ways.
But even more interestingly, people with significant brain damage at birth nevertheless do better at language than any other species on the earth, under intense training. That suggests to us that the system has even been over designed a little bit to be prepared with a kind of redundant system in case things go wrong.
Young children with the whole left hemisphere damaged or removed at a very early age can nevertheless acquire language with half a brain, so to speak. The wrong half, for that matter. That again suggests the system is somehow over designed. That can only occur if there’s been a long evolutionary history, I think, in which language has played a significant role, suggesting to me that at least, language-like processes are not new, are not recent....
The few times that language has turned into writing, become externalized in the world, almost all began with a kind of pictorial mode. The problem with pictures is that, of course, they’re not like sound, not like language. The advantage of pictures is that they’re much closer to what they represent, in some sense. We can even look at some of these ancient writing systems and make some educated guesses and it turns out usually they’re wrong for interesting reasons. One of those reasons is that almost as soon as picture-like writing came about people began to use the pictures to also represent sounds, the name of things, a rebus, like you find in a coke bottle cap, in which you’re trying to figure out the meaning of something by using a picture to represent a sound also.
Most of the pictorial writing systems very quickly took on sound iconism, that is sound likeness as well. The word for that thing has a sound and I can use that sound element to put together with other sound elements. What happened was that kind of system got co-opted very quickly. Now in one sense you can say, well this is a disadvantage because the interpretation of the meaning might be easier if you could see the picture, if you got something about the picture of it right away, because vision is one of the ways we remember objects and relationship. On the other hand, it’s broken up into bits and pieces, it’s separated. And of course it’s quite distant from speech. Speech has recoded this in a whole different way. What seems to have happened in most of the world’s written forms, not all of them, but even those that still have a bit of pictorial nature to them, even have acquired this feature, and that is, that they have become representations of the speech stream itself, and only secondarily of something in the world.
What that tells us is that these have become adapted to us. That is, our constraints - the things that we do well automatically. They're not any longer adapted to what we do easily, but are not so flexible and facile as language, which has this kind of evolutionarily instantiated support system that makes it relatively effortless. So something that is more encoded than a picture is actually easier for communication in this regard. It’s still a kludge, it’s still an external invention for which brains weren’t well designed. I think the evolution of writing systems is a hint to what the brain design is. They’re not perfect, they’re also satisfices. There are also other factors that are influencing them but they are evolving towards a better fit to what we do automatically and easily....
I think one of the really interesting things about writing is how long it takes before we’re mature enough, before our brain is ready to acquire it and to do it easily because the learning process there is clumsy. With respect to speech or even sign languages, the learning process is already set up. From a very early age, in fact, we’re able to do things that, if one was just to look at the complexity of the task, would seem infinitely more complex than interpreting these letters and figuring out how letters map to sounds. I think that that’s a real paradox that makes the reading and writing problem stand out....
So, to some extent, the very specialization that we have for acquiring language and acquiring it at an early age amplifies this difference in human brains. That brains come in to the world, so to speak, expecting to be bombarded with language at an early age as part of its organizing principle, and that human brains that don’t get that experience, in effect, are not getting the normal nutritive experience of their expected evolutionary anticipation of this kind of environment that they’re about to fall into....
The analogy that comes to mind is the analogy of the various kinds of computer operating systems that have occurred over the past few decades - one of the things that I think we often times use as a model for language, and is typically the model that many linguists take home. When you’re trying to come up with a rule system to describe a natural language it turns out to be enormously complicated, lots of tweaks and special cases and rules that are very arcane that no one really knows themselves – that one has to discover in the study of language.
As a result, I think many people have looked at language as though it’s this kind of engineered, clumsy interface that we have to acquire. If you look at it like a computer interface written by engineers that has to be used by non-engineers, it looks as though it’s this impossible thing to learn - that it takes a whole lot of work to do it. The result is, it looks as though you have to have special training in advance to do it. Before you get there you have to have gone through the class, you have to have read the manual, and then you can figure it out, or you have to have the crib sheet in your brain to begin with....
The interesting feature of that is that to some extent writing is still like that early computer interface. It requires these complicated learning processes in which we learn the sounds first, then we learn the words. We eventually learn to make these complicated decodings and we probably automate a lot of it so we actually don’t even see the letters much anymore as we read through it. This is why when I look at my own writing I often times don’t catch my own typos. I’ve automatically begun to just see things as whole words or, in fact, even whole phrases. I look past them because they’re so easy for me....
One of the things I want to say about that is that I think it’s interesting that in the history of recent linguistics we have modeled language processing on a Turing Machine. The original generative grammar that Noam Chomsky pioneered and that made a radical advance in the field is modeled on the way computers work, that is a simple set of instructions and rules to generate something. Almost certainly that’s radically different than the way our brains work. The question is, can we use that descriptive tool, that engineering model, to go back and try to figure out how brains are doing this? Or, in fact, does that engineering model actually get in the way of figuring out what the brain’s logic is as opposed to Turing Machine logic?...
Language is a case in which it doesn’t make sense for an individual to have a language. A unique individual language is a non sequitur, a non starter. It’s something that is systemic from the beginning. In the same sense, what we’ve learned about evolution and what we’ve learned about embryological development is they are also systemic. They’re not instructed from single genes that say you must be this shape. They’re the result of the genes setting in motion competitions, growth processes, interactions, communications, and out of those relationships emerge the structures that we see in embryos. Similarly, out of the relationships of ecosystems come the various niches and organism designs....
Even when we learn individual words we don’t, as a sort of caricature would suggest, we don’t hear the word, ask somebody what it means, and know automatically how to use it. We do sometimes ask people what things mean but usually we learn words in context. We learn what words mean by progressively differentiating their meaning, by catching the nuance, the connotation, the way they fit with things. And we know when somebody doesn’t quite understand the meaning of a word, especially a technical word, because they often times use it in a sort of rote way that’s out of context, in which it misses some crucial connections because they haven’t, in fact, differentiated it....
What prompted me to write the book, and prompted a lot of my thinking was not the readings of experts in the field but, in fact, an encounter with an eight year old in one of my son’s classes. I went in as a dad, telling the classroom what dad does for a living to give kids an idea of different occupations and that sort of thing. I talked about my studies of the brain, my interest in language and so on. In that process I said the standard things such as that language is a unique form of communication that only humans do, and we must have a lot of changes in our brain that make it possible, that we’re specialized for this purpose and that we are the only species that does this.
This prompted a young kid to say, ‘Well, but don’t dogs and cats have language? Dog language and cat language and bird language and so on?’ And I said, ‘Well, metaphorically yes, they communicate, but their way of communicating is very different than ours.’ I went on to explain those differences but that didn’t seem to satisfy this questioner who came back again and said, ‘Okay, I get it. So they have simple languages and we have complicated languages.’ I thought for a second and I said, ‘No, it’s not like that. It’s not that they’re simple languages, they’re very different from language. There aren’t things like verbs and nouns, and so on and so forth. Language is really quite different and the way it combines things is quite different. So it’s not really simple, it’s complicated.
For some reason that was a troubling answer as well. The question that came back to me was, ‘Why? Why don’t animals have simple languages?’ I managed to get myself out of that question by sort of confabulating the way an adult does, and said, ‘Well, of course we don’t understand all of those things and what a very complicated question’ – then moving on....
But that question why really bothered me. It bothered me for a simple reason and that is that all of the explanations that we had cooked up for this didn’t seem to answer that simple question. If language was just very complicated and you needed a really good brain with lots of learning devices to acquire it, then animals ought to have simple languages. But in fact, they have something very different than languages, a whole lot more like laughter and sobbing and smiles and frowns, which we also have. A simple language also wouldn’t require you to have a kind of crib sheet in your brain, a language device, a universal grammar that allowed you to say, ‘Oh, it’s one of those. I got that already’ because a simple language wouldn’t need all that support because you wouldn’t have to worry about it.
There wasn’t, in the standard explanations of intelligence, of special language devices, and so on and so forth, any story that answered the reverse question - not why we have language, but why aren’t there simple languages out there? Why don’t animals just slightly less intelligent than us have just a slightly simpler language? And those less intelligent than them have a slightly simpler language than that? Or why aren’t there dozens of languages with different language devices, in different species, that are all like language but very, very different and not inter translatable? We don’t find something like that. That suggested to me that we were on the wrong track....
So the problem of negotiation, especially in a group that has to cooperate, is that we have to find ways of representing the ‘what if’. It’s not that we’ve thought of the ‘what if’, necessarily. But there are alternatives that aren’t there now. What if I’m not here? Will this change the mating strategy? What if I’m off at the kill site and there are other males back here? Can I rely on females to avoid them? Can I rely on that male to respect the fact that I’m the dominant animal? I’m not there to defend my mating access. Will that change things? What can I rely on?
I think what we can rely on, and what we still rely on is what amounts to the gossip and the tattletale side of things. That is, we use other eyes who know that there are agreements, that we have expectations, and that we have come to some sort of agreement to keep things on the up and up. Keep things even, so to speak, so that we can both cooperate and compete in some sort of a balance that doesn’t tear the group apart....