langbrain banner

Interview with Sydney Lamb

In 1998, Professor Qilong Cheng, now of the Shanghai University of Science and Technology, interviewed Sydney Lamb by e-mail. The interview has been published in Chinese in the Chinese journal Foreign Language Teaching and Research 1999/2, pp. 61-64.

Following is a transcript of the original interview, which occurred in two parts.

Part I
Date: Mon, 22 Jun 1998

Dear Professor Cheng --
  I am glad to have this opportunity to answer your questions.

> 1. Sydney Lamb, as one of the great modern linguists, you have founded Cognitive Stratificational linguistics. Why is your theory called "cognitive stratificational linguistics"?

I first called it stratificational linguistics, beginning in about 1966, to draw attention to the fact that linguistic structure, according to the theory, is stratified; that is, it consists of several subsystems, which I also called stratal systems. Later I realized that this theory was different from others in another very important respect: We were trying to understand the linguistic system as it is represented in the brains of actual people, while other theories were trying to represent grammar as some kind of abstraction based on analysis of linguistic productions but not with any attention to the actual minds or brains of people. Therefore, I thought it would be appropriate to distinguish this theory by labelling it with the term "cognitive".

More recently, since the term 'cognitive' is now being used by other linguists for other theories even though they have not shown how their accounts of linguistic structure are related to the brain, I have started using the term 'neurocognitive linguistics' to distinguish my theory from these other "cognitive" theories.

> 2. What is your view of language?

There are various ways to look at language, not all of them necessarily in conflict with one another. To me the most interesting aspect is the linguistic system of the human brain. If we can understand that system it will help us to understand how thinking works and how our minds work in general, including in their attempts to understand the world and ourselves. Therefore I try to study that system.

> 3. What are the aims of your theory?

The aim is to construct a theory of the linguistic system of the human brain. Such a theory should have two parts: First, an account of how linguistic information is represented in the cerebral hemispheres of the brain; second, an account of how the system operates. The first requirement includes two parts: the large-scale and the small-scale. On the large scale we are interested in the various linguistic subsystems and where they are located in the cortex of the typical person, and how they are interconnected. On the small scale we are interested in the details of how information is represented at the level of neurons and their interconnections.

The second requirement, how the system operates, also has two parts: first, the operations which make it possible for people to produce and understand speech; second, the learning processes which enable the brain to acquire and enrich its linguistic information.

> 4. In some of your books and papers we read that Hjelmslev had substantial influence on your work. Could you say something about it?

The most important idea of Hjelmslev in my thinking is his discovery that linguistic structure is not made up of symbols or objects of any kind but rather of relationships. When I incorporated this discovery into my own developing ideas back in 1964, I was able to understand that the whole linguistic system is a network of relationships. This made it possible, a few years later, to begin exploring the relationships between this abstract relational network of linguistic structure and the network of neurons in the brain which embodies our linguistic knowledge.

> 5. What is your most important criticism of Bloomfield and the neo-Bloomfieldians?

It is that they were too heavily influenced by behavioral psychology, so that they tried to study language in isolation from the brain. This unfortunate attempt left them not only unable to understand the true nature of language, it also kept linguistics isolated from the rest of science.

> 6. There are quite a few papers published in China introducing Systemic Functional Grammar and Generative Grammar. How does your theory differ from them?

First, Systemic Functional Grammar. My theory is related to this one and it has influenced my thinking. The notation system used for relational networks is an adaptation of the network notation developed by Halliday. My theory is also a functional theory of language. The main difference is that Systemic Functional Grammar looks at language primarily in its relation to society and social interaction, whereas neurocognitve linguistics looks at language primarily in its relation to the brain.

Generative grammar is quite different. It is not a functional theory, and it is very heavily influenced by the methods of Bloomfield and the neo-Bloomfieldians. This fact is ironic since Chomsky criticized behavioral psychology and said that he was interested in the mind. Yet his methods, which were largely derived from those of his neo-Bloomfieldian teacher Zellig Harris, maintained the non-functional approach of Bloomfield, and its concentration on the forms or expressions of language, with relative neglect of the relationships between form and function. So although Chomsky was correct in thinking that it is important to consider the mind, he never quite succeeded in actually considering it. Nowhere in his writings do we find any real discoveries about the mind, and nowhere do we find him trying to incorporate any findings of neuroscience into his theories of language. In short, generative grammar neglects both of the requirements for a neurocognitive theory of language that I mentioned in my answer to your third question. And so his kind of linguistics remains isolated from the rest of science, and in fact he even criticises empiricism.

Part II
Date: Thu, 16 Jul 1998

Dear Professor Cheng:
  It is a pleasure to offer answers to these interesting questions.

> 7. You mentioned something about mind\brain in your response to Question 6. Many "tough-minded" scholars may think that mind can be realized by brain. On the other hand, some "soft-minded" scholars may consider mind and brain are separate things. Even some others (say, many AI people) assert that between mind and brain there is a third system, a cognitive information system. Could you please say something about your view on this mind\brain issue?

First I have to say that 'mind' is a non-technical term that has come down to us from a tradition of many centuries' duration, during most of which time people had very little understanding of the anatomy and physiology of cognition. And so we shouldn't expect this term to be very useful, and in fact it is more likely to be confusing and misleading, especially since it means different things to different people.

Therefore, I usually avoid using this term altogether. Instead of taking traditional terms and trying to find valid meanings for them, I prefer to explore phenomena and then assign suitable terms as they become needed for concepts developed during the exploration. In doing so, I have not encountered a need to use the term 'mind' for anything, although I do use terms like 'brain', 'cognitive system', 'information system'. These last two are abstract terms for structures specified in the relational network theory. So they too are not really important. What is important is specifying the structure of the network, and the operations of transmitting activation and building connections and adjusting thresholds.

In my answer to your question 6, I mentioned the term 'mind' only in connection with Chomsky, since he has used this term.

> 8. Could you explain the term "linguistic information system" used in your theory?

Ah, this relates to my answer to the last question. This term is used just to introduce the idea that people must have information that makes it possible for them to learn and use languages; and all of that information together may be considered a system; and so we can give it the name 'linguistic information system'. But then we must specify what it consists of and how it works. That is what the theory is about. It aims to specify the structure and operation of the system, and how it gets developed in children, and how it is implemented by neurological structures and processes.

> 9. Some linguists set rationalism against empiricism, as they emphasize the importance of linguistic universals in language learning. Could I say that in your view they are somewhat complementary?

Yes. I see no conflict between these two modes of exploration.

> 10. In what sense that your linguistic theory is a functional one?

The term 'functional' in linguistics, as I understand it, is used mainly in contrast to the term 'formal' in the first of its two main uses, although some functionalists contrast it, mistakenly in my opinion, with both of these uses. The two uses of 'formal' are (1) emphasis on linguistic expressions (forms) and on relationships that can be found among them, largely without regard to their meanings/functions; (2) emphasis on the use of formal notation. It is because of (1) that formal linguistics tends to place great emphasis on syntax, even though syntax constitutes a relatively minor part of the linguistic information system.

My kind of linguistics is functional in that I am interested in the meanings and other functions of linguistic expressions, and especially since I believe (along with other functionalists) that the only way to understand the relationships of linguistic expressions to other linguistic expressions is by means of understanding their meanings/functions. On the other hand, unlike some functionalists, I have no objection to the use of formal notation where it can be useful.

> 11. Can I say as a cognitive model, your relational network can account for the social aspects of language, but in what way those aspects are probably accounted for?

Yes, the relational network, as a cognitive model, should in principle account for all of the knowledge of a typical person -- we have to say 'in principle' since that knowledge is too extensive to ever be described in detail. Now, that knowledge includes knowing things like meanings and discourse functions of lexemes as well as the social conventions and structures of the community, as well as the ways in which different lexemes and different styles of discourse and different concepts influence social relationships and interpretations. In other words, social interaction in general depends upon the cognitive systems of the people who are interacting and would not be possible without the knowledge of those people. For example, each person in a conversation maintains an internal model of the other persons involved and makes predictions about how they are likely to react to something that might be said. And so we need the cognitive orientation in order to understand why and how the social functions of language operate.

> 12. What role a corpus may play in your theoretical framework?

First we may recognize the difference between analytical linguistics and neurocognitive linguistics. These two approaches are complementary. Analytical linguistics is concerned mainly with analyzing speech and texts, and studying the occurrence and uses of lexemes and other linguistic units; also with writing grammars and dictionaries. Neurocognitive linguistics is concerned mainly with trying to understand the structure and operation of the neurocognitive linguistic system. Neurocognitive linguistics must draw upon and take account of the results of analytical linguistics, while also meeting the requirements of (1) operational plausibility, (2) developmental plausibility, and (3) neurological plausibility.

Now, the use of a corpus is very valuable, one might say essential, in analytical linguistics. Therefore, we consider the corpus to be very valuable also for neurocognitive linguistics, since we need and use the findings of analytical linguistics. On the other hand, the actual task of analyzing the corpus belongs to analytical linguistics but not to neurocognitive linguistics.

> 13. Could I say that in your framework conceptual system is separate from the sememic stratum? If so, is it proper to say that discourse should be treated firstly in the conceptual system?

The term 'sememic' has had somewhat different meanings at different stages of development of the theory. According to the current definition, the conceptual system is part of the sememic system, but the sememic system also includes other kinds of information, such as perceptual. It includes all of the knowledge and information that the linguistic system has access to, and that is just about all the information that a person can be aware of. In an earlier version of the theory, we used the term 'gnostemic' for what I am now calling 'sememic'. We can say that the sememic system has several subsystems. It is very complex.

Discourse phenomena, which have been considered much more by Halliday and the Systemic-Functional school than by us, involve both sememic amd lexicogrammatical subsystems.

> 14. Could "pragmatic principles" (such as CP, PP, etc.) much discussed also be accounted for in the conceptual system?

These too involve what we now call the sememic system. There is some evidence that pragmatic phenomena are processed mainly in the right hemisphere.

> 15. Can I say that in your theory the relational notation system is very important because it represents some important neurocognitive aspects of language?

The relational network notation was devised first for understanding relationships among linguistic units, especially functional relationships. It became apparent that the entire linguistic structure was a system of relationships, containing no symbols or other objects like words, phonemes, morphemes as these are usually understood.

Now, if the entire system is a network of relationships, and if the system is in the brain, then we would like to know how the brain implements the network structure. It is encouraging to learn that the brain is itself a network of neurons, and it is also encouraging to find that some of the elementary components of the relational network have resemblance to properties of neurons. But these encouraging observations do not in themselves justify any direct or simple translation of network structures into neural structures. They only encourage us to explore what the relationship might be. Such exploration has provided a hypothesis of that the basic modules of relational networks, which are called 'nections', may be implemented in the brain by cortical columns. A cortical column typically consists of around 100 neurons.

> 16. Did you develop a kind of device to make it possible for the relational network to learn a language and to acquire more knowledge?

I have a hypothesis about this, described sketchily in some recent publications. We now need for someone to build a computer program to test it. Actually, two of my students have built programs for this purpose, but the results have been inconclusive. Now we need a better program.

> 17. Could you tell us the important achievements in the development and application of your linguistic theory?

One important achievement, I think, is the demonstration that the linguistic system is made up entirely of relationships. This finding was anticipated by Saussure and Hjelmslev, but they only proposed it as a point of view, never demonstrated that it could actually work.

Another is the bridge we are building between linguistics and neuroscience. If successful, this achievement will allow linguistics to be integrated with the rest of science so that it will no longer be isolated from science as it has been up to now.

Another is our theory of how the brain represents, uses, and learns information. If correct, this constitutes a solution to what many neuroscientists consider one of the most important problems of neuroscience.

We also propose explanations for many phenomena that linguists and psychologists have found puzzling, such as the ability of people to form new sentences they have never heard before (an ability described but not explained by generative grammar), and the ability of children to produce, by analogy, word forms they presumably have not heard since they are not part of adult language, like 'brang' as the past tense of 'bring' (compare 'sing' and 'sang', 'ring' and 'rang').

> 18. As a conclusion, I would like to ask you: what you think are the important topics that the neurocognitive linguistics should work on?

One great present need is for a computer program to test the theory -- not only for testing the learning hypothesis but also hypotheses of details of network structure and how activation spreads through the network.

Another important opportunity is for research on correlating findings from brain imaging with our hypotheses, and conducting new experiments in brain imaging to test some of our hypotheses.

A third area lies more within the field of linguistics as traditionally practised. We need to have more work done on showing how the relational network system accounts for and explains phenomena that have been analyzed using other linguistic frameworks such as Halliday's Systemic-Functional Grammar.

> Thank you very much for your great assistance and your enlightening thinking!

You are welcome!


Go to Menu (for non-frame version.)

This page was last modified on 24 April 2000.

© 2000 Rice University. This document, or any portion hereof, may be used for non-commercial informational purposes only. Any copy of this document, or portion hereof, must include the copyright notice ( in its entirety.