(introspective music) - Good afternoon. I'm William Lester, Professor of Chemistry and chair of the Hitchcock Committee. I'm pleased along with
the graduate council to introduce Dr. Noam Chomsky as part of the Charles M. and Martha
Hitchcock Lecture Series. In appreciation for
their generous bequests, I'd like to tell you something about how the Hitchcock Lectures
came to UC Berkeley. Dr. Charles Hitchcock was
a physician for the army, he moved west during the gold rush and settled in San
Francisco where he opened a private practice that thrived. In 1885, he established a professorship here at Berkeley as an expression of his long-time interest in education. His daughter, Lillie Hitchcock Coit greatly expanded her
father's original gift, making it possible for
us to present a series of public lectures. The Hitchcock Professorship
is one of the most cherished endowments of the
University of California, recognizing the highest distinction of scholarly thought and achievement. Thank you Lillie and Charles, and now a few words about
our speaker, Noam Chomsky. (audience applauds) Dr. Chomsky is no stranger
to the Berkeley campus. In fact, he spoke on this very spot to an overflow crowd several years ago, and ever since then we have
wanted him to come back. So we're honored that he
has accepted our invitation to be a Hitchcock professor. Noam Chomsky is one of the
most influential linguists of our time. He has dominated several
fields in linguistics over the past 40 years. He is credited with changing the focus of linguistics from a concern with methods of classification to a search
for explanatory principles and his numerous writings on
the subject document this. But Noam Chomsky's
influence extends beyond the realm of academia,
he is a forceful critic of US foreign policy and a spokesperson against the abuse of power. Noam Chomsky received
his PhD in linguistics from the University of
Pennsylvania in 1955. He has been a professor at MIT since 1961. He has written numerous
books and has received various honorary degrees,
awards, and honors. I am very pleased to
present Dr. Noam Chomsky. (audience applauds) - Thank you. Well, I ended yesterday by describing the so-called principles
and parameters approach, which gave the first serious conception, to my knowledge, of how the tension between descriptive and
explanatory adequacy might be resolved. In this sense, it's fair
I think to regard it as the first genuine proposal as to what a theory of language might be. Whether that step turns out
to have been right or wrong, it offered the kind of
a liberating perspective which led to an explosion of
highly productive inquiry, and it opened new
problems to investigation, also revitalized the study
of language acquisition along new lines and other
language-related areas of the cognitive sciences. To the extent that outstanding
problems can be solved, not a simple matter, but to the
extent that that's possible, we will have a conception
of the initial conditions that map experience to
the state that's attained by the faculty of language,
to what I called yesterday the internal language, the I-language, that determines the infinite array of sound-meaning
combinations that the speaker of a language has available for use, the potential expressions of the language. These initial conditions include, but are not exhausted by,
the genetically-determined initial state of the language organ. In addition, there's the third factor that I mentioned yesterday,
general principles that are not coded in the initial state, just as properties of protein folding are not coded in the
genome but are available to be exploited in the
growth of the organism. Now there is, I think,
some reason to believe that understanding of language has at last reached the
level that makes it feasible to address these questions, that is, to pursue the Galilean intuition that nature is somehow perfect. In this case asking not
only what the properties of language are, but why language has these properties and not others. That inquiry, in recent
years, has been called the minimalist program. Notice that that program
is theory independent, whatever you think is the
right theory of language you may or may not decide
to pursue these questions. The initial state, it can be disaggregated into elements that have
a principled explanation and others that remain unexplained at this level of analysis, they have to be attributed, therefore,
to something independent, perhaps an evolutionary accident where properties of the
brain that remain unknown and would have to be
investigated along similar lines. The principled elements
of the initial state are the conditions that are imposed on the faculty of language by the systems with which it interacts. These are what are called
interface conditions. If language is to be usable at all, its design must satisfy these conditions. That is, the information that
is in the internal expressions that are generated by the language have to be accessible to
other internal systems, that includes language external
but person internal systems, including sensory-motor
systems and conceptual systems that enter into thought and action. So we can therefore
restate the deeper problem of determining why language
has the peculiar properties that it does, insofar as the
properties of an I-language can be accounted for in
terms of interface conditions and general principles of
computational efficiency and the like, they have
a principled explanation. And we will have verified
the Galilean intuition of perfection of nature in this domain. Well, if one proceeds this far,
you face a challenging task. The task is to examine
every device, principle, idea, technique, that is
employed in characterizing languages and try to
determine to what extent it can be eliminated in
favor of a principled account in terms of general conditions
of computational efficiency and the interface conditions
that the organ must satisfy to function at all. That's been a focus of a good deal of work in the past few years, and there are, by now,
rather plausible accounts in these terms of some basic processes of the computational system of language. These systems that previously had been stipulated as descriptive technology but can now be reduced to principles of computational efficiency. And also more complex properties
of individual languages have, to an interesting
extent, been derived from close examination of
the way principled mechanisms function under minimization of search and other principles of
efficient computation. Furthermore, what appear
to be radical differences among languages have in a
number of interesting cases been reduced to something
close to the optimal conclusion that would means that the
internal computational processes are identical, the
phenomenal differences result not from how the mind computes but rather how the internal
objects that it constructs are related to external events
by sensory-motor systems. So the interesting cases involve
a category of properties, features with no independent meaning that had not really been noticed before, though they're crucial, it appears. And differences in how
the apparent diversity of languages seems to be traceable to a very significant effect,
the slight differences in the way these uninterpretable features are externalized, are used
by the sensory-motor system. It's as if the mind
computes in a fixed way but with varying effects
at the mouth and the ear. The general observation, not this far, but the general point was
a very natural consequence of the principles and parameters approach, it was recognized from the
moment that it came to be formulated with at least some clarity. Actually, I discussed it near here in 1979 at Kant Lectures at Stanford. At that time I referred to classic work, biological work on regulatory mechanisms by Jacob and Monod, two Nobel laureates, quoting their observations
about how slight changes in regulatory mechanisms,
utilizing differently the same structural
information might yield enormous phenomenal differences. Their examples, differences
between a butterfly and a lion or a worm and a whale, basically identical
organisms but with minor changes in how regulatory
mechanisms function, same mechanisms. Similarly, slight changes in parameters left open in the fixed
schematism of language might lead to what appear to
be very different systems. Apologies here for quoting myself in 1979, it was actually in the years that followed that such ideas really
did begin to bear fruit in the study of language. Similarly, in the biological sciences there have been quite dramatic progress along similar lines. Such conclusions would
have greatly pleased Alan Turing I think, he's
known to philosophers and linguists, logicians,
computer scientists, mainly for his work in other areas, but some of his best-known
work was on morphogenesis and the effects of
chemical and physical law on the possible course of evolution. These are matters on which
he took a rather strong stand and that have become a central topic in contemporary biology. One fundamental problem that remains is why there is parametric
variation at all in the computational system, so why isn't there just one language? There are some possible answers
that have been suggested, but the general problem in my
opinion is still mysterious, it's a problem that eludes our grasp, but you can think of possible answers. One might also hope to
approach the question of evolution of language in these terms. It's rather curious to
compare the enormous number of pages devoted to the
evolution of human language with work on evolution
of much simpler systems, say, bee communication. A couple of years ago
asked a biologist friend to do a database search on that topic, and he could come up with very little. Apparently, the subject is
considered too difficult to pursue, although bees are
just a wonderful subject, they only have 800,000 neurons, a brain the size of a grass seed, many different species that
yield rich comparative data, a very brief gestation period, lifespan, no need to obtain consent
forms for experiments, and so on, every conceivable advantage. But, nevertheless, the
problem's considered far too hard to study. The question of how the
human language faculty, it might have evolved,
is vastly more obscure and hard to study. One can imagine various scenarios, a number have been proposed, some of them are maybe suggestive. But some leading evolutionary biologists, Richard Lewontin is the most prominent, they have argued strongly that the problem is entirely beyond reach
in terms of anything remotely like present understanding. He, Lewontin, at least extends that to evolution of human
mental faculties generally. Well, if we, nevertheless,
want to pose the problem of evolution of language, we have to begin with a recognition shared by everyone that the faculty of language
is not a distinct entity, it's not like a box in the human brain with a single location
or a single function, whatever that term is supposed to imply. The faculty of language
surely recruits processes, capacities, physiological mechanisms, that have evolved quite independently, and it could turn out in principle that there's nothing in
the faculty of language that's specific to language, that the faculty is just the specific form of organization of
elements that are recruited to constitute this organ of the body, to borrow Randy Gallistel's term again. If the minimalist approach to language has real prospects of success, and at least my judgment is that it does, we would expect to find
that language crucially involves interface conditions
and computational processes, some of these at least may
have homologous structures in other primates, maybe even beyond. If that's the case, then
the study of language will involve, in part, the
evolution of these elements, the evolution of the way they're organized in the language faculty and whatever may have
evolved independently in the last flick of an
eye in evolutionary time, in human evolution, and
it's extremely short. There is intriguing
work, very recent work, on primate perceptual systems
that could bear on the topic. For example, a rather surprising discovery that untrained tamarin monkeys distinguish among some language types
in ways that are analogous to newborn human infants, in all cases without any experience,
though perhaps they use different cues, that's
under investigation. Additional work is proceeding
along similar lines also investigating what
Harvard primatologist Marc Hauser calls wild
minds in natural settings, also experimental work that he
and his laboratory are doing on a number of topics including
computational capacities of monkeys on language-like problems. It could turn out that
the core computational properties of natural
language are found elsewhere in the animal world, perhaps in what perceptual psychologists call the rules of visual intelligence that are used to create what we see, quoting. Conceivably, you might even find them in systems as remote as insect navigation, that possibility and others
like them can't be excluded and they do suggest avenues of research into the evolution of at
least some of the factors that enter into human language avenues that might be promising. My view, at least, if so,
that would be the first time that the problem of evolution of language becomes a really serious problem. Well, to proceed further
along in any of these areas would require a good deal
of background discussion that I can't take time for here. And, of course, this is where
the topic gets interesting, but I'll put that aside and just restate the guiding intuition of
the minimalist program, basically Galilean, insofar
as this program succeeds we will be able to conclude
that principles of the kind that one finds, say,
principles that hold for, say, sphere packing in early cell division or formation of polyhedral cells of polyhedral surfaces of viruses or stability of body forms and symmetries, maybe optimal wiring of neural systems, just general principles of
physical and mathematical laws, may also hold for an organ
that's a very recent product of evolution and appears
to be a crucial component of whatever it is that's
specific about human beings, certainly at the core of their existence, their thought and their interaction. Although we have to bear in mind that the classical problems
of the theory of mind, the ones that preoccupied
Descartes, Newton, and others, they remain as obscure as ever, and the problem of unification, well, it takes a wholly different form after the Newtonian revolution,
as I mentioned yesterday that problem remains unsolved. The gaps look as unbridgeable today as they did for chemistry and physics not many years ago, 70 years ago. Well, let me return to
the initial question that I began with yesterday, the questions that are sometimes called the problems of intentionality, that is, how does language,
this biological entity, how does it relate to
the rest of the world? One of these is the unification problem, how do accounts of the brain in terms of computational systems relate to others, say in terms of cells? As long as that question is unanswered, there's a crucial explanatory gap in accounts of behavior,
whether it's insect navigation or human language, there's
a gap in explaining, accounting for how computations
eventuate in action. Notice that this is over and above the questions that are not even posed, that is questions of choice of action, so like why does a cricket
turn left instead of right? What about the second
problem of intentionality, that is, the use of language
to talk about the world, as when I say that I read
Darwin's Descent of Man referring to a book. There are parallel problems
on the receptive side, but let's focus attention on the questions as they arise for the speaker. These questions have two aspects, one is how does the sensory-motor system use the instructions provided
by the internal language to carry out gestures? That would be articulatory gestures in the case of spoken language, and second, how does the system of thought use these instructions
to talk about a book or a river or the crisis in Argentina or whatever it may be? It's this second aspect that's held to be particularly vexing, but perhaps one can gain some insight into it by asking about the less contentious interface
with sensory-motor systems. So we now put aside the explanatory gap that holds for all animal behavior and we ask how the phonetic information that's generated by the internal language is accessed and used by
the sensory-motor system for articulation,
alternatively perception, but I'm putting that aside. Let's adopt a conventional but, in fact, non-trivial assumption,
namely that the phonetic information of each expression is encoded in a single internal generated object, what's called its phonetic form or phonetic representation,
here you have to be careful to divorce the notion representation of any connotations drawn
from other uses of the term as in the classical theory of ideas or in discussion of
representational art, for example. These problems have
been studied intensively with hi-tech equipment for half a century and long before that in other ways, but understanding remains quite limited. The problems are not easy, well, maybe everybody's
overlooked an easy solution which works like this. We all agree that the lexical item, any lexical item, say book, has a sound and sounds are perfectly
robust and familiar notions so we can tell without difficulty whether some event is the sound of book or the sound of river. Unproblematically, we
can speak of the various events that are sounds of
book, of the word book, on various occasions of use. So let's just say that
the internally generated phonetic representation of the word book just picks out a sound or
many sounds if we like, that relation we can
give a name, we can call, we can give that relation a name, let's call it phonetically
denote, p denote, let's say that that relation holds between the internal object and the sound. Its properties will be the properties of the technical notion, denote or refer, that's devised for formal systems, say the relation between the numeral three and the platonic entity three in a system of formal arithmetic
of the Fregean variety, for those of you who know
some logic or metamathematics. Further inquiries about the sound don't have to bother us,
they can be forwarded to the physics department,
perhaps they'll tell us that the sound is some
indescribable, four-dimensional space-time construct based
on motions of molecules. Perhaps inquiries can be forwarded to the sociology department too if we want to make the
task even more hopeless by introducing some unexplained notion of common language that
involves norms and conventions, attitudes, experts, and so on. Well, however we proceed there aren't any ontological problems, that is, no problems about existence. No one doubts the
existence of the particular motion of molecules that
I produce when I say book, so there's no problem about
the existence of sounds, the notion is perfectly robust, to borrow the standard philosophical term. Problems of communication
are very easily solved, you identify what I'm
saying because your internal representation p denotes
the same sound as mine or at least a sufficiently similar one. Okay, without preceding we've now solved the problems to which acoustic and articulatory phoneticians have devoted so much labor without much success. Well, it's clear enough why this solution has never been proposed, it's a bad joke. The new theoretical
notions, sound, and p denote are completely useless, they
merely restate the problem in vacuous terminology,
leaving the problems, in fact, even in more
worse shape than before. The proposal has all the virtues
of theft over honest toil to borrow Bertrand Russell's aphorism. The actual course pursued is much harder, we have to discover the elements of the phonetic representation
of the word book and others, the properties
that constitute it. Those are commonly called its features, and we have to determine
how feature complexes are accessed by the sensory-motor systems and related to external events, maybe ultimately motions of molecules. There's no determinant
mind-external object that's picked out by an element of the phonetic
representation as its sound. Rather, these internal
objects provide information, it's accessed by other systems to yield or interpret mind-external
entities in a variety of ways that depend on
circumstances, expectations, and a host of other factors. In fact, the real-world
problem is so intricate and involves so many variables that no one even contemplates studying it, actual inquiry keeps to highly idealized experimental situations as
always in the natural sciences. Well, let's now consider an easy answer to the problems that
arise on the meeting side, following point by point the preceding ridiculous account of the
sound side of language. I made up the other one,
but the one I'm going through now is real, unfortunately. We all recognize that the
word book has a meaning just as it has a sound. The entity book, whatever it is, has an internal semantic representation that incorporates all
information about its meaning that's determined by the language just as the internal
phonetic representation incorporates all language-determined information about sound. And we know exactly what the word book, or its semantic representation, picks out in the world, namely books just as the phonetic representation picks out the sound or the
sounds of the word book. So we can therefore set up a relation that's a counterpart of p
denote, let's call it s denote, meaning semantically denote, s denote holds that relation, holds between the internal semantic representation of the expression book and books. You understand that I'm
talking about books not tables because your word book, s,
denotes the same things as mine. And the child acquires
the s denotation relation by virtue of causal
properties of the world that relate external phenomena
to mind-internal entities, let's call them concepts
to use the standard term. Those who are concerned about the status of the external objects that are s denoted that don't have to have any qualms, these are some indescribable constructions based on whatever physics
tells us about the world, and we can forward further inquiries to the physics department or
maybe the sociology department, again, if we want to make
it even more hopeless. There aren't any problems
about the existence of books, those are the things that are
on bookshelves and tables. It's true that if you and I both took Darwin's Descent of Man out of the library there is a question about whether
one book was taken or two, but we can settle this any way we like. And if someone is concerned
that books are simultaneously abstract and concrete and have a host of other extremely odd
properties when you look closely, we can answer robustly that
that's just what books are, it's a problem of metaphysics and not semantics or cognitive science. Well, it should be clear that
something's gone badly wrong, we're back to theft
rather than honest toil. For one thing, referring is
something that people do, not words, that was stressed 50 years ago by Oxford philosopher Peter Strawson, and known before, of course. And like other human actions, referring is a highly intricate action. It's specific to circumstances,
it has normative aspects, the act of referring succeeds or fails in ways that depend on a
wide variety of conditions. The act need not even involve terms that have some circumstance
independent relation to the referential intention. So, for example, I can refer to India without using any word
or having any thought that has any independent
connection to India, whatever India is. If such fundamental properties
of referring, the act, are omitted from consideration, one may be studying something, but it's not the problem of
intentionality or aboutness. Well, a fair response that could be given is that exactly in other cases we have to idealize if we hope to gain some grasp of reality. We have to abstract away
from a welter of complexity to focus on the properties
of core notions, in this case s denote or p denote, and that's a reasonable response, but it's a promissory note. As in other cases, it has to be justified by showing how the idealization yields some insight and explanatory power and doesn't merely reformulate the original dilemmas in misleading ways. That doesn't seem an easy
task in the present case, to put it rather mildly. In fact, I think it really is theft rather than honest toil in both cases, the case that's never even discussed, the sensory-motor side, and the case that's pretty standard, the
analog on the semantic side. Well, how can we approach the
problem of what's happening when we think or talk about the world? It's possible, and in my view likely, that the study of sound provides
some useful clues to that. In that inquiry there isn't
any reference-like relation between an element of
phonetic representation and a mind-external entity. Rather, the speaker or
hearer employs the systems of language use to access
the phonetic representation, the internal object, so as to produce and interpret organism-external events. And perhaps something similar
is true on the meaning side, so Smith uses an expression to refer when his attention is focused
on some parts of the world which he views from the
complex perspectives that are provided by
internalist semantics, very much as in the case of sound. Features of the internal item, book, in Smith's internal
language, provide information that's used by other cognitive faculties, constraining the ways that Smith uses it to talk about the world,
think about the world differently from other words. Smith succeeds in communicating with Jones to the extent that Jones
attends to related parts of the world and has
appropriately-related perspectives and understanding of
circumstances and background. Similarly, Jones's ability to perceive what Smith is uttering
depends on his ability to map the noises that he hears to his own internal language. And these are always
matters of more or less, not yes or no. Well, that's the point of view that I've personally
regarded as reasonable since I began thinking seriously about these topics about 50 years ago. At that time influenced by Oxford ordinary language philosophy
and later Wittgenstein. But as I later learned, the approach has traditional antecedents. There's an important 18th century critique of the theory of ideas
based on the observation that the phrase, he has an idea, should not be understood on the model of, he has a diamond, invoking
a reference-like relation between the term idea and
some extramental entity, that's putting aside for the moment whether that's even a proper
step in the case of diamond, I think it's not. Rather, the phrase, he has an idea, 18th century commentators pointed out the phrase, he has an idea,
means something like he thinks. The phrase is what Gilbert Ryle called a systematically
misleading expression in an influential contribution to 20th century ordinary
language philosophy, actually resurrecting
a traditional critique. And the same conclusion
holds for a belief, a thought, desire, other terms
of so-called folk psychology as expressed in its
English language version, which is far from universal,
in fact, rather idiosyncratic. The basic insights generalized
to the whole vocabulary and even more dramatically
to more complex expressions constructed from lexical items. If this is anything like correct, most of the work that's
going in philosophy of mind, philosophy of language, and
theoretical cognitive science is just off on totally a wrong track, as my opinion has been for years. The roots of these
insights are far deeper, in fact, they extend
well beyond misleading analogic interpretations of surface form as in the case just mentioned, the 18th century case resurrected by Ryle. Perhaps the first study of these matters was by Aristotle, he asks, for example, what's the nature of the house? And he concludes, quoting Aristotle, "We can define a house as
stones, bricks, and timbers" "in terms of its material constitution" "or as a receptacle to shelter
chattels and living beings" "in terms of function and design." So I may think that the place
that I call home is a house, but I could be wrong, it
could really be a library in which some odd people
spend much of their time. And, in fact, someone
entering it for the first time might be pardoned for
reaching that conclusion. The answer depends on
choice of perspective and on circumstances which
I might not even know. So if that thing was
designed to be a library and is characteristically
used that way while I'm gone, in fact, much of the time I'm there, then perhaps it really is a library, it's not a house, contrary
to what I thought, or perhaps it's really a garage or maybe it's an oddly constructed and misplaced paperweight
belonging to a giant and so on, there simply is
no mind-independent truth of the matter, and material constitution, as Aristotle recognized,
is only one factor in reaching answers. We can also integrate the factors of material constitution
and function design, that's in Aristotle's terms,
combining matter and form, and we can bring in other factors too, some of which he also explored. Well, for Aristotle these
are questions of metaphysics, that is, the way the world is. Notice that he talked
about defining house, the thing, not the word house. From the 17th century there's
been a very reasonable tendency to reformulate such analyses in epistemological, conceptual terms, that is, as properties of something real, but the construction and interpretation of experience that's provided
by our cognitive capacities, that's in our cognoscitive powers in 17th century terminology, which include the internal semantics of the language. And, in fact, there was
quite illuminating discussion of these issues by Hobbes,
Cudworth, Locke, many others, sometimes adopting David
Humes' principle that, I'm quoting him, that
the identify with which we ascribe to things is
only a fictitious one established by the mind,
not a peculiar nature belonging to what we were talking about. Well, in all of these matters, textual interpretation is uncertain, but the general idea seems clear enough and very plausible. The house that Smith lives in or the books that he's reading surely do not have their
strange and quite intricate properties by virtue of some
mind-independent constitution. And the properties really
are strange and intricate, as soon as you look at the
meanings of the words carefully, dictionaries have nothing
to say about this, they just presuppose it. Rather, they have these
properties by virtue of the way Smith and others think in
particular circumstances and the internal meanings of the terms in which these thoughts are internally or sometimes externally expressed. These devices, in turn, are a property of fixed and shared internal human nature as are other aspects of
their lives and being. The semantic properties of expressions are used to think and talk about the world in terms of perspectives that are made available by the resources of the mind, rather in the way that the sounds of language seem to function, in the latter case as everyone assumes. I've tried to show
elsewhere and won't review that these conclusions are supported by descriptive observations
that are in a tradition, 17th, 18th century tradition that's been much too long forgotten
and greatly reinforced when we look more
closely into the meanings of even the simplest words,
including those investigated in the 17th, 18th century philosophy, British empiricism and others, words like tree or river or person or names of substances like water or the most elementary devices
of reference and anaphora, probably every aspect of language. All are far more intricate than
has commonly been supposed, so much so that they must come to us from the original hand of
nature, in David Humes' phrase, and hence must be fundamentally
the same for all languages. Now, these topics have been addressed in illuminating work by Julius
Moravcsik over at Stanford on what he calls the
ideational theory of meaning, bringing classical sources to bear on contemporary issues
of language and thought, others have pursued them too. And the approach seems to me to have a lot to recommend it. Well, what about Fregean systems
that are usually adopted? That is, systems based on
either relation of denote or refer, holding between
linguistic objects and extralinguistic entities. It seems to me that these
are reasonable enough for Frege's specific purposes, most important for what
really interested him, namely formal studies
of mathematical objects, at least understood
platonistically in his terms. The ideas also seem
appropriate as a kind of a normative ideal for the specific human enterprise of science. That is, one hopes that
such notions as say, black hole or oxygen or
electromagnetic field will pick out something in the world, in the mind-independent world. And we hope that the same will be true of the internal entities
and computational principles that are postulated in the
study of insect navigation or visual perception or human language. There's also some evidence
that animal communication, outside of humans, is based
on a notion of representation that's similar to the
invented technical concept of reference that's familiar
in the study of formal systems. Here, the concept
representation is understood as isomorphism, that is,
a one-to-one relation between mind-brain processes and an aspect of the environment to
which these processes adapt the animal's behavior. As, for example, when an
ant picks out the corpse of a conspecific by its odor. Actually, I'm quoting here,
again, Randy Gallistel in the comprehensive
introduction to a series of essays on animal representation. Well, if the picture that
I've just briefly reviewed is anywhere near accurate, it could turn out that
the use of language, human language, to refer or in other ways is totally different from
animal communication systems and this is in numerous other respects. Human language might not have
any denotational semantics, just an intricate form of pragmatics along with very rich internal syntax that includes what's
usually called semantics but ought to be called syntax, it's the study of
internal representations. Of this amalgam, the parts
that we can currently hope to understand in most depth
are the internal syntax, what's called semantics. A theory of human action
that would bear on in some revealing way
on the act of referring is far more remote than
comparable theories for much simpler organisms and actions, domains in which the problem
is scarcely even entertained because it's understood
to be far too complicated. Well, so far I've kept the language, but in principle everything I've said should carry over to the study
of other mental qualities. In practice, however, the
difficulties mount very quickly, language appears to be relatively isolated from other cognitive capacities. Here I'm referring to its
structure, not its use or the particular component
which are integrated into the structure which
could be individually shared with other cognitive faculties
or even other species as I had mentioned before. One of the reasons why
language is a good topic for study for inquiry into the mind is its essential role in human affairs, but another is that it is indeed or appears to be relatively isolated. When we turn to other aspects of mind, for example, our moral nature, it's much harder to isolate
components for separate study, that means to abstract them
from reflective thought and variety of other factors. Nonetheless, those topics,
the study of our moral nature, have been subjected to
investigation in various ways. Now, that includes interesting
thought experiments, actual experiments with children in the last couple of years
and comparative studies. Not uncommonly, the real
world offers illustrations of how these faculties function, sometimes with very painful choices. Issues like that test our moral faculties and may help us to discover
something about their nature. Sometimes this perspective is counterposed to what's called a relativistic one which holds in an extreme form that apart from their basic physical
structure humans have no nature, they have only history,
or that their thought and behavior can be
modified without limit. Nothing like this can
be even close to true if taken literally, though seems to be what is sometimes said in aversion due to Richard Rorty
history on anthropology, I'm quoting him, show that humans have extraordinary malleability, we're coming to think of ourselves as the flexible, protean,
self-shaping animal rather than as having specific instincts. Now there can be no moral
progress in human affairs, just different ways of looking at things. We should put aside the vain effort of exploration of our moral nature or reasoned argument about it, and we should keep to what he
called manipulating sentiments if we happen to be for or against torture or massacre, for example. I suspect I'm misinterpreting
because it's hard to believe that the words are intended to mean what they seem to say. Well, such proposals have
evoked a good deal of criticism, Oxford philosopher Galen Strawson, quoting related statements of Rorty's on the relevance of the
extralinguistic world, the truth, he asks whether
the nonsense might be less bad if it didn't build in such
an astonishing contempt for the reality of human suffering. And his conclusion is it's just as bad. A recent paper on the
philosophical foundations of human rights by Matthias Mahlmann discussing Rorty and others points out that, quoting him, nobody would've taken a nazi seriously who had claimed in 1945
that the sole basis for the moral condemnation
of the Holocaust by the rest of the world is just due to some kind of culturally
relative emotional manipulation based on shrewdly devised
sentimental stories. If that's so, and I assume it is, we want to understand why, and if nobody really means no normal human being, that leads us back to the hard questions of intrinsic human nature. This notion of unique human malleability is not at all novel, it's been
a fairly conventional view, at least back to the
beast-machine controversies that were inspired by Descartes. And, for example, the
argument of James Harris, British philosopher in 1740, that unlike animals and machines, the leading principle of man is multiform, originally uninstructed,
pliant and docile. The idea that human-alleged
weakness of instinct leads to vast variety
and extreme malleability has had a long and, in fact,
inglorious history ever since. Well, with no metric and
little understanding, it's hard to know what to
make of those judgments. But whatever merit they have, they cannot offer an alternative to the conception that I just outlined. No one doubts that a
person's understanding and judgment and values and goals reflect acquired cultures,
norms, conventions, and so on. But these aren't mind-external entities, they're not acquired by taking a pill, they're constructed by the mind on the basis of scattered
and constructed experience, and they're constantly
applied in circumstances that are novel and complex. And to these facts and their significance were discussed 250 years ago by David Hume who observed, I'm quoting him, that the number of our duties
is in a manner infinite. Therefore, just as in other
parts of the study of nature we must seek a few general principles upon which all our notions
of morals are founded, principles of human nature
that are original instincts of the human mind that are
perhaps enhanced by reflection but steadfast and immutable as components of fixed human nature. Hume, notice, is articulating the basic idea behind generative grammar in a different cognitive
domain and centuries earlier. Like Adam Smith, Hume
took sympathy to the, what he called a very powerful
principle in human nature, one of our original instincts and the grounding of much else. That idea was reconstructed
in a Darwinian framework about a hundred years ago by the anarchist natural historian Peter Kropotkin in what I think should be
taken as the founding work in what's nowadays called
evolutionary psychology. And there's recent work that suggests some possible evolutionary scenarios. Well, there's little reason to suppose that the variety of cultural outcomes reflect significant variety
of genetic endowment so we're back to the situation we faced in the study of language
or the visual system or any other basic property of organisms. It's necessary to account for the richness and specificity of outcome on the basis of shared intrinsic nature
tolerating variation but within a highly structured range as throughout the biological world. Well, to this picture we should add other conceptions that have
been studied in recent years and also resurrecting
the 17th century origins of modern science. One strand is the recognition
that our innate capacities are only latent, that is
they have to be triggered by experience to be manifested. Then they're manifested in ways determined by our intrinsic nature,
much as susceptibility to a disease is innate, though the disease requires an external trigger. That's actually the analogy
that Descartes suggested in discussing innate
ideas, one of the reasons why Locke's famous critique
is beside the point, that's long recognized. Another idea that was fruitfully examined in the 17th century is that the phenomena of the world around us do not in themselves constitute experience. They become experience for
us as they are constructed by our modes of cognition. I'm quoting 17th century work, these, they must therefore conform to these modes of cognition. This is 17th century, picture
came into modern thought through Kant's version of it. These modes of cognition
are a distinctive property of our nature, they differ
for different organisms. They're what the evolutionary
biologist Konrad Lorenz called the biological a priori, in work that I, in fact,
discussed here 35 years ago. That's also true of the
rich mental constructions that we call cultures,
norms, and conventions insofar as they're shared by groups that interact in complex ways. Well, still assuming that each child is intrinsically capable of acquiring any culture over a very broad range, the process of mental
construction of experience and interpretation of it is based on the common genetic constitution which must be rich to the
extent that the outcomes are highly structured and constrained in ways that do not
simply reflect features of the environment,
basically Hume's observation. It seems unavoidable that the so-called relativistic approaches
must be profoundly innatist, at least that they're willing to address the issues of nature acquisition and use of attained
systems, Hume's questions. If so, then they fall
together with the study of visual or linguistic systems or other properties of organisms, hard to see any serious distinctions here, or any interpretation
under which relativistic approaches differ from the most
highly innatist approaches. Well, one last word on the
import of any conclusion, it has to be a very tentative
conclusion about human nature, one way to assess the
importance of such conclusions is to observe how deeply
they enter into conceptions of right and justice and the
struggles that they engender. It's very easy to illustrate
from personal relations to international affairs. More generally, every approach
to how human relations should be arranged,
whether it's revolutionary or reformist or committed to stability, every such approach is
based on some conception of human nature, at least implicitly. If it has any claim to moral standing, it's advanced with the claim
that it's beneficial to humans, meaning because of their intrinsic nature, at least as one of its crucial qualities. We should face honestly
the fact that our ignorance which is profound, but yet
recognize that we have no choice but to proceed on the
best tentative assumptions we can reach. For many of the classical mysteries, quite extraordinary bodies of doctrine have been developed in the
past several hundred years, they're some of the greatest achievements of the human intellect, they also have far-reaching
implications for human life. And there have been some remarkable feats of unification as well,
sometimes very surprising in how they turned out as in the cases I discussed yesterday. How remote the remaining
mountain peaks are, even where they are,
one can scarcely guess. Within the range of feasible inquiry there's plenty of work to be done in understanding mental
aspects of the world, including human language. We would do well, however, in my opinion, to keep in some corner of our minds Hume's conclusion about
nature's ultimate secrets and the obscurity in which they ever did and ever will remain, and
particularly the reasoning that led him to that
judgment and the confirmation of that reasoning in
the subsequent history of the hard sciences. Now, these are matters that are too easily forgotten I suspect and that merit serious reflection, perhaps someday even
constructive scientific inquiry. Thank you. (audience applauds) - We have a number of questions that we'd like to put to you. - Okay. - First of these is, when
you say that language is a matter of computation, do you mean in the observer-relative or the observer-independent sense? And do humans really
start off in the world with the requisite concepts to express things like carburetor and bureaucrat? - Well, as far as the
first part is concerned we could ask the same question
about insect navigation. So take a look at, say, the
study of insects by scientists. You find that they attribute to the insect highly intricate computational systems which enable, say, bees or ants to do things that are
far out of our range. We can't determine the azimuth of the sun as a function of the time of year and day or do dead reckoning the way
an ant and so on and so forth. The explanations of these that are offered in the literature are
computational systems. Are they independent of the observer? Well, in the sense that any science is independent of the observer, so when you have a model of
the, say, planetary system, in a sense it's relative to the observer. That is, we can't get out of
our skins, that's hopeless. But the enterprise of science is an effort to get out of our skins as much as we can. You try to construct
an account of the world that is as observer-independent
as we can manage. Now, there are limits to this, for example, studied in quantum physics, but we can put that aside. So yes, it's observer-independent in the sense in which
any scientific construct is observer-independent, no different. It's true whether it's
computational systems of insects, whether it's
computational systems of humans, whether it's the planetary system or anything else that
science tries to understand. As for carburetor and bureaucrat, I have to say that my Jerry Fodor was a little offended by
the fact that the statement you quoted was attributed
to me in a recent article, actually it's his, his proposal, and he wrote to me that he's sorry that it came up because he can prove that the notion carburetor isn't innate since he doesn't even know
how to spell it, you know, but the fact is that he was
making a very serious point. His point was that we
have to somehow account for the fact that terms like carburetor and, what was the other,
bureaucrat, we do understand, just as we understand river and tree and person and very simple words, and we understand them on the basis of extremely limited evidence. We have a rich and complex
understanding of them. So we're back to the problem of why we grow arms rather than wings. You can't just wave your hand about it and say, well, it's the culture because then we have to ask
how did we acquire the culture? And, as I said, it's not by taking a pill. The culture is a construction of the mind on the basis of scattered experience. So the answer that Fodor,
Fodor takes a stronger position than I do, I don't
want to defend his position, but in general this is the problem, what is it about the nature of our minds, the intrinsic nature of our minds that allows us to acquire concepts like river, person, tree, water, book, carburetor, bureaucrat,
even though we have very scattered experience? And that's the general problem
of developmental biology. And the only answer that anyone knows is the one I just quoted from Hume, talking about our moral
nature 250 years ago, and the one that's
assumed by every biologist for everything except the
human mental qualities. It's got to come out, come somehow from the intrinsic nature of our minds. And in that respect it's innate, the fact that it's difficult to accept is a sign of our irrationality because it's got to be that
way unless it's a miracle. It's either a miracle or it's
pretty much along the lines that Fodor suggested, maybe
not as extreme as his position. - Thank you. Semi-related, the second question is, is it a good idea or not to search for evolutionary precursors to language whether in cognitive,
sensory, motoric systems? - Yeah, it's a good idea, and, in fact, if you can find any, fine. In fact, I suppose the
question was written before the comments I made
- Absolutely. - Well, yeah, I think it's a fine idea. In fact, I think we're
just getting to the point where it may be possible to do it. If, in fact, something like the program that I very sketchily outlined, if something like it
turns out to be correct, that is if we can give
a principled explanation for significant aspects of language in terms of interface conditions and computational principles, then it makes perfect sense to search not only for, I wouldn't
call them precursors, for homologues, homologous
structures in other organisms that happen to be like this. And as I say, we might
find very remote ones. We might, I don't think anyone expects it, but you can't exclude the possibility that we might discover that the basic computational principles of language show up in insect navigation,
it's not impossible. In fact, it's not even crazy. It wouldn't be a precursor to language like bees aren't a precursor to humans. Remember, species are not
precursors of one another like bees are no more precursors of us than we are of bees, just
different evolutionary lines. And if you can find sensory or perceptual or other computational components in the nature of other organisms that have some potentially
homologous relationship to human language, yeah,
that would be interesting, of course, that's what
evolutionary studies are about. - [William] Well, further
proof that these were written before you spoke is
given the next question, why are children geniuses of language between ages one-and-a-half and four, yet other cognitive abilities
are still developing at a much more concrete,
much less abstract level? - It's kind of like asking
why children undergo puberty around age 12 instead of age 35, that's the way we're designed, you know? We're designed so that certain things happen at particular times. In that respect we're like every other organism in the world. Binocular vision, being
able to focus your eyes and use that for depth
perception and so on, now that comes along
at around four months, so we might ask, well, why does
it come along at four months and not at one month or six years? Well, the answer to that
would have to be something about the way we're
genetically programmed. All of these questions are
much too hard to answer, nobody knows the answers
to any questions like that even for primitive organisms, what we call primitive, like insects, now which are by no means primitive. But to try to answer the
question for language is far beyond reach, nobody knows how genetic instructions
create an organism. That's one of the reasons
why the Human Genome Project is looked at with some
skepticism by many biologists. Now, you could know
the entire human genome and that would tell you
very little about humans, because you don't know how you go from genetic instructions to an organism. That's an incredibly complex process, only tiny bits of which are understood. And this is just one of
those many questions, all of them very poorly understood. - [William] Somewhat different
direction is indicated by the next question,
if the study of language is a branch of natural science and language consists of a
lexicon plus computations, how would you respond to Sorel's critique that computation is not a
notion of natural sciences like mass or photosynthesis
but an abstract mathematical notion which
is entirely relative to the observer? - Same, well, first of
all it's not relative. It's relative to the
observer in the same way that any scientific construct is. Yes, in that sense, of course it is. Is it an abstract mathematical concept? Sure it is, just like Fourier analysis calculus, anything that's
used in the sciences is an abstract in the advanced sciences, is an abstract mathematical concept that is applied to particular situations. That's why Newton and Leibniz
had to invent calculus in order to study the effects of motion, and the same is true here. Computational systems are particular kinds of mathematical notions that appear to be very well-adapted to problems like, say, insect navigation, bee communication, human language, and some other things. So the point is correct, yeah, it is an abstract mathematical concept used in these areas to
give us the best theories that we can find. If somebody can find a better theory that doesn't involve computation, great, let's have a look at that, for insects or humans or anything else. But until then we just work with the best theories we can find. Notice that this is very similar to what I talked about
yesterday concerning chemistry. Up until very recently, in
fact, less than a century ago, 1920s, leading scientists
described chemistry in exactly the same way. They said it's just an
invented computational system, it's a calculating device, it
yields the phenomenal effects that you see in experiments,
but it's not real, can't be real, it's just something that people invented in
order to have some way of organizing the phenomena that they discover in experiments. Why can't it be real? Well, because it couldn't
be reduced to physics. Which was true, couldn't
be reduced to physics, reason, physics was wrong, so it couldn't be reduced to physics. After physics was radically revised by the invention of
development of quantum physics, Linus Pauling succeeded in giving a quantum theoretic account
of the chemical bond, he won the Nobel Prize for it. It's 1935, it's not that long ago And then chemistry and
physics were unified and everyone recognized
that all of this talk about calculating devices
was complete nonsense. Chemistry was the best
theory you could devise for dealing with these problems, and that's the only notion
of reality that we have, we don't have any other notion. The only notion of reality we have is whatever our best theories
provide, there's nothing else. So chemistry was as real as you like, happened to be more real than physics which was on a wrong track. When physics changed they became unified, you get the contemporary science. In this case, the best theories we have as far as anyone knows are
computational theories, so that's reality. If neurophysiology doesn't
integrate with them and we don't know where the problem is, it may be in the brain sciences, it may be in the computational theories, maybe everything's on the wrong track, that's often happened in
the history of science too. You know, you don't know the answers until you find the answers. But there are no questions
about the reality and calculating devices
and observer relativity and so on, they simply do not arise for reasons which are well understood. We have a couple of centuries of history of the hard sciences from which we ought to be able to learn some
lessons, including these. And they apply in this
case as in earlier cases. - [William] As a theoretical chemist, thank you for transcending
some boundaries. Moving along. You mentioned yesterday
that a number of theories of biolinguistics are, quote, pointless or obviously misguided. Why then do other scholars
continue to pursue them? (audience laughs) - Gotta ask the other scholars. I can't, it depends, I mentioned
a few and gave some reasons why I think they're misguided. Incidentally, those were not
theories of biolinguistics, those were theories of language which deny that language
is a part of the world. - [William] I see. - I think that's seriously misguided. - [William] You have disqualified Quine's radical translation
paradigm, calling it a new, quote, mythological, epistemological
dualism, closed quote. Could you explain this more? - Well, the radical translation, there's a little misunderstanding of this, the radical translation paradigm, the fact that, this will
only be intelligible to those of you who know this stuff. But Quine's conclusion about what he calls indeterminacy of translation
that comes from this paradigm, I don't disagree with that conclusion. I think the conclusion is correct, it just seems to be a truism. That's what I thought when
I was a student of Quine's in graduate school 50 years ago, and we argued about it
then and I still think so, it's true, but it's a truism. In fact, it's just one example of underdetermination
of theories by evidence, known to be a truism at least since Hume. Quine's argued that there's
nothing more involved, but I don't think he's demonstrated it, you can look at the literature and see. As for the radical translation paradigm, that's an approach to
the study of language that we would never adopt in
the study of anything else, say, insect navigation or any topic, you would never adopt that paradigm. In that case you have to ask why do you oppose it for human language? Well, as far as I can see, that's a kind of methodological dualism. It's saying we somehow have to approach higher human mental faculties by methods that we would never use anywhere else in studying the physical world. Now, that's not old-fashioned
metaphysical dualism, it's something much worse,
methodological dualism. Metaphysical dualism, you
know, mind-body dualism, that was a scientific theory. It was a substantial scientific theory, it was the early scientific revolution, Galileo through Newton, it was
a serious scientific theory that was a conception of matter in terms of a mechanistic
picture of the world, what's called contact mechanics,
common sense mechanics, things interact if they're in contact, what we all intuitively believe even if we've taken a physics course and know that it's false,
we still believe it because we can't help it, it's our nature. So contact mechanics was developed, gave a picture of matter. There were clearly some things
it couldn't accommodate, it's Descartes' point and he was correct. He therefore postulated
a different principle to account for them, that's science, you know, it's the
right way to do science. He was shown to be wrong
in a surprising way, namely matter doesn't exist, okay, therefore you can't, doesn't matter, it's of no interest that other phenomena don't fall within the concept of matter because nothing does. Okay, so that ends metaphysical dualism, there's no more mind-body problem. Methodological dualism is just pernicious, there's no reason why
we should study human higher mental faculties in suicidal ways which we would never use
in studying other topics, why should we do that? That's a kind of a methodological dualism which has no saving graces
as far as I can see. - [William] Thank you
so much, and thank you for a couple of wonderful presentations, please join me in thanking Dr. Chomsky. (audience applauds) (introspective music)