AN INTERVIEW WITH JUAN URIAGEREKA
(Part I)
Ayesha Kidwai: In your own work you have consistently sought to contextualize the project of generative syntax within the larger framework of human biology and evolution. While this is without question the context of Chomsky’s inquiries, how much do you think that this consciousness of the larger picture should guide the everyday practice of the ‘ordinary syntactician’? If it is also your impression that this consciousness is true of only a fraction of syntacticians, do you think that a neglect of the larger picture poses a problem?
Juan Uriagereka: There’s an old American saying that goes “You pays
your money, you makes your choice!” In my view linguistics is a “special
science”, in pretty much the way chemistry or for that matter biology is.
If you wish, it is part of the study of “complex systems” in nature, and
“complexity” is, well, complex. Of course, language is so central to the
human experience that it may not seem complex to any of us, but
professionals of this business know better. The question is, then, what to
do about that. Some of us – and certainly Chomsky is the best example –
feel that to gain a real understanding of linguistic structuring you must
fit it within “the larger picture”. This is mainly to help us decide what,
among competing and reasonable theories, could fit a criterion of adequacy
that makes one or more approximate the truth. So here you go into more or
less traditional concerns (and I mean “traditional” in the sense that
these worries take you back to the Enlightenment, the Renaissance or even
Classical Antiquity): what sorts of structures can a child acquire, what
sort are realistic for a human mind, what could have emerged in the
history of our species, and so on and so forth. Without criteria like
these, it is hard for me to even imagine how we could decide on the
details of “the right theory”, among other things because we simply can’t
experiment with a human the way a chemist does in the lab, or even the way
the biologist can with plants or animals. As a result of that, a structure
that would surely reveal itself if we started knocking out genes or taking
away parts of brains, or whatever, has to be studied with much more
subtlety and interdisciplinary ingenuity. But that’s neither here nor
there: it’s just (good) science.
And no, “neglecting the larger picture” poses no such problem as such,
any more than, say, neglecting doing experiments would for a chemist.
This is not a matter of principle, just one of practice. Ultimately, only
time will tell whether structures postulated in ignorance of all of this
will in the end be more adequate than those we postulate informed on
biochemical, physical or mathematical concerns. There are no guarantees
for success in science, anymore than there are in art. The proof is always
in the pudding.
AK: With the developments in the minimalist program however it appears that the big picture context has attained greater prominence than ever before. Two sorts of constraints now apply for explanatory adequacy (and beyond): (i) those grounded within the language faculty, such as considerations of economy and computational complexity, and (ii) those that operate in domain-general ways, such as “bio-physical laws”. Is it not the case that while recourse can be taken to some intuitions about language to discover those in (i), the ones in (ii) lie outside the domain of linguistics per se? How can such constraints then be understood and ‘discovered’ by ‘ordinary syntacticians’? And how is such work to be of the same order of value in a cross-disciplinary fashion, as for example, the modularity thesis was three decades ago?
JU: First let me clarify the idea of “beyond explanatory adequacy”.
Chomsky classically postulated three levels of adequacy for linguistic
theory – I’d say for any scientific theory. (i) Observational, (ii)
descriptive and (iii) explanatory. (i) is the minimum criterion for
empirical science, obviously. (ii) already requires some sort of coherent
edifice from which you can deduce the facts observed, and of course make
new predictions and so on. (iii) is where the real claims of truth lay,
though. Each science presumably has a way of distinguishing among
(roughly) descriptively adequate theories, whether it is elegance,
reduction to some other science, or whatever you think should decide the
competition among theoretical alternatives. In the case of linguistics,
Chomsky suggested, reasonably, that the very first explanatory criterion
should be language acquisition by infants, in “poverty of the stimulus”
conditions. It is pointless to have a fancy descriptive theory if there is
no way a child could possibly fix its open variables in realistic time. It
took us a few decades to even come close to addressing such a criterion,
and I think it is fair to say that it was only with the advent of the
Principles and Parameters model that we came even close to being able to
explicate how children acquire the diversity of languages we experience in
the world. Don’t get me wrong: I don’t think that is a finished project –
in short, we don’t know yet, in full detail, how it is a child can achieve
this phenomenal task. But at least we have an “in principle” answer, and
we pretty much know, or at least strongly suspect, that if we continue
working on this line of research, we’ll get a better and better picture.
That said, have we cracked the linguistics nut completely? Do we have,
now, an explanatory theory? Well, it depends on the demands you impose.
Physics in the nineteenth century was thought to be pretty much finished
and understood, a matter of finessing the details of the forces and
interactions among what was then thought to be atomic units. By
challenging that system – not giving it up, mind you: just asking
questions about its foundations – people like Einstein and Poincaré first,
and of course later on all the particle-physics guys, Schroedinger,
Heisenberg, Dirac and so on, the whole bunch, told us how much we were
missing by not going beyond the obvious. It’s probably a model story for
us too. Sure, we can stop at classical “explanatory adequacy”, just as a
number of physicists these days would still not go into superstring theory
or whatever the latest model brings, “branes” or whathaveyou. That’s
fine. But then again it’s perfectly legitimate to keep attempting to unify
the forces, ask about reducing the elementary particles to more basic
objects, and so on. Same in linguistics, going “beyond explanatory
adequacy” just means pushing the envelope, and sure, then Minimalism makes
a couple of interesting, and tentative, suggestions about where to find
answers.
Now, once that much is accepted, your question is whether “ordinary
syntax” can make progress here, and whether some of the minimalist
concerns don’t just lay outside linguistics. The thing is, I’m afraid I
don’t quite share the presuppositions. Again, let’s go back to physics,
chemistry or biology, those sciences where clear progress has been made
during the last few centuries. Does, say, String Theory look like ordinary
physics? Well, I can tell you I studied a bit of relativity in school, and
that did look like the ordinary physics I knew and could operate with,
with Newton’s laws extended in ways that, though difficult to follow
operationally, are conceptually quite simple. I can still read that stuff.
However, if you show me any quantum physics beyond, say, Schroedinger’s
equation – which I can still trace, in some form at least, to more
familiar stuff involving kinetic and potential energy, that sort of thing
– I have to start asking help from physicists. And even when they explain
things to me, I don’t feel I have a good grounding of what’s really going
on, and why, say, the relation between scalar and vectorial values does
make such a difference in how given variables interact, and so on. But
that’s “ordinary physics” too, or I should even say, that’s most of what
kids now study in college, right?
Same with chemistry: I’d doubt Levoisier would have the foggiest idea of
what, say, Linus Pauling was doing, although of course ultimately what
Pauling achieved allowed for the explanation of the phenomena Levoisier
had observed – of course via Dalton, Mendeleiev, etc. Not to talk about
what goes on today in organic chemistry, where not just simple “valence”
is at issue, but all sorts of arcane topological shapes that various
compounds achieve, which affects just about anything one can imagine in
“chemical form”, if you wish. And imagine what Darwin would say about
“ordinary molecular biology”… So the issue is true of all sciences, as
they progress: they become more intriguing, interactive, and harder to
work with. So what do you do then? What scientists have always done: When
Newton needed the calculus, he invented it. The idea was good enough for
Leibniz to also, apparently, have come up with it, independently. And
that’s ultimately the right measure: if Minimalism is in some sense right,
things will start falling into place. Units and interactions will start to
make sense, whether Chomsky postulated them or someone else did, simply
because they are good ideas about a complex phenomenon.
Let me clarify one more thing: you don’t “discover” any of this.
Theoretical science is not really about discovery, but about evaluation,
following criteria along the lines discussed above. How does one come up
with a theoretical edifice, like, say, Minimalism? Certainly not via any
discovery, something that, say, Chomsky one day finds in a lab or after
encountering a language in the Amazon. Rather, first of all, you are part
of a scientific enterprise – linguistics is clearly one of them, going
back to Panini, if not the scholars who first allowed us to segment speech
into writing systems. Second, you begin to develop theoretical edifices,
like for instance the one Saussure and the Russian formalists put together
in the early twentieth century, which is where most of us come from in
contemporary linguistics. Third, you test the hell out of them: you
squeeze, stretch, bend, and even break what they say, until you
approximate a better understanding of what they may be telling you.
This is one respect in which I think many scholars deeply misunderstand
Chomsky’s approach to science. They often say that he “changes his mind”.
I think that’s rubbish: he rarely does, the sorts of assumptions I’m
discussing now have been his forever, so far as I can tell. What he does
change are the ways in which his theories work. Why? Well, because he
learns new facts, like all of us do, and because he’s human after all, and
he may have missed a step here or taken too fast a turn there. The
greatest geniuses make mistakes. Ramanujan, for instance, was apparently
the most gifted mathematician you could encounter – but he occasionally
didn’t get things right. And that’s in a formal science; imagine in an
empirical one. For instance, Einstein made mistakes he didn’t know he was
making (attempting to reduce quantum physics to relativity, which everyone
nowadays will tell you cannot possibly work, the opposite reduction being
the only one with a chance) and he even thought he made mistakes where
there wasn’t one (like the Cosmological Constant, which apparently has
something to it).
So Chomsky, or the rest of us, is no different: we try our best, and part
of that process is not regarding our models as dogma. They are just ideas.
I’d be very worried about a theoretical model that has remained constant
for decades! It reminds me of Borges’s character, Pierre Menard, who
wanted to write Don Quixote in the twentieth century, after all that has
happened since Cervantes’s did… So can anyone seriously keep the exact
same model after, say, the advent of Chaos Theory? How about the mapping
of the genome or the proteome? How about developments in Topology,
Category Theory and so on? I think it would be absurd, if these other
adjacent sciences continue to make progress that surely will be of
relevance to anyone who studies language, or more generally “form” in this
universe.
So that brings me to the last aspect of your question: Will the concepts
we work with in Minimalism have the impact that Modularity had? I don’t
have a crystal ball, but it seems to me that it would be irresponsible for
other scientists out there not to look at what we do in linguistics,
especially at the level we operate with in Minimalism. Let me tell you a
little secret: there are not so many great ideas out there in Complexity
Theory… I mean, the field is fascinating, and researchers have now
convinced most of us that there is such thing as predictable
unpredictability (chaos), and perhaps even the emergence of stability
within that mess (complexity), God only knows how. But that’s the key: we
only have certain intuitions about how form can stabilize under, say,
dissipative conditions of the sort Ilya Prigogine studied. Some progress
might come from the sorts of studies that I guess go back to René Thom in
topology, singularities, catastrophes and all that, themselves arguably
the sort of thing that Poincaré had foreseen. And some might also come
from computational studies, the sort perhaps Stephen Wolfram has pushed,
which again go back to earlier things by Mandelbrot and so on. But I see
absolutely no reason why we, linguists, cannot also contribute to this
project. In fact, Neils Jerne already observed that in his Nobel laureate
lecture, in 1984. And people like David Searls are making extremely
intriguing use of computational linguistics concepts (like the Chomsky
Hierarchy) to give us some insight into such hard nuts to crack as the
“protein folding” puzzle, after having studied linguistics with the Penn
Group.
In my view this is only the beginning. I have profited the most from
working with people in the edges of linguistics, applying relevant notions
to adjacent fields, whether it is philosophy, computer science,
psychology, neurology, or even computational biology. These scientists
have asked my collaboration as a boring syntactician, and in order to
better understand the very complex problems they were dealing with. In
fact, I’d say our department at Maryland is constructed based on exactly
this premise, which was shared throughout two long decades and by three
very competent chairs (Lightfoot, Crain and Hornstein). I don’t think this
is accidental, just the normal consequence of interdisciplinary research,
which linguistics, and certainly the Minimalist Program, continues to be
at the forefront of.
AK: The Hauser, Chomsky & Fitch work of recent years has had a tremendous impact on scientific inquiry in general and ethology and biology. While you have already begun to explore the implications of the “recursion-only” thesis in your own work, many other linguists have been skeptical, holding the claim to be too strong. Most notable of the aggrieved have been Ray Jackendoff and Steven Pinker, whose critique of the thesis specifically is that it flies in the face of the findings of generative linguists over the past five decades – that syntax and morphology are special to language, and therefore to humans. Do you agree with them that the HCF thesis is a rejection of past knowledge? If not, can you please elaborate upon how this thesis may be seen as an outcome of the progress made by generative inquiry over the last fifty years?
JU: I totally disagree with that critique. Chomsky, Hauser and Fitch,
aside from being top-notch scientists in their respective fields, happen
to be extremely well informed and responsible scientists, who are just
aware, to start with, of current developments on the Evo-Devo project in
biology – not to speak of current developments within linguistic theory.
Let’s take the matter head on. Suppose it were the case that, say, the
FOXP2 gene recently found to be implicated in language (and let’s assume
for the sake of argument that it is, and indeed seriously so) had the
basic same role it has now in an ancestral organism. To be specific – and
this is probably close to true – we know FoxP2 (in lower-case to
distinguish it from the human version) is present in song-birds, distant
from us some three hundred million years of evolution. Moreover, we know
that even among the birds, while all manifest this gene in roughly the
same neural circuitry, only three major groups (singing birds, humming
birds, and parrots) have the relevant feedback loops that lead to what
you may think of as “vocal learning”. To the best of anybody’s knowledge,
(relatively) similar brain anatomy within the mammals is found in bats and
cetaceans, and then you don’t find it among the apes until you hit us (and
perhaps Neanderthals).
Alright: so what does this mean for the autonomy of syntax, if true? It
really means zilch. It is perfectly possible, indeed likely, that human
syntax is nothing but new strategies with old tools, the way Stephen Jay
Gould would have expected for all biological innovation. To tell you a
possible story, much in the spirit of the HCF paper, it could have been
that an ancestral gene (FoxP2) was responsible for regulating a “parsing
window” of the sort that you need to identify bird-song. This is very
likely going on in zebra finches, at least, who we know through the work
of Constance Scharff, express FoxP2 in the acquisition circuit (early on
in life) and moreover in the production circuit (later in life). In the
second instance, as Stephanie White and collaborators have recently shown,
whether the gene is down or up-regulated depends on whether the bird is
singing alone, practicing as it were, or singing to the female, attempting
to get one particular song through that, if communicated, can lead to
mating. So there you go: arguably if the “operational memory” window (to
put it in terms sympathetic to Michael Ullman’s conjectures in this
regard) is narrowed, the bird enters some sort of an improvisation mood,
within relevant limits – a bit like Charlie Parker practicing in the
back-alley. In contrast, if the memory window is widened, the bird now
basically sings an aria, Pavarotti style, to his beloved, hoping for the
best. Say that’s true (it’s the only explanation I can even imagine for
the observed behavior). So we have a regulating gene that widens and
narrows operational memory, whatever that is. A gene, by the way, which we
know can be turned off, even within birds, for whatever reason –indeed
most birds don’t have any of this, even if it is part of an ancestral
lineage shared with mammals, so pretty far back.
Now enter the mammals, and for some reason a few of those again make use
of the gene. I don’t know of any specific studies – at least as careful
as the finch ones – that tell us what the gene is doing in dolphins or
bats. But whatever it happens to be doing there, it may have nothing to do
with what it does for birds (in all likelihood), or for that matter us.
Our version of the gene is novel, possibly mutated within the last couple
of hundred thousand years, if calculations made by the Leipzig group are
correct (if so, possibly not even Neanderthals had this allele, although
we’ll know soon enough, as their genome, I’m told, is in the process of
being mapped). What does FOXP2 do for us? In short, we don’t know. We do
know, however, that when the gene mutates, as in the famous KE family, the
result is dismal inflectional morphology, non-existent concord, absence of
complex embedding, and so on. (I wish, by the way, that all the data
available on this poor people were examined again from current
perspectives, but I also feel bad about bothering them any more.) That
seems important, and possibly very structural.
Of course, it’s true also that the affected members have difficulties
with the serialization of gestures and things of that sort, including
mathematical tasks. But we have no idea whether those are parasitic in
syntax, in some sense or another, as seems likely. At any rate, you can’t
start knocking the gene out – the way you do with mice – to see what
happens then, and you can’t chop people’s heads off – they way you do
with finches – to see whether the caudate nucleus is swamped with the
FOXP2 protein as we talk, write, sing or whatever. So we go with other
methods, and come up with conjectures, and the usual, which is the way
Michael Ullman made his bet that this gene is regulating operational
memory. So fine, say all of this is true for the sake of argument. Have we
questioned the autonomy of syntax?
I really don’t see how. In all likelihood, what we are doing with FOXP2 is
something like linearizing in parsing the complex structures that, say,
Theta Theory (or your favorite structuring mechanism) gives us. It is
very possible, I’d say even probable given all we’ve found about chimps
and bonobos in the last couple of decades, that they too have some form of
syntactic structuring mechanism, at least as a system of thought. So fine:
we possibly, then, have, the Theta Theory of a chimp coupled with the
linearization parser of a finch. If you wish, to put it as a T-shirt
slogan, again in the HCF spirit: chimp + finch = human. Alright, I’m being
provocative – but it’s to make a simple point. Although we have no idea
what the function is that relates (the relevant neuroanatomy or molecular
biology of) a chimp and a finch, we do know that it is this sort of
combinations that give you new species, or new structures within them, or
new behaviors within given structures, etc. That is the Evo-Devo project,
in short. The realization that things couldn’t be as smooth as Darwin
expected (with nineteenth century tools and preconceptions), and perhaps
even as bumpy as Gould claimed (with classical twentieth century, and in
his case historicist assumptions). Who knows what the details are here,
but this is what we’re getting over and over at this level: modify a
regulating gene for some reason, and although in most instances you just
don’t get a viable organism (legs start popping out of eyes or weird,
monstrous things like that) occasionally you get the fountain of youth. So
you stick to it (“you” meaning some species), and then other changes
hopefully follow that finesse the new structure/behavior.
Is the new organ, circuit, mechanism or whatever “autonomous”? Well, how
can it not be, if nothing with those characteristics had been around
before in evolutionary history? I mean, take bird-song. Is that an
autonomous faculty? To the best of my knowledge it pretty much is, and it
has little to do with other so called systems of communication, like
whatever octopi do, or the bees, or even plants (which in very serious
respects communicate too). That approach seems to me to be rather
hopeless, at least at this stage of our understanding of “communication”.
Many, the majority of interesting things you see out there, are pretty
autonomous. And syntax is nothing but another such autonomous mechanism,
organ, or whatever you want to call it. Moreover a pretty recent one, it
seems. So I really don’t see even the source of the criticism to the HCF
thesis. And yes: I do think the thesis – if it can be called so, I’d
rather refer to it as a program, or even a set of questions – is the
result of progress in syntactic theory. To start with, we now understand
recursion much better than we did half a century ago. And all the notions
I alluded to in the previous paragraph, some coming from parsing, others
from notions like linearization, and perhaps even the minimalist emphasis
on these kinds of features of the system manifesting themselves optimally.
I should add, by the way, that another criticism that is often raised in
this regard is the contention that biology is not optimal. This hides a
misunderstanding, first, of the sense of optimality we seek within
minimalism. It is emphatically not functional optimality of the classical
sort in New-Darwinism. Indeed, most of us are probably prepared to admit
that the language faculty is pretty sub-optimal in a variety of
functionalist regards, for instance with respect to center-embedding or
the issues arising from garden paths. But the form of the system is where
you see, if you wish, structural optimality, of the sort you expect in
physics. That suggests that the very physics of the system are involved in
that optimality, if the observation turns out to be correct. Now, does
biology have optimality of that sort? You bet. Think, for instance, of
the scaling law that West, Brown and Enquist have beautifully modeled in
fractal terms or for that matter the old observations about Fibonacci
patterns that go back at least to the Renaissance. Only a dull professor
can fail to see the structural optimality of a sunflower! And the thing
is although the same mathematical function describes what you see in a
peacock’s display, or in a virus coating. Note: in each of these instances
the arrangement serves a totally different function: maximal display of
features for I guess – photosynthesis in one instance, ventilation in
another, whatever’s going on with viruses, and so on.
And that’s just nature out there, around us, all over the place. I don’t
even think it is worth emphasizing this point, but obviously it seems
contentious, still, in many camps – which I find truly perplexing. The
questions go back to Aristotle if not the Pre-Socratics, and were fully
articulated by Leibniz, Goethe, and the great morphologists of the
eighteen and nineteen centuries. And in a nutshell, although obviously
Darwin gave as a great paradigm to talk about all of this, he didn’t – he
couldn’t – give us all the answers, whether they would lie in biological
workings within genes, chromosomes, cells, individuals, populations, and
their respective dynamics, and basic physico-chemical and computational
constraints, ultimately mathematical restrictions, and on and on. It’s a
“complex” system, life is. Is it really surprising that this wonderful
array is at play? That in some areas it is really serendipitous while in
others it’s truly elegant?
AK: A significant aspect of your research in recent years has been the
puzzle of language evolution. In a paper with Massimo Piatelli-Palmerini,
‘The Immune Syntax: The Evolution of the Language Virus’, you make some
interesting conjectures about the evolution of language. In the paper, you
make the suggestion that humans (with FLB and minus FLN) would have had a
“proto-language” [that] allowed for elementary grounded messages,
involving indexicals (or names) combined with immediate locations, or even
salient characteristics, an “ability, in its most rudimentary form, is not
specifically linguistic.” The emergence of language, you conjecture, is
the emergence of Narrow Syntax, and a consequence of the transformations
that such proto-language(s) underwent once the changes in the human
version of the gene FOXP2 gave humans a significantly improved
“phonotactic kit”, hence a massively expanded phonetic vocabulary.” My
question is how the first moment of language emergence is to be conceived.
As Richard Lewontin has argued, this moment is extremely difficult to
conceptualise: “Evolution by natural selection occurs when individuals
within a species possess a trait that gives them a reproductive or
survival advantage within the species that lack the trait. It is an
explanation of how a new trait spreads within a species, not how the
species may replace other species once a trait has been incorporated. Just
because a trait may be of advantage to a species when all of its members
possess it, it does not follow that a single individual who first showed
the trait in the species would leave more offspring. Thus a species that
possesses linguistic competence may indeed take over the earth as a
consequence of the technological and managerial capabilities that are the
result(s) of language, but in a species lacking linguistic competence,
the rudimentary ability to form linguistic elements by a few individuals
may be taken as a sign of difference that causes them to be expelled or
even killed.”
As far as I can judge, this first moment – and the mortal consequences
for the first human+FOXP2 -- remains as problematic for your fable as
anybody else’s. While it is not that Lewontin’s doubts necessitate a bar
on all speculation, the point is an important hurdle for all such
conjecture. Your thoughts?
JU: I couldn’t agree more with Lewontin’s worry. I entirely share it, and
this is actually one of the main reasons Massimo and I went into the virus
scenario. (The idea of the virus, by the way, must be credited to William
Burroughs, although in his colorful style he qualified it as “from outer
space” – which I have no evidence for, one way or the other.) In that
piece you mention, we didn’t develop that in any way, but in later pieces,
and the book we’re now writing, we fully embrace the virtual necessity of
the virus scenario to address the Lewontin worry. We want to cut the
Gordian Knot: produce a whole sub-species with the relevant
characteristics, literally. In short, if you get a population infected by
some species specific and tissue specific virus, and this alters their
genome in relevant ways, then in a matter of one generation (not eons) you
have the population you need. That way they cannot be “expelled or even
killed”, or at the very least they could fight back as a unit.
Now, the question here is how it is possible for a virus, or any such
“horizontal” transmission of genetic material, to do the trick. But this
we know: we have a text-book instance in the adaptive immune system, which
is relatively recent (going back to our common ancestor with jaw-endowed
fish, like sharks). Apparently we didn’t get this system through normal
“vertical” genetic inheritance. It came out as a result of a massive
infection which did not kill us (“us” meaning our ancestor with sharks),
but made us stronger. This is rare, of course – in this instance because
the innate immune system, which we share with our common ancestor with
bacteria, would have attempted to prevent it – but it is not impossible in
the least. In fact, more and more it seems clear that much of what looks
like “junk DNA” has this sort of origin (about 50% of our genome in the
human instance, by present count – and growing). So our conjecture is that
some insertion of that sort may have given us a crucial element in the
emergence of the language faculty, given a sub-species with roughly the
properties your question mentioned. (If you wish, chimp-like Theta Theory
or whatever else turns out to be the correct “proto-language” step, if
that’s even the right notion.)
This, incidentally, is testable. In particular, if something along these
lines is biochemically real – not just a metaphor – then one should see
its mark in the junk DNA of the organism, in relevant regions. This is the
reason I have become more and more skeptical that FOXP2, in particular, is
the smoking gun here. I say that because, although to my knowledge nobody
has yet examined its non-coding region (the Leipzig group analyzed the 3%
or so that codes the protein), judging from the similarities with chimps
on the segments examined, this doesn’t look like a “viral-infection” kind
of variation, which should be quite massive. FOXP2 is actually one of four
genes in the FOXP family, all regulating genes, one actually involved in
the immune system (FOXP3, if I recall correctly). Apparently they all
“talk” to each other, so they probably, to same extent, act jointly. Which
is to say two things: (i) that a viral insertion on any of these genes
would arguably have the desired effects, and (ii) nobody has the foggiest
idea how it is that these genes, any of them, work, let alone how they do
so in combination. I mean, say Ullman is right and somehow these guys (or
at any rate FOXP2), does regulate some crucial aspect of operational
memory. Precisely how is that done? No clue, and no obvious way to go yet,
as it is not clear, to start with, what the hell is operational memory,
what sort of “thing” it is, and where it resides, and so on. Let alone how
it affects computations of various degrees of complexities within the
Chomsky Hierarchy. So for all those reasons Massimo and I like to talk of
Mr. X, meaning some gene (or gene sequence) that somehow got infected, as
per the logic of the Lewontin worry. Again, in principle this guy could be
found, in part by comparing those regions of the human genome with lots of
transposon activity (a landmark of “horizontal” transmission) to
comparable chunks of the chimp genome, to find discrepancies. This is
being done, I believe, although not everyone out there focuses on “junk
DNA” – I think in fact few scientists do.
At any rate, we expect Mr. X to also give us another bizarre property of
human language: atomism, which is probably at the heart, both, of our
amazing ability to have our lexicons explode by age three or earlier, and
the capacity to turn something into a symbol, or if you wish, less
poetically, a “bag of tricks”. I mean, I could, in Goedel fashion really,
take any chunk of the table in front of me and bag all its contents
(literally) and think of that as a symbol. Or the contents of half this
page, or whatever else you want. That’s a distinctive human ability, to
the best of my knowledge, which I think we are now in the position of
understanding a topology for. (I happen to think this has to do with the
ability we also have to transform Euclidean spaces into non-Euclidean
ones, by warping them.) So perhaps Mr. X has to do with that sort of
ability, which we got infected for. This is less crazy than it may seem.
It all depends on exactly what sort of space codes memories, and I mean
space not just in the obvious physical sense, but actually also in the
mathematical sense, a set of local relations. Some manifold, if you wish,
that somehow allows us to store information (where “us”, again, might be
quite an ancestral species); “information bubbles”, as it were. Needless
to say, for a story along those lines to be even remotely correct,
perhaps the place where the presupposed memory resides is not neural
networks (where it is unlikely that matters of direct viral infection,
topological warpings of the protein-folding sort, and so on, would happen,
at least literally). More likely it should be found, as Randy Gallistel
has occasionally remarked, “at the bottom” (to use Richard Feynman’s quip
about there being “room” down there…). This is within neurons, not outside
them or in their interconnections. I know this does sound crazy, but the
truth is that the cell’s guts are a priori much more promising as a model
of what the Turing architecture may look like in a “biological computer”
than anything else we’ve found so far. I mean, there is a place where “a
processor and a memory tape” can be quite literally encountered, right?
But anyway, if anything like this is even remotely on track, then FOXP2
was probably recruited by the consequences that Mr. X may have had in our
brains, whatever those really were.
That’s what I’d say from my present perspective: that Mr. X somehow helped
us evolve the characteristic atomistic bagging of symbols, which probably
related to our ability – as children, lost at puberty – to literally
“latch” onto new words upon a trivial exposure, of one or two token
instances. It’s almost as if the whole network of Universal Grammar were
deployed by the child every time a new word comes in at the right context,
and then some sort of imprinting forces the virtually permanent storage of
that word in that brain. As I say, this ability doesn’t reach into
adulthood – try learning a new word with one exposure, the way my four
year old does! Anyhow, it is not hard to fathom that this is what may have
had a viral origin, and that as a consequence we had a vocabulary
explosion that required phrasal combinations, the way Martin Nowak and
Partha Niyogi have argued, and then parsing all that wonderful mess would
be harder than parsing “proto-language”. At that point I could fathom that
the ancestral FOXP2, used by birds for parsing complex songs out of simple
strings of sounds, could have come in handy. Nature is full of clever
moves like that, for instance when forming eyes several times anew, but
with some of the same basic genes, like the Eyeless gene, stuff that forms
the proper invagination to capture images, etc. There may be
different ways to do this (the lens could be outside the, as it were,
“camera obscura” or inside, simple or complex, etc.), but the basic
tool-kit is essentially the same.
AK: In your conjecture, the direct effect of FOXP2 is the birth of morphology, which is metaphorically, a viral intruder. The introduction of morphology in turn triggers an immune syntax response to eliminate it, and over at least a generation, “this sort of response, which forces the system into an antigen-antibody dependency that only a context-sensitive system can model (one where the derivational history can be accessed…) amounts to the fixation of a morphosyntactic parameter.” How would you respond to the question that this conception sees the perfect Language to be the proto-language, and it conceives of NS as the means of eliminating the noise created by FOXP2? If so, why would a competing individual with just FLB not be preferred?
JU: Well, two things. First, that particular scenario was one of the two
we outlined in our recent work – the other one, more literal, I just
discussed. But in either one, you are quite correct in noting that the
role morphology is given is of an “imperfection”. In effect, this is the
Viral Theory that Chomsky introduced a few years ago, and people like
Howard Lasnik and his collaborators have interestingly explored. So we are
using “viral” in two ways here. One is the literal, biological meaning.
The other one is a linguistic viral, who I myself introduced when
presenting Chomsky’s theory in Rhyme and Reason: the system must expel (a
kind of) morphology – and do it fast. I suggested that this looks like
what the adaptive immune system does in antigen-antibody interactions. My
hope was, and still is, that eventually this metaphor may be reduced, and
perhaps even that some of the mysteries of “uninterpretable features” can
be understood this way, or at least approximated rationally. For it should
be clear that these sorts of features are quite peculiar, particularly in
a system that seeks perfection in the interfaces of sound and, especially,
meaning. What is “uninterpretability” doing there?
So let’s go to the bottom of this issue: Do I regard that as a problem?
This is a deep one, and Chomsky himself, over the years, has gone back and
forth on the matter (“uninterpretable features” for him being an
imperfection at points, and following from “virtual conceptual necessity”
at others). Of course I have no idea which of these opposing views is the
correct one, but I really have no conceptual issues with the
“imperfection” one. Put it this way: if there were a perfectly perfect
universe, with, say, as many protons as anti-protons, the interactions
themselves would annihilate the known universe. It is because,
apparently, one in I don’t know if it is ten billion interactions is
imbalanced in the direction of matter that we have the characteristic
asymmetry that gives us “us” (where “us” is now everything, quite
literally).
Or take another familiar example, those pretty eddies that Prigogine
studied: you have a water flow, augment the velocity of the fluid, drop a
pebble, and bingo! Eddies (in well understood conditions, mind you, which
eventually lead to turbulence if you continue augmenting the velocity of
the fluid). Take away the pebble, the “imperfection”, and there goes your
system. So anyway, sometimes imperfections are actually necessary to
produce a certain structure in terms of the reaction the system provides.
The reaction is perfect, but the imperfection is not part of the system. I
don’t see a problem with that approach, and suspect that – while the
“total perfection” is a good one for so-called conservative
phase-transitions – this “limited imperfection” is a good one for
so-called dissipative phase-transitions.
I don’t know, call me old-fashioned – or just old – but I kind of like it
that imperfection too has a role in all of this, no matter how small.
Again, God knows if any of this is true, but my ego gets a nice lift when
I figure that even imperfect things have a prayer. But there we’d go into
my politics, and this is probably a good place to end.