Professional Documents
Culture Documents
in Search of A Mechanism From The Brain To The Mind A Conversation With Chris Frith 1St Edition Howard Burton Online Ebook Texxtbook Full Chapter PDF
in Search of A Mechanism From The Brain To The Mind A Conversation With Chris Frith 1St Edition Howard Burton Online Ebook Texxtbook Full Chapter PDF
https://ebookmeta.com/product/pushing-the-boundaries-a-
conversation-with-freeman-dyson-1st-edition-howard-burton/
https://ebookmeta.com/product/a-matter-of-energy-biology-from-
first-principles-a-conversation-with-nick-lane-1st-edition-
howard-burton/
https://ebookmeta.com/product/ocean-enlightenment-a-conversation-
with-edie-widder-1st-edition-howard-burton/
https://ebookmeta.com/product/astrophysical-wonders-a-
conversation-with-scott-tremaine-1st-edition-howard-burton/
Our Human Variability A Conversation with Stephen
Scherer 1st Edition Howard Burton
https://ebookmeta.com/product/our-human-variability-a-
conversation-with-stephen-scherer-1st-edition-howard-burton/
https://ebookmeta.com/product/beyond-mirror-neurons-a-
conversation-with-greg-hickok-1st-edition-howard-burton/
https://ebookmeta.com/product/inflated-expectations-a-
cosmological-tale-a-conversation-with-paul-steinhardt-1st-
edition-howard-burton/
https://ebookmeta.com/product/indiana-steinhardt-and-the-quest-
for-quasicrystals-a-conversation-with-paul-steinhardt-1st-
edition-howard-burton/
https://ebookmeta.com/product/understanding-adhd-a-conversation-
with-stephen-hinshaw-ideas-roadshow-conversations-1st-edition-
howard-burton/
Ideas Roadshow conversations present a wealth of candid insights from some of the
world’s leading experts, generated through a focused yet informal setting. They are
explicitly designed to give non-specialists a uniquely accessible window into
frontline research and scholarship that wouldn’t otherwise be encountered through
standard lectures and textbooks.
Over 100 Ideas Roadshow conversations have been held since our debut in 2012,
covering a wide array of topics across the arts and sciences.
Introduction
The Conversation
I. Becoming a Psychologist
II. Probing Agency
III. The Active Brain
IV. Ideal Bayesian Operators
V. In Search of a Mechanism
VI. Humanistic Hubris
VII. Free Will
VIII. The Very Big Picture
IX. Final Thoughts
HB: I’d like to talk a little bit about your intellectual origins. And I’d like
to ask you specifically about your influences in psychology—I know
that you’ve written a little bit about that—but I want to go further back
than that and begin with your interest in science writ large. My
understanding is that you began reading natural sciences at
Cambridge, and I was curious to know if science was an all-consuming
passion of yours from an early age.
CF: Well, certainly from an early age, I was very interested in nature and
doing birdwatching, but also going up into the attic and mixing things
together and seeing what happened. And I’m told that at the age of 12
when asked what I wanted to be I said, “A research scientist”.
HB: Really? From the age of 12?
CF: But I don’t think I knew what that meant, because certainly no one
in my family was like that. My father was a classicist, so I did lessons in
Greek. And in fact I was the last generation who, in order to read
science at Cambridge, had to do a Latin exam—going back to the good
old days when everything was in Latin. At school, I basically did maths
and physics, which I was good at. But in my gap year, I learned about
this mysterious subject called cybernetics, particularly as applied to
people—which is sort of control theory.
So I arrived at university, and of course at Cambridge all you can do is
natural sciences—there was no such thing as cybernetics, the nearest
thing was psychology. But you couldn’t do psychology in your first year.
So my background is in applied maths, physics and what we used to call
then “min and crys”, which is mineralogy and crystallography. Then I
did psychology as what they call a “half-subject” in the second year, and
then I specialized in it in the third. But that means that my
undergraduate-level psychology is based on one and a third years of
work. And in fact, the “min and crys” turned out to be very useful
because you had to learn about three-dimensional maps and things like
that, which suddenly became relevant 20 years later with brain
imaging.
My plan at the end was to go on and do work with someone called
Donald MacKay, who was one of the very early people doing
information theory as applied to the brain—what we’d now call
“computational neuroscience”—and in fact I had spoken to him and he
had taken me on to do a PhD, but luckily I failed to get a good enough
degree, which is a constant storyline for me.
So I went and did clinical psychology as a way back into doing
research—which is also fascinating because today you would do
research as a way into clinical psychology, but then it was the other way
around. There was only one course at the time, which was in the
Maudsley Hospital in South London, a 13-month course in abnormal
psychology. So I’m now technically an abnormal psychologist. And they
rapidly said, “Yes, yes, very good. We think you should not see patients.
Why don’t you do a PhD?”
And I agreed with that because in order to see patients you have to
believe that you can really help them, and I wasn’t sure that I could. In a
way there’s a conflict between doing research—where you have to
doubt everything—and being a clinician where you have to be
confident. There’s always that conflict I think.
So I did my PhD with Hans Eysenck, which was quite interesting, and
it sort of set off from there.
And one of the very lucky things at that time for me was that—as I
claim—we had the very first computer in a psychology department in
the country, which was in 1965. So I learned to program in machine
code as part of my PhD.
HB: I imagine that, aside from whatever you learned in particular at the
time, doing that played a more general role in exposing you to
computationally-oriented thinking and made you more receptive than
others, perhaps, to the idea of moving even further in that direction
later on.
CF: Yes.
HB: Getting back to what you were saying before, what was your father
the classicist’s reaction to your announcement at the age of 12 that you
were going to be a research scientist? Was he pleased? Bemused?
Intrigued?
CF: Oh, I think he was pleased. And in fact, the chap who told me about
cybernetics was a friend of my father who was another teacher—he
was in biology or something, I believe—and we used to see him quite a
lot. He was a very entertaining chap who would regularly tell us all
about the current developments in science.
My parents were very open-minded because while I said that I
wanted to be a research scientist, my younger brother—now known as
Fred—announced that he was going to be a rock guitarist.
HB: That was okay too?
CF: That was okay too—and he still is.
HB: And when you were at school and interested in the natural sciences
and mathematics before you went off to Cambridge, did you have any
particularly influential teachers who stimulated you?
CF: Well, I guess the most influential was the sixth form maths teacher. I
just loved the idea that you could solve things with these equations.
And you knew for certain when you had done it correctly, as opposed to
writing an English literature essay, say, where different people naturally
had different views and opinions on things. In fact, as it happens the
English teacher and I seemed to disagree on all possible topics, so that,
in a sense, switched me over.
HB: Was he a model of the English professor in Making Up the Mind:
How the Brain Creates Our Mental World?
CF: Quite possibly.
It’s interesting because at university I did particularly well in “min
and crys”, because the lecturers there were really strict and told you
what to do and made sure you did it. But the psychology was
marvellous fun: I was particularly lucky because I had the very good
fortune to be lectured by Richard Gregory and Donald Broadbent—who
were the main people who started off cognitive psychology—Larry
Weiskrantz—who was very big in monkey physiology—and my direct
tutor was someone called John Steiner, who at that time was, well, I
guess what you call a “born-again Skinnerian”, so he was completely
behaviourist. So I got this extraordinary mixture of behaviourism plus
the new thing: cognitive psychology.
HB: I imagine that there were many behaviourists at the time.
CF: Well, in Cambridge, it never really caught on to quite the same
extent as in the States, for example.
And interestingly John Steiner subsequently became a psychoanalyst,
which always fascinates me because I think there’s an interesting
relationship between behaviourism and psychoanalysis, because both
are about how everything is determined by your early experiences in
some sense.
HB: You gave this self-effacing anecdote about how you hadn’t been
successful enough at Cambridge to enable you to move directly into
what later became computational biology—and I’m not sure I’m going
to take that a hundred percent on your word—but I’m guessing that
before you began that clinical program in abnormal psychology at
Maudsley Hospital, you were quite interested in abnormal psychology
or aspects of different psychological conditions?
CF: I don’t quite remember how it started, but I was particularly
interested in schizophrenia and I read lots of books about it. Later,
when Hans Eysenck was producing the second edition of his enormous
handbook of abnormal psychology his students were handed out
different chapters and I got the one on perception, which was mostly
about schizophrenia. What particularly fascinated me then, and still
does, is the problem of hallucinations and delusions.
It’s easy enough to understand in principle if you’ve got affected
regions of your brain why you become blind or deaf or can’t understand
a concept or something, but it’s very difficult to understand why you
start seeing things that aren’t there or believing things that are
obviously not true—although scientists are quite good at that too, as it
happens, but that’s another matter.
And so, I was always interested in questions like, Can we think about
a mechanism? and, How do we relate this to normal functioning? What is
it that could go wrong in normal functioning that can make you start
seeing things that are not there or hearing people talking about you,
when they’re not?
HB: And was this perspective—focusing on brain functioning and
specific mechanisms in the brain—was this something that was
somewhat iconoclastic at the time?
CF: Well, thinking about the brain, for me, came somewhat later. But
certainly the early cognitive stuff was very much about thinking about
the mechanisms or cognitive processes or information processing that
underlies all our different abilities.
So that was iconoclastic in relation to behaviourism, because you’ve
started thinking about what’s inside “the black box”, but at that stage
we were just talking about cognitive processes. And of course the
neuropsychologists, whom I came into contact with a bit later—people
like Elizabeth Warrington and Tim Shallice and John Morton—were
interested in asking, If somebody has a lesion in the brain, what does that
tell us about cognitive processes? They weren’t really interested in what
it tells you about the brain, at that stage.
HB: So tell me how your work on schizophrenia evolved and what the
prevailing views on it were at the time.
CF: Well, after my PhD I did several years as a postdoc—which was
marvellous because I more or less did whatever I liked: in those days
money was not a problem somehow, which I never quite understood,
but that’s the way it was. But then I joined this Medical Research
Council unit run by Tim Crow, where specifically the main question we
had to answer was, What’s the biological basis of schizophrenia?
And this was somewhat iconoclastic because there was this
extraordinary distinction between “functional psychosis” and “organic
psychosis” in old-fashioned psychiatry. According to this view, an
“organic psychosis” is where there was clearly something wrong with
the brain, while a “functional psychosis”—which was how
schizophrenia was regarded—either meant, There must be something
wrong with the brain but we don’t know what it is, or, There’s nothing
wrong with the brain.
So it was almost as if this was not a brain disorder—and you had
people like Ronnie Laing, who was saying that it was caused by society
or something like that—it was a response to an abnormal society—or
you had other people saying it’s caused by peculiar interactions in the
family. And both of these ideas, in a sense, faded away because there
wasn’t much empirical evidence to support them. And one of the first
things we did in this unit with Eve Johnstone was one of the first ever
structural brain imaging of schizophrenic patients using modern
technology, which in those days was what they called CAT-scans.
HB: When was this exactly?
CF: It was published in 1976. And these scans showed that chronic
patients with schizophrenia had enlarged ventricles, which was not due
to any treatment or things like that. And that, I think, had a big impact
on switching the belief towards the view that, This is really a brain
disorder which we need to explore.
But just to give you an idea of the problems, there’s this famous DSM
statistical manual (Diagnostic and Statistical Manual) for deciding how
you diagnose people. And in DSM-3, which is what we had at that time,
you have counter-indications. It said, To get a diagnosis of schizophrenia
you have to have these hallucinations and delusions and various other
things, but there must be no known brain disorder. So as soon as you find
any brain disorder it ceases to be regarded as schizophrenia. This has
changed.
HB: Quite the question-begging that.
CF: Yes. This has changed.
And the other thing we did—which I’m still very proud of—involved
research on anti-psychotic medication.
They first discovered the anti-psychotic medication in ‘55, I think, by
accident. They all turned out to be dopamine-blocking drugs. So we did
an experiment with one of the standard treatments that was something
called flupenthixol. Flupenthixol is interesting because it has two
isomeric forms: one of these forms blocks dopamine, while the other
one—which has lots of other effects—doesn’t; so you could do a very
tight comparison. And it turned out that indeed, yes: the one that
blocked the dopamine receptors reduced severity of symptoms over
the course of four weeks, whereas the other one was no different from
placebo.
So there are various interesting things in that result. First of all, it
seems to be very specific to this dopamine blocking—which again,
relates it to the brain. And that still seems to be true: I don’t think
they’ve progressed that much on that score.
Secondly, there were positive results for everyone, including the ones
on placebo.
And thirdly, that this effect was specifically on hallucinations and
delusions and not on the so-called negative symptoms: the retardation
and poverty in speech and things like that. So in a sense, you were
finding an effect, of dopamine blocking, which is clearly a very low-level
brain type intervention with an extremely high level of subjective
experience. So the key question then became, How do we bridge the
gap?
Questions for Discussion:
HB: That brings up the important idea of how to bridge low-level and
high-level gaps in all sorts of ways, including, but not limited to
schizophrenia. But I’d like to stay with schizophrenia for a little while
and continue to examine the evolution of our understanding of that
condition.
But first, I’d like to go back and talk about the societal understanding
and appreciation of this disease, because my sense is that has also
changed enormously since the time when you first started doing your
research. Not to imply that everybody’s got a full understanding of the
situation today, but my sense is that in the popular consciousness the
word means something quite different today than it did then. Is that a
fair statement, you think?
CF: Well, I don’t really know what that word meant to people in the 50s
and 60s, if they knew the word at all. I mean, they knew that there were
“mad people” who lived in these big asylums, and you occasionally saw
them on the street talking to themselves and they probably thought
they were a bit dangerous. And I’m not sure that, on the whole, that’s
changed all that much.
I mean, I’m not quite sure how true it is nowadays, but certainly in
the recent past if you say were to say the word “schizophrenia”, people
would think, Split-mind and multiple personality—which of course is
completely wrong.
It’s a rather funny term, because the word “schizophrenia” does
actually mean “split-mind”, but it meant that there was a split between
your different faculties, like emotion and reason and motor and
perception. A classic example would be what was sometimes called “a
peculiar effect”: for example, you’d say to the patient, “Your mother is
very ill”, and he’d laugh. It would be that sort of splitting, not the
multiple personality that people believe about.
And then, of course what happened was that the big asylums were
closed down so that these people were no longer “over there”
somewhere—they were in hostels or on the streets or whatever. The
other thing that has happened in my lifetime, of course, is that in the
olden days if you were to see someone walking on the street talking to
themselves, they were thought to be schizophrenic. Now they’re most
likely to be on the phone.
HB: And they may or may not be schizophrenic in addition.
CF: That’s right.
HB: Let’s get back to this question of our scientific understanding of
things. Perhaps you could just give me a sense of how our
understanding of schizophrenia has evolved in the past 30 or 40 years
and why?
CF: Certainly—but of course this will naturally be my perspective on
things.
As I said earlier, I was interested in this problem of hallucinations
and delusions, and there’s one particular delusion that I became very
interested in, which is the delusion of control, which occurs in about
16% of cases. This is where the patient says, “I’m not in control of my
actions. Some alien force is causing me to do things”. It could be even
simple things like lifting up the glass and drinking.
It’s very difficult to find examples, interestingly—you’d think people
would collect these symptoms, but they all come from more or less
three papers, one of which is mine.
So you have a patient saying something like, “The force is causing me
to move”—this is pre-Star Wars, incidentally—and there’s also
something called “thought insertions”, which is even more peculiar and
yet to be solved, where the patient says, “There are thoughts coming
into my mind, which are not mine”. It’s very odd, because how can a
thought be in your mind and not be yours? After all, is there a little label
that comes with each thought saying “mine”?
But when you think about action, that’s much easier, because, for
example, if I hear a voice, it could be me talking, or it could be you
talking to me, and in a sense we need a label so that I know what I’m
hearing is my voice and not yours.
And this takes us right back to my big hero, Hermann von Helmholtz,
who pointed this out in relation to eye movements. When I move my
eyes, obviously things jump about on my retina so there’s movement
on the retina, but it’s due to me. And I have to be able to distinguish
between movement on the retina due to me and movement on the
retina due to something actually moving in the world.
And he basically said—I can’t remember his exact terminology
—“Because there’s a message involved—you’re sending a message to
your eye muscles to move the eye—you can use that signal as a way of
determining what the corresponding movement is due to, whether it’s
you or the world”. And he has this simple experiment: if you poke your
eyeball with your finger, carefully, to make it move, the world appears to
jump about, but you can use that signal to your eye muscles to
determine that the phenomenon in question is internal to you and not
in the world.
So I took that up and thought, Maybe what goes wrong in
schizophrenia is that this signal—this normal signal that tells you that
it’s your movement or your action—doesn’t arrive for some reason.
And we did various experiments, but the one I particularly like was
done by Sarah-Jayne Blakemore when she was doing a PhD with me.
She recognized that this obviously relates to tickling, as you can
immediately see, because we know that you can’t tickle yourself—Larry
Weiskrantz published something about this in Nature in the early
1970s. The question is why, and the answer is gained using this
Helmholtzian argument that because you can predict exactly what
you’re going to feel when you tickle yourself, it’s suppressed.
So Sarah-Jayne and I took this into the scanner and we had various
clever bits of equipment so you could tickle yourself directly or
indirectly. And she showed that if you introduce a delay of a hundred
milliseconds or so—if you’re holding a rod and tickling yourself with it,
you can introduce a delay—then it feels more ticklish.
But the nice thing was that she then tried this out on people with
schizophrenia. And indeed for the ones with delusions of control, if you
ask them to rate how ticklish it feels, there was no difference between
them tickling themselves and Sarah-Jayne tickling them. And we did
more sophisticated things after that. So that seemed to fit: we were
beginning to come up with a more mechanistic story of what might be
going wrong.
HB: So as I understand it there’s a natural focus on this question of
agency. When you and I are lifting up glasses of water or deciding to
look out the window or whatever, we have no doubt whatsoever that
it’s we, broadly defined—and, hopefully we’ll get to what that means in
a moment—who are actually doing that, but that the idea is that people
at least with some particular form of schizophrenia have a difficult time
with this whole concept.
CF: Yes. And there have been similar studies demonstrating this. Judith
Ford did some very nice work on hallucinations where she showed that,
while we normally suppress the sound of our own voice—which you
can measure with EEG and so forth—this was not happening to the
same extent in people who are prone to delusion. So you’re getting a
similar story.
And the obvious question now is, So what about dopamine?
In parallel with this, there were very exciting developments in the
dopamine story—mostly, I think, due to Wolfram Schultz at Cambridge,
who is looking at monkeys. He showed that there are neurons in the
middle of the brain—in the ventral tegmental area—which release
dopamine. He was measuring activity in these neurons and he could
show that they were actually predicting reward. In other words, if the
monkey gets an unexpected reward, these neurons fire.
You can then do some conditioning, so that the monkey learns that a
certain signal—a light flash—tells it there’s a reward coming.
Now this light flash, of course, is a signal of unexpected reward, so
what happens is that the neurons would fire immediately after the
monkey sees the flash rather than receiving the reward itself, since the
reward is now entirely predicted. On the other hand if the reward
doesn’t come as expected, then the firing rate of those neurons actually
decreases.
So you have a very nice mechanism here which is telling you whether
you’re being rewarded or not. This led to the development of early
forms of computational neuroscience, where you can have a very nice
story of how learning occurs on the base of whether your reward goes
up or down, and you can then learn to attach rewards to signals and so
on and so on.
It used to be viewed as simply a reward-mechanism—dopamine was
released when you were rewarded—but it’s now much more
sophisticated. It’s actually a signal of reward prediction error, as they
call it, which is used in learning.
And to this you add a so-called Bayesian perspective: you have prior
expectations and you have evidence, and then you update your model
of the world on the base of it.
HB: I’d like to get into those Bayesian aspects you were just mentioning
in more detail shortly, but first I’m going to back up and ask a few more
general sorts of questions.
Correct me if I’m wrong, but my sense is that through your work in
schizophrenia—or at least your work in schizophrenia combined with
other work—you’ve been led to appreciate the important way that the
brain acts and conditions information in such a way that we can learn
from it, namely this idea of prediction and reinforcement.
So at some level, it seems to me that you have used schizophrenia as
a window to better understand what’s occurring more generally, insofar
as you’re saying, “Oh, in these circumstances it doesn’t seem to be
working quite as well as it should. Let’s try to understand what’s
happening there.” Is that a fair way of looking at it?
CF: Yes. I think certainly one of my basic beliefs would be that we
should study the abnormal systems in order to learn about how it
works in the normal case. In a sense the abnormal system is somewhat
of a “simplification”, and I’ve been very influenced by
neuropsychologists studying patients with known lesions.
For example, alongside what I’ve just been talking about was the
discovery of things like “blindsight” and the famous patient, DF, that
David Milner and Melvyn Goodale studied. This is a patient who, due to
carbon-monoxide poisoning, has damaged her temporal lobe. She is
technically assessed, I believe, as effectively blind: she can see, but she
can’t recognize objects on the basis of their shape.
The fascinating thing that Milner and Goodale recognized is that she
wouldn’t be able to tell you that this thing in front of me is a mug, and
she wouldn’t be able to tell you where the handle is, but she can reach it
correctly. That gives you the idea that there are these two roughly
independent streams, one of which is for recognizing what the things
are and one of which is for reaching and grasping; and in the normal
case they’re all tied up together and it’s very difficult to separate them
out, but in the abnormal case, you can start to see these fractionations.
Questions for Discussion:
1. What are some of the key assumptions behind the idea that a
close examination of abnormal cases can shed light on generic
brain processing mechanisms?
HB: I’d like to talk more now about our general picture of how the brain
is operating.
Let me start off with what I’ll call the naive view—which we
understand now is not the best picture of what’s going on—but for the
longest time, I think, people did look at things this way. And the naive
view seems to be we have these receptors that correspond to our
senses and sense data impinges itself upon us through these various
receptors, which is how we get information about the world around us.
There’s this rather awkward little step which is elided in all of that,
which is how this is actually processed in our mind’s eye, but if we just
forget about that for a moment, the idea is that we’re going around the
world with our eyes open, say, and so photons hit our retina and there
is corresponding electrical stimulation and so forth in our brain. In
other words the brain is somehow this big receiver of information.
CF: Yes.
HB: And my sense is that things are considerably more complicated
than that.
CF: That’s right.
HB: Hence my calling it “the naive view”. So perhaps I could get you to
just give us a clear sense of what we now believe and why, very much in
keeping with what you were just discussing.
CF: Yes. So I would characterize the earlier version as a “feed-forward”
version. The evidence comes from the senses. Then it goes to a higher
level area that determines the shape, that goes to a higher level area
that sorts out what object in particular it is, and so on.
Take reading. You can say that there are marks on the page, which can
then be interpreted as letters, then converted to words, then
recognized as sentences; and then in the old-fashioned diagram box
there are diagrams that denote “the place that sentences go when
they’re understood”.
Again, Helmholtz, I think, was the first to see that this is clearly
wrong.
It’s partly because he realized that it’s simply too long, in
physiological terms: even though nerve conduction is rather slow, the
time it takes to recognize what an object is, is ten times slower. And
realized that there was something he called “unconscious inferences”
that were going on—we get this experience that this is the object, but
we’re not aware of how much work the brain has done to arrive at this
point. So the interesting question is, What is this work, exactly?
And this is where the idea of “predictive coding” or “the Bayesian
approach” comes in. And there are two aspects to this.
The first is that you have to have a prior expectation.
From my past experience, for example, I have a very good idea of
what’s likely to be on this table beside me. And what I use the evidence
from my senses to do is to evaluate to what extent that prior
expectation was right or not. If it’s right, that’s fine. If it’s wrong, I have
to slightly change what I think is out there—which leads to the idea
that most of the time, since our prior expectations are right, we’re not
actually taking any account of what’s out there. I think I say somewhere
in my book, Making Up The Mind—which I probably stole from
somewhere else—that basically our perception is “a hallucination
mildly constrained by reality”.
A nice example of this, if you’ll allow me to jump about a bit, is the
phantom limb. How on earth can someone have a phantom limb when
there’s no limb actually there? You can say, “Well, what motor control
theory tells us is that when I perform an action, I have a prediction of
where my limb is going to be and what it’s going to feel like”. Most of the
time, my experience of the world is not what my limb is, it’s my
prediction about where it will be and what it will feel like. So the person
with the phantom limb still has all these predictions and things intact,
and that’s what I think results in the phantom limb phenomenon.
But the second point is that not only do we have expectations prior
to the evidence from the senses, but we spend a lot of time doing things
to the world. And this is where I’m very much influenced by my friend,
Daniel Wolpert, who proudly says when he goes to meetings, “I am an
engineer”—he’s now in the engineering department in Cambridge.
And he’s also, as I am, a motor chauvinist because prior to us, there
was a sense that everything was perception. If we go back to Hubel and
Wiesel, they would say, “We know a great deal about visual perception”;
and if one of them would draw a picture of the brain, most of it would
be the visual system.
But we would say in contrast, “No, action is what the brain is all
about. If you don’t have action, you’re going to die.”
Daniel has this nice anecdote he likes to tell—I’m not sure if it’s
actually entirely true, but it doesn’t really matter, it’s illustrative.
There’s some sort of creature like a sea squirt that, in its larval form,
swims about and finds food, but when it matures into an adult, it
immediately attaches itself to a rock and never moves again. And the
first thing it does is it dissolves its brain, because it doesn’t need it
anymore—just like someone who’s being given tenure, incidentally.
It’s worth mentioning that Helmholtz also pointed this out when he
described how we perceive how far away things are. Well, you can use
something called parallax, which is basically to move from side to side,
and you find that things further away move less than the nearer things.
So you’re using your action, you’re making predictions about what will
happen when you act, to find out more about the world.
HB: Yes. And getting back to this overall approach that I was
mentioning earlier of using abnormal circumstances or systems as a
window into normal brain function, in Making Up The Mind you
specifically highlight the role that visual illusions can play in helping us
get a deeper understanding of how our normal act of perception works
by examining precisely how it is somehow being interfered with
through these illusions.
CF: Yes, that’s right. As we were saying in this new formulation
everything depends on prior expectations; and there are, of course,
circumstances where your prior expectations are completely wrong,
and I think that most illusions are indicating where this point is. So
many of them depend on seeing something as being 3D when it’s really
2D, but a particularly nice one is something called the “Hollow-Mask
illusion” that Richard Gregory used to great effect.
This is where you have a hollow mask, but if you look at it from
behind when the nose is actually pointing away from you, you cannot
help but see it sticking out towards you. And if you mount it on a rod
and rotate it, you see that when it’s facing you it rotates normally, but
when it’s facing away from you and you’re looking at it from behind it
seems to start rotating in the other direction once the illusion kicks it
because you’re now misperceiving the nose as sticking out.
And I would say that this is due to have an incredibly strong prior
expectation that faces stick out which completely overrides the
evidence we are being presented with to our senses. Now what is quite
interesting is that in schizophrenia, this illusion is much less strong. So
it seems that there’s something peculiar about the way they integrate
their prior expectations in their evidence.
Questions for Discussion:
1. Why do you think that many of those who still cling to what
Howard calls the “naive view” of the brain work on the
neuroscience of vision? How might the field of vision be
regarded as “a victim of its own success”? Readers interested in
this issue might want to compare Chapter 7 of Minds and
Machines with Duke neuroscientist Miguel Nicolelis with
Chapter 3 of Vision and Perception with Stanford
university vision scientist Kalanit Grill-Spector.
HB: For the moment, however, I’d like to concentrate on the apparent
distinction between our brains and the processing that they’re doing,
and our conscious minds—a distinction that strikes me as a major
theme throughout Making Up The Mind.
I’m certainly willing to accept the fact that there are irrational people
—after all, the evidence seems to be overwhelming. Whether or not
that means they have irrational brains, as well as irrational minds, I
don’t know, but the point is that this distinction certainly seems to
exist for even reasonably rational people.
So if we focus on that we can now forget about focusing on Bayesian
probability per se—that was an example of trying to look at the
distinction.
When we talk about whether or not we’re convinced of something—
our beliefs, our desires, all the rest of that—we’re talking about
“ourselves”. We all have a fairly clear understanding of what that means
—even if we can’t specify it logically or physiologically—and that’s very
different, of course, than looking at brain activity in an fMRI machine.
Now, what I detect from you is some ambiguity—I’m not accusing
you of anything other than what every reasonable human being has
grappled with throughout the dawn of history, so this is not particularly
directed at you—but there is this obvious ambiguity that we’re all
battling with, it seems to me.
If we are materialists, we are naturally inclined to say something like,
“We don’t believe in a soul or ‘soul-stuff’; we believe that at some level
there’s nothing other than physical stuff out there, and therefore the
brain must cause the mind—and we also have all sorts of other evidence
for those conclusions, ranging from lesions to the brains to how people
behave under narcotics and so forth.”
So there are all sorts of reasons to believe that the brain and the
mind are causally connected, but the question is, Well, how does it work
exactly? Or even approximately, for that matter.
At one point in Making Up The Mind, you say admirably humble
words to the effect of, “This leads us to the question of consciousness, and
I’m not going to look so much at consciousness because that’s too
difficult. I’m going to look at what it’s for and go through an evolutionary
pathway and so on.”
But I want to put you back on the hook for a moment now and simply
ask you, “OK, look, we’ve got these two things: the brain and the mind.
How are they linked up?”
CF: Well, that’s what I’ve been thinking about, mostly, since writing that
book. And it’s very much to do with what I was saying before about the
personal and subpersonal and how they relate, but also I’ve become
more and more interested in culture. In some ways, the brain is not
enough on its own—it’s almost like a tool.
And one thing I think about is that, genetically speaking, our brains at
birth are no different from the brains of people from something like
200,000 years ago, when people were making these crude stone tools
and so on. But adult brains today I would suspect, are very different
from the brains of people 200,000 years ago, because the brain is very
plastic and culture affects the brain.
I was involved in the famous Taxi Drivers Study, showing that the
hippocampus, or portions of the hippocampus, of a London taxi driver
increases in volume as a result of learning The Knowledge—the
rigorous mental map of London that is required for all successful
drivers. I’ve been involved in a study which shows that Italian brains
are different from English brains because the spelling of Italian and
English is so different.
So there are innumerable things in modern culture that will make our
brains very different from what they were 200,000 years ago. A lot of
what we mean by “culture” is at the personal level: it depends on
communication, the interactions between people, which create
traditions and so on, that are then fed back into the system.
But to talk to people and describe our experiences and how the mind
works is quite difficult.
So this is becoming a bit speculative—there’s some nice work,
slightly controversial from Dijksterhuis and colleagues in the
Netherlands, where they show that making a complicated decision
which involves taking into account 12 different variables, you could do
it actually better if you didn’t think about it.
And there’s another study that says that if you have to recognize a
face and you’re asked to describe it, that actually makes you worse at
recognizing it. So the idea is that our subpersonal brain is extremely
good at handling a very complicated multi-dimensional structure, but
as soon as things get up to the personal level—which from my point of
view means we have to talk to other people about it—we have to
simplify it, we have to reduce the number of dimensions.
So we can do tricks, like making them richer. But in essence, we have
to reduce the number of dimensions. And if you choose the wrong
dimensions, you’re going to make the wrong decision. And I think that
might be the sort of thing that’s happening in these problem-solving
cases we spoke about earlier. But it’s this talking to each other—
including things like how the mind works—which creates culture and
feeds back into us.
HB: Let me just interject for a moment, because while I have no
problem whatsoever in being speculative, I don’t want to lose the
thread. So let me try to be a little bit more concrete.
I’m immersed in a culture. And as a result of this immersion, I need to
be communicating with you and others, and that communication
necessitates that I bounce ideas off you, predict how you might
respond and so forth, all of which requires me to use my wonderful,
ideal Bayesian operator brain.
CF: Yes.
HB: And by interacting with you and utilizing this prediction-
confirmation process—which presumably also includes some higher
level aspects of empathy and so forth—my brain is also evolving and
it’s changing its structure, resulting in things like, as you said earlier,
how Italians have a slightly different brain than anglophones, say.
So all of that, at least on a hand-wavy level, I’m okay with. But I don’t
know if that brings me any closer to this sense of how I’m linking my
“me”—my personal level—with “my brain”—the subpersonal level. Do
you see my problem?
CF: Yes, yes.
Well, the way I look at it is that there are signals coming up from the
subpersonal level, into the personal level, which we can actually talk to
people about. And likewise, when people tell us things that somehow
influence how the subpersonal level works.
So an example of a signal would be, I have a sense of effort: I feel how
hard I’m having to work to do a particular task. And in many tasks, the
longer I do it, the harder it seems to be. And I’m aware of this sense of
effort. And I can tell someone, “This feels very effortful”. And indeed, I
will say, eventually, something like, “I’m too tired. I can’t do this
anymore”.
Now there was a very nice experiment—not by me, first from Carol
Dweck’s group. They were investigating this phenomenon called “ego-
depletion”—which is very famous and well-established, but now under
attack—that if you do a difficult mental task or if you have to inhibit
yourself from eating nice food, it exhausts your mental resources and
you will have difficulty with another executive task immediately
afterwards.
Now what these people did is they added an additional group. So they
had similar experiments, but now there were two groups. One group is
told when you do a difficult mental task, it’s like a muscle and you will
feel tired and it will be difficult to do something further. The other
group is told when you do a difficult mental task, you will feel energized
and ready for more work. And lo and behold the people who were told
that they would feel tired, made more Stroop errors after the executive
task and the people who were told they would be energized, made
fewer Stroop errors after the task.
So that seems to me to be a direct example of how you being told
how your mind works, influences your behaviour at this low level.
HB: So that’s a sign of concrete interaction between these different
levels. But then, what I’d really like to know is what’s...
CF: The mechanism?
HB: Yes.
CF: Well, now this is becoming extremely hand-wavy.
There’s another experiment that I think gets us a bit closer to the
mechanism. You know that there are all these “common goods
games”—you interact with someone and you have to learn whether
they’re trustworthy or not. If you invest money and they give you some
money back then they’re more trustworthy, and if they didn’t give you
the money back they are less trustworthy, and you can have a
completely standard learning algorithm using prediction errors: if they
give you more money back, their trust goes up, and if they give you less
money back, the trust goes down.
You can see these prediction errors happening in the brain where the
dopamine is. That’s all fine. The interesting thing is that, if you tell
them, “This is a very trustworthy person”, they stop noticing the
prediction errors both behaviourally and in terms of brain function. So
they’re less influenced by the actual behaviour of the person, you see
less prediction-error activity in the middle of the brain and in the
Bayesian terminology if you say the mechanism here is that you’ve
been given the prior information that this is a trustworthy person, so
you down-weight how much attention you pay to their actual
behaviour.
HB: This is almost analogous to the tickling thing.
CF: Yes—that’s a good point, I hadn’t thought of that. With respect to
the question of mental effort and related issues, it’s a matter of
interpretation, but I think you can explain it all on these very high-level
priors (prior information). So the issue of ego-depletion is fascinating
because now almost everybody in the world who’s read the books and
knows about it has this high-level prior, but you can change this.
And there are other interesting, analogous experiments on free will,
where you can say to people, “No sensible people these days believe in
free will. Francis Crick has shown in his book that there’s no such thing:
everything is predetermined”. And if you tell people that, they believe it
—which you can assess through a follow-up questionnaire—they will
then be more likely to cheat on tests, they will become less helpful in
social situations—”
HB: Because they are now convinced that they’re not responsible.
CF: Yes. Another thing that is fascinating to me involves this
phenomenon called “post-error slowing”. If you do a reaction time task,
for example, after you’ve made an error you will slow down, because
you’re monitoring yourself and you’re saying “I’m doing it too fast”. And
it turns out that people who don’t believe in free will showed less post-
error slowing, and the amplitude of their readiness potential in the
brain becomes smaller. I think the prior here is about how much top-
down control an individual believes she has over her behaviour. So if
you now believe you have less top-down control—”
HB: Or none.
CF: Yes—or none—that will have a considerable impact. So I think
these are hints of the kinds of mechanisms that are involved, and my
current research project is precisely about that: trying to discover
something about these mechanisms.
Questions for Discussion:
Whilst we thus behold the entire failure of all that was foretold
against the system, it is a subject of just felicitation to its friends, that
all their anticipations of its benefits have been fulfilled, or are in
progress of fulfillment. The honorable gentleman from South
Carolina has made an allusion to a speech made by me, in 1824, in
the other House, in support of the tariff, and to which, otherwise, I
should not have particularly referred. But I would ask any one, who
can now command the courage to peruse that long production, what
principle there laid down is not true? what prediction then made has
been falsified by practical experience?
It is now proposed to abolish the system, to which we owe so much
of the public prosperity, and it is urged that the arrival of the period
of the redemption of the public debt has been confidently looked to
as presenting a suitable occasion to rid the country of evils with
which the system is alleged to be fraught. Not an inattentive observer
of passing events, I have been aware that, among those who were
most early pressing the payment of the public debt, and upon that
ground were opposing appropriations to other great interests, there
were some who cared less about the debt than the accomplishment of
other objects. But the people of the United States have not coupled
the payment of their public debt with the destruction of the
protection of their industry, against foreign laws and foreign
industry. They have been accustomed to regard the extinction of the
public debt as relief from a burthen, and not as the infliction of a
curse. If it is to be attended or followed by the subversion of the
American system, and an exposure of our establishments and our
productions to the unguarded consequences of the selfish policy of
foreign powers, the payment of the public debt will be the bitterest of
curses. Its fruit will be like the fruit
“Of that forbidden tree, whose mortal taste
Brought death into the world, and all our woe,
With loss of Eden.”