CHAPTER 1 The Problem

How does the brain produce adaptive behaviour ? In
attempting to answer the question, scientists have discovered two
sets of facts and have had some difficulty in reconciling them.
The physiologists have shown in a variety of ways how closely the
brain resembles a machine : in its dependence on chemical
reactions, in its dependence on the integrity of anatomical paths,
and in many other ways. At the same time the psychologists
have established beyond doubt that the living organism, whether
human or lower, can produce behaviour of the type called 4 pur-
poseful ‘ or ‘ intelligent ‘ or ‘ adaptive ‘ ; for though these words
are difficult to define with precision, no one doubts that they
refer to a real characteristic of behaviour. These two character-
istics of the brain’s behaviour have proved difficult to reconcile,
and some workers have gone so far as to declare them incom-

Such a point of view will not be taken here. I hope to show
that a system can be both mechanistic in nature and yet
produce behaviour that is adaptive. I hope to show that the
essential difference between the brain and any machine yet made
is that the brain makes extensive use of a principle hitherto
little used in machines. I hope to show that by the use of this
principle a machine’s behaviour may be made as adaptive as we
please, and that the principle may he capable of explaining even
the adaptiveness of Man.

But first we must examine more closely the nature of the
problem, and this will be commenced in this chapter. The suc-
ceeding chapters will develop more accurate concepts, and when
we can state the problem with precision we shall be not far from
its solution.



Behaviour, reflex and learned

1/2. The activities of the nervous system may be divided more
or less distinctly into two types. The dichotomy is perhaps an
over-simplification, but it will be sufficient for our purpose.

The first type is reflex behaviour. It is inborn, it is genetically
determined in detail, it is a product, in the vertebrates, chiefly
of centres in the spinal cord and in the base of the brain, and it is
not appreciably modified by individual experience. The second
type is learned behaviour. It is not inborn, it is not genetically
determined in detail (more fully discussed in S. 1/9), it is a product
chiefly of the cerebral cortex, and it is modified markedly by the
organism’s individual experiences.

1/3. With the first or reflex type of behaviour we shall not be
concerned. We assume that each reflex is produced by some
neural mechanism whose physico-chemical nature results inevit-
ably in the characteristic form of behaviour, that this mechanism
is developed under the control of the gene-pattern and is inborn,
and that the pattern of behaviour produced by the mechanism is
usually adapted to the animal’s environment because natural
selection has long since eliminated all non-adapted variations.
For example, the complex activity of ‘ coughing ‘ is assumed to
be due to a special mechanism in the nervous system, inborn and
developed by the action of the gene-pattern, and adapted and
perfected by the fact that an animal who is less able to clear its
trachea of obstruction has a smaller chance of survival.

Although the mechanisms underlying these reflex activities are
often difficult to study physiologically and although few are known
in all their details, yet it is widely held among physiologists that
no difficulty of principle is involved. Such behaviour and such
mechanisms will not therefore be considered further.

1/4. It is with the second type of behaviour that we are con-
cerned : the behaviour that is not inborn but learned. Examples
of such reactions exist in abundance, and any small selection
must seem paltry. Yet I must say what I mean, if only to give
the critic a definite target for attack. Several examples will
therefore be given.

A dog selected at random for an experiment with a conditioned



reflex can be made at will to react to the sound of a bell either
with or without salivation. Further, once trained to react in
one way it may, with little difficulty, be trained to react later in
the opposite way. The salivary response to the sound of a bell
cannot, therefore, be due to a mechanism of fixed properties.

A rat selected at random for an experiment in maze-running
can be taught to run either to right or left by the use of an appro-
priately shaped maze. Further, once trained to turn to one side
it can be trained later to turn to the other.

A kitten approaching a fire for the first time is unpredictable
in its first reactions. The kitten may walk almost into it, or
may spit at it, or may dab at it with a paw, or may try to sniff
at it, or may crouch and ‘ stalk ‘ it. The initial way of behaving
is not, therefore, determined by the animal’s species.

Perhaps the most striking evidence that animals, after training,
can produce behaviour which cannot possibly have been inborn
is provided by the circus. A seal balances a ball on its nose for
minutes at a time ; one bear rides a bicycle, and another walks
on roller skates. It would be ridiculous to suppose that these
reactions are due to mechanisms both inborn and specially per-
fected for these tricks.

Man himself provides, of course, the most abundant variety of
learned reactions : but only one example will be given here. If
one is looking down a compound microscope and finds that the
object is not central but to the right, one brings the object to
the centre by pushing the slide still farther to the right. The
relation between muscular action and consequent visual change
is the reverse of the usual. The student’s initial bewilderment
and clumsiness demonstrate that there is no neural mechanism
inborn and ready for the reversed relation. But after a few days
co-ordination develops.

These examples, and all the facts of which they are representa-
tive, show that the nervous system is able to develop ways of
behaving which are not inborn and are not specified in detail
by the gene-pattern.

1/5. Learned behaviour has many characteristics, but we shall
be concerned chiefly with one : when animals and children learn,
not only does their behaviour change, but it changes usually for
the better. The full meaning of i better ‘ will be discussed in



Chapter 5, but in the simpler cases the improvement is obvious
enough. ‘ The burned child dreads the fire ‘ : after the experi-
ence the child’s behaviour towards the fire is not only changed,
but is changed to a behaviour which gives a lessened chance of
its being burned again. We would at once recognise as abnormal
any child who used its newly acquired knowledge so as to get
to the flames more quickly.

To demonstrate that learning usually changes behaviour from a
less to a more beneficial, i.e. survival-promoting, form would
need a discussion far exceeding the space available. But in this
introduction no exhaustive survey is needed. I require only
sufficient illustration to make the meaning clear. For this pur-
pose the previous examples will be examined seriatim.

When a conditioned reflex is established by the giving of food
or acid, the amount of salivation changes from less to more. And
the change benefits the animal either by providing normal lubri-
cation for chewing or by providing water to dilute and flush away
the irritant. When a rat in a maze has changed its behaviour so
that it goes directly to the food at the other end, the new behaviour
is better than the old because it leads more quickly to the animal’s
hunger being satisfied. The kitten’s behaviour in the presence of
a fire changes from being such as may cause injury by burning to
an accurately adjusted placing of the body so that the cat’s body
is warmed by the fire neither too much nor too little. The circus
animals’ behaviour changes from some random form to one deter-
mined by the trainer, who applied punishments and rewards.
The animals’ later behaviour is such as has decreased the punish-
ments or increased the rewards. In Man, the proposition that
behaviour usually changes for the better with learning would
need extensive discussion. But in the example of the finger
movements and the compound microscope, the later movements,
which bring the desired object directly to the centre of the field,
are clearly better than the earlier movements, which were dis-
orderly and ineffective.

Our problem may now be stated in preliminary form : what
cerebral changes occur during the learning process, and why does
the behaviour usually change for the better ? What type of
mechanistic process could show the same property ?

But before the solution is attempted we must first glance at the
peculiar difficulties which will be encountered.



1/6. The nervous system is well provided with means for action.
Glucose, oxygen, and other metabolites are brought to it by the
blood so that free energy is available abundantly. The nerve
cells composing the system are not only themselves exquisitely
sensitive, but are provided, at the sense organs, with devices of
even higher sensitivity. Each nerve cell, by its ramifications,
enables a single impulse to become many impulses, each of which
is as active as the single impulse from which it originated. And
by their control of the muscles, the nerve cells can rouse to
activity engines of high mechanical power. The nervous system,
then, possesses almost unlimited potentialities for action. But
do these potentialities solve our problem ? It seems not. We
are concerned primarily with the question why, during learning,
behaviour changes for the better : and this question is not
answered by the fact that a given behaviour can change to one
of lesser or greater activity. The examples given in S. 1/5,
when examined for the energy changes before and after learning,
show that the question of the quantity of activity is usually

But the evidence against regarding mere activity as sufficient
for a solution is even stronger : often an increase in the amount of
activity is not so much irrelevant as positively harmful.

If a dynamic system is allowed to proceed to vigorous action
without special precautions, the activity will usually lead to the
destruction of the system itself. A motor car with its tank full
of petrol may be set into motion, but if it is released with no driver
its activity, far from being beneficial, will probably cause the
motor car to destroy itself more quickly than if it had remained
inactive. The theme is discussed more thoroughly in S. 20/12 ;
here it may be noted that activity, if inco-ordinated, tends merely
to the system’s destruction. How then is the brain to achieve
success if its potentialities for action are partly potentialities for
self-destruction ?

The relation of part to part

1/7. It was decided in S. 1/5 that after the learning process the
behaviour is usually better adapted than before. We ask, there-
fore, what property must be possessed by the neurons, or by the
parts of a mechanical ‘ brain ‘, so that the manifestation by

5 B


the neuron of this property shall result in the whole animal’s
behaviour being improved.

Even if we allow the neuron all the properties of a living
organism, it is still insufficiently provided. For the improvement
in the animal’s behaviour is often an improvement in relation to
entities which have no counterpart in the life of a neuron. Thus
when a dog, given food in an experiment on conditioned reflexes,
learns to salivate, the behaviour improves because the saliva
provides a lubricant for chewing. But in the neuron’s existence,
since all its food arrives in solution, neither ‘ chewing ‘ nor ‘ lubri-
cant ‘ can have any direct relevance or meaning. Again, a rat
learns to run through a maze without mistakes ; yet the learning
has involved neurons which are firmly supported in a close mesh
of glial fibres and never move in their lives.

Finally, consider an engine-driver who has just seen a signal
and whose hand is on the throttle. If the light is red, the
excitation from the retina must be transmitted through the
nervous system so that the cells in the motor cortex send impulses
down to those muscles whose activity makes the throttle close.
If the light is green, the excitation from the retina must be
transmitted through the nervous system so that the cells in the
motor cortex make the throttle open. And the transmission is
to be handled, and the safety of the train guaranteed, by neurons
which can form no conception of ‘ red ‘, ‘ green ‘, ‘ train ‘, ‘ signal ‘,
or ‘ accident ‘ ! Yet the system works.

1/8. In some cases there may be a simple mechanism which
uses the method that a red light activates a chain of nerve-cells
leading to the muscles which close the throttle while a green light
activates another chain of nerve-cells leading to the muscles which
make it open. In this way the effect of the colour of the signal
might be transmitted through the nervous system in the appro-
priate way.

The simplicity of the arrangement is due to the fact that we
are supposing that the two reactions are using two completely
separate and independent mechanisms. This separation may well
occur in the simpler reactions, but it is insufficient to explain the
events of the more complex reactions. In most cases the ‘ correct ‘
and the ‘ incorrect ‘ neural activities are alike composed of excita-
tions, of inhibitions, and of other changes that are all physiological,



so that the correctness is determined not by the process itself but
by the relations which it bears to the other processes.

This dependence of the ‘ correctness ‘ of what is happening at
one point in the nervous system on what is happening at other
points would be shown if the engine-driver were to move over to
the other side of the cab. For if previously a flexion of the elbow
had closed the throttle, the same action will now open it ; and
what was the correct pairing of red and green to push and pull
must now be reversed. So the local action in the nervous system
can no longer be regarded as 4 correct ‘ or ‘ incorrect ‘, and the
first simple solution breaks down.

Another example is given by the activity of chewing in so
far as it involves the tongue and teeth in movements which must
be related so that the teeth do not bite the tongue. No move-
ment of the tongue can by itself be regarded as wholly wrong, for
a movement which may be wrong when the teeth are just meeting
may be right when they are parting and food is to be driven on
to their line. Consequently the activities in the neurons which
control the movement of the tongue cannot be described as either
4 correct ‘ or ‘ incorrect ‘ : only when these activities are related to
those of the neurons which control the jaw movements can a cor-
rectness be determined ; and this property now belongs, not to either
separately, but only to the activity of the two in combination.

These considerations reveal the main peculiarity of the problem.
When the nervous system learns, it undergoes changes which
result in its behaviour becoming better adapted to the environ-
ment. The behaviour depends on the activities of the various
parts whose individual actions compound for better or worse into
the whole action. Why, in the living brain, do they always
compound for the better ?

If we wish to build an artificial brain the parts must be specified
in their nature and properties. But how can we specify the
4 correct ‘ properties for each part if the correctness depends not
on the behaviour of each part but on its relations to the other
parts ? Our problem is to get the parts properly co-ordinated.
The brain does this automatically. What sort of a machine can
be ^//-co-ordinating ?

This is our problem. It will be stated with more precision in
S. 1/12. But before this statement is reached, some minor topics
must be discussed.



The genetic control of cerebral function

1/9. The various species of the animal kingdom differ widely
in their powers of learning : Man’s intelligence, for instance, is
clearly a species-characteristic, for the higher apes, however well
trained, never show an intelligence equal to that of the average
human being. Clearly the power of learning is determined to
some extent by the inherited gene-pattern. In what way does
the gene-pattern exert its effect on the learning process ? In
particular, what part does it play in the adjustments of part to
part which the previous section showed to be fundamental ?
Does the gene-pattern determine these adjustments in detail ?

In Man, the genes number about 50,000 and the neurons number
about 10,000,000,000. The genes are therefore far too few to
specify every neuronic interconnection. (The possibility that a
gene may control several phenotypic features is to some extent
balanced by the fact that a single phenotypic feature may require
several genes for its determination.)

But the strongest evidence against the suggestion that the
genes exert, in the higher animals, a detailed control over the
adjustments of part to part is provided by the evidence of S. 1/4.
A dog, for instance, can be made to respond to the sound of a
bell either with or without salivation, regardless of its particular
gene-pattern. It is impossible, therefore, to relate the control of
salivation to the particular genes possessed by the dog. This
example, and all the other facts of which it is typical, show that
the effect of the gene-pattern on the details of the learning process
cannot be direct.

The effect, then, must be indirect : the genes fix permanently
certain function-rules, but do not interfere with the function-rules
in their detailed application to particular situations. Three
examples of this type of control will be given in order to illustrate
its nature.

In the game of chess, the laws (the function rules) are few and
have been fixed for a century ; but their effects are as numerous
as the number of positions to which they can be applied. The
result is that games of chess can differ from one another though
controlled by constant laws.

A second example is given by the process of evolution through
natural selection. Here again the function-rule (the principle of



the survival of the fittest) is fixed, yet its influence has an infinite
variety when applied to an infinite variety of particular organisms
in particular environments.

A final example is given in the body by the process of inflam-
mation. The function-rules which govern the process are gene-
tically determined and are constant in one species. Yet these
rules, when applied to an infinite variety of individual injuries,
provide an infinite variety in the details of the process at particular
points and times.

Our aim is now clear : we must find the function-rules. They
must be few in number, much fewer than 50,000, and we must
show that these few function-rules, when applied to an almost
infinite number of circumstances and to 10,000,000,000 neurons,
are capable of directing adequately the events in all these circum-
stances. The function-rules must be fixed, their applications

(The gene-pattern is discussed again in S. 9/9.)

Restrictions on the concepts to be used

1/10. Throughout the book I shall adhere to certain basic
assumptions and to certain principles of method.

The nervous system, and living matter in general, will be
assumed to be identical with all other matter. So no use of any
1 vital ‘ property or tendency will be made, and no deus ex machina
will be invoked. No psychological concept will be used unless
it can be shown in objective form in non-living systems ; and
when used it will be considered to refer solely to its objective
form. Related is the restriction that every concept used must
be capable of objective demonstration. In the study of man
this restriction raises formidable difficulties extending from the
practical to the metaphysical. But as most of the discussion
will be concerned with the observed behaviour of animals and
machines, the peculiar difficulties will seldom arise.

No teleological explanation for behaviour will be used. It
will be assumed throughout that a machine or an animal behaved
in a certain way at a certain moment because its physical and
chemical nature at that moment allowed it no other action. Never
will we use the explanation that the action is performed because
it will later be advantageous to the animal. Any such explanation



would, of course, involve a circular argument ; for our purpose
is to explain the origin of behaviour which appears to be teleo-
logically directed.

It will be further assumed that the nervous system, living
matter, and the matter of the environment are all strictly deter-
minate : that if on two occasions they are brought to the same
state, the same behaviour will follow. Since at the atomic level
of size the assumption is known to be false, the assumption implies
that the functional units of the nervous system must be sufficiently
large to be immune to this source of variation. For this there is
some evidence, since recordings of nervous activity, even of single
impulses, show no evidence of appreciable thermal noise. But
we need not prejudge the question. The work to be described
is an attempt to follow the assumption of determinacy wherever
it leads. When it leads to obvious error will be time to question
its validity.


1/11. The previous section has demanded that we shall make no
use of the subjective elements of experience ; and I can antici-
pate by saying that in fact the book makes no such use. At
times its rigid adherence to the objective point of view may
jar on the reader and may expose me to the accusation that I am
ignoring an essential factor. A few words in explanation may
save misunderstanding.

Throughout the book, consciousness and its related subjective
elements are not used for the simple reason that at no point have I
found their introduction necessary. This is not surprising, for the
book deals with only one of the aspects of the mind-body relation,
and with an aspect — learning — that has long been recognised to
have no necessary dependence on consciousness. Here is an
example to illustrate their independence. If a cyclist wishes to
turn to the left, his first action must be to turn the front wheel
to the right (otherwise he will fall outwards by centrifugal force).
Every practised cyclist makes this movement every time he
turns, yet many cyclists, even after they have made the move-
ment hundreds of times, are quite unconscious of making it.
The direct intervention of consciousness is evidently not necessary
for adaptive learning.



Such an observation, showing that consciousness is sometimes
not necessary, gives us no right to deduce that consciousness
does not exist. The truth is quite otherwise, for the fact of the
existence of consciousness is prior to all other facts. If I perceive
— am aware of — a chair, I may later be persuaded, by other
evidence, that the appearance was produced only by a trick of
lighting ; I may be persuaded that it occurred in a dream, or
even that it was an hallucination ; but there is no evidence in
existence that could persuade me that my awareness itself was
mistaken— that I had not really been aware at all. This know-
ledge of personal awareness, therefore, is prior to all other forms
of knowledge.

If consciousness is the most fundamental fact of all, why is it
not used in this book ? The answer, in my opinion, is that
Science deals, and can deal, only with what one man can demonstrate
to another. Vivid though consciousness may be to its possessor,
there is as yet no method known by which he can demonstrate his
experience to another. And until such a method, or its equivalent,
is found, the facts of consciousness cannot be used in scientific

The problem

1/12. It is now time to state the problem. Later, when more
exact concepts have been developed, it will be possible to state the
problem more precisely (S. 8/1).

It will be convenient, throughout the discussion, to have some
well-known, practical problem to act as type-problem, so that
general statements can always be referred to it. I select the
following. When a kitten first approaches a fire its reactions are
unpredictable and usually inappropriate. Later, however, when
adult, its reactions are different. It approaches the fire and seats
itself at that place where the heat is moderate. If the fire burns
low, it moves nearer. If a hot coal falls out, it jumps away.
I might have taken as type-problem some experiment published
by a psychological laboratory, but the present example has
several advantages. It is well known ; it is representative of a
wide class of important phenomena ; and it is not likely to be
called in question by the discovery of some small technical flaw.

With this as specific example, we may state the problem



generally. We commence with the concepts that the organism
is mechanistic in action, that it is composed of parts, and that
the behaviour of the whole is the outcome of the compounded
actions of the parts. Organisms change their behaviour by
learning, and change it so that the later behaviour is better
adapted to their environment than the earlier. Our problem is,
first, to identify the nature of the change which shows as learning,
and secondly, to find why such changes should tend to cause
better adaptation for the whole organism.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s