THE PROCESS OF LEARNING IN ANIMALS AND INFANTS
In the present chapter I wish to consider the processes by which,
and the laws according to which, an animal’s original repertoire of
reflexes is changed into a quite different set of habits as a result
of events that happen to it. A dog learns to follow his master in
preference to anyone else; a horse learns to know his own stall in
the stable; a cow learns to come to the cow-shed at milking time.
All these are acquired habits, not reflexes; they depend upon the
circumstances of the animals concerned, not merely upon the congenital
characteristics of the species. When I speak of an animal “learning”
something, I shall include all cases of acquired habits, whether or not
they are useful to the animal. I have known horses in Italy “learn” to
drink wine, which I cannot believe to have been a desirable habit. A
dog may “learn” to fly at a man who has ill-treated it, and may do so
with such regularity and ferocity as to lead to its being killed. I do
not use learning in any sense involving praise, but merely to denote
modification of behaviour as the result of experience.
The manner in which animals learn has been much studied in recent
years, with a great deal of patient observation and experiment. Certain
results have been obtained as regards the kinds of problems that have
been investigated, but on general principles there is still much
controversy. One may say broadly that all the animals that have been
carefully observed have behaved so as to confirm the philosophy in
which the observer believed before his observations began. Nay, more,
they have all displayed the national characteristics of the observer.
Animals studied by Americans rush about frantically, with an incredible
display of hustle and pep, and at last achieve the desired result by
chance. Animals observed by Germans sit still and think, and at last
evolve the solution out of their inner consciousness. To the plain
man, such as the present writer, this situation is discouraging. I
observe, however, that the type of problem which a man naturally sets
to an animal depends upon his own philosophy, and that this probably
accounts for the differences in the results. The animal responds to
one type of problem in one way and to another in another; therefore
the results obtained by different investigators, though different, are
not incompatible. But it remains necessary to remember that no one
investigator is to be trusted to give a survey of the whole field.
The matters with which we shall be concerned in this chapter belong to
behaviourist psychology, and in part to pure physiology. Nevertheless,
they seem to me vital to a proper understanding of philosophy, since
they are necessary for an objective study of knowledge and inference. I
mean by an “objective” study one in which the observer and the observed
need not be the same person; when they must be identical, I call the
study “subjective.” For the present we are concerned with what is
required for understanding “knowledge” as an objective phenomenon. We
shall take up the question of the subjective study of knowledge at a
later stage.
The scientific study of learning in animals is a very recent growth;
it may almost be regarded as beginning with Thorndike’s _Animal
Intelligence_, which was published in 1911. Thorndike invented the
method which has been adopted by practically all subsequent American
investigators. In this method an animal is separated from food, which
he can see or smell, by an obstacle which he may overcome by chance. A
cat, say, is put in a cage having a door with a handle which he may by
chance push open with his nose. At first the cat makes entirely random
movements, until he gets his result by a mere fluke. On the second
occasion, in the same cage, he still makes some random movements, but
not so many as on the first occasion. On the third occasion he does
still better, and before long he makes no useless movements. Nowadays
it has become customary to employ rats instead of cats, and to put them
in a model of the Hampton Court maze rather than in a cage. They take
all sorts of wrong turnings at first, but after a time they learn to
run straight out without making any mistake. Dr. Watson gives averages
for nineteen rats, each of which was put into the maze repeatedly,
with food outside where the rat could smell it. In all the experiments
care was taken to make sure that the animal was very hungry. Dr.
Watson says: “The first trial required on the average over seventeen
minutes. During this time the rat was running around the maze, into
blind alleys, running back to the starting point, starting for the food
again, biting at the wires around him, scratching himself, smelling
this spot and that on the floor. Finally he got to the food. He was
allowed only a bite. Again he was put back into the maze. The taste of
the food made him almost frantic in his activity. He dashed about more
rapidly. The average time for the group on the second trial is only a
little over seven minutes; on the fourth trial not quite three minutes;
from this point to the twenty-third trial the improvement is very
gradual.” On the thirtieth trial the time required, on the average,
was about thirty seconds.[3] This set of experiments may be taken as
typical of the whole group of studies to which it belongs.
[3] Watson, _Behaviorism_, pp. 169–70.
Thorndike, as a result of experiments with cages and mazes, formulated
two “provisional laws,” which are as follows:
“The Law of Effect is that: of several responses made to the same
situation, those which are accompanied or closely followed by
satisfaction to the animal will, other things being equal, be more
firmly connected with the situation, so that, when it recurs, they
will be more likely to recur; those which are accompanied or closely
followed by dissatisfaction to the animal will, other things being
equal, have their connections with that situation weakened, so that,
when it recurs, they will be less likely to recur. The greater the
satisfaction or discomfort, the greater the strengthening or weakening
of the bond.
“The Law of Exercise is that: Any response to a situation will, other
things being equal, be more strongly connected with the situation in
proportion to the number of times it has been connected with that
situation and to the average vigour and duration of the connections.”
We may sum up these two laws, roughly, in the two statements: First, an
animal tends to repeat what has brought it pleasure; second, an animal
tends to repeat what it has often done before. Neither of these laws is
at all surprising, but, as we shall see, there are difficulties in the
theory that they are adequate to account for the process of learning in
animals.
Before going further there is a theoretical point to be cleared up.
Thorndike, in his first law, speaks of satisfaction and discomfort,
which are terms belonging to subjective psychology. We cannot observe
whether an animal _feels_ satisfaction or _feels_ discomfort; we can
only observe that it behaves in ways that we have become accustomed to
interpret as signs of these feelings. Thorndike’s law, as it stands,
does not belong to objective psychology, and is not capable of being
experimentally tested. This, however, is not so serious an objection
as it looks. Instead of speaking of a result that brings satisfaction,
we can merely enumerate the results which, in fact, have the character
which Thorndike mentions, namely, that the animal tends to behave so
as to make them recur. The rat in the maze behaves so as to get the
cheese, and when an act has led him to the cheese once, he tends to
repeat it. We may say that this is what we mean when we say that the
cheese “gives satisfaction”, or that the rat “desires” the cheese.
That is to say, we may use Thorndike’s “Law of Effect” to give us an
objective definition of desire, satisfaction, and discomfort. The law
should then say: there are situations such that animals tend to repeat
acts which have led to them; these are the situations which the animal
is said to “desire” and in which it is said to “find satisfaction”.
This objection to Thorndike’s first law is, therefore, not very
serious, and need not further trouble us.
Dr. Watson considers one principle alone sufficient to account for
all animal and human learning, namely, the principle of “learned
reactions.” This principle may be stated as follows:
_When the body of an animal or human being has been exposed
sufficiently often to two roughly simultaneous stimuli, the earlier of
them alone tends to call out the response previously called out by the
other._
Although I do not agree with Dr. Watson in thinking this principle
alone sufficient, I do agree that it is a principle of very great
importance. It is the modern form of the principle of “association”.
The “association of ideas” has played a great part in philosophy,
particularly in British philosophy. But it now appears that this is
a consequence of a wider and more primitive principle, namely, the
association of bodily processes. It is this wider principle that is
asserted above. Let us see what is the nature of the evidence in its
favour.
Our principle becomes verifiable over a much larger field than the
older principle owing to the fact that it is movements, not “ideas”,
that are to be associated. Where animals are concerned, ideas are
hypothetical, but movements can be observed; even with men, many
movements are involuntary and unconscious. Yet animal movements and
unconscious involuntary human movements are just as much subject to
the law of association as the most conscious ideas. Take, _e.g._ the
following example (Watson, p. 33). The pupil of the eye expands in
darkness and contracts in bright light; this is an involuntary and
unconscious action of which we only become aware by observing others.
Now take some person and repeatedly expose him to bright light at
the same moment that you ring an electric bell. After a time the
electric bell alone will cause his pupils to contract. As far as can be
discovered, all muscles behave in this way. So do glands where they can
be tested. It is said that a brass band can be reduced to silence by
sucking a lemon in front of it, owing to the effect upon the salivary
glands of its members; I confess that I have never verified this
statement. But you will find the exact scientific analogue for dogs in
Watson, p. 26. You arrange a tube in a dog’s mouth so that saliva drops
out at a measurable rate. When you give the dog food it stimulates the
flow of saliva. At the same moment you touch his left thigh. After a
certain length of time the touch on the left thigh will produce just
as much saliva without the food as with it. The same sort of thing
applies to emotions, which depend upon the ductless glands. Children
at birth are afraid of loud noises, but not of animals. Watson took a
child eleven months old, who was fond of a certain white rat; twice at
the moment when the child touched the rat, a sudden noise was made just
behind the child’s head. This was enough to cause fear of the rat on
subsequent occasions, no doubt owing to the fact that the adrenal gland
was now stimulated by the substitute stimulus, just like the salivary
glands in the dog or the trumpet player. The above illustrations show
that “ideas” are not the essential units in association. It seems that
not merely is “mind” irrelevant, but even the brain is less important
than was formerly supposed. At any rate, what is known experimentally
is that the glands and muscles (both striped and unstriped) of the
higher animals exhibit the law of transfer of response, _i.e._ when
two stimuli have often been applied together, one will ultimately call
out the response which formerly the other called out. This law is one
of the chief bases of habit. It is also obviously essential to our
understanding of language: the sight of a dog calls up the word “dog”,
and the word “dog” calls up some of the responses appropriate to a real
dog.
There is, however, another element in learning, besides mere habit.
This is the element dealt with by Thorndike’s “Law of Effect.” Animals
tend to repeat acts which have pleasant consequences, and to avoid
such as have unpleasant consequences. But, as we saw a moment ago,
“pleasant” and “unpleasant” are words which we cannot verify by
objective observation. What we can verify by observation is that
an animal seeks situations which in fact have had certain results,
and avoids situations which in fact have had certain other results.
Moreover, broadly speaking, the animal seeks results which tend to
survival of itself or its offspring, and avoids results which tend in
the opposite direction. This, however, is not invariable. Moths seek
flames and men seek drink, though neither is biologically useful. It
is only approximately, in situations long common, that animals are so
adjusted to their environment as to act in a way which is advantageous
from a biological standpoint. In fact, biological utility must
never be employed as an explanation, but only noticed as a frequent
characteristic, of the ways in which animals behave.
Dr. Watson is of the opinion that Thorndike’s “Law of Effect” is
unnecessary. He first suggests that only two factors are called for in
the explanation of habit, namely, _frequency_ and _recency_. Frequency
is covered by Thorndike’s “Law of Exercise”, but recency, which is
almost certainly a genuine factor, is not covered by Thorndike’s two
laws. That is to say, when a number of random movements have finally
resulted in success, the more recent of these movements are likely to
be repeated earlier, on a second trial, than the earlier ones. But Dr.
Watson finally abandons this method of dealing with habit-formation in
favour of the one law of “conditioned reflexes” or “learned reactions”.
He says (_Behaviorism_, p. 166):
“Only a few psychologists have been interested in the problem. Most
of the psychologists, it is to be regretted, have even failed to see
that there is a problem. They believe habit formation is implanted by
kind fairies. For example, Thorndike speaks of pleasure stamping in
the successful movement and displeasure stamping out the unsuccessful
movements. Most of the psychologists talk, too, quite volubly about the
formation of new pathways in the brain, as though there were a group
of tiny servants of Vulcan there who run through the nervous system
with hammer and chisel digging new trenches and deepening old ones.
I am not sure that the problem when phrased in this way is a soluble
one. I feel that there must come some simpler way of envisaging the
whole process of habit formation or else it may remain insoluble.
Since the advent of the conditioned reflex hypothesis in psychology
with all of the simplifications (and I am often fearful that it may be
an oversimplification!) I have had my own laryngeal processes [_i.e._
what others call “thoughts”] stimulated to work upon this problem from
another angle.”
I agree with Dr. Watson that the explanations of habit-formation which
are usually given are very inadequate, and that few psychologists have
realised either the importance or the difficulty of the problem. I
agree also that a great many cases are covered by his formula of the
conditioned reflex. He relates a case of a child who once touched a
hot radiator, and afterward avoided it for two years. He adds: “If we
should keep our old habit terminology, we should have in this example
a habit formed by a single trial. There can be then in this case no
‘stamping in of the successful movement’ and ‘no stamping out of the
unsuccessful movement.’” On the basis of such examples, he believes
that the whole of habit-formation can be derived from the principle of
the conditioned reflex, which he formulates as follows (p. 168):
_Stimulus X will not now call out reaction R; stimulus Y will call out
reaction R (unconditioned reflex); but when stimulus X is presented
first and then Y (which does call out R) shortly thereafter, X will
thereafter call out R. In other words, stimulus X becomes ever
thereafter substituted for Y._
This law is so simple, so important, and so widely true that there
is a danger lest its scope should be exaggerated, just as, in the
eighteenth century, physicists tried to explain everything by means of
gravitation. But when considered as covering all the ground, it seems
to me to suffer from two opposite defects. In the first place, there
are cases where no habit is set up, although by the law it should be.
In the second place, there are habits which, so far as we can see at
present, have a different genesis.
To take the first point first: the word “pepper” does not make people
sneeze, though according to the law it should.[4] Words which describe
succulent foods will make the mouth water; voluptuous words will have
some of the effect that would be produced by the situations they
suggest; but no words will produce sneezes or the reactions appropriate
to tickling. In the diagram given by Dr. Watson (p. 106), there are
four reflexes which appear to be not sources of conditioned reflexes,
namely sneezing, hiccoughing, blinking, and the Babinski reflex;
of these, however, blinking, it is suggested (p. 99) may be really
itself a conditioned reflex. There may be some quite straightforward
explanation of the fact that some reactions can be produced by
substitute stimuli while others cannot, but none is offered. Therefore
the law of the conditioned reflex, as formulated, is too wide, and it
is not clear what is the principle according to which its scope should
be restricted.
[4] Dr. Watson apparently entertains hopes of teaching babies
to sneeze when they see the pepper box, but he has not
yet done so. See _Behaviorism_, p. 90.
The second objection to Dr. Watson’s law of habit, if valid, is more
important than the first; but its validity is more open to question. It
is contended that the acts by which solutions of problems are obtained
are, in cases of a certain kind, not random acts leading to success by
mere chance, but acts proceeding from “insight”, involving a “mental”
solution of the problem as a preliminary to the physical solution.
This is especially the view of those who advocate _Gestaltpsychologie_
or the psychology of configuration. We may take, as typical of their
attitude on the subject of learning, Köhler’s _Mentality of Apes_.
Köhler went to Tenerife with certain chimpanzees in the year 1913;
owing to the war he was compelled to remain with them until 1917, so
that his opportunities for study were extensive. He complains of the
maze and cage problems set by American investigators that they are such
as _cannot_ be solved by intelligence. Sir Isaac Newton himself could
not have got out of the Hampton Court maze by any method except trial
and error. Köhler, on the other hand, set his apes problems which could
be solved by what he calls “insight”. He would hang up a banana[5] out
of reach, and leave boxes in the neighbourhood so that by standing
on the boxes the chimpanzees could reach the fruit. Sometimes they
had to pile three or even four boxes on top of each other before they
could achieve success. Then he would put the banana outside the bars
of the cage, leaving a stick inside, and the ape would get the banana
by reaching for it with the stick. On one occasion, one of them, named
Sultan, had two bamboo sticks, each too short to reach the banana;
after vain efforts followed by a period of silent thought, he fitted
the smaller into the hollow of the other, and so manufactured one stick
which was long enough. It seems, however, from the account, that he
first fitted the two together more or less accidentally, and only then
realised that he had found a solution. Nevertheless, his behaviour when
he had once realised that one stick could be made by joining the two
was scarcely Watsonian: there was no longer anything tentative, but a
definite triumph, first in anticipation and then in action. He was so
pleased with his new trick that he drew a number of bananas into his
cage before eating any of them. He behaved, in fact, as capitalists
have behaved with machinery.
[5] Called by Köhler “the objective,” because the word
“banana” is too humble for a learned work. The pictures
disclose the fact that “the objective” was a mere banana.
Köhler says: “We can, from our own experience, distinguish sharply
between the kind of conduct which, from the very beginning, arises out
of a consideration of the characteristics of a situation, and one that
does not. Only in the former case do we speak of insight, and only
that behaviour of animals definitely appears to us intelligent which
takes account from the beginning of the lie of the land, and proceeds
to deal with it in a smooth continuous course. Hence follows this
characteristic: _to set up as the criterion of insight, the appearance
of a complete solution with reference to the whole lay-out of the
field_.”
Genuine solutions of problems, Köhler says, do not improve by
repetition; they are perfect on the first occasion, and, if anything,
grow worse by repetition, when the excitement of discovery has worn
off. The whole account that Köhler gives of the efforts of his
chimpanzees makes a totally different impression from that of the
rats in mazes, and one is forced to conclude that the American work
is somewhat vitiated by confining itself to one type of problem,
and drawing from that one type conclusions which it believes to be
applicable to all problems of animal learning. It seems that there
are two ways of learning, one by experience, and the other by what
Köhler calls “insight”. Learning by experience is possible to most
vertebrates, though rarely, so far as is known, to invertebrates.
Learning by “insight”, on the contrary, is not known to exist in any
animals lower than the anthropoid apes, though it would be extremely
rash to assert that it will not be revealed by further observations on
dogs or rats. Unfortunately, some animals--for instance, elephants--may
be extremely intelligent, but the practical difficulty and expense of
experimentation with them is so great that we are not likely to know
much about them for some time to come. However, the real problem is
already sufficiently definite in Köhler’s book: it is the analysis of
“insight” as opposed to the method of the conditioned reflex.
Let us first be clear as to the nature of the problem, when described
solely in terms of behaviour. A hungry monkey, if sufficiently near
to a banana, will perform acts such as, in circumstances to which it
has been accustomed, have previously enabled it to obtain bananas.
This fits well with either Watson or Thorndike, so far. But if these
familiar acts fail, the animal will, if it has been long without food,
is in good health, and is not too tired, proceed to other acts which
have never hitherto produced bananas. One may suppose, if one wishes to
follow Watson, that these new acts are composed of a number of parts,
each of which, on some former occasion, has occurred in a series which
ended with the obtaining of the banana. Or one may suppose--as I think
Thorndike does--that the acts of the baffled animal are random acts,
so that the solution emerges by pure chance. But even in the first
hypothesis, the element of chance is considerable. Let us suppose that
the acts A, B, C, D, E, have each, on a former occasion, been part of
a series ending with success, but that now for the first time it is
necessary to perform them all, and in the right order. It is obvious
that, if they are only combined by chance, the animal will be lucky if
it performs them all in the right order before dying of hunger.
But Köhler maintains that to anyone watching his chimpanzees it was
obvious they did not obtain “a composition of the solution out of
chance parts”. He says (pp. 199–200):
“It is certainly not a characteristic of the chimpanzee, when he is
brought into an experimental situation, to make any chance movements,
out of which, among other things, a non-genuine solution could arise.
He is very seldom seen to attempt anything which would have to be
considered accidental in relation to the situation (excepting, of
course, if his interest is turned away from the objective to other
things). As long as his efforts are directed to the objective, all
distinguishable stages of his behaviour (as with human beings in
similar situations) tend to appear as complete attempts at solutions,
_none_ of which appears as the product of accidentally arrayed
parts. This is true, most of all, of the solution which is finally
successful. Certainly it often follows upon a period of perplexity or
quiet (often a period of survey), but in real and convincing cases,
the solution never appears in a disorder of blind impulses. It is
one continuous smooth action, which can be resolved into its parts
_only by the imagination_ of the onlooker; in reality they do _not_
appear independently. But that in so many ‘genuine’ cases as have been
described, these solutions as wholes should have arisen from mere
chance, is an entirely inadmissible supposition.”
Thus we may take it as an observed fact that, so far as overt behaviour
is concerned, there are two objections to the type of theory with
which we began, when considered as covering the whole field. The first
objection is that in cases of a certain kind, the solution appears
sooner than it should according to the doctrine of chances; the second
is that it appears as a whole, _i.e._ that the animal, after a period
of quiescence, suddenly goes through the right series of actions
smoothly, and without hesitation.
Where human beings are concerned, it is difficult to obtain such
good data as in the case of animals. Human mothers will not allow
their children to be starved, and then shut up in a room containing
a banana which can only be reached by putting a chair on the table
and a footstool on the chair, and then climbing up without breaking
any bones. Nor will they permit them to be put into the middle of a
Hampton Court maze, with their dinner getting cold outside. Perhaps
in time the State will perform these experiments with the children of
political prisoners, but as yet, perhaps fortunately, the authorities
are not sufficiently interested in science. One can observe, however,
that human learning seems to be of both sorts, namely that described
by Watson and that described by Köhler. I am persuaded that speech is
learnt by the Watsonian method, so long as it is confined to single
words: often the trial and error, in later stages, proceeds _sotto
voce_, but it takes place overtly at first, and in some children until
their speech is quite correct. The speaking of sentences, however, is
already more difficult to explain without bringing in the apprehension
of wholes which is the thing upon which _Gestaltpsychologie_
lays stress. In the later stages of learning, the sort of sudden
illumination which came to Köhler’s chimpanzees is a phenomenon with
which every serious student must be familiar. One day, after a period
of groping bewilderment, the schoolboy knows what algebra is all
about. In writing a book, my own experience--which I know is fairly
common, though by no means universal--is that for a time I fumble and
hesitate, and then suddenly I see the book as a whole, and have only
to write it down as if I were copying a completed manuscript.
If these phenomena are to be brought within the scope of behaviourist
psychology, it must be by means of “implicit” behaviour. Watson makes
much use of this in the form of talking to oneself, but in apes it
cannot take quite this form. And it is necessary to have some theory
to explain the success of “implicit” behaviour, whether we call it
“thought” or not. Perhaps such a theory can be constructed on Watson’s
lines, but it has certainly not yet been constructed. Until the
behaviourists have satisfactorily explained the kind of discovery which
appears in Köhler’s observations, we cannot say that their thesis is
proved. This is a matter which will occupy us again at a later stage;
for the present let us preserve an open mind.