Chapter 3 Two Early Connectionist Theories

1
Chapter 3Two Early Connectionist Theories

• Presented By
• Shayna N. Lamkin

2
Chapter Highlights
Introduction Pavlovs Conditioning Basic
Principles of Pavlovian Conditioning
Excitation and Inhibition Applications and
Implications Thorndikes Early Connectionism
Thorndikes Basic Ideas Elaborations
Comparison of Thorndike with Pavlov
Thorndikes Place in Psychology
3
Introduction
First Psychological laboratory was founded by
Wilhelm Wundt in Germany, 1897. A marking
point at which modern scientific psychology was
placed on a definite institutional footing. Wundt
and his colleagues wanted to understand human
sensations and thoughts and feelings. They
wanted to take the continuous flux of conscious
awareness and analyze it into its basic
components.
4
Are memory images the same as sensations?
Are feelings a special kind of sensation or are
they something really different? How is the
intensity of a sensation related to the intensity
of the physical stimulus that produces it?This
kind of psychology, developed in Germany, became
to a great extent the standard for the rest of
Europe and for America.
5
Pavlovs Conditioning
1904-Ivan P. Pavlov (1849-1936) won the Nobel
Prize in physiology and medicine for his work on
digestion. His research consisted of opening a
fistula in the wall of a dogs stomach. He
noticed that the stomach secretions he was
studying were first triggered not by food
reaching the stomach but by chewing or even just
the sight of food, and he began to find this
anticipatory secretion the most interesting
aspect of the digestive process. Through this new
line of research (Pavlov 1960, original date
1927), he became even more famous as the father
of conditioning.
6
The experimenter starts with a stimulus (the
unconditioned stimulus) that will reliably elicit
a specific response (the unconditioned
response). Pavlovs research-the unconditioned
stimulus was meat powder and the unconditioned
response was salivation. What was to become the
conditioned stimulus could be any of great
variety of stimuli a bell, a ticking metronome,
a triangle drawn on a large care, and so on If
this stimulus was presented repeatedly just
before the meat powder, it too came to elicit
salivation, the conditioned response, and it thus
became a conditioned stimulus.
7
Since the term conditioning came to be applied
quite broadly, this particular kind of
conditioning, being first studied, came to be
called classical conditioning. The terms
unconditioned stimulus and conditioned stimulus
may seem a bit odd, considering what they
mean. What is meant is that they mean is
unconditional stimulus (not conditional on any
previous training), while the bell (or whatever
stimulus precedes the meat) is a conditional on
having been paired with the meat. Emphasizing
that although Pavlovs own studies mostly used
food as the unconditioned stimulus, there was no
such limitation either in his theory or in other
research on Pavlovian conditioning.
8
Unpleasant (or noxious) stimuli have been used as
unconditioned stimuli as much as have pleasant
ones. Examples range from disliking a food
because eating it followed by illness, to hating
a person who mistreated you, to being afraid of
horses after being thrown by one.
9
Basic Principles of Pavlovian Conditioning
The most basic principle of Pavlovian
conditioning is that the more often the
conditioned stimulus has been presented just
before the unconditioned stimulus, the greater is
the tendency for the conditioned stimulus to
produce the conditioned response. The resulting
change indicates the acquisition of
conditioning. This change represents two laws of
acquisition, one stating that with more pairings
of the conditioned and the unconditioned
stimulus, the probability that a conditioned
response will occur increases, and the other
stating that the size of the conditioned
responses increases (in Pavlovs research, the
number of drops of saliva).
10
A second basic principle of conditioning is
extinction. Extinction means that after
acquisition, the more successive times the
conditioned stimulus is then presented without
the unconditioned stimulus, the weaker the
tendency to make the conditioned response will
become. As with acquisition, this law is really
two laws, one stating that the more times in
succession the conditioned stimulus has been
presented alone, the smaller the conditioned
response will be, and the other stating that the
more such presentations there have been, the less
often any conditioned response will occur when
the conditioned stimulus is presented.
Extinction might thus be considered simply a
process of unlearning the conditioning.
11
Two other principles of conditioning are very
basic. One is generalization. This means that
after conditioning, the dog will make the
conditioned response not only to the exact
conditioned stimulus with which it was trained,
but also to the other stimuli that resemble
its. Another principle is discrimination, which
is related to generalization in much the same way
that extinction is to acquisition.
Discrimination requires training with two
conditioned stimuli, one of which is always
followed by the unconditioned stimulus (as in
acquisition) an the other never followed by it
(as in extinction). As training continues, there
will be less and less conditioned responding to
the one not followed by the unconditioned
stimulus, until eventually there is none.
12
Excitation and Inhibition
Pavlov suggested that the laws of conditioning
could be explained by the joint actions of two
main processes in the brain excitation and
inhibition. Excitation is a process of arousal,
one that tends to make responses happen, whereas
inhibition is a process of suppression, one that
tends to prevent responses from occurring. Both
therefore operate in opposition to each other.
Of the two, excitation plays a much greater part
in producing conditioning, but inhibition is
needed to explain many of the specific ways in
which conditioning works.
13
An example of conditioning, in which a bell as
the conditioned stimulus is presented to the dog
just before meat is presented as the
unconditioned stimulus. Both of these stimuli
produce excitation in the cerebral cortex of the
brain, each at a particular spot in the cortex
appropriate to that stimulus. Since the food is
something important to the dogs survival,
whereas the bell is a biologically less important
stimulus, the excitation produced by the food is
the stronger of the two.
14
According to Pavlov, excitation is then drawn
from the location of the cortex where the bell is
represented to the location where the food is
represented. This effect occurs, he said,
because of two general tendencies. One is the
tendency for weaker excitation to be drawn toward
the location of stronger excitation. The other
is the tendency for excitation that occurs first
to be drawn toward the location of excitation
that occurs slightly later. Each time the bell
is presented just before the food, the excitation
from the conditioned stimulus is drawn to the
location of the excitation from the unconditioned
stimulus, and as a result the connection between
these two regions in the cortex gets stronger.
15
Excitation, Pavlov said, will occur not only at
the place of the cortex appropriate for the
conditioned stimulus, but also at the place where
the unconditioned stimulus is represented. This
excitation, in turn, will produce a response
similar to the unconditioned response.
Excitation, he said also has an automatic
tendency to irradiate-that is, to spread out from
its original focus in all directions over the
surface of the cortex. The ideas of Pavlovs was
based on direct physiological information.
16
Applications and Implications
Pavlov believed that the principles of
conditioning could be used to explain a variety
of phenomena. He related these principles to
personality, considering that one of the most
fundamental differences among dogs-and among
humans-is the balance between excitation and
inhibition. Excitatory personalities tend toward
too much unrestrained activity (When in doubt,
do something, do anything!), whereas inhibitory
personalities tend toward unresponsiveness (When
in doubt, the safest thing to do is
nothing.). Pavlov considered conflict between
excitation and inhibition to be the basis of
neurosis.
17
Consider the case of the dogs that had learned a
discrimination between a circle followed by food
and an ellipse followed by no food. After they
had learned, on further trials the ellipse was
gradually changed in shape to resemble more and
more closely a circle, so that it was harder and
harder to make the discrimination, until finally
it became impossible to see any difference
between the two stimuli and the discrimination
broke down completely. As a result, the dogs had
both excitatory and inhibitory tendencies toward
the seemingly identical figures, and no way to
tell which tendency was appropriate. Pavlov
noted that excitatory dogs would respond to both
stimuli, inhibitory dogs to neither.
18
This failure to discriminate was not just a calm
adoption of the same response to both stimuli.
Rather, the dogs barked, tried to leave the
experimental room, and generally appeared
anxious, frustrated, and upset. These symptoms
seemed so similar to those of humans in difficult
conflict situations that Pavlov labeled the
syndrome experimental neurosis. Interceptive
conditioning, in which either the conditioned or
the unconditioned stimulus, or both, is presented
directly to one of the internal organs. The
responses that get conditioned by such
interceptive stimulation are also responses of
the internal organs or their blood supplies.
19
Example-Cold water as the unconditioned stimulus
can make the blood vessels in the wall of the
stomach constrict, and this unconditioned
response can then be trained as a conditioned
response to some conditioned stimulus. Such
conditioning brings unconscious bodily
processes. Semantic conditioning, the
conditioning of meaning or conditioning in which
the conditioned stimulus is verbal and the effect
of the conditioning is to modify its meaning for
the learner. Whereas bells and lights and other
such conditioned stimuli make up a first
signaling system, language, Pavlov said, becomes
a second signaling system.
20
Thorndikes Early Connectionism
The idea that pleasure and pain are important
determiners of behavior has a distinguished
history in psychology. It forms the basis of the
theory of psychological hedonism that was
developed by the English philosopher Jeremy
Bentham in the eighteenth century and adopted by
a number of other British philosophers.
According to this view, we all do those things
that give us pleasure and avoid those that give
us pain. However, it remained for Edward L.
Thorndike (1874-1949) to make a similar, more
behavioristic view central to the psychology of
learning.
21
Thorndike was a pioneer in experimental animal
psychology. Instead of relying on stories about
the intelligent feats of this or that animal, he
took animals into the laboratory, presented them
with standardized problems, and made careful
observations of how they solved their problems.
Animal Intelligence (1898), his monograph, is
still one of the classics in the field. His
most widely quoted studies involved cats in a
problem box. A hungry cat was confined in a cage
with a tempting morsel of fish outside. The cat
could open the door by pulling a loop of string
hanging inside the cage. Usually a cat went
through a long process of walking around, clawing
the sides of the cage, and other responses
before it pulled the loop of string and was able
to leave the cage.
22
On successive tests in the cage, the cats took
shorter and shorter times to pull the string.
Even after several experiences of opening the
door by pulling the string, on a given trial a
cat would still spend considerable time in other
behavior before pulling the string. This led
Thorndike to conclude that the cats learning to
pull the string involved not an intelligent
understanding of a relation between string and
pulling and door opening but a gradual stamping
in of the stimulus-response connection between
seeing the string and pulling it. At the time
Thorndike published these studies, they were
radical in two respects their careful
observation of animal behavior under controlled
conditions and their concern with the gradual
strengthening of responses.
23
They were Thorndikes answer to the argument
about whether animals solve problems by reasoning
or by instinct. By neither, said Thorndike, but
rather by the gradual learning of the correct
response.
24
Thorndikes Basic Ideas
The question on whether what animals and people
learn is connections or cognitions, Thorndike was
firmly on the connectionist side. Some of
Thorndikes views on learning had changed during
the course of his career, one that did not change
was his conviction that what we learn are
stimulus-response bonds. These bonds, in his
view, were connections in the nervous system
between incoming stimuli and the responses that
those stimuli produce. He did not use the term
intervening variables, but he treated these
bonds as intervening variables in his
interpretation of learning. Why does a given
stimulus produce a given response rather than
some other response, or none? Because there is a
bond between that stimulus and that particular
response.
25
The question is What does learning consist of?
The forming and strengthening and weakening of
these stimulus-response bonds. His theory is
thus a clearly connectionist theory. The other
question-What part does reinforcement play in
the formation of these bonds? Thorndike was
again clear it plays a crucial role. His
primary law of learning was the law of effect.
This law states that in order for a
stimulus-response bond to be formed or s
strengthened, the response must occur in the
presence of the stimulus and then be followed
quickly by a satisfier (reinforcer). In other
words, making that response in the presence of
that stimulus must have a satisfying effect
hence the name, law of effect.
26
Likewise, if the response occurs in the presence
of the stimulus but is followed by an annoyer,
the stimulus-response bond will be weakened. The
fact that satisfiers and annoyers are essential
elements in the learning process, according to
Thorndike, make his, in addition to being a
connectionist theory, a reinforcement theory as
well. Thorndikes emphasis here was heavily on
what the cats did, rather than on what they
thought or felt, in other words, on their
behavior. His use of the terms satisfier and
annoyer may seem too subjective for such an
objective approach. However he defined those two
words in a quite objective way.
27
By satisfying state of affairs is meant one
which the animal does nothing to avoid, often
doing things which maintain or renew it. By an
annoying state of affairs is meant one which the
animal does nothing to preserve, often doing
things which put an end to it-Thorndike Thorndike
says nothing here about the animals feelings,
only about what the animal does. If satisfiers
give the animal pleasure or annoyers give it
displeasure, that is not part of Thorndikes
theory, which defines satisfiers and annoyers by
their effects on the animals behavior, not on
its feelings.
28
Later on, Thorndike modified the law of effect to
make satisfiers much more important than
annoyers. Reward, he decided, strengthens
connections, but punishment does not directly
weaken them. If punishment is effective in
weakening the tendency to do something, it is
primarily because it makes behavior more variable
and thus gives some new response a chance to be
rewarded. He based this change in interpretation
on a variety of animal and human research,
including analysis of biographical information.
Thorndikes view that reward strengthens behavior
directly whereas punishment weakens it indirectly
has been accepted by a number of subsequent
psychologists.
29
Elaborations
Thorndike believed firmly throughout his career
that the strengthening of stimulus-response bonds
by satisfiers is central in all learning. When
he began his research with animals, he regarded
the sort of trial-and-error learning described
above as only one of several possibilities. From
consulting with various experts on animal
training, he expected that his cats would learn
faster to pull the loop of string if he held the
cats paw in his hand, moved it so that it pulled
the loop of string down, and thus showed the cat
that the door of the cage would then open and let
the cat out to where the food was.
30
However, his research eventually convinced him
that neither of these methods would work, that
animals would learn only by trying various
responses, being reinforced for only one of them,
and thereby gradually learning the successful
response. Though subsequent researchers have had
some success with the other two methods-passive
movement and imitation-the question of how widely
they occur when the experiment is done
appropriately is still being argued. In
Thorndikes career, he stated a number of laws of
learning, of which the law of effect is by far
his best known. One of the others was the law of
exercise, which stated that a stimulus-response
bond was strengthening from the addition of
reinforcement.
31
To the extent that this law of exercise was
valid, then the law of effect would refer only to
the additional strengthening from the addition of
reinforcement. However, his research suggested
that any strengthening of a bond by practice
alone, without reinforcement, was quite small.
The law of exercise thus became less important in
his theory, overshadowed by the law of effect.
Nevertheless, it retained some importance as a
factor in memory, where repeatedly saying two
words together would strengthen the tendency to
remember one when hearing the other, even without
any obvious satisfier following it. Thus
Thorndike, though he stated and enthusiastically
advocated the law of effect, did not completely
rule out learning in the absence of reinforcement.
32
Thorndikes concern with education led him to
take considerable interest in the topic of
transfer, the effect of learning one thing on
subsequently learning something else. Usually the
transfer is positive, meaning that the second
learning is more rapid than it would be without
the first learning, though sometimes the first
learning interferes with and slow down the
second, which constitutes negative transfer. In
Thorndikes time it was widely believed that
certain difficult subjects learned in
school-Latin and math were favorite
examples-produced positive transfer to a wide
variety of other learning, much as though they
were strengthening the brain by exercising it, or
at least teaching the student to think logically
in some very general way.
33
The truth or falsity of this belief was obviously
important to educators and Thorndike set out to
test it. He concluded that learning any material
transfers to learning any other material only the
degree that the two overlap. Thorndikes
considerable influence in educational circles
thus came out in favor of teaching students the
specific knowledge and skills they would need,
rather than other knowledge and skills designed
to produce general mental improvement.
34
Comparison of Thorndike with Pavlov

• Both set out to study learning in an objective,
  scientific way.
• Both analyzed not only learning but all behavior
  into responses to stimuli.
• Both put a good deal of emphasis on studies of
  animals, but were nonetheless strongly concerned
  with applying what they learned to the betterment
  of humankind.
• They were also marked similarities in the laws of
  learning that they proposed.
• Both saw gradual strengthening of
  stimulus-response connections.
• Both noted that when this reinforcement was
  removed, the learned response was extinguished.

35
Their systems differed in at least two ways. One
is the type of learning on which they focused.
In Pavlovian conditioning, a new stimulus (the
conditioned stimulus) comes to elicit a response
that is very similar to the one already elicited
by the unconditioned stimulusa new stimulus
comes to elicit an old response. For
Thorndikes cats, the new behavior of pulling the
loop of string was learned as a response to the
same stimuli of the puzzle box that had been
there all along-an old stimulus became attached
to a new response.
36
Pavlov considered it of major importance to
figure out what was happening in the cerebral
cortex during conditioning. Thorndike was little
concerned with the underlying physiology. He
assumed that stimulus-response bonds represented
some reality in the nervous system, but it did
not matter to him what in particular was going on
in the brain as these bonds were strengthened and
weakened. This distinction is not particularly
important. Both Pavlov and Thorndike were using
intervening variables inferred from behavior, and
their value lies in how adequately they explained
the behavior.
37
Thorndikes Place in Psychology
Thorndike was a man of great energy and of wide
interests within psychology. He was a pioneer
not only in both animal and human learning but
also in the psychology of education and of
individual differences. In all of these areas he
was interested both in pure science and in
applications, and his contributions included
methods, theory, and data. There are thus a
number of aspects of American psychology for
which a look at Thorndikes work makes a
reasonable starting point. Instead what he did
most effectively for the development of learning
theory was to provide a few strong, simple ideas
for subsequent psychologists and educators to
use, build on, or in some cases reject.
38
For well over half of the century since he
published his 1898 monograph on animal learning,
the majority of American learning theorists have
followed his lead, making Thorndikes pioneering
ideas more precise.