Tolman'sSign Learning
The
theory of Edward C. Tolman
(1886-1959) was called purposive behaviorism in his major systematic work, Purposive behavior
in
animals and men (1932b). Later he (and others) called it a
sign-gestalt theory, a sign-significate theory, or an expectancy theory.
These later terms all emphasize the cognitive nature of the theory,
which distin-guishes it in certain respects from the stimulus-response theories
of Thorndike, Guthrie, Skinner, and Hull. The designation sign learning provides a
satisfactory short name, abbreviating sign-gestalt and sign-significate, while
calling attention to the cognitive reference within the theory.
Tolman acknowledged the complex
affiliations of his system-with Watson's behaviorism, McDougall's hormic
psychology, Woodworth's dynamic psychology, and gestalt psychology (in both
classical and Lewin's forms). He later recognized a number of parallels between
his system and the probabilistic functionalism developed independently by
Brunswik (Tolman and Brunswik, 1935). Still later, the influence of
psychoanalysis began to be noticeable (1942, 1943). Because of these complex
affiliations it is appropriate to consider his system following the treatment of other behaviorisms, and before turning
to the nonbehavioristic theories.
Despite
some shifts in vocabulary, Tolman held firm to his main tenets during the
twenty-seven years between the time his book appeared and his death.
1. His
system was a genuine behaviorism, and as such rigidly
rejected introspection as a method and "raw feels" as data for
psychological sci-
ence. When he made reference to
consciousness, to inventive ideation, and the like, he was talking about
interpretations of observed behavior. He did not accept "verbal
report" as a dodge by which to smuggle con-sciousness in through the back
door.
2. The
system was a molar, rather than a molecular, behaviorism. An act of behavior has distinctive properties all its own, to be identified
and described irrespective of whatever muscular, glandular, or neural
proc-esses underlie it. The molecular facts of physics and physiology upon which behavior rests
have identifying properties of their own, which are not the properties of
behavior as molar. This meant for Tolman an inde-pendence from physiology, a characteristic which he shared
with several of the writers whom we have considered.
3. The
system was a purposivism, but of a
sort to avoid the implica-tions of a teleological metaphysics. It was a
purposivism because it recognized that behavior is regulated in accordance with
objectively deter-minable ends. It was not mentalistic; purposes are not those
of a self-conscious mind. I t is not in agreement with teleological points of
view which make effects take precedence over and determine their causes.
The strongest rejection was of
American structuralism, because struc-turalism was dependent upon introspection
of the most offensive sort, that known as Beschreibung. Watsonian behaviorism
was almost as vigor-ously rejected, because it was not only molecular but
tended to neglect the problems of goal-seekingbehavior.
THE SYSTEMATIC POSITION
Behavior
as Molar
The descriptive properties of
molar behavior are the most general characteristics of behavior which would impress
themselves upon an in-telligent onlooker without presuppositions and before any
attempt to explain how the behavior comes about.
First, behavior is goal-directed.
It is always a getting-toward something, or a getting-away from something. The
most significant description of any behavior is what the organism is doing,
what it is up to, where it is going. The
cat is trying to get out of the box, the carpenter is building a house (or
earning a living), the musician is seeking acclaim for his vir-tuosity. The particular movements involved
are less descriptive of the molar behavior than is the goal toward which or
away from which the movements lead. This feature characterizes molar behavior
as purposive.
Second, the behavior makes use of
environmental supports as means-objects toward the goal. The world in which
behavior goes on is a world
1 In this
Tolman follows Perry (1918).
S
(OBO) Cumulative nature and number of trials2
P
Pattern of preceding and
succeeding maze units
11. Individual
DifferenceVariables
H
Heredity
A Age
T Previous training
E Special endocrine, drug or vitamin conditions
It is possible to study the
effect of such variables on resulting behavior. The usual learning curve is a
plot of the functional relationship under stated conditions. These are the
behavioral "facts" about learning. It is the effort to explain the facts
which leads to theories.
Tolman's explanation rests on intervening variables.
These are in-ferred processes between the independent variables (stimuli, etc.)
and the dependent variables (responses, etc.). The preliminary list as
pre-sented coordinates one intervening variable with each of the environmen-tal
variables.
|
Intervening Variable
|
Environmental
Variable
|
|
|
Demand
|
correlated with
|
Maintenance schedule
|
|
Appetite
|
correlated with
|
Appropriateness of goal object
|
|
Differentiation
|
correlated with
|
Types and modes of stimuli provided
|
|
Motor
skill
|
correlated with
|
Types of motor response required
|
|
Hypotheses
|
correlated with
|
Cumulative nature and number of
|
|
|
|
trials
|
|
Biases
|
correlated with
|
Pattern of
preceding and succeeding
|
|
|
|
maze units
|
Although the intervening variables
may sound subjective, in principle each can be given objective definition and
measurement through a defin-ing experiment in which everything else is held
constant except the cor-relative environmental variable while that one is
systematically varied. Demand, for example, may be expected to increase with
the number of hours since feeding, but the relationship between food
deprivation and demand is not a simple one. It must be studied empirically. The
same holds for each of the intervening variables.
Having thus established a basis
for inferring the value of the interven-ing variable from the antecedent
conditions, the next stage in theory con-struction is to find the equations
relating intervening variables to behav-
2 2 (OBO) i s
a shoithand formula which means
some consequence or summation of previous expeiiences in which
one occasion (0) has led through behavior (B) to another
occasion (0).The occasions are such features as a choice-point, a
goal at the left, and so on.
|
195
|
ioral outcomes, as these intervening variables
simultaneously take on different values.
In spite
of the clear outline of what a systematic theory ought to be, Tolman nowhere
attempted quantitative predictions paralleling those of Hull, so that his conjectures
have not in that sense been put to the testes This does not mean that his experiments were
unrelated to his theory. There were, in fact, many predictions, but they assert
that one path will be preferred to another, that under one set of circumstances
the problem will be easier than under another set, and so on. The dimensional
analy-sis which completes the function was not provided, and Hull's
conscien-tious efforts are instructive in showing how difficult that task
proves to be.
SIGN LEARNING
Sign Learning as an Alternative to Response
Learning
Stimulus-response theories, while stated with
different degrees of sophistication, imply that the organism is goaded along a path by internal and external
stimuli, learning the correct movement sequences so that they are released
under appropriate conditions of drive and environmen-tal stimulation. The
alternative possibility is that the learner is following signs to a goal, is
learning his way about, is following a sort of map-in other words, is learning
not movements but meanings. This is the con-tention of Tolman's theory of sign
learning. The organism learns sign-significate relations; it learns a behavior
route, not a movement pattern. Many learning situations do not permit a dear
distinction between these two possibilities. If there is a single path with food at the end and
the organism runs faster at each opportunity, there is no way of telling
whether his responses are being stamped in by reinforcement or whether he is
guided by his immanent purposes and cognitions.
Because both stimulus-response
and sign learning so often predict the same behavioral outcome, it is necessary
to design special experiments in which it is possible to favor one theory over
the other. Three situations give strong support to the sign-learning
alternative. These are experi-ments on reward expectancy, on place learning,
and on latent learning.
1. Reward
expectancy. One of the earliest and most striking observa-tions
on reward expectancy was that of Tinklepaugh (1928). I n his ex-periment, food
was placed under one of two containers while the monkey was looking but
prevented from immediate access to the cans and food. Later the monkey was
permitted to choose between the containers and
No comments:
Post a Comment