26
THOMAS
ROWLAND
AND
CARSON
MCGUIRE
Following the first experiments, Harlow continued his studies, presenting a set
of eight monkeys with a series of discrimination problems. The monkeys were
given fifty trials to learn the discrimination of one color from another, then another
fifty trials to learn to discriminate between two different colors. Another problem
involved discriminating a particular container by shape, i.e. square
vs.
round. The
results confirmed Harlow’s earlier hypothesis that an organism develops “learning
sets,” i.e. that one can learn to learn. In the first three trials, the organism’s per-
centage of successful discriminations was not significantly greater than chance,
and successful discriminations were made in only about seventy-five percent of
the instances after six trials, but the amount of successful discrimination
after
experience with the problem
rises dramatically. After much experience, only the
ht
instance is chance
or
lower, with a sharp increase to almost perfect discrimi-
nation thereafter. The animals learned how to solve discrimination problems.
Harlow’s “educated” monkeys eventually needed only one trial to make the proper
solution.
Using the same set of monkeys, Harlow (1949) complicated the environment
and demonstrated that organisms can
learn
to
solve reversal problems.
The reversal
problem involved switching the relevant stimulus after an initial learning experience,
i. e. in the
first
few trials the color of the cover was rewarded, but a reversal was
made and the unrewarded color became the relevant and rewarded stimulus. At
the point of reversal, the monkeys made many mistakes in discrimination but after
experience with the problem, the percentage of correct respqnses rose to ninety-
seven percent correct. In addition, after having learned the reversal problem, the
monkeys were eventually able to achieve almost perfect performances in similar
problem situations after the first trial.
Further extending his work, Harlow (1951a) applied
his
concept of
oddity
problem.
In such a problem, the organism is presented with three stimulus objects,
two of which are identical. The previous relevant stimuli of perceptual cues (color)
and position now had to be ignored by the organism, and response had to be made
to one shape in comparison to two other shapes which themselves had to be com-
pared. As was expected, the monkeys made many errors in the initial phases of the
problem, but they showed progressive transfer from their previous experience with
discrimination problems to the situation in which they were placed.
At
the end of
Harlow’s training experiences, the monkeys were making ninety percent correct
discriminations on the
first
trial.
That learning sets may operate as relatively
isolated central
units
was demon-
strated by Harlow (1949) in an experiment with six monkeys with previous ex-
perience with discrimination training but no experience with positional discrimi-
nation. After an initial block of problems involving object discriminations, the
monkeys were presented with an equal block of problems involving right-left
positional discriminations. Twenty-five blocks of fourteen problems were alter-
nated between object discrimination sets and positional discrimination sets. The
results showed that the original object discrimination remained at a high level of
accuracy after
a
small decline with the second set which may be explained as inter-
ference from the first set of positional discrimination problems. Throughout the
experience the performance on the previously unexperienced positional discrimi-
nation problems improved steadily until with the last set of such problems the
animals were making eighty-five percent correct discriminations.