- The purpose of this study was to investigate the memory factors
associated with preschoolers' performance in a two-choice, uncertain
outcome decision situation. Choice behavior was examined under
two conditions: one, (NMA), in which the child was required to
rely on his memory to establish an appropriate prediction of the
occurrence of two events; and one (MA), in which the child was provided
with a visual record of events as they actually occurred, which
he might scan in order to establish an appropriate prediction. Five
major hypotheses were tested:
1. Childrens stable-state strategies will not differ in the
MA vs. the NMA condition for either sex.
2. Children of both sexes will make a smaller proportion of
errors in arriving at a stable-state strategy in the
memory aid condition vs. the non-memory aid condition.
3. Children of both sexes will stabilize earlier in the series
in the memory aid condition vs. those in the non-memory
4. The proportion of children who stabilize at a level higher
than the common median for the two event distributions
will be larger in the 75:25 event condition than in the 65:35
event condition; the proportion will not differ for boys and
girls, but will be significantly greater than the proportion
expected by chance.
5. There is no difference in the predicted and observed
stable-state strategies of subjects for either sex within
The sample used for the comparison consisted of 39 middle-class preschool children chosen on the basis of availability. Subjects
were randomly assigned to all experimental conditions in an
effort to control for bias due to individual differences, practice, and
The apparatus was a pegboard with 106 holes. The two events
were defined by the position of a peg in the board, an "up" position
or a "down" position. The occurrence of the two events was determined
by one of eight random schedules. The subjects served as
their own controls across the memory conditions.
The data for the first, third, fourth and fifth hypotheses were in the form of p values, the proportion in which the subject chose
the most frequently occurring event. The data for the second hypothesis
were subjects' errors in choices of the events as they actually
occurred in the first 80 "learning" trials.
The first hypothesis, which predicted that subjects would perform
at a higher level when provided with a memory aid, was not
tenable as the data were found to support the alternative hypothesis
that subjects without the memory aid exhibited a more adequate performance.
Girls performing without the memory aid exhibited a
higher stable-state performance than when they were provided with a
memory aid. For boys, stable-state behavior in the two memory
conditions did not differ.
The test of the second hypothesis, which pertained to accuracy
of performance, showed that proportion of errors across the first
80 trials was similar for both sexes, whether or not a memory aid
The third hypothesis, regarding rate of learning, was untestable
since none of the experimental groups stabilized. The learning
curves generated from the median data indicated that learning did
occur in the early part of the task and that in the final block of trials,
performance deteriorated in the memory aid condition and improved
in the non-memory aid condition in relation to the level achieved
midway through the task.
Results from the test of hypothesis four, used to determine
whether or not children employed probability rules in selecting the
more frequent event in the stable-state, indicate that subjects did
not change in performance in the 75:25 event distribution in relation
to their performance in the 65:35 event distribution. This was the
case whether or not a memory-aid was provided. In fact, only
three of the experimental groups performed at a level better than
chance, all of which were performances in the non-memory aid condition;
all memory aid performances plus the performance of the
boys in the non-memory aid 75:25 event distribution failed to exceed
The quantitative test of the Siegel mathematical model of
choice behavior, provided through hypothesis five, indicated that the
model's predictions of stable-state strategies did not differ significantly
from the observed stable-state strategies for any of the
groups in either memory condition. That is, the Siegel model allowed
for quantitatively accurate predictions of stable-state choice
behavior of both sexes in this preschool sample, whether or not a
memory aid was provided.
The discrepancy in performance found in favor of the non-memory aid condition was tentatively explained in terms of the inability
of preschoolers to use the complex information made available
through the memory aid and in terms of the concept of utility of variability. The fact that only three quarters of the subjects performed
at a better than chance level (and these only in the non-memory aid condition) severely limited the information yielded by
the study and the comparisons that could be made with other studies.
In general, the results are consistent with the findings of Weir's
1967 study in which a memory aid was used in conjunction with reinforcement.
The study points up the need for further investigation
of memory factors as they relate to probability learning, asymptotic
performance and an adequate memory aid, with regard to choice
behavior of preschool children.