Children's Understanding of Probability Concepts

Author(s): Harry W. Hoemann and Bruce M. Ross

Source: Child Development, Vol. 42, No. 1 (Mar., 1971), pp. 221-236
Published by: Wiley on behalf of the Society for Research in Child Development
Stable URL: http://www.jstor.org/stable/1127077
Accessed: 27-06-2016 09:40 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1127077?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Society for Research in Child Development, Wiley are collaborating with JSTOR to digitize,
preserve and extend access to Child Development

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 CHILDREN'S UNDERSTANDING OF PROBABILITY
CONCEPTS

HARRY W. HOEMANN and BRUCE M. ROSS

Center for Research in Thinking and Language,

Catholic University of America

HOEMANN, HARRY W., and Ross, BRUCE M. Children's Understanding

of Probability Concepts. CHILD DEVELOPMENT, 1971, 42, 221-236. 4 ex-
periments administering probability judgment tasks are reported using child
Ss ranging in age from preschool to early adolescence. Experiment 1 showed
equivalent results with probability and proportionality instructions when
judgments were performed between 2 circles with different black and white
proportions. Experiment 2 showed that fewer correct probability than propor-
tionality judgments occurred when Ss judged a single circle. It was concluded
that the 2-circle task does not require probability concepts, since Ss need
not construct probability ratios to succeed. These results confirm those of
Piaget and Inhelder. Experiments 3 and 4 modified the 2-circle task to re-
quire use of probability concepts and administered a probability task with
double arrays of discrete objects. Results were comparable to those found for
the single-circle task. Researchers who have claimed that preschool children
use probability concepts are criticized since their experimental tasks have
been similar to the unmodified 2-circle task of Experiment 1.

Several recent studies (Davies 1965; Goldberg 1966; Yost, Siegel, &
Andrews 1962) have ostensibly shown that under favorable conditions
children as young as CA 4 and 5 show some understanding of probability
concepts. This result is in opposition to Piaget and Inhelder (1951) who
attribute the beginning of probability knowledge only to children at the

This investigation was supported by research grants NICHHD-02026 and Social

Rehabilitation Services 3111-S. Thanks are expressed to Jo Anne Brown for assistance in
Experiment 4. Parts of this report were presented at the 1968 and 1969 Eastern Psychologi-
cal Association meetings in Washington, D.C., and Philadelphia. Author Hoemann's ad-
dress: Department of Psychology, Bowling Green State University, Bowling Green, Ohio
43402.

[Child Development, 1971, 42, 9.1-936. ? 1971 by the Society for Research in Child Development, Inc.
All rights reserved.]

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

level of concrete operations, approximately CA 6 to 8, and postpone the

possession of consistently adequate knowledge until the level of formal
operations. Neither Piaget and Inhelder nor the later investigations have,
however, been concerned with the extent to which their probabilistic ex-
perimental tasks are in fact solved by the use of probability concepts. It
is our contention that despite the high face validity of the tasks that have
been used to determine knowledge about children's probability concepts,
probability considerations in these tasks are confounded with other vari-
ables-such as the ability to discriminate perceptual differences accurate-
ly-that are not germane to the assessment of the child's understanding
of probability. If this is the case, it considerably weakens the argument
of most American studies, since they have taken the minimal criterion of
above-chance performance as evidence that the preschool child possesses
some rudimentary knowledge of probability.
The present investigation consists of four experiments. In the first
two experiments a separate assessment was made of the ability to per-
form accurate proportionality judgments, an ability which can be mea-
sured independently of probability considerations even though it is re-
quired for correct probability choices. Where errors for probability
judgments closely matched those made while performing under propor-
tionality instructions, it was concluded that probability concepts were
not being measured. Experiment 1 required a choice between two odds
arrays and Experiment 2 a choice within a single array. Past results have
uniformly found two-array tasks easier (Goldberg 1966; Piaget & Inhel-
der 1951; Yost et al. 1962). Piaget and Inhelder have offered an explana-
tion for this result which we could partially test by using the above
design of separate proportionality and probability instructions with each
type of task.
Besides the probability task of Experiment 2, two additional prob-
ability tasks were administered. In Experiment 3 we administered a
modified two-array task in an attempt to make it comparable in difficulty
to the single-array task. This attempt was successful. In Experiment 4
the modified two-array task was altered by the use of different materials
in order to extend our results to a task that offered choices between col-
lections of discrete objects instead of the area-size comparisons required
in the earlier experiments. Children of preschool, concrete, and formal
operational ages were used as Ss in the first two experiments, and children
of CA 7 and 11 in Experiments 3 and 4. Ages were selected on the basis of
appropriateness for revealing the correlation between operational stage
and probability knowledge given the level of difficulty of our experimen-
tal tasks. A group of older deaf children also was administered the task
in Experiment 4 to determine whether language deficiency had a mea-
surable effect on performance.

222

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

EXPERIMENT 1

Method

Subjects.-The Ss in this and subsequent experiments, with the ex-

ception of the deaf Ss, were enrolled in private and parochial schools in
Washington, D.C., and adjacent Maryland suburbs. The Ss were from
predominantly middle-class homes and making normal progress in school.
In Experiment 1 there were separate probability and proportionality
instruction groups of 20 Ss, 10 boys and 10 girls, within 12 months of the
mean ages 4-7, 6-0, 7-5, and 10-6. Thus there were eight groups in all
with a total of 160 Ss.
Design and rationale.-The task required of S was to choose one of a
pair of paper disks 12 inches in diameter placed in front of him. Each
disk had a different pattern of black and white wedges. With probability
instructions the S was required to choose which disk he preferred to spin
for the named color E designated on a particular trial (either black or
white) by means of a plastic spinner attached to a dowel inserted in the
center of each disk. With proportionality instructions spinners and
dowels were removed and S was asked to choose which disk had the most
black (white) in it. With both types of instructions disks were changed
every trial. Note that with this spinner task the correct answer is exactly
the same for a specified disk under either proportionality or probability
instructions. To determine the correct answer also does not involve
counting, as the ordinal "more than" judgment is sufficient to provide
the correct answer.
Odds variables.-Three odds difference levels between disks were
administered: -, 1, and -. These odds differences can be expressed as a
proportionality ratio between disks, using the proportion of the desired
color that covered the surface of each disk as a basis. Each difference
level was administered for 12 trials, with three different disk comparisons
administered at each odds difference. For the - odds difference one circle
was covered - by the designated color versus -i of the same color for the
other circle (- minus - equals 4); other comparisons at the ? odds differ-
ence were I versus 1 and 1 versus -. For the i difference level, compari-
sons between disks were - versus , versus 3, and - versus -. For the

difference level, comparisons were - versus -, 9 versus 1, and 1 versus

1. Note that with the 1 and 1 difference levels in some trials the desig-
nated color was in the majority in both circles, while in other trials it was
in the minority in both circles.
Controls.-There were 36 trials with half the trials having black as
the correct choice and half white; likewise half the trials had the left
spinner as correct choice and half the right. An additional variable was
also introduced in the configuration of the disks. Half the trials had solid

223

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 CHILD DEVELOPMENT

black and white wedges constituting the disks; the remaining trials had
the black and white segments of each disk broken up into several alter-
nating wedges so that all of the black or white in a given disk was not
contiguous. Specifically, for the other 18 trials a given black-white pro-
portionality in one circle was divided into two wedges of each color in
one circle and four wedges of each color in the other. All wedges of the
same color in the same disk were of the same size. A multi-wedge example
of a versus , comparison with black in the majority displayed four al-
ternating wedges of -I- black and - - white in one circle versus eight alter-
nating wedges of - black and - white in the other circle. The variable
of number-of-wedges in the disks was interspersed randomly in the trial
sequence.
Apparatus and procedure.-The disks were 20 pairs of 12-inch circles
of heavy drawing paper composed of black and white wedges, with India
ink used in coloring the black wedges. The disks were stacked on two
adjacent pieces of 14-inch square plywood on a table in front of S. In-
serted dowels were - X 3 inches in length, and the spinners were plastic
circles 4 inches in diameter with a black point painted against a white
background. The first two pairs of disks were used for instruction and six
practice trials. The pairs used for practice included disks having gray
wedges with the color of the alternating wedges white on one disk and
black on the other. Thus possible versus impossible choices of both
"black" and "white" were included in the practice trials. The S was in-
formed that these initial trials were practice trials. Each of the 18 re-
maining disk pairs appeared twice in the course of the 36 experimental
trials, once with black and once with white as the E-specified color.
Either a black or a white 3 X 4-inch card was held up by E to accompany
the verbal designation of the specified color for each trial.
All Ss in this and subsequent experiments were tested individually.
Under both probability and proportionality instructions S was told at
the beginning of a session that he was to play a game with E.
With probability instructions children were told: "I want you to look
at the two spinners very carefully and show me which one you will spin to
make the pointer point to black (white)." Although verbal responses were
permitted, pointing to the chosen spinner was also required of Ss. The S
then spun the spinner on the disk he had pointed to. Responses and out-
comes were recorded by E, and favorable outcomes, even when incorrect
from a probability standpoint, were verbally acknowledged by E as suc-
cessful predictions. With proportionality instructions where there were no
spinners present, children were told: "I want you to look at these two
circles very carefully and show which circle has the most black (white)."
No indication as to whether S was correct on any trial was made by E
with proportionality instructions so that knowledge of results would not
affect judgments. Instead general encouragement was given the child

224

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

throughout the course of each session. Under both instructions a pair of

disks was removed after each trial exposing the next pair.

Results

Mean percentages of correct responses for the independent prob-

ability and proportionality groups at each age are presented in table 1,
broken down by odds differences. Further, the differences between
proportionality and probability instruction groups are shown for each
comparison. Unless otherwise specified, median tests were used through-
out these experiments in performing significance tests. Median tests were
judged more appropriate than other nonparametric tests which tend to
produce spurious significance with the large number of ties and small
score variance generally obtained. The obtained differences between com-
bined proportionality and probability groups are not significant at any
age. Under both instructions with increasing odds differences, increasing

TABLE 1

MEAN PERCENTAGES CORRECT FOR PROPORTIONALITY

AND PROBABILITY INSTRUCTION CHOICES FOR THE
Two-SPINNER TASK IN EXPERIMENT 1

ODDS DIF- CHRONOLOGICAL AGE

FERENCE AND

INSTRUCTIONS
41 6 7? 101
I.

Proport ..... 92 99 100 100

Prob........ (83) (93) (98) (100)
Diff....... 9 6 0

4'

Proport...... 80 95 97 99
Prob..... . . . (80) (87) (94) (97)
Diff....... 0 8 3 2

Proport...... 72 88 94 96
Prob...... (62) (82) (89) (93)
Diff........ 10 6 5 3

Combined:
Proport...... 81 94 97 99
Prob......... (75) (87) (94) (97)
Diff....... 6 7 3 2

225

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

accuracy in performance is apparent. Since performance on the non-

probabilistic task adequately predicts results, it can be tentatively con-
cluded that probability concepts do not contribute to correct solutions.
One reason that this conclusion must be qualified is that the task is
rather easy, at least for the older age groups. A related objection is that
difference scores may not be a sensitive measure even at younger ages,
since children who possess some low-level probability concepts may al-
ways be able to make the direct and obvious inference from proportions
to probability predictions when they can perform the correct proportion-
ality judgments. However, if situations of a rather similar type can be
shown to produce significant discrepancies between proportionality and
probability judgments, the conclusion that probability concepts are not
required with the two-spinner task would be strongly sustained. Related
situations that do show significantly poorer performance under prob-
ability instructions (with older children also) will be considered in sub-
sequent experiments.
Between adjacent age groups a significant difference was found only
between proportionality groups at CA 4- and 6 (x2 = 13.33; df = 1;
p < .01), although the probability difference also approached significance
between these ages (p < .10). Among the three odds differences only pro-
portionality at CA 41 and at CA 6 showed significant differences (p <
.001 by Friedman two-way analyses of variance). Note that if the prob-
ability task had been administered by itself, these significant odds differ-
ences would have been inappropriately attributed to differences in risk-
taking behavior.
A supplementary experiment was carried out with the same proce-
dure used in the main experiment except for certain differences in disk
comparisons. There were 60 children, with 20 children again in each age
group; testing was carried out at three ages with mean CAs of 4-6, 6-6,
and 8-6. There was a difference between this and the main experiment in
wedge patterns, but this difference was unimportant for the present dis-
cussion. The two major variables of interest are that (a) the odds differ-
ence - was substituted for - while keeping the differences - and 1, and
(b) the same children performed the proportionality judgments after
first making probability judgments with the identical disk pairs. With
one exception neither the probability nor the proportionality judgments
of this experiment were more than 10 percent different from those of the
previous experiment at the odds differences of - and 1. As expected, for
both proportionality and probability judgments in comparing the closest
comparable ages less accuracy was attained for the s than - odds. On
the basis of these results, it appeared reasonable in the next experiment
with a single-spinner situation to use the more economical technique of
having the same Ss perform probability judgments followed by propor-
tionality judgments.

226

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

EXPERIMENT 2

Method

Subjects.-Twenty Ss, 10 boys and 10 girls within 6 months of the

mean ages 4-3, 5-3, 6-2, 7-2, 8-2, 12-2, and 13-2 for a total of 140 Ss, were
tested in Experiment 2 successively with both the proportionality and
probability instructions.
General Design.-The task required of Ss was to indicate nonverbally
with a single spinner which color the pointer would point to when it
stopped, black or white. Subsequently in the same session Ss in the five
younger age groups were administered the proportionality instructions
and asked to indicate which color, black or white, comprised the larger
total area. Odds differences were arranged so as to replicate those in
Experiment 1: . black versus - white and vice versa for the ' odds differ-
ence, . versus - for the 1 odds difference, and 9 versus 7 for the - odds
difference. There were eight trials at each odds difference for a total of
24 trials. As an additional variable there were six trials with solid black
and white wedges, six trials with two black and two white wedges, six
trials with three black and three white wedges, and six trials with four
black and four white wedges. Disks with two black wedges and disks
with four black wedges were symmetrical. The trial sequence, identical
for both the probability and proportionality instructions, was determined
by a random ordering of odds differences, number of wedges, and color of
the correct response.
Procedure.-In the probability instruction Ss were told, "I want you
to spin this pointer, but first show me what color it will point to when it
stops." Favorable outcomes were reinforced verbally regardless of the
correctness of the prediction from a probability standpoint. The spinner
was removed by S after each trial, and E removed the uppermost disk,
exposing the disk underneath for the next trial. In the porportionality
instruction the spinner and its dowel were removed, and Ss were required
to point to the majority color. Responses were recorded by E without
reinforcement.

Results

Major results for Experiment 2 are presented in table 2, where means

and differences are presented in the same way as was done for the two-
spinner task of Experiment 1. In table 2, however, proportionality and
probability measures at a given age were obtained from the same children
and no proportionality judgments were obtained at CA 19 and 13, as it
was assumed that these Ss would give errorless results. This means that
for some comparisons only percentages correct under probability instruc-

227

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

TABLE 2

MEAN PERCENTAGES CORRECT FOR PROPORTIONALITY AND PROBABILITY INSTRUC-

TION CHOICES FOR THE SINGLE-SPINNER TASK IN EXPERIMENT 2

CHRONOLOGICAL AGE
ODDS DIFFERENCE
AND INSTRUCTIONS 4 5 6 7 8 12 13

Proport ............ 74 97 100 99 100

Prob.............. (52) (69) (81) (85) (82) (94) (96)

Diff .............. 22 28 19 14 18 ...

Proport ............. 70 92 98 97 100

Prob............... (56) (66) (78) (81) (76) (86) (87)
Diff.............. 14 26 20 16 24 ...

Proport ............. 68 82 99 93 98
Prob.............. (61) (57) (71) (70) (73) (84) (88)
Diff............... 7 25 28 23 25 ...

Combined:
Proport ............. 70 90 99 97 99
Prob............... (56) (64) (77) (79) (77) (88) (90)

Diffr.............. 14 26 22 18 22

tions can be evaluated. It must first be established, however, that Ss are

to some extent applying probability concepts. To test this possibility in
Experiment 1, we set up the criterion of a significant difference between
results obtained by proportionality and probability instructions. Evalu-
ating this criterion in the present experiment, there were significantly
more errors in the combined probability results as compared to the
combined proportionality results at each age from 4 through 8 (p < .01
or greater using Wilcoxon signed-rank tests). Therefore, unlike our con-
clusions about the double-spinner task, we can claim that the single-
spinner task does test for probability concepts.
At the same time it would not be valid to claim that Ss at CA 4 were
applying probability concepts correctly since their combined mean of cor-
rect judgments is only 56 percent, which is not significantly above chance
performance. With ages 5 through 8 it is appropriate to point out that
combined proportionality judgments are 90 percent or better, but, unlike
Experiment 1, the reaching of the proportionality performance ceiling
does not correspondingly improve probability judgments.
For the sake of uniformity and to avoid this ceiling effect in assess-

228

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

ing age differences in ability to apply probability, we used the probability

instruction scores rather than difference scores in making between-
age probability comparisons. The only significant difference for combined
probability was between CA 8 versus 12 (x2 = 6.42; df = 1; p < .025).
When odds differences under probability instructions are considered,
it can be observed that shorter odds at all ages produce fewer correct
responses with probability instructions, except at CA 4 where probability
responding was not significantly above chance. However, the range was
usually narrow, and only at CA 7 was there a significant odds difference
(p < .01 by Friedman two-way analysis of variance).

COMPARISON BETWEEN EXPERIMENTS 1 AND 2

As pointed out in the introduction, Piaget and Inhelder (1951, pp.

160-161) give a theoretical explanation of the results found in comparing
Experiment 1 with Experiment 2, namely, that it is easier for children to
choose between two odds ratios the one with more likely odds than to
choose the more likely event in a single-odds ratio. Briefly put, Piaget
and Inhelder claim that to perform the single-odds task correctly the
child must decompose all possible outcomes into a complete set of frac-
tions representing each different outcome. Ratios are then constructed
and compared using the favorable outcomes as the numerator and the
total number of possible outcomes as the denominator of each odds
ratio. Thus even where only two outcomes are possible a two-step pro-
cess is required.
With a double-odds task, on the other hand, no decomposition of
possible outcomes is necessary as long as either the numerator or the
denominator is the same in both odds ratios. This explanation, which
was given by Piaget and Inhelder for children's performance with double-
odds ratios of discrete objects, appears to apply as well to the calculation
of magnitude ratios in the present two-spinner task. To be correct in the
two-spinner task the child need only compare numerator magnitudes,
that is, compare across circles the magnitudes of the color designated by
E. This direct magnitude comparison is satisfactory because the two spin-
ners always have the same denominators since, necessarily, the sum of
the sectors of every circle must always be equivalent. Empirical support
for the supposition that only comparisons between the E-designated
color were performed comes from the close correlation in Experiment 1
between results with probability as compared to proportionality instruc-
tions at each odds level.
When the single-spinner task is considered, the result that prob-
ability instructions produced poorer results than proportionality instruc-
tions does not of itself prove that the Piaget and Inhelder explanation of

229

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 CHILD DEVELOPMENT

the process involved in single-odds probability predictions is correct.

But their explanation of the difficulty found with single-odds tasks sug-
gests that a double-odds task would be of the same order of difficulty if
the child performing the double-odds task also had to deal with fractions.
The attempt to equate the difficulty of the double-spinner task with that
of the single-spinner task and thereby make the former task more nearly
a probability task in fact as well as appearance supplies the raison d'etre
for Experiment 3.

EXPERIMENT 3

The only procedural change for the two-spinner task in this experi-
ment as compared to Experiment 1 was that on each trial Ss were re-
quired by E to compare different rather than the same colors in the two
circles (e.g., spin for black in the left circle or spin for white in the right
circle). Faced with this task, the child's easiest course to obtain a correct
solution is simply to compare the magnitudes of the E-designated unlike
colors in each circle. Therefore performance with the modified double-
spinner task can be just as good as in Experiment 1.
The main point is, however, that to justify direct magnitude compar-
isons between circles the child must have made some use, if only implicit,
of fractional comparisons. For direct comparison of designated magni-
tudes of unlike colors is only legitimate if the child has some realization
that the denominators in the designated-color to complete-circle ratios of
each disk are equal. This fact should be especially transparent with the
present task since the two disks are always the same size. Nevertheless, it
was our hypothesis that even this minimal formulation of geometrically
represented fractions goes beyond the operations used by the younger
children in Experiment 1. Thus we predicted that results with the modi-
fied two-spinner task should be more like the results obtained with the
single-spinner task of Experiment 2 than the double-spinner task of
Experiment 1.

Method

Subjects.-Two independent groups of 7-year-old Ss, 13 in each

group, were tested in Experiment 3. The mean age for both groups was
7-3, with an 8 to 5 sex division in each group.
Procedure.--One group was administered a Modified Double Spin-
ner Condition and the other a Single Spinner Condition. The Modified
Double Spinner Condition was the same as that received by the Ss in
Experiment 1 who performed under probability instructions, except that
Ss were required to compare dissimilar colors: "Will you spin here for

230

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

black [E placed 1-inch square pieces of black poster board on the black
wedges of the designated disk] or here for white? [E placed 1-inch square
pieces of white poster board on the white wedges of the designated disk]."
The position and color of the correct response were in a random sequence
with odds differences and number of wedges occurring in the same order
as in Experiment 1. The procedure for the Single Spinner Condition was
exactly the same as in Experiment 2 except that there were eight trials
each at 1, , , and -L difference levels administered in a random order.

Results

The means for odds differences of the two groups in Experiment 3

are presented in the left side of table 3 along with the means for 7-year-
olds taken from table 2 for comparison purposes. Omitted is the result
that children in the Single Spinner Condition were 99 percent correct, with
the - odds difference which was not administered in any other condition.
It can be seen that the means for combined odds are strikingly close for
all three groups. In the Experiment 3 single-spinner group there is one
inversion as to better performance with shorter odds. But except for
this, performance at each odds level is quite similar. It can be concluded
that even with relatively small N's the double-spinner task can be equated
with the single-spinner task.

EXPERIMENT 4

The possibility still remains that the reported results apply only to
the spinner situation. Experiment 4 was administered to extend the pres-
ent findings to a situation with two collections of discrete objects, making

TABLE 3

MEAN PERCENTAGES CORRECT FOR PROBABILITY CHOICES IN EXPERIMENTS

3 AND 4 WITH COMPARISON DATA FROM EXPERIMENT 2

ExP. 9 ExP. 3 ExP. 4: DISCRETE OBJECTS

Single Single Double

Spinner Spinner Spinner CA 11, CA 11,
ODDS CA 7 CA 7 CA 7 CA 71 Hear Deaf
DIFFERENCE (N= 90) (N= 13) (N= 13) (N= 90) (N= 17) (N= 14)
............... 85 87 94 82 99 93
4............... 81 91 84 78 86 90
S ............... 70 68 79 73 84 85
Combined........ 79 89 83 78 87 89

231

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

it more analogous to the tasks used by previous American investigators

as well as those used by Piaget and Inhelder. In contrast to Experiment
3, older 11-year-old children (and verbally deficient deaf children) were
included as Ss to determine if task results could be obtained like those
found for the older children with the single-spinner task in Experiment 9.

Method

Subjects.-Three independent groups of Ss were administered the

task. A group of 20 7-year-olds about evenly divided as to sex was drawn
from the first and second grades of a parochial school in a lower-middle-
class neighborhood in Washington, D.C. Ages ranged from 7-6 to 7-11
with a mean of 7-8.
A group of 11-year-olds from a middle-class parochial school, nine
boys and eight girls, was also administered the task. The age range was
from 10-11 to 12-5, with a mean of 11-7. A third group consisting of
eight boys and six girls was administered the task at a state residential
school for the deaf. All these Ss were either congenitally deaf or had
suffered the onset of deafness in infancy. Except for one boy of 9-4, the
age range was from 10-2 to 12-3, with an overall mean of 11-2.
Procedure.-Two 1-gallon glass jars 51 inches in diameter and 10
inches high with openings 4 inches across were substituted for the two
disks. The jars, which were transparent with minimal distortion, were
placed on a table in front of S with an approximate 6-inch separation.
Pink and green table tennis balls were used to form the odds collections
in each jar. As an example, if a - difference level were desired, the trial
might be between six pink and two green balls in the left jar and one pink
and seven green balls in the right jar, with E asking: "Would you rather
pick a green ball in this jar [pointing to the left one] or a pink ball in
this one [pointing to the right one]?" In this way the same order of odds
comparisons was performed as in Experiment 3; as previously noted this
order also matched that of Experiment 1 except for the specification of
different colors. Thus there were again 36 trials, 19 at each odds level-I,
1, and i-with half the trials requiring "pink" and half "green" as the
correct choice, and half the left jar and half the right. Note that the balls
were never stacked but lay flat on the bottom of the jars so that they
were easy to count.
After the child pointed to the jar with the color he preferred to draw
for, the jar was covered with a white opaque cloth that had a pull draw-
string. The E shook the heavy jar and set it down in front of the child,
who either closed his eyes or looked away while he reached in through
the opening in the cloth with one hand and drew out a single ball. The
balls were rearranged by E after each trial. The first four trials were prac-

232

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

tice trials with even choices (e.g., two pink and two green in the left jar
and four pink and four green in the right jar) administered to make sure
Ss understood the instructions. On all trials a star was recorded on the
answer sheet in view of the child whenever he drew out the specified
color.
With deaf children special care was taken to insure that they under-
stood instructions. Printed labels of "pink ball" and "green ball" were
matched to a ball of the appropriate color initially, and on every subse-
quent trial a pink ball was put in front of the jar where S was required to
predict "pink" and a green ball in front of the jar where he was required
to predict "green" by pointing.

Results

The means for odds differences of the three groups in Experiment 4

are presented in the right half of table 3. Although consistent improve-
ment was shown for the older as compared to the younger 7-year-olds,
differences among combined totals were not statistically significant.
Results for the 7-year-olds were very similar to those for the other 7-
year-old groups in previous experiments shown in table 3. Similarity
between the deaf and hearing 11-year-olds was also great, and this simi-
larity can be extended to include the 12- and 13-year-old groups in table
2. In fact, it can be observed that among these four older groups in tables
2 and 3 the maximum range at any odds difference is a surprisingly small
4 percent.
From such a high degree of similarity the conclusion seems warranted
that the discrete object task is equivalent to the double-spinner task of
Experiment 3 and that, therefore, the discrete object task involves the
application of probability concepts. This conclusion rests partly on
inference, since proportionality errors were not measured independently.
However, it seems safe to assume that since Ss could simply compare the
balls of the designated colors by counting, near-perfect scores would have
been attained under proportionality instructions.

DISCUSSION

The present experiments suggest that sucessfully choosing the more

favorable odds need not give an index of probability knowledge. Partic-
ularly if direct magnitude comparison can take place, accurate perfor-
mance may tap the same skill as choosing the larger of two cookies, a
task at which all preschool children are adept. It is possible, nevertheless,
to maintain that above-chance performance in a probability-structured
task of any variety does indicate some rudimentary probability knowl-

233

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

edge. A difficulty with this conception, as with cookie choice, is that

children perform much too well. For example, in Experiment 1 we found
that for children as young as CA 4-, failure to discriminate proportions
accurately appeared to be the main constraint on correct performance
under probability instructions.
In opposition to several recent investigators, we have taken the
position that not all tasks which are nominally probability tasks require
use of probability concepts. Some guarantee is necessary that Ss con-
struct and compare fractions or ratios other than obvious perceptual
magnitudes before we can accept that probability concepts are being ap-
plied. In the single-spinner task of Experiment 2 and the contrasting
color two-array tasks of Experiments 3 and 4 we believe that we have met
this criterion. In Experiment 2 we were able to show directly that prob-
ability instructions do contribute a significant error component over and
above that contributed by proportionality discrimination errors. In
Experiments 3 and 4 no independent assessment of proportionality
errors was made, but it seems reasonable to assume that proportionality
errors would be very few by CA 7, just as in Experiments 1 and 2.
Our three quite different probability assessment tasks administered
in Experiments Q, 3, and 4 produced similar percentages of correct
responding at comparable ages. These results were achieved without the
fairly elaborate controls for color preference and odds displays used by
several previous investigators (Goldberg 1966; Yost et al. 1962). Of
course, only with Experiment 2 did we administer to children of preschool
age what we consider a probability task, and they did not perform sig-
nificantly above chance. But there is an interesting methodological point
to consider here. If preschool children can be swayed by color preferences,
how strong can their presumed grasp of probability principles be? Our
interpretation is that young children do indeed sometimes prefer color to
quantity with no probability inference being involved in either case.
A subject-group variation was the administration of the discrete
odds task in Experiment 4 to a group of 11-year-old deaf Ss. Their per-
formance equaled that of the hearing Ss at all odds levels. In a previous
paper (Ross 1966), deaf children had been found to lag behind hearing
children at this age in performing a probability task, although there was
little difference by CA 15. In the previous probability task, successive
ball draws were predicted using as a container an opaque box that hid
the balls from sight after an initial viewing. It is hypothesized that the
use of a visible ball display in Experiment 4 accounts for the deaf
children bettering their performance so that they were the equal of hear-
ing children at a younger age than that found in the previous study.
Our conclusions support the process explanation put forward by
Piaget and Inhelder as to why choice between two arrays is easier than
prediction with a single array. In addition, the greater concentration on

234

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
 HARRY W. HOEMANN AND BRUCE M. ROSS

this comparison in the present study permits us to extend their conclu-

sions. Piaget and Inhelder published only representative protocols rather
than sampling age levels, and they therefore were concerned chiefly with
qualitative descriptions. For instance, they considered the one-array
versus two-array comparison only for a few children with an age range
from 6-1 to 7-6. Our data present results for groups of children at younger
and older ages. Thus we showed in Experiment I that preschool children
of CA 4, performed at the 75 percent level on the two-array task that al-
lowed a direct magnitude comparison, while this level was not reached
until CA 6 in the single-array task of Experiment 2. More important,
progress was sufficiently slow with the single-array task and the compa-
rable modified two-array tasks so that, while children in Experiment 1
attained nearly 90 percent success at CA 7?, this level was not reached in
Experiments 2 and 4 until the onset of the formal operational period at
CA 11 to 13.
Note that there is nothing absolute about these comparisons since
odds can be adjusted to further increase comparison difficulty, as was
reported in the supplementary experiment performed in conjunction with
Experiment 1 where - odds differences were administered. It is also
likely that tasks that can be solved directly by magnitude comparisons
may increase in difficulty when interspersed with problems that cannot
be solved so easily; one here recalls the findings reported with water jar
problems where easy problems were solved by a more complicated pro-
cedure if Ss were set for harder problems (Luchins 1942). A mixed order
of relatively easy and hard problems is in fact shown in some of the pro-
tocols printed by Piaget and Inhelder.
In testing the hypothesis in Experiment 1 that the child was only
depending on magnitude discrimination, we were following the suggestion
by Piaget and Inhelder that the child was generally getting the correct
answer for the wrong reason-wrong from the standpoint of assessing the
child's understanding of probability concepts. We recognize, though,
that it is a matter of definition as to what one calls probability knowledge.
Whereas magnitude discrimination is considered by us as only a precur-
sor, by others it may be considered a genuine probability concept. Our
view clearly accords more with Piaget's cognitive theory in that it puts
the child's earliest glimmers of understanding with regard to probability
concepts at the same developmental stage as the child's first envisioning
of a concept of chance. Piaget and Inhelder's data show that the "dis-
covery of chance" does not occur until the onset of the period of con-
crete operations when the child is 6 to 8 years old. Magnitude estimation,
by contrast, is almost completely mastered by this age and therefore
plays no further role in the elaboration of chance and probability con-
cepts that continues through adolescence.

235

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms
CHILD DEVELOPMENT

REFERENCES

Davies, C. M. Development of the probability concept in children. Child Development, 1965,

36, 779-788.
Goldberg, S. Probability judgments by preschool children: task conditions and perfor-
mance. Child Development, 1966, 37, 157-168.
Luchins, A. S. Mechanization in problem solving: the effect of Einstellung. Psychological
Monographs, 1949, 54 (Whole No. 248).
Piaget, J., & Inhelder, B. La genbse de l'idde de hasard chez l'enfant. Paris: Presses Univer-
sitaires de France, 1951.
Ross, B. M. Probability concepts in deaf and hearing children. Child Development, 1966,
37, 917-997.
Yost, P. A.; Siegel, A. E.; & Andrews, J. M. Non-verbal probability judgments by young
children. Child Development, 1962, 33, 769-780.

236

This content downloaded from 128.163.2.206 on Mon, 27 Jun 2016 09:40:30 UTC
All use subject to http://about.jstor.org/terms