Journal of Personality and Social Psychology

1966, Vol. 3, No. 6, 611-618

AN EXPERIMENTAL TEST OF THE SEQUENTIALITY

OF DEVELOPMENTAL STAGES IN THE CHILD'S
MORAL JUDGMENTSx
ELLIOT TURIEL 2
Yale University
2 developmental propositions of Kohlberg's theory of moral judgments were
tested: (a) that the stages form an invariant sequence, and, thus, more learning results from exposure to the stage directly above one's level than to stages
further above; (b) that passage from 1 stage to the next involves integration of the previous stages, and, thus, more learning results from exposure to
the stage directly above than to the stage 1 below. First, Ss' stages were
determined in a pretest. 44 Ss of Kohlberg's Stages 2, 3, and 4 were equally
distributed among 3 experimental groups and 1 control group. In the treatment conditions, Ss were exposed to either the stage 1 below, 1 above, or 2
above the initial dominant stage. The control group was not administered a
treatment condition. In a posttest the influence of the treatment conditions
was assessed. The results confirmed the hypotheses since exposure to the stage
directly above was the most effective treatment.

Moral development has been approached

from different viewpoints. Developmental
theories such as Piaget's (1948) focus on
the cognitive processes underlying moral responses and assume that the organization of
these processes is different at different stages
of development. The greater part of developmental research on morality has stemmed
from Piaget's theory of moral stages, stages
supported only to a limited extent by subsequent investigations (see Kohlberg, 1963b).
Kohlberg (1958, 1963a) has postulated the
following set of moral stages, which are based
on children's reasoning in response to hypothetical moral conflicts (Kohlberg, 1963a) :
Stage 1: Punishment and obedience orientation.
Stage 2: Naive instrumental hedonism.
Stage 3: Good-boy morality of maintaining good
relations, approval of others.
Stage 4: Authority-maintaining morality.
study is based on a dissertation presented
to Yale University in candidacy for the degree of
Doctor of Philosophy. It was conducted while the
author held a United States Public Health Service
predoctoral fellowship. The author wishes to express
his gratitude to the members of the dissertation
committee: Edward Zigler, Irvin Child, Merrill Carlsmith, and Robert Abelson. The author is also
indebted to Lawrence Kohlberg for his invaluable
advice and aid. Thanks are due to Rita Senf for
her critical reading of the manuscript.
2
Now at Bank Street College of Education and
the Center for Urban Education, 33 West 42nd
Street, New York, New York.

Stage 5: Morality of contract and democratically

accepted law.
Stage 6: Morality of individual principles of conscience [pp. 13-14].

While space does not permit a detailed

definition of Kohlberg's stages nor of his
methods for the elicitation and stage classification of responses, the Method section
should clarify the nature of his data.
Kohlberg postulated that his stages define
a sequence normally followed by each individual. The sequence of the stages is hypothesized to be invariant, with the attainment
of a mode of thought dependent upon the
attainment of the preceding mode, requiring
a reorganization of the preceding modes of
thought. Evidence for this hypothesis (Kohlberg, 1963a, pp. 15-17) consists, first, of
findings of age differences, in various cultures,
consistent with the notion of sequence and,
second, of findings of a "Guttman quasisimplex" pattern in the correlations between
the various types of thought, a pattern
expected if they form a developmental order.
While this evidence supports the validity
of the stages as forming a fixed sequence,
there has been no experimental evidence. The
aim of the present research was to subject
Kohlberg's hypotheses to an experimental test.
In particular, the concept of developmental
sequence suggests some hypotheses regarding
developmental change and learning of new

611

ELLIOT TURIEL

612

moral concepts. The plan of the study was

to select subjects at varying developmental
stages, expose them to moral reasoning that
differed from their dominant stage, and then
test the amount of learning and generalization of the new concepts. First, part of the
Kohlberg moral judgment interview was administered to determine the subject's dominant stage. With the remaining part of the
Kohlberg interview, the subject was then exposed to concepts corresponding to a stage
differing from his own. Some subjects were
exposed to the stage that was one below their
own, some to the stage one above, and some
to the stage two above. Finally the subject
was retested on the entire interview. If Kohlberg's stages do form a fixed developmental
sequence, so that the attainment of a mode
of thought is dependent on the attainment
of the preceding mode, then it is expected
that subjects exposed to the stage directly
above their dominant stage would show more
usage of that stage on the retest than would
subjects exposed to stages two above or one
below.
Thus this study was designed to test the
following two hypotheses:
1. That Kohlberg's stages form an invariant sequence so that an individual's existing
mode of thought determines which new concepts he can learn. It was expected that subjects exposed to reasoning corresponding to
a stage directly above their dominant stage
would be influenced more than those exposed
to reasoning corresponding to a stage further
above.
2. That each stage represents a reorganization of the preceding stages, and in effect is
a displacement of those stages. If each stage
is a reorganization of the preceding stages,
rather than an addition to them, then a
tendency to reject lower stages would be expected, so that subjects exposed to a stage
one above would be influenced more than
those exposed to a stage one below their own.
METHOD
Subjects
This experiment used 44 seventh-grade boys from
the New Haven public schools, between the ages
of 12-0 and 13-7. These boys, chosen at random
from the school files, were from the middle socio-

economic class, as determined by their parents'

occupation and education level.

Scoring Methods
An individual's developmental stage is determined
by using Kohlberg's (1958) moral judgment interview, which contains nine hypothetical conflict
stories and corresponding sets of probing questions.
The following story is an example:
In Europe, a woman was near death from a
special kind of cancer. There was one drug that
the doctors thought might save her. It was a
form of radium that a druggist in the same town
had recently discovered. The drug was expensive
to make, but the druggist was charging ten times
what the drug cost him to make. He paid $200
for the radium and charged $2,000 for a small
dose of the drug. The sick woman's husband,
Heinz, went to everyone he knew to borrow the
money, but he could only get together about
$1,000, which is half of what it cost. He told the
druggist that his wife was dying and asked him
to sell cheaper or let him pay later. But the
druggist said: "No, I discovered the drug and I'm
going to make money from it." So Heinz got
desperate and broke into the man's store to steal
the drug for his wife. Should the husband have
done that?
Two scoring procedures are available for determining a subject's scores on each of the six stages.
(The stage with the highest score represents his
dominant stage.) The first, a more global method,
involves the use of rating forms devised by Kohlberg
(19S8). A second scoring procedure uses detailed
coding forms (Kohlberg, 1958) for each of the nine
situations of the interview. These coding forms were
constructed and standardized on the basis of responses given by a large number of subjects. Each
response listed in the coding forms has a stage
assigned to it. A subject's responses to a given situation are divided into "thought-content" units, and
each unit is assigned to a stage, as determined by
the stage classification of that unit in the coding
form. In this way the total number of units assigned
to each stage is determined.

Design and Procedure

There were three steps in the experimental procedure. The subject's dominant stage was determined
by a pretest interview. In the experimental session
subjects were exposed, through role playing, to concepts that were either one below, one above, or two
above their initial dominant stages. These experimental treatments will be referred to as 1, +1, and
+2 treatments, respectively. The treatment groups
were equated on IQ via the Ammons Full-Scale
Picture Vocabulary Test. In a posttest interview the
subjects' stage scores were reassessed to determine
the influence of the treatment.
Pretest selection interview. During the first meeting each subject was individually administered six

STAGES IN MORAL JUDGMENTS

of the nine situations of the Kohlberg interview in
order to determine his initial stage scores. A tentative
assessment of each subject's scores was made using
Kohlberg's global rating forms. Only those subjects
whose scores on the dominant stage were twice as
large as their scores on the next most dominant stage
were retained. In all, 21 subjects were discarded,
while the 48 retained were equally distributed among
Kohlberg's Stages 2, 3, and 4.
Since the global rating system did not provide the
sensitivity desired for the experiment, the protocols
of subjects retained were rescored using Kohlberg's
detailed coding forms. Only those subjects who then
scored higher on their dominant stage, as determined
by the global ratings, than on any other stage were
retained. Four subjects were thus discarded, leaving
a total of 44.
Experimental treatment conditions. All subjects of
a given dominant stage were randomly assigned to
the control group or to three experimental groups
(JV = 11 per group). In the experimental treatments,
administered 2 weeks after the pretest, subjects were
exposed to moral reasoning in individual roleplaying situations with an adult experimenter. In one
treatment the reasoning presented was one stage
below the initial dominant stage (1 treatment);
the second treatment group was exposed to reasoning
that was one stage above (+1 treatment); and in
a third treatment the reasoning presented was two
stages above (+2 treatment). Members of the control group were not seen by the experimenter for
any kind of treatment.
Through role playing of the three remaining stories
of the Kohlberg interview, experimental subjects
were exposed to the new moral concepts. After each
story was read the subject played the role of the
main character in the story, and as the main character he was to seek advice about the problem from
two friends. The experimenter played the parts of
the two friends. The subject first asked one "friend"
for "advice," with that friend's advice favoring one
side of the conflict, and then asked the second
friend, who favored the other side of the conflict.
The reasoning was always at the stage appropriate
to the subject's treatment condition. All the arguments used in the role playing were constructed by
closely following the coding forms and thus are
based on specific coded responses.
Illustrative examples of the treatment-condition
arguments are based on the Kohlberg situation in
which the husband's conflict is between stealing a
drug or letting his wife die. The following two arguments, containing Stage 3 reasoning, represent what
a Stage 2 subject in the +1 treatment was exposed
to in this situation:
(a) You really shouldn't steal the drug. There
must be some better way of getting it. You could
get help from someone. Or else you could talk the
druggist into letting you pay later. The druggist
is trying to support his family; so he should get
some profit from his business. Maybe the druggist
should sell it for less, but still you shouldn't just
steal it.

613

(6) You should steal the drug in this case.

Stealing isn't good, but you can't be blamed for
doing it. You love your wife and are trying to
save her life. Nobody would blame you for doing
it. The person who should really be blamed is the
druggist who was just being mean and greedy.
The experimenter, while administering the treatment, did not know the subject's stage, since he had
not scored the pretest, and did not know the experimental group of the subject; hence administration
of the treatments was blind. The only exceptions
were subjects exposed to "Stage 1" concepts who
must have been in Stage 2, and those exposed to
"Stage 6," who must have been Stage 4 subjects.
The possibility of the experimenter recalling the subjects' stages since he previously interviewed them is
unlikely because there were many lengthy interviews,
and because a subject's stage is determined using
the scoring guides.
Pastiest interview. The posttest consisted of the
six pretest situations plus the three situations of the
experimental treatments; it was administered to the
experimental subjects 1 week after the treatment,
and to the control subjects 3 weeks after the pretest. Subjects were called to the experimental room
individually, where they were told they would be
asked questions regarding stories similar to the ones
they had previously heard. (The repetition of some
of the stories and questions did not seem to affect
the subjects' willingness to respond. They generally
responded with the same interest and concentration
as in the pretest.)

Reliability and Scoring of Protocols

The results reported in this paper are based on
the scores obtained through the detailed coding. The
interviews were coded by the experimenter more
than a year after their administration. The scorer
had no knowledge of the identity of the protocol
he was coding, nor of its experimental condition.
All the pretests were scored separately from the
posttests. The coding was carried out on a situationby-situation basis rather than on a subject-bysubject basis; after all subjects' responses to the
first situation were coded, all subjects' responses to
the next situation were coded, and so on.
One estimate of the reliability of Kohlberg's detailed coding system is based on the independent
coding by two judges of responses obtained from
17 subjects not used in this experiment. Scores for
each subject consisting of the percentage of the
statements falling into each of the six stages were
calculated. A weighted score per subject was then
obtained by multiplying the number of points at
each stage by the number of the stage, summing
these products, and dividing the sum by the total
number of points. The product-moment correlation
between the scores of the two judges was .94.
A measure of interjudge agreement on the scoring
of the subjects in this experiment was obtained
from the correlation between the scores of the
author, who used the detailed coding system, and

ELLIOT TURIEL

614

those of another scorer who used the global rating

system. Under both scoring systems a subject receives a number of points on each stage, which can
be converted into a single score by the procedure
described above. A product-moment correlation of
.78 was found for the original 48 subjects. Since the
two scoring systems differ slightly, this correlation
is a conservative estimate of the interjudge reliability
of the detailed coding system.

RESULTS
The analysis of the posttest interview,
which included all nine moral judgment
situations, was divided into the following two
parts:
1. Stage scores were obtained from the
posttest responses to the three situations used
in the treatments and not in the pretest. Since
the experimental subjects were directly influenced on those three situations, these scores,
which will be referred to as "direct scores,"
represent the amount of direct influence of
the treatment.
2. Posttest stage scores for the six situations used in the pretest represent the amount
of indirect influence, or the tendency to generalize the treatment influence to situations
differing from those on which subjects were
directly influenced. The measure reflecting indirect influence is the difference between a
subject's pretest and posttest scores on each
stage. These change scores will be referred to
as "indirect scores."
TABLE 1
MEAN DIRECT POSTTEST STAGE SCORES (!N PROPORTIONS) ON THE STAGES ONE BELOW (1), THE
SAME As (0), ONE ABOVE (+1), AND
Two ABOVE (+2) THE PRETEST
DOMINANT STAGE
Stage level
relative to
pretest
dominant
stage"
-1

+1
+2

Condition groupsb
i
treatment

+2
+1
treatment treatment

.336n
.283

.183,,
.346

.13121

.26622

.057

.102

.20913
.374
.14S2,
.099

Control
.240,4
.395

.12224
.085

Note.Dunnett < tests were computed for each boldface

figure against each of the other three figures in the same row.
Tests significant at the .05 level, Group 11 > 13; at the .025
level,
11 > 12; at the .005 level, 22 > 21, 22 > 23, 22 > 24.
n
Each subject had received pretest scores at each developmental stage, the highest of these indicating his dominant
stage. On the posttest, for each individual the proportion of his
total
score was calculated for each level listed in the left column.
b
N 11 in each group.

Direct Scores
The analysis of the direct scores involved
the percentage of usage for each subject of
the stage that is: one below the initial dominant stage ( 1 scores), at the same stage
as the initial dominant stage (0 scores), one
above the initial dominant stage ( + 1 scores),
and two above the initial dominant stage
( + 2 scores).3
The hypothesis was that an individual accepts concepts one stage above his own dominant position more readily than he accepts
those two stages above, or those one stage
below. Two specific hypotheses result from
this general hypothesis that the +1 treatment
would be the most effective: (a) that the +1
treatment causes more movement to +1 than
the +2 treatment causes movement to +2
or the 1 treatment to 1, and (b) that the
+ 1 treatment causes more 4-1 movement
than does any other treatment.
Test 0} Hypothesis a. Table 1 presents
(in boldface type), for each experimental
group, the mean amount of usage of concepts
at the same stage as that of the treatment
condition. Table 1 also presents the control
group mean scores on the stages that are one
below ( 1 scores), one above ( + 1 scores),
and two above ( + 2 scores) their dominant
stage.
The experimental groups' scores may not
reflect solely the influence of the experimental
manipulations. To determine how much of
these scores reflects factors other than the
treatments, it is necessary to correct for the
change that would have occurred independently of the experimental manipulations. The
best estimate of this change is provided by
the control group, which had no treatment.
It may be assumed that the scores of the
control group are due to statistical regression
and other artifactual sources.4
8
The other scores, such as those of the stage
two below or three above the dominant stage, are
not reported because they did not show significant
differences between the groups and do not add to
the understanding of the problem.
4
It may be a function of skewness that the 1
score of the control group was considerably larger
than the +1 or +2 scores. Of a subject's series of
scores one stage has the largest score while its
adjacent stages have the next largest scores, with
the more distant stages to the dominant stage having

STAGES IN MORAL JUDGMENTS

The experimental groups' scores were corrected by subtracting from those scores the
corresponding control group scores. This subtraction was done in the following manner:
The 1 mean of the control group was subtracted from the 1 mean of the 1 treatment group; the +1 mean of the control
group was subtracted from the +1 mean of
the +1 treatment group; the +2 mean of
the control group was subtracted from the
+2 mean of the +2 treatment group. The
three corrected means ( 1 = .096, +1 =
.144, + 2 = .014) obtained in this way are
presumably free of artifacts and thus represent the amount of influence of the experimental treatments.
The corrected means show that, as hypothesized, the direct influence of the +1
treatment was greater than that of the other
two treatments. The corrected mean of the
+ 1 treatment group was shown to be significantly greater than the corrected mean of the
+2 treatment group by a one- tailed t test
(t = 3.55, p < .DOS).5 The one-tailed t test
of the difference between the corrected means
smaller scores. The subjects of this experiment tended
to use the stages below the dominant stage more
than those above, resulting in a positively skewed
distribution on the six situations of the pretest.
When the other three situations are included, more
usage of the stage directly below the dominant stage,
resulting in less skewness, would be expected.
The control group and the experimental groups
were originally very similar. There were no significant differences between the combined scores of
the experimental groups and the scores of the control
group, with the t values ranging from .10 to .65.
We also compared each experimental group with the
control group and found no significant differences.
'Having subtracted the appropriate control score
from the experimental condition score we then computed a t test for the difference between the corrected
means. The standard error for this / test is complicated
by the fact that we subtracted correlated groups from
independent groups. However, the appropriate
standard error may be shown to be :
where :
si2 = the MSw for the +1 scores multiplied by 2/n
s = the MSW for the +2 scores multiplied by 2/n
n = the number of subjects in each group
TCICJ = the correlation between the +1 and +2 scores
of the control group.
(We are indebted to Robert Abelson and Merrill
Carlsmith for the derivation of this expression.)

615

of the +1 treatment group and the 1 treatment group reached a borderline level of
significance (t = 1.43, p < .10).
The corrected mean of the 1 treatment
group was significantly greater than the corrected mean of the + 2 treatment group
(* = 2.03, p< .05).
Test of Hypothesis b. We have demonstrated that the amount of usage of the treatment condition stage was greater in the +1
treatment group than in the other two experimental groups. While this result is necessary
to demonstrate the greater influence of the
+1 treatment, the +1 scores of the +1 treatment group must also be compared with the
+ 1 scores of the other groups.
Table 1 contains the +1 scores of each of
the four groups. The differences between the
+1 score of the +1 treatment group and the
+ 1 scores of the other groups were tested
using Dunnett's t statistic, which is appropriate in simultaneously testing one group mean
against each of several others (Winer, 1962).
These t tests indicated that the +1 treatment
was the most effective condition in moving
subjects up one stage, since the +1 score of
the +1 treatment group was significantly
larger than the +1 scores of any other group
(Table 1).
Other findings. Table 1 also presents the
1, 0, and +2 scores for the four groups.
The 1 score of the 1 treatment group was
larger than the 1 scores of the other groups.
However, the Dunnett t test indicates that
the difference between the 1 score of the
1 treatment group and the 1 score of the
control group did not reach significance
(t= 1.66). The differences between the -1
score of the 1 treatment group and the
1 scores of the +1 and the +2 treatment
groups were both significant (Table 1).
Using Dunnett t tests, comparisons of the
+2 score of the +2 treatment group with the
+2 scores of the control group (t < 1), of
the 1 treatment group (t = 1.16), and the
+ 1 treatment group (t < 1), indicated that
the +2 treatment did not show a significant
effect.
Congruent with the hypothesis, the control group and the +2 treatment group
showed the greatest usage of the dominant

ELLIOT TURIEL

616

stage (0 scores). An analysis of variance comparing the control and the + 2 treatment
groups on the one hand, with the 1 and
+ 1 treatment groups on the other hand,
showed a significant difference (F = 4.72,
dj- 1/32, p< .OS).
Conclusions regarding the direct scores, (a)
The +1 treatment had a direct effect, an
effect greater than that of either the 1 or
+ 2 treatment, (b) Although not reaching
an acceptable significance level, there was
some suggestion that the 1 treatment had
an effect in moving subjects down one stage.
(c) The +2 treatment did not show a significantly greater effect than the control
condition or the other experimental treatments in moving subjects up two stages.

(-.002). The one-tailed t test of the difference between the corrected means of the +1
treatment group (.052) and the 1 treatment
group (.012) reached a borderline level of
significance (t = 1.46, p < .10). Although
the +1 score of the +1 treatment group was
larger than the +1 scores of the other groups,
none of these differences was significant. No
other relevant differences were significant.
DISCUSSION

The analysis of the direct scores showed

that the +1 treatment was the most effective
of the three treatments, with the +2 treatment being the least effective. The similarity
between the patterns of the indirect and the
direct scores suggests that the differential
effect of the treatments represented something
Indirect Scores
more than memorization of the specific verThe analysis of the indirect scores was balizations used in the treatments and that
similar to that of the direct scores. The in- some change occurred in generalized moral
direct score is not a rating of responses on concepts. This conclusion remains tentative
the posttest, but rather a measure of change since the results on the indirect scores were
from pretest to posttest. For each subject's minimally significant and since the same
stage scores we subtracted the pretest from interview form was used in the test-retest
the posttest scores and obtained change procedure.
The findings support Kohlberg's schema of
scores.
stages
as representing a developmental conAs indicated by Table 2, the pattern of
results of the indirect scores was consistent tinuum, in which each individual passes
with the hypotheses and with the results on through the stages in the prescribed sequence.
the direct scores. The evidence is only sug- If the stages do form a developmental
gestive since significant findings were mini- sequence, then it should be easier for subjects
mal. A one-tailed t test (t = 2.70, p < .025) to understand and utilize concepts that are
showed that the corrected mean of the +1 directly above their dominant stage than
treatment group (.052) was significantly concepts that are two stages above.
The developmental interpretation is also
larger than that of the +2 treatment group
strengthened by the finding that subjects
assimilated the next higher stage more readily
TABLE 2
MEAN INDIRECT POSTTEST STAGE SCORES (IN PROPOR- than the lower stage, even though they could
TIONS) ON THE STAGES ONE BELOW ( 1), THE
understand the concepts of the lower stage
SAME As (0), ONE ABOVE (+1), AND
as well as, if not better than, those of the
Two ABOVE (+2) THE PRETEST
higher stage. Hence, we have an indication
DOMINANT STAGE
that the attainment of a stage of thought
Condition groups'1
involves a reorganization of the preceding
Stage level
relative to
modes of thought, with an integration of each
pretest
dominant
-1
+2
+l
Control
previous stage with, rather than an addition
stage
treatment treatment treatment
to, new elements of the later stages.
-1
0
+1

+2
a

+ .057
-.045
-.004
-.016

N = 11 in each group.

+ .001
-.043
+ .045
-.002

+ .009
-.022
+ .016
+.008

+ .045

-.061
-.007
+.010

Causal Factors of Changes in Stage

The subjects exposed to the stage one
above their dominant stage did learn to use

STAGES IN MORAL JUDGMENTS

some new modes of thought. A factor causing
the use of new modes of thought may be
cognitive conflict. Indeed, Smedslund's work
with the concept of conservation (Smedslund,
1961a, 1961c, 1961d) indicates that cognitive
conflict may lead to reorganization of structure. The concept of cognitive conflict is
similar to the concept of disequilibrium,
which Piaget and Inhelder have presented
rather obscurely (Inhelder & Piaget, 1958;
Piaget, 19SO). They seem to be saying that
movement from one structure to the next
occurs when the system, by being challenged,
is put into a state of disequilibrium. Thus
change in structure would involve the establishment of a new equilibrium after the
occurrence of disequilibrium.
Such a viewpoint is relevant to our study.
Since subjects were exposed to new modes of
thought through arguments justifying both
sides of a moral conflict, they did not really
receive solutions to the problems. Such a
situation, which exposed subjects to cogent
reasons justifying two contradictory positions,
could have resulted in cognitive conflict
arising from an active concern with both sides
of the issue. When the arguments were too
"simple," as in the 1 treatment, the subjects may not have become actively involved.
When the arguments were too "complicated,"
as in the +2 treatment, the subjects may
not have understood them. However, exposure
to concepts one stage above, concepts within
a subject's grasp, allowed him contact with
new contradictory ideas requiring thought.
Perhaps coping with concepts that had some
meaning to the subjects led to new modes
of thinking, or to a greater use of the stage
that was one above the initial stage.
Related Studies
A study having a direct relation to the
present research, by Bandura and McDonald
(1963), attempts to demonstrate that Piaget's
(1948) sequence of moral development
changes is a function of reinforcement contingencies and imitative learning. The study
assumed that Piaget's stages of moral development could be defined as a stage of
"objective responsibility" (moral judgment in
terms of the material damage or conse-

617

quences), followed by a stage of "subjective

responsibility" (judgment in terms of intention). Following one of Piaget's procedures,
Bandura and McDonald assigned children to
stages in terms of responses to paired storied
acts, one being a well-intentioned act resulting in considerable material damage, and the
other a maliciously motivated act resulting in
very little material damage.
Their experimental treatments attempted
to influence the subjects by reinforcing adult
models who expressed judgments in opposition
to the child's orientation, and by reinforcing
any of the child's own responses that run
counter to his dominant mode. Two measures
of learning of the opposite orientation were
obtained: the amount of learning during experimental treatment, and a posttest response
to new stories immediately following the
treatment. They showed that children could
be influenced to judge on the orientation opposite to their initial one. Bandura and McDonald viewed this evidence as "throwing
considerable doubt on the validity of a developmental stage theory of morality."
An adequate test of a stage theory of
morality must deal with stages that are truly
representative of mental structure rather than
with specific verbal responses. Empirical tests
of Piaget's moral judgment theory indicate
that the stages do not meet the necessary
criteria (Kohlberg, 1963b). However, the
Bandura and McDonald study does not provide an adequate test of Piaget's theory because his two stages are not those of objective
and subjective responsibility, but rather are
those of heteronomous and autonomous orientations. The heteronomous and autonomous
stages are each represented by 11 observable
aspects (Kohlberg, 1963b) of children's definitions of right and wrong, of which the
dimension of objective-subjective responsibility is only one. By studying only one
dimension as manifested in children's choices
between two alternatives Bandura and McDonald dealt with isolated surface responses,
and not with the concept of stage or mental
structure. In their experimental treatment
one of two possible answers was reinforced
Therefore, the induced changes did not represent underlying structures, but instead repre-

618

ELLIOT TURIEL

sented switches to what the subjects thought

were the correct answers.6
Another important deficiency in their procedure was the administration of the posttest
immediately after the experimental treatment.
As Smedslund (1961b) has demonstrated, the
test of duration over time is a main criterion
for distinguishing between cognitive structure
and superficially learned responses. There was
a small decrease in subjective responses given
by objective children from the experimental
treatment to the posttest, while there was no
such decrease in the objective responses of
subjective children. This finding, that downward movement was more stable than upward
movement, is in contrast to our findings, in
which upward movement was more stable. It
is not surprising that the learning of surface
verbal responses related to a lower stage can
be retained for a short time. It is interesting
that the learning of responses related to a
higher stage was not entirely retained, even
for such a short period of time.
In the present research we have worked
with responses assumed to reflect mental
structure and have found that the concept
of developmental stage or mental structure
has much relevance to the understanding of
children's moral thinking. We suggest that
the effectiveness of environmental influences
depends on the relation between the type of
concept encountered and developmental level.
6

It must be pointed out that in the contrasting

pairs of stories a well-intentioned act always resulted
in much material damage, while the maliciously
motivated act always resulted in little material
damage. A child in the objective stage could easily
have learned to more frequently designate, as being
worse, that story which contained less material
damage, thinking that it was the expected answer.
Thus he could have given the "higher stage" answer
without having learned the concept of intention.

REFERENCES
BANDURA, A., & MCDONALD, F. J. Influence of social
reinforcement and the behavior of models in
shaping children's moral judgments. Journal of
Abormal and Social Psychology, 1963, 67, 274-281.
INHELDER, B., & PIAOET, J. The growth of logical
thinking from childhood to adolescence. New
York: Basic Books, 1958.
KOHLBERG, L. The development of modes of moral
thinking in the years ten to sixteen. Unpublished
doctoral dissertation, University of Chicago, 1958.
KOHLBERG, L. The development of children's orientations toward a moral order: I. Sequence in the
development of moral thought. Vita Humana,
1963, 6, 11-33. (a)
KOHLBERG, L. Moral development and identification.
In H. W. Stevenson (Ed.), Yearbook of the National Society for the Study of Education: Pt. I.
Child psychology. Chicago: University of Chicago
Press, 1963. Pp. 277-332. (b)
PIAGET, J. The moral judgment of the child. (Orig.
publ. 1932) Glencoe, 111.: Free Press, 1948.
PIAGET, J. The psychology of intelligence. (Orig.
publ. 1947) New York: Harcourt, Brace, 1950.
SMEDSLUND, J. The acquisition of conservation of
substance and weight in children: II. External
reinforcement of conservation of weight and of the
operations of addition and subtraction. Scandinavian Journal of Psychology, 1961, 2, 71-84. (a)
SMEDSLUND, J. The acquisition of conservation of
substance and weight in children: III. Extinction
of conservation of weight acquired "normally"
and by means of empirical controls on a balance.
Scandinavian Journal of Psychology, 1961, 2, 8587. (b)
SMEDSLUND, J. The acquisition of conservation of
substance and weight in children: V. Practice in
conflict situations without external reinforcement.
Scandinavian Journal of Psychology, 1961, 2, 156160. (c)
SMEDSLUND, J. The acquisition of conservation of
substance and weight in children: VI. Practice on
continuous vs. discontinuous material in problem
situations without external reinforcement. Scandinavian Journal of Psychology, 1961, 2, 203-210.
(d)
WINER, B. J. Statistical principles in experimental
design. New York: McGraw-Hill, 1962.
(Received June 1, 1965)