Professional Documents
Culture Documents
Generalized Expectancies I&E PDF
Generalized Expectancies I&E PDF
609, 1966
Psychological M o n o g r a p h s : General and A p p l i e d
G E N E R A L I Z E D E X P E C T A N C I E S FOR I N T E R N A L VERSUS
E X T E R N A L C O N T R O L OF R E I N F O R C E M E N T
JULIAN B. ROTTER
University of Connecticut1
The effects of reward or reinforcement on preceding behavior depend in
part on whether the person perceives the reward as contingent on his own
behavior or independent of it. Acquisition and performance differ in situa-
tions perceived as determined by skill versus chance. Persons may also differ
in generalized expectancies for internal versus external control of reinforce-
ment. This report summarizes several experiments which define group dif-
ferences in behavior when Ss perceive reinforcement as contingent on
their behavior versus chance or experimenter control. The report also de-
scribes the development of tests of individual differences in a generalized
belief in internal-external control and provides reliability, discriminant va-
lidity and normative data for 1 test, along with a description of the results
of several studies of construct validity.
nal versus external control of reinforce- act to determine choice behavior along
ment. with the value of potential I'einforcements.
These generalized expectancies will result
77; eo re t tea I 13 a ck gro un d in characteristic differences in behavior in
Social learning theory (Rotter; 1954. a situation culturally categorized as chance
1955, 1960) provides the general theoreti- determined versus skill determined, and
cal background for this conception of they may act to produce individual differ-
the nature and effects of reinforcement. In ences within a specific condition.
social learning theory, a reinforcement Specific expectancies regarding the causal
acts to strengthen an expectancy that a nature of behavior-outcome sequences in
particular behavior or event will be fol- different situations would also affect be-
lowed by that reinforcement in the future. havior choice. From social learning theory
Once an expectancy for such a behavior- one would anticipate that the more clearly
reinforcement sequence is built up the and uniformly a situation is labeled as skill
failure of the reinforcement to occur will or luck determined, in a given culture, the
reduce or extinguish the expectancy. As lesser the role such a generalized expectancy
an infant develops and acquires more ex- would play in determining individual differ-
perience he differentiates events which are ences in behavior.
causally related to preceding events and
those which arc not. It follows as a general Related Conceptions
hypothesis that when the reinforcement In learning theory it has been recognized
is seen as not contingent upon the subject's that differences in subject behavior are re-
own behavior that its occurrence will not lated to task differences along a dimension
increase an expectancy as much as when of skill and chance. Goodnow and Post-
it is seen as contingent. Conversely, its man (1955) and Goodnow and Pettigrew
nonoccurrence will not reduce an expect- 11955.), for example, present data to
ancy so much as when it is seen as con- show that probabilistic learning theory is
tingent. I t seems likely that, depending not applicable where the subject feels that
upon the individual's history of reinforce- the occurrence of the reinforcement is law-
ment, individuals would differ in the de- ful. Wyckoff and Sidowsky (1955) simi-
gree to which they attributed reinforce- larly felt that their subjects' behavior
ments to their own actions. changed when they no longer felt that the
Expectancies generalize from a specific task was a '''guessing" problem. Cohen
situation to a series of situations which are (1960) has extensively studied differences
perceived as related or similar. Conse- in subjects' behavior or strategy in choice
quently, a generalized expectancy for a and skill games noting the tendency for
class of related events has functional prop- the ''gambler's fallacy" to appear in
erties and makes up one of the important chance games .. . an effect opposite to the
classes of variables in personality de- usual effect of reinforcement. A somewhat
scription. Harlow's (1949) concept of high- different approach to chance and skill
level learning skills seems similar to this task differences is assumed by Feather
notion that individuals differ in learned (1959) who felt that motivation was less-
generalized expectancies involving rela- ened in chance tasks as compared to skill
tionships between a wide variety of be- tasks. In general, however, a theoretically
haviors and their possible outcomes. A based, systematic study of chance and
generalized attitude, belief, or expectancy skill differences in acquisition and per-
regarding the nature of the causal relation- formance has not been made prior to the
ship between one's own behavior and its series of studies to be reported here.
consequences might affect a variety of The literature of personality theory does
behavioral choices in a broad band of life contain discussions of a number of varia-
situations. Such generalized expectancies bles which may have some relationship
in combination with specific expectancies to the one of major concern in this paper.
Internal versus Exterx.il Control of Reinforcement
The significance of the belief in fate, sociologists (Nettler, 1957; Srole. 1956)
chance, or luck has been discussed by vari- have developed a crude individual meas-
ous social scientists over a long period ure of alienation.
of time. Most of their concern, however, In psychology, White (1959) in discuss-
has been with differences among groups ing an alternative to drive reduction has
or societies rather than individuals. Typ- noted how the work of many authors has
ical of an early discussion of this kind converged on a belief that it is chararcter-
is that of Veblcn (1899), who felt that a istic of all species to explore and to at-
belief in luck or chance represented a tempt to master the environment. He has
barbarian approach to life and was gen- labeled this concept competence. While
erally characteristic of an inefficient so- White was not specifically interested in
ciet}'. Although Veblen was not concerned individual differences he has noted that
with individual differences, his discussion such a motive or drive is not explained
implied that a belief in chance or luck as by primary drive and although perhaps not
a solution to one's problems was charac- as strong as sonic primary drives it is mod-
terized by less productivity and, conse- crate in strength and persistence. Angyal
quently, bears some parallel to the hy- (1941) has also noted the significance of
pothesis that a belief in external control the organism's motivation towards auton-
of reinforcements is related to a general omy, or the active mastery of the environ-
passivity. Veblen also stated, "In its simple ment.
form the belief in luck is this instinctive There arc a number of other psychologi-
sense of an unscrutable, teleological pro- cal variables which appear to bear some
pensity in objects or situations." In relationship to the concept under investi-
other words, Veblen states that the belief gation. Some of these are undoubtedly
in luck is related to or similar to a general related, but, for others, it is possible that
belief in fate. the relationship is more apparent than
More recent!}', Merton (1946) has dis- real.
cussed the belief in luck more or less as a Perhaps one of the major conceptions
defense behavior, as an attempt "to serve which bears some relationship to the be-
the psychological function of enabling peo- lief in internal versus external control of
ple to preserve their self esteem in the face reinforcements is that of need for achieve-
of failure." He states it "may also in some ment. The work of McClelland, Atkinson,
individuals act to curtail sustained en- Clark, and Lowell (1953) and of Atkinson
deavor," or, in other words, he too suggests (1958) and their colleagues working pri-
a relationship between passivity and the marily with adults, and Crandall (1963)
belief in chance or luck. with children, suggests that people who are
The concept of alienation which has high on the need for achievement, in all
played an important role in sociological probability, have some belief in their own
theory for many years does seem related at ability or skill to determine the outcome
a group level to the variable of internal- of their efforts. The relationship is prob-
external control. The alienated individual ably not linear, however, since a person
feels unable to control his own destiny. high on motivation for achievement might
He is a small cog in a big machine and at not be equally high on a belief in internal
the mercy of forces too strong or too vague control of reinforcement, and there may
to control. Marx, Weber, and Durkheim be many with a low need for achievement
placed great importance on this concept, who still believe that their own behavior
and more recently Merton (1949) has determines the kinds of reinforcements
stressed its importance in the study of they obtain.
asocial behavior. Seeman (1959) has Another variable which may bear some
linked the concept of alienation as it refers genuine relationship to the variable of in-
to •poicerlessness, to internal-external con- ternal versus external control of reinforce-
trol as a psychological variable. Some ment is the concept of "field determined"
Julian B. Rotter
versus "body oriented." The work of Wit- ever, not with this variable at all but only
kin, Lewis, Hertzman, Machover, Meis- with the question of whether or not an in-
sncr, and Wapner (1954) suggests that dividual believes that his own behavior,
people can be ordered on a continuum, in skills, or internal dispositions determine
.some perception experiments, describing what reinforcements he receives. While the
whether they derive most of their cues conformist (the opportunist, in particular)
from the field or from internal sources. A who is actively trying to learn and adjust to
study by Linton (1955) suggests that people the rules of the society he lives in is at
who are "field oriented" or "field depend- one end of Riesman's continuum, he is
ent" tend to be more conforming. However, likely to be in the middle of the continuum
unpublished data of the author indicate with which we are concerned.
no relationship between an individual meas- Finally, a word should be said about the
ure of internal-external control and the general concept of causality. This psycho-
Gottschalk Figures Test, a measure fre- logical dimension is one which has been
quently used as an operation for "field neglected for some time, although it is one
determined." of the strong interests of Piaget (1930),
Perhaps less clear is the relationship of who studied how the notion of causality
internal versus external control of rein- developed in children. Pepitone (1958) has
forcement to the notion of "ego control." recently discussed several aspects of the
Although the concept of ego control is not attribution of causality in social interac-
always defined similarly, it seems to con- tions. However, individual differences in
tain the ideas of confidence and ability to how causality is assumed to relate events
deal with reality. While it seems likely has not been a subject of investigation.
that the individuals at both extremes of It would seem that some relationship
the internal versus external control of would exist between how the individual
reinforcement dimension are essentially views the world from the point of view of
unrealistic, it is not as likely that the peo- internal versus external control of rein-
ple toward the middle of the distribution forcement and his other modes of per-
are less confident. We do have indications, ception of causal relationships.
however, that the people at either extreme
of the reinforcement dimension are likely Studies of Complex Learning
to be maladjusted by most definitions, and, The notion that individuals build up
to the extent that ego control is another generalized expectancies for internal-ex-
type of definition of maladjustment, it ternal control appears to have clear im-
would bear some curvilinear relationship plications for problems of acquisition and
to the variable we are concerned with here. performance. If a human can deal with
Similar to the conception of Witkin ct future events with the use of verbal sym-
al. is that of Riesman (1954), who has bols and can perceive an event as following
attempted to describe an apparently com- a preceding behavior of his own, then the
parable distinction. Riesman's concep- strength of that connection will depend at
tion is based on the degree to which people least in part on whether or not he feels
are controlled by internal goals, desires, there is a causal or invariable relationship
etc. versus the degree to which the}r are between his behavior and the event. Once
controlled by external forces, in particular a person has established a concept of
social forces or conformity forces. Al- randomness or chance the effects of rein-
though this variable may bear some rela- forcement will vary depending upon what
tionship to the one under investigation, it relationship he assigns to the behavior-
should be made clear that the apparent reinforcement sequence.
relationship is not as logical as it appears. A person who is looking for an unusual
Riesman has been concerned with whether brand of tobacco and is finally able to find
the individual is controlled from within or it will return to the same place where he
from without. We are concerned, how- was reinforced before when he needs to-
Internal versus External Control of Reinforcement
bacco again. However, an individual who menter without the subject's knowledge of
needs money and finds a five dollar bill in such control. Two strategies are used. The
the street is not likely to return to that first is to provide a relatively ambiguous
spot to look for a five dollar bill when he task under two conditions, one in which
needs money. A behavior of looking on the the subject is instructed that it is skill de-
ground may be strengthened to some de- termined. Obviously in these studies we
gree in the latter case. However, the in- are dealing with a continuum in which in
dividual is selective in what aspects of his one situation the task is likely to be per-
behavior are repeated or strengthened ceived as relatively more skill determined.
and what aspects are not, depending upon The second strategy is to present different
his own perception of the nature or caus- tasks which are also surreptitiously con-
ality of the relationship between the rein- trolled by the experimenter and which are
forcement and the preceding behavior. defined as skill and chance essentially
In its simplest form, our basic hypothesis through previous cultural experience. For
is that if a person perceives a reinforce- example, although they have certain prob-
ment as contingent upon his own behavior, lems of comparability, dice throwing is
then the occurrence of either a positive or generally recognized as a chance task,
negative reinforcement will strengthen or while solving arithmetic problems and
weaken potential for that behavior to re- throwing darts are generally recognized
cur in the same or similar situation. If as skill tasks.
he sees the reinforcement as being outside The first of these studies was under-
his own control or not contingent, that is taken by Phares (1957). Phares used
depending upon chance, fate, powerful color matching as an ambiguous task and
others, or unpredictable, then the preced- instructed half of the subjects that the
ing behavior is less likely to be strength- task was so difficult as to be a matter of
ened or weakened. Not only will there be a luck and the other half of his subjects that
difference of degree but also a difference, success was a matter of skill and that pre-
in some instances, in the nature of the vious research had shown that some people
function as the result of a series of trials. were very good at the task. The subjects
I t is evident that if this analysis is correct matched samples to finely graded stand-
then different kinds of learning paradigms ards. He used a second task of matching
or situations are going to produce different lines of slightly varying lengths to stand-
kinds of learning functions. A learning situ- ards placed on cards at different angles.
ation such as that in which the experimenter For both tasks a fixed order of partial
arbitrarily determines the right response for reinforcement (right or wrong) was used
whether or not food is given, regardless of and the measure of expectancy was the
the behavior of the subject, will produce a number of chips a subject would bet on
different kind of learning than one where the his probability of being correct on the suc-
subject believes his behavior determines ceeding trial.
whether or not the reinforcement will oc- Phares found, as hypothesized, that the
cur. In other words, learning under skill increments and decrements following suc-
conditions is different from learning under cess and failure, respectively, were signifi-
chance conditions. cantly greater under skill instructions
To test this hypothesis a series of stud- than under chance instructions. Rein-
ies was undertaken comparing verbal ex- forcements under skill conditions had a
pectancies for future reinforcement under greater effect on raising or lowering ex-
conditions of chance and skill learning. pectancies for future reinforcements. He
In this group of studies it has been neces- also found that subjects shifted or changed
sary in order to compare skill and chance their expectancies more often under skill
learning tasks directly to provide a sim- conditions. (Another measure of the same
ilar sequence of reinforcement in both data described above.) Finally he showed
cases which was controlled by the experi- a strong trend toward unusual shifts in
Julian B. Rotter
expectancies, that is, up after failure or A further check on these studies was
down after success (the gambler's fallacy! made by Rotter, Liverant, and Crowne
under chance conditions. The significance (1961), who studied the growth and ex-
of this last finding was marginal (*p = tinction of expectancies in chance-con-
.07, two-tailed test). trolled and skilled tasks. This study in-
This study was followed by one by volved using two tasks, one the ESP task
James and Rotter (1958). In this study referred to above and, the second, a
the emphasis was on the extinction of motor task presumably involving steadi-
verbal expectancies. Under conditions of ness which would typically be perceived
partial and 100% reinforcement an extra- as a skill task. Again in both tasks simi-
sensory perception (ESP) type of task was lar serruences of reinforcement were used.
used with experimenter control, and the In this case instructions were identical, the
exact same sequence of 50% partial rein- difference in the cultural perception of the
forcement was given to two groups of sub- tasks being the experimental variable.
jects (two other groups had 100% rein- This study utilized eight groups, four
forcement) for 10 training trials. Two chance and four skill with 25%, 50%, 75%,
groups were told that guessing in the task and 100% reinforcement over eight train-
had been shown by scientists to be entirely ing trials followed by an extinction series.
a matter of luck, and two groups were told This study confirmed the previous findings
that there was evidence that some people of both Phares and James and Rotter.
arc considerably skilled at the task. While During the training trials, subjects (except
the groups did not differ significantly at for 100% reinforcement groups) showed
the end of the training trials, the chance greater increments or decrements following
and skill groups did differ significantly success and failure respectively under skill
in the number of trials to extinction. Ex- conditions than under chance conditions.
tinction was defined as stating an expect- Major differences in extinction were ob-
ancy of 1 or 0 on a scale of 10 for three tained independently of the expectancy
consecutive trials. levels at the end of the training trial. In
this study, extinction curves for the two
The interesting thing about the results groups crossed over completely, so that all
of this investigation was that the usual of the findings of the James and Rotter
findings of superiority of partial over studies were replicated, but, in addition,
100% reinforcement in trials to extinction 100% reinforcement took significantly
was true only of the group with chance longer to extinguish than 50% reinforce-
instructions, but under skill conditions the ment in the skill task. Differences between
mean number of trials to extinction for the groups were smaller at the 25% and
100% reinforcement was longer (22.9) than 75% reinforcement schedules than at 50%
under 50% reinforcement (19.8). Trials to or 100%. The latter findings were inter-
extinction under partial reinforcement preted as suggesting that at the 25% and
were significantly longer for chance than the 75% levels the chance task was being
for skill instructions, and trials to ex- rewarded or reinforced more often or less
tinction for 100% reinforcement were sig- than could be accounted for by chance
nificantly longer for skill than for chance alone. The frequency of reinforcement it-
instructions. The findings were interpreted self may tend to make the task appear
to indicate that under chance conditions more like a skill task. In the case of 100%
the extinction series was interpreted as a chance reinforcement, however, the abrupt
change in the situation, a disappearance of change from continuous positive rein-
previous lucky hits in the 100% reinforce- forcement to continuous negative rein-
ment condition but not in the 50% rein- forcement suggests a change in the nature
forcement conditions. For the subjects with of the situation.
skill instructions, the greater the previous
reinforcement the longer it took the sub- The question could arise as to whether
ject to accept the fact that he was not or not differences in extinction patterns
able to do the task successfully. would be the same with a behavioral cri-
IXTKRXAL VERSUS EXTERNAL COXTHOL OF REINFORCEMENT
terion other t h a n verbalized expectancies. one with skill instructions, were given 75%
To test this, Holclen and Rotter (1962) reinforcement for a sequence of eight
again used the ESP task instructing one trials and then were tested for generaliza-
group of subjects that it was a skill task tions of expectancies by having one trial
and the other group that it was deter- on the new second task. Two other groups
mined entirely by luck. Subjects were were given the same 75% reinforced eight
given two dollars in nickels and told they training trials followed by a series of ex-
could bet a nickel on each trial on whether tinction trials, then given a 5-minute rest
or not they would succeed until they and given two additional trials on the
wished to discontinue and keep the re- same task. These latter groups were
maining money or until they ran out of examined for "spontaneous recovery."
nickels. Three groups all given 50% par- James' findings again replicated the dif-
tial reinforcement were used, one with skill ferences between chance and skill groups
instructions, one with chance instructions, in the growth of acquisition of expectan-
and one with ambiguous instructions. Re- cies. He found, as hypothesized, signifi-
sults showed a clear difference, with the cantly greater generalization of expectan-
subjects given chance instructions and cies from one task to another under skill
those who were not told it was either a instructions than under chance instruc-
chance or a skill task having significantly tions. He also found more "spontaneous
more trials to extinction (almost twice as recovery" under skill instructions, but the
many) than the skill group, Extinction difference in this case only approached
was defined as voluntarily quitting the significance.
experiment.
Bennion's .study of the effect of variabil-
An unpublished dissertation by Ben- ity in scores on a task can be interpreted
nion (1961) using the same tasks as in as defining one of the conditions which
the Rotter, Livcrant, and Crowne study, make for the perception that the task is in
rather than instructions to produce the fact skill or chance determined. Other con-
skill and chance difference, examined a ditions affecting such a perception were
partial reinforcement sequence that was studied b}r Blackman (1962). Blackman
predominantly positive but in which re- used the well-replicated finding that under
ported scores differed in variability to two chance conditions extinction in a 50%
groups. Overall mean score and frequency reinforcement sequence is likely to be
of success and failure as denned by the considerably longer than under skill con-
experimenter were controlled. Bennion hy- ditions. In a counterbalanced design he
pothesized that greater variability of scores used numerous sequences of presumably
either under chance or skill conditions random appearing lights, controlling for
would produce results similar to that of the the percentage of reinforcement. The task
difference between the chance and skill was one of attempting to predict whether a
conditions. There would be greater respon- red or a green light would appear on the
siveness in changes in expectancy to success following trial. He varied the length of se-
and failure under the less variable condi- quences in which the same light would ap-
tions. He found support for this hypothesis pear consecutively, and he varied the de-
as well as replicating the difference in re- gree of patterning from presumably purely
sponsiveness under chance and skill condi- random through an easy pattern to a com-
tions obtained by Phares and by Rotter, plicated pattern. Extinction began when
Liverant, and Crowne in previous studies. the red light ceased to go on, and the meas-
ure of extinction was based upon the elim-
In another unpublished dissertation,
ination of subject predictions of red re-
James (1957) studied some of the same
sponses. He found, as he hypothesized,
variables and in addition the generaliza-
that the length of sequences significantly
tion of expectancies and the "spontaneous
affected the number of red responses in ex-
recovery" of expectancies. He used both
tinction and the expectancies associated
a line-matching and an angle-matching
with them. The longest sequences ex-
task. Two groups, one with chance and
Julian B. Rotter
tinguished more quickly. Similarly, the taken before and after the training. He
easy but not the complicated pattern, found, as hypothesized, that the recogni-
which was apparently not perceived, also tion thresholds dropped significantly
resulted in quicker extinction. These re- more in the skill-instructed than in the
sults are interpreted to indicate that chance-instructed groups although they
longer sequences and recognizable patterns had had the same number of shocks on
suggest to the subject that there is not a the same trials and for the same nonsense
random pattern but an experimenter-con- syllables. Phares concluded that subjects
trolled one. Consequently, when extinction who feci they have control of the situation
begins and the red light no longer appears are likely to exhibit perceptual behavior
the subject interprets the situation as one that will better enable them to cope with
in which the experimenter has changed the potentially threatening situations than
sequence of lights. If, however, the sub- subjects who feel chance or other noncon-
ject interprets the original sequence as trollable forces determine whether or not
random, he will persist much longer before their behavior will be successful.
extinguishing on anticipation that the red Investigations of differences in behav-
light will appear again. ior in skill and chance situations provide
Implications from the studies of Rotter, relatively clear-cut findings. When a sub-
Livcrant, and Crowne (1961), Bennion ject perceives the task as controlled by the
(1961J, and Blackmail (1962) can be experimenter, chance, or random condi-
summarized. Subjects are more likely to tions, past experience is relied upon less.
sec a sequence of reinforcement as not be- Consequently, it may be said that he learns
ing chance controlled when the percentage less, and under such conditions, he may
of reinforcement significantly deviates indeed learn the wrong things and develop
from a 50-50 percentage in a right-wrong a pattern of behavior which Skinner has
situation, when the sequence of reinforce- referred to as "superstitious." These stud-
ments appears to have a pattern, when ies strongly imply that the interpretation
unusually long sequences of one of two of investigations of acquisition and per-
alternative events occur, and when vari- formance must be made in light of the
ability of performance is minimal in a position on the continuum of complete
task allowing for scoring along a contin- chance control to complete skill control at
uuum. which the particular task falls. Differences
A somewhat different variable was in- in learning are not merely a matter of de-
vestigated by Phares (1962), who studied gree but also of nature or kind as indi-
perceptual thresholds for shock-associated cated by the dramatic reversal of extinc-
stimuli in chance-controlled versus skill tion curves as demonstrated by Rotter,
situations. Phares used a tachistoscopie Liverant, and Crowne (1961). Perhaps
exposure of nonsense syllables, some of more important are the implications for
the stimuli being accompanied by shock. the learning theory favored by psycholo-
The skill group was told that the shock gists in general. Such theory is often based
could be escaped by pressing the correct upon experimental paradigms which in-
button which could be learned. The chance volve experimenter control. That is, they
group was instructed that they could press use tasks where the experimenter decides
any of the sequence of buttons and this in a more or less arbitrary fashion when
may or may not avoid the shock depend- he will reinforce or where he will reinforce
ing upon chance. The skill group was run but not ones where the subject feels that
first and then the chance groups. In this his own performance determines primarily
way the experimenter could control the whether or not he will be successful at the
number of shocks, so that he was able to task. However, many, if not the majority
match the chance group with the skill of learning situations of humans in every-
group in the total number of shocks ob- day life situations, are in fact perceived as
tained during the 10 training trials. Recog- skill controlled. The direct application of
nition thresholds for the syllables were theories of learning based upon experi-
Internal versus External Control of Reinforcement
menter controlled tasks to such learning (1957) dissertation, previously referred to.
is in grave doubt. James revised Phares' test still using a
Although there is no direct proof that Likert format and wrote 26 items plus
''experimenter control" is equivalent to filler items based on the items which ap-
"chance control," it would seem logical peared to be most successful in the Phares
that the subject perceiving no regularity study. He similarly hypothesized that
or predictability to the reinforcement within each of his groups, regardless of
would regard it similarly. This conclusion chance or skill instructions, those individ-
is supported by the earlier mentioned uals who scored toward the external end of
Blackman (1962) experiment... a typical the continuum would behave in each group
"experimenter control" paradigm . . . where in the same way as the difference between
he obtained longer extinction times when the chance group and the skill group for
the training sequence of partially reinforced all subjects. James was able to find low but
red and green lights lacked discernible pat- significant correlations between his test
terns. I t is also supported by the fact that and behavior in the task situation. Exter-
tasks with chance instructions produce the nal subjects had smaller increments and
same kind of differences between 100% and decrements following success and failure,
50% partial reinforcement in extinction generalized less from one task to another,
rates as do the typical experimenter control and recovered less following the period of
experiments. extinction. They also tended to produce
more unusual shifts (up after failure and
Internal versus External Control down after success) in expectancy.
as a Personality Variable The James-Phares scale has been used
in some research involving correlates of
Development of Measures of Internal-Ex-
individual differences in a generalized
ternal Control
expectancy for internal-external control.
The first attempt to measure individual However, the late Shephard Liverant in as-
differences in a generalized expectancy or sociation with J. B. Rotter and M. Seeman
belief in external control as a psychologi- undertook to broaden the test; develop
cal variable was begun by Phares (1957) subscales for different areas such as
in his study of chance and skill effects on achievement, affection, and general social
expectancies for reinforcement. Phares de- and political attitudes; and control for
veloped a Likert-type scale with 13 items social desirability by the construction of a
stated as external attitudes and 13 as in- new forced-choice questionnaire. The earli-
ternal attitudes. The scale was developed est version of this scale included a hundred
on a priori grounds, and he found some forced-choice items, each one comparing
suggestive evidence with his first crude an external belief with an internal belief.
attempt at measuring individual differ- The scale was item analyzed and factor
ences that prediction of behavior within analyzed and reduced to a 60-item scale b}'
a task situation was possible. In particu- Liverant on the basis of internal con-
lar, lie found that the items stated in an sistency criteria.
external direction gave low predictions, Item analysis of the 60-item scale indi-
approaching statistical significance, that cated that the subscales were not generat-
individuals with external attitudes would ing separate predictions. Achievement
behave in a similar fashion as did all sub- items tended to correlate highly with social
jects when placed in a chance situation desirability, and some subscales corre-
versus a skill situation. That is, they lated with other scales at approximately
tended to show more unusual shifts, the same level as their internal con-
smaller magnitude of increments and decre- sistency. On this basis, items to measure
ments, and a lower frequency of shifts of more specific subareas of internal-external
expectancy in any case than did subjects control were abandoned.
who scored low on these 13 items.
Data were collected for a large group
Phares! work was followed by James' of subjects to provide item correlations
10 Julias B. Rotter
with the Marlowe-Crowne Social Desir- the subjects' expectations about how rein-
ability Scale (Crowne & Marlowe, 1964). forcement is controlled. Consequently, the
The overall correlation of the scale with test is considered to be a measure of a
the Social Desirability scale for different generalized expectancy. Such a generalized
samples ranged from .35 to .40 which was expectancy may correlate with the value
deemed to be too high. Reduction and the subject places on internal control but
purification of the 60-item scale was none of the items is directly addressed to
undertaken by S. Liverant, J. B. Rotter, the preference for internal or external con-
and D. Crowne. Validity data from two trol.
studies were used along with internal Test data on the I-E scale have been
consistency data. Item validity for most of obtained in a series of samples. Results
the items was available from a study of are summarized in Table 2. Where no
Seeman and Evans on tuberculosis pa- source is given, the data have been col-
tients who had evidenced greater self-effort lected by the author and are being reported
towards recovery versus those who were here for the first time.
more passive. Item validity for the pre- Internal consistency estimates are rela-
diction of individual differences in trials to tively stable as shown in Table 2. While
extinction in the previously cited study of these estimates are only moderately high
Rotter, Liverant, and Crowne (1961) was for a scale of this length, it should be
also available. In this final revision, word- remembered that the items arc not ar-
ing of some items was changed to make ranged in a difficulty hierarchy, but rather
the items appropriate for noncollege adults are samples of attitudes in a wide variety
and upper level high school students. of different situations. The test is an addi-
By eliminating those items which either tive one and items are not comparable.
had a high correlation with the Marlowe- Consequently, split-half or matched-half
Crowne Social Desirability Scale, a propor- reliability tends to underestimate the in-
tional split so that one of the two alterna- ternal consistency. Kuder-Richardson re-
tives was endorsed more than 85% of the liabilities arc also somewhat limited since
time, nonsignificant relationship with other this is a forced-choice scale in which an
items, or a correlation approaching zero attempt is made to balance alternatives
with both validation criteria, the scale was so that probabilities of endorsement of
reduced to 23 items. The final version of either alternative do not include the more
the scale, the one on which most of the extreme splits.
subsequent data to be reported are based, Test-retest reliability for a 1-month
is a 29-item, forced-choice test including period seems quite consistent in two quite
6 filler items intended to make somewhat different samples. The somewhat lower
more ambiguous the purpose of the test. reliabilities for a 2-month period may be
This measure will be referred to in the partly a function of the fact that the first
remainder of this article as the I-E scale. test was given under group conditions and
The I-E scale is presented in Table 1. the second test was individually ad-
Instructions for administration are pre- ministered. In the studies of test-retest re-
sented in Appendix A. Riserial item cor- liability, means for the second administra-
relations with total score ivith that item tion typically dropped about 1 point in
removed are given for 200 males, 200 fe- the direction of less externality.
males, and the combined group. I t can be Correlations of the 60-item I-E scale
seen that these are moderate but consistent. with the Marlowe-Crowne Social Desir-
The letter preceding the external choice in ability Scale were obtained in a number of
every item is italicized. The score is the college student samples and typically
total number of external choices. ranged between —.35 and —.40. The at-
A careful reading of the items will make tempt to reduce this correlation in the new
it clear that the items deal exclusively scale was moderately successful. The cor-
with the subjects' belief about the nature of relations for the new scale range from
the world. That is, thev are concerned with — .07 to —.35. The greater range may re-
INTERNAL VERSUS EXTERNAL CONTROL OP REINFORCEMENT 11
TABLE 1
The I-E Scale w i t h C o r r e l a t i o n s o f Each Item w i t h T o t a l Score, E x c l u d i n g
t h a t Item
Riscrial item correlations
200 F 400 M + F
La. Children get into trouble because their parents punish
them too much. (Filler)
b. The trouble with most children nowadays is that their
parents are too easy with them.
2.a. Many of the unhappy things in people's lives are
partly due to bad luck. .265 .250 .260
b. People's misfortunes result from the mistakes they
make.
3.a. One of the major reasons why we have wars is because
people don't take enough interest in politics.
b. There will always be wars, no matter how hard people
try to prevent them. .214 .147 .182
4.a. In the long run people get the respect they deserve in
this world.
b. Unfortunate^, an individual's worth often passes un-
recognized no matter how hard he tries. .238 .344 .289
5.a. The idea that teachers are unfair to students is non-
sense.
b. Most students don't realize the extent to which their
grades are influenced b}'accidental happenings. .230 .131 .179
6.a. Without the right breaks one cannot be an effective
leader. .345 .299 .319
Capable people who fail to become leaders have not
taken advantage of their opportunities.
7.a. No matter how hard you try some people just don't
like you. ' .200 .262 .229
b. People who can't get others to like them don't under-
stand how to get along with others.
8.a. Heredity plays the major role in determining one's
personality. (Filler)
b. I t is one's experiences in life which determine what
they're like.
9.a. I have often found that what is going to happen will
happen. .152 .172 .164
b. Trusting to fate has never turned out as well for me as
making a decision to take a definite course of action.
lO.a. In the case of the well prepared student there is rarely
if ever such a thing as an unfair test.
b. Many times exam questions tend to be so unrelated to
course work that studying is really useless. .227 .252 .238
11.a. Becoming a success is a matter of hard work, luck has
little or nothing to do w i t h i t .
b. Getting a good job depends mainly on being in the
right place at the right time. .391 .215 .301
12.a. The average citizen can have an influence in govern-
ment decisions.
h. This world is run by the few people in power, and there
is not much the little guy can do about it. .313 .222 .265
13.a. When I make plans, I am almost certain that I can
make them work.
b. I t is not always wise to plan too far ahead because
many things turn out to be a matter of good or bad
fortune anyhow. .252 .285 .271
14.a. There are certain people who are just no good. (Filler)
b. There is some good in everybody.
12 Jct.iax B. Rottkh
T A B L E 1—Continued
Hiserial item correlations
Item
200 M 200 K 400 M + F
15.a. I n my case getting what I want has little or nothing
to do with luck.
b. Many times we might just as well decide what to do by
— flipping a coin. .369 .209 .288
16.a. Who gets to be the boss often depends on who was
""" lucky enough to be in the right place first. .295 .318 .307
b. Getting people to do the right thing depends upon
ability, hick has little or nothing to do with it.
17.a. As far as world affairs are concerned, most of us are
~~ the victims of forces we can neither understand, nor
control. .313 .107 .357
b. By taking an active part in political and social affairs
the people can control world events.
18.a. Most people don't realize the extent to which their
~ lives are controlled by accidental happenings. .258 .362 .310
b. There really is no such thing as "luck."
19.a. One should always be willing to admit mistakes. (Filler)
b. It is usually best to cover up one's mistakes.
20.a. It is hard to know whether or not a person really likes
" you. .255 .307 .271
b. How many friends you have depends upon how nice a
person you are.
21.a. In the long run the bad things that happen to us are
~~ balanced by the good ones. .108 .197 .152
b. Most misfortunes are the result of lack of ability,
ignorance, laziness, or all three.
22.a. With enough effort we can wipe out political corrup-
tion.
b. It is difficult for people to have much control over the
— things politicians do in office. .226 .224 .227
23.a. Sometimes I can't understand how teachers arrive at
~ the grades they give. .275 .248 .255
b. There is a direct connection between how hard I study
and the grades I get.
24.a. A good leader expects people to decide for themselves
what they should do. (Filler)
b. A good leader makes i t clear to everybody what their
jobs are.
25.a. Many times I feel that I have little influence over
— the things that happen to me. .521 .440 .480
b. I t is impossible for me to believe that chance or luck
plays an important role in my life.
26.a. People are lonely because they don't t r y to be friendly.
b. There's not much use i n trying too hard to please
~ people, if they like you, they like you. .179 .227 .195
27.a. There is too much emphasis on athletics in high school. (Filler)
b. Team sports are an excellent way to build character.
28.a. What happens to me is my own doing.
b. Sometimes I feel that I don't have enough control over
™ the direction my life is taking. .331 .149 .238
29.a. Most of the time I can't understand why politicians
_ behave the way they do. .004 .211 .109
b. I n the long run the people are responsible for bad gov-
ernment on a national as well as on a local level.
Note.—Score is number of underlined items.
TABLE 2
I n t e r n a l - E x t e r n a l C o n t r o l T e s t D a t a : R e l i a b i l i t y and D i s c r i m i n a n t V a l i d i t y
Internal consistency
Ohio State University Split half 50 M .65
Elementary psychology Spearman-Brown 50 F .79
students
Sample 1 100 Combined .73
Kuder-llichardson 50 M .70
50 F .76
100 Combined .73
Ohio State University Kuder-Richardsou 200 M .70
Elementary psychology 200 F .70
students
400 Combined .70
National stratified sample Kuder-Richardson 1000 Combined .69 Franklin (1963)
Purdue opinion poll M & F approxi-
mately
10th, 11th, and 12th grades Equal iVs
Test-retes t relia bility
Ohio State University 1 month 30 M .60
Elementary psychology Group administration 30 F .83
students
60 Combined .72
Prisoners 1 month 28 M .78 Jossor (1964)"
Colorado Reformatory
Ohio State University 2 months 63 M .49
Elementary psychology 1st group administra-
students tion 54 F .61
2nd individual admin-
istration 117 Combined .55
Corre ation with Marlowe-C •owne Social Desirabilityr Scale
Ohio State University 166 M -.16
Elementary psychology 140 F -.32
students
306 Combined -.21
T A B L E 2—Continued
11 Personal communication.
TABLE 3
Means and S t a n d a r d D e v i a t i o n s o f I-E Scores f o r Samples o f S e v e r a l P o p u l a t i o n s
Florida State University I Group 116 Combined 9.05 3.66 Gore and Rotter
Negro students, psychology Experimental B2M, 54F (1963)
classes I
National stratified sample,! Various 1000 Combined 8.50 3.74 Franklin (1963)
Purdue opinion poll ( M &F
10th, 11th, and 12th grades j Approximately
equal jVs
18-year-old subjects from! Individual 32 M ! 10.00 i 4.20 Crowne and Conn
Boston area I 25 F j 9.00 j 3.90 (1965)"
57 i Combined I 9.56 ! 4.10
a Personal communication.
which, at least for many subjects, would with such test variables as adjustment,
result in maladjustment in our society. social desirability or need for approval.
In substance, the relationship between and intelligence are low for the samples
I-E scores and adjustment would not be a studied and indicate good discriminant
linear or a clear one from a theoretical validity.
point of view. We might expect seriously
maladjusted groups to have more vari- Multi-method Measurement
ability on I-E scores and probably more
Campbell and Fiske (1959) have indi-
frequently to have high scores in the direc-
cated the importance of multimethod meas-
tion of externality. Within a relatively
urement in the determination of construct
homogeneous (normal) group such as un-
validity of personality tests. Earlier
selected college students or high school
studies with the 60-item scale of the
students theoretical expectation would be
forced-choice I-E test typically produced
for a low linear correlation.
correlations between .55 and .60 with the
Several samples of Ohio State elemen- earlier James-Phares Likert-type scale.
tary psychology students have been ex- The largest sample studied was that of
amined for the relationship between the Blackman (1962), who obtained a correla-
I-E scale and the Rotter Incomplete Sen- tion of .56 for his 151 elementary psychol-
tences Blank (Rotter & Rafferty, 1950). In ogy student subjects. Florence Johnson
general, linear correlations have not been
(1961) obtained a correlation of .58 for
significant, and, while some curvilinear
120 subjects,
correlations have been significant, they are
Two studies of nonquestionnaire ap-
not U-shaped distributions and cannot be
proaches to the measurement of internal-
explained simply. Ware (1964)3 found a
external control have been made with the
correlation of .24 between the I-E scale
23-item scale. Adams-Webber (1963) com-
and the Taylor Manifest Anxiety scale for
pared the forced-choice I-E scores with
his 111 subjects (significant at the 5%
scores from a story-completion test. The
level). Efran (1963) used a shortened
story beginnings involved a central char-
form of the Taylor Manifest Anxiety scale
acter who initiates an "immoral" course of
and of the I-E scale and examined the
action. Scoring was based upon whether
relationship for 114 combined male and
the consequences of this act in the story
female tenth-, eleventh-, and twelfth-
completions appeared to follow from the
grade high school students. flis obtained
individual's behavior or were caused by it
correlation was .00.
or were more a function of external condi-
In summary, the test shows reasonable tions or agents. Judges rated story endings
homogeneity or internal consistency, par- from a crude manual. Adams-Webber ana-
ticularly when one takes into account that lyzed his data by dividing his 103 sub-
many of the items are sampling a broadly jects into groups based on the number of
generalized characteristic over a number external endings for his three story com-
of specific or different situations. How- pletions. Analysis of variance indicated a
ever, at least with the relatively homoge- highly significant difference among the
neous sample studied the test is limited in groups (p = <.001). The "projective" test
ability to discriminate individuals. Other of tendency to see punishment for moral
populations may provide a greater spread transgression as being externally imposed
of scores but for college students in the or as being the result of the immoral be-
middle 50% of the distribution the test is havior was significantly related to I-E
more suitable for investigations of group scale scores.
differences than for individual prediction.
Whether or not a more refined measure of In a study of academic failure, Cardi
such a broad characteristic can be de- (1962) developed a measure of internal-
veloped is an open question. Relationships external control from a semistructured
interview which ranged from 35 minutes to
3 Personal communication, 1964. an hour. Judges' ratings following a man-
18 Ji'Li.ix B. Rotter
ual were correlated with I-E scale scores fied themselves as Republican, 20 who
obtained at an earlier time and inde- identified themselves as Democrats, and
pendently of the interview. As in the 45 who said they were independent or not
Adams-Webber study the judges' ratings identified with either major party. With
were satisfactorily reliable. She obtained a the earlier 60-item scale Johnson (1961)
biserial correlation of .61 (p — <.002) for found no significant differences in I-E scale
her 25 subjects between subjects rated scores between 73 college student- sup-
high or low from the interview data and porters of Nixon and 47 supporters of
I-E scale scores. The variable being Kennedy. However, such results may not
studied here is capable of reliable meas- be typical of another geographical area
urement by a variety of test methods. where there are sharper differences in poli-
tical libcralncss between the two major
Social Class Differences in Internal-Ex- parties.
ternal Control
When the Warner scale based on father's Children's Tests of Internal-External Con-
occupation was used, studies of the Ohio trol
State samples of elementary psychology Three measures of internal-external con-
students did not show significant social- trol for children have been devised. The
class differences. However, the college first of these by Bialcr (1961) was modi-
student population utilized was highly fied from the James-Phares scale. It is a
homogeneous. Similarly, Gore and Rotter 23-itcm questionnaire with yes-no re-
(1963) failed to find significant social-class sponses. With younger children the items
differences in a somewhat lower social class are read, and the child answers yes or no.
but similarly homogeneous grouping at a A typical item is, "Do you really believe
southern Negro college. Studies with a kid can be whatever he wants to be?"
younger or noncollege age samples, how- Crandall, Katkovsky, and Preston (1962)
ever, have shown differentiation. Franklin developed a scale (Intellectual Achievement
(1963) recorded a significant relationship Responsibility—IAR) for ''self-responsibil-
between higher socioeconomic class and ity" in achievement situations. The items
internality based on his national stratified dealt with whether or not the child felt that
sample of 1,000 cases. Battle and Rotter he, rather than other persons, usually caused
(1963). using Negro and white sixth- and the successes and failures he experienced
eighth-grade children and a projective in intellectual achievement situations. The
type test, did find a significant social-class child chose between alternatives as in the
effect with race and intellectual level con- following example: "Suppose you did better
trolled. There was also a significant effect than usual in a subject at, school. Would it
for race, but most of the variance was probably happen (a) because you tried
accounted for by an interaction in which harder, or lb) because someone helped
the lower-class Negroes were considerably you?"
more external than the groups of middle- A third test for children, more projective
class Negroes or upper- or lower-class in nature, was developed by Battle and
whites. This finding is similar to the Lef- Rotter (19631. This test presented the
court and Ladwig (1965) study of Negro subject with six- situations modeled on the
and white prisoners, most of whom were Rosenzweig picture frustration approach.
lower socioeconomic class. The child was told how he would fill in
the balloon, as in comic strips, for an out-
Political Affiliation line drawing: for example, where one child
One analysis of the relationship of the is saying, "How come you didn't get what
test to political identification has been you wanted for Christmas?". A reliable
made. This study included 114 Ohio State scoring manual for this test was developed.
elementary psychology college students. This projective measure correlated signifi-
No significant differences were found in canth- (-42) with the Bialer questionnaire
the mean scores of 49 students who identi- in a group of 40 white and Negro children.
Internal versus External Control op Reinforcement 19
Data obtained from all three of these tests same direction, namely, that externals tend
will be referred to in the following section. to produce more unusual (or gambler's
fallacy) shifts. Battle and Rotter (1963)
Personal Correlates of a Gen- also found that the Bialer scale signifi-
eralized Expectancy f o r I n - cantly predicted the number of unusual
ternal-External Control: shifts for sixth- and eighth-grade Negro
Construct Validity of the and white children.
I-E Measures Liverant and Scodel (1960) examined
the preferences for bets in a dice-throwing
Performance in Controlled Laboratory
situation using the earlier 60-item version
Tasks
of the I-E scale. They found that subjects
The first investigations of individual scoring toward the internal end of the
differences in the I-E variable were made scale tended to prefer intermediate prob-
in connection with learning or performance ability bets or extremely safe bets over the
tasks in which skill and chance instructions long shots and that they tended to wager
were given. Phares (1957) obtained pre- more money on safe as against risky bets
diction bordering on significance for 13 when compared to those subjects scoring
chance-oriented items for individual differ- at the external end of the continuum.
ences in the size of the increments or In general, individual prediction in com-
decrements following success and failure petitive laboratory situations for college
and in the frequency of unusual shifts students has been only partially successful.
within conditions: that is, those shifts in Apparently, the rather narrow range of
expectancy where the subject raises his internal-external control attitudes in col-
expectancy after a failure or lowers it after lege students and the strong situational
a success. determination of the competitive labora-
James (1957), in the previously cited tory tasks limits prediction. The behavior
experiment, found low but significant pre- most susceptible to individual prediction is
diction with revised scale and individual that which deals most directly with risk
differences within each condition. In each taking and expectancies of the real influ-
case the behavior of externals differed from ence of luck as demonstrated by belief in
that of internals in the same way that the the gambler's fallacy.
overall population differed under chance
instructions as compared with skill instruc- Attempts to Control the Environment
tions. James found that the size of incre- In the following sections for the sake of
ments and decrements in expectancies fol- brevity subjects in the upper half of the
lowing reinforcement, the frequency of distribution of scores on the I-E scale or
unusual shifts, the tendency to generalize other measures will be referred to as exter-
from one task to another, and the number nals and those in the lower half as internals.
of trials to extinction were significantly I t should be made clear that we are dealing
related to his questionnaire of internal- here with only one variable affecting be-
external control. havior and that we are not implying a
Later studies using task differences typology of any kind. In fact, in some of
rather than instructions and in some cases the studies involving college populations
other types of tasks have not been as suc- those subjects being characterized as rela-
cessful as James in predicting individual tively more external may in fact be more
differences within conditions using either internal on the average than the mean of
the James-Phares scale or the more recent the population at large.
I-E scale. More consistent prediction has Perhaps the most important kind of
been made of the frequency of unusual data to assess the construct validity of the
shifts during a controlled reinforcement internal-external control dimension involves
sequence. Several investigations have found the attempts of people to better their life
this difference to be significant or near conditions, that is, to control their environ-
significant and the trend is always in the ment in important life situations. I t is in
20 Jri.i.w B. Hotter
this sense that the I-E scale appear* to attempts to change, and others not to,
measure a psychological equivalent of the must have been related to their own gen-
sociological concept of alienation, in the eralized expectancy that their behavior
sense of powerlessncss. The first study of could, in fact, effect a change in the preju-
this type was undertaken by Seeman and dice which surrounded them as well as
Evans (1962). who employed a revision of other variables. While this study had strong
the 60-item I-E scale not too different from face validity, in that the students who
the later-developed 23-item scale. They signed up for these activities expected to
investigated the behavior of patients in a take part in them, no follow-ups were
tuberculosis hospital, measuring how much made as to whether or not they actually
they knew about their own condition, how did take part. A study by Strickland
much they questioned doctors and nurses (1965) investigated activists in a Negro
about their own condition, and how satis- civil rights movement in a different state
fied they were with the amount of feedback in comparison to Negroes matched for ed-
they were getting about their medical ucation and socioeconomic status who did
status. They used 43 matched pairs of not take part in such issues. She again
white male patients, each pair being found a significant difference with the
matched for occupational status, education. activists more internal on the I-E scale.
and ward placement. As hypothesized, they Phares (1965) in a more stringent test
found that the internals knew more about of a generality of internal-external control
their own condition, questioned the doctors attitudes selected two samples, one inter-
and nurses more (according to doctors' and nal and one external, on the I-E scale but
nurses' independent ratings), and expressed matched for the attitudes towards main-
less satisfaction at the amount of feedback taining fraternities and sororities on cam-
or the information they were getting about pus. He instructed both groups to act
their condition from the hospital person- as experimenters to change the attitudes
nel. of other students. He found, as hypothe-
Seeman (1963) followed this study with sized, that his internal subject-experi-
one of reformatory inmates, investigating menters were significantly more successful
memory for various kinds of information in changing attitudes of others than the
which they were exposed to in incidental external subject-experimenters, who did
fashion. He found a significant correlation. not differ significantly in the amount of
independent of intelligence, between in- change achieved from a control group who
ternality-externality and the amount of were not subject to any influence condition.
information remembered about bow the In two separate studies the author has
reformatory was run, parole, and long- investigated the relationship between peti-
range economic facts which might affect tion signing and internal-external control.
the persons after they left the reformatory. In both instances, subjects were given the
Gore and Rotter 11963) obtained signed opportunity to sign petitions pro or con
commitments from students at a southern some issue such as Red China's being ad-
Negro college regarding activities to be mitted to the United Nations or pro or con
undertaken during vacation in behalf of postseason football games, on the pretext
the civil rights movement. Students who that only by providing both alternatives
were willing to take part in a march on the could the petitions be passed out in classes.
state capitol or to join a freedom riders' I t was hypothesized that internality on the
group were clearly and significantly more test would relate to signing the petitions
internal than those who were only willing in cither direction versus nonsigning. In
to attend a rally, were not interested in both cases, the I-E scale failed to predict.
participating at all, or avoided even filling who would be signers and who would not.
out the requested form. Since these were Whether or not the signing of a petition
all-Negro students who must have had high under classroom conditions involves other
involvement in the integration issue, the variables which were not taken into ac-
willingness of some to take part in active count and masked anv internal-external
INTERNAL VERSUS EXTERNAL CoKTROL OF REINFORCEMENT 21
control variance was n o t clear. I n a n y case, counted for by the two factor analyses of
the test failed to predict petition signing the test. In the factor analyses done by the
under these conditions. author, several factors involving small but
A recent investigation in this area in- significant variance were isolated, each
volves a different cultural population. See- involving only two or three items with
man (1964)' studied workers in Sweden significant loadings. These factors, however,
with a translated version of the I-E scale. were highly specific and did not constitute
Seeman's results seem to point clearly to broad enough subareas to appear useful
the fact that membership in unions versus to the author. I t must also be remem-
nonmembership, activity within the union, bered that most items involving achieve-
and general knowledge of political affairs ment had to be dropped from the scale
were all significantly related to intern a lity. because of their apparently great suscepti-
Correlations were low but significant and bility to social desirability influence.
held up when controlled for variables such Internal-External Control and Achieve-
as education, age, and income. ment Motivation
Perhaps related to this feeling that one
can control the environment is also a feel- I t would seem a logical extension of the
ing that one can control himself. Some stud- notion of internal-external control that
ies of the relationship of internal-external those at the internal end of the scale would
control to smoking perhaps arc relevant. show more overt striving for achievement
Straits and Sechrest (1963) found that non- than those who felt they had little control
smokers were significantly more internal over their environment. However, there
than smokers, and James, Woodruff, and are two limitations on the potential
Werner (1965) replicated that finding and strength of this relationship, particularly as
in addition reported that following the it applies to college students or adults. One
Surgeon General's report, among male of these is that among college students and
smokers, those who quit and did not return adults, particularly with males, there arc
to smoking in a specified period of time more defensive externals or people who
have arrived at an external view as a de-
were more internal than those who believed
fense against failure but who were origi-
the report but did not quit smoking. The
nally highly competitive. Many such peo-
difference was not significant for females
ple still maintain striving behavior in
who apparently were motivated by other
clearly structured competitive situations
variables including, for example, the tend- but defensively account for failures by
ency to gain weight when not smoking. expressed external attitudes. The other
This group of studies lends strong and limitation is one of specificity in that in-
relatively consistent support to the hypoth- ternal-external control attitudes are obvi-
esis that a generalized expectancy—that ously not generalized across the board, and
one can affect the environment through in the highly structured academic achieve-
one's own behavior—is present in at least ment situation there is probably more
two different cultures, can be reliably specificity determining response than in
measured, and is predictive of logical be- other kinds of situations. With children
havioral construct referents. who have less experience in the competitive
While significant correlations have been academic situation, a higher relationship
referred to throughout this section, it could be anticipated than with a select
should be made clear that for the most part population of college students.
they are low and leave room for much in
the way of specific attitudes in the par- The Crandall et al. (1962) scale (IAR)
ticular areas of behavior that were in- developed for use with children is specific
vestigated. Perhaps some explanation may to the achievement area and Crandall et al.
be called for here for the fact that variance did find free play achievement behavior
due to the specific situation was not ac- and achievement test scores in boys but not
in girls related to test scores. Neither the
* Unpublished manuscript. Children's Manifest Anxiety scale nor the
Julian B. Rotter
Thematic Apperception Test (TAT) data had shown that some people were
achievement measure predicted either boys' very good at it. All subjects were then
or girls' achievement behavior. given eight trials of which 75% were posi-
Cellura5 found a direct relationship be- tively reinforced, followed by an extinction
tween the SEA academic achievement test, series of no correct answers until their
with IQ partialled out, of lower socioeco- verbalized expectancies reached 1 or 0 for
nomic status boys and the IAR scale. two consecutive trials. Within the skill and
Crandall's subjects were predominantly chance groups, subjects were divided into
middle class, and Cellura's were all in internals and externals at the median. De-
Hollingheads' fourth and fifth categories. cision time was measured for all subjects
In Franklin's (1963) study of high school from the time they were given the sample
students involving the national stratified. for judging until they selected a standard.
sample of 1,000, he hypothesized 17 rela- The subject was unaware that he was being
tionships of the I-E scale to reported evi- timed. Analysis of variance produced a
dences of achievement motivation. These significant interaction. Internals took longer
included such things as early attempts to to decide on a matching standard under
investigate colleges, intention to go to col- skill conditions than did externals but took
lege, amount of time spent doing homework, a shorter time under chance conditions than
parents' interest in homework, etc. He did externals. Most of the difference was at-
found a significant relationship in the pre- tributed to the internals who had very long
dicted direction in 15 of his 17 relation- decision times under skill conditions and
ships. very short times under chance conditions,
these being significantly different. The ex-
A study by Efran (1963) produced an
ternals took longer under chance conditions
indirect but extremely interesting indica-
than under skill conditions but the differ-
tion of the relationship between striving
ence was not significant. The result not only
for achievement and internal-external con-
shows the greater involvement of internals
trol. Using a balanced-order controlled
under skill conditions but in general sug-
procedure he studied high school students'
gests that internals tend to value reinforce-
tendency to forget (repress) failures versus
ments for skill much more than chance, and
successes and found that the tendency to
if the opposite cannot be said for the more
forget failures was significantly related to
external subjects of this study, it is at least
scores towards the internal end of the di-
clear that there is no significant differen-
mension. It is possible that the functional
tiation for them.
value of a defensive tendency towards ex-
ternality is indicated by these findings. The In summary, the expected relationship
results suggest that the external has less between the tendency to perceive what
need to "repress" his failures since he has happens to a person as dependent upon
already accepted external factors as deter- his own actions and greater motivation in
mining his success and failure to a greater achievement is generally supported al-
extent than those subjects scoring as more though prediction was not consistent for
internal on the I-E control scale. boys and girls using the Crandall et al. scale
A study by Rotter and Mulry (1965) with children.
also supports the stronger motivation of Internal-External Control and Resistance
internals in achievement situations. In this to Subtle Suggestion
study, 120 male and female unselectcd One other area of construct validity has
subjects were placed in an angle-matching been investigated in some depth. This in-
situation of extreme difficulty. Half of the volves the variables of independence, sug-
subjects were instructed that the task was gestibility, and conformity as related to
so difficult as to be chance determined and internal-external control. I t seems that in-
half that it was difficult but that previous ternals would be more resistive to manipu-
"Unpublished manuscript, "Tnternality as a lation from the outside if, in fact, they arc
determinant of academic achievement in low aware of such manipulation. If they were
SKS adolescents." Syracuse University, 1963. aware, they would feel deprived of some of
Internal versus External Control op Reinforcement 23
their control of the environment. Externals were aware and did not condition and
expecting control from the outside would those subjects who were aware and did
be less resistive. One special consideration condition. As expected, the subjects who
here, however, is in the area of conformity. were aware and did not condition were
If the internally oriented person perceives considerably more internal than those who
that it is to his advantage to conform he were aware and did condition.
may do so consciously and willingly with- The study by Getter involving a some-
out yielding any of his control. It is only what different technique produced a fairly
where it might be clearly to his disadvan- large number of latent conditioners. That is,
tage that he would resist conformity pres- subjects who showed no significant evi-
sures. dence, of conditioning during the training
An investigation of the latter hypothesis trials, but during extinction when no rein-
was carried out by Crowne and Liverant forcement was given, showed a significant
(1963). They studied unselected college rise in the reinforced response. Again,
students, dividing them at the median into these subjects were significantly more in-
internals and externals and observing them ternal than either subjects who did not
in an Asch conformity situation. Under one show such latent conditioning among non-
set of conditions the usual Asch instruc- conditioners or who conditioned during the
tions were used. In the second set of con- training trials.
ditions, subjects were given a certain Both of these studies suggest a kind of
amount of money and allowed to bet on negativism to external manipulation on the
each of their judgments. Subjects could part of internals. However, a study by
choose to bet or not to bet and could deter- Gore (1962) helps to clarify this issue.
mine the amount they were willing to bet Gore used an experimenter influence para-
on each judgment. In the normal Asch digm in which she presented TAT cards to
situation, there were no differences be- three groups of subjects, ostensibly to
tween internals and externals in the determine which cards produced longer
amount of yielding. However, under bet- stories. One condition involved overt influ-
ting conditions the internals yielded signifi- ence in which she specified which card she
cantly less than the externals. They also thought was the best. The second condition
bet more on themselves when going against involved subtle influence in which she pre-
the majority than did externals on their sented the same card, saying to the sub-
independent trials. The internals had no jects and smiling, "Now let's see what you
significant differences between their bets on do with this one." The third condition was
conforming and independent trials, but the a control condition of no influence. She also
externals bet significantly less on inde- used unselected subjects dividing them at
pendent trials than they did on trials on the median into internals and externals.
which they yielded. Her results showed no significant differ-
Other tests of the tendency to yield to ences between internals and externals
external influence were obtained from stud- under the overt suggestion condition and
ies of Strickland (1962) and Getter (1962) control condition, but under the subtle sug-
relating scores on the I-E scale to verbal gestion condition the internals produced
conditioning. On the basis of a thorough significantly shorter stories than did ex-
postexperimental interview, Strickland di- ternals and significantly shorter stories
vided her subjects into those who were than did control subjects in the no sugges-
aware of the reinforcement contingency tion condition. I t is under the subtle sug-
and those who were not. In addition, of gestion conditions they reacted by telling
those who were aware she divided those shorter stories or were, in fact, negativis-
who conditioned from those who did not. tic. However, in the overt condition, there
While she found no overall relationship were no traces of this negativism. Appar-
between conditionability and I-E scale ently, when given the conscious choice the
scores, she did find large and significant internal is not resistive. However, when he
differences between those subjects who is aware that an attempt is being made to
24 Julian B. Rotter
.subtly manipulate him he does become re- within this group there was a significant
sistive. relationship between intelligence and ex-
The four studies taken as a whole sup- ternality. This relationship was counter to
port one another. The individual who per- the socioeconomic level findings. I t was the
ceives that he does have control over what more intelligent Negro children in the lower
happens to him may conform or may go socioeconomic level who were the most
along with suggestions when he chooses external. The findings are based on a small
to and when he is given a conscious alter- A' and may be regarded as only suggestive.
native. However, if such suggestion or at- However, what they imply is that the per-
tempts at manipulation arc not to his ception of limited material opportunities
benefit or if he perceives them as subtle and of powerful external forces is one
attempts to influence him without his variable making for an external attitude.
awareness, he reacts resistively. The find- Similarly, Cellura (See Footnote 5) found
ings have considerable significance for the that with both the Bialer scale and the
general area of persuasion and propaganda. Crandall et al. IAR scale the parents of the
more external children had significantly
Antecedents of Internal-External Attitudes lower education levels.
Relatively little work has been clone on
One investigation with college student
antecedents for developing attitudes of in-
subjects attempted to relate orthodoxy of
ternal versus external control. The con-
religious beliefs to internal-external con-
sistent indication that lower socioeconomic
trol (Rotter, Simmons, it Holdcn, 19617)
level groups are more external allows for a
using the McLean scale which measures be-
number of alternative explanations. Graves lief in the literalness of the Bible. No rela-
(1961) predicted and found differences tionship was found. With other college
among Ute Indians, children of Spanish- student groups there were no significant
American heritage, and whites in an iso- differences between individuals of one
lated triethnic community. As he expected religious faith versus another. Interviews
the Indians are most external, the Spanish- with individual subjects in a college popu-
Americans in the middle, and the whites lation at least suggest that religion may well
more internal. The implication is for direct have a role in the development of internal
cultural teaching of internal-external atti- or external attitudes. However, it is the
tudes since, in fact, the Spanish-Americans specific emphasis that is placed upon the
were financially more deprived than the role of external fatalistic determination by
Indians although it is true that the Indians parents which is more likely to determine
who were partially supported by the the attitude than the abstract doctrines of
government had fewer occupational out- the sect. These studies bearing on the prob-
lets. In support of this interpretation of lem of antecedents of internal-external
the influence of direct teaching, Shirley beliefs are indirect, and work needs to be
lessor (1964)" found a correlation of .38 be- done in this area of investigation. One ob-
tween mothers' coded answers to interview vious antecedent worthy of study would
questions of internal-external attitudes and be the consistency of discipline and treat-
responses of their high school children to a ment by parents. Clearly it would be ex-
23-item questionnaire similar to the 23- pected that unpredictable parents would
item scale described here. Her sample in- encourage the development of attitudes of
cluded 81 pairs drawn from the same external control.
cross-cultural sample.
In the previously cited study by Battle Summary
and Rotter (1963), the highly external The studies reported here represent an
group was the low-socioeconomic-level Ne- unusually consistent set of findings. For
gro group in contrast to middle-class Ne- most findings there are replications some-
groes or whites of cither lower- or middle- times in other laboratories, sometimes with
class identification. Interestingly enough, other kinds of populations, and sometimes
" Personal communication, 1964. 7 Unpublished manuscript, 1961.
Internal versus External Control of Reinforcement 25
APPENDIX A
INSTRUCTIONS FOR T H E I-E SCALE
This is a questionnaire to find out the way in by the examiner on the answer sheet, then finish
which certain important events in our society reading these directions. Do not open the book-
affect different people. Each item consists of a let until you are told to do so.
pair of alternatives lettered a or b. Please se- Please answer these items carefully but do
lect the one statement of each pair (and only not spend too much time on any one item. Be
one) which you more strongly believe to be the
sure to find an answer for every choice. Find the
case as far as you're concerned. Be sure to se-
number of the item on the answer sheet and
lect the one you actually believe to be more true
rather than the one you think you should choose black-in the space under the number f or 2
or the one you would like to be true. This is a which you choose as the statement more true.
measure of personal belief: obviously there arc I n some instances you may discover that
no right or wrong answers. you believe both statements or neither one. I n
Your answers to the items on this inventory such cases, be sure to select the one you more
are to be recorded on a separate answer sheet strongly believe to be the case as far as you're
which is loosely inserted in the booklet. R E - concerned. Also try to respond to each item
M O V E T H I S A N S W E R S H E E T NOW. Print independently when making your choice; do not
your name and any other information requested be influenced by your previous choices.
APPENDIX B
TABLE B l
D i s t r i b u t i o n of I-E Scale Scores f o r 575 M a l e s and (505 Female Ohio State
E l e m e n t a r y Psychology Students
Males" Fem;ilesb
I-E score / Cum. % I-E score / Cum. ^
21 1 100.00
20 1 100.00 20 1 99.83
19 1 99.83 19 3 99.07
18 4 99.65 18 7 99.17
17 10 98.90 17 10 98.02
16 10 97.22 16 8 96.30
15 10 95.48 15 17 95.04
U 15 93.74 14 23 92.23
13 31 91.13 13 37 88.43
12 32 85.74 12 31 82.31
11 32 80.17 11 42 77.19
10 49 74.61 10 42 70.25
9 S3 66.09 9 64 63.31
8 73 56.87 8 53 52.73
7 52 44.17 7 50 43.97
6 52 35.13 6 06 35.70
5 41 26.09 5 37 24.79
I 43 18.96 4 42 18.68
3 29 11.48 3 37 11.74
2 22 6.43 2 22 5.62
1 10 2.61 1 8 1.98
0 5 0.87 0 4 0.66
»A' = 575; Mean = 8.15; SD = 3.88.
b N = 605; Mean = 8.42; SD = 4.06.
REFERENCES
Adams-Webber, J. Perceived locus of control of Axgyal, A. Foundations for a science of personality,
moral sanctions. Unpublished master's thesis, New York: Commonwealth Fund, 1941.
Ohio State University, 1963.
IxtEHXAL VERSUS EXTERNAL CONTROL OK REINFORCEMENT 27
Atkixsox, J. AY. (Ed.) Motives in fantasy action prediction of subject compliance to experi-
and society. Princeton: D. Van Nostrand, 1958. menter bias. Unpublished doctoral dissertation,
Battle, Esther S., & Rotter, J. B. Children's feel- Ohio State University, 1962.
ings of personal control as related to social class Gore, Pearl Mayo, & Rotter, J. B. A personality
and ethnic group. Journal oj Personality, 1963, correlate of social action. Journal oj Personality,
31,482-490. 1963,31,58-64.
Benxiox, R. C. Task, trial by trial score variability Graves, T. D. Time perspective and the deferred
of internal versus external control of reinforce- gratification pattern in a tri-ethnic community.
ment. Unpublished doctoral dissertation, Ohio Unpublished doctoral dissertation, University of
State University, 1961. Pennsylvania, 1961.
Bialer, I. Conceptualization of success and fail- Harlow, H. F. The formation of learning sets.
ure in mentally retarded and normal children. Psychological Review, 1949, 56, 51-65.
Journal oj Personality, 1961, 29, 303-320. Holden, K. B., & Rotter, J. B. A nonverbal meas-
Blackman, S. Some factors affecting the perception ure of extinction in skill and chance situations.
of events as chance determined. Journal, oj Psy- Journal oj Experimental Psychology, 1962, 63,
chology, 1962, 54, 197-202. 519-520.
Campbell. D. T., it Fiske, D. W. Convergent and James, W. H. Internal versus external control of
discriminant validation by multitrait-multi- reinforcement as a basic variable in learning
mcthod matrix. Psychological Bulletin. 1959, 56, theory. Unpublished doctoral dissertation, Ohio
81-105. State University, 1957.
Cardi, Miriam. An examination of internal versus James, W. H., & Rotter, J. B. Partial and 100% re-
external control in relation to academic failures. inforcement under chance and skill conditions.
Unpublished master's thesis, Ohio State Uni- Journal oj Experimental Psychology, 1958, 55,
versity, 1962. 397-403.
Cohen, J. Chance, skill and luck. Baltimore: Pen- James, W. I-L, Woodruff, A. B., & Werner, W.
guin Books, 1960. Effect of internal and external control upon
Ckaxdall, V. J. Achievement. In Harold W. Steven- changes in smoking behavior. Journal oj Con-
son et ah (Eds.), Xational Society jot the Study sulting Psychology, 1965, 29, 184-186.
oj Education Yearbook: Part I. Child psychol- Johnson, Florence A". Political attitudes as re-
ogy. Chicago: Univer. Chicago Press, 1963. lated to internal and external control. Unpub-
Craxdall, V. J., Katkovsky, W., & Preston, Axxe. lished master's thesis, Ohio State University.
Motivational and ability determinants of young 1961.
children's intellectual achievement behaviors. Ladwig, G. W. Personal, situational and social
Child Development, 1962, 33, 643-661. determinants of preference for delayed reinforce-
Crowxe, D. P., & Liverant, S. Conformity under ment. Unpublished doctoral dissertation, Ohio
varying conditions of personal commitment. State University, 1963.
Journal oj Abnormal and Social Psychology, 1963, Lefcourt, H. M., & Ladwig, G. W. The American
66,547-555. Negro: A problem in expectancies. Journal oj
Crowne, D. P., & Marlowe, D. The approval Personality and Social Psychology, 1965, 1, 377-
motive. New York: Wiley, 1964. 380.
Efran, J. S. Some personality determinants of Lixtox, Harriet R. Dependence on external in-
memory for success and failure. Unpublished fluence : Correlations in perceptions, attitudes
doctoral dissertation, Ohio State University, and judgment. Journal oj Abnormal and Social
1963. Psychology, 1955, 5 1 , 502-507.
Feather, N. T. Subjective probability and de- Liverant, S., & Scodel, A. Internal and external
cision under uncertainty. Psychological Review, control as determinants of decision making un-
1959, 66,150-164. der conditions of risk. Psychological Reports,
Franklin, R. D. Youth's expectancies about in- 1960, 7, 59-67.
ternal versus external control of reinforcement McClelland, D.. Atkinson, J. W., Clark, R. A.,
related to N variables. Unpublished doctoral dis- & L o w e l l , E. L. The achievement motive.
sertation, Purdue University, 1963. New York: Appleton-Century-Crofts, 1953.
Getter, H. Variables affecting the value of the re- Mertox, H. Mass persuasion. New York: Harpers,
inforcement in verbal conditioning. Unpublished 1946.
doctoral dissertation, Ohio State University, Mertox, R. Social structure and anomie. In So-
1962. cial theory and social structure. Glencoe, 111.:
Goonxow, Jacqueline J., & Pettigrew, T. F. Ef- Free Press, 1949, 125-149.
fects of prior patterns of experience upon Nettler, G. A measure of alienation. American
strategies and learning sets. Journal oj Experi- Sociologiccd Review, 1957, 22, 670-677.
mental Psychology, 1955, 49, 381-389. Pepitoxe, A. Attributions of causality, social at-
Goodnow, Jacqueline J., & Postman, L. Probabil- titudes and cognitive matching processes. I n
ity learning in a problem-solving situation. Jour- R. Taguiri & L. Petrillo (Eds.), Person per-
nal oj Experimental Psychology, 1955, 49, 16- ception and interpersonal behavior. Stanford.
22. Calif.: Stanford University Press, 1958. Pp. 258-
Gore. Pearl Mayo. Individual differences in the 276.
28 Jclian B. Rotter