You are on page 1of 6

, Brhav. The-r. & Exp. Psychrut Vol. 23. No. 4. pp. ?Sl-256. IYY?. wi1S-7916/92$5.IXl+ Il.

lXl
Prmtcd m tircat Bntain Pergamon Prcrs Ltd.

SELF- .EFFICACY: A PREDICTOR BUT NOT A CAUSE OF BEHAVIOR

RUSSELL M. F. HAWKINS
University of South Australia

Summary - The concept of self-efficacy, as expounded by Bandura as part of his Social


Cognitive Theory, has made considerable impact in the psychological literature. It is argued
that self-efficacy is a useful hypothetical construct for predicting behavior, but it has no valid
claim to being a cause of behavior. Claims for self-efficacy as a causal agent have failed to
acknowledge that self-efficacy itself is an epi-phenomenon of performance. Conventional
learning theory explanations of observed performance levels are shown to be more
parsimonious than accounts relying on the concept of self-efficacy.

In 1978, Kazdin asserted that “the develop- ing behavior but has no claim to being a cause
ment of empirically based constructs such as of behavior.
self-efficacy should be encouraged and allowed Bandura (1986) has presented what he calls a
to grow considerably before being subjected to Social Cognitive Theory (SCT) which is his
overly zealous criticism” (Kazdin, 1978, p. most comprehensive statement of “an integra-
184). In 1991, Corcoran suggested that it may tive theoretical framework to explain and
how be the time to re-examine the utility of the predict psychological changes achieved by dif-
construct of self-efficacy. ferent modes of treatment”. This theory has
Lee (1989) has argued that the ability of the evolved through a variety of revisions and
concept of self-efficacy to explain human be- elaborations (e.g., Bandura, 1977, 1980,
havior is largely illusory. Nevertheless, the 1982a, b, 1986, 1989). The notion of self-
illusion must be quite compelling since a efficacy is central to this theory which assumes,
PsychLit search revealed an annual average of that psychological procedures, whatever their form,
close to 100 articles on self-efficacy from 1983 serve as means of creating and strengthening expecta-
to the present. The utility of the concept has tions of personal efficacy efficacy expectations are a
major determinant of peoples’ choice of activities, how
been widely debated, with Corrigan (1990) much effort they will expend, and of how long they will
suggesting that sophisticated statistical techni- sustain effort in dealing with stressful situations”
ques provide the basis for an “empirically valid (Bandura, 1977, pp. 193-194, emphasis added).
model” incorporating the self-efficacy concept. Much of Bandura’s experimental work on
Lee responded by suggesting that since many self-efficacy relates to what he has called “the
theories of psychology (including those that causal contribution of self-efficacy to action”
incorporate the self-efficacy concept) rely on (Bandura, 1982a, p. 126). For example, Ban-
unobservable variables, they are “useless dura (1982a) reported modifying self-efficacy
in understanding or predicting behavior, from an initially low level to a higher level and
although they may be seductive ways of talking he showed that performance varied as a func-
about that behavior afterwards” (Lee, 1990, p. tion of the subject’s self perceived efficacy. At
143). The present paper will argue that self- first analysis, Bandura’s case seems well sup-
efficacy has a certain utility in terms of predict- ported. If systematic changes to an antecedent

Requests for reprints should be addressed to Russell M. F. Hawkins. Department of Psychology, Magi11 Campus
University of South Australia. Lorne Avenue, Magill, SA 5072, Australia.

251
252 RUSSELL M. F. HAWKINS

variable reliably lead to predictable changes in the changes in behavior. There was a clear and
a consequent variable, causation appears to close relationship between self-efficacy and
have been demonstrated. Other authors have performance (Bandura. 1977). The difficulty
also supported the idea that self-efficacy is a with this experiment lies not in the results,
causal agent (e.g., Earley and Lituchy, 1991). which have been replicated in a large number
There is certainly an abundance of evidence of analogous studies, but in their interpreta-
indicating a relationship between self-efficacy tion. The main argument of this paper will be
and behavior. For example, Devins and Ed- that the attribution of causal properties to self-
wards (1988) found self-efficacy to be the most efficacy, as described in Bandura’s publica-
important contributor to effective smoking self tions, is inappropriate since self-efficacy is
change. Bandura (1977) has shown that people merely a convenient hypothetical construct.
avoid activities that they believe exceed their one inferred to summarize observed consisten-
coping capabilities, and Bandura (1982a) has cies of behavior, and one best restricted to use
cited studies which show that self judgements in a metaphorical sense to facilitate communi-
of efficacy relate to how much effort people cation.
will expend in the face of obstacles and
difficulties. Poser (1978) has argued that even An Alternative Explanation of the Generic Self-
if “verbal reports of self-efficacy predict later efficacy Experiment
performance, [this] does not in itself establish
the changes in self confidence ‘cause’ the Since Bandura’s model regards “enactive
behavioral change” (p. 197). attainments” as one of the most powerful
determinants of self-efficacy, experiments
The Generic Self-efficacy Experiment which utilized the procedure of using “enactive
attainments” to modify self-efficacy will be
In an experiment representative of the type discussed first. “Enactive attainments” (Ban-
used to support Bandura’s causal analysis, dura, 1982a) have also been called “perform-
snake phobic subjects had the level, strength ance accomplishments” and the term is essen-
and generality of their self-efficacy expecta- tially a euphemism for “direct experience”.
tions measured (Bandura, 1977). The subjects In the Bandura (1977) experiment, the treat-
designated the tasks they expected they could ment procedure was an example of “enactive
complete from a list of performance tasks attainment”. This treatment, by virtue of its
arranged in order of increasing threat; they being concrete behavioral evidence of mastery
rated the strength of their expectations and experience, is the reason for the improvement
finally rated their efficacy expectations for in self-efficacy. Bandura has made this point
snakes similar to the ones used in treatment as himself:
well as for dissimilar snakes. These measures
were taken prior to treatment for the phobia, enactive attainments provide the most influential
source of efficacy information because it can be based
following treatment but before the behavioral on authentic mastery experiences. Successes heighten
posttest. and after completing the behavioral perceived self-efficacy; repeated failures lower it (Ban-
posttest. The treatment consisted of either dura, 1982a, p. 126).

engaging in progressively more threatening “Enactive attainments” are essentially just


interactions with a boa constrictor, or observ- behaviors. We know that behaviors which are
ing the therapist performing the progressive closely followed by positive reinforcers will
interactions. increase in their frequency or intensity and
The results of this experiment were that the behaviors which are not reinforced will extin-
greater the increments in self perceived effi- guish. Purely functional analyses of behavior
cacy as a result of the treatment, the greater now routinely lead to modification programs
A Predictor but not a Cause of Behavior 253

which change behavior essentially by mod- assumption that self-efficacy is an independent


ifying various response contingencies. One variable. The salient independent variable is
lesson from the success of these programs is actually the amount of training used to produce
that it is not necessary to create or infer the rise in self-efficacy. The mechanism used to
intervening hypothetical constructs relating to improve self-efficacy was itself a form of
unobservable mental states in these subjects. training towards improved performance. It is
To make the point clearly, it is not useful to therefore not surprising that subsequent per-
suppose that a rat in a Skinner box improves formance improved and it is not necessary to
during magazine training because of a blossom- utilize the idea of self-efficacy to explain the
ing sense of self-efficacy; it is sufficient to improvement. Borkovec (1978) has also criti-
observe simply that when positive reinforce- cised self-efficacy theory utilizing a similar
ments are given contingent on certain be- argument to the one above. He concluded that
haviors, those behaviors will increase. since existing learning principles can adequ-
In the Bandura (1977) experiment, the treat- ately account for behavioral change without
ment itself involved procedures which can be recourse to unobservable cognitive events,
viewed as extinction procedures for snake self-efficacy is better viewed as a reflection of
phobia. Systematic desensitization to a snake behavior change than a cause.
phobia involves a progression through a hier-
archy of fearful events. What essentially mat- Further Considerations of Causation
ters is the subject’s continued tolerance of the
objects (i.e., his or her behavior). The conven- As well as “enactive attainment”, Bandura
tional explanation of systematic desensitization has listed three further sources of information
is that each hierarchical step is a small one, that which lead to perceptions of self-efficacy.
the anxiety is extinguished as the consequences These are “vicarious experience”, “verbal per-
of the behavior are found to be innocuous and suasion” and “physiological state”.
that repetition of this process through a care- “Vicarious experience”, although inferior to
fully designed hierarchy eventually extin- direct experience in terms of modifying anxiety
guishes the original phobia. At no stage in the and avoidance problems, is nevertheless an
therapy is it necessary to invoke the concept of effective source of learning and performance
self-efficacy at all. If it were to be evoked, it (Roper, Rachman and Marks, 1975). Bandura
could be equally regarded as a consequence of has described the causal role of self-efficacy in
behavior (that is, it is the result of success at mediating the level of performance after vica-
hierarchy level x) as a cause of behavior (that is rious experience but, while there is no quarrel
the cause of success at hierarchy level x + 1). with the extensive literature which shows that
Thus the outcome can be explained adequately modeled behavior leads to learning in an
without recourse to mentally inferred states. observer, the concept of self-efficacy seems
Bandura (1982a) reported modifying self- redundant for the same reasons as detailed in
efficacy from an initially low level to a higher the discussion of “enactive attainment”.
level, and he showed that performance varied “Verbal persuasion” is Bandura’s third
as a function of the subject’s self perceived category. Bandura says that verbal persuasion
efficacy. The results of this experiment were will lead to increased self-efficacy which will
displayed graphically following the convention lead “to successful performance if the height-
of placing the independent variable (in this ened appraisal is within realistic bounds” (Ban-
case, self-efficacy) on the abscissa. The im- dura, 1982a, p. 127). The major problem with
pression conveyed is certainly that as levels of this category is that once again the concept of
self-efficacy were improved, so too was per- self-efficacy is not required to explain the
formance. Again, the difficulty is in Bandura’s behavior. Verbal persuasion may lead people
251 RUSSELL M. F. HAWKINS

to attempt tasks they might otherwise avoid for convenience of communication, but it is not
(e.g., a parent persuades a child to put his or formally required at all. A functional analysis
her face in the water as part of a swimming of antecedent and consequent conditions which
lesson) or it may lead to people making a has no recourse to what a radical behaviorist
greater effort. The improved performance can might term the “mystical” concept of “self-
be understood in a straightforward fashion as efficacy” is more parsimonious than an expla-
being the result of the increased willingness to nation which does not need to invoke such a
try a new task or to try harder with an existing concept (Skinner, 1053). This is not to suggest
task. that the term “self-efficacy” is meaningless, but
Kirsch (1982) has demonstrated that the it would be better used with the clear under-
behavior of snake phobic subjects is not so standing that it is simply a metaphor.
much a matter of belief in one’s uhility to As has already been noted, Bandura has at
accomplish a task as of response willingrzex~. times conceded that there are reciprocal links
Kirsch demonstrated that the use of hypotheti- between self-efficacy and performance thus
cal financial incentives influenced people’s acknowledging that behavior may determine
reported willingness to undertake a task re- self-efficacy. For example.
quiring an approach to a feared stimulus.
pcrformancc mastery. in turn. can boost perceived aclf-
Kirsch (19X.5) has further argued that since efficacy in a mutually enhancing process. It is not as
questionnaires about people’s ability to though self percepts of efficacy affect future pcrform-
approach a feared stimulus actually mcasurc ancc hut play no role whatnoevcr in earlier perform-
:lncc attainments (Bandura, lOX2a. p.12X).
their willingness to approach the stimulus.
Bandura’s examples of snake-approach self- However. in an attempt to re-establish the
efficacy are inconsistent with his method of primacy of self-efficacy he further stated.
measuring the construct. Question5 about causal ordering of factors ark in
Perhaps in anticipation of the criticism that enactivcly based influences when interactive proccsscs
arc treated as linear sequential ones and causally prior
the notion of self-efficacy is superfluous to any
self-efficacy determinants of past performance accom-
explanation of observed performance and in plishments go unmeasured cvcn in the casts 01
further argument to support the role of self- enactively instated self-efficacy. performance is rwt rkc~
efficacy as a cause of performance. Bandura ~wx~si.s of t/w ~ausul chuin (Bandura. 19X2a. p. 12X.
emphasis added).
has said:
Bandura supported this quotation with the
Information that is relevant for judging personal idea that self perceived learning efficacy affects
capabilities - whether conveyed enactively. vicari-
ously. persuasively or physiologically - is not
how much effort is invested in an activity and
enlightening. Rather it becomes instructive only thus determines behavior. This ignores the
through cognitive appraisal (Bandura, 1982;r. p. 127). probability that previous performance out-
Cognitive appraisal of course includes media- comes may have determined the level of
tion via a person’s self-efficacy notions; thus perceived self-efficacy. This is not an insoluble
self-efficacy is accorded a pivotal place in the chicken and egg problem. Self-efficacy must
understanding of human behavior. The point have an origin and it seems likely that early
here which he has not addressed, however, is behavior is critical in this regard. At some early
whether or not information needs to “become stage responses were emitted, which were or
instructive” for improved performance to re- were not reinforced. The reinforced responses
sult. A veritable plethora of operant condition- strengthened over time and these represent the
ing experiments has shown that not even rudiments of performance. Thus learning
consciousness, let alone an evaluation of self- theory may explain even the development of
efficacy. is required for learning to occur. The sophisticated behavior. Cognitive appraisal of
idea of self-efficacy may be a “handy” concept behavior would have occurred after the fact,
A Predictor but not a Cause of Behavior 2.55

not before it. Implicit in Bandura’s quotation perception theory (Bern, 1970) postulated that
(see previous page), is the idea that if the in fact attitudes are only inferred from be-
origins of self-efficacy are not in performance, havior after the event.
they must be a genetic endowment of some It is suggested that self-efficacy ratings are a
type. This is a provocative idea for which consequence not a cause, of behavior. While
evidence has not yet been provided. Bandura has pointed out that self-efficacy can
In his historical review, Kirsch (1986) has be measured before the behavior of interest
both clarified the concept of self-efficacy and and therefore cannot be an effect, his time
shown how it is a derivation of Rotter’s (1966) sample may not be appropriate. That is, the
social learning theory. In its clarified form, level of self-efficacy is not suddenly present
self-efficacy, as a proven predictor variable “de nova” but has been determined by prior
(O’Leary, 1985), may be worth measuring in events. A further analogy will clarify this point.
clinical situations. However, the reason for the The behavioral concept of stimulus control
success of self-efficacy as a predictor need not includes the idea that certain stimuli set the
be because it causes subsequent behavior but occasion for particular responses (e.g., a ring-
because it is an index of the performance ing telephone signals that a certain behavior is
history of past successes and failures (i.e., a likely to be reinforced). This does not mean
running average). To illustrate the point fur- that a SD causes the behavior. In fact the SD
ther, one might invent a concept called “mean has acquired its potency from previous learned
runs scored in cricket matches”. Such a con- associations in much the same way that it is
struct may be measurable (i.e., you could ask a now argued that self-efficacy is a product of
person for a verbal report of their best estimate past performance assessments.
of their average run rate). This information In a critique of social science’s attempts to
may very well serve to predict effectively demonstrate causality, Ryder (1987) has made
performance in a future cricket match but it is a number of points which seem apposite to the
clearly not the cause of the future behavior. self-efficacy issue. In Ryder’s concept of
“animism”,
one constructs a myth of an entity not immediately
Difficulties associated with the attribution of seen, that “causes” what is seen, and that is invested
causation to hypothetical constructs with a life of its own. An explanatory concept of
some sort, a mythology. is invented, and then is so
loved by its inventor that is seems to have life and be
Bandura has emphasized the notion of self- real, at least to the inventor. Animistic causality
efficacy as an explanatory, causative mediating can, of course. mean nothing more than summarizing a
variable which is central to the understanding set of observations by giving them a name and then
investing the name itself with a life of its own.
of the quality of performance. But what is self- Suppose a person’s activity seems well described by
efficacy? Essentially it does not exist; it is a saying that the person is sleepy almost all the time. The
hypothetical construct which is invoked to help sleepiness is then “explained” by inventing a property
that “causes” sleepiness, let us say “soporific propen-
understand a pattern of observed behavior. In sity” or SP. Since SP does nothing else but cause
this sense it is analogous to the concept of an something that is only a summary label for observa-
“attitude”, another hypothetical construct of tions, and has no obvious claim to being anything other
than an alternative summary label, it can be said that
supposedly great explanatory value. However, SP is a pure case of an animistic and vacuous usage of
in the field of attitude research a controversy causality. Consider a man who regularly acts in the
has long raged over whether attitudes cause ways that clinicians are likely to describe as obsessio-
nal. Having thus described what he does, is something
behavior or not. While it may be intuitively added by saying that he is an obsessional personality
attractive to suppose that they do and several type? Why does Joe sleep with men? Because he is
consistency theorists (e.g., Festinger, 1957) gay. you dummy. Oh. I see (Ryder, 1987. pp. 98-101).

have maintained that this is so, Bern’s self Bandura’s attribution of causality to the
2x1 RUSSELI_ M. F. HAWKINS

concept of self-efficacy can be rejected on the Christina Lee. Journal of Behavior Therapy and Experi-
mental Psychiatry, 2/, 141-142.
grounds of parsimony. Where alternative ex- Devins. G. M., & Edwards, P. J. (1988). Self-efficacy and
planations hold, a hypothesis involving a smoking reduction in chronic obstructive pulmonary
theoretical term not defined independently of disease. Brhaviour Research and Therapy, 26. 127-135.
Earley, P. C., & Lituchy, T. R. (1991). Delineating goal
the behavior giving rise to it is always less and efficacy effects: A test of three models. Journcrl of
satisfactory than a functional account which Applied Psychology, 76. X1-98.
involves no special concept at all. As Moore Festinger. I,. (1957). A Iheory o/ cognitive di.~soname.
Stanford: Stanford University Press.
(1984) has suggested, Kazdin, A. E. (1978). Conceptual and assessment issues
raised by self-efficacy theory. Advances in Behavrour
science is often hailed as the antidote to the mystical,
Research and Therapy. I. 177-185.
the occult and the supernatural. Some temporary
Kirsch, I. (19%). Early research on sell-efficacy: what wc
obstacles may arise when someone claims something to
already know without hnowing we knew. Journal of
be a cause, only to find out that the putative “cause”
.Soc,iul and (‘linkal P.~vcholog~, 4. 33Y-3.58.
isn’t really a “cause” after all, but these setbacks are
Kirsch. I. (1985). Self-efficacy and expectancy: old wine
only temporary (p. 80). with new labels. Journal of Personality and Social
Psychology, 4Y. 823-X30.
Kirsch, I. (1982). Efficacy expectations or response predic-
References tions: The meaning of efficacy ratings as a function of
task characteristics. Journal of Per.sonulity and Social
Bandura, A. (1977). Self-efficacy: toward a unifying Psychology, 42, 132-136.
theory of behavioral change. P.c&o/o~iccr/ Ket~ie~~, X4, Lee, C. (1990). Theoretical weaknesses: Fundamental
IYI-215. flaws in cognitive-behavioral theories arc more than a
Bandura, A. (lY80). Gauging the relationship between problem of probability. Journal of Behavior Therapy
self-efficacy. judgement and action. Cqqnirive 7’hercrp) and Experimenlal Psychiatry, 21, 143-145.
and Research, 4. 263-268. Lee, C. (1989). Theoretical weaknesses lead to practical
Bandura. A. (1982a). Self-efficacy mechanism in human problems: the example of self-efficacy theory. Journal of
agency. American P.~ychologist. 37, 122-147. Rehavior Therapy and Experimental P.\ychlarry. 20. I lS-
Bandura, A. (1982b). The assessment and predictive 123.
generality of self percepts of efficacy. Journal of Be- Moore. J. (IYX4). On behaviorism, knowledge. and causal
havior Therapy and Experimental Psvchiatry, I.?. 19S explanation. The Psychological Record, .14. 73-97.
199. O’Leary, A. (1985). Self-efficacy and health. Behaviour
Bandura, A. (19X6). Social Joundarions o/rhought: a social Research and Therapy. 23. 337-45 1.
cognitive theory. Englewood Cliffs, New Jersey: Prentice Poser. E. G. (lY78). The self-efficacy concept: Some
Hall. theoretical. procedural and clinical implications. Ad-
Bandura. A. (IYXY). Human agency in social cognitive t’ances irl Behaviour Research und Therapy, I, IY.%202.
theory. American Psycho1o,qis/, 44. 1175-I 1x1. Roper, Cr.. Rachman, S.. and Marks, I. (lY7S). Passive
Bcm, D. J. (1970). BeliejS. arrirudev and hrrmar~ af;lNfr.s. and participant modelling in exposure treatment of
Belmont: Brooks/Cole. obseasivc-compulsive neurotics. Behavlour Research and
Borkovcc. T. D. (lY7X). Self-efficacy: cause or rellcction Therupy. 13, 271-279.
of behavioral change. Ad~~c~rrccs in Behat?orrr Revetrrch Rotter. J. B. (1966). Gcncralized expectancies for internal
aud Theralq 1, 163-170. versus external control of reinforcement. Psychological
Corcoran. K. J. (19Yl). Efficacy. “skills.” reinforccmcnt Monographs. X0 (Whole No. 609).
and choice behavior. American Psychologkr. February, Ryder, R. G. (19X7). The realistic therapist: modesiy and
ISS-1.57. relativism in therapy and re.pearch. Newbury Park: Sage.
Corrigan. P. W. (1990). Theoretical weaknes\ in behavior Skinner. B. F. (1953). Science and human behavior. New
theory is no more than statistical variance: response to York: Macmillan.

You might also like