The Behavior Analyst 1987, 10, 1-15 No. 1 (Spring)

A Behavior-Analytic Critique of Bandura'sSelf-Efficacy Theory

Anthony BiglanOregon Research Institute

A behavior-analytic critique of self-efficacy theory is presented. Self-efficacy theory asserts that efficacyexpectations determine approach behavior and physiological arousal ofphobics as well as numerous otherclinically important behaviors. Evidence which is purported to support this assertion is reviewed. Theevidence consists ofcorrelations between self-efficacy ratings and other behaviors. Such response-responserelationships do not unequivocally establish that one response causes another. A behavior-analytic al-ternative to self-efficacy theory explains these relationships in terms ofenvironmental events. Correlationsbetween self-efficacy rating behavior and other behavior may be due to the contingencies ofreinforcementthat establish a correspondence between such verbal predictions and the behavior to which they refer.Such a behavior-analytic account does not deny any of the empirical relationships presented in supportof self-efficacy theory, but it points to environmental variables that could account for those relationshipsand that could be manipulated in the interest of developing more effective treatment procedures.Key words: Self-efficacy expectations, correspondence, behavior analysis, rule-governed behavior, pho-

bias, private events

This paper presents a critique of self-efficacy theory from a behavior-analyticstandpoint. Such an account of the phe-nomena that self-efficacy theory address-es is appropriate for at least two reasons.First, the role the environment plays indetermining behavior that is said to re-quire cognitive explanations needs ex-plicating. Self-efficacy theory is an ex-emplar of a class of behaviorally-basedtheories that explain behavior in termsofcognitive constructs and that explicitlydeemphasize the role ofthe environmentin determining behavior. The theory hasbeen applied to diverse aspects ofhumanbehavior, including phobias (Bandura,1982a), smoking cessation (Condiotte &Lichtenstein, 1981), smoking prevention(McAlister, Perry, & Maccoby, 1979), so-

Preparation of this paper was supported in partby grants HD15825 and HD13409 from the Na-tional Institute of Child Health and Human De-velopment, and CA38273 from the National Can-cer Institute. The views expressed are those of theauthor and are not intended to reflect those of anyof these agencies. The author would like to thankDonald Baer, Wesley Becker, James Coyne, Mi-chael Dow, Hyman Hops, Edward Lichtenstein, andScott McConnell for their helpful comments on ear-lier versions of the manuscript. Special thanks toLisa James for her editorial assistance. Requests forreprints should be sent to Anthony Biglan, OregonResearch Institute, 1899 Willamette Street, Suite2, Eugene, Oregon 97401.

cial skills (Kazdin, 1979), arithmetic skills(Bandura & Schunk, 1981), depression(Zeiss, Lewinsohn, & Munoz, 1979), andpain tolerance (Glasgow et al., 1982).Clearly, many investigators believe thatthe theory provides a useful account ofhuman behavior. Such an approach hasprobably been preferred to behavioranalysis because of the widely-held be-liefs that behavior analysis does not en-compass private behavior (Zettle &Hayes, 1982) and that it is impossible,in principle, to account for human be-havior without cognitive constructs(Bandura, 1969, 1977a, 1977b). Thepresent paper provides a counterargu-ment to these beliefs by explicating therole of the environment in determining(a) the behaviors that self-efficacy expec-tations are said to explain, (b) the behav-iors (both overt and private) that are saidto constitute self-efficacy expectations,and (c) the correspondence betweenstatements and ratings ofexpectation andsubsequent behavior.A second reason for the present paper

is to develop further the behavior-ana-lytic account of socially significant clin-ical behaviors. Behavior analysts havebeen disinclined to discuss complex hu-man behavior unless they have empiricalevidence in keeping with accepted meth-odological criteria (Marr, 1984). This

1

2 ANTHONY BIGLAN

disinclination is unfortunate because be-havior analysis, when applied, has madeimportant contributions to the amelio-ration of important problems (Kazdin,1979). A behavior-analytic account ofthetreatment of complex human behaviorcould be similarly fruitful, yet in largepart has not been undertaken.The paper is divided into four sections.

In the first, the self-efficacy construct isdescribed. The second section reviews theevidence that has been presented in sup-port of self-efficacy theory. In the thirdsection, a behavior-analytic account ofthe same phenomena is presented. Thepaper concludes by discussing the im-portance of returning to an emphasis onthe manipulation ofenvironmental vari-ables in clinical research.

THE SELF-EFFICACY CONSTRUCTThe concept ofself-efficacy is designed

to account for the effects of differentmodes of clinical treatment (Bandura,1977a). Self-efficacy is defined as "theconviction that one can successfully ex-ecute the behavior required to producethe outcomes" (Bandura, 1977a, p. 193).According to this view, any interventionshown to affect a target problem, such asphobic behavior, has its effects becauseit increases people's perceptions of self-efficacy. Efficacy expectations are said todetermine subsequent behavior, both interms ofthe initiation and the persistenceof coping behavior (Bandura, 1977a).Self-efficacy influences a person's choiceof activities and behavioral settings and"can affect coping efforts." Efficacy ex-pectations "determine" how much effortand persistence people show. Self-effi-cacy, however, is not the sole determi-nant of behavior; the person must alsohave "appropriate skills and adequate in-centives" (Bandura, 1 977a).The factors that are said to increase

self-efficacy are precisely the ones thathave been shown to affect outcome inclinical treatment research. Those factorsinclude performance accomplishmentssuch as those produced by participantmodeling and in vivo desensitization. Vi-carious experiences, in which the client

observes a live model or imagines a mod-el, are said to increase efficacy expecta-tions. Verbal persuasion, such as thatinvolved in self-instruction and psycho-therapy, also has some effect on self-ef-ficacy. Finally, techniques that reduceemotional arousal are said to affect self-efficacy. These include imaginal or "sym-bolic" desensitization.

EVIDENCE RELEVANT TOSELF-EFFICACY THEORY

Congruence Between Self-EfficacyRatings and Subsequent BehaviorThe evidence most frequently offered

in support of the causal status of self-efficacy is that self-efficacy ratings are re-lated to subsequent behavior. Consider-able evidence of this sort exists, althoughit is not as consistent or strong as hassometimes been asserted. Self-efficacyratings have been shown to predict post-treatment approach behavior of phobicsregardless of the treatment proceduresthat were used (Bandura & Adams, 1977;Bandura, Adams, & Beyer, 1977; Ban-dura et al., 1980); ratings have beenshown to predict behavior, whether theassessments were done in the midst oftreatment or at the end (Bandura &Adams, 1977); and the ratings predict be-havior toward stimuli other than thosefor which treatment was conducted (Ban-dura et al., 1977). As for target behaviors,self-efficacy ratings have been shown topredict behavior for agoraphobics whowere receiving treatment (Bandura et al.,1980), relapse among participants insmoking cessation programs (Baer, Holt,& Lichtenstein, 1986), the assertivenessof people receiving treatment (Kazdin,1979), physical stamina (Weinberg, Yu-kelson, & Jackson, 1980), career choice(Betz & Hackett, 1981), and exercise(Bandura & Cervone, 1983).A number of recent papers have cast

doubt on the assertion that self-efficacyratings are as highly or consistently re-lated to subsequent behavior as initialreports suggested. Kirsch (1980; Kirsch& Wickless, 1983) has shown that the"microanalytic" method used to assesscongruence between self-efficacy ratings

SELF-EFFICACY THEORY 3

and subsequent behavior in many studies(e.g., Bandura & Adams, 1977; Banduraet al., 1977, 1980) produces spuriouslyhigh estimates of congruence. The meth-od involves assessing the degree to whicha subject's behavior on a behavioralavoidance test is consistent with the sub-ject's indications of self-efficacy for en-gaging in those same behaviors. Kirsch(1980) points out that, since both ofthesescales are Guttman scales, subjects willtypically have only one point on eachscale at which they go from endorsing orcompleting the item to not endorsing orcompleting it. For example, people mightsay that their efficacy was high for com-pleting the first 7 of 10 steps in an ap-proach test, and they might actually com-plete 8 of those steps. The microanalytictechnique involves examining the pro-portion of these items on which efficacyagrees with actual behavior. In the ex-ample given, the person would have acongruence of .90, since efficacy wouldagree for steps 1 through 7 (where thesubjects both said they could do the stepsand did them) and on steps 9 and 10(where the subjects both said that theycould not complete the items and failedto complete them). Item agreement is en-tirely a function ofthe point on each scaleat which the person changes from en-dorsing or completing the item to notendorsing or completing the item. As-sessing agreement on each of the pointsof the scale is superfluous and suggestsgreater congruence than is appropriate.Not surprisingly, studies using this meth-od of assessing congruence have oftenyielded estimates ofcongruence of .80 ormore (cf. Bandura & Adams, 1977). Moreappropriate correlational analyses typi-cally account for 30% or less of the vari-ance in subsequent performance. Thus,on empirical grounds alone, there are rea-sons to doubt the causal role of the self-efficacy construct.The causal role of self-efficacy is also

said to be shown by the fact that self-efficacy ratings are more highly related tosubsequent behavior than are samples ofprevious behavior. For example, Ban-dura et al. (1980) found that agorapho-bics' performance on a post-treatment

approach test was predicted better by self-efficacy ratings than by performance onan approach test during treatment. Sim-ilar results have been found in other stud-ies (Bandura et al., 1977). Some recentstudies, however, have failed to show sucha relationship. Baer et al. (1986), for in-stance, found that self-efficacy ratings didnot predict later smoking behavior whenthe variance associated with currentsmoking level was taken out. Lane andBorkovec (1984) found that changes inself-efficacy ratings predicted self-report-ed anxiety, but not physiological and be-havioral measures.

The Relative Effectivenessof TreatmentsTreatment procedures that differ in ef-

fectiveness also differ in the degree towhich they affect self-efficacy ratings. Thisevidence has been offered in support ofthe causal role of self-efficacy. Banduraet al. (1977) showed that participantmodeling produced higher self-efficacyratings and more approach behavior thanmodeling, which in turn produced higherself-efficacy ratings and superior ap-proach behavior than a no-treatmentcontrol. Bandura and Adams (1977)treated 10 highly phobic subjects withsystematic desensitization, and then pro-vided supplementary participant mod-eling to the nine subjects who had notbeen completely successful in the desen-sitization. The participant modelingboosted self-efficacy ofthese subjects and,at the same time, increased their ap-proach behavior.

Self-Efficacy and ArousalThe value of the self-efficacy construct

has also been advocated on the basis ofevidence that purports to show that self-efficacy causes arousal (Bandura, 1982a).One type of evidence offered in supportof this assertion is that ratings of self-efficacy correlate with ratings of antici-pated fear and of fear during approachtests (cf. Bandura et al., 1980). This evi-dence has been criticized because no di-rect physiological measures of arousalwere obtained. In response to this con-

4 ANTHONY BIGLAN

cern, Bandura, Reese, and Adams (1982)assessed the correspondence between self-efficacy ratings and blood pressure andcardiac acceleration. They manipulatedthe strength of self-efficacy ratings forspecific approach tasks through "enac-tive mastery." The ratings were associ-ated with blood pressure and cardiac ac-celeration in anticipation and duringperformance ofthe approach tasks. Whenfurther treatment produced maximumratings of self-efficacy strength for alltasks, physiological reactions to thesetasks were reduced. In a similar study,Bandura et al. (1985) treated spider pho-bics with modeling until each subject hadan approach task for which they ratedtheir self-efficacy as being strong, a taskfor which they rated their self-efficacy asbeing moderate, and one for which theylabeled their self-efficacy as being weak.Subjects' levels of epinephrine and nor-epinephrine during each approach taskwere predictable from these self-efficacyratings.

A BEHAVIOR-ANALYTICALTERNATIVE

Response-Response Relationshipsand Causation

Evidence presented in favor of self-ef-ficacy's causal status necessarily involvescorrelations between ratings of self-effi-cacy and other aspects of behavior, in-cluding physiological responses (Hayes &Brownstein, 1986; Moore, 1984a, 1984b).Some of the evidence appears to involveexperimental manipulations of self-effi-cacy, however, and thus seems to escapethe charge that the evidence is only cor-relational. For example, some studiesmanipulate self-efficacy levels throughtreatment and then assess the relation-ship between self-efficacy and overt be-havior (cf. Bandura et al., 1982). Otherstudies appear to show that manipula-tions of self-efficacy ratings produce cor-responding changes in rated physiologi-cal arousal (Bandura et al., 1982, 1985).In all these studies, a sequence of eventsis presented as follows: (a) self-efficacyratings are experimentally manipulatedthrough treatment; (b) when specific

strengths of self-efficacy are achieved forspecific approach tasks, behavior orarousal is evaluated; (c) self-efficacy rat-ings are related to behavior or arousal.Changes in behavior and arousal are con-cluded to be due to changes in self-effi-cacy.The attribution of the mediating role

of self-efficacy is an arbitrary interpre-tation of this sequence of events. The se-quence can alternatively be viewed as in-volving the effects of treatmentmanipulations on both self-efficacy rat-ings and other behavior. The relationshipbetween self-efficacy ratings and otherbehavior is inevitably correlational, be-cause both are responses of the same or-ganism (Moore, 1984a, 1984b). Any pro-cedure that affects self-efficacy is aprocedure applied to the same organismthat exhibits overt behavior and re-sponds physiologically. Thus, when en-vironmental variables are manipulatedin order to affect self-efficacy ratings, theenvironment that affects these other as-pects of behavior is also being manipu-lated.Construing this sequence of events as

a process whereby self-efficacy affectsovert behavior or physiological arousalsuggests a preference among self-efficacytheorists for the attribution ofcausal sta-tus to response variables. The preferenceamong behavior analysts is to include inthe account a description ofthe environ-mental variables that determine bothtypes of responses and any relationshipsamong them (Biglan & Kass, 1977;Moore, 1984a, 1984b; Skinner, 1953,1945, 1974; Zettle & Hayes, 1982). Hayesand Brownstein (1986) have argued thatneither view is inherently correct, but thatthe two approaches reflect different goalsand different criteria for the truth ofstate-ments. The behavior-analytic view is acontextualist "world view" that empha-sizes prediction and control as the criteriafor the truth of statements in its system;explanation is considered to be achievedto the extent that statements lead to suc-cessful prediction and control. Self-effi-cacy theory reflects a world view that ac-cepts as true statements that generateaccurate predictions in situations other

SELF-EFFICACY THEORY 5

than ones in which the statements werederived. A phenomenon is taken to beexplained to the extent that it corre-sponds to a set of theoretical statements.Thus, self-efficacy and behavior-analyticapproaches to response-response rela-tionships may be seen as stemming fromdifferent world views.Whether the behavior-analytic ap-

proach to such problems is ultimatelypreferable is a matter of how one eval-uates prediction and control as criteriafor the truth of statements. Hayes andBrownstein (1986) point out that the em-phasis on prediction and control leads tothe specification of environmental vari-ables that account for all responses(whether verbal or nonverbal, private orobservable) and for the relationships be-tween responses. The following sectionsprovide just such an account for thephenomena addressed by self-efficacytheory. It takes nothing away from theself-efficacy account, but it leads to thespecification of variables outside the or-ganism which can, at least in principle,be manipulated. The account is specu-lative. But, to the extent that it delineatesexternal variables that affect both mea-sures ofthe self-efficacy construct and be-haviors that are correlated with self-ef-ficacy measures, then it may lead to moreeffective clinical interventions.

Self-Efficacy and Verbal BehaviorRatings, statements, and thoughts of

self-efficacy may be seen as forms ofver-bal behavior in which people predict oth-er behavior. Rather than assuming thatsuch behavior reflects an underlyingcause, behavior analysts see the environ-ment as the critical factor in the deter-mination of ratings, thoughts, and state-ments of efficacy, as well as overt verbaland nonverbal behavior and physiolog-ical responses. One type ofresponse (e.g.,a thought about one's efficacy) can func-tion as a discriminative stimulus or aconsequence of another type ofbehavior(e.g., approaching a snake), but suchfunctional relationships are assumed tobe a product of environmental events.Discriminative control is not assumed to

be exercised only by thoughts, state-ments, or ratings of self-efficacy; any or-ganismic event can come to function asa discriminative stimulus or a reinforcerfor another organismic event, providedthat reinforcement follows the contingentoccurrence of the two events, at least in-termittently.A response class of saying and doing.

People experience numerous instances inwhich their statements about what theyare doing, have done, or are about to doare followed by consequences. Typically,the consequences are based, at least inpart, on the correspondence between thedescription of behavior and the actualbehavior. For example, a person whopromises to do something for a friendmay be thanked or praised for makingthe promise, but praise may turn to dis-approval if the promise is not subse-quently fulfilled. Not all of the relevantconsequences are social. People who canaccurately predict how long it will takethem to complete a task may be rein-forced when the accuracy of their pre-dictions allows them to plan their timeefficiently and/or to avoid "having toolittle time" to finish a task. People alsohave a history of reinforcement for ac-curately reporting what they have done.Verbal descriptions of past behaviorachieve reinforcements from others, suchas attention and praise, and are rein-forced when they facilitate the individ-ual's later behavior, for example, whenpeople remember where they left off ona task they worked on earlier. Inaccuratereports of past behavior are branded aslies and are otherwise punished.What likely emerges from such a his-

tory is a response class ofsaying and doingin which verbal descriptions ofbehaviorand the behavior described correspondclosely with each other. The occurrenceofa verbal description ofbehavior to oc-cur in the future (e.g., promises and state-ments ofintent) increases the probabilityof the subsequent behavior. By the sametoken, the occurrence of overt behaviormakes it more likely that subsequent ver-bal descriptions of that behavior will oc-cur (e.g., recollections). Evidence has beenprovided in a number of studies that this

6 ANTHONY BIGLAN

type of correspondence is brought aboutby its consequences (Israel & Brown,1977; Paniagua & Baer, 1982; Risley &Hart, 1968; Whitman et al., 1982). Theinvolvement of such a process in muchbehavior in the natural environment issuggested by studies showing that, oncecorrespondence is established, nonverbalbehavior can be maintained by reinforc-ing reports of the behavior, even whenthose reports are sometimes inaccurate(Baer et al., 1984; Whitman et al., 1982).The analysis ofresponse classes ofsay-

ing and doing is applicable to the behav-ior of people who fear specific objects.For these people, verbal statements abouttheir phobic behavior are probably mem-bers of a response class of predictivestatements and associated actions re-garding the object. Subjects in self-effi-cacy experiments have had numerous ex-periences in which their descriptions andpredictions offearful behavior have beenreinforced. People who say, "I can't gocamping, I'm too afraid ofsnakes," avoidthe aversive situation. If such reports areinconsistent with their past behavior, thatinconsistency is likely to be pointed outin aversive ways. Ifthey subsequently en-gage in behavior that is inconsistent withtheir stated fear, the inconsistency maybe noted in ways that are aversive, andthey may be expected to begin to do thingsthat they have heretofore avoided. Theymay also achieve the more generalizedreinforcement of"being right" when theiravoidance behavior matches their pre-dictions of fearful behavior (Hayes et al.,1985). Avoidance of phobic stimulicould reinforce predictive statements,even those statements that no one elseheard. For example, the behavior ofthinking about one's fear ofbridges couldbe maintained if it led to a person avoid-ing crossing bridges.

Stimulus control ofthe response class.The stimuli that control this responseclass of predictions and nonverbal be-havior appear to include external stim-ulus situations, such as the presence of aphobic stimulus and stimuli that havebeen paired with the phobic stimulus,private events such as physiologicalarousal and thoughts of the object, and

questions regarding likely future behav-ior. Thus, people's reports that-they willbe fearful on airplanes may be under thestimulus control of the verbal stimulus,"Why don't you fly to Arizona for thewinter?" as well as the physiologicalarousal evoked by that stimulus.When the controlling stimuli are ac-

cessible to others, the process by whichsuch stimulus control is established andmaintained can readily be imagined. Achild cries and says, "I won't go in theyard," as a dog barks at him from behinda fence. Adults may reinforce both theverbal and the avoidance behavior bycomforting the child and allowing him toavoid the dog. In this way, the presenceof yards with big dogs is established as acontrolling stimulus for verbal responsessuch as "I won't go in the yard," or "I'mafraid," and the corresponding nonver-bal responses.The processes by which private stim-

ulation, such as a pounding heart, comesto control predictions offearful behaviorand associated behavior are more diffi-cult to understand. Others do not havedirect access to private events, yet whenpeople report their fears they are to someextent responding to events that take placewithin them. Skinner (1945, 1957) hasidentified four processes through whichprivate events such as a pounding heartor covert speech could come to be a dis-criminative stimulus for a verbal re-sponse, such as "I am anxious." First,others may observe public events thataccompany the private event. In the ex-ample just given, the presence of thebarking dog may accompany the phys-iological arousal that is occurring whenthe child says, "I'm afraid to go near thatdog." Thus, when others reinforce thestatement, "I can't go near that dog," theydo so in the presence of physiologicalevents. On future occasions, the arousal,by itself, may be a stimulus for the verbalresponse, "I can't do that." Second, oth-ers may reinforce verbal reports of fearon the basis of the observable behaviorofthe person. For example, ifa child criesand cowers in the presence of the dog,the response, "I'm scared," will be rein-forced on the basis of the observable be-

SELF-EFFICACY THEORY 7

havior. In the future, this response maybe more likely on occasions when onlythe private physiological events that orig-inally accompanied the observable be-havior are present. Third, people maycome to respond to their own privatestimulation through metaphorical exten-sion of responses to observable stimuli.For example, people may describe them-selves as "agitated" because their movingaround a great deal resembles the con-dition of external stimuli in the presenceofwhich the response "agitated" has beenreinforced. The private stimulation thataccompanies this moving around maythen come to control the response.Fourth, and finally, people's descriptionsofthemselves may be under the stimuluscontrol of their covert behavior, such ascovert verbal statements. For example,people may say that they are fearful ofairplanes in response to covert verbalstatements, like, "The plane is going tocrash." Such stimulus control is estab-lished when the people are reinforced forlabelling themselves as fearful in thepresence ofovert verbal statements ofthesame type.The temporal gap between controlling

stimuli andpredictions. One objection tothe analysis presented here might be thatthe length of time between people's ex-periences with phobic objects and theirsubsequent predictions, and the length oftime between predictions and subsequentbehavior, make it necessary to posit acognitive construct like self-efficacy.Bandura (1977b) has made such an ar-gument with respect to imitative behav-ior: Cognitive processes are said to berequired to account for the fact that mod-eled behavior can be imitated long afterit has been observed. Deguchi (1984) hasargued that there is no reason why grad-ual lengthening of the gap between themodel and the imitation could not occur,such that reinforcement occurs afterlonger and longer delays. Descriptions ofwhat a person will do may initially occurin the presence of the variables control-ling the response and perhaps the incip-ient behavior. The temporal gap betweenthe stimuli that control the prediction andthe prediction could become longer, as

long as reinforcement were achieved. Theidea that the time between predictionsand later behavior could be lengthenedis suggested by studies in which reportsof behavior were shown to control be-havior 22 hours later (cf. Risley & Hart,1968).Summary. From a behavior-analytic

standpoint, self-efficacy ratings arethought ofas one type ofverbal behavior.Although subjects may not previouslyhave been reinforced for making ratingsof their self-efficacy, they have made nu-merous predictions about their phobic(and nonphobic) behavior, and the con-sequences of those predictions have es-tablished a response class that includesboth the statements of what they will dowith respect to the phobic object and whatthey actually do. Ratings and other pre-dictive statements are probably under thestimulus control of both external stimuliand private events such as physiologicalresponses, covert speech, and perceptualbehavior. At the same time, the conse-quences for correspondence between ver-bal behavior about approach and avoid-ance offeared stimuli and the subsequentbehavior make it likely that people willbehave in a manner consistent with theirpredictions of approach or avoidance.

The Precision ofBehavior AnalysisBehavior analysis makes it possible to

be precise about when specific covertevents are thought to have occurred. Big-lan and Kass (1977) noted that cognitiveconstructs typically fail to distinguish be-tween events that occur within the or-ganism at a given time and place, andhypothetical constructs that go beyondan inference that an event has taken place.People may be said to have high self-efficacy for springboard diving, either be-cause they report thinking things like, "Ican do a jack-knife," or because they areseen to perform such a dive. A conceptsuch as self-efficacy may be appealing inpart because people think specificallyfrom time to time that they can do things.If people typically behave without en-gaging in any covert verbal behavior, itgoes unnoticed. Distinguishing between

8 ANTHONY BIGLAN

instances when such thoughts occur andwhen they do not reduces the tendencyto assume that all behavior can be under-stood as resulting from such cognitions.Behavior analysis distinguishes among

ratings of self-efficacy, covert verbalstatements about self-efficacy, and overtverbal statements about self-efficacy.When people rate themselves as high inself-efficacy, they may or may not be re-porting on specific verbal behavior inwhich they have engaged regarding theirefficacy. For example, people who ratethemselves as unable to touch a snakemay or may not actually think to them-selves that they will be unable to touchthe snake at some point prior to makingthe rating.These distinctions are important for at

least two reasons. First, they may promptgreater precision in both theory and re-search about when ratings are actuallyunder the stimulus control of covert andovert verbal behavior. Such precisioncould improve our ability to change im-portant aspects of clients' verbal behav-ior. Second, clarification that self-efficacyratings are not necessarily under thestimulus control of verbal statementspoints to the control that other eventssuch as physiological responses and en-vironmental events have over these rat-ings. Bringing these events into the ac-count makes it more likely that they willbe used in efforts to change self-efficacyratings and, more importantly, to changeavoidance behavior.

Reinterpretation ofthe Evidencefor Self-Efficacy Theory

Correspondence between efficacy rat-ings and later behavior. From a behavior-analytic standpoint, the correspondencebetween self-efficacy ratings and later ap-proach behavior is due to the contingen-cies that have placed a client's predic-tions about phobic behavior and actualphobic behavior into the same responseclass. Such correspondence does not in-dicate that the ratings play a causal role,or that an underlying causal entity is as-sessed by the ratings. People who touchsnakes during participant modeling will

be more likely to say that they will touchsnakes in the future, and will be morelikely to touch them to the extent thatthey are reinforced by events such as low-ered arousal and feedback from the ther-apist that they have succeeded. The con-tingencies involved in treatment modifyboth types of behavior.The fact that, for phobic behavior, self-

efficacy ratings have sometimes beenfound to predict later approach behaviorbetter than prior behavior should not besurprising. Both ratings and later behav-ior are functions of the consequences ofprior behavior; they are not simply func-tions ofthe prior behavior. Thus, severalclients might engage in exactly the sameapproach behavior, but have differentconsequences for that behavior. For ex-ample, one might become physiological-ly aroused and another might not becomearoused. ' This difference in conse-quences would be expected to producediffering levels of both self-efficacy rat-ings and of subsequent approach behav-ior. Under these circumstances, self-ef-ficacy ratings would be more highlycorrelated with later approach behaviorthan would prior approach behavior, butthat correlation would not imply that self-efficacy determines later behavior.Although physiological arousal is not

the only consequence of approach be-havior, it may be important when ac-counting for the fact that self-efficacy rat-ings predict subsequent approachbehavior better than does prior approachbehavior. In some studies, approach be-havior was uniformly high across sub-jects at the end of treatment, but vari-ability in self-efficacy ratings remainedand the ratings were correlated with per-formance on a subsequent approach test(e.g., Bandura et al., 1977). Variability inarousal, however, may have remained.Unfortunately, arousal was not assessed.Such differences in arousal could havefunctioned as controlling stimuli for self-efficacy ratings and as consequences thataffected later approach behavior. Thus,subjects who picked up the snake andwere aroused would presumably be lesslikely to engage in such approach behav-ior in the subsequent test, and less likely

SELF-EFFICACY THEORY 9

to indicate that they felt efficacious, thanwould those who picked up the snake anddid not become aroused.

In studies where congruence was found,subjects reported no arousal to imaginedphobic stimuli (although here, too,arousal was not directly assessed) (e.g.,Bandura & Adams, 1977). In these stud-ies, variability among subjects in past ap-proach behavior and the consequences ofthat behavior presumably remained, sincethe approach behavior of these subjectswas not directly modified during treat-ment. This remaining variability in thecontingencies could account for the vari-ability in both self-efficacy ratings andlater approach behavior. In sum, whileself-efficacy ratings have been shown tocovary with subsequent behavior in a va-riety ofcircumstances, uncontrolled vari-ables probably account for the variabilityin both ratings and in later behavior.Thus, although the data are consistentwith the assertion that self-efficacy playsa causal role, they by no means entail thatconclusion.

Self-efficacy ratings as controllingstimuli for later behavior. A number ofwriters have argued that the act of mak-ing a rating affects subsequent behavior(Borkovec, 1978; Eysenck, 1978; Kaz-din, 1979; Lang, 1978), although nonehas discussed the relationship in termsof stimulus control. A series of studieshave purported to show that making rat-ings does not affect subsequent behavior.Bandura et al. (1980), for instance, ran-domly assigned six of twelve subjects toa condition in which self-efficacy ratingswere not made. Because the mean per-formance of these subjects did not differfrom that for six subjects who did ratetheir self-efficacy, the authors concludedthat making self-efficacy ratings does notaffect subsequent behavior. The study isirrelevant to the present issue. Evaluatingthe effects ofmaking ratings on the meanbehavior of subjects has nothing to dowith congruence. Having made self-effi-cacy ratings -on whatever level -the in-dividual will be more likely to performclose to that predicted level due to pastsocial consequences for such correspon-dence. This does not imply that making

self-efficacy ratings necessarily raisessubjects' subsequent performance- orlowers it. It only implies that, havingmade a rating, the subject will tend tobehave in a manner consistent with thatrating.

Telch et al. (1982) purported to testwhether "social demand for consistency"affects the congruence between self-effi-cacy ratings and approach test perfor-mance. In a low demand condition, sub-jects were led to believe that their self-efficacy ratings would not be seen by theexperimenter, while in a high demandcondition, subjects were told that theirratings would be reviewed and that the"ratings would be used to validate pro-gress in the mastery of fears" (Telch etal., 1982). Prior to treatment, the con-gruence between self-efficacy and perfor-mance was significantly higher for thelow-demand condition than it was for thehigh-demand condition. Following treat-ment, congruence was similarly high inboth conditions, and had apparently in-creased since pretreatment. The authorsconcluded that the congruence betweenself-efficacy and performance could notbe due to social demand for consistency,since the high-demand condition actuallyproduced lower congruence at the outset.

This study is not relevant to evaluatingthe hypothesis that the act of making arating is a controlling stimulus for laterbehavior. It did not really manipulate so-cial demands for consistency. The ma-nipulation of a social demand for con-sistency might have been achieved ifsomesubjects had been told something like:

Self-efficacy ratings turn out to be surprisingly poorpredictors ofhow you will perform. We are askingyou to fill this out to demonstrate to you how in-accurate your beliefs about your performance ca-pabilities are.

Under these circumstances, subjectsmight have lower congruence betweenratings and subsequent approach behav-ior because of the implication that rein-forcement would not be contingent ontheir acting in correspondence with whatthey had said. The Telch et al. (1982)study manipulated the salience ofthe rat-ings as an evaluation of the subjects.

10 ANTHONY BIGLAN

Those in the high-demand condition weretold that their ratings would be used to"evaluate progress." Subjects' motiva-tion to show progress in their ratings wasthus enhanced. The lower congruence atpretreatment for this group was probablydue to subjects' systematically underes-timating their approach behavior to meetthe demand to show "progress." Telch etal. (1982) indicated that subjects in thiscondition did indeed underestimate theirability to perform compared to othersubjects-a result that was not predictedby the authors.

Telch et al. (1982) argued that, if self-efficacy ratings predict behavior evenwhen the ratings are made in private (theirlow-demand condition), then the act ofmaking the rating could not affect sub-sequent behavior. At first blush, that ar-gument appears consistent with the find-ings of Hayes and his colleagues, whohave shown that, for a variety of thera-peutic procedures, the instructions,models, or coping statements given tosubjects are effective only when subjectsare aware that the therapist has access tothese events (Hayes et al., 1985; Nelsonet al., 1983; Rosenfarb & Hayes, 1984;Zettle & Hayes, 1983). Ifdifferential con-sequences occur for correspondence be-tween predictions about behavior andlater behavior, some congruence betweenself-efficacy ratings and later behavior canbe expected, even when ratings are madein complete privacy. Congruence shouldbe higher when others have access to somemembers of the response class of predic-tive behavior. Thus, a person's interac-tions with a therapist during treatmentwould be expected to raise congruence,since treatment consists of instanceswhere the person is instructed to engagein various approach behaviors, agrees todo so, engages in the behavior, and isreinforced. This treatment sequencewould account for why congruence in theTelch et al. (1982) study is higher aftertreatment than before -even when theself-efficacy ratings are made in privateon both occasions: Treatment strength-ened the correspondence between pre-dictions and actual behavior.

Further support for this position is

provided by the study by Bandura andCervone (1983), in which subjects in thecontrol condition received no feedbackregarding their performance and did notset goals. These subjects were not ex-posed to contingencies for matching theirself-efficacy ratings (or any other predic-tive statements) with their behavior. Inaddition, the task was sufficiently ambig-uous to make it unlikely that experiencewith the task itself could have enhancedthe match between predictions and be-havior. Under these conditions, self-ef-ficacy ratings did not correspond to sub-sequent behavior.Gauthier and Ladouceur (1981) com-

pared congruence between self-efficacyratings and approach behavior betweena group of subjects who completed theratings privately and a group of subjectswho publicly stated their ratings to theassessor. Congruence did not differ be-tween the groups. In this case, "private"apparently meant filling the ratings outand leaving them for the assessor. Theact of making the rating may not havebeen effectively private, since subjectscould presumably discern that their rat-ings would be seen by the experimenter.Under these circumstances, the very actof indicating a particular level of efficacyis a discriminative stimulus for later be-havior. (These authors also claimed thatthe fact that the two groups did not differin mean performance undermined thehypothesis that making ratings affects lat-er performance. As discussed above,mean group differences would not be ex-pected.)

Relative effectiveness of treatments.From a behavior-analytic standpoint, therelative effectiveness of different treat-ments is not due to their relative effecton the self-efficacy construct, but to theirrelative effect on arousal and approachbehavior. Participant modeling is moreeffective than desensitization in modi-fying approach behavior because subjectsactually practice the approach behaviorsthat they will be asked to perform atpost-treatment assessment. Imaginal de-sensitization has a smaller effect on ap-proach behavior than does modeling be-cause subjects only imagine approach

SELF-EFFICACY THEORY 11

responses and do not practice them. Im-provements in the subsequent approachbehavior of subjects who received desen-sitization are likely due to a reduction inarousal and perhaps the covert approachrehearsal. Because arousal during imag-ination is probably one consequence thataffects later approach behavior, its re-duction increases the probability of ap-proach behavior. Desensitization mayalso increase the probability ofapproachbehavior ifpeople have a history ofbeingreinforced for imagining their behaviorand then performing accordingly. Nei-ther reducing arousal nor imagining ef-fective performance, however, are aspowerful in modifying approach behav-ior as actual practice of the behavior.Thus, these treatments differ in effec-tiveness because they differ in the degreeto which they modify targeted behavior.

Self-efficacy theorists take the corre-lation between change in self-efficacy andchange in approach behavior as evidencethat changes in the self-efficacy constructlead to changes in approach behavior.From the standpoint of behavior analy-sis, changes in self-efficacy ratings simplyreflect changes in this form of verbal be-havior, not some underlying cause of allphobic behavior. Why are changes in theratings related to changes in behavior?The answer: Because changes in self-ef-ficacy ratings are also effects ofthe changesin arousal and approach behavior thatare produced by these treatments. Sinceself-efficacy ratings are partly under thestimulus control of approach behaviorand arousal, treatments that directly af-fect approach behavior and arousal, suchas participant modeling, have a greatereffect on self-efficacy ratings than do oth-er treatments. This set ofresults does notdemonstrate that treatment effects occurbecause treatment affects self-efficacy-it simply demonstrates that self-efficacyratings are affected by these treatmentsin a manner similar to the way that thetreatments affect approach behavior.

Self-efficacy and arousal. The correla-tion of self-efficacy ratings with measuresofarousal does not demonstrate that self-efficacy causes arousal, any more thanself-efficacy's correlation with overt be-

havior entails that it determine overt be-havior. It simply shows that people's ver-bal predictions of what they can do arerelated to their arousal. From a behavior-analytic view, this correspondence isprobably due to two factors.

First, both self-efficacy and subsequentarousal are functions of prior experiencewith the phobic stimulus. The arousal aperson experiences would be expected tobe correlated with the arousal they ex-perienced on previous exposures to thephobic stimulus. Moreover, the arousalexperienced during the previous expo-sure is one controlling stimulus for sub-jects' self-efficacy ratings. Thus, one rea-son why self-efficacy ratings are correlatedwith later arousal is that both self-efficacyratings and the later arousal ratings arerelated to the arousal that was experi-enced on the previous occasion when theperson was in the presence of the phobicstimulus.

Second, people's history of reinforce-ment for correspondence between pre-dictions and later behavior make it likelythat arousal ratings (and perhaps actualarousal) will correspond to their efficacyratings. That is, after people have indi-cated that their efficacy expectations wereat any given level, later arousal wouldtend to be in line with those expectationsbecause of people's extensive history ofreinforcement for having predictions oftheir own behavior conform to later be-havior. People who are phobic have pre-sumably had numerous experienceswhere they said that they would be anx-ious if they did something. Failure to beanxious in the relevant circumstanceswould prompt others to question themin aversive ways.

THE ROLE OF THEENVIRONMENT

Deemphasis ofthe Environment inSelf-Efficacy Theory

To some, self-efficacy theory may beattractive precisely because it deempha-sizes the environment. According toBandura (1977b), the environmentalcontrol of behavior has been overem-

12 ANTHONY BIGLAN

phasized and cognitive determinants havebeen underemphasized:A valid criticism of extreme behaviorism is that,in a rigorous effort to avoid spurious inner causes,it has neglected determinants of behavior arisingfrom cognitive functioning. (1977b, p. 40)The third distinctive feature of social learning

theory is the central role it assigns to self-regulatoryprocesses. People are not simply reactors to externalinfluences. They select, organize, and transform thestimuli that impinge upon them. (Bandura, 1977b,p. vii)

Bandura's (1977a) original statement ofthe self-efficacy construct was prefacedby a" . . . reconceptualization ofhumanlearning and motivation in terms of cog-nitive processes" (p. 193). The effects ofmodels, reinforcements, and instructionsare seen as cognitively mediated. Thus,the construct is advocated in part becauseit focuses attention on cognitive deter-minants of behavior.Research conducted to test self-efficacy

theory illustrates the shift of investi-gators' attention away from environmen-tal determinants ofbehavior. In the stud-ies by Bandura and his colleagues onphobic behavior, the effects ofpreviouslyvalidated interventions on self-efficacyratings have been tested, and the abilityofthese ratings to predict subsequent be-havior have been examined. The focushas been on the self-efficacy construct.These studies replicated earlier work re-garding which treatment procedures af-fect phobic behavior, but they have testedno new ways in which the environmentthat therapists provide to phobics can af-fect phobic behavior. Moreover, otherinvestigators who have tested self-effi-cacy theory have not identified or testednew ways in which therapeutic proce-dures might affect clinical problems. Anumber of tests of self-efficacy as a pre-dictor ofsmoking relapse have been con-ducted (Baer et al., 1986; Condiotte &Lichtenstein, 1981; DiClemente, 1981),but no treatments uniquely derived fromself-efficacy theory have been proposed.Replication of the treatment of phobiasusing cognitive restructuring and guidedexposure, where the central focus is oncorrelations between self-efficacy ratingsand performance, have been evaluated

(Biran & Wilson, 1981). A good deal ofresearch has been conducted regardingthe effects ofconditions under which self-efficacy ratings are made on subsequentperformance (Bandura et al., 1980; Gau-thier & Ladouceur, 1981; Telch et al.,1982). In essence, the theory has prompt-ed investigators to show that already-validated treatments are related to self-efficacy, which is in turn related topost-treatment behavior. Rather thanpromoting new ways ofchanging clients'environments, the theory has directed at-tention toward providing post hoc ac-counts ofwhy previously validated treat-ment procedures are effective. Theprocedures themselves were derivedwithout benefit of self-efficacy theory.

The Importance ofthe EnvironmentA focus on the effects of the environ-

ment on behavior is critical. One can eas-ily lose sight ofthe extraordinary growthofour knowledge about human behaviorin the past twenty years. This growth isdue to the examination of the way inwhich environmental manipulations af-fect behavior. In clinical psychology andpsychiatry, treatment programs have beendeveloped that beneficially affect simplephobias (Rachman, 1978), agoraphobia(Gelder& Marks, 1966), depression (Big-lan & Dow, 1981), exhibitionism (Mal-etsky, 1980), stuttering (Brady, 1971), so-cial skills (Curran, 1977; Hersen &Bellack, 1976), sexual dysfunction (Lo-Piccolo & LoPiccolo, 1978), and maritaldiscord (Jacobson & Margolin, 1979). Forchildren, programs have been identifiedthat affect aggressive behavior (Patter-son, 1982), social isolation (Hops, 1982),and bedwetting (Christopherson & Ra-poff, 1980). Mentally retarded people arebeing taught to live independently (Shal-ock & Harper, 1978). Educational pro-grams have been developed that achievegreater improvements in intellectual skillsand in reading, mathematics, and lan-guage among disadvantaged children thanwas possible twenty years ago (Becker &Carnine, 1980).A conclusive demonstration that these

developments were due to a shift away

SELF-EFFICACY THEORY 13

from construct-oriented approaches andtoward analysis of the effects of the en-vironment is not possible. Concurrentwith the development of more effectiveinterventions, however, was a good dealof critical examination of earlier at-tempts to explain behavior in terms ofpersonality constructs (Mischel, 1968).Theoretical analysis of environment-be-havior relations received increasing at-tention at the same time that practicalprogress was being made (Kazdin, 1978b).Whatever the theoretical analyses mayhave been that prompted these develop-ments, treatment programs in the areasenumerated above involve the specifi-cation of particular behavioral targets(including private events) and the ma-nipulation of the environment. Evenwhen the effects of the intervention arepresumed to be cognitively mediated, thetreatment itself still consists of alter-ations of the person's environment (e.g.,discussions and instruction).

THE AS YET UNFULFILLEDPROMISE OF BEHAVIOR

ANALYSISA theoretical approach like behavior

analysis is an essential step in shiftingattention back to developing more effec-tive treatments. The approach explicateshow presumed cognitive mediators areaffected by the environment, and howtheir effect on other aspects of behaviorderives from the environment. Thus, theapproach demonstrates that it is possibleto examine how the environment affectsall aspects ofthe behavior ofpeople withclinical problems-including their men-tation (Biglan & Kass, 1977). If the anal-ysis is correct, it is only through the en-vironment that self-efficacy ratings andimportant clinical targets can be changed.Up to this point, the contributions of

behavior analysis to the understanding ofcomplex human behaviors such asdepression and phobias have been lim-ited. A number ofbehavior analysts havenoted that the approach has not fulfilledits promise (Skinner, 1953, 1957, 1969,1974) to increase the understanding andcontrol of important human behaviors

(Hops, 1985; Marr, 1984; Michael, 1984;Zettle & Hayes, 1982). Behavior analystshave, however, begun to address clinicalissues. For example, a series of studiesby Hayes and his colleagues shows thata variety of procedures often consideredto be examples of self-generated or cog-nitive control of behavior owe their ef-fects to the social environment (Hayes etal., 1985; Nelson et al., 1983; Rosenfarb& Hayes, 1984; Zettle & Hayes, 1983).At the same time, therapies that are saidto be cognitive have been explained interms of rule-governed behavior (Zettle& Hayes, 1982). Most important, newmethods of treatment have been derivedbased on analyses of rule-governed be-havior relevant to emotions and the con-tingencies controlling rule governance(Hayes, in press).The present paper is in keeping with

these developments. If the analysis pre-sented here has provided a plausible ac-count of self-efficacy findings in behav-ior-analytic terms, it may serve as acounter-argument to the prevailing em-phasis on response-response relation-ships. More effective methods oftreatingclinical problems will be achieved whenwe have a clearer account of the ways inwhich all behavior-public or private,and motor, physiological, or verbal-isaffected by the environment.

NOTES1. Recent discussion by Hayes (in press) on the role

of anxiety in avoidance behavior suggests thattreatments for anxiety disorders may have twodifferent effects. They may lower arousal andthey may change the stimulus control that arousalhas over overt behavior. Thus, people may be-come more willing to approach feared objectsfor two distinct, but not mutually exclusive rea-sons: (a) they experience less physiological arousalin the presence of the object and are thus notstimulated to avoid it and (b) they experiencearousal but, due to the treatment contingenciesto which they have been exposed, become will-ing to approach the object because the stimuluscontrol of arousal over avoidance behavior hasbeen weakened. Hayes (in press) argues that mostcurrent treatments of fears tend to emphasizethe former mechanism. He proposes an ap-proach in which the latter mechanism is em-phasized. Ifhe is correct, one would predict thatboth self-efficacy ratings and approach behaviorcould be increased without reducing arousal.

14 ANTHONY BIGLAN

REFERENCESBaer, J. S., Holt, C. S., & Lichtenstein, E. (1986).

Self-efficacy and smoking re-examined: Con-struct validity and clinical utility. Journal ofCon-sulting and Clinical Psychology, 54, 846-852.

Baer, R. A., Williams, J. A., Osnes, P. G., & Stokes,T. F. (1984). Delayed reinforcement as an in-discriminable contingency in verbal/nonverbalcorrespondence training. Journal ofApplied Be-havior Analysis, 17, 429-440.

Bandura, A. (1969). Principles of behavior mod-ification. New York: Holt, Rinehart, & Winston.

Bandura, A. (1977a). Self-efficacy: Toward aunifying theory ofbehavioral change. Psycholog-ical Review, 84, 191-215.

Bandura, A. (1977b). Social learning theory. En-glewood Cliffs, NJ: Prentice-Hall.

Bandura, A. (1982a). Self-efficacy mechanism inhuman agency. American Psychologist, 37, 122-147.

Bandura, A. (1982b). The assessment and pre-dictive generality of self-percepts of efficacy.Journal ofBehavioral Therapy andExperimentalPsychology, 13, 195-199.

Bandura, A., & Adams, N. E. (1977). Analysis ofself-efficacy theory of behavioral change. Cog-nitive Therapy and Research, 1, 287-310.

Bandura, A., Adams, N. E., & Beyer, J. (1977).Cognitive processes mediating behavioral change.Journal ofPersonality and Social Psychology, 35,125-139.

Bandura, A., Adams, N. E., Hardy, A. B., & How-ells,G.N. (1980). Testsofthegeneralityofself-efficacy theory. Cognitive Therapy and Research,4, 39-66.

Bandura, A., & Cervone, D. (1983). Self-evalua-tive and self-efficacy mechanisms governing themotivational effects of goal systems. Journal ofPersonalityandSocialPsychology, 45, 10 17-1028.

Bandura, A., Reese, L., & Adams, N. E. (1982).Micro-analysis of action and fear arousal as afunction of differential levels of perceived self-efficacy. Journal of Personality and Social Psy-chology, 43, 5-21.

Bandura, A., & Schunk, D. H. (1981). Cultivatingcompetence, self-efficacy, and intrinsic interestthrough proximal self-motivation. Journal ofPersonality and Social Psychology, 41, 586-598.

Bandura, A., Taylor, C. B., Williams, S. L., Mef-ford, I. N., & Barchas, J. D. (1985). Catechol-amine secretion as a function ofperceived copingself-efficacy. Journal of Consulting and ClinicalPsychology, 53, 406-414.

Becker, W. C., & Camine, D. (1980). Direct in-struction: An effective approach to educationalintervention with the disadvantaged and low per-formers. In B. B. Lahey & A. E. Kazdin (Eds.),Advances in clinical child psychology (Vol. 2, pp.429-473). New York: Plenum.

Betz, N. E., & Hackett, G. (1981). The relation-ship ofcareer-related self-efficacy expectations toperceived career options in college men andwomen. Journal of Counseling Psychology, 28,399-410.

Biglan, A., & Dow, M. G. (1981). Toward a sec-

ond-generation model: A problem-specific ap-proach. In L. P. Rehm (Ed.), Behavior therapyfor depression (pp. 97-121). New York: Academ-ic.

Biglan, A., & Kass, D. J. (1977). The empiricalnature of behavior therapies. Behaviorism, 5, 1-15.

Biran, M., & Wilson, G. T. (1981). Treatment ofphobic disorders using cognitive and exposuremethods: A self-efficacy analysis. Journal ofCon-sufting and Clinical Psychology, 49, 886-899.

Borkovec, T. D. (1978). Self-efficacy: Cause orreflection of behavioral change? Advances in Be-haviour Research and Therapy, 1, 163-170.

Brady, J. P. (1971). Metronome-conditionedspeech retraining for stuttering. Behavior Ther-apy, 2, 129-150.

Christopherson, E. R., & Rapoff, M. A. (1980).Pediatric psychology: An appraisal. In B. B. Lah-ey & A. E. Kazdin (Eds.), Advances in clinicalchildpsychology (Vol. 3, pp. 311-329). New York:Plenum.

Condiotte, M. M., & Lichtenstein, E. (1981). Self-efficacy and relapse in smoking cessation pro-grams. Journal of Consulting and Clinical Psy-chology, 49, 648-658.

Curran, J. P. (1977). Skills training as an approachto the treatment of heterosexual-social anxiety:A review. Psychological Bulletin, 84, 140-157.

Deguchi, H. (1984). Observational learning froma radical-behavioristic viewpoint. The BehaviorAnalyst, 7, 83-95.

DiClemente,C.C. (1981). Self-efficacyandsmok-ing cessation maintenance: A preliminary report.Cognitive Therapy and Research, 5, 175-187.

Eysenck, H. J. (1978). Expectations as causal ele-ments in behavioral change. Advances in Behav-iour Research and Therapy, 1, 17 1-175.

Gauthier, J., & Ladouceur, R. (1981). The influ-ence of self-efficacy reports on performance. Be-havior Therapy, 12, 436-439.

Gelder, M., & Marks, I. M. (1966). Severe ago-raphobia: A controlled prospective trial of be-havior therapy. British Journal ofPsychiatry, 112,309-319.

Glasgow, R. E., Klepac, R. K., Dowling, J., & Rokke,P. D. (1984). Extensions of self-efficacy: Pre-dicting pain tolerance. Unpublished manuscript,University of North Dakota: Williston, ND.

Hayes, S. C. (in press). A contextual approach totherapeutic change. In N. Jacobson (Ed.), Cog-nitive and behavior therapies in clinical practice.New York: Guilford.

Hayes, S. C., & Brownstein, A. J. (1986). Men-talism, behavior-behavior relations, and a be-havior-analytic view of the purposes of science.TheBehaviorAnalyst, 9, 175-190.

Hayes, S. C., Brownstein, A. J., Zettle, R. D., Ro-senfarb, I., & Korn, Z. (1986). Rule-governedbehavior and sensitivity to changing conse-quences of responding. Journal of the Experi-mental Analysis ofBehavior, 45, 237-256.

Hayes, S. C., Rosenfarb, I., Wulfert, E., Munt, E.D., Korn, Z., & Zettle, R. D. (1985). Self-re-inforcement effects: An artifact ofsocial standard

SELF-EFFICACY THEORY 15

settings? Journal of Applied Behavior Analysis,18, 201-214.

Hersen, M., & Bellack, A. S. (1976). Social skillstraining for chronic psychiatric patients: Ratio-nale, research findings, and future directions.Comprehensive Psychiatry, 17, 559-580.

Hops, H. (1982). Social skills training for sociallywithdrawn/isolated children. In P. Karoly & J.Steffen (Eds.), Enhancing children's competen-cies. Lexington, MA: Lexington Books.

Hops, H. (1985). [A review ofMarital interaction:Analysis and modification]. The Behavior Ana-lyst, 8, 235-246.

Israel, A. C., & Brown, M. S. (1977). Correspon-dence training, prior verbal training, and controlof nonverbal behavior via control of verbal be-havior. Journal ofApplied BehaviorAnalysis, I0,333-338.

Jacobson, N. S., & Margolin, G. (1979). Maritaltherapy: Strategies based on social learning andbehavior exchange principles. New York: Brun-ner/Mazel.

Kazdin, A. E. (1978). History of behavior modi-fication: Experimental foundations of contem-porary research. Baltimore: University Park Press.

Kazdin, A. E. (1979). Sociopsychological factorsin psychopathology. In A. S. Bellack & M. Hersen(Eds.), Research andpractice in social skills train-ing (pp. 41-74). New York: Plenum.

Kirsch, I. (1980). "Microanalytic" analyses of ef-ficacy expectations as predictors of performance.Cognitive Therapy and Research, 4, 259-262.

Kirsch, I., & Wickless, C. V. (1983). Concordancerates between self-efficacy and approach behaviorare redundant. Cognitive Therapy and Research,7, 179-188.

Lane, T. W., & Borkovec, T. D. (1984). The in-fluence of therapeutic expectancy/demand on self-efficacy ratings. Cognitive Therapy and Research,8, 95-106.

Lang, P. J. (1978). Self-efficacy theory: Thoughtson cognition and unification. Advances in Be-haviour Research and Therapy, 1, 187-192.

LoPiccolo, J., & LoPiccolo, L. (1978). Handbookofsex therapy. New York: Plenum.

Maletzky, B. M. (1980). Assisted covert sensiti-zation. In D. J. Cox & R. J. Daitzman (Eds.),Exhibitionism: Description, assessment, andtreatment (pp. 187-251). New York: Garland.

Marr, J. (1984). Conceptual approaches and is-sues. Journal ofthe Experimental Analysis ofBe-havior, 42, 353-362.

McAlister, A., Perry, C., & Maccoby, N. (1979).Adolescent smoking: Onset and prevention. Pe-diatrics, 63, 650-658.

Michael, J. (1984). Verbal behavior. Journal ofthe Experimental Analysis ofBehavior, 42, 363-376.

Mischel, W. (1968). Traits and states as con-structs. New York: Wiley.

Moore, J. (1984a). On behaviorism, knowledge,and causal explanation. The Psychological Re-cord, 34, 73-97.

Moore, J. (1984b). On privacy, causes, and con-tingencies. The Behavior Analyst, 7, 3-16.

Nelson, R. O., Hayes, S. C., Spong, R. T., Jarrett,

R. B., & McKnight, D. L. (1983). Self-rein-forcement: Appealing misnomer or effectivemechanism? Behavioral Research and Therapy,21, 557-566.

Paniagua, F. A., & Baer, D. M. (1982). The anal-ysis of correspondence training as a chain rein-forceable at any point. Child Development, 53,786-798.

Patterson, G. R. (1982). A social learning ap-proach to family intervention: Vol. 3. Coercivefamily process. Eugene, OR: Castalia.

Rachman, S. J. (1978). Fear and courage. SanFrancisco: W. H. Freeman.

Risley, T. R., & Hart, B. (1968). Developing cor-respondence between the nonverbal and verbalbehavior of preschool children. Journal of Ap-plied Behavior Analysis, 1, 267-281.

Rosenfarb, I., & Hayes, S. C. (1984). Social stan-dard setting: The Achilles heel of informationalaccounts of therapeutic change. Behavior Ther-apy, 15, 515-528.

Shalock, R., & Harper, R. (1978). Placement fromcommunity-based mental retardation programs.American Journal ofMental Deficiency, 83, 240-247.

Skinner, B. F. (1945). Operational analysis ofpsy-chological terms. Psychological Review, 52, 270-277.

Skinner, B. F. (1953). Science and human behav-ior. New York: Macmillan.

Skinner,B.F. (1957). Verbalbehavior. NewYork:Appleton-Century-Crofts.

Skinner, B. F. (1969). Contingencies ofreinforce-ment: A theoretical analysis. New York: Apple-ton-Century-Crofts.

Skinner, B. F. (1974). About behaviorism. NewYork: Knopf.

Telch, M. J., Bandura, A., Vinciguerra, P., Agras,S., & Stout, A. L. (1982). Social demand andcongruence between self-efficacy and perfor-mance. Behavior Therapy, 13, 694-701.

Weinberg, R. S., Yukelson, D., & Jackson, A.(1980). Effect of public and private efficacy ex-pectations on competitive performance. JournalofSport Psychology, 2, 340-349.

Whitman, T. L., Scibak, J. W., Butler, K. M.,Richter, R., & Johnson, M. R. (1982). Improv-ing classroom behavior in mentally retarded chil-dren through correspondence training. Journal ofApplied Behavior Analysis, 15, 545-564.

Zeiss, A., Lewinsohn, P., & Munoz, R. F. (1979).Nonspecific improvement effects in depressionusing interpersonal skills training, pleasant activ-ity schedules, or cognitive training. Journal ofConsulting and Clinical Psychology, 47, 427-439.

Zettle,R.D.,&Hayes,S.C. (1983). Effectofsocialcontext on the impact of coping self-statements.Psychological Reports, 52, 391-401.

Zettle, R. D., & Hayes, S. C. (1982). Rule-gov-erned behavior: A potential theoretical frame-work for cognitive-behavioral therapy. Advancesin Cognitive-Behavioral Research and Therapy,1, 73-118.