Some current dimension of applied behavior analysis baer, wolf, risley
JOURNAL OF APPLIED BEHAVIOR ANALYSIS 1968, 1, 91-97 NUMBER I (SPRING, 1968) SOME CURRENT DIMENSIONS OF APPLIED BEHAVIOR ANALYSIS DONALD M. BAER, MONTROSE M. WOLF, AND TODD R. RISLEY THE UNIVERSITY OF KANSAS The analysis of individual behavior is a tions, it seems, are. Analytic behavioral ap-problem in scientific demonstration, reason- plication is the process of applying sometimesably well understood (Skinner, 1953, Sec. 1), tentative principles of behavior to the im-comprehensively described (Sidman, 1960), provement2 of specific behaviors, and simul-and quite thoroughly practised (Journal of taneously evaluating whether or not anythe Experimental Analysis of Behavior, 1957 changes noted are indeed attributable to the-). That analysis has been pursued in many process of application-and if so, to whatsettings over many years. Despite variable parts of that process. In short, analytic be-precision, elegance, and power, it has resulted havioral application is a self-examining, self-in general descriptive statements of mecha- evaluating, discovery-oriented research pro-nisms that can produce many of the forms cedure for studying behavior. So is allthat individual behavior may take. experimental behavioral research (at least, The statement of these mechanisms estab- according to the usual strictures of modernlishes the possibility of their application to graduate training). The differences are mat-problem behavior. A society willing to con- ters of emphasis and of selection.sider a technology of its own behavior appar- The differences between applied and basicently is likely to support that application research are not differences between thatwhen it deals with socially important behav- which "discovers" and that which merely "ap-iors, such as retardation, crime, mental illness, plies" what is already known. Both endeavorsor education. Such applications have ap- ask what controls the behavior under study.peared in recent years. Their current num- Non-applied research is likely to look at anyber and the interest which they create appar- behavior, and at any variable which may con-ently suffice to generate a journal for their ceivably relate to it. Applied research is con-display. That display may well lead to the strained to look at variables which can bewidespread examination of these applica- effective in improving the behavior undertions, their refinement, and eventually their study. Thus it is equally a matter of researchreplacement by better applications. Better to discover that the behaviors typical of re-applications, it is hoped, will lead to a better tardates can be related to oddities of theirstate of society, to whatever extent the behav-ior of its members can contribute to the good- 2If a behavior is socially important, the usual be-ness of a society. Since the evaluation of what havior analysis will aim at its improvement. The so-is a "good" society is in itself a behavior of cial value dictating this choice is obvious. However, it can be just as illuminating to demonstrate how aits members, this hope turns on itself in a behavior may be worsened, and there will arise occa-philosophically interesting manner. However, sions when it will be socially important to do so. Dis-it is at least a fair presumption that behav- ruptive classroom behavior may serve as an example.ioral applications, when effective, can some- Certainly it is a frequent plague of the educationaltimes lead to social approval and adoption. system. A demonstration of what teacher procedures produce more of this behavior is not necessarily the Behavioral applications are hardly a new reverse of a demonstration of how to promote posi-phenomenon. Analytic behavioral applica- tive study behaviors. There may be classroom situa- tions in which the teacher cannot readily establish high rates of study, yet still could avoid high rates of Reprints may be obtained from Donald M. Baer, disruption, if she knew what in her own proceduresDept. of Human Development, University of Kansas, leads to this disruption. The demonstration whichLawrence, Kansas 66044. showed her that would thus have its value. 91
92 DONALD M. BAER et al.chromosome structure and to oddities of their which society shows in the problems beingreinforcement history. But (currently) the studied. In behavioral application, the behav-chromosome structure of the retardate does ior, stimuli, and/or organism under study arenot lend itself to experimental manipulation chosen because of their importance to manin the interests of bettering that behavior, and society, rather than their importance towhereas his reinforcement input is always theory. The non-applied researcher may studyopen to current re-design. eating behavior, for example, because it re- Similarly, applied research is constrained lates directly to metabolism, and there areto examining behaviors which are socially im- hypotheses about the interaction between be-portant, rather than convenient for study. It havior and metabolism. The non-applied re-also implies, very frequently, the study of searcher also may study bar-pressing becausethose behaviors in their usual social settings, it is a convenient response for study; easy forrather than in a "laboratory" setting. But a the subject, and simple to record and inte-laboratory is simply a place so designed that grate with theoretically significant environ-experimental control of relevant variables is mental events. By contrast, the applied re-as easy as possible. Unfortunately, the usual searcher is likely to study eating because theresocial setting for important behaviors is are children who eat too little and adults whorarely such a place. Consequently, the analy- eat too much, and he will study eating insis of socially important behaviors becomes exactly those individuals rather than in moreexperimental only with difficulty. As the convenient ones. The applied researcher mayterms are used here, a non-experimental anal- also study bar-pressing if it is integrated withysis is a contradiction in terms. Thus, ana- socially important stimuli. A program for alytic behavioral applications by definition teaching machine may use bar-pressing be-achieve experimental control of the processes havior to indicate mastery of an arithmeticthey contain, but since they strive for this con- skill. It is the arithmetic stimuli which aretrol against formidable difficulties, they important. (However, some future appliedachieve it less often per study than would a study could show that bar-pressing is morelaboratory-based attempt. Consequently, the practical in the process of education than arate of displaying experimental control re- pencil-writing response.3)quired of behavioral applications has become In applied research, there is typically acorrespondingly less than the standards typi- close relationship between the behavior andcal of laboratory research. This is not because stimuli under study and the subject in whomthe applier is an easy-going, liberal, or gen- they are studied. Just as there seem to be fewerous fellow, but because society rarely will behaviors that are intrinsically the target ofallow its important behaviors, in their cor- application, there are few subjects who auto-respondingly important settings, to be manip- matically confer on their study the status ofulated repeatedly for the merely logical com- application. An investigation of visual signal fort of a scientifically sceptical audience. detection in the retardate may have little im- Thus, the evaluation of a study which pur- mediate importance, but a similar study in ports to be an applied behavior analysis is radar-scope watchers has considerable. A somewhat different than the evaluation of a study of language development in the re- similar laboratory analysis. Obviously, the tardate may be aimed directly at an immedi- study must be applied, behavioral, and ana- lytic; in addition, it should be technological, "Research may use the most convenient behaviors conceptually systematic, and effective, and it and stimuli available, and yet exemplify an ambition should display some generality. These terms in the researcher eventually to achieve application to are explored below and compared to the cri- socially important settings. For example, a study may seek ways to give a light flash a durable conditioned teria often stated for the evaluation of behav- reinforcing function, because the experimenter wishes ioral research which, though analytic, is not to know how to enhance school childrens responsive- applied. ness to approval. Nevertheless, durable bar-pressing for that light flash is no guarantee that the obvious Applied classroom analogue will produce durable reading be- havior for teacher statements of "Good!" Until the The label applied is not determined by the analogue has been proven sound, application has not research procedures used but by the interest been achieved.
SOME DIMENSIONS OF APPLIED BEHAVIOR ANALYSIS 93ate social problem, while a similar study in typical reliability will not always be possible.the MIT sophomore may not. Enhancement The reliable use of human beings to quantifyof the reinforcing value of praise for the re- the behavior of other human beings is antardate alleviates an immediate deficit in his area of psychological technology long sincecurrent environment, but enhancement of the well developed, thoroughly relevant, and veryreinforcing value of 400 Hz (cps) tone for the often necessary to applied behavior analysis.same subject probably does not. Thus, a pri- A useful tactic in evaluating the behavioralmary question in the evaluation of applied attributes of a study is to ask not merely, wasresearch is: how immediately important is behavior changed? but also, whose behavior?this behavior or these stimuli to this subject? Ordinarily it would be assumed that it was the subjects behavior which was altered; yetBehavioral careful reflection may suggest that this was Behaviorism and pragmatism seem often to not necessarily the case. If humans are ob-go hand in hand. Applied research is emi- serving and recording the behavior undernently pragmatic; it asks how it is possible study, then any change may represent ato get an individual to do something effec- change only in their observing and record-tively. Thus it usually studies what subjects ing responses, rather than in the subjects be-can be brought to do rather than what they havior. Explicit measurement of the reliabil-can be brought to say; unless, of course, a ity of human observers thus becomes notverbal response is the behavior of interest. merely good technique, but a prime criterionAccordingly a subjects verbal description of of whether the study was appropriately be-his own non-verbal behavior usually would havioral. (A study merely of the behavior ofnot be accepted as a measure of his actual be- observers is behavioral, of course, but prob-havior unless it were independently substan- ably irrelevant to the researchers goal.) Alter-tiated. Hence there is little applied value in natively, it may be that only the experimen-the demonstration that an impotent man can ters behavior has changed. It may be reported,be made to say that he no longer is impotent. for example, that a certain patient rarelyThe relevant question is not what he can say, dressed himself upon awakening, and conse-but what he can do. Application has not been quently would be dressed by his attendant.achieved until this question has been an- The experimental technique to be appliedswered satisfactorily. (This assumes, of course, might consist of some penalty imposed unlessthat the total goal of the applied researcher the patient were dressed within half an houris not simply to get his patient-subjects to after awakening. Recording of an increasedstop complaining to him. Unless society probability of self-dressing under these condi-agrees that this researcher should not be tions might testify to the effectiveness of thebothered, it will be difficult to defend that penalty in changing the behavior; however, itgoal as socially important.) might also testify to the fact that the patient Since the behavior of an individual is com- would in fact probably dress himself withinposed of physical events, its scientific study half an hour of arising, but previously wasrequires their precise measurement. As a re- rarely left that long undressed before beingsult, the problem of reliable quantification clothed by his efficient attendant. (The at-arises immediately. The problem is the same tendant now is the penalty-imposing experi-for applied research as it is for non-applied menter and therefore always gives the patientresearch. However, non-applied research typi- his full half-hour, in the interests of precisecally will choose a response easily quantified experimental technique, of course.) This errorin a reliable manner, whereas applied re- is an elementary one, perhaps. But it suggestssearch rarely will have that option. As a re- that in general, when an experiment proceedssult, the applied researcher must try harder, from its baseline to its first experimental phase,rather than ignore this criterion of all trust- changes in what is measured need not alwaysworthy research. Current applied research reflect the behavior of the subject.often shows that thoroughly reliable quantifi-cation of behavior can be achieved, even in Analyticthoroughly difficult settings. However, it also The analysis of a behavior, as the term issuggests that instrumented recording with its used here, requires a believable demonstra-
94 DONALD M. BAER et al.tion of the events that can be responsible for so long as the social setting in which the be-the occurrence or non-occurrence of that be- havior is studied dictates against using themhavior. An experimenter has achieved an repeatedly. Indeed, it may be that repeatedanalysis of a behavior when he can exercise reversals in some applications have a positivecontrol over it. By common laboratory stan- effect on the subject, possibly contributing todards, that has meant an ability of the ex- the discrimination of relevant stimuli in-perimenter to turn the behavior on and off, volved in the problem.)or up and down, at will. Laboratory standards In using the reversal technique, the experi-have usually made this control clear by dem- menter is attempting to show that an analysisonstrating it repeatedly, even redundantly, of the behavior is at hand: that whenever heover time. Applied research, as noted before, applies a certain variable, the behavior is pro-cannot often approach this arrogantly fre- duced, and whenever he removes this vari-quent clarity of being in control of important able, the behavior is lost. Yet applied behav-behaviors. Consequently, application, to be ior analysis is exactly the kind of researchanalytic, demonstrates control when it can, which can make this technique self-defeatingand thereby presents its audience with a prob- in time. Application typically means produc-lem of judgment. The problem, of course, is ing valuable behavior; valuable behaviorwhether the experimenter has shown enough usually meets extra-experimental reinforce-control, and often enough, for believability. ment in a social setting; thus, valuable be-Laboratory demonstrations, either by over- havior, once set up, may no longer be depen-replication or an acceptable probability level dent upon the experimental technique whichderived from statistical tests of grouped data, created it. Consequently, the number of re-make this judgment more implicit than ex- versals possible in applied studies may be lim-plicit. As Sidman points out (1960), there is ited by the nature of the social setting instill a problem of judgment in any event, and which the behavior takes place, in more waysit is probably better when explicit. than one. There are at least two designs commonly An alternative to the reversal techniqueused to demonstrate reliable control of an may be called the "multiple baseline" tech-important behavioral change. The first can nique. This alternative may be of particularbe referred to as the "reversal" technique. value when a behavior appears to be irre-Here a behavior is measured, and the measure versible or when reversing the behavior is un-is examined over time until its stability is desirable. In the multiple-baseline technique,clear. Then, the experimental variable is ap- a number of responses are identified and mea-plied. The behavior continues to be mea- sured over time to provide baselines againstsured, to see if the variable will produce a which changes can be evaluated. With thesebehavioral change. If it does, the experimen- baselines established, the experimenter thental variable is discontinued or altered, to see applies an experimental variable to one ofif the behavioral change just brought about the behaviors, produces a change in it, anddepends on it. If so, the behavioral change perhaps notes little or no change in the othershould be lost or diminished (thus the term baselines. If so, rather than reversing the just-"reversal"). The experimental variable then produced change, he instead applies the ex-is applied again, to see if the behavioral perimental variable to one of the other, aschange can be recovered. If it can, it is pur- yet unchanged, responses. If it changes at thatsued further, since this is applied research point, evidence is accruing that the experi-and the behavioral change sought is an im- mental variable is indeed effective, and thatportant one. It may be reversed briefly again, the prior change was not simply a matter ofand yet again, if the setting in which the be- coincidence. The variable then may be ap-havior takes place allows further reversals. plied to still another response, and so on. TheBut that setting may be a school system or a experimenter is attempting to show that hefamily, and continued reversals may not be has a reliable experimental variable, in thatallowed. They may appear in themselves to each behavior changes maximally only whenbe detrimental to the subject if pursued too the experimental variable is applied to it.often. (Whether they are in fact detrimental How many reversals, or how many base-is likely to remain an unexamined question lines, make for believability is a problem for
SOME DIMENSIONS OF APPLIED BEHAVIOR ANALYSIS 95the audience. If statistical analysis is applied, The problem again will be to make such anthe audience must then judge the suitability analysis reliable, and, as before, that mightof the inferential statistic chosen and the pro- be approached by the repeated alternate usepriety of these data for that test. Alternatively, of different values on the same behavior (re-the audience may inspect the data directly versal), or by the application of different val-and relate them to past experience with simi- ues to different groups of responses (multiplelar data and similar procedures. In either baseline). At this stage in the development ofcase, the judgments required are highly quali- applied behavior analysis, primary concern istative, and rules cannot always be stated prof- usually with reliability, rather than with para-itably. However, either of the foregoing de- metric analysis or component analysis.signs gathers data in ways that exemplify theconcept of replication, and replication is the Technologicalessence of believability. At the least, it would "Technological" here means simply thatseem that an approach to replication is better the techniques making up a particular behav-than no approach at all. This should be es- ioral application are completely identifiedpecially true for so embryonic a field as be- and described. In this sense, "play therapy"havioral application, the very possibility of is not a technological description, nor is "so-which is still occasionally denied. cial reinforcement". For purposes of applica- The preceding discussion has been aimed tion, all the salient ingredients of play ther-at the problem of reliability: whether or not apy must be described as a set of contingen-a certain procedure was responsible for a cor- cies between child response, therapist re-responding behavioral change. The two gen- sponse, and play materials, before a statementeral procedures described hardly exhaust the of technique has been approached. Similarly,possibilities. Each of them has many varia- all the ingredients of social reinforcementtions now seen in practice; and current ex- must be specified (stimuli, contingency, andperience suggests that many more variations schedule) to qualify as a technological pro-are badly needed, if the technology of impor- cedure.tant behavioral change is to be consistently The best rule of thumb for evaluating abelievable. Given some approach to reliabil- procedure description as technological isity, there are further analyses of obvious value probably to ask whether a typically trainedwhich can be built upon that base. For exam- reader could replicate that procedure wellple, there is analysis in the sense of simplifi- enough to produce the same results, givencation and separation of component processes. only a reading of the description. This isOften enough, current behavioral procedures very much the same criterion applied to pro-are complex, even "shotgun" in their applica- cedure descriptions in non-applied research,tion. When they succeed, they clearly need to of course. It needs emphasis, apparently, inbe analyzed into their effective components. that there occasionally exists a less-than-pre-Thus, a teacher giving M & Ms to a child cise stereotype of applied research. Where ap-may succeed in changing his behavior as plication is novel, and derived from princi-planned. However, she has almost certainly ples produced through non-applied research,confounded her attention and/or approval as in current applied behavior analysis, thewith each M & M. Further analysis may be reverse holds with great urgency.approached by her use of attention alone, the Especially where the problem is applica-effects of which can be compared to the ef- tion, procedural descriptions require consid-fects of attention coupled with candies. erable detail about all possible contingenciesWhether she will discontinue the M Sc Ms, of procedure. It is not enough to say what isas in the reversal technique, or apply atten- to be done when the subject makes responsetion with M Sc Ms to certain behaviors and R1; it is essential also whenever possible toattention alone to certain others, as in the say what is to be done if the subject makes themultiple baseline method, is again the prob- alternative responses, R2, R3, etc. For exam-lem in basic reliability discussed above. An- ple, one may read that temper tantrums inother form of analysis is parametric: a dem- children are often extinguished by closingonstration of the effectiveness of different the child in his room for the duration of thevalues of some variable in changing behavior. tantrums plus ten minutes. Unless that pro-
96 DONALD M. BAER et al.cedure description also states what should be ter of degree: an increase in those childrendone if the child tries to leave the room early, from D- to C might well be judged an im-or kicks out the window, or smears feces on portant success by an audience which thinksthe walls, or begins to make strangling sounds, that C work is a great deal different than Detc., it is not precise technological description. work, especially if C students are much less likely to become drop-outs than D students.Conceptual Systems In evaluating whether a given application The field of applied behavior analysis will has produced enough of a behavioral changeprobably advance best if the published de- to deserve the label, a pertinent question canscriptions of its procedures are not only pre- be, how much did that behavior need to becisely technological, but also strive for rele- changed? Obviously, that is not a scientificvance to principle. To describe exactly how question, but a practical one. Its answer isa preschool teacher will attend to jungle-gym likely to be supplied by people who must dealclimbing in a child frightened of heights is with the behavior. For example, ward person-good technological description; but further to nel may be able to say that a hospitalizedcall it a social reinforcement procedure re- mute schizophrenic trained to use 10 verballates it to basic concepts of behavioral devel- labels is not much better off in self-helpopment. Similarly, to describe the exact se- skills than before, but that one with 50 suchquence of color changes whereby a child is labels is a great deal more effective. In thismoved from a color discrimination to a form case, the opinions of ward aides may be morediscrimination is good; to refer also to "fad- relevant than the opinions of psycholinguists.ing" and "errorless discrimination" is better.In both cases, the total description is ade- Generalityquate for successful replication by the reader; A behavioral change may be said to haveand it also shows the reader how similar pro- generality if it proves durable over time, ifcedures may be derived from basic principles. it appears in a wide variety of possible envi-This can have the effect of making a body of ronments, or if it spreads to a wide varietytechnology into a discipline rather than a of related behaviors. Thus, the improvementcollection of tricks. Collections of tricks his- of articulation in a clinic setting will prove totorically have been difficult to expand system- have generality if it endures into the futureatically, and when they were extensive, diffi- after the clinic visits stop; if the improved ar-cult to learn and teach. ticulation is heard at home, at school, and on dates; or if the articulation of all words, notEffective just the ones treated, improves. Application If the application of behavioral techniques means practical improvement in importantdoes not produce large enough effects for behaviors; thus, the more general that appli-practical value, then application has failed. cation, the better, in many cases. TherapistsNon-applied research often may be extremely dealing with the development of heterosexualvaluable when it produces small but reliable behavior may well point out there are sociallyeffects, in that these effects testify to the op- appropriate limits to its generality, once de-eration of some variable which in itself has veloped; such limitations to generality aregreat theoretical importance. In application, usually obvious. That generality is a valuablethe theoretical importance of a variable is characteristic of applied behavior analysisusually not at issue. Its practical importance, which should be examined explicitly appar-specifically its power in altering behavior ently is not quite that obvious, and is statedenough to be socially important, is the essen- here for emphasis.tial criterion. Thus, a study which shows that That generality is not automatically accom-a new classroom technique can raise the grade plished whenever behavior is changed alsolevel achievements of culturally deprived chil- needs occasional emphasis, especially in theclren from D- to D is not an obvious exam- evaluation of applied behavior analysis. It isple of applied behavior analysis. That same sometimes assumed that application has failedstudy might conceivably revolutionize educa- when generalization does not take place intional theory, but it clearly has not yet revo- any widespread form. Such a conclusion haslutionized education. This is of course a mat- no generality itself. A procedure which is ef-
SOME DIMENSIONS OF APPLIED BEHAVIOR ANALYSIS 97fective in changing behavior in one setting analysis will make obvious the importance ofmay perhaps be easily repeated in other set- the behavior changed, its quantitative charac-tings, and thus accomplish the generalization teristics, the experimental manipulationssought. Furthermore, it may well prove the which analyze with clarity what was responsi-case that a given behavior change need be ble for the change, the technologically exactprogrammed in only a certain number of description of all procedures contributing tosettings, one after another, perhaps, to ac- that change, the effectiveness of those proce-complish eventually widespread generaliza- dures in making sufficient change for value,tion. A child may have 15 techniques for and the generality of that change.disrupting his parents, for example. The elim-ination of the most prevalent of these maystill leave the remaining 14 intact and in REFERENCESforce. The technique may still prove both val- Journal of the Experimental Analysis of Behavior.uable and fundamental, if when applied to Bloomington: Society for the Experimental Analy- sis of Behavior, 1957-.the next four successfully, it also results in Sidman, Murray. Tactics of scientific research. Newthe "generalized" loss of the remaining 10. In York: Basic Books, 1960.general, generalization should be pro- Skinner, B. F. Science and human behavior. Newgrammed, rather than expected or lamented. York: Macmillan, 1953. Thus, in summary, an applied behavior Received 24 December 1967.