Behavior Analysis, Relational Frame Theory, and the Challenge of Human Language and Cognition

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by MURAL - Maynooth University Research Archive Library

The Analysis of Verbal Behavior 2003, 19, 39–54

Behavior Analysis, Relational Frame Theory, and the Challenge of Human Language and Cognition: A Reply to the Commentaries on Relational Frame Theory: A Post- Skinnerian Account of Human Language and Cognition

Steven C. Hayes, University of Nevada, Reno, and Dermot Barnes-Holmes and Bryan Roche, National University of Ireland, Maynooth

Answers to a series of commentaries are presented and the challenge Relational Frame Theory (RFT) presents to behavior analysis is explicated. RFT is a behavior analytic theory, based on extensive behavior analytic data, which appeals only to known principles to explain arbitrarily applicable relational responding. The claim that such responding is operant must be answerable within behavior analysis. RFT has too much empirical support for the field to avoid this challenge. If the answer is “yes,” behavior analysis seems destined to enter a new era.

Any disciplinary approach is composed of tively simple organism–environment interac- multiple levels of scientific activity: assumptions studied largely in non-human organisms. tions, methods, techniques, concepts, research Behavior analysis has been enormously suc- questions, research answers, and the like. Ap- cessful, but that very success may make it more proaches that are long lived, and particularly difficult to take new directions in areas that those associated with significant intellectual have not been as successfully addressed. figures, also have substantial histories and tra- Relational Frame Theory seems to be a test ditions, complete with heroes and tales of case. RFT is making explicitly behavior ana- former battles won. As traditions become more lytic claims based on behavior analytically sen- extended, new directions become harder to sible research data. Nevertheless, it suggests assimilate, especially if these new directions significant changes in some of the methods, challenge previously dominant methods or techniques, concepts, research questions, and questions, or if the views of historical figures research answers of the field. For example, if are challenged. RFT is correct, we need many more basic hu- Behavior analysis is one of the longer-lived man laboratories; we need new research meth- traditions in psychology. Behavior analysis ods; we need much more experimental research sought nothing less than a comprehensive ac- on language processes; and many of Skinner’s count of the psychological activity of organ- specific ideas (e.g., about the role of human isms, particularly and ultimately, the complex emotion; about the nature and function of self- behavior of human beings, based on a func- knowledge; about the need to understand pri- tional, monistic, scientific approach. Behavior vate events in order to understand overt be- analysis adopted some unusual tactics in pur- havior; and so on) are challenged in whole or suing this goal, including the emphasis on rela- in part. This is a difficult possibility for behavior analysis to face head on, and it helps explain a sense of discomfort that is palpable Please address correspondence concerning this throughout the commentaries. We understand article to Steven C. Hayes, Department of Psychol- ogy / 296, University of Nevada, Reno, Reno, NV the discomfort, and have felt it ourselves, but 89557-0062 (email: [email protected]) or to we do not think behavior analysis can avoid Dermot Barnes-Holmes, Department of Psychology, that discomfort without harm. The challenge National University of Ireland, Maynooth, Relational Frame Theory presents to behavior Maynooth, Co. Kildare, Ireland (email: analysis as a whole must be faced and resolved, [email protected]). one way or the other. 39 40 HAYES, BARNES-HOLMES, AND ROCHE

The Challenge of Relational Operants be merely a molar description of that which we are trying to understand rather than an ex- It is precisely because RFT is a behavior planation of what we are trying to understand.” analytic theory that the challenge is so great. (p. 17) RFT claims that verbal behavior is a particular If RFT used a new principle to explain rela- kind of operant behavior: arbitrarily applicable tional operants, it might be appropriate for be- relational responding. Considerable detail is havior analysts to resist these claims until devoted to delineating the kind of operant we “more conservative” alternatives based on ex- are speaking about. RFT researchers have pro- isting principles could receive more attention. vided several lines of evidence that support this That is not the case, however. claim and year after year more data come in. The principles and procedures used to ex- To our knowledge, there has not yet been a sour plain relational operants in RFT are identical note in the data that have been generated on to those used to explain any operant: contacted this point so far. consistencies in contingencies across multiple It should not be easy for behavior analysts exemplars. The explanation of arbitrarily to dismiss a credible claim that a particular type applicably relational operants appeals to no of operant exists. Behavior analysis as a field new principles at all. It is only after assuming must have concepts to decide if something can that the operant described by RFT exists that or cannot be an operant, and if a particular op- we then claim that a new principle exists. The erant is conceptually possible, methods and concept of a new principle of behavior is not techniques to determine empirically if it is in used to explain the process that gives rise to fact. If such is not the case, behavior analysis relational operants, rather it is used to explain is not a field worthy of the name “operant psy- the implications of such an operant. This dis- chology.” tinction fundamentally changes the nature of RFT is now 18 years old. It has spawned the challenge RFT presents to behavior analy- more basic human operant work than almost sis as a whole. It may be scientifically conser- any theory put forward during that time. Sooner vative to avoid using new explanatory prin- or later the field must reach a conclusion about ciples, but it is not scientifically conservative the central claim of RFT. Arbitrarily applicable to use old labels for phenomena that fail to fit relational responding either is an operant or it the definitional limits for those labels. is not. That conclusion does not have to be What is new about RFT? We point out that reached tomorrow, but it cannot be put off in- “the RFT approach invokes a purely functional definitely. If critics believe that relational op- concept of an operant” (p. 146) rather than one erants are conceptually impossible, they must that relies on topography, but this is not any- say why. If they believe it is an empirical ques- thing new. We use terms like “generalized op- tion and the question has not yet been an- erants” or “overarching operant class” to em- swered, they should be able to say precisely phasize the functional nature of the operant but why the existing data are insufficient and pre- we are also careful to note that “The concept cisely what additional data are still needed. of a response class with an infinite range of topographies is a defining property of operant A New Behavioral Principle? behavior, and has been from the very begin- ning (e.g., see Skinner, 1938, p. 33-41).” (p. One basis for avoiding this challenge is the 147), and that “No new type of operant is sup- suggestion that new principles are implicated posed by these terms—the qualifiers are merely by RFT, but closer examination shows that this to avoid confusion.” (p. 147). Several of the cannot be used to delay a decision about the commentators note that we appeal to a special central claim of RFT. In several places in the kind of operant, which is true, but only in the current reviews it is said that RFT theorists sense that we define the relational operant. That appeal to new principles to explain verbal phe- operant is not special because it is nomena. For example, Malott says that we seem “overarching”—as Spradlin correctly notes, to “feel it more parsimonious to explain a re- this qualifier is just a way to emphasize sult in terms of one new principle of behavior Skinner’s functional definition of the operant— than two existing principles” adding “My con- rather is it special because of its particular func- cern about this new principle is that it seems to tion altering features. That, however, is a re- REPLY TO COMMENTARIES 41 sult—the process that is supposed to establish claim is both empirical and conceptual. those features is not special in any way. In the book an example is offered: For behavior analysts, claiming that arbitrarily applicable relational responding is op- A discriminative stimulus is a stimulus in the erant behavior is an explanation. The explana- presence of which there has been a greater prob- tion could be empirically wrong, of course, but ability of reinforcement for a given behavior to criticize such a claim on the grounds that it than in its absence. Suppose a child is rewarded is “merely a description” (Malott) should be for waving when the word “dog” is heard. The word “dog” is a discriminative stimulus. Sup- rejected by behavior analysts on pose, however, that the child is now taught to metatheoretical grounds. Behavior analysts say “dog” given the word D-O-G, and to point describe contingencies and their results. This at actual dogs given D-O-G. Suppose that as a is explanation for behavior analysts, and claim- result of this training the child now waves upon ing that something is an operant is a testable seeing a dog. Such an outcome has repeatedly claim about process. been seen in the literature (e.g., Hayes et al., Salzinger approaches this same type of criti- 1987). The dog cannot be a discriminative cism in the following quote: stimulus because the child has no history of greater reinforcement for waving in the pres- What is reinforced then is the relational re- ence of dogs than in the absence of dogs. The sponse. Sidman’s (2000) analogy is at least effects cannot be stimulus generalization be- equally attractive, viz. p. 144). “An equivalence cause there are no formal properties that are relation can be thought of as a bag that con- shared between the word and actual dogs. The tains ordered pairs of all events that the con- effect cannot be due to classical conditioning tingency specifies; the bag can be shaken and because it would require an appeal to backward the elements mixed without regard to any spa- conditioning. The effect cannot be due to com- tial or temporal relations among them.” That pounding because “dog” and dogs have not leaves unspecified what the mechanism is that even occurred together. seems to transcend spatial and temporal relations. (p. 8) Relational Frame Theory suggests that the performance is due to a learned process that trans- It is not clear why Salzinger believes that formed these discriminative functions. In nor- Sidman’s “bag theory” is equally attractive. mal discriminative control, the stimulus func- Sidman’s bag theory could apply only to tion is learned, but not the process itself. In equivalence, not to all of human language, and contrast, the derived performance is discrimi- it is at best a behavioral result, not a process. native-like, but it is not discriminative. These discriminative-like effects seem to depend on He is right to note that Sidman’s idea leaves a learned process of altering behavioral pro- unspecified what transcends spatial and tem- cesses, and that is something that is not cov- poral relations, but RFT is clear on this point. ered by an existing technical term. Despite the RFT claims that what transcends spatial and conservatism of an RFT approach, therefore, temporal relations is the relational operant it- ... in our analysis, verbal events (and relational self. This is a process claim and it points di- frames) instantiate a newly identified behav- rectly to what is new in an RFT account. ioral process (Hayes and Hayes, 1992). (p. 46) RFT claims that: If operants exist that are focused on and al- the instrumental behavior of relational fram- ter how behavioral processes themselves func- ing alters the functions of behavioral processes tion, then we need a new term because no ex- ... If Relational Frame Theory is correct, the isting term will do. This is a simple conceptual alteration of these behavioral processes was claim, and we see no logical flaw in it. But the itself a learned process. Said another way, re- “if ... then” nature of the claim clearly sepa- lational framing is operant behavior that affects rates the issue of implications (the “then” part the process of operant learning itself. We know of the statement) from the issue of process (the of no term for such an effect. (p. 45) “if” part of the statement). Let’s dream up a new operant to show the This quote shows clearly that we use a “new point in another area. Suppose it were possible principle” only as a label for a new effect, not to combine operant and classical conditioning for the processes that created that effect. This in the following way. Imagine that an animal 42 HAYES, BARNES-HOLMES, AND ROCHE is presented with several CS–UCS pairings. pacting on classical conditioning in this way Subsequently, the UCS is presented alone and because the scientist is “making up new prin- reinforcement is delivered contingent upon ciples.” The account did no such thing. The whether the animal responds in a way that was “new principles” were needed to describe the originally evoked by the CS. For example, in previously unknown impact of known prin- our imaginary experiment, reinforcers would ciples, thus making the issue more empirical be delivered to Pavlov’s dogs if they showed than theoretical. It is certainly not a mere philo- some minimal auditory response when pre- sophical matter, as if the data on this new op- sented with food powder. Over scores of tri- erant could be rejected on the basis of broad als, varying in the magnitude of the response philosophical or terminological preferences. required and varying across different specific This parable parallels rather precisely the kinds of CS and UCS events, imagine that this current situation with regard to RFT. It helps operant contingency trained our subjects in explain why we believe that behavior analysis some contexts to show strong CS related re- cannot walk away from this challenge un- sponses to the UCS following any new set of scathed. If relational operants of the sort RFT CS–UCS pairings. Now imagine that those imagines exist then behavior analysis is in a animals who have been through our training new era. For that not to occur, relational oper- are, unbeknownst to an experimenter, mixed ants must either be shown to be incoherent in in with naive subjects in a classical condition- principle, or not to exist empirically, or for the ing study examining backward conditioning. implications to be incoherent in principle, or Two very different response patterns result. for the implications not to exist empirically. Some subjects (the naive ones) show little or Any empirical or conceptual resolution will be no CS-related responses to the UCS. Some positive for behavior analysis if it is true to this subjects (our trained subjects) show strong and own approach. If, however, the field simply reliable responses of this kind. If no one knew ducks the empirical and conceptual challenge of the different histories in our two sets of sub- presented by a major theory within its ranks jects it might be reasonable to conclude that by (falsely, we believe) treating this challenge “sometimes classical conditioning leads to as if it is mere a matter of preference, termi- strong CS-related responses to the UCS but we nology, or philosophy, the field will be risking do not know why it sometimes does and some- its soul in the name of keeping its identity. RFT times does not.” It would be scientifically irre- is simply too well established as a behavior sponsible to make this same claim if the differ- analytic theory to walk away from in that way. ent histories were known, however. Suppose that someone suspected the true Relational Responding as an Active Process source of the effect, and conducted a series of experiments that showed that operant histories Another putative reason for avoiding the could indeed have an effect of this kind. It challenge that Relational Frame Theory pre- would be perfectly proper and scientifically sents to the field of behavior analysis is the conservative to ask for evidence on this point. suggestion that the concept of arbitrarily ap- But suppose that the scientist pointed out, very plicable relational responding, as an active reasonably, that if operant contingencies could process, does not fit with traditional behavior- have this effect on classical conditioning we analytic thinking. McIlvane summarizes this would need a new term to talk about it. Purely argument as follows: descriptive terms like “inverse CS–UCS relations” would not do because they merely described the result, not the source of the result. Another major departure from the abstraction Classical conditioning terms such as “backward analysis is that RFT views relational learning classical conditioning” would not do because as behavior. One does not merely exhibit rela- they would imply that the process involved was tional stimulus control. One “relates.” By contrast, the traditional behavioral analysis of ab- classical conditioning per se, whereas the key straction has been the narrowing of stimulus process was operant and only the preparation control. Thus, relational responding in RFT was classical. If a new term were used, how- seems to be an active rather than a passive pro- ever, it would not be “conservative” for scien- cess. This characteristic has set many tradition- tists to reject the idea of such an operant im- ally trained behavior analysts to head scratch- REPLY TO COMMENTARIES 43

ing as they tried to fit the theory within a be- allows it to discriminate between the relevant havior analytic framework.” (p. 30). features of the task (responding to one event in terms of another based on a contextual cue) The main difficulty here is that in behavior and the irrelevant features (the actual physical analysis we do not normally separate the dis- properties of the related objects).” (p. 26) crimination from the response. When a rat presses a lever in the presence of a green light The final clause, in parentheses, serves to this is the discriminative response—we don’t highlight that the traditional definition of rela- say that the light was discriminated and then tional stimulus control and abstraction does not this caused the rat to press the lever. apply to RFT—in fact, the abstraction in RFT Behavior analysts are suitably wary of us- involves learning in some contexts to ignore ing behaviors as causes of other behaviors. the physical properties of the stimuli as the Focusing on behavior–behavior relations can basis for a relational response. easily work against more complete experimen- An arbitrarily applicable relational response tal and applied analyses in which current and is not, therefore, a direct response to some com- historical contextual variables are identified mon physical property across two or more and manipulated to achieve the scientific goals stimuli. Rather, it involves the discrimination of prediction-and-influence (Hayes & of a contextual cue that brings to bear a rel- Brownstein, 1986). In the context of arbitrarily evant history of non-arbitrary and arbitrary re- applicable relational responding, however, ac- lational responding. In this sense, the cue tive terminology highlights rather than ob- causes an active relational response. McIlvane scures the historical and current contextual is correct when he states, “Relational frame variables responsible for this type of relational theory might be briefly summarized as an ex- control. This is a subtle but fundamentally im- tension of the traditional behavior analytic ac- portant issue for RFT and warrants further count of abstraction” (p. 29). However, this ex- elaboration. tension is substantive and RFT should not, Relational stimulus control in behavior therefore, be confused with relational stimu- analysis is typically thought of as a process of lus control as traditionally defined. Interest- ingly, Osborne appears to agree that there has abstraction [that] may develop when a set of been some confusion in this regard; “Behavior otherwise physically different discriminative analysts have probably used the phrase rela- stimuli have a physical property in common. tional responding to denote the act of relating For example, having learned via discrimina- but didn’t realize at the same time that they tion training to respond differentially to a red were speaking of relational framing” (p. 23). flower, a red car, a red ball, and other red items, From our perspective, therefore, the the learner may subsequently respond differ- overarching and extended history of contextu- entially to other red items without the need for ally controlled non-arbitrary and arbitrary re- explicit discrimination training. (McIlvane, p. 29, emphasis added) lational responding that is required for relational framing to develop is so important that In effect, the traditional definition of rela- it seems wise to highlight this feature of RFT tional stimulus control in behavior analysis by referring to relational framing as an active draws heavily upon the abstraction of common or behavioral process. physical properties, and in one sense the or- It is worth noting that this way of talking ganism that learns to abstract such properties could also be applied to traditional discrimi- is responding directly to them. Arbitrarily ap- nation learning. The classic perception study plicable relational responding, however, does by Held and Hine (1963), using new-born kit- not involve direct stimulus control by non-ar- tens, demonstrated that even seeing vertical bitrary physical properties; instead the rela- lines can be conceptualized as operant behav- tional response is arbitrarily applicable and ior (see Catania, 1998). However, because the under the control of a contextual cue or cues. operant of seeing is established so early and so In the book, we say it this way: thoroughly in the behavioral history of the organism, it is easy to forget that the first part of In order to abstract the behavior of relating, any visually based discrimination response in- the organism must be exposed to training that volves the operant of seeing the stimulus. On 44 HAYES, BARNES-HOLMES, AND ROCHE balance, there are no good pragmatic reasons tific simply because they will not necessarily for remembering this fact (at least in behavior generalize across tips of evolutionary branches, analysis), and thus we rarely say that the his- any more than, say, the finding that operant tory of reinforcement for seeing stimuli leads conditioning does not apply to bacteria would to an active discriminative response. For arbi- limit the validity of such findings in birds and mammals. All that is needed is that behavioral trarily applicable relational responding, how- researchers must not be so wedded to this stra- ever, the slide into traditional ways of thinking tegic assumption that alternative strategies can- about relational stimulus control (i.e., in terms not be pursued. of common physical properties) is so steep that, at least for now, it seems wise to differentiate Relational frame theory is oriented toward hu- it in this way—as an active behavioral process man language and cognition. Whether the rich- involving an overarching operant history that ness and complexity of human and language in principle could stretch back for decades. cognition may yet be shown to be in the same functional class as behavior studied with other The Issue of Continuity organisms is yet to be determined. Only empirical research, not assumptions, will resolve this issue. This is precisely the view adopted Another issue raised by some of the com- by RFT (p. 145). mentators that seems to have caused some con- sternation is the relationship between Rela- By adopting an agnostic position with respect tional Frame Theory and the continuity as- to the continuity assumption, RFT, contrary to sumption. For example, McIlvane asks “Why Osborne’s claim, is consistent with Skinner’s do relational frame theorists not more directly early thinking on this issue. Within behavior consider arguments that behavior may be the analysis, we point out that product of phylogenic contingencies” (p. 35); and Osborne states, “It is at this level, that of The traditional focus on nonhumans was based continuity or discontinuity between homo sa- on the idea that the principles of behavior iden- piens and all other animals, that RFT wishes to tified with such populations would be gener- divide from Skinner, and this is potentially a ally applicable to humans.... This form of the huge divide” (p. 23); and Spradlin notes “Their continuity assumption was a strategic assump- position implies discontinuity between the be- tion: it was a means to an end. It was not a havior of humans and nonhumans” (p. 3). Im- categorical assumption—that is, one that is fun- damental to the conceptual coherence of the plicit in McIlvane’s comment, and explicit in field. This is why Skinner warned that “We can both Osborne’s and Spradlin’s, is the view that neither assert nor deny discontinuity between RFT is self-consciously opposed to the conti- the human and the subhuman fields so long as nuity assumption. we know so little about either” (p. 442, 1938), The role of this assumption in behavior and that “It is possible that there are properties analysis, and its relationship to RFT is a com- of human behavior which will require a differ- plex issue (Hayes, 1987a, 1987b; see Dymond, ent kind of treatment” (p. 442, 1938). (p. 145) Roche, & Barnes-Holmes, in press, for a detailed discussion), but the RFT position was It should be emphasized that the “continuity clearly stated in the book—the theory remains assumption” we are speaking about is the stra- agnostic: tegic research assumption made by early behavior analysts: it is not the one that built in evolutionary theory: Arbitrarily applicable relational responding occurs readily, even with human infants and The evolutionarily sensible form of continuity with difficulty or not at all with nonhumans. assumes that new contains old. There is no After thirty years of behavior analytic research guarantee or assumption that old contains new. on derived stimulus relations, that statement is Biological evolution itself would be turned into still true. We do not need to take the stand that nonsense by such an assumption. When we nonhumans will never show derived stimulus look across tips of evolutionary branches we relations in order to begin to launch an exten- are not looking back in time: we are always sive and coherent program of basic research looking both back in time (to the point at which into these processes in human beings. The find- specific species differentiated) and forward in ings will be no less useful and no less scien- time to the present. Thus, discontinuity across REPLY TO COMMENTARIES 45

present day species would in no way contra- and-influence and avoid getting sidetracked dict a biologically sensible form of the conti- into well-worn and long since barren furrows nuity assumption. (p. 145). of scientific debate concerning the nature–nature controversy. How many animal studies will Adopting an agnostic position with regard have to be conducted before negative evidence to the continuity assumption does not imply is taken as final? (cf. Lionello-De-Nolf & that RFT underestimates the importance of Urcuioli, 2002). And if evidence is finally ob- phylogenic contingencies. Relational frame tained that proves unequivocally that a non- theory, like any behavior analytic account, as- human can demonstrate the basic equivalence sumes that all learned behaviors have their ul- relation, what then? How important will this timate origins in phylogenic contingencies. research be for mainstream psychology, or sci- Furthermore, RFT assumes that these contin- ence more generally? As behavior analysts, we gencies may be partly responsible for derived are at a crucial stage in our history—we need relational responding: to choose our research agendas very carefully and not waste time on wild goose chases (see If ... language and cognition will not yield en- Malott, 1991). tirely to principles derived from nonhuman research, it means simply that processes emerged in one evolutionary branch and not another. Parsimony What was new could be extremely small, and yet produce huge differences in behavioral out- Another issue that appeared to be of con- comes. Metaphorically, a person standing at the cern to some of the commentators was an ap- edge of a cliff may step forward an inch and parent lack of parsimony in RFT. Malott, for fall hundreds of feet. The step was not large: example, suggests that his own chaining ac- only the outcome was large. count of symmetry and transitivity is more par- If this is the conclusion we are eventually driven simonious than RFT because the former is cast to, what kind of processes could produce such “in terms of the basic principles of behavior” an effect? The ability of a listener to derive a (p. 17) and McIlvane remarked “One problem bidirectional relation readily from multiple ex- that I had with RFT:ALC’s discussions of trans- emplars would be a ready nominee. (p. 146) formation of functions was the general lack of an obvious relationship to the basic processes RFT does not place phylogeny center stage of behavior analysis” (p. 33). in the theory, however. Focusing on phylog- One can only appeal to parsimony as a mea- eny, and more specifically on whether or not sure of theoretical elegance and viability when nonhumans are capable of demonstrating the two relatively adequate explanations co-exist. simplest forms of derived stimulus control, RFT has made significant progress by using seems counterproductive to the analytic goal particular behavioral concepts applied to an of the theory. McIlvane may well be correct in operant conception of relational responding. If arguing that such a focus could lead to “de- we are correct, relational operants require that bate” like that “paralleling similar discussions we rethink and some of our units of analysis. in psycholinguistics,” but RFT is chasing what The “natural lines of behavioral fracture” for we see as a far more important prize than mere verbal humans may be fundamentally differ- discussion. RFT aims to provide a naturalistic, ent from those that we find with non-verbal monistic, and purely functional-analytic ac- organisms. Without a clear focus on the alter- count of human language and cognition that natives, behavior analysis could well cut itself readily facilitates meaningful change within on Occam’s razor. these two domains in both basic and applied Several of the commentators suggested al- research settings. This is something that ternatives—either their own or others. In our psycholinguistics, and cognitive psychology view, however, none of the currently available more generally, has not achieved, and prob- accounts can readily describe, let alone explain, ably never will given their analytic goals and all of the data that has been generated in RFT research strategies. studies (see Barnes, 1994; Barnes & Roche If RFT is to make a contribution that other 1996; Barnes-Holmes, Healy, & Hayes, 2000; psychological traditions have not, we need to Barnes-Holmes, et al., 2001; Dymond & stay focused on the analytic goal of prediction- Barnes, 1995; Hayes & Barnes, 1997; Hayes 46 HAYES, BARNES-HOLMES, AND ROCHE

& Wilson, 1996; O’Hora, Roche, Barnes- basic stimulus functions? Surely not, since we Holmes, & Smeets, 2001). It is to these alter- can make up new relations at a whim (see page natives that we now turn. 40 in the book). At the present time, it is difficult to judge Sidman’s position on this, because Alternative Accounts the work on non-equivalence relations within that tradition is so sparse and theoretically con- One way of responding to the challenge pre- fused. On the one hand, for example, Carrigan sented by Relational Frame Theory is to ex- and Sidman (1992) stated “Difference, oppo- plore alternative accounts, and a number of the sition, and negation ... are never equivalence commentators have indeed suggested alterna- relations; they do not exhibit reflexivity, and tive views. We will list each of these here and need not exhibit transitivity” (p. 185), but respond briefly to what has been offered. Sidman has also argued that these other rela- Sidman’s Account. One of the commentators tions may be interpreted as equivalence rela- (McIlvane) suggests that there is little differ- tions under contextual control. In the book, we ence between Sidman’s treatment of equiva- quoted Sidman on this very point: “‘The fact lence relations and RFT: that a stimulus pair can be brought via contextual control into such differing relations as The main difference between Sidman’s think- same, opposite, different, and so forth, can be ing and that articulated in RFT:ALC is whether handled by any formulation of equivalence that complex forms of stimulus control are selected recognizes the role of context’ (1994, p. 561).” by or constructed via reinforcement contingen- We then went on to point out that “This sen- cies. Once the selectionist vs. constructivist issue is ultimately settled, I anticipate fairly few tence appears to be Sidman’s only treatment points of major disagreement between of multiple stimulus relations.” Sidman’s followers and the thinking repre- McIlvane, in his commentary, referred to re- sented in RFT:ALC (p. 33). search on ordinal classes as an example of Sidman’s approach to non-equivalence rela- Furthermore, as noted earlier, Salzinger ar- tions, but this work begs the same question: gued that Sidman’s “bag analogy” is at least are ordinal relations another biological given equally attractive as the concept of a relational or contextually controlled equivalence classes? frame. The key difference, then, seems to be Furthermore, in the book we point out that RFT that for Sidman, the contingencies select a has generated data that does not appear to yield phylogenically established basic stimulus func- to equivalence or ordinal-class based analyses tion, but for RFT the contingencies construct (pp. 59-62; see also Dymond & Barnes, 1995, learned or overarching operant classes. pp. 182-183). At the present time, therefore, it Theoretically, we don’t think this difference is extremely difficult to compare directly makes a difference—as argued in the previous Sidman’s account with that of RFT. The two section, all learned behaviors have their ori- models, as Spradlin points out, appear to call gins in phylogenic contingencies. We also for very different research programs, and pointed out, however, that in practical terms Sidman’s treatment of non-equivalence rela- the different foci on phylogeny versus ontog- tions, which are the sin qua non of RFT, re- eny may have massive implications for the re- main at best underdeveloped and at worst con- search programs that develop from the two fused and contradictory. theoretical perspectives. Interestingly, Spradlin, who devotes much of his commentary to com- Malott’s Chaining Account paring and contrasting Sidman’s approach with RFT, seems to agree with our view when he In the commentary provided by Malott an writes, “... the research agendas implied by the explanation of symmetry and transitivity is pre- two models are very different” (p. 3). sented in terms of a behavioral chain. The ex- A related problem in brushing over the dif- planation rests critically upon a vocal reper- ferences between the two accounts is that toire that is accidentally reinforced as a tact Sidman has focused almost exclusively on during the symbolic matching training (i.e., the stimulus equivalence, which he argues is a ba- echoic “Mark” is reinforced in the presence of sic stimulus function or biological given. But the photograph). Like Horne and Lowe’s what of other derived relations? Are these, too, (1996) naming hypothesis, therefore, Malott’s REPLY TO COMMENTARIES 47 account seems to require a vocal response or diational or Pavolvian account? We invite tact, which mediates the derived relational re- Malott and others to build out their theoretical sponse. Even the relatively limited data that is and empirical analyses to a level comparable currently available suggests that this type of to that of RFT to see if the alternatives can vocal chaining explanation is inadequate. achieve something that our theory cannot. This Shusterman and Kastak (1993), for example, would certainly invigorate the discipline of provided some evidence to suggest that exem- behavior analysis. plar training with a sea lion (without any differential name training) can produce symme- Attributes, Rules, and Logic try and transitivity in a matching-to-sample task. Furthermore, Lipkens, Hayes, and Hayes In the commentary provided by Salzinger, (1993) reported a longitudinal study in which seven formal attributes of verbal behavior were symmetrical responses between a heard name presented, with a view to seeking out the vari- and a seen object were observed in an infant ous characteristics that distinguish it from all who failed tests for differential naming. More other behavior. In doing so, Salzinger implies recently, Carr, Wilkinson, Blackman, and that this strategy provides a viable alternative McIlvane (2000) have reported equivalence in to RFT, which has involved identifying one learning disabled individuals with very limited characteristic. Later in his commentary language abilities (but years of experience with Salzinger also suggests that rule-governed be- MTS procedures). Finally, a current research havior and formal logic may provide an alter- project in the second author’s lab has obtained native explanation for derived relational re- good evidence for equivalence responding in sponding. In our view, these cannot be consid- a mute autistic child (but with an extensive his- ered viable alternatives to RFT. tory of matching-to-sample training). These Listing the formal attributes of verbal behav- types of findings clearly suggest that derived ior may be a useful exercise early in a research relational responding (at least simple forms) program, but such a list will never provide the may occur in the absence of an expressive vo- technical or functional-analytic definition of cal repertoire. Insofar as Malott’s chaining ac- verbal behavior that we have sought to develop count requires such a repertoire, his mediational in RFT. If we are to use the term verbal in a explanation for symmetry and transitivity technical sense in behavior analysis, we need seems untenable. to delineate it functionally from all other classes It is also worth noting that mediational ac- of non-verbal behavior. The list Salzinger pro- counts of stimulus equivalence have also been vides, like the definition by Hockett, may help around since the 1930’s, and they failed. orient us towards important functional proper- Sidman (1994), himself, abandoned them early ties, but it cannot substitute for a genuine func- on in the equivalence research program. Re- tional-analytic definition or analysis. In the cently, Tonneau (2001) has argued that we book we are precise in our definition of verbal should give the mediational concepts of Pav- behavior: lovian conditioning another go, but he also points out that his Pavlovian approach to any- Relational frame theory takes the position that thing beyond the most simple transfer perfor- derived stimulus relations constitute the core mances will be too complex to yield to experi- of verbal behavior. Verbal behavior is the ac- mental analysis (p. 123). tion of framing events relationally. Both speak- It may be possible to construct a mediational ers and listeners engage in verbal behavior. account of the human language and cognitive When a speaker frames events relationally and abilities that are covered in RFT. However, no produces sequences of stimuli as a result, the one has done it yet in over a century of media- speaker is engaging in verbal behavior. In more tional theorizing. If someone does, we will see lay terms, we say that the speaker is speaking if it moves behavior analysis significantly with meaning. If the same formal stimuli are produced but not because the speaker has closer to the goals of prediction-and-influence framed events relationally (e.g., when a parrot in the domains of human language and cogni- repeats what is said), then no verbal behavior tion in both basic and applied settings. Only is involved. Verbal meaning, in this approach, this outcome will decide which is the best op- is not a mental event, nor an inference, nor a tion to take—the RFT approach or some me- simple effect. It is a highly specified behav- 48 HAYES, BARNES-HOLMES, AND ROCHE

ioral process (see the earlier section on the na- healthy, to disagree with the RFT explanations ture of relational frames for that specification) that we offer of verbal behavior, rules, and (pp. 43–44). logic. Nevertheless, we also believe that it is behaviorally nonsensical to argue that non- Salzinger and others may not agree with the functionally defined concepts, such as rules and RFT functional-analytic definition of verbal logic, can be used to explain the functionally behavior, but in so doing it is incumbent upon defined concepts of RFT. Doing so is like ar- them to provide an alternative functional defi- guing that the concept of the operant should nition––providing a list of formal characteris- be explained using the lay concepts of inten- tics will not suffice. tion and purpose. Appealing to rule-governed behavior or formal logic as possible alternatives to RFT falls victim to the same problem. Both of these con- SUMMARY cepts do not have clear functional-analytic definitions, and thus they cannot be used as the basis for a behavior-analytic explanation. In As we looked over the various alternative short, rule-governance and formal logic them- accounts it became apparent that none of them selves require explanation, and it was for this attempt to cover the entire domain of human reason that we subjected them to RFT analy- language and cognition. Even Sidman’s treat- ses in the book. Once again, we were relatively ment of equivalence relations, which is now precise in our functional-analytic treatments of over thirty years old, is still focused almost these concepts. With respect to rule-governed exclusively on one particular derived relation behavior we suggested the following: and seems to have moved progressively further away from where it started—a functional In RFT there is a clear difference between non- analysis of human symbolic control (Barnes- verbal and verbal regulation. Rule-governed be- Holmes, Hayes, & Roche, 2001). Other pos- havior is a subset of verbal regulation. The term sible alternatives, such as Malott’s (and becomes more likely to be used when the ver- Tonneau’s) chaining and mediational accounts, bal antecedent is a relational network or a com- are likewise limited in their foci and have their parison of such networks, and especially when historical roots in analyses that have not comparison between a verbal antecedent and worked very well empirically or conceptually. the verbal construction of ongoing events is part Furthermore, none of the accounts, apart from of the source of control over behavior regulation. Rule-governance is also more likely to be Sidman’s, have a robust research program used when the non-arbitrary features of the linked to them. And the remaining alternatives, environment are abstracted and transformed, which appeal to formal definitions of verbal and when the verbal network is generally ap- behavior and to the concepts of rules and logic, plicable (p. 108). ipso facto, fail to provide genuine functional- analytic explanations. In contrast, RFT has And with respect to logic, we argued as fol- made significant progress, both conceptually lows: and empirically, in tackling human language and cognition from a purely functional-analytic Logic, from an RFT perspective, is in essence perspective. a relational activity that involves the derived Much more remains to be done. One direc- transformation of function in accordance with tion for future research could involve devel- multiple stimulus relations. Although specific oping another viable and data-driven functional examples of relational framing may sometimes appear logical (e.g., if A is the same as B and B account that attempts to grapple with the broad is the same as C, then A is the same as C), logic range of issues covered by RFT. We would does not provide an explanation for relational welcome any genuine attempt to step up to this framing. Instead, it is RFT that provides the challenge, which we believe would be an ex- basis for a behavioral explanation of logical tremely healthy sign for our field. reasoning, including instances in which individuals fail to reason logically (p. 191). More Minor Points and Misunderstandings

We should emphasize that we think it en- Before moving forward with the closing sec- tirely reasonable, and indeed intellectually tions of our reply, we feel that it will be useful REPLY TO COMMENTARIES 49 to clarify a number of minor points and ad- pate in a frame of opposite, and A is established dress possible misunderstandings that we iden- as a punishing stimulus, B may acquire a rein- tified in the commentaries. Because these are forcing function (i.e., if A and B are opposite mostly minor points of clarification, we will and A is bad, then B must be good). (pp. 31- make them relatively brief and refer the reader 33, 150). to the relevant pages of the book for further Resistance to the idea of a new principle is elaboration and detail. Furthermore, in the in- understandable, particularly the transformation terests of brevity, in general we will not pro- of function, because it seems to impact upon vide direct quotations from the commentators, every other known behavioral principle, and but simply identify the relevant individual(s) thus RFT is revolutionary in this sense and page number(s) from which the point was (Spradlin). For example, a respondently con- derived. ditioned appetitive CS can be transformed into The Distinction between a Contextual Cue an aversive CS if it comes to participate in the and a Discriminative Stimulus (McIlvane & appropriate relational frames. It is for this rea- Salzinger). Relational frame theory reserves the son, we suspect, that McIlvane, following a term contextual cue for those stimuli that con- similar line of thinking to that of Malott’s chain- trol (bring to bear) a history of arbitrarily ap- ing example, suggests that transformation of plicable relational responding. We use the term function may parallel “the concepts of behav- contextual cue, rather than discriminative ioral resurgence, spontaneous interconnection stimulus because; (1) it brings to bear a spe- of repertoires, and other similar phenomena” cific history of relational responding, and thus (p. 33). From this perspective, the transforma- the relational response cannot be traced directly tion of function, to be readily understood, to the physical properties of the stimuli, and should be broken down into smaller or more (2) a contextual cue can itself acquire its con- traditional analytic units. Of course, any be- trolling properties via relational framing, and havioral researcher is free to engage in this in- thus it does not fit the standard definition of a terpretive exercise (although we note that no discriminative stimulus (see pp. 25–27, 32–33, behavior analyst has attempted to undertake 62–64). this task in any serious way). As stated previ- Transformation of Function (McIlvane & ously, however, it is our position that the power Osborne). In RFT, the transformation of func- and utility of RFT is derived, in large part, from tion is the result of an overarching history of the way in which it reconceptualises the ana- arbitrarily applicable relational responding, and lytic units of operant psychology for the treat- this effect does not seem to fit with any of the ment of human language and cognition. If oth- traditional technical concepts in behavior ers feel that the concept of transformation of analysis. If three stimuli, A, B, C, participate function is too molar or descriptive (e.g., in an equivalence relation, and A is established Malott), it is incumbent upon them to develop as a discriminative stimulus through a history alternative, coherent treatments that prove to of differential reinforcement, given appropri- be equally, if not more productive than RFT, ate contextual cues both B and C may acquire in the empirical and conceptual domains cov- similar discriminative-like properties in the ered by RFT. absence of any direct reinforcement. When this Stimulus Generalization as an Agreed upon occurs, we say that the functions of the B and Process (McIlvane). Many of the core concepts C stimuli have been transformed in accordance in behavior analysis have extended histories with the mutually and combinatorially entailed of conceptual and empirical research attached relations. The term transformation denotes that to them with certain key issues remaining un- the previously “neutral” functions of the B and resolved, and as McIlvane correctly points out C stimuli have been changed or altered in a stimulus generalization is one of these. How- specific way based on an extended history of ever, the same could be said for so many other arbitrarily applicable relational responding, concepts, too, including the very definition of rather than through direct training, primary operant versus respondent behaviors (e.g., stimulus generalization, respondent condition- Donahoe, Burgos, & Palmer, 1993). To engage ing, and the like. Stimulus functions can also at this level in the RFT book, however, would be transformed in accordance with non-equiva- have undermined one of the primary purposes lence relations; for example if A and B partici- of the text—to broaden the behavior-analytic 50 HAYES, BARNES-HOLMES, AND ROCHE research agenda and to attempt wider commu- non-verbal behavior, which none of the com- nication with those who have virtually forgot- mentators picked up on. For example, in the ten that behavior analysis even exists (p. 1). book we distinguish between verbal and non- We do not suggest, however, that the study of verbal contextual control over arbitrarily ap- stimulus generalization (and numerous other plicable relational responding; in the former issues, too) are inherently unimportant and re- case, the contextual cue itself participates in a solved to everyone’s satisfaction—merely that relational frame, but in the latter case it does the broadly agreed-upon usage of such terms not (this distinction may have some bearing on in behavior analysis is sufficiently well estab- conditioning without awareness, the develop- lished to allow us to move forward with the ment of well-practiced skills, and the like; see RFT research agenda. pp. 88-91). In short, we need to distinguish be- Common Sense Examples of Relational Net- tween verbal and non-verbal contextual con- works of Familiar, Meaningful Stimuli Are Less trol over relational framing, and nonverbal Arbitrary (McIlvane). In the book we provide behavior (in which relational framing is com- a number of common-sense examples of rela- pletely absent). Where the boundaries lie and tional networks (i.e., based on the types of se- how one crosses from one domain to the other mantic relations found in natural languages). is largely an empirical matter. In fact, the talk The purpose of these examples was to illus- of boundaries might be misleading. It may be trate to the non-behavioral reader that the net- more useful and accurate to talk of degrees of works typically found in derived relations stud- “verbalness” in any given human action, rather ies could well provide a valid model of the than discrete categories of verbal versus non- types of relational networks found in natural verbal behavior. This would certainly be in language (e.g., pp. 52-57). To the average de- keeping with RFT’s contextualistic and devel- rived stimulus relations researcher these ex- opmental focus (p 151–154). amples may well seem course, but in the ser- vice of communicating with a wider audience we felt it was a useful rhetorical device. How- The Goal: An Account of Language ever, and this is extremely important, in no way did we imply that natural language relational One generic “misunderstanding” that seemed networks are somehow less arbitrarily appli- to emerge from time to time across some of cable than laboratory-induced networks com- the commentaries is particularly troublesom— posed of nonsense syllables and/or abstract the idea that RFT is a theory of stimulus equiva- stimuli. Indeed, we gave explicit examples to lence rather than a theory of human language the contrary (p. 55). To make such a claim and cognition. For example, McIlvane states: would undermine the RFT definition of verbal behavior itself (p. 43–44). Arbitrarily Applicable Relational Respond- Given the voluminous body of behavior analytic work on relational learning summarized ing Is All that Humans Do (Osborne). From in RFT:ALC, Horne and Lowe’s article, and the RFT perspective, it is almost certain that as Sidman’s (1994) book, one would expect there a history of relational framing increases, it will to have already published a comprehensive, predominate in the human behavioral reper- detailed treatment of theories of stimulus toire. In fact, many religions, some therapies, equivalence and related phenomenon. How- and the Acceptance and Commitment Therapy ever, such a treatment has not yet appeared, (Hayes, Strosahl, & Wilson, 1999) work in par- and the authors chose not to attempt one in ticular, are devoted, in part, to undermining ex- RFT:ALC. (p. 32, emphasis added) cessive verbal or arbitrary relational control in those domains in which it seems to create hu- Furthermore, as pointed out previously, man suffering (pp. 214–219, 244–251). For Spradlin devoted much of his review to com- RFT, therefore, humans can engage in both paring and contrasting Sidman’s (1994) book verbal and non-verbal behavior, although ex- with the current volume (although he clearly cessive levels of the former may be the basis recognizes the different implications arising for a wide range of psychological problems. It from the two works), and Salzinger compared should also be noted, however, that there are Sidman’s bag analogy for equivalence classes subtleties in distinguishing between verbal and with the RFT concept of the relational response. REPLY TO COMMENTARIES 51

This “equivalence-centric” approach to RFT, Nelson (1986) study has not been replicated, although inappropriate, is understandable given except for the ambiguous results of Barnes, the history of the theory and the research his- McCullagh, and Keenan (1990). First, we don’t tories of some of the commentators (two of agree that the Barnes, et al. results were am- whom are leaders in the field of stimulus biguous. The one child who failed to demon- equivalence). Nevertheless, some of the criti- strate equivalence also failed to demonstrate cisms voiced by the commentators can be symmetry between her expressive and recep- traced to their thinking of RFT as a theory of tive repertoires—the other children who passed stimulus equivalence, and thus in our view the equivalence test possessed both of these these comments seem off target. For example, repertoires. Second, we are not sure if this ques- both McIlvane and Osborne bemoan the lack tion is an extension of comments McIlvane and of engagement with, and references pertaining co-authors previously made (Carr et al., 2000) to, other behavior-analytic thinking on rela- that the training results for Devany et al. (1986) tional learning. In the Preface to the book, how- were unusual, or if McIlvane is speaking about ever, we pointed out that the “scholarly engage- the RFT-relevant testing findings from that ment” material could be found in various other study. If the former, we would note only what RFT articles, some of which contain detailed Carr et al. (2000) agree to: every component comparisons of RFT with alternative accounts of the training results have been shown else- (e.g., Barnes, 1994; Barnes & Roche 1996; where. If the latter, we would note that corre- Barnes-Holmes, Healy, & Hayes, 2000; lational studies of this kind are only one method Barnes-Holmes, et al., 2001; Dymond & useful in confirming a general relationship be- Barnes, 1995; Hayes & Barnes, 1997; Hayes tween derived stimulus relations and language. & Wilson, 1996; O’Hora, et al., 2001) and that Subsequently, experimental studies have pro- the broader purposes of the book would not be vided stronger systematic replications of the served by repeating these arguments in detail. RFT approach to this area. Some of this work Many of the arcane and abstruse “in-house” is described briefly in the book (p. 183–191), issues on relational learning and equivalence and other full research articles are either cur- seem unimportant as measured against the goal rently in preparation or under submission to of an account of human language and cogni- relevant journals. tion as opposed to a theory of stimulus equiva- We agree with McIlvane that longitudinal lence as such. studies with normally developing infants are The same general response applies to the required, but that applies with equal force to criticism that RFT has focused too much at- the theoretical ideas of other researchers in this tention on normal adults, without dealing with broad area. The only currently published study the relevant pre-experimental histories of this type was conducted in the RFT lab at (McIlvane; see also Spradlin). Because RFT is Reno; Lipkens, et al., 1993, and other RFT re- concerned with all areas of human language search is currently underway, including stud- and cognition, all humans, young and old, are ies using event-related potentials (ERPs) to get relevant and important for the research agenda. around the experimental problems associated The origin of stimulus equivalence is but one with behavioral procedures using infants. In very small part of RFT. Even if the equivalence our view, however, the true test of the approach research community manages one day to work is whether or not it leads to greater prediction- out whether or not equivalence is a behavioral and-influence across the entire domain of hu- primitive, we will need to understand how this man language and cognition. Human infant single behavioral repertoire feeds into the vast studies are only a small part of the research universe of human language and cognition. It agenda being pursued. is our view that science does not always do RFT, as the book title declares, is a theory of well by waiting until a specific and perhaps human language and cognition. In fact, from not very important question has been answered the language-and-cognition focused perspec- before moving onto important areas of inquiry. tive of RFT, one could turn the infant-studies A similar argument applies to the question argument on it head and ask why has there been concerning the early RFT correlational studies no published research conducted with senior using language-disabled children. For example, citizens. Our response is that we have plans in McIlvane asks why the Devany, Hayes, and that direction too, but the field clearly needs 52 HAYES, BARNES-HOLMES, AND ROCHE many more human operant laboratories to and b) to affirm that we believe that this re- mount the entire RFT research program. analysis can be done from within the behavior analytic tradition that Skinner established, de- Post-Skinnerian Behavior Analysis? spite the fact that the reanalysis leaves many of Skinner’s specific ideas behind. Some of the comments seem more concerned Interestingly, none of the commentators try with past identity than in assessing the chal- to show that the implications we see for spe- lenge of RFT per se, and particular commenta- cific Skinnerian concepts are incorrect. To the tors react strongly to the “post-Skinnerian” la- extent that the challenge is met, it takes the form bel we attach to Relational Frame Theory. For of suggesting that other concepts might account example, Osbourne says “As much as the au- for some small part of relational learning (e.g., thors of RFT appear to want to separate them- Malott, McIlvane, Salzinger). However, none selves from Skinner—witness their subtitle— of the authors argue that these alternatives there are interesting parallels that they cannot could account for the range of RFT data al- avoid” (p. 20). Osbourne and others spend time ready produced. Furthermore, the commenta- arguing that RFT owes much to Skinner’s tors sometimes seem to use traditional Skinne- work. rian analyses in ways that themselves would This is undoubtedly true, and hardly surpris- fundamentally challenge Skinner’s ideas if RFT ing. The book was written by behavior ana- data were explained by them. Salzinger argues lysts. Indeed, the book is dedicated to Skinner that RFT might be accounted for by rule-gov- and Kantor. Fortunately, we were very precise erned behavior, for example, but since human about what we meant by the term “post-Skin- infants show derived relational responding nerian”: (e.g., Lipkens, et al., 1993) this account would have the effect of claiming that human infants This book is behavior analytic but it is also post- with minimal verbal abilities are following Skinnerian. Calling it that is meant both to do rules—not a happy result for Skinner’s account honor to a great intellect of the twentieth cen- of rule governance. tury, and to acknowledge that this volume rep- resents a form of that approach that is quite different in many of the specific topical domains of interest to human psychology. If a CONCLUSION behavioral psychologist adopts a Relational Frame Theory perspective, the world of human events suddenly changes. A whole host of top- In considering the five commentaries to- ics that have been written about and studied gether, the following themes seem quite strong. within the behavioral tradition now are viewed There is general agreement about the need to differently.... [following a list of 19 items] deal with human language and cognition from When a list of this kind becomes long enough it is time to stop and note that something has a behavioral perspective. Furthermore, there happened. For us, this book is such an event. appears to be general agreement that RFT may (pp. 253-254) well have something of value to offer in this regard. The reviewers also recognise that RFT Based on what we actually say, there should is a behavioral account, and all of them pointed be no confusion about RFT as a “post-Skinne- to links with other areas of behavior analysis. rian” account. It is not an attempt to separate McIlvane and Spradlin see considerable over- from Skinner, nor merely to note, as Osbourne lap with Sidman’s work on equivalence classes, also suggests, that we are writing after Skin- Salzinger with rule-governed behavior, Malott ner is dead. It is not “anti-Skinnerian” or “anti- with his work on linguisitic productivity, and behavioral” (as one track of Salzinger’s word Osborne sees RFT as a natural extension of associations lead to). Spradlin has it right: the Skinner’s work. sub-title “suggests that while the account has However, the main criticism of RFT by three its roots in Skinner’s work, that it progresses of the commentators seems to be focused on beyond Skinner’s work” (p. 5). We use the term the way in which the theory draws on the con- “post-Skinnerian” a) to note that if RFT is cor- cept of the operant. Both Malott and McIlvane, rect, many specific Skinnerian analyses related for instance, call for an account in terms of ba- to complex human behavior need to be redone, sic behavioral processes. Malott attempts to REPLY TO COMMENTARIES 53 provide one example, in terms of stimulus-re- If the answer is “Yes” to each of these ques- sponse chains, and McIlvane mentions “behav- tions, then the field as a whole will have to ioral resurgence, spontaneous interconnection deal with the wide-ranging and revolutionary of repertoires, and other similar phenomena....” implications that arise from this relational op- Salzinger points to rule-governed behavior and erant. Behavior analysis will have unquestion- logic but does not attempt to explain either at a ably entered the “post-Skinnerian” era because, technical level. In contrast, Osborne seems sat- in effect, behavioral psychology will have to isfied with RFT as an operant account to such re-examine the impact of a wide variety of be- an extent that he argues that there really is noth- havioral processes in verbal organisms that ing very new here. In every case, however, there have hitherto provided the bedrock upon which is no attempt to deal directly with multiple our science is built. Stepping up to this chal- stimulus relations and the transformation of lenge is exactly what RFT attempts to do, but function that have been the central focus of RFT stepping up to the challenge of these five ques- studies for the past 13 years. When examples tions is something that the entire field of be- are given, such as that provided by Malott, or havior analysis can no longer comfortably the stimulus class definition provided by avoid. Osborne (p. 18), they concern only frames of coordination and thus can yield (in terms of REFERENCES description if not explanation) to class-based analyses. Barnes, D. (1994). Stimulus equivalence and We believe that it is telling that none of the relational frame theory. The Psychological reviewers attempted to provide a description, Record, 44, 91–124. let alone an account, of multiple stimulus rela- Barnes, D., McCullagh, P. D., & Keenan, M. tions—the very heart of RFT research. We have (1990). Equivalence class formation in non- been asking this question of our colleagues for hearing impaired and hearing impaired chil- some time now (e.g., Hayes & Barnes, 1997), dren. The Analysis of Verbal Behavior, 8, and few seem willing even to try, and then only 19–30. in the most minimalist fashion. For example, Barnes, D. & Roche, B. (1996). Relational Tonneau (2001) mentions the issue in relation frame theory and stimulus equivalence are to his account and then admits that he believe fundamentally different: A reply to that the domain is too complex to yield to an Saunders. The Psychological Record, 46, experimental analysis cast in his terms (see 489–508. Barnes-Holmes & Hayes, 2002). Barnes-Holmes, D., Healy, O., & Hayes, S. C. If behavior analysis, as a field, is to face the (2000). Relational frame theory and the re- challenge that RFT presents, the following lational evaluation procedure: Approaching questions will need to be answered. human language as derived relational responding. In J. C. Leslie and D. E. Blackman (Eds.), Experimental and applied analysis 1. Are we satisfied that an operant is a basic of human behavior (pp. 149–180). Reno, behavioral process? NV: Context Press. 2. If so, can we define and empirically iden- Barnes-Holmes, D., Hayes, S. C., & Roche, B. tify operant behavior? (2002). The (not so) strange death of stimu- 3. If we can, can we define and empirically lus equivalence. European Journal of Be- identify traditional relational responding havior Analysis, 2, 35–41. (based on formal properties of relata) as Carr, D., Wilkinson, M., Blackman, D. E., & operant behavior? McIlvane, W. J. (2000). Equivalence classes 4. If so, can we define and empirically iden- in individuals with minimal verbal reper- tify arbitrarily applicable relational re- toires. Journal of the Experimental Analy- sponding (as defined in our book) as op- sis of Behavior, 74, 101–114. erant behavior? Carrigan, P., & Sidman, M. (1992). Conditional 5. If so, is it the case empirically that this re- discrimination and equivalence relations: A lational operant modifies stimulus func- theoretical analysis of control by negative tions established by other learning pro- stimuli. Journal of the Experimental Analy- cesses? sis of Behavior, 58, 183–204. 54 HAYES, BARNES-HOLMES, AND ROCHE

Catania, A. C. (1998). Learning (4th ed.). Up- visually guided behavior. Journal of Com- per Saddle River, NJ: Prentice-Hall. parative and Physiological Psychology, 56, Devany, J. M., Hayes, S. C., & Nelson, R. O. 872–876. (1986). Equivalence class formation in lan- Horne, P. J., & Lowe, C. F. (1996). On the ori- guage-able and language-disabled children. gins of naming and other symbolic control. Journal of the Experimental Analysis of Journal of the Experimental Analysis of Be- Behavior, 46, 243–257. havior, 65, 185–241. Donahoe, J. W., Burgos, J. E., & Palmer, D. C. Lakatos, I. (1970). Falsification and the meth- (1993). A selectionist approach to reinforce- odology of scientific research programmes. ment. Journal of the Experimental Analysis In I. Lakatos & A. Musgrave (Eds.), Criti- of Behavior, 60, 17–40. cism and the growth of knowledge (pp. 91– Dymond, S., & Barnes, D. (1995). A transfor- 196). Cambridge: Cambridge University mation of self-discrimination response func- Press. tions in accordance with the arbitrarily ap- Lionello-De-Nolf, K. M., & Urcuioli, P. J. plicable relations of sameness, more-than, (2002). Stimulus control topographies and and less-than. Journal of the Experimental tests of symmetry in pigeons. Journal of the Analysis of Behavior, 64, 163–184. Experimental Analysis of Behavior, 78, 467– Hayes, S. C. (1987a). Upward and downward 495. continuity: It’s time to change our strategic Lipkens, G., Hayes, S. C., & Hayes, L. J. assumptions. Behavior Analysis, 22, 3–6. (1993). Longitudinal study of derived stimu- Hayes, S. C. (1987b). Language and the in- lus relations in an infant. Journal of Experi- compatibility of evolutionary and psycho- mental Child Psychology, 56, 201–239. logical continuity. Behavior Analysis, 22, Malott, R. W. (1991). Equivalence and rela- 49–54. tional frames. In L. J. Hayes & P. N. Chase Hayes, S. C., & Barnes, D. (1997). Analyzing (Eds.), Dialogues on verbal behavior (pp. derived stimulus relations requires more 41–44). Reno, NV: Context Press. than the concept of stimulus class. Journal O’Hora, D., Roche, B., Barnes-Holmes, D., & of the Experimental Analysis of Behavior, Smeets, P. M. (2001). Response latencies 68, 265–270. to derived stimulus relations: Testing two Hayes, S. C., & Brownstein, A. J. (1986). Men- predictions of relational frame theory. The talism, behavior-behavior relations and a Psychological Record, 52, 51–75. behavior analytic view of the purposes of Schusterman, R. J., & Kastak, D. (1993). A science. The Behavior Analyst, 9, 175–190. California sea lion (Zalophus californius) Hayes, S. C., & Hayes, L. J. (1992). Verbal is capable of forming equivalence relations. relations and the evolution of behavior The Psychological Record, 43, 823–839. analysis. American Psychologist, 47, 1383– Sidman, M. (1994). Equivalence relations: A 1395. research story. Boston, MA: Authors Co- Hayes, S. C., Strosahl, K. & Wilson, K. G. operative. (1999). Acceptance and Commitment Sidman, M (2000). Equivalence relations and Therapy: An experiential approach to be- the reinforcement contingency. Journal of havior change. New York: Guilford Press. the Experimental Analysis of Behavior, 74, Hayes, S. C., & Wilson, K. (1996). Criticisms 127–146. of relational frame theory: Implications for Skinner, B. F. (1957). Verbal behavior. New a behavior analytic account of derived York: Appleton-Century-Crofts. stimulus relations. The Psychological Tonneau, F. (2001). Equivalence relations: A Record, 46, 221–236. critical analysis. European Journal of Be- Held, R., & Hein, A. (1963). Movement-pro- havior Analysis, 2, 1–128. (Includes com- duced stimulation in the development of mentary.)