key: cord-1043392-h5f43jl9
authors: Dixon, Mark R.; Belisle, Jordan
title: Selection by Consequence: A Response to Hayes & Fryling (2019)
date: 2020-10-13
journal: J Contextual Behav Sci
DOI: 10.1016/j.jcbs.2020.10.003
sha: 1b1041127b8c6fa685a92b0c0f714b3de6d48aeb
doc_id: 1043392
cord_uid: h5f43jl9

The paper by Hayes and Fryling (2019) seeks to inform readers that the Kantorian system of Interbehaviorism has been misunderstood and misrepresented by contextual behavior scientists. Furthermore, these authors suggest that much is to be gained by embracing the system developed by Kantor, most importantly that being large scale system building efforts. We disagree with this position, and find the Kantorian system to be of questionable relevancy and at risk of potential extinction within the behavioral community. We also have concerns that perhaps the insights provided by Hayes and Fryling will fail in recruitment of additional members to the Interbehavioral cause. Although the overarching theme of Interbehaviorism is present within emerging dynamical approaches to behavior science, adopting the vernacular of Kantor may be unnecessary to continue his tradition. We recommend allowing empirical selection to run its course in determining the eventual fate of Kantor’s Interbehaviorism.

control behavior, the IBF is made up of interacting components that together produce psychological events. Three major components of the IBF are the contact medium, setting and setting events, and the interbehavioral history. Although these components are described as occurring within the IBF, the IBF cannot be reduced to these individual parts due to complex interactions at lower levels of analysis. This is not unlike the three-body problem facing physicists such as Newton and Laplace and eventually "solved" by Poincaré at the end of the nineteenth century (Chenciner, 2015) . The three-body problem occurs when attempting to predict the orbital behavior of three planetary bodies. Although prediction of two bodies can be easily achieved, the addition of the third body produces sufficient complexity at lower levels such that making predictions given knowledge of the component parts of the system becomes impossible. Poincaré proposed, however, that prediction may be forthcoming by treating the movement of the three bodies as one singular higher-order event that paved the way for research on dynamical systems and chaos. Unlike dynamical systems theories that develop quantitative and testable models of the dynamic and chaotic evolution of complex systems, Kantor's Interbehaviorism provided only a narrative description of the IBF. Irreducibility does not and should not equate to non-testability if attempting to remain within scientific discourse (Popper, 1963) . To the contrary, because of the chaotic interaction of events at lower levels of analysis within physics models, analysis or testability of those lower level events is immensely challenging except for within highly controlled contexts where variability is artificially constrained. To achieve testability requires focusing on higher-order patterns of the system, and in so doing, generating testable hypotheses about the evolution of that system. Irreducibility was already a major aspect of Einstein's theory of relativity when Kantor developed his IBF explanation for psychological events. This approach represented at-best a J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 7 metaphorical extension of other field theories that were "in the air" at that time. Because of the irreducibility problem, higher-order evolving patterns in systems are examined and used to make predictions, however no clear higher-order patterns are quantitatively described within Kantor's work in such a way as could be empirically tested. Indeed, as noted by Hayes and Fryling (2019), Kantor's account is too complex for such an experimental analysis, but what does that leave in terms of empirical strategy to vet this or that theory scientifically? Other behavior scientists and developmental psychologists, however, took the work being done on dynamical systems and applied it directly to behavioral events in order to deal with system complexity. Thalen (1985) developed several studies to show how infant motor movement develops dynamically, resulting in gross individual differences that are self-organizing and highly susceptible to small differences at initial conditions. Kelso (1985) has since extended this work by looking at the dynamical selforganization of behavioral and neural pattern generation, showing multistability and bifurication.

In both lines of research, interdependent mechanisms allow for testable predictions and have been used to describe the considerable individual differences observed across people. In more recent applications of dynamical systems modelling, this can be achieved by using artificial intelligence and running simulations until stable predictions are achieved (Bruzzo & Vimal, 2007) . Using Artificial Intelligence in Medical Epidemiology (AIME), researchers were able to predict the outbreak of the COVID-19 virus with up to 88 percent accuracy in Malaysia and Brazil (Allen, 2020, Medical Expo) . Undoubtedly, a virus outbreak is maintained and propagated by very many lower-level events that interact in complex ways. And, predicting who will be infected and at what time is likely impossible. However, AIME like other AI systems are designed to detect complex higher-order patterns by making predictions and learning from errors, not unlike operant learning. In so doing, the AI system becomes increasingly more J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 8 accurate. Self-evolving models such as AIME extend directly from dynamical systems theory (Devaney, 2018) . Kantor's work and the emergence of dynamical systems theories were occurring simultaneously, however the testability of dynamical systems compared against an unclear link to empirical testing for Kantor's IBF may explain why behavior scientists have avoided the IBF as a possible explanation for behavior, yet new theories are incorporating the work of dynamical systems modelling.

In our own work, we have examined the apparent self-organization of verbal relational behavior from an RFT perspective consistent with Relational Density Theory (RDT; Belisle & Dixon, 2020, in press). RDT provides quantitative predictions about the self-organization of relational frames, allowing for the model to be refined in light of new empirical evidence. RDT attempts to explain higher order evolving properties of relational networks quantitatively in order to make predictions that can be falsified empirically and ultimately to refine the model. strength of individual relations (i.e., distance between stimuli), and the distance between classes which may provide a quantitative description of coherence (i.e., pre-experimental relatedness of two or more classes). Within RDT, relational gravity proposes that classes with greater mass (Rm) and with smaller distances between them (Rd, relational coherence) are more likely to merge, expressed by the equation: F = (Rm1 * Rm2) / Rd. Indeed, results showed a successful class merger for a "coherence" group where the targeted merger were classes that were preexperimentally close compared against a "non-coherence" group where the merger targeted classes that were pre-experimentally distal. We provide RDT not to suggest superiority to Kantor's IBF, as indeed relational behavior comprises only a small part of the totality of the IBF.

Rather, RDT was developed from existing, testable ideas such as BMT and dynamical systems theories, and thus it should be concerning that other novel approaches designed to deal with the modern challenges of complexity may remain uninfluenced by Kantor. Yet we must disclose that we have read Kantor's work prior to RDT development, leading to the possibility that maybe at an implicit level, there could have been a slight influence by Kantor, as our history now included this content. Indeed, the first author of the present paper has engaged in considerable discussion of Kantor and was the graduate advisor of the second author, both of whom developed RDT. The degree to which this discussion contributed or was needed for the development of a dynamical model cannot be ascertained nor assumed; however, Kantor's assumptions persist within the surviving vernacular of other scientific approaches (e.g., density, volume, gravity, etc.).

Kantor's Interbehaviorism is often thought as an alternative to the radical behavioral approach. Even Skinner (1989) himself commented "why have they (interbehaviorists) not built a nest of their own to lay their eggs in?" We believe that interbehaviorism really should not be J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 10 considered an entirely unique idea or approach within the broader scientific community, honestly may not have even been that novel within the radical behaviorist community. Although the initial writings of Kantor historically precede those of Skinner, the repeated beckoning for the radical behaviorists to pay attention to Kantor came well after both scholars had become contemporaries. As noted elsewhere by Skinner, "… any unit of operant behavior is to a certain extent artificial. Behavior is the coherent, continuous activity of an integral organism. Although it may be analyzed into parts… we need to recognize its continuous nature in order to solve certain common problems" (Skinner, 1953, p. 116) . The seemingly discrete units of the threeterm-contingency may therefore be best viewed as a tool to isolate specific mechanisms for targeted intervention. By doing so, the behavior scientist can make testable predictions about how manipulation of any single part of the complex and continuous environmental event will influence behavior. Perhaps this is why Hayes and Fryling (2019) claim that the methods used for experimental work by the Interbehaviorist are the same ones used by the radical behaviorist.

If so, then what is the difference? If nothing, then what is the utility of adopting the more complex model and vernacular with a potential loss in empirical testability? As we describe below, the lack of testability in Kantor's Interbehavioral approach creates the real risk of problems within a scientific account of human behavior.

Fryling. In their current paper the authors make the claim that perhaps Kantor's most important contribution to the field of behavioral psychology is that of the notion of stimulus function. Kantor discusses, and these authors recollect, how stimulus objects as physical things may acquire interbehavioral functions given interaction between the organism and these objects. Once such interaction has occurred, the objects themselves become more complex, as they now possess additional psychological J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 11 functions. Take for example a rock that is initially nothing more than a stimulus object in the world. Upon approaching by a person, the interbehavioral event between the rock and the organism occurs, which renders the rock as now containing stimulus function. In this case the rock was needed to keep a breeze from blowing closed an office door. This rock, is now functionally altering the world, and may be thought of possessing a stimulus function. With additional interaction, this same person using the rock to block the door closure may engage in additional interbehavior with the rock, as it is gray like their father, who was considered the "rock" of the family. Now, upon glancing down to see the rock, memories of childhood start to occur. While a provocative interpretation, the same sequence of events can be explained through a Skinnerian system using stimulus discrimination and perceptual behavior (seeing in the absence of the thing seen), or placed into layman terms to aid in ease of discourse. In the latter dialogue, the rock reminds them of their father because it is grey and strong, just like their dad.

Parsimony of explanation is perhaps not a universally embraced value within philosophy, yet when seeking explanations that can eventually be put to the empirical test, parsimony tends to matter.

Arguably, "stimulus function" was the point of emphasis of work based on respondent conditioning of Pavlov and Watson. Generally, "functional" approaches in science describe how one event causes (elicits or evokes) another event. Within the basic respondent model, a conditioned or unconditioned stimulus elicits a conditioned or unconditioned response, and as noted by Watson (1920), higher-order conditioning across the lifespan of an individual is likely immensely complex. Indeed, the "Activity Stream" proposed by Watson highly resembled an behaviors. For example, a mother that provides nutrients and warmth becomes a conditioned stimulus eliciting appetitive emotional affect in the baby. Should this occur very early in life (a bifurication within the complex system), stimuli subsequently paired with the mother are more likely to obtain those same eliciting functions. If, however, nutrients and warmth are not provided early, those same functions may fail to transfer leading to complications later. Small changes at initial conditions influence the evolution of the system considerably, just like in a dynamical system. This model was entirely based on the eliciting effects of stimulus function.

Skinner extended this account primarily by introducing the idea of the operant and therefore behavior function -focusing on changes in the environment produced by behavior.

However, a critical aspect of this theory is that reinforcement alters the evocative potential of stimuli in the environment -reinforcement alters stimulus function. At the most basic level is the three-term contingency, however Skinner too discussed the potential complexity and interaction of many environmental and behavioral events that contribute to stimulus control. For example, when examining private verbal behavior, the speaker and listener are contained within the same skin, where response evoking stimuli are produced by the person themselves (Skinner, 1957) .

Recombination also provided a rudimentary example of relational framing, describing how patterns of verbal behavior may combine to produce novel verbal utterances without direct reinforcement. The ideas at the roots of relational framing was mentioned by Skinner within Verbal Behavior (1957) under the auspice of association, and this model of responding to stimulus functions and not just objects was eventually matured upon by Relational Frame Theory with immense success (Hayes, Barnes-Holmes, & Roche, 2001) . We do not contend that the idea of "stimulus function" is not immensely important within our field; however, others were developing this idea in absence of the Interbehavioral version. Maybe it is because these other J o u r n a l P r e -p r o o f whether the model has been used to solve new challenges -does it survive the test of time, or does it die in favor of new models? Selectionism is a cruel arbitrator over disagreements, and in the case of Kantor's Interbehaviorism, time has not been favorable. However, more contemporary models that incorporate dynamical systems such as developed in the work of Thelen and Kelso, retain much of the essence of Kantor's ideas. These approaches make no mention of the contact medium, setting and setting events, and the interbehavioral history. They do mention bifurication, self-organization, and susceptibility to initial conditions. Discussion of Kantor's work may have participated in our synthesis of RFT and dynamical systems within RDT, perhaps implicitly. The specifics of Interbehavioral system appear to be increasingly selected out of the behavioral account, but the general message retains within more complex models that extend empirically from the work of the likes of Watson, Skinner, Pavlov, and others. Functional contextualist models provide an exemplar of this empirical progression (RFT as a basic model and ACT as an applied treatment model, see Belisle, 2020) as an opposite case example. Even though potentially faulty rules adopted by many behavior analysts led to initial hesitation or outright refusal to allow RFT and ACT models into the field, other fields working to improve the human condition (clinical psychology, social work, education) quickly adopted these models with great success. Empirical demonstrations of stimulus relations that supported the ideas of RFT grew at an exponential rate in the 2000s (O'Connor et al., 2017) eclipsing Skinner's verbal behavior approach, and ACT is being touted as a transdiagnostic solution to many psychological or behavioral disorders. And, despite initial hesitation, as noted by Belisle, Paliliunas, et al. (2020) , research utilizing RFT models with children has grown exponentially within major applied behavior analytic journals, including children with and without disabilities.

This work incorporates dynamics and successfully moves from theory to impact. Can better J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 21 outcomes be achieved using Interbehavioral models? Time will tell and selection will be the arbiter of the end of this story. We contend that limited empirical work completed so far appears to favor extinction as the eventual outcome. Just as the Shakers needed sex to reproduce, so too do theories need empirical data to ensure generational survival.

Successful working has not yet achieved within Interbehaviorism. The original article proposes, "scientific domains are completely free from absolutes, ultimates, and universals" (Hayes & Fryling, 2019) . Is this statement itself an absolute? Irony aside, science is first and foremost predicated on specific assumptions that are held absolutely. For example, in a physical or natural science, physicalism operates as a base assumption. That is, only physical events can cause other physical events. Pavlov, Watson, Skinner, Kantor, and contemporaries represent an extension of scientific physicalism within the context of developing a natural science of human behavior. At a basic level, theories or models that violate the assumptions of physicalism can and should be dismissed from the outset, which runs in contradiction to more dualistic or mentalistic accounts more pervasive in other branches of psychology. Pragmatism is another assumption that, though not held by all scientists, is an assumption held up by those self-proclaimed as functional contextualists or as pragmatists more generally. This assumption extends upon physicalism so that not all physically adherent theories are held equal simply because they exclude non-physical explanations. Being a "natural" science or philosophy of behavior is insufficient. The theory or model has to actually work to solve problems, and new theories must demonstrably work better than older more established theories to be adopted by the larger scientific and applied community. Kuhn, in Scientific Revolutions (1962) , extends this idea by proposing that the adoption of more successful theories or models is not an intentional choice of scientists, rather consumers of science will select technologies produced from the more J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 22 successful models. Over time, the successful scientist will gravitate towards approaches that allow him or her to work more successfully within their world and eventually to abandon old models in favor of newer models. If we accept pragmatism as our second absolute assumption, the non-adoption of Kantor's models might been seen as proof that these models are not superior to anything that existed prior or that has been since developed.

Kantor did not invent the "bias" of the observer in the observed. However Hayes and Fryling express this point as a core feature of the interbehavioral system that the rest of science fails to detect. "Moreover, while it is the case that Interbehaviorists would agree that no one escapes the effects of his or her personal and cultural histories, this does not mean that scientific knowledge is personal and ephemeral." Pp. xx, "The empirical methods employed by interbehaviorists are the same as those used in the sciences more generally Pp. xx," and "The practices of the investigative sub-domain of behavior science, known as the experimental analysis of behavior, are among the empirical methods employed by Interbehavioral Psychologists…". Pp XX. We assume the authors are referring here to the methods used by other radical behavior analysts, and if true, it appears that the interbehaviorist needs to shape shift into radical behaviorist in order to accomplish verification of phenomena. We believe statements such as these, and the conclusions that result indicate a concerning shortcoming of Interbehaviorism.

We applaud Hayes and Fryling (2019) for their recent well-written and convincing contribution to scientific discourse around the role of interbehaviorism for the behavior scientist.

But what is interbehaviorism? We believe that it is an unprovable and untestable, narrative about J o u r n a l P r e -p r o o f SELECTION BY CONSEQUENCES 23 how the psychological world might work. And when if someone tries to test interbehaviorism, they destroy its assumption of disrupting a field of interaction. We believe that you can't have it both ways -say you are a science, but not be placed under the truth criterion of science. And you can't say you are a philosophy of science but say nothing about philosophy. What is left is a body of work that has inspired some to place elements of the system into the account of human behavior (field theory), but the totality that Kantor hoped for has never been actualized, and we have concerns that it never will. A 100 plus years have passed since Kantor's original writings, and science has for better or worse, moved on. Like Brother Arnold and Sister June, the last two remaining Shakers (Blakemore, 2017) , eventually even the most inspirational movements that fail to gain adopters will come to an end. Only time will tell as to if Kantor's contributions will eventually follow the fate of the Shakers, but as we look at the general lack of data before us, the odds are pretty high that they will.

J o u r n a l P r e -p r o o f

2020: New decade, new disruption

Relational frame theory: A post-Skinnerian account of human language and cognition

There are only two Shakers left in the world

Self: An adaptive pressure arising from self-organization, chaotic dynamics, and neural Darwinism

Poincaré and the three-body problem

Sketch of JR Kantor's psychological interbehavioral field theory

An introduction to chaotic dynamical systems

John Dewey: The later works

The gravitational equations and the problem of motion

The grand design

Interbehaviorism and interbehavioral psychology