Model Dependent Realism and the Rule-Governed Behavior of Behavior Analysts: Applications to Derived Relational Responding
Pick the relational-frame model that predicts best in your clinic, not the one claiming to be true.
01Research in Context
What this study did
Belisle (2020) wrote a theory paper. He asked how behavior analysts should pick a model of derived relational responding. He borrowed model-dependent realism from physics. The idea is simple: pick the model that works best in your setting. Do not worry if it mirrors reality.
The paper walks through how we choose among RFT, stimulus equivalence, and newer hybrid models. It says we should test each model like an engineer tests blueprints. The one that predicts client progress wins.
What they found
There is no single true model of derived relational responding. Different models win in different labs and clinics. Model-dependent realism tells you to stop hunting for the real one. Just use the model that gives you clean data and better client outcomes.
The paper shows how to weigh three things: how well the model predicts, how simple it is, and how easy it is to teach to staff.
How this fits with other research
Alonso‐Álvarez et al. (2018) seem to disagree. Their data say equivalence plus exclusion can explain same and opposite performances without new relational frames. This looks like a clash, but it is not. Belisle (2020) would simply treat their account as another model. If it predicts better in your clinic, use it.
Marin et al. (2024) extend the same theme. They warn that equivalence shown in quiet labs may vanish in noisy homes. Belisle (2020) agrees: pick the model tested in the setting you actually work in.
Schoneberger (2016) tried to fix the realism problem with Rortian pragmatism. Belisle (2020) sidesteps the debate entirely. He says forget about truth; focus on what works.
Why it matters
You run an ABA clinic. One staff member swears by RFT, another by stimulus equivalence. Stop arguing. Run a quick pilot with each model. Track which one produces faster emergent relations for your clients. Adopt the winner and move on. Model-dependent realism saves time and ends turf wars.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Run a five-trial probe with two models; keep the one that yields more correct untaught responses.
02At a glance
03Original abstract
A fundamental assumption within radical behaviorism is that all human behavior, including the rule-governed behavior of scientists, can be understood within a functional account. I propose that models of human behavior can be best described as a set of rules that are selected by behavior analysts to solve applied challenges, rather than descriptions of nature as it “truly exists.” Model dependent realism (MDR) developed within the field of physics may provide useful criteria that could allow behavior analysts to more accurately track the relative probability of success of a given model within applied contexts. As a case example, I examine dispersive models of derived relational responding in terms of the criteria outlined within MDR, and I describe a preliminary level-scaling account of derived relational responding that encompasses several models in pursuit of a unified account. The account is context dependent and adopts a pragmatic truth criterion, consistent with assumptions within functional contextualism and radical behaviorism as an overarching rule governing the behavior of our applied subfield.
Perspectives on Behavior Science, 2020 · doi:10.1007/s40614-020-00247-x