Journal of Contextual Behavioral Science (JCBS)
Volume 40, April 2026
Authors
Matthias Raemaekers, Martin Finn, Jan De Houwer
Key Findings
- Computational modelling approaches can facilitate research on complex forms of AARR.
- RFT can guide researchers developing a computational model of AARR.
- We describe a partially hardcoded computational model of AARR inspired by RFT.
- We simulate behavioral data from a conceptual replication of Steele and Hayes (1991).
- Implications of our approach and future research directions are discussed.
Abstract
Relational Frame Theory (RFT) has inspired a considerable body of research demonstrating fundamental aspects of arbitrarily applicable relational responding (AARR) purported to be the building blocks of human language and cognition. Current empirical research has certain limitations, however. Computational modelling allows researchers to circumvent certain practical limitations to studying (the development of) complex forms of AARR and force them to scrutinize the theory, yet limited research has sought to develop computational models of AARR. RFT can guide researchers developing computational models of AARR by emphasizing its operant and context-sensitive nature and by clarifying the nature of the learning history that is required for its development. We describe a partially hardcoded computational model of AARR inspired by RFT. By combining computational reinforcement learning with hardcoded relational knowledge, we simulate AARR as observed in a conceptual replication of the seminal Steele and Hayes (1991) study. Our results add to a growing literature modeling relational behavior and illustrate that RFT can be useful for researchers interested in computational modelling. Implications and future research directions are discussed.