Introduction: Relational Frame Theory (RFT) posits derived relational responding (AARR) as the foundation of language and cognition. The Hyperdimensional, Multilevel (HDML) framework operationalizes complexity as a key dimension influencing AARR. This study examined the interaction between task complexity and repeated measurement on AARR performance within a dynamic paradigm.

Methods: Sixty participants were tested using the clinical simulation paradigm, including symbolic stimuli (e.g., ")))" and "!!!") and contextual cues (OST, UNR, ALB), in a between-subjects design. Task complexity (Low vs. High) was manipulated, and performance was assessed across two consecutive measurement points (Set1, Set2). Data from 5,077 trials were analyzed using Generalized Linear Mixed Models (GLMM) with Gamma and Binomial distributions in SPSS.

Results: For response time (RT), a significant Measurement Point × Complexity interaction was observed (β = 0.260, p < .001). Simple slopes analysis revealed a crossover pattern: RT increased from Set1 to Set2 in low-complexity trials (3.89 ms to 4.28 ms) and decreased in high-complexity trials (4.12 ms to 3.49 ms). For accuracy, only the main effect of Measurement Point was significant (β = -1.243, OR = 0.288, p < .001), indicating a general decline at the second measurement. The main effect of Complexity (p = .837) and the interaction (p = .291) were not significant for accuracy.