The Implicit Relational Assessment Procedure (IRAP) is increasingly used in applied and clinical settings, and yet many of the procedural variables of the measure have not been subjected to a systematic analysis. One such variable is the type of rules that are employed when instructing the IRAP and the effect this might have on participants’ performances. In the current article, three experiments were used to assess the impact of three different types of rules or instructions on IRAP performance. The instructions varied in the degree to which they specified parts of the relational network being assessed by the IRAP. The findings showed that the type of rule presented to participants during an IRAP can have a dramatic effect on the strength and direction of the trial-type effects that are produced by the measure. Furthermore, the type of instructions employed appear to interact with the order in which the IRAP blocks are presented (history-consistent versus history-inconsistent). The findings indicate that the behavioural dynamics that occur when participants complete an IRAP require extensive and systematic experimental and conceptual analyses, and this work will likely have an important bearing on research seeking to investigate the predictive validity of the IRAP in applied research settings.