In order to replicate Hughes and Barnes-Holmes’s (2011) study, 12 undergraduate students were presented with two novel stimuli (the words “Cug” and “Vek”) in one of three attitude-induction conditions. In the first condition, a training IRAP provided feedback that shaped the participants’ brief-and-immediate relational responding over multiple discrete trials. In the second condition, a written narrative presented the two potential attitude objects (the words “Cug” and “Vek”) as equivalent to a series of either positive or negative trait adjectives. In the third condition, participants were given both the training IRAP and the written narrative. Having completed the training phase, the participants then took a testing IRAP in which the words “Cug” and “Vek” were presented as being either “Similar” or “Opposite” to a range of verbal stimuli with either positive or negative connotations. As the results of Experiment 1 indicate, all three of the experimental manipulations generated differential relational responding, yet there was no repeat of the weakening in the ‘IRAP effects’ that was observed by Hughes and Barnes-Holmes. To find out if the engineered response biases persisted over a longer timescale than the 20 min it took to run the first testing IRAP, a second experiment was conducted in which the remaining participants (n = 11) retook the testing IRAP five days later (Experiment 2a). The results of Experiment 2a indicate that the response biases that were engineered in the first session persisted into the second session; furthermore, there was a weakening of the ‘IRAP effects’ across the three pairs of test blocks in all of the attitude-induction conditions. To find out if the engineered response biases could be strengthened by further training, following the second testing IRAP (Experiment 2a), the participants were presented with the same attitude-induction procedure they had been given during Experiment 1; they were then asked to complete a third and final testing IRAP (Experiment 2b). The results of Experiment 2b indicate that response biases can be strengthened through further training. However, the ‘IRAP effects’ recorded in Experiment 2b were broadly weaker than the ‘IRAP effects’ recorded in the other two experiments.