The current study examined the malleability of implicit attitudes using the Implicit Relational Assessment Procedure (IRAP). In Experiment 1, "similar" and "opposite" were presented as response options with the sample terms "old people" and "young people" and various positive and negative target stimuli. Results showed significantly faster response latencies for consistent (e.g., Similar Positive-Young People) compared to inconsistent tasks (e.g., Similar-Positive-Old People). Explicit measures did not correlate with this IRAP effect. Experiment 2 determined whether prior exposure to pictures of admired and disliked old and young individuals had an impact on IRAP performance. Results revealed that pro-old exemplars reduced the pro-young IRAP effect, but reversed the ann-old effect, and this held for 24 h; explicit measures were largely unaffected. The findings suggest that the IRAP provides an informative measure of attitude change following pro- versus anti-exemplar training.