Individuals with contamination concerns show aberrant interpretational and attentional processes. Yet, it is unclear whether threat-related associations play a causal role in anxiety symptoms and attentional bias. The objective of our study was to investigate if training implicit associations affects stress reactivity and attention in the context of contamination concerns. In a double-blind randomized design, we used a modified Implicit Associations Task (IAT) to train associations between contamination and danger in a non-clinical sample (N = 121). Dependent measures were a brief-IAT to assess changes in associations, contamination-related behavior approach tasks, and a spatial cueing task to measure attentional bias. Results show that training successfully modified implicit associations. However, there were no transfer effects on approach behavior or attention. Findings suggest that the modified IAT is a useful task for training implicit associations, but that transfer to other domains (attention and behavior) is limited. Limitations and future implications are discussed.