Worry is a verbal and abstract thought activity with only little mental imagery involved. It has been shown that this processing bias leads to hampered emotional processing of worry topics so that worry is maintained in the long run. However, there is some evidence that mental imagery during worrying has the opposite effect: It leads to stronger emotional reactions than verbal thoughts and thereby fosters emotional processing of worry topics. In the present study, we examined whether training mental imagery reduces pathological worry. We compared the effect of a novel training in mental imagery (TMI; n = 37) on various worry-related outcomes to a control training in verbal thinking (TVT; n = 38) and a waiting-list control group (n = 36) in a sample of pathologically high worriers. Both trainings showed significant within-group decline regarding e.g. pathological worry activity, worry-related impairment and anxiety assessed both one and five weeks after the training. Cohen's d on different outcome measures ranged from .39 to 1.17. However, unexpectedly, there were no significant differences between groups on any worry- or affect-related outcome measures. In, sum our training in general mental imagery skills turned out to be a promising intervention for the reduction of pathological worry. However, we found a training in verbal thinking to also have unexpected beneficial effects. The theoretical and clinical implications of these findings and methodological limitations of our study are discussed. •The efficacy of a training in mental imagery (TMI) to reduce worry was evaluated.•TMI was compared to a ‘training in verbal thinking’ and a wait list control group.•TMI led to significantly reduced pathological worry and associated outcomes.•Also verbal thinking during worrying showed unexpected positive effects.•The effects were most pronounced in highly anxious and worry-prone subjects.