A commonly suggested solution to reduce misinterpretations of verbal probability expressions in risk communications is to use a verbal-numerical (mixed format) approach, but it is not known whether this increases understanding over and above a purely numerical format. Using the ‘which outcome’ methodology (Teigen & Filkuková, 2013), we examined the effect of using verbal, numerical and mixed communication formats, as well as investigating whether marking outcomes as salient would alter the outcomes people perceived as ‘unlikely’ or having a 20% chance of occurring. We observed no effect of saliency, but replicated previous findings, with general preference for values at the high end of a distribution (including maximum/above maximum values) present in both verbal and mixed communication formats. This demonstrates the relevance of these findings for real-world consequential risk communication. Whilst the estimates differed between the mixed and numerical formats, we found that the mixed format yielded the more accurate estimates.