[Context and motivation] Dynamically adaptive systems are increasingly conceived as a means to allow operation in changeable or poorly understood environments. [Question/problem] This can result in the selection of solution strategies based on assumptions that may not be well founded. [Principle ideas/results] This paper proposes the use of claims in goal models as a means to reason about likely sources of uncertainty in dynamically adaptive systems. Accepting that such claims can’t be easily validated at design-time, we should instead evaluate how the system will behave if a claim is proven false by developing a validation scenario. [Contribution] Validation scenarios may be costly to evaluate so the approach we advocate is designed to carefully select only those claims that are less certain, or whose falsification would have serious consequences.