This paper reviews recent results obtained in the medical diagnosis field by adding to a coherent inference process qualitative constraints. Such further considerations turn out to be significant whenever a basic lower-upper conditional probability assessment induces extension bounds too vague to take any decision. Three general types of qualitative judgements are proposed and fully described. They do not constitute a “panacea” to solve any problematic situation, but their application can considerably improve inferences results in specific fields, as two practical applications show. Keywords coherent inference, conditional exchangeability, qualitative constraints, diagnosis procedures