Abstract. In the early stages of model driven development, models are frequently incomplete and partial. Partial models represent multiple possible concrete models, and thus, they are able to capture uncertainty and possible design decisions. When using models of a complex modeling language, several well-formedness constraints need to be continuously checked to highlight conceptual design flaws for the engineers in an early phase. While well-formedness constraints can be efficiently checked for (fully specified) concrete models, checking the same constraints over partial models is more challenging since, for instance, a currently valid constraint may be violated (or an invalid constraint may be respected) when refining a partial model into a concrete model. In this paper we propose a novel technique to evaluate well-formedness constraints on partial models in order to detect if (i) a concretization may potentially violate or (ii) any concretization will surely violate a well-formedness constraint to help engineers gradually to resolve uncertainty without violating well-formedness. For that purpose, we map the problem of constraint evaluation over partial models into a regular graph pattern matching problem over complete models by semantically equivalent rewrites of graph queries.