Résumé
Metamodeling allows to capture domain knowledge through the definition of its structure (concepts and relations between them) and its constraints (logical expressions) often written in OCL. The OCL constraints added to a metamodel are of two types: 1) domainrelated constraints: they differ from one domain to another and are expressed based on the knowledge of experts; and 2) those that are added to the majority of metamodels to precise some inaccurate structures that may cause problems when instantiating models. We call these structures Metamodel Inaccurate Structures (MIS). In this paper, we performed an empirical study in order to point out the metamodel inaccurate structures. As a first step, a study is conducted on a set of OCL constraints taken from the UML metamodel to investigate the relation between structure and constraints. Then, to confirm our findings, we realised a quantitative analysis in order to count the occurrences of constraints that complete inaccurate structures. We believe that our results can help designers in the quest of creating or refactoring metamodels and specifying constraints that precisely capture domain knowledge to ensure consistency of the derived artifacts.