We compare the performance of naive data modelers in modeling association, generalization, and aggregation relationships with the relational and object-oriented data models. We first develop research hypotheses based on the properties of expressiveness, minimality, and unique semantic interpretation to analyze the effectiveness of the two models. We then test our hypotheses in anexperiment with 22 naive modelers. The findings of our study support the notion that, to be effective, a data model should satisfy these three properties