Fortunately, the entire structure of relational database management is built on a solid theoretical underpinning that makes the operations used to associate the proper attributes with the proper entities predictable, repeatable, and formally correct.
This process is referred to as normalization. Normalization is not an art form; it is a science based on provable formal concepts such as dependency theory. The normal forms and their derivations operate strictly within a series of inference rules and formal logical operations drawn from formal logic and set theory. This should not be taken to mean that only one correct logical database design can be developed for a set of relations. In fact, it is fairly easy to create numerous formally correct logical designs for the same set of relations. The point is that it is possible to develop both correct and incorrect logical designs, and by following the rules of normalization and dependency theory, you can always develop a correct logical design.
Unfortunately, the determination of entities and their relationships is subjective, unpredictable, unrepeatable, and more of an art form than a science. In spite of the subjectivity of entity-relationship-attribute analysis, the normalization process almost always rescues any blunders made in the determination of entities and relationships, so it all works out in the end.
See Chapter 5: “The Normalization Process” and the references cited there for more information about normalization and dependency theory.