Information Systems Research, 1996, Volume 7,
Issue 2, Page 137-162.
Alongstanding debate in the data modeling literature pertains to whether the grammars used to generate conceptual schemas should sustain a distinction between entities and attributes. The grammars used to generate entity-relationship diagrams and object-oriented conceptual models, for example, provide separate constructs for representing entities and attributes. The grammars used to generate binary data models, however, provide only a single construct for representing both entities and attributes. To sharpen the focus of the debate, a multi-trial free-recall experiment was conducted with database designers who had been trained primarily in a binary conceptual schema design methodology. In the experiment, the designers were first shown conceptual schema diagrams based on a binary model. The designers were then asked to recall the diagrams. Throughout their training as designers, they had been admonished to eschew any distinction between entities and attributes. Moreover, the diagrams they were shown in the experiment did not make a distinction between entities and attributes. Their recall protocols seemed to show, however, that they were considering some elements of the diagrams to be entities and others to be attributes Their memory structures appear to reflect, therefore, that they perceive entities and attributes to be two distinct constructs.