A comparison of data models and data retrieval methodologies

Previous studies on database data models have produced conflicting results. It was hypothesized that reported results differed due to different methodologies used to study the effect of data model on information retrieval. To test this hypothesis three methods of retrieval were utilized. The first was information search through instance diagrams (no language condition) which used the same method as Brosey and Shneiderman (1978). In the other two conditions, a query language (SEQUEL) was used and subjects either wrote queries (Offline) or typed queries into a computer (Online). Relational and hierarchical data models were compared. Sex was used as a blocking variable and all subjects answered both easy and hard questions (Query Difficulty). Correctness of retrieval results showed a main effect for method of retrieval as hypothesized. The No Language group scored higher than the Offline group who in turn scored higher than the Online group. It was hypothesized that the relational data model would result in better performance. Unexpectedly, there was a significant main effect for data model with the hierarchical model resulting in higher correctness of retrieval scores than the relational data model. (See more in text.)