The motivation for this study was that database query facilities are not effectively meeting casual users' needs. A solution to this problem is especially important due to the increasing number of potential users falling into the classification of "casual user". There is considerable controversy revolving around the question of which elements and/or which combination of elements within the casual users' environment are necessary to provide an effective man-machine interface. This study is intended to extend the basic knowledge relating to the effect of different data models on casual users' performance and confidence in writing database queries. The data models used to present the external view of the database in this study are the relational, hierarchical and network models. The experiment involves a written test designed to permit the evaluation of the subject's ability to comprehend and retrieve information from a database by writing English-like queries. The subject's performance in writing the queries is based on errors made in (1) writing the specification portion, (2) writing the condition portion, (3) writing the navigation portion, and (4) the use of the language. Overall performance is represented by the sum of the four component scores. The subject's confidence is a self-reported value on a scale of one to five. The group using the relational model performed significantly better than the group using the hierarchical model when writing the specification and navigation portions of the queries. The absences of a significant difference in overall performance among the data model groups supports the technique used to evaluate specific aspects of the queries independently. No statistically significant differences in confidence level were detected among the data model groups. This is an indication that the group using the hierarchical model may have been overly confident, in view of their poorer performance.