With the increased demand for the utilization of computerized information systems by business users, the need for investigating the impact of various user interfaces has been well recognized. It is usually assumed that providing the user with assistance in the usage o-f a system would significantly increase the user's productivity. There is, however, a dearth of systematic inquiry into this commonly held notion to verify its validity in a scientific fashion. The purpose of this study is to investigate the impact of system-provided user assistance and complexity level of the problem on novice users' productivity in specifying database queries. The study is theoretical in the sense that it presents an approach adopted from research in deductive database systems to attack problems concerning user interface design. It is empirical in that it conducts an experiment in a controlled laboratory setting to collect primary data for the testing of a series of hypotheses. The two independent variables are system-provided user assistance and problem complexity, while the dependent variable is the user's query specification productivity. Three measures are used as separate indicators of query specification productivity: number of syntactic errors, number of semantic errors, and time required for completing a query task. Due to the lack of a well-defined metric for user assistance, the study first presents a generic classification scheme for relational query specification. Based on this classification scheme, two quantitative metrics for measuring the amount of user assistance in terms of prompts and defaults were developed. The user assistance is operationally defined with these two metrics. Four findings emerge as significant results of the study. First, user assistance has a significant main effect on all of the three dependent measures at the 1 percent significance level. Second, problem complexity also has a significant impact on the three productivity measures at the 1 percent significance level. Third, the interaction effect of user assistance and problem complexity on the number of semantic errors and the amount of time for completion is significant at the 1 percent level. Fourth, Although this interaction effect on the number of syntactic errors is not significant at the 5 percent level, it is at the 10 percent level. More research is needed to permit a thorough understanding of the issue of user interface design. A list of topics is suggested for future research to confirm or to modify the findings of this study.