A translating-computer-interface approach to providing a common, or virtual-system, mode of access to a network of heterogeneous online bibliographic retrieval systems has been investigated. Enhanced access to such systems by end users has been demonstrated through test usage of an experimental interface. A table-driven, rule-based message interpreter that provides a dynamic and flexible means for handling the automation of the interconnection of computer and human retrieval protocols has been found useful in implementing the interface. Experimental analysis has demonstrated the success of various computer-assisted instructional techniques as well as indicated basic tradeoffs and limitations in their implementation in working systems. Further development of interface systems appears promising and possibilities for further research have been outlined in a number of areas. (Author)