This paper analyzes the hardware and software features that would be desirable in a computer-based semantic network system for representing biology knowledge. It then describes in detail a prototype network of molecular biology knowledge that has been developed using Filevision software and a Macintosh computer. The prototype contains about 100 linked concept maps. Design principles are delineated. For example, the decision was made not to include direct links between members of a set such as "RNA" and "DNA." To move from one to the other in the network it is necessary to move up the hierarchy to "nucleic acids" and back down. This requirement is pedagogically attractive and reduces clutter. Many other such principles are required to assure consistency in map generation. Working with information in the computer-based semantic network provides an experience heretofore unavailable. Project members find it is fun, and somewhat challenging even for creators of the database, to find paths from point X to point Y in the database. Moving through the net requires an integration of concepts that seems qualitatively different from that encountered in most other interactions with academic subject matter. (Author)