With respect to the effects of types of knowledge on human diagnostic performance, the results of several experiments claimed that training with procedural knowledge is more effective than training with principle knowledge. However, more useful results would be attained by investigating when and how the principles of system dynamics is valuable for diagnosis. Accordingly, we conducted an experiment to reevaluate the value of principle knowledge in two problem situations. A simulator system, named DLD, to diagnose an electronic device was created. It is a context-free digital logic circuit which includes forty-one gates of three basic types. The experiment investigated the effects of principle knowledge over common procedural knowledge. The experimental results showed that the effects of principle knowledge is dependent on the complexity of diagnostic situations. This adds up on experimental evidence against the presumed ineffectiveness of principle knowledge and forward reasoning in fault diagnosis. The results also suggest the source of the usefulness of principle knowledge.