We study the effectiveness of n-detection test sets based on sition faults in detecting defects that affect the timing behavior of a circuit. We use path delay faults as surrogates for unmodeled defects, and show that the path delay fault coverage achieved by an n-detection transition fault test set increases significantly as n is increased. We also introduce a method to reduce the number of tests included in an n-detection test set by using different values n for different faults based on their potential effect on the defect coverage. The resulting test sets are referred to as variable n-detection test sets.