The development of test cases is an important issue for testing software, communication protocols and other reactive systems. A number of methods are known for the development of a test suite based on a formal specification given in the form of a finite state machine. In this paper, we overview and experiment with these methods to assess their complexity, applicability, completeness, fault detection capability, length and derivation time of their test suites. The experiments are conducted on randomly generated specifications and on two realistic protocols called the Simple Connection Protocol and the ITU-T V.76 Recommendation.