For the cylinder-selective monitoring of combustion cycles in spark-ignition engines, the dynamic exhaust-gas pressure is analyzed. A time domain based diagnostic system for misfire detection has been developed and tested on data measured in a BMW 750i, V-12 engine. It uses features of the suitable low-pass-filtered exhaust-gas pressure signal by calculating differences of the locally determined extrema. For the detection and localization of all misfire combinations a simple inference system in the form of linguistic rules is used. It is shown that even within the operating areas of high engine speeds and low loads on engines with a high number of cylinders good classification rates can be obtained. In addition to the development of the method, experiments were carried out to examine the robustness and restrictions of the approach under special operation conditions and at the limits of the engine as well as phenomena occurring in coupled exhaust-pipes with mufflers and combustion-related influences.