We present in this article a method of validation for "serial transaction". The serial transaction model has been proposed in order to validate a concrete real-time application. This model is typically a task reading serial information (RS232, CAN,): several instances are identical and read an unitary part of a serial packet, these tasks have the same WCET, offset shifting, priority and relative deadline. In addition, the last task of a transaction has to deal with the whole packet, and is typically longer, but has a longer relative deadline, and a lower priority. The method of validation that we present here uses the concept of reverse transaction deduced from the serial transaction to analyse.