This paper proposes a probabilistic approach to diagnose intermittent faults in a wireless sensor network (WSN). Though system level diagnosis has become an important technique to diagnose faults in wired networks, its applicability to wireless network has not been well studied. Though, intermittent faults are most probable in WSN, diagnosis of such faults has not been well explored. The algorithm proposed in this paper has been evaluated through analytical means and is simulated assuming a 512 sensor nodes interconnected to form an arbitrary network topology. The results show that accurate diagnosis of intermittent faults in WSN is feasible. Keywords: distributed diagnosis, wireless sensor networks, intermittent faults, synchronous systems, Identifiability.