Error Propagation Analysis (EPA) is a technique for understanding how errors affect a program's execution and result in program failures. For this purpose, EPA usually compares the traces of a fault-free (golden) run with those from a faulty run of the program. This makes existing EPA approaches brittle for multithreaded programs, which do not typically have a deterministic golden run. In this paper, we study the use of likely invariants generated by automated approaches as alternatives for golden run based EPA in multithreaded programs. We present Invariant Propagation Analysis (IPA), an approach and a framework for automatically deriving invariants for multithreaded programs, and using the invariants for EPA. We evaluate the invariants derived by IPA in terms of their coverage for different fault types across six representative programs through fault injectionexperiments. We find that stable invariants can be inferred in allsix programs, although their coverage of faults depends on theapplication and the fault type. © 2017 IEEE.