Given a behavior of interest in the program, statically determining the corresponding responsible entity is a task of critical importance, especially in program security. Classical static analysis techniques (e.g. dependency analysis, taint analysis, slicing, etc.) assist programmers in narrowing down the scope of responsibility, but none of them can explicitly identify the responsible entity. Meanwhile, the causality analysis is generally not pertinent for analyzing programs, and the structural equations model (SEM) of actual causality misses some information inherent in programs, making its analysis on programs imprecise. In this paper, a novel definition of responsibility based on the abstraction of event trace semantics is proposed, which can be applied in program security and other scientific fields. Briefly speaking, an entity is responsible for behavior B, if and only if R is free to choose its input value, and such a choice is the first one that ensures the occurrence of B in the forthcoming execution. Compared to current analysis methods, the responsibility analysis is more precise. In addition, our definition of responsibility takes into account the cognizance of the observer, which, to the best of our knowledge, is a new innovative idea in program analysis.