Abstract: Let X_1,X_2,\cdots be a sequence of i.i.d. random variables. Shao (1997) established a self-normalized large deviation without any moment assumption. However, the proof of the upper bound was quite complicated. In this note we give a much simpler proof.