As more and more data are expressed and stored in XML format, the study on XML keyword retrieval becomes the focus of IR (information retrieval) and Database. This paper gives and proves some properties of SLCA (smallest lowest common ancestor), which is the key concept of XML keyword retrieval. It also introduces anew XML keyword retrieval algorithm, Nearest Pair, on the basis of the properties above. This algorithm uses the iterative bi-search technology to look for nearest pairs, which can decrease the assistant computation by one order of magnitude. The experimental results show that Nearest Pair outperforms the existing mainstream algorithms in most cases.