RNA interference (RNAi) provides a powerful technique for the derivation and analysis of loss-of-function phenotypes in vertebrate cells, where alternative approaches are either arduous or frequently ineffective. RNAi, however, is not always totally specific, and confirming the specificity, and hence the validity, of data obtained using RNAi therefore forms an essential component of experiments that rely on this technique. Here I briefly review the potential pitfalls associated with RNAi, and then suggest experimental approaches that can be used to maximize the specificity of RNAi or to confirm that the data obtained are indeed valid.