We investigate in this paper the anchor selection problem with consideration of anchor location uncertainty, which arises naturally from error propagation problem in successive or distributed localization processes. We establish an applicable criterion for selecting anchors from sensors under any prescribed amount of anchors, by minimization the lower bound of transformation error term in Ash-Moses’s error decomposition. This criterion quantitatively explain the strategy of deploying the anchors around the perimeter of the network, in order to minimize the propagation of anchor location error. We also discuss propositions such as asymptotical optimality, under several situations. Simulations illustrate the performance of the perimeter deployment strategy with varying anchor location uncertainty. This work can be regarded as complement of existing works under the assumption of accurate anchor location.