We present a study of cursors for selecting 2D-projected 3D targets. We compared a stereo- and mono-rendered (one-eyed) cursor using two mouse-based and two remote pointing techniques in a 3D Fitts' law pointing experiment. The first experiment used targets at fixed depths. Results indicate that one-eyed cursors only improve screen-plane pointing techniques, and that constant target depth does not influence pointing throughput. A second experiment included pointing between targets at varying depths and used only "screen-plane" pointing techniques. Our results suggest that in the absence of stereo cue conflicts, screen-space projections of Fitts' law parameters (target size and distance) yield constant throughput despite target depth differences and produce better models of performance.