BEHIND
Vocalists at proper scale (ie the visual auditory image is human scale or slightly larger) should be behind the plane. Exaggerated scale or vocals that are too forward indicate peaks in high-frequencies and can cause destruction of a 3d stereo image.
Agree with many here who say the recording, etc. are influential. Also, agree that test tracks are invaluable for calibrating a system.
(not to get involved with the toe-in discussion, but the "mid channel" of the stereo signal (where many vocals live) often emanates from the side of drivers and LESS toe-in can increase the mid channel content, like vocals. So toe-in and speaker position definitely are very important to vocal depth.)