Yep. This is what I mean by the second possibility - there should be a low enough noise floor (and I don't just mean noise from the systems, but across the entire experiment - environmental noise, participant noise, etc.) that a difference should be detectable if present. One can use a staircase...