Change detection in the flicker paradigm: Do faces have an advantage?