Someone said they use extended listening in the dark to avoid light
stimulation and do hear differences. The key is not the level of
illumination but the knowledge of which is in the system. It is suggested
that visual knowledge of which gear is active provides the overlap in the
hearing and seeing parts of the brain, in addition to the anticipation
bias of which is thought to provide some difference beforehand. Dark
listening only eliminates one of the perception altering sources. The
thing to do would be dark listening without knowing which bit of gear is
active, which would remove the anticipation bias also. One way to remove
the visual input would be to have both bits of gear contained in boxes of
identical appearence, or just put a cloth over the connections obscuring
which is active; as has been suggested. If removing this knowledge
produces results tending to random, as it apparently does, then we know
knowledge of what is active is a perception distorting input.

If someone wants to test level of illumination as a varible, it can be
tested, but I fear it a waste of time. This has been done already in
another form. It has been shown that blind folk do no better in hearing
differences then do sighted people when knowledge of what gear is active
is absent.