Of course it isn't difficult to conduct. It is, however, apparently more difficult to do so having done the controls I outline above. I say that because, in all the DBTs I hear about, almost none show that they have proven the sensitivity of the test to be adequate.
Any conclusions drawn from a test which hasn't been shown to be sensitive enough to measure what you are looking for, are essentially meaningless. Yet they still get trotted out. So much for intellectual rigour.
Not necessarily, m'lud.
The point is to see if listeners can reliably (in a statistical sense) 'distinguish' *in the circumstances of the test*. Thus the aim would be to have a test setup of the kind where some listeners have said thay believe they *can* 'distinguish' (or feel they may be able to) when, say the test is run *without* being blind. Ideally, one they also would regard as a 'good' system.
The aim here is to determine if 'independent' awareness of what you're listening to makes a difference to the results. i.e. to check for biases due to information *other than the actual sounds*.
So *if* someone believes that have/can hear a 'difference' in such a situation when the comparison is *not* blind, then it may be an appropriate setup for a blind test.
Yes, it can be useful to run the test both ways. But given a system that most people would think a 'good' one, they may reasonably conclude that any difference which vanishes in a 'blind' test may be too small to fuss much over.
Similarly, yes, it's reasonable for other to *repeat* this test using different systems to see if they can find evidence that a change of system alters this result and shows that in such a case people *can* hear a 'difference' even when run blind.
Science uses replication, etc, to check for any problems with results from any individual experiment. The point of blind versus non-blind is to determine if having info *other* than the sound alone affects the outcome. A change of system is to probe any other factors which might have been systematic to one example.
Alternatively you run the entire argument the other way if no-one has any idea if they might be able to tell. i.e. you can run the test 'blind' and then, if you wish, run it 'non blind' again. But again if the two results differ the implication is that this is down to not having anything other than the 'sound' to go on in the blind test.