I thought I had posted this idea for a protocol earlier, but a search on my username didn't turn it up.
This depends on the claim being that the spirit voices are in the recordings before anyone listens to them.
Come up with a numbered list of ten questions for the spirits that have known answers. Simple things. like "what is 2+2?" "What city are we in?" "What country does my aunt live in?" Stuff like that. It's okay if flaccon knows the answer, too.
Each question should be on a different subject, so that answers can be easily associated with questions. You wouldn't want two questions to which the answer was a city, for example.
Flaccon should ask the question, recording the answer to a numbered file, but not listen to the answer yet.
After asking all the questions, and recording all the answers, flaccon should leave the room and have someone else rename the files to A through J using dice. For the A file, roll the dice, and rename whatever number comes up to A. There should be a list made so that we know what number is in which letter-named file. Continue in this way for B, C, D, etc. If the number is 11 or 12, or the number that comes up has already been renamed, roll the dice again. Obviously, after renaming to I, there will only be one file that can be J.
Now the person doing the renaming should leave the room with the lists of questions and what the answers have been renamed to.
Flaccon can listen to file A, and write down the answers she hears in file A.
When she's satisfied that she's heard everything correctly, she can continue on to B, and so on.
If she wants to go over them any number of times to make sure she heard correctly, that's fine too.
After she's satisfied that she's heard all the answers that are there, and written it all down, the list of questions and the renaming list can be brought back in and the questions and answers compared.
Now, if the test were being done with a human that knew the answers, there would be a ten for ten correspondence of the questions to the correct answers.
I don't see any reason to think spirits should do worse.
But let's give them a couple of wrong associations just because. If 7 out of the 10 responses are correct, call it a win for the spirits.
If other people can see holes in this, please point it out. I don't see any easy way to calculate odds for this, though.
My expectation for a result would be that there would be some correct answers heard, but not in the file responding to the question.