Here's jsfisher's protocol with the suggested improvements incorporated.
1. Alderbank selects a card from his pack at random and allows flaccon to view it, without looking at it himself
2. flaccon asks the spirits to identify it, then she listens to the recording playback to hear the card correctly identified. She repeats this up to four times, until she is satisfied that she has a usable recording. If she has still not succeeded then that card and recording are discarded
3. If the recording is deemed acceptable, Alderbank will seal the card in a numbered envelope (still without looking at it) and set it to one side. The recording file should be named CARD1.WAV and placed in a directory with no other files on flaccon's computer. flaccon should make her own private note of which card was identified in each file.
4. Repeat steps 1-3 until at least 3 usable recordings - CARD1.WAV, CARD2.WAV and CARD3.WAV - are available. If after 90 minutes this has not proved possible then the test protocol is deemed unworkable and the test abandoned
5. The computer is left set up so the individual recordings can be easily played for flaccon's friend. Alderbank looks over the computer screen and the general area around it for any (presumably unintended) clues to the cards' identities.
6. flaccon leaves and the friend is brought in.
7. Alderbank will open the first envelope and show the first card to the friend. He will then replay the first recording, and the friend will verify that he/she can hear it being identified in the recording. If he/she cannot verify this then the test protocol is deemed unworkable and the test abandoned.
8. Alderbank will then replay each of the other two recordings for the friend, one at a time, and allow the friend to name the corresponding card from the recording. If this takes longer than 20 minutes that card is abandoned. If the friend thinks they have heard the card's identification successfully Alderbank will write the card name provided by the friend on the corresponding envelope. Both suit and number of the card must be identified. ETA: If the friend cannot hear the identity of at least one of the two cards in its corresponding recording then the test protocol is deemed unworkable and the test abandoned
9. Once all three cards have been handled, flaccon will be brought back in and Alderbank will open the envelopes and permit both flaccon and the friend to compare the cards to what the friend heard on the recordings.
Success criteria: at least one of the two cards should be successfully identified (both suit and number) for further investigation to be deemed warranted