Some first thoughts are:
Create a pool of targets from which to be selected.
This pool should be as homogenous as possible. i.e. make sure they are all the same sex and age etc. (at least make them similar in the info the reader will get).
Choose 6 subjects from a larger pool completely at random.
Number them randomly from 1 to 8 (if there are two dummies included).
Give the details to the reader who will prepare the 8 readings.
The intended targets of the readings will be noted by the person running the test.
Give all eight readings to each of the six participants.
They have to choose the reading which they think is the best match to themselves.
Once the results are in a score is calculated on how many people matched up the correct reading.
Odds of success can be set arbitrarily. In most preliminary trials of this type odds of 1000:1 against are usually set. This is high enough to make it unlikely for success to be a chance occurrence and the odds also allow for some slackness in the testing procedure.
In this case scoring 4 or more has odds of 1:337
Scoring 5 or more has odds of 1:6,096
I would say that a score of 5 or 6 would be needed to be considered as meaningful (worthy of further, more rigorous consideration).
Is that the sort of set up that describes this test?
If so, I'm sure that we can refine the details as required.