while i am very skeptical on any such “astral projection” concepts, there are some ways you can go about experimenting if you feel so inclined.
inasmuch as communication is possible, we have to have what would be called a communication channel. there is some theoretical basis to decide wether two emitters/receptors are really communicating or not. consider this situation :
A acts as emmiter and it sends the signal S. you verify if B acting as receptor in fact receives signal S. if it does not, they are obviously not communicating. however, even if it does, it could be they are still not communicating, inasmuch as S might be an artefact of the medium (parasites for instance in the case of an antenna) or of B himself (most likely cause for a human mind).
If you go about it making A send signal S1 and B signal S2, you have not improved much, but just applied the situation above twice. it is in fact rather equivalent from a “does there a channel exist perspective” with sending S twice in a row.
It is obvious that making S2 a function of S1 or vice versa wouldnt make much difference.
In the particular case of human minds, picking a particularly bad chosen aleator signal (such as asking an english speaker to communicate a japanese text) could be a worng approach, because it might inhibit communication even more than it inhibits cheating.
the easiest way to check, i suppose, woudl be to pick obvious but not very verbalised concepts, such as some colors, some universals (truth beauty etc) and some states (hunger, sleepiness, fear) for your signals. pick say four well distinct ones in each cathegory. write some simple rules (ie if you feel red, you answer beauty) that are different for the two subjects.
then give one of them a starting point, and note what they say they receive. do say 4 such sessions over 4 distinct days, change the rules each day. ask them to write down if they make mistakes (ideally they dont). give them simple procedures (ie in the interval 15:00 15:05 you listen. in the interval 15:05 15:10 you answer etc etc) let them do about 16 or 24 such back and forth. pick rules and start point so you will have long periods (see below)
then pick up your data and analize it. first off see whats called penetration. see how many exchanges before they lose it. do averages, count averages.
then see “pick up” factor, this is the most important bit really. say we have this, row one correct row 2 actual notes, letters for cathegories, numerals for respective signal :
A1 B3 B2 C1 A2 C2 C3 A1 B3 … (as you see we have reached period, from now on it will just repeat same sequence. ideally use a longer periood, 8ish to 12ish)
A1 B3 B2 C2 C3 A1 B3 B2 C2
see the way they always mistake the B2 C2 instead of the correct B2 C1 ? but after that they pick up and go about it right, untill the next B2. this proves they are communicating, but you have a problem with defining the rules. if they went like
A1 B1 C3 A3 B2 B1 C2
situation would be very different. see for instance B1 went to C3 and later on to C2. tyhis sort of self contradiction makes us belive B1 is never heard. Also see how they never actually follow our rules ?
do some processing on your data, to notice self contradictions, pick ups (like in 1st example) and do statistical sampling. ie see for each term how likely is for another term to follow. if you get aproximately the same for all terms, in our example 8-8.5% for any of them, its a random distribution. if they seem to clog around a speciffic answer (positive correlation, like for instance correct answer is B2 and you get 50% B, 15% A and 35% C and furthermore, of that 50% B’s another 60% are 2’s) and that answer is the right answer, they are likely to be in fact communicating. how likely ? well depends how well the data corellates.
hope this helps.