When attempting to record electronic voice phenomena, we try something a little different.
Many investigators prefer to conduct "EVP sessions" to record information only during a question and answer period. This can be useful, especially if many people are present and it's necessary to make sure they all remain quiet. However, I believe that spirits, just as the living, may have different personalities. Personally, I'd be intimidated by being asked questions and have everyone waiting for an answer -- especially if they scream and make a big fuss if I were to say something.
So I'm not a fan of "EVP sessions." We leave at least one static recorder or a video cam with sound somewhere. Then each of us keeps an audio recorder with us at all times. And we leave them on the whole time, not just during a question/answer period.
I talk and ask questions, but also make general comments about various things, or sort of narrate what I'm doing: "Now I'm walking down the hall....I wonder what's behind this door....It feels cold in here...." etc. During this narration is when I've recorded some of the most interesting sounds. We try to ask questions that would need more than a simple "yes" or "no" answer, as many sounds can be mistaken for a one-syllable word. An audible breath can sound like "yes" if you're listening for such an answer. Look for longer words or phrases when possible.
We don't taunt, judge or attempt to shame the dead. I prefer a non-confrontational approach. Not "Shame on you for doing whatever" but rather "Why did you feel a need to do whatever?"
I think it's always a polite thing to thank them for speaking with you afterwards, whether or not you think you caught anything.
Leaving multiple recorders on the whole time makes a whole lot to review later. 4 recorders going during a 2-hour exploration gives me 8 hours of audio to review. Nobody said this would be easy. We both listen separately, and sometimes hear completely different things. Also, get second and third opinions on what is heard, without first revealing what you think you heard. In our past videos I presented text of what I was hearing, and this is not a good idea. It sets up a confirmation bias and the brain will automatically hear what is suggested. In the future I will not include the text until afterwards.
Anytime investigating outdoors it's important to use a windscreen over the microphone. It's frustrating to have to throw out the audio because of wind noise that can sound like whispering.
Presently, it is impossible to certify without a doubt that a voice or whisper is of paranormal origin. Nobody can know absolutely that it's not something else like a stray radio or cell phone signal, or who-knows-what. At least not at this time with the technology available. But sometimes it sounds like an intelligent and very relevant response or comment. Pareidolia, perhaps? Look it up.
Note: I see so many teams presenting an EVP as "evidence" when they are in a room full of people talking, laughing, moving around and breathing. I think those should be discarded when you can't account for each person present at every moment and what sounds they might be making. If they're just having fun and blabbing, they aren't that serious about this. For this reason I will not consider any audio "evidence" if more than about 4 people are present.