This is the story of how I set out to catch an EVP, caught one, and still don't believe in them.

EVP is short for "Electronic Voice Phenomena". EVPs are human-sounding voices of unexplained origin which appear on a recording device. TAPS and other paranormal investigators offer EVPs as evidence of paranormal phenomena. Normally, an investigator gathers EVPs by turning on a voice recorder and a source of white noise, such as a fan or running water. (Some people believe that ghosts create EVPs by manipulating the white noise.) TAPS, like many EVP investigators, uses inexpensive recorders designed for taking dictation. These are known as "IC recorders". (More on those later.)

I thought it would be unfair to criticize EVPs without trying to get some personal experience, so I ran a small experiment. I put a digital recorder in the laundry room, turned on the ceiling fan, shut the door, and let it run for an hour. The results were kind of dull; an hour's worth of pure fan noise, without any disembodied voices to speak of. You see, I'd used an Edirol R-09, which is designed for doing live recordings of music.  To get the experience of listening to an IC recorder, I needed to do some more work.

First, I needed to have a generally crappier recording. Most IC recorders record at about the same sound quality as a walkie-talkie, so I started by down-sampling my recording to 8,000 Hz. That left me with a slightly less distinct recording of a ceiling fan, which was still not very exciting.

Second, I wanted to put the sound through a speech compressor. Although an IC recorder may look like a regular tape recorder, it's actually a very specialized piece of equipment. An IC recorder is designed to record the human voice, and the way it stores information works very well for voices. When you give it other sounds, it does its best, but occasionally it can turn non-voice sounds into voicelike ones. I used the public domain "Speex" codec to compress my downsampled fan noise, uncompressed it, and sat down to listen.

First, I found what sounded like an impressive EVP, but turned out to be my wife's voice, heavily distorted, asking, "Can I turn the tape off, dear?"  The speech compressor turned the fan noise into a sound like a jackhammer. Taking advice from www.aaevp.com, I tried to listen "through" it. After a while, I started to hear what sounded like voices, singing.

The AAEVP FAQ says:

I am hearing music in the noise coming from my box fan. How is this possible?

We cannot give you a definitive reason why you hear the music, only a hypothesis. The sound of a fan is often used by EVP experimenters to record the phenomenal utterances. We think the sound, as physical energy, is transfigured into voice in electronic equipment in much the same way that ectoplasm, which seems to be a "near-physical" form of energy, is transfigured in a Spiritualist séance to form faces over the medium. We also think similar principles are involved. [...]

It seems more likely to be harmonics off the fan's hum being turned into voicelike sounds by the speech compression, but you be the judge.

Next, I tried listening for speech. One thing that I hadn't realized about EVP work is the sheer amount of white noise that you have to listen to to find an EVP. After listening to white noise for a while, you start to make things up, or at least leap to conclusions. On my second listen, I found a point, about eight minutes into the recording, where a voice seems to say "They can't keep him." It's at best what they call a "Class C" EVP– inaudible without headphones, and very ambiguous even with them. Listening to it at the time, the voice seemed clearly there, although hard to make out. I've listened to it since, and I don't always hear it. Still, not bad for an hour's worth of recording time. On the original tape all we have is the fan whirring away.

So, Stephen, I hear you asking (after listening to that fan noise for hours, I can hear almost anything), how do you know that it's not an actual ghost recording an actual EVP? And if I replied to voices in my head, I'd point out that the original recording contained no voices until I downsampled and compressed it. Unless the ghost has invaded my computer, those voices come from turning noise into speech.

There are more experiments to run:

1. Repeat the experiment with one of the brand of voice recorders used by TAPS.
2. Which kinds of sounds are the most likely to produce EVP? Is white noise indeed the best? Put different original recordings through this process and listen for voices.
3. Are there certain phonemes more likely to be produced by non-voice sounds?
4. Experiment more heavily with noise reduction.

To give some credit, some EVP practitioners are aware of some of these issues, and are taking some steps to deal with them through making backup recordings. There's no sign of TAPS using these methods, though, and even they would not necessarily eliminate all errors.

I can't prove that EVPs aren't ghostly voices, and I haven't tried very hard to do so in this small tale. I'm trying to show some alternative ways for voices to have gotten on the tape. To me, it seems as though the techniques that EVP practitioners have developed over the years are carefully honed to increase the likelihood of fooling oneself. I suspect that EVP-gathering methods are chosen because they "get results", not because they are objective or coherent.