The voices told me to write this.

No they didn't. The voices rarely say anything of note.

I'm not cracking up. I'm talking about the ghostly voices recorded on tape and referred to as EVP's (electronic voice phenomena). They're called that because it sounds serious and scientific, and because those who strive to appear serious and scientific think they'll sound like amateurs if they call them what they really are. The voices of the dead. Making up a scientific name doesn't change the nature of something. A biologist might say 'Nereis diversicolor' while the rest of us call it a 'ragworm'. It's the same thing. Don't be scared to say what you're really studying.

These recordings are second only to photographs as solid evidence of spirit activity. In fact, nothing beyond sound and vision will convince any sceptic.

Now, the problem has always been that if the voice is clear on the tape, why did nobody hear it at the time? It's hard to give credence to the idea that spirits somehow 'write' their speech directly onto the tape, so we're left with a recording that shouldn't be. A voice that the tape heard but nobody else did. A highly directional sound source aimed straight at the microphone. So directional that even if someone held the microphone straight out in front of them, none of that sound would pass it by and reach their ears.

Now, who couldn't pick huge holes in an idea like that?

Yet the voices get recorded. So where do they come from?

New Scientist, 22nd September issue, had an interesting article on how the brain processes information from the ears, eyes, etc. It seems we don't get as much information about the world as we think.

Ears aren't perfect at hearing. Fortunately the brain is pretty good at filling in the gaps. The impulses it gets might be full of errors but the error-correction neurons are good at their jobs. That's why when someone's talking against a loud background noise, you can make out what they're saying even though parts of their speech are drowned out. The brain fills the gaps.

However, if you record a sentence, then replace parts of it with silence, nobody can make it out. There's no background noise, so the brain has no reason to try to work out the missing parts.

Play that same sentence against a white-noise background and you can understand it. Now, the brain assumes the noise is interfering and fills the gaps.

Where's this leading? Well, suppose you hear something in a silent room. A few whispered sounds. No sentence structure, nothing to interpret, it could be a fly-buzz or a creaking board. A breeze through a badly-fitted window.

Or it could just be a faint and broken voice, forcing itself through the barriers of our senses. Trying to be heard, but not getting through sufficiently to be interpreted as a voice.

The recorder picks that up. When you play it back, you have to turn up the volume to hear anything because these voices are very low. Any recorder will hiss at high volume. A tape will hiss so hard you might not hear a thing, but sometimes, just sometimes, that white-noise background will trigger your brain's error-correction and you'll hear that voice.

The voices are there. You did hear them. Your brain didn't recognise it as a voice, and dismissed it. Add in some white noise and there it is.

If, by some quirk of genetics, your brain's error-correction works all the time without needing a background noise, then you'll hear the voices yourself.

Of course, if you ever admit to that, you'd better get used to life in a rubber room. Modern society isn't all that tolerant of mutants.