Sine-Wave Speech is created by taking regular speech and essentially reducing it down to sine waves, absent all the complexities that make normal speech understandable.
Listen to a sample and you’ll hear nothing but whistles and beeps.
On the other hand, listen to the original sound (ordinary speech) and then listen to exactly the same sample again.
The theory is that the brain’s very perception of the sound is colored by its past experiences of listening to similar sounds. You perceive, at a low level, that you’re hearing speech, and therefore you can decipher it. I suppose it’s not unlike how we adjust to listening to people with heavily accented speech, but it’s so much more extreme that it’s scary.
Listen to four more samples on that page, and not only will you get the same effect or all four, but you’ll even get better at deciphering the sine-wave version the first time around.