I've read some skepticism about this, primarily from those pointing out that most of the little piezoelectric speakers in phones don't have sufficient frequency response to reproduce the relevant 18kHz - 20kHz sounds. Seems like "vibrate" function might be able to pull this off nearly as well.

Well, if anything, a smaller diameter speaker will be more efficent at higher frequencies unless someone has found a way to dampen it in some way. Most tweeters for stereos are often no larger than one inch because of this reason -- and we have very little reason to attempt to reproduce noises in excess of 20KHz because we cannot hear them and it has psychological effects at that point. The audiophiles will argue this one until they are blue in the face, claiming that music doesn't sound the same unless recorded and reproduced with 96KHz or better sampling frequencies, citing harmonics in the 50KHz range from something as simple as a trumpet.

What I find a good bit more interesting is to look at a spectral analysis of older synthesizer music (like Nine Inch Nail's Pretty Hate Machine) and look at the range around 12KHz to 20KHz. Since the sampling rate is 44.1KHz, you should see no energy above 22KHz anyway (Nyquist theorem), but what is fascinating is how much of this music has a constant spike at 15KHz or so. When broadcasting, FM radio uses a pilot at 19KHz, which means you can expect some notching around that point but most of the upper frequency harmonics are due to percussion devices -- cymbals. The spike I saw probably has to do with the EQ or model of synth that Trent was using for the album.

I know my high frequency hearing isn't where it used to be (open mike night at Chez Catonic with the PA up too loud) but I can still hear some TVs. Interesting idea the kids have come up with. :)

RE: Pupils perform 'alarming' feat | Metro.co.uk