It is possible for a device to record something at say 20khz and the (audible) harmonics of that signal to be non linearly boosted during playback.I'm going to provide semi serious answer to what is really silly topic.
For the OP story to be true, one would have to assume that this "Olympus digital recorder" device, sensitivity aside, is not only capable of recording wider range of frequencies than those audible to human ear (which on its own wouldn't be impossible, if a little out of spec of regular microphone one would expect in such device), but also upon playback it would have to be actively converting those frequencies to audible range and transposing sounds voiced at pitches outside hearing threshold by several octaves to something you would now recognise as human voice. That's not going to happen in this kind of portable device.
It would require an unstable amplifier/filter though.
There are sounds that we are more attuned to depending on context, like a twig snapping in a wood. The opposite might also be true.