A few nights ago I had the shortwave receiver on, and there was someone working another station I didn't hear. Eager to test my little Fourier transform utility, I captured some audio from my laptop microphone and fed it into the FFT. At one point the signal consisted of a long run of "dit"s, which is makes the audio signal quite spectrally pure. Here's a part of the result, featuring that stream of dits:
The audio carrier is visible at about 1002Hz, and then a number of sidebands each side, showing how the carrier is modulated with dits running at about six every second. Since the signal itself was pretty clean, a number of odd-numbered harmonics are also visible: the 3rd, 5th, and beyond, depending on your standards of evidence. This reflects a square wave's spectrum: odd harmonics only.
Pop quiz: does the human ear work in the time domain or frequency domain?