Hairs/groups of hairs just do single frequencies each. The shape of the cochlea (akin to a folded up tapering tube) creates resonant nodes so that certain frequencies will always have nodes across the hairs you associate with the frequency. All the hairs themselves should be roughly identical and only respond to intensity, not frequency. It's hard however to damage hairs where low frequency nodes occur as transmission of low frequencies to the cochlea is significantly attenuated.
Like I said earlier, it's a bizarre system! It is because of this we cannot hear 2 very similar frequencies simultaneously, particularly if one is quieter than the other, and hence the mp3 algorithms could be developed to exploit this by removing such data. It presumably is why cochlea implants are in no way of the same quality as our hearing, as a normal microphone wouldn't interface with this system where each frequency has a seperate 'channel'.