Recording at 192 kHz is entirely useless, unless for some reason you need to record extreme ultrasonic information, like bat sonar. When it comes to actual audible information within the range of human hearing, there is no detail that lower rates cannot capture. Higher sample rates do not improve anything within this audible range—they just allow gentler anti-alias filters to be used, as well as giving more headroom if further digital processing needs to be applied later on. Even 96 kHz is overkill for many applications, though it may be helpful in some cases.

Really the results depend on the quality of the conversion that is used. Top quality A/D converters will give better results even at 44.1 kHz than any lesser design could do at a higher sample rate. The important considerations to look for are the quality of the anti-alias filter, which aims to prevent high frequency non-harmonic distortion while minimizing time-smearing artifacts, and the signal-to-noise ratio of the converter chip and its surrounding circuitry. If the equipment is designed and implemented well enough, the sample rate will ultimately prove to be mostly irrelevant.

According to Dan Lavry, who designs high-end converters, the theoretical 'ideal' sample rate is right around 60 kHz. This allows for capturing all audible detail while still leaving enough room for successful anti-alias filtering without creating any significant artifacts. However, since this rate was never adopted as a standard, the most practical rates to choose from are either 48 or 88.2, in my estimation.

It is also worth noting that many digital processing plugins will oversample to higher rates to perform their calculations and then downsample back to the original rate again. So in many cases it isn't even necessary to have recorded at a higher rate in the first place, since the oversampling gives the needed mathematical headroom while not wasting hard drive space by recording ultrasonic information without purpose.