A scientist doesn't literally need to perceive the nuances of rhythmic quality in order to work on the problem of measuring it, but it would certainly help.
Here's how I look at the problem.
What is rhythmic quality and where does it come from? Let's talk live music, let's talk what musicians do. A musician interested in Mozart probably has no trouble sensing the nuances of rhythmic quality that apply to Mozart's music (typically of Mozart is the need to differentiate phrases but using a different quality). Musicians are doing subtle things with varying time and dynamics. These things can vary on a micro level or a macro level. Also involved, possibly, is variation in timbre. When multiple instruments are playing, they may be doing slightly different things.
But can a musician describe exactly what they are doing in terms that can be measured? Probably not. However, I know of one guy who has investigated this. (There are probably more.) He is Manfred Clynes and he wrote software to try to imitate musicians. See http://en.wikipedia.org/wiki/Manfred_Clynes. His software puts in variations in tempo and dynamics, on both micro and macro scales using a kind of fractal structure.
I am not 100% impressed with Clynes' work. I think he oversimplifies it, and he's in a hurry to claim a revolutionary discovery. But anyway it suggests lines of future investigation.
Now consider the distortions in audio reproduction introduced by microphones and speakers. Mike positioning also changes the sound.
I hear these changes/distortions as messing with the rhythmic quality, and so do many people. The same way my brain perceives rhythmic quality in live music, the same pattern-recognition circuits, get involved in evaluating reproduced sound. Seems to me factors like group delay, tonal balance, direct-to-reflected-sound ratio, and transient compression could affect perceived rhythmic quality.
So how do we go about measuring rhythmic quality? It almost certainly will be a multi-dimensional measurement (more than one number). We start by making some guesses what it involves. Start with measuring variations in time and dynamics. Say we perform a bit of music in ten ways, X1, X2, .. X10. Using their brain, we have someone group these by similarity. Maybe X1, X3, X8 are similar. X2,X4 are similar. Etc.
Then we look for a multi-dimensional measurement that, in essence, groups X1 through X10 in a similar way. If we hear X1, X3, etc as similar, and the measurement gives close sets of numbers for each of those, then we are onto something.
We might have to proceed more slowly, looking for one dimension at a time.
Now I predict someone is going to tell me that audio systems can't change tempo or introduce subtle timing variations. Folks, it's about how it is perceived. A system can certainly obscure or overemphasize transients. Group delay could smear a transient. The perception of rhythmic quality is essentially a perception of a pattern across a bunch of transients. If you obscure or overemphasize a few of those, you change how they are perceived.
I'm not saying that a system can make a piece sound like a faster tempo. But it can affect whether that tempo makes musical sense.
