In a nutshell, closer instruments are brighter than far-away instruments, and their reverb is less diffuse as well as brighter.
 
For an extreme example of distance-related high-frequency absorption, think of distant thunder versus a lightning strike across the street.

I second roger's point about having different libraries.  Although they sometimes are panned more or less the same, there are enough exceptions that you need to have a system. I use a combination of Garritan and, believe it or not, Cakewalk pans. For Woodwinds, I think the old Cake values are perfect:
Flutes: 6% left
Oboes 7% right
Eng Horn 13% right
Clarinets 9% left
Bassoons 10% right
 
For brass, I use Cake's values for
 
Trumpets 7% right
Trombones 26% right
Tuba  45% right
I put the horns at 22% left, a change from where Cake had them (center).
 
These values sound great for these instruments!
 
For strings, I use a Garritan pans:
 
Violin 1: 44% left
Violin 2 30% left
Violas center or 39% right
Cellos 50% right
Basses 70% right
 
I put the harp at 18%, a Cake value - sounds perfect
For percussion, I generally have cymbals et al around 37-40% left, and timpani at 50% left.
 
Where did I get these values? Actually I can't remember, but they work!

Thank you Michael.
I might add that pan is not everything because it is two-dimensional. It is necessary to add depth, that is, a three-dimensional simulation.
So I preferred to use the Virtual Sound Stage plugin, because it does a correct placement simulation inside a stage, including the early reflection of different rooms.
After that, I submit the result to a good Convolution Reverb, such as Altiverb, but via Send, to regulate the intensity of the Reverb Tail and with the pre-post fader, to simulate the distance of the microphone, as in microphone layouts for recording of real orchestras.
This way I can, for example, get the sound of the trombones more to the back of the stage, ie further from the microphone, even if the library is completely dry, as in my case I use the SampleModeling for brass.

I've thought about getting that, but the necessity of applying this complicated (and, I would guesss, CPU-intensive) technonogy to each instrument is too much for me. I do try to use reverb for spatial depth, but probably am not very successful with it. We all have our sweet spot, between untra-simplicity and uber-complexity, where we decide to hang out. for me, it's definitely closer to the simplicity end. I do want people to get a good idea of what my music should sound like if a real orchestra was playing, but I'm not after total realism. Too much work for me. Besides, the more technical details of it all elude me. I NEED to keep it simple.