Of course it depends on teh acoustical crossover function, which may or may not (probably not!) be the same as the electrical crossover.
The best case (IMHO) is a Linkwitz-Riley crossover, where both drivers are in phase with each other throughout the crossover frequency. There will be a null at crossover at +/- 90 degrees if their acoustic centers are 1/2 wavelength apart at the crossover frequency. These nulls will cause the image to wander as the reflected sound won't have the same frequency response as the direct sound. So ideally, they should be less than 1/4 wavelength apart; that's about 1.5 inches center to center.
So unless the midbass is 2" outer diameter and the tweeter 1", you can't achieve the ideal. Since the ear seems to be more sensitive to horizontal images than vertical, it is common practice to place them in a vertical arrangement. If they are one wavelength apart (6 inches), there are nulls at +/- 30 degrees, corresponding roughly to the floor and ceiling bounce - a popular choice among theoretically-inclined designers, and about as far apart as you can get with good integration.