We outline a method by which the angular radii of giant and main sequence stars located in the Galactic Bulge can be measured to a few percent accuracy .
The method combines comprehensive ground-based photometry of caustic-crossing bulge microlensing events , with a handful of precise ( \sim 10 \mu { as } ) astrometric measurements of the lensed star during the event , to measure the angular radius of the source , \theta _ { * } .
Dense photometric coverage of one caustic crossing yields the crossing time scale \Delta t .
Less frequent coverage of the entire event yields the Einstein timescale t _ { E } and the angle \phi of source trajectory with respect to the caustic .
The photometric light curve solution predicts the motion of the source centroid up to an orientation on the sky and overall scale .
A few precise astrometric measurements therefore yield \theta _ { E } , the angular Einstein ring radius .
Then the angular radius of the source is obtained by \theta _ { * } = \theta _ { E } ( \Delta t / t _ { E } ) \sin { \phi } .
We argue that the parameters t _ { E } , \Delta t, \phi , and \theta _ { E } , and therefore \theta _ { * } , should all be measurable to a few percent accuracy for Galactic bulge giant stars using ground-based photometry from a network of small ( 1m-class ) telescopes , combined with astrometric observations with a precision of \sim 10 \mu { as } to measure \theta _ { E } .
We find that a factor of \sim 50 times fewer photons are required to measure \theta _ { E } to a given precision for binary-lens events than single-lens events .
Adopting parameters appropriate to the Space Interferometry Mission ( SIM ) , we find that \sim 7 minutes of SIM time is required to measure \theta _ { E } to \sim 5 \% accuracy for giant sources in the bulge .
For main-sequence sources , \theta _ { E } can be measured to \sim 15 \% accuracy in \sim 1.4 hours .
Thus , with access to a network of 1m-class telescopes , combined with 10 hours of SIM time , it should be possible to measure \theta _ { * } to 5 \% for \sim 80 giant stars , or to 15 \% for \sim 7 main sequence stars .
We also discuss methods by which the distances and spectral types of the source stars can be measured .
A byproduct of such a campaign is a significant sample of precise binary-lens mass measurements .