The entropy region of a probabilistic graph with respect to a Cartesian 
representation of its vertex set is defined, and shown to be elemental in 
several zero-error distributed source coding and distributed computation 
problems. Inner and outer bounds are derived and discussed in special cases.