Why 'observable' in general relativity may be the wrong concept

Applying the gauge-theory concept of 'observable' to general relativity has quietly caused decades of confusion in both classical and quantum gravity. By separating two distinct meanings of gauge transformation — local and global — this paper argues diffeomorphism invariance only qualifies as a gauge symmetry in the global sense, where the notion of observable doesn't cleanly apply. The popular idea that all physical content in general relativity lives in correlations between fields turns out to be unfounded under this analysis.