A similar confusion can be found in discussions of the related problem of defining observables in canonical general relativity. The claim gets its traction from the fact that it is very difficult to construct observables in canonical general relativity, while (apparently) it is relatively straightforward in the standard Lagrangian description. (See, e.g., Curiel, 2009, pp. 59–60, for an explicit statement of this claim. Curiel cites a theorem of Torre, 1993, to the effect that there can be no lo