Well, since the topic has been raised, there's another question I'm curious about.
I think in the U. S. publication is the key to copyright. For sure, unpublished letters can be copyrighted when they surface, even if they were written over a century ago.
Millet's Le Semeur was painted in 1850, but when was it "published?" What constitutes publication for an oil painting which isn't mass-produced? Is it published when it's exhibited in a gallery? If so, then no problem (for most famous classic old paintings).
Otherwise, do you need to research when the first printed reproduction was published? Truly "slavish" color reproductions weren't cheap and common (and used in textbooks and encyclopedias and magazines) until, oh, maybe the 1930s.
So, if "published" means "printed" then printed reproductions that are old enough to be in the public domain would be hand-tinted engravings, or chromolithographs (or whatever the Currier and Ives things are called), or black-and-white photos. Which are significantly different enough from a color photo of the same painting.
Just wondering...