Thanks for all the replies. I think quite a few of them offer important clues, particularly those who spoke from first- or second-hand experience of working in bronze.
Thanks to @Alessan and others who mentioned classical statues. This is quite relevant to me because my wife and I just spent three weeks in Europe, including Florence, Venice, and Athens, and saw lots of classical sculptures, including Michelangelo’s David.
However, I’m not sure I’m convinced that the presence or absence of an expression is major factor in the unreal quality we’re observing. ISTM that most of the clues we get about emotions come from the eyes, and that the difference between a toothy smile and a grimace is mostly in the eyes and eyebrows.
For instance, here’s a less familiar view of David, looking straight into his face.
I think you’ll agree that his tension and anxiety, as well as his determination, are seen mostly in his eyes and eyebrows, much less so in the mouth.
(Seeing him in person, this view was the biggest revelation for me. We’ve all seen the standard photo of him, straight on, with him looking to the right. But that angle doesn’t convey this expression at all. It’s only when you go around and look him in the face that you see what’s really happening with him. So I was rather gratified to read that Michelangelo did not intend him to be seen the way he is displayed in the museum, but from the side, as Goliath would have seen him.)
Oh, and BTW, not all classical statues eschewed emotional expressions. (click to see the full image)
Obviously, as others have pointed out, not every sculptor is Michelangelo, and working in marble is quite different from working in bronze. To the extent that it is harder to render eyes effectively in bronze, that certainly would add to the uncanny valley effect.
One thing I note is that many of the bronzes lack fine detail, and have a sort of balloonish quality to the features, reminiscent of Jeff Koons’s animals. The Fonz and Oprah, among the examples above, seem this way. (Of course, their shiny finish, missing in the Ginsburg piece, contributes to the effect.)
But others, like the Queen, seem quite detailed. For those that have some experience in bronze, is this entirely an artistic choice or am I right in assuming that the former style is easier to achieve technically?
But all these contributing factors aside, I think there’s still something missing, and I’m reasonably sure it’s what I mentioned in the OP: the artists aren’t working from life, but from 2D images.
I wonder if high tech would do a better job: laser scan the subject, 3D-print a model, then use that to make a bronze casting. Would that work better? (Is it even safe to laser scan a person with their eyes open?)
Of course, this would remove most of the artistic input to the project, but would it also eliminate or reduce the uncanny valley effect?