I recall looking through the viewfinder with one eye and around it with the other, figuring it would match at the "natural" size. It was about 55mm. But that's not necessarily right...
Plus it depends on the nature of the print! Look at the final print. Say, a 4 by 6 photo, at reading distance. Hold it up, keeping the distance to the eye the same, and it should look exactly like a wire frame (a window) would, in its original position.
So it depends on the size of the print and the viewing distance. Cropping changes that, meaning you need a shorter lens if you plan margins to crop later. The modern computer viewing is probably different than "print", and even 4 by 6 is not what was used to come up with that.
If you want people to not look funny, use a definite telephoto length.
The back of the eye is not flat, and the projection is not "corrected" (but the mapping of which pixel is where un-does the projection effect) so there's really no such thing without special equipment. However at reading distance the scanning of the macula over the "window" gives an effect that is pretty close to flat, except that you have two eyes and they can't both match at the same time, and the perception is corrected for eye placement vs head rotation axis and that shows visible differences if you trace a window vs hold up a normal photo.
But to be precise about what is meant, and to show that it's accurate, the "window" is the definition to use. That's what movie directors are doing when they hold their hands out to define corners of a frame.
If the print is held so that a person's face is life-size (put it where the window would be), it looks objectionable if you are closer to the print/window than you would be to look at the person normally.