I'm glad to hear surprise at finding out how far back the flower was tilted.
I did not intend to withhold the geometry -- it was just very late at night (2:26 am) and I was too tired to do more. But in retrospect, withholding the geometry was the perfect thing to have done because it caused people to react to what's actually in the images instead of having their interpretation colored by prior knowledge.
The images look flat to me too. I used to be surprised by that, but finally I came to grips with how perspective cues actually work and realized that this sort of image essentially doesn't have any. Many other macro images don't either, for all the reasons that Chris talks about, and maybe a couple more besides.
<lecture mode on>
Perspective cues rely on scale differences between foreground and background. The angle of view gets involved, but fundamentally what matters most is the ratio between depth of field and distance to subject.
When you shoot a flower in a garden with normal, wideangle, and telephoto lenses, it's easy to tell the difference between those shots by comparing the size of foreground and background flowers. If two similar flowers are 1 foot different in depth, and you're standing 2 feet away from the closer one, then the size ratios are 3:2. Stick on a wideangle, move in to 1 foot, and the size ratio increases to 2:1. Switch to a telephoto, back off to 10 feet, and the ratio drops to 1.10:1.
A ratio of 2:1 is compelling and 3:2 is pretty obvious. But 1.10:1 is a lot harder to see, so the shot from 10 feet looks flat. Shoot it from 100 feet through a telescope and it will look a little flatter, but that size ratio between foreground and background flowers never goes below 1.00:1 no matter how far back you move.
With gardens, streets, and portraits, many setups have big ratios of depth to distance so perspective is a strong cue.
With studio macro, it's more common to have small ratios so perspective gets a lot weaker.
In the case I illustrated, the distance from entrance pupil to focus point is around 60 mm. Since the depth of subject is about 6.5 mm, the size ratio is only about 1.11:1 even for features at the very front and back of the in-focus range. For features not at the very front and very back, it's even less.
Brian's setup is larger than mine, but the ratios are similar. Hence my comment in his thread, "so I'll bet that very similar images could have been shot even with a much longer lens". Yes, switching to a longer lens would have diminished the perspective cues even farther. But given how weak those cues already were, I think the difference would not have been large.
<lecture mode off>
same angle of view as looking through a 237mm lens on 35mm at infinity
A lot narrower than I expected.
But very typical of macro. This is another case where relying on experience in other regimes leads to the wrong intuition. In that respect it's similar to some other issues we've discussed, like the effects of sensor size and apertures in macro versus landscape photography.
ChrisR wrote:rjlittlefield wrote:it's hard for me to tell that there even is a difference, let alone what it is.
Even if you make them both full screen width?
Yes. I can say this quite confidently because I did my stacking at full screen 1680x1050, and I had to be very careful to crosscheck the source frame numbers against my notes because the overall appearance was so similar.
Again, I emphasize that the differences are easy to see in the animation that flashes directly between them. That display technique taps into the piece of our visual system that evolved to give us early warning of jumping tigers and falling trees. It's a movement detector and it's finely honed. The differences are "obvious".
But when images are viewed side by side, or one at a time, differences have to get detected by other parts of the visual system that are far less effective. There is a phenomenon called "change blindness", with a horde of experiments demonstrating that people very often do not notice differences between two images unless they see the change happen, even when the differences are blatantly obvious once they are noticed. Google search on
change blindness demos and you'll find some great demos using images that are identical except for isolated differences.
The perspective images here are a little different from the typical change blindness demos in that the perspective differences come as slight changes in scale spread across the whole image, but the concept is very similar. When I see the two images separately, the task I'm faced with is essentially to evaluate each image on its own, label it according to what I think the perspective is, and then compare the labels. Problem is, the differences in the perspective are too small for me to evaluate reliably, hence I'm stuck not being able to tell whether there even is a difference. Show them to me side by side and I can puzzle it out, but that's a matter of cognition, not perception.
--Rik