Here's how I see it: I look at a button as defined by its shape as "perceived touchable area". A circle enclosed in a square has a smaller area than the squar

Here's how I see it: I look at a button as defined by its shape as "perceived touchable area". A circle enclosed in a square has a smaller area than the square enclosing the circle. You're seeing the new buttons as circles (thus smaller area), even though behind the scenes (as your little experiment has demonstrated) the buttons are still squares (real touchable area) on a grid. I would also argue the eyes have a harder time parsing the layout, as the visual boundaries between the circular buttons have more complex shapes than the old UI, where the boundaries were very thin lines. If you want to run another "experiment" – albeit theoretical – imagine we shrink the "perceived touchable area" of each button even further, say we get rid of the circular background and keep solely the digit displayed on the button. Your eyes will now perceive the visual representation of the button, as only the digit, and I would say our initial impulse is to touch what we see and not some imaginary boundary we know exists.