In the slides, it was argued that you cannot run a variance argument with shape that replaces squares with circles.

In the slides, it was argued that you cannot run a variance argument with shape that replaces squares with circles. Yet later it was argued that you can run a variance argument with shape just by observing that the shape of ordinary objects varies depending on the angle from which you look at them. How do we reconcile these seemingly opposite claims? 

A

B

C