Interpolated frames are always problematic for factual analysis because the computer doing the interpolation is working from an algorithmic protocol which generalizes how to calculate changes of position, but the specifics of the image may cause the computer to miscalculate (anatomically), especially on Patty, when some of her features are at the grain level of the film, and grain patterns can vary, so the computer is interpolating changes in grain pattern, not changes of true anatomy. I've taken two identical frames of Patty, but copied in two independent instances, and by overlapping them, and cycling back and forth from one scan to the other, I can create the appearance of facial movement, even in the same frame (which of course has no natural movement. The apparent false movement is the fluctuation of the grain pattern.
Any proper analysis of Patty requires a consideration of how close the studied feature is to the grain pattern. Generally the smaller the feature, the greater the likelyhood of grain noise being the source, not true image detail. I've seen other researchers neglect this variable, and produce flawed analysis, so as a general rule, when I see something about her anatomy that is curious or questionable, I look for verification of the same feature in other frames. The more it shows in other frames, the more likely it is real, and not grain noise.