How exactly did the nose get that long? A critical rethinking of the Pinocchio effect and how shape changes relate to landmarks

The Pinocchio effect has long been discussed in the literature on geometric morphometrics. It denotes the observation that Procrustes superimposition tends to distribute shape changes over many landmarks, even though a different superimposition may exist for the same landmark configurations that concentrates changes in just one or a few landmarks. This is widely seen as a flaw of Procrustes methods. Visualizations illustrating the Pinocchio effect use a comparison of the same pair of shapes superimposed in two different ways: in a resistant-fit superimposition that concentrates the shape difference in just one or a few landmarks, and in Procrustes superimposition, which distributes differences over most or all landmarks. Because these superimpositions differ only in the non-shape aspects of size, position and orientation, they are equivalent from the perspective of shape analysis. Simulation studies of the Pinocchio effect usually generate data, either single pairs or larger samples of landmark configurations, in a particular superimposition so that differences occur mostly or exclusively at just one or a few landmarks, but no steps are taken to remove variation from size, position and orientation. When these configurations are then compared with Procrustes-superimposed data, differences appear and are attributed to the Pinocchio effect. Overall, it is ironic that all manifestations of the Pinocchio effect in one way or another rely on differences in the non-shape properties of position and orientation. Rigorous thinking about shape variation and careful choice of visualization methods can prevent confusion over this issue.