“…When measurement error has precise directions in shape space which can be modeled (even based on a subset of specimens during a preliminary study), it can often be removed from the data. This strategy—which is accomplished by projecting observations to the subspace orthogonal to a given vector in multivariate space (Gharaibeh, ; Valentin, Penin, Chanut, Sévigny, & Rohlf, )—has been fruitfully used on empirical datasets to remove artefactual variation due to position of the head in pictures of human faces (Gharaibeh, ) and body arching in fish (Franchini et al., ; Fruciano, Tigano, & Ferrito, , ; Fruciano, Franchini, Kovacova, et al., ; Ingram, ; Valentin et al., ), as well as variation due to sexual dimorphism (Fruciano et al., ). Similar procedures could also be used to estimate the amount of variation realistically attributable to measurement error.…”