The taxonomy of peculiar galaxies is confusing and possibly inconsistent, but studies using objective classifications calibrated by simulations offer a way forward. These studies show that contamination by bandshifted late-type galaxies into samples of "peculiar" galaxies is small until z ~ 1.5, at which point the observed fraction of irregular/peculiar/merging systems is already much higher than predicted on the basis of no-evolution models. We therefore conclude that "morphological K-corrections" are a second-order effect, and do not account for the proportion of irregular/peculiar/merging systems seen on deep HST images. At higher redshifts contamination by bandshifted spirals may be more important, although genuinely protogalactic systems are definitely seen. At least some "chain galaxies" are protogalaxies - strongly starbursting systems forming their first generation of stars - and not not edge-on low surface brightness systems.
I thank my collaborators Richard Ellis, Jarle Brinchmann, Karl Glazebrook, Andy Fabian, Sidney van den Bergh, Nial Tanvir, and Basilio Santiago for their many contributions to the projects described in this article. I am also grateful to Simon Lilly and the rest of the CFRS team for useful discussions, and for permission to describe results in advance of publication.