A comment on my recent post about alvarezsaur arctometatarsals made me realise there was a bit more scope for talking about this issue of the origins of this structure. Arctometatarsalian pedes are known in tyrannosaurs, troodontids, alvarezsaurs, ornithomimids and oviraptorosaurs. In other words a pretty big selection of derived theropods have at least some taxa with this condition.
This begs the obvious question of whether or not this is homologous? After all, this could easily plot on a cladogram as originating before the tyrannosaurs and being maintained throughout the derived theropods only to be lost in therizinosaurs and droimaeosaurs + birds. However, there are two good reasons to think that this convergence and not homology with the characteristic being acquired multiple times.
First off are the details of the actual pes. Although obviously we would expect different lineages to adapt and modify such a structure independently, there are some very clear differences between quite how the middle metarsal is pinched and in which way by which of the surrounding elements. It’s subjective of course but they do look quite different in form.
Secondly and more importantly, the character is not shared by all of the taxa in those various groups and especially not basal forms. Early tyrannosauroids like Dilong and Guanlong don’t have it, nor does the basal alvarezsauroid Haplocheirus, and it’s not present in at least some early troodontids, ornithomimids or oviraptorosaurs. So it is not a simple plot of acquisition before tyrannosaurs and occasional lost, but instead more parsimonious to infer that it has been gained independently multiple times (each with a slightly different form) than been lost multiple times in all those basal forms (i.e. lost in each of Guanlong and Dilong and any other tyrannosaurs or even lost at the base and then acquired again) and been modified repeatedly along each lineage.
What has drive this convergence is likely the benefits of such a structure. Work on the functional morphology of such a foot suggests that it increases running efficiency and may also provide increased turning ability. In short, this is a feature of active runners, something that certainly matches at least some other anatomical specialisations seen a number of these groups.