LIKELIHOOD TESTS OF GENERAL PHYLOGENETIC HYPOTHESES: HOW MANY TIMES DID BELLEROPHONT MOLLUSCS EVOLVE?

Workers disagree about the basic phylogenetic relationships of bellerophont molluscs. One view posits that bellerophonts arose at least twice, once at the base of gastropod phylogeny and again among monoplacophorans. Another view is that bellerophonts are a monophyletic group of basal gastropods. Still other views posit that bellerophonts represent a single clade of monoplacophorans or that bellerophonts are derived gastropods with secondary bilateral symmetry. The different inferred phylogenies reflect different ideas about biological implications of shell and muscle scar characters that, in turn, yield different ideas about relative phylogenetic information of those characters. These ideas range from treating some characters as diagnostic of basic groups and others as reflecting homoplasy rather than phylogeny, to treating characters as nearly equally informative about phylogeny. Therefore, testing the different phylogenetic scenarios really is a question of testing the different character evolution scenarios.

Likelihood can test alternative character evolution and phylogenetic parameters simultaneously by measuring how well data match the expectations of the hypotheses. Likelihood tests also allow different types of data (e.g., stratigraphic and morphologic data) to test hypotheses while calculating likelihoods in very different manners. I analyzed 80+ Cambro-Ordovician species of bellerophonts, limpets and anisostrophic gastropods. The analyses began with 86 characters and 168 derived states (including 6 muscle characters with 11 derived states), although 10 shell characters were eliminated after separate analyses rejected independence. I also used over 700 taphonomically and geographically controlled occurrences to estimate tree likelihoods given stratigraphic data.

The best polyphyletic tree (with one bellerophont group paraphyletic to anisostrophic gastropods and another among monoplacophorans) and accompanying character evolution parameters are significantly better than the best monophyletic tree and accompanying character evolution parameters. The character evolution parameters maximizing both hypotheses are similar (e.g., four general rates each), but show noteworthy differences in rates for characters key to earlier phylogenetic inferences.