Abstract

Premise of research. Zosterophylls are a major constituent of Siluro-Devonian tracheophytefloras worldwideand gave rise to the morphologically diverse lycopsid clade. Despite this pivotal position in the plant tree of life,relationships within the group have remained poorly understood. Furthermore, anatomical data have been under-sampled in previous phylogenetic analyses including zosterophylls. With this in mind, we addressed the relationshipsamong zosterophylls using a data set that maximized anatomical characters and employing phylogenetic and pheneticmethods.Methodology. A matrix of 40 total characters (12 of which are anatomical) and 19 zosterophyll species withknown anatomy, representing 17 genera, was compiled and analyzed using parsimony-constrained phylogeneticmethods, clustering, and ordination methods.Pivotal results. PheneticanalysesshowsensitivitytotaxonsamplingandsupporttheplacementofRenaliaamongthe zosterophylls but do not support taxonomic inferences strongly congruent with those supported by phylogeneticanalyses. Phylogenetic analyses consistently recover two major clades: one lacks internal resolution and comprises thebulk of the zosterophyll taxa included in the analyses; the other clade includes the zosterophyllVentaruraand thelycopsidSengelia, often accompanied byDiscalisandTrichopherophyton. Analyses using subsets of characters (onlymorphology or anatomy) recover trees that differ from those obtained using the total set of characters (morphologyand anatomy).Conclusions. The incongruence between the results of the total character analyses and those using only morphol-ogy oronlyanatomy highlights theimportanceofbroadening thesampling ofmorphological characterspace. Becauseboth anatomy and morphology are part of the identity and evolutionary history of a species, the relationships recov-ered by the inclusion of both morphological and anatomical characters are more likely to reflect natural evolutionaryrelationships. Breadth of character sampling and not the amount of phylogenetic resolution should be the primarycriterion for selecting among alternative hypotheses of relationships.