The Rhizomorphales, the most derived portion of the lycopsid (clubmoss) Glade, is now represented only by the diminutive genus Isoetes. However, during their Late Palaeozoic acme the rhizomorphic lycopsids exhibited a wide range of architectures and body sizes, from recumbent pseudoherbs to trees 40 m high. All possessed the rhizomorphic syndrome: a centralized rootstock and secondary thickening, reflecting an inescapable developmental constraint of bipolar determinate growth. These features in turn allowed acquisition of the tree habit by the lycopsids, independently of the physiologically and ontogenetically distinct lignophyte Glade that includes the seed-plants. Differences among lycopsid genera in the number and size of four major growth modules – rhizomorph, stem, lateral branches (two positionally distinct submodules), and isotomous crown branches – resulted from differences in the relative size, number and equality of dichotomies of the apical meristems.

A detailed experimental cladistic analysis of the best known fossil rhizomorphic lycopsids demonstrates extensive iteration among several distinct growth architectures characterizing ten genera. Scenarios can be constructed for the type of morphogenetic transitions necessary to (1) derive one genus from a putative ancestor or (2) explain the relationship of two genera relative to a putative outgroup. The scenarios are best formulated within a synthesis of terminology devised primarily by zoologists to describe via size–shape trajectories various modes of evolutionary–developmental change: heterotopy, heterochrony sensu lato, and allometric modifications. Many examples of these phenomena are evident among the rhizomorphic lycopsids, and can be explored by reconstructing hypothetical ancestors occupying interior nodes of the cladogram. Iterative origination of the small-bodied genera from tree-sized ancestors is inferred, by various forms of paedomorphosis and decreases in the number of developmental stages that are sufficiently profound to locally perturb perceived phylogenetic relationships. This study highlights several problems of cladistic analysis in general and of fossils in particular, especially the significance for character polarization of both perceived primitiveness and large phenetic gaps. Many phylogenetic studies of several extant angiosperm clades also imply frequent architectural transitions, but few suggest repeated origins of non-trees from trees.

A simple model for control of development focuses on D-genes: switches that control morphogen production. This perspective emphasizes the importance of a series of inter-related hierarchies reflecting ontogenetic time, size, burden and complexity within species, and phylogeny among species. The increasingly evident simplicity and common origin of D-gene control in all living organisms is used to formulate a neoGoldschmidtian paradigm of instantaneous, non-adaptive, saltational speciation via teratological ‘hopeful monsters’ that escape the constraints of developmental canalization. The scenario does not require large-scale mutations or high levels of fitness in the mutants - merely temporary release from competitive selection by establishing the new lineage in a vacant niche. The model is more appropriate to higher plants than to the more developmentally complex and constrained higher animals, and more appropriate to fossil plants than to their more ecologically complex and developmentally constrained living descendants. Future progress in understanding transitions in plant form requires reciprocal illumination between such scenarios and empirical observations of gene expression analyzed in a cladistic context.