A PHYLOGENETIC STRUCTURE FOR THE MONOCOTYLEDONS, AS INFERRED FROM CHLOROPLAST DNA RESTRICTION SITE VARIATION, AND A COMPARISON OF MEASURES OF CLADE SUPPORT

Phylogenetic relationships within the monocots were analyzed by cladistic analysis of restriction site and length variation in the inverted repeat region of the chloroplast genomes of 52 taxa. Nine equally most parsimonious trees were identified and 38 clades were resolved by the strict consensus tree. A clade consisting of Acorus and Gymnostachys (traditionally included in Araceae) was resolved as the sister group of all other monocots. Among the remaining monocots a dichotomy was resolved between a clade that includes Alismatanae and Aranae (i.e., Araceae sens. strict.), and another that includes all remaining monocots. Within the latter group, Lilianae, including Pandananae, Cyclanthanae, and Velloziales, constitute a paraphyletic group, within which is nested a clade in which Arecanae are resolved as the sister group of a Bromelianae-Commelinanae-Zingiberanae complex. Bootstrap analysis, Bremer support (i.e., ''decay'') analysis, and successive character removal were conducted to generate three measures of clade stability. Despite overall concordance between the three measures, several clades had relatively strong support as assessed by one method and weak support as assessed by another. Three procedures were employed for Bremer support analysis, including the ''subparsimonious search procedure'' that is employed most frequently. Each of the other two procedures involves a series of subanalyses, each targeting a single clade resolved in all equally most parsimonious trees. The ''clade deresolution procedure'' estimates Bremer Support for a target clade by determining lengths of cladograms that differ from equally most parsimonious cladograms only by the lack of resolution of that clade. The ''constrained inconsistency search procedure'' estimates Bremer Support for a target clade by a series of constrained cladistic analyses, each with monophyly forced for a clade that is inconsistent with the target clade. The subparsimonious search procedure was the most successful of the three by one measure, the number of clades for which it provided the most accurate (i.e., lowest) estimate of Bremer support. However, for some clades for which this procedure did not produce the most accurate estimate, its estimates were several steps higher than those of the other two methods.