CONTINUOUS TRACK ANALYSIS - A NEW PHYLOGENETIC AND BIOGEOGRAPHIC METHOD

Continuous track analysis (CTA) can depict reticulate evolutionary pat terns in phylogenetics and biogeography. A reticulate connection impli es convergence, hybridization, or introgression in an evolutionary gra ph of taxa and implies dispersal in an evolutionary graph of biogeogra phic areas. CTA finds graphs that (1) have a minimal number of connect ions and (2) imply that most character states or taxa have distributio ns or tracks across taxa or areas (objects) that are continuous, i.e., can be traced across the connections among the objects including that state without traveling through any other objects. Continuous tracks imply either that character states in phylogenies have unique evolutio nary origins or that taxa in biogeographic analyses are monophyletic. Relatively simple graphs usually cannot imply completely continuous tr acks. Therefore, CTA graphs seek to minimize the number of track fragm ents, which are locally continuous parts of a track; tracks with more than one fragment are discontinuous. Minimizing fragments is the same as minimizing character-state transitions only if there are no reticul ations. Because hypoi hetical ancestors do little to reduce the number of fragments, CTA tends to place known taxa or areas at internal node s. A heuristic algorithm analogous to tree bisection-reconnection is u sed to find highly parsimonious CTA graphs. In phylogenetic analyses, CTA employs a special complementary binary coding convention that sere ndipitously solves the missing characters/missing data problem. Althou gh the problem of ancestors ''inheriting'' states from hybrid descenda nts is irrelevant if reticulations merely represent convergence patter ns, CTA includes an optional algorithm that avoids such instances by e xplicitly identifying ancestors and descendants. CTA was compared with standard parsimony analysis using a data set of 17 Neogene species of North American fossil hipparionine horses. CTA separates the three ma jor clades and illustrates their convergent features with reticulation s, whereas standard parsimony analysis groups the three in an unresolv ed polytomy. CTA also minimizes the number of hypothetical, unsampled ancestors and lineages.