Measuring Evolutionary Constraints: A Markov Model for Phylogenetic Transitions Among Seed Dispersal Syndromes

I introduce a Markov probabilistic model of transitions among discrete morphological states as a method for describing and testing nonrandom patterns of evolutionary change. The Markov model assumes one-generational dependency, i.e., that the future direction of evolutionary change depends on the current morphology of a species, not on any history of changes. This model is very flexible, allowing for any number of discrete states to describe morphology, yet permit rigorous testing of even complex evolutionary hypotheses. I apply this model to changes in seed dispersal mechanisms within 571 genera of Neotropical plants, using cladistic methods to infer the ancestral and derived states within each genus. I then test a series of progressively more complex hypotheses about the constraints that might shape the patterns of observed evolutionary transitions: 1) no transition constraints; 2) all dispersal mechanisms are equally labile evolutionarily; 3) the probability of particular evolutionary transitions among dispersal mechanisms depends on the descendant state but not on the ancestral state; 4) transition probabilities differ among pairs of dispersal mechanisms, but are reciprocal within such pairs. More complex hypotheses matched the data significantly better than did simpler hypotheses. However, only one of the hypotheses (reciprocal transitions) fit the observed data and then only for the most cautious interpretation of the frequencies of transitions within genera. These results suggest that evolutionary transitions among major adaptive syndromes are indeed ordered, and the observed patterns of transitions suggest possible reasons for such macroevolutionary structure.