Ancestral state reconstruction sheds new light on the loss of divarication hypothesis on New Zealand's outlying islands

Abstract The New Zealand flora is remarkable in many respects, but one of its most notable features is the convergent evolution of many plant lineages towards a divaricate habit. The adaptive significance of divarication remains controversial, but it has usually been considered a response to harsh climates or a defence mechanism against the diverse array of Ratites (large browsing birds) that once populated New Zealand. The loss of divarication hypothesis posits that divaricate‐related traits should be lost on New Zealand's outlying islands, as these islands were never reached by Ratites. Like most evolutionary components of the island syndrome, this hypothesis was tested through the pairwise comparison method, that is, by comparing traits of island endemics to those of their closest relative(s) on the mainland, assuming that the most recent common ancestor (MRCA) was divaricate. This assumption was challenged by recent results but has never been tested. Here, we test this assumption. We collated data for all but one genus comprising at least one divaricate species and one island endemic in the New Zealand region. We used ancestral range reconstruction to determine which endemics colonised the islands from New Zealand. We then used ancestral state reconstruction to test whether each MRCA was divaricate and to quantify the degree of morphological evolution in leaf size and branching angle (i.e. significant divaricate traits) on islands. Twenty‐one out of 29 MRCAs were estimated as most likely non‐divaricate, meaning that in most cases divarication evolved on the mainland and was never present on New Zealand's outlying islands. Island endemics also evolved larger leaves and smaller branching angles regardless of whether the MRCA was divaricate or not. Synthesis . These findings show how most island lineages were never divaricate and advance our understanding of the evolution of divarication in the New Zealand region. In addition, they have broader implications for the field of island biogeography, as they highlight the limits of the pairwise comparison method and provide a way to overcome them.