Ancestral‐area algorithms are unreliable

ABSTRACT We examine two standard phylogenetic–biogeographic patterns and how these are interpreted by current ancestral‐area algorithms. In the first pattern, a basal, paraphyletic grade is restricted to one part of its clade's overall range. In the second pattern, the clades of a group overlap in one central area but are allopatric elsewhere, resulting in a ‘star pattern’. Ancestral‐area algorithms will calculate a localised centre of origin for both patterns in the area of overlap. Yet both patterns can also be derived by vicariance (causing the allopatry) and subsequent range expansion by normal dispersal (causing the overlap). In this model, ancestors of allopatric clades were already widespread and polymorphic before the modern clades began to diverge. The overlap region is not a centre of origin, it is a region of secondary range expansion. There is no reason to assume that basal grades occupy ancestral areas or habitats. There is also no need to use a priori areas in biogeographic analysis. Instead, the distributional relationships among the clade localities can be examined directly. Distribution has emerged as a critical factor in molecular systematics, as many clades are defined more easily by their geographic distribution than by traditional morphological characters. Thus, it makes sense to review the precise geometry of the distributions in a study group, including details of allopatry, disjunction, overlap and others. A distinction is drawn here between secondary clade overlap that has developed following the origin of the clades, and primary clade overlap (true sympatry) that developed with the origin of the clades. In current practice, distribution maps have been replaced in biogeographic analysis by outputs from ancestral‐area analyses. But if the main clades recovered in a study are mapped, simple, distinctive features, such as allopatry, marginal overlap and disjunction often become evident, and we encourage authors to map their clades. The patterns discussed here, for example, the star pattern centred on New Zealand, are repeated in unrelated groups. Thus, the regions interpreted in ancestral‐area analysis as centres of origin can be re‐interpreted as phylogenetic‐biogeographic breaks (nodes) in widespread ancestors. The method proposed here – ‘allopatry indicates vicariance; overlap indicates dispersal’ – is simpler to apply than the current algorithms, and the results are much simpler, with a single cause explaining many distribution patterns, rather than each component clade of a biogeographic pattern having a separate history.