Our analysis of the biogeography of the Campanulidae indicates that the early evolution of this major angiosperm clade took place in the Southern Hemisphere, perhaps initially in Australasia. Specifically, the Aquifoliales, Asterales, Bruniales, Escalloniaceae, and Apiales all appear to have originated in the Southern Hemisphere, despite the fact that many of these groups are today most diverse in the Northern Hemisphere. It is difficult to infer the ancestral environment in which campanulids initially diversified. However, it is possible that they occupied, at an early stage, more seasonal and temperate climates, especially in view of the high paleo-latitudes of Gondwana at that time  and presumed seasonal fluctuations in temperature and photoperiod. Adaptation to such climates may be reflected in a possible synapomorphy of the campanulid clade , namely xylem vessel elements with predominately scalariform perforation plates, which are today most commonly found in cold climates .
There remains, however, the possibility that the campanulids originated in the Northern Hemisphere or, perhaps, were even widespread across both the Northern and Southern Hemispheres. A Northern Hemisphere origin would argue for the Southern Hemisphere as a center of survival of early-diverging lineages rather than a center of origin. Perhaps the early northern representatives of basal campanulid lineages all went extinct, possibly because the environments that they occupied became too rare. And, perhaps the representatives of these lineages that moved into the Southern Hemisphere survived there because these environments remained intact. Such a scenario would gain support if there were ancient fossils of these lineages in the North Hemisphere. In general, our inferred southern origin for particular clades is consistent with current evidence from the fossil record. For example, there are fossils that not only place Asteraceae in South America in the early Paleogene, but also suggest that they were derived from ancestors that were also Gondwanan in origin (Figure 2; [95, 142, 143]). However, there is one case that could imply an early northern origin. The Late Cretaceous flowers of Silvianthemum suecicum and Bertilanthus scanicus would place campanulids in Sweden in the Late Santonian-Early Campanian (~85 Ma; ). These fossils share several curious features with modern day Quintinia (Paracryphiaceae), such as secretory hairs on the floral surface, styles with radiating palisade cells, the postgenital union of the 3–4 apocarpous styles, and parietal placentation [145, 146]. There are, however, several other important features that do not fit, such as stamens opposite the petals, dorsifixed anthers, and, in the case of the Sylvianthemum, having 10 stamens in two whorls. These characters raise the possibility that these fossils are not directly related to Quintinia, but belong elsewhere within Asteridae (e.g., within Cornales). In fact, our biogeographic analyses might be viewed as supporting such a conclusion. Extant Quintinia are today largely restricted to Australasia and our analysis never identified a scenario for Paracryphiaceae involving the Holarctic. At the present time there is no definitive fossil evidence placing any of the early campanulid lineages in the Southern Hemisphere during the Cretaceous, but neither is there conclusive fossil evidence placing them in the Northern Hemisphere. In the absence of such fossil evidence, a southern origin for the campanulids appears to be the best-supported hypothesis.
A corollary of a southern origin hypothesis is that there have been movements within the campanulids from the Southern Hemisphere into the Northern Hemisphere. At this stage we can estimate only a minimum number of such movements; a better understanding of the real number will require more extensive sampling and biogeographic analyses within each of the major lineages. Likewise, identifying pathways of movement from the Southern Hemisphere to the Northern Hemisphere will require fine-grained analyses. If our speculation is correct that the initial diversification of campanulids occurred in more temperate climates of the Southern Hemisphere, it is possible that some early lineages were preadapted to occupy temperate zones in the Northern Hemisphere, and that they tracked such habitats northward when relevant connections allowed successful movement into the Northern Hemisphere (cf. ). In any case, northward movements would have been facilitated during the Paleogene by several corridors, such as the connection of South America to North America through the paleo-Caribbean (as postulated for Asteraceae: [111, 113]), and the collision of the Arabian Peninsula with Eurasia during the late Eocene (~35 Ma; ), which corresponds, for example, to an inferred northward movement within Apiaceae (median = 28 Ma, 95% HPD: 18–40 Ma).
The movement of the Dipsidae (the least inclusive clade containing Paracryphiaceae and Dipsacales; ) into the Northern Hemisphere during the Late Cretaceous is perhaps the most difficult to explain. We infer that the range of the ancestor of the Dipsacales and the Paracryphiaceae came to encompass the Austral region and the Southeast Asia between 98 and 73 Ma when there was quite certainly not a direct connection between these areas. A closer connection between Australasia and Southeast Asia was established only much later, in the Paleogene [108, 109]. It is noteworthy that a Southeast Asian pathway is consistent with recent evidence on basal relationships and biogeographic patterns in Adoxaceae, where there are now indications of the existence of early-diverging Southeast Asian lineages in both Sambucus and Viburnum. Recent evidence in Viburnum, in particular, points to a possible Southeast Asian origin and several movements from there into temperate Asia . If our biogeographic scenarios and age estimates were correct, the establishment of a distribution including the Austral region and Southeast Asia in the Cretaceous would most likely have entailed long distance dispersal. However, the inference of such an event prior to any direct connection could also be pointing to an instance where our age estimates, at least for the origin of the Dipsacales, are simply too old.
Importantly, our dating analyses indicate that the campanulids, and the major lineages within the clade, existed far enough back into the Cretaceous for their diversification to have been influenced by the break-up of Gondwana. As might be expected, therefore, we have been able to identify several disjunctions within early diverging campanulid lineages (e.g. Bruniales and the split between Goodeniaceae and Asteraceae + Calyceraceae; Figure 3) that are consistent with the hypothesis that the separation of the Gondwanan landmasses played a causal role. By “consistent with” we specifically mean that the geographic disjunction, the inferred biogeographic scenario, and the dating of the lineage splitting events are all consistent with the vicariance hypothesis. With respect to the dating, in particular, the inferred range of dates for the lineage splitting event must overlap the range of dates assigned to the relevant geological event based on other evidence. Of course, the need to be consistent in all three ways does highlight the difficulties of pattern-based approaches such as the one adopted here. Although this would appear to be the most stringent test, it is possible that some apparently positive cases will be better explained in the end by dispersal when, for example, taxon sampling is improved. At the same time, data from other sources (e.g., on dispersal limitations) could further bolster a positive conclusion. In any case, the approach we have taken here provides a baseline for more detailed analyses of individual clades.
Viewed from this perspective, it is clear why it is difficult to confidently attribute clade diversification to ancient continental movements. First, plant groups that are old enough to have been influenced by these events are also quite likely to have moved around a lot since that time, becoming widespread across several biogeographic regions (e.g. Ilex; Figure 2). Such movements will tend to cover up the initial geographic disjunction and, therefore, confound the inference of ancestral areas. We simply will not be able to recognize many genuine cases of ancient vicariance by virtue of subsequent geographic movements. Along these lines, dispersal limitations in some lineages (possibly Bruniales?) may allow a Gondwanan signal to persist and thus bias in favor of detecting such disjunctions. Second, as we move that far back in time, confidence intervals on age estimates tend to become wider. On the one hand, this will increase the chance of false overlap with the geological event in question, potentially yielding false positive results. In fact, this will be especially true for clades where the quality of the fossil record is poor. This will increase confidence intervals further and make almost any disjunction between continents appear to have occurred within the window of a particular geological event. On the other hand, as the range of possible lineage splitting dates widens, many additional geological and climatological events might be responsible, so that causal factors become harder to identify with confidence. This can be especially problematic when there is the added uncertainty associated with the age of the continental disjunctions themselves, as is the case with the breakup of Gondwana . In view of these potentially confounding issues, in general, it is remarkable that we have uncovered even a single possible case of vicariance.