(c) Vicariance
The vicariance hypothesis implies that Rhipsalis arose prior to the breakup of Gondwana - or at least before substantial water barriers had developed between its constituent parts - and had by then spread over much of the supercontinent. Estimates for the timing of this breakup vary, but there is a reasonable consensus about the sequence of events, even if the precise dates are not always known. The following summary is based on that given in Storey (1995) (Fig. 2). [See Table 1 for the geological time scale.] The first stage (in the Middle Jurassic, probably between 180 and 160 Ma [million years ago]) seems to have been the formation of a seaway between West Gondwana (Africa and South America) and East Gondwana (Antarctica, Australia, India and New Zealand) with Madagascar (which was attached to India) beginning to part from its original position near Kenya/Somalia. The second stage (Early Cretaceous, about 130 Ma) saw the beginning of the split between South America and Africa, and rifting between India and the rest of East Gondwana. The final stage (Late Cretaceous, 100-80 Ma) was the splitting off of India from Madagascar (and its eventual collision with Asia), and the separation of Australia, New Zealand and Antarctica from one another.

At the risk of some over-simplification, the small islands of the western Indian Ocean, ie the Comores, Seychelles and Mascarenes, can be regarded as the "crumbs" left behind after the carve-up of the Gondwana "cake".

Besides its occurrences in former parts of West Gondwana, Rhipsalis is also present in Central America, the Greater Antilles and Florida. I suppose the standard explanation would be that the last two occurrences are the result of bird-dispersal from South America, and as the distances involved are not particularly great and there are conveniently placed islands along the presumed path, this does have a certain plausibility. There are other possibilities, however. First, the Greater Antilles are thought to be the remnants of an isthmus or archipelago that formerly existed between North and South America and was subsequently swept eastwards by seafloor-spreading commencing in the Late Cretaceous (see Gibson and Nobel 1986: 252, fig. 11.18). The narrow neck of land connecting North and South America at the present day is a much more recent feature.

According to the scenario advanced here, Rhipsalis had spread into the rain forests of "proto-Central America" by the Late Cretaceous (probably much earlier) and was then dispersed by geological rather than biological processes to its present position on the eastern edge of the Caribbean, along with several other cacti (eg Hylocereus, Epiphyllum and Melocactus). The presence of R. baccifera in Florida is particularly intriguing, as the genus occurs nowhere else in North America. Florida is quite close to Cuba and probably within flying distance of at least some rain forest birds, but there is evidence that it is an "exotic" terrane, a small piece of Gondwana that formed somewhere near Venezuela and later split off and drifted away, finally becoming "sutured" to North America (Opdyke et al 1987). Of course the occurrence of Rhipsalis in Florida may be purely coincidental, but I see it as confirmatory evidence for the Gondwanic origin of the genus.

If Rhipsalis is a Gondwanic genus, then the cactaceae must have appeared quite early in the history of the angiosperms (flowering plants), contrary to the impression given in some popular books that they are a relatively young group. [The most extreme example of this would have to be the statement by W.T. Marshall (in Marshall & Bock 1941:1) that cacti are "possibly as young as ten thousand years"!] Barthlott (1979:39) suggested that cacti arose in the Cretaceous, but went on to say the "first divisions within the family probably took place towards the end of the Tertiary Period" [ie within the last few million years]; unfortunately, he did not give any reasons for this view, but he may have been influenced by geological evidence suggesting arid and semi-arid habitats became particularly well developed during the late Tertiary. He was also very sceptical about the suggestion that Rhipsalis is a Gondwanic genus, stating that at the time South America and Africa parted "there was no such plant as Rhipsalis in the present-day sense" (Barthlott 1979:222). [I refrain from asking the obvious question!]

Now, there is much controversy about the origin of angiosperms, but all paleobotanists agree that they were present in the Early Cretaceous, at about 135 Ma. Most claim that they appeared only slightly earlier than this, and that there are no reliable pre-Cretaceous records, but there are certainly reports of considerably older alleged angiosperms extending back at least to the Middle Jurassic (Bathonian Stage, 168-166 Ma) (summarised by Shields 1988: table 4), and of pre-Cretaceous nectar-feeding insects. Ren (1998) recently reported fossil Brachycera (short-horned flies) with a long proboscis - an adaptation for nectar feeding - from the Late Jurassic (143-150 Ma) of China and on this basis suggested angiosperms originated substantially earlier, in the Middle Jurassic (159-180 Ma).

Shields (1988) pointed out the close association between Lepidoptera (butterflies and moths) and angiosperms and that as the earliest known lepidopteran is of earliest Jurassic age, he concluded that flowering plants arose near the Triassic-Jurassic boundary (c. 205 Ma). [He went on further, claiming that dicotyledons and monocotyledons evolved independently, and that the former arose in what is now Queensland; ie they have a Gondwanic origin.] Some authors favour an origin in the Late Triassic (225-205 Ma), a time when dinosaurs, mammals and possibly birds also appeared (Crane et al, 1995). The earlier they arose, the easier it is to account for the present-day distribution of angiosperms in general - not just cacti. An extreme conclusion would be that Rhipsalis appeared in the Jurassic prior to rifting between Madagascar and Africa, ie as early as 180-160 Ma, but these two land-masses seem to have remained in close proximity (possibly even in contact) for a long time afterwards, allowing at least some interchange of plants and animals.

I suggest that Rhipsalis managed to reach Madagascar and Ski Lanka (and India?) after the development of the seaway between them and the rest of East Gondwana (Antarctica, Australia and New Zealand), ie not before about 130 Ma. This would explain why Rhipsalis doesn't occur in Australia, despite the presence of apparently suitable tropical rain forests in Queensland, but of course another possibility is that it did manage to spread across all of Gondwana, only to become extinct except in the warmer parts of the supercontinent. Note that the South Pole was close to the southern edge of Gondwana for much of the Jurassic and Early Cretaceous, and was not far from its present position in the Lat Cretaceous when Australia and New Zealand were at high latitudes. I also suggest that Rhipsalis was present in India, and has become extinct there as a result of climatic change or through human activity.
An early origin for the Cactaceae is consistent with its current diversity (about 2000 species in about 90 genera). As Gibson and Nobel (1986: 18) point out it is a "very large family - one of the largest in the New world and the second largest family that is essentially restricted to the New World (second to the pineapple family, Bromeliaceae.)"

If the vicariance scenario outlined above is correct, then not only is Rhipsalis of great antiquity, but the R. baccifera group is similarly ancient and is an example of "bradytely", ie unusually slow evolution. Other examples of bradytely include crocodiles and "living fossils" such as horseshoe crabs, tuataras and the okapi.

Evolutionary implications

(a) The place of Pereskia in cactus evolution
Earlier I said that other things being equal, the most widespread member of a group is likely to be the most primitive (or least derived) member of that group. "What about Pereskia?" is the obvious response. Surely it is the least derived cactus genus and it or something very similar must have been the stem-group for the family, yet it is less widely distributed that Rhipsalis. In fact, Pereskia occurs naturally in Central America, Cuba and Haiti, and in northern South America, from Columbia south to Peru and northern Argentina, a geographic distribution not too different from that of the American species if Rhipsalis. Pereskia and Rhipsalis species, however, occupy quite different habitats - the former are shrubs, trees or climbers living in open situations, often with more typical ground-dwelling cacti, whereas the others are (with a few important exceptions to be discussed later) epiphytic in rain forests.

It is certainly not difficult to point to primitive characters in Pereskia. One of the less obvious is the way in which it carries out photosynthesis: Pereskia exhibits C3 photosynthesis (the metabolic pathway present in most dicotyledons) rather than CAM (crassulacean acid metabolism) which is characteristic of other cacti. [there is little doubt that CAM is the more derived condition, but it seems to have evolved independently in more than one major group of plants.] However, it needs to be pointed out that the species currently assigned to Pereskia form a heterogeneous group with both primitive and derived character-states. The presence of leaves in Pereskia species is invariably cited as a primitive character, but the leaves vary greatly in size - from only 1 cm in length (P. portulacifolia) to 20 cm (P. bleo); most species have a distinct petiole (stalk) but some are sessile. The stem is non-succulent in some species (eg P. sacharosa), succulent in others (eg P. pititache) (Gibson & Nobel 1986: 41). Floral characters are even more variable - some species (P. aculeata) have a superior ovary (ie with the ovary placed above the stamens and other floral parts - generally considered to represent the primitive condition in angiosperms), but in others (P. grandiflora) the ovary is inferior (ovary below the stamens - the normal condition in cacti.) The flowers may be sessile or pedunculate, and solitary or in clusters. [They also vary considerably in colour and size, but probably no more so than in some other cactus genera.] As Buxbaum (1955:184) noted there is considerable variation in Pereskia fruits as well - they range in size from about 4 to 60 mm in diameter, and in shape from spherical to conical (P. bleo) to polyhedral. Some species have a thick pericarp (the outer coating), in others the pericarp is thin and the fruit is mostly filled with pulp. Some fruits are juicy and edible, others hard. Small wonder that some European botanists recognize a segregate genus Rhodocactus Knuth for those species with an inferior ovary. For some reason this has never gained general acceptance among American or British cactologists, but there seems to be some support for subdividing Pereskia into two or more groups or even subgenera (which is at least recognition that the genus is heterogeneous).

Then there is the problem of spines. All Pereskia species have spines, some as long and vicious as almost any in the rest of the family. Now we tend to take cactus spines for granted and along with the presence of areoles, pretty well defining Cactaceae, except for a few (presumably "advanced") genera in which they have become reduced in size or have disappeared completely. Did the earliest cacti have spines? The answer depends on whether we can identify the ancestral group for cacti, or at least its sister (most closely related) group. Gibson and Nobel (1986: 243-251) discussed this problem in considerable detail and concluded Cactaceae are most closely related to Portulacaceae, Didiereaceae and Basellaceae. Portulacaceae - which includes succulents such as Anacampseros, Ceraria and Portulacaria - is a cosmopolitan temperate to tropical family, but is best represented in the Americas and South Africa. Gibson and Nobel (1986:247) point out that the family includes "two aberrant woody species", the alpine New Zealand cushion plant Hectorella caespitosa, and the similar Lyallia kerguelensis from Kerguelen Island in the Indian Ocean. [New Zealand botanists place these species in their own family, Hectorellaceae, but this makes little difference to Gibson and Nobel's overall idea.]

The Didiereaceae (including Decaryia and Alluaudia) are confined to semi-arid regions of Madagascar and include very spiny cactus-like shrubs and trees. Basellaceae is a small family of vines, mostly from tropical America but also occurring in Africa, southern Asia and New Guinea. All of these families have typically small flowers with a superior ovary, but only Didiereaceae have spines. the fact that the Didiereaceae are restricted to Madagascar makes it unlikely that this family is the one most closely related to cacti, in which case the spines have probably evolved independently in this group.

All this means of course is that the living species of Pereskia have departed to varying degrees from the ancestral cactus, which I assume was xerophytic (but had not evolved CAM photosynthesis), had well-developed leaves, lacked spines, possessed areoles, and had relatively small flowers with a superior ovary. [Other attributes of the hypothetical ancestral cactus are discussed in detail by Gibson and Nobel (1986: 235-238), who maintain that it had spines.] The variation in the living species of Pereskia is hardly surprising given its assumed antiquity. On present evidence it would take a brave botanist (I'm neither) to deny that Pereskia (or at least a very similar genus) is ancestral to other cacti. One who did question Pereskia's place in cactus evolution was Croizat who had no hesitation in taking on the whole biological establishment whenever he felt like it. Unfortunately, he seems to have been influenced more by biogeographic considerations (the fact that Pereskia is much less widely distributed than Rhipsalis) than by plant morphology (Croizat 1961:758).

I also said "other things being equal", and of course things rarely are in the real world. Species (and by extension genera) are going extinct all the time on a geological scale, and it is possible that Pereskia was once at least as widespread as Rhipsalis, but has subsequently disappeared everywhere except in tropical America. There are many examples of once widespread groups of organisms that have a much more restricted geographic distribution at the present day, eg coelacanths, rhynchocephalians (tuataras), and ginkgos. During the Miocene (23-5.5 Ma) eucalypts and acacias flourished in Central Otago, only to become extinct later, probably in response to late Tertiary cooling.

An alternative possibility is that owing to its ecological requirements Pereskia has always been restricted to parts of America and never managed to spread into Africa (and other parts of Gondwana), either because suitable habitats did not exist there or because it lacked an effective means of dispersal. Hallam (1985) reviewed what was known at that time of climates during the Mesozoic. His maps (Hallam 1985: figs. 7, 8) show most of East Gondwana as "dry" during the Late Jurassic, but by the Late Cretaceous the dry areas are restricted to parts of western South America and western and northern Africa, the remainder being "seasonally wet" or "wet". Unfortunately, he does not provide a map for the Early Cretaceous, the period critical to this discussion, but it seems likely that humid areas increased in extent as seaways developed between the constituent parts of Gondwana, and East Gondwana moved into warmer climes. Pereskia may have been prevented from spreading from South America into Africa because of intervening rain forests, in contrast to Rhipsalis, which would have found ideal conditions for dispersal.

In the third (and final) part of this article I will discuss the place of Rhipsalis in cactus evolution and present my ideas on the history of the family.

Part 3