I've been giving the background to the question, "Were there Cretaceous anthropoids?" ("The problem in a nutshell", "What is an anthropoid?")
One of my sources has been the 2004 book by K. Christopher Beard, The Hunt for the Dawn Monkey: Unearthing the Origins of Monkeys, Apes, and Humans. I'm not going to do a full review of the book. It has much of value in it, especially as a description of Beard's major find, Eosimias, and the way that it revised our understanding of anthropoid origins. I'll relate the current understanding of Eosimias (or at least, mine) in the next installment.
In Beard's telling, a central issue in the pre-Eosimias scientific framework was the identification of some omomyids as early tarsiers.
By the early 1990's, most paleoprimatologists accepted tarsiers as the sister group of the anthropoids. The Tarsier-Anthropoid clade identified by this hypothesis is called the haplorhine clade. The haplorhines, as their name implies, are linked by soft tissue characters of the nose, but also some aspects of the face, eyes, bony structure of the ears, and other characters. This used to be an area of much greater controversy before molecular data gained the resolution it has today. Now, we have lots and lots of genetic material from tarsiers and other prosimians, and it is clear that tarsiers are closer to anthropoids than other prosimians.
That doesn't necessarily settle the issues with fossils that share some tarsier features. We don't have their genes, and so we have only morphology to test the hypothesis that they are in fact tarsier relatives. This is important, because if we knew that a fossil represented an ancient tarsier relative, then we could conclude that tarsiers had in fact diverged from anthropoids by that time. And if we know that the anthropoids had diverged from tarsiers, then we know that some stem anthropoid must have existed by the time of that tarsier fossil.
This is exactly what Beard had suggested of a fossil omomyid called Shoshonius, from the Eocene of Wyoming.
[Shoshonius] allows us to predict that the anthropoid lineage extends back in time at least fifty million years, even if no anthropoid fossils come close to being that old. Paleontologists refer to such missing segments of the fossil record as "ghost lineages." We are forced to admit their existence because of the shape of the evolutionary tree, along with the age and position of key fossils that adorn it. Thus, if Shoshonius lies on the same major branch of the family tree that includes tarsiers, and if anthropoids lie on a separate but adjacent branch, the continuity of evolutionary descent requires us to deduce that the anthropoid lineage is at least as old as Shoshonius. To argue otherwise is to deny the geometry of the evolutionary tree (for example, by claiming that Shoshonius is in the wrong place) or to advocate the spontaneous generation of anthropoids (Beard 2004:165).
I like the straightforward starkness of Beard's description. Deny that anthropoids are 50 million years old, and you may as well believe they originated by spontaneous generation! I wouldn't go so far; the traits in question may have evolved convergently in some omomyids and in tarsiers -- in other words, it might not be such a stretch to think that Shoshonius is in the "wrong place". But you get the idea. Once you find a member of a crown group in the fossil record, the crown ancestor must be older than that fossil. If you don't have fossils of the other lineages in the crown group, that means these are all ghost lineages.
One might object that ghost lineages are unparsimonious. It would seem like a cladogram that necessitates long ghost lineages should pay some kind of penalty compared to alternatives that don't require organisms to persist for millions of years without leaving any fossil trace.
But the unpleasant truth is that for every well-documented series of fossils that shows a lineage evolving over millions of years, there are many others for which the fossils are few and far between. Paleontologists have devoted a disproportionate effort finding and describing primate fossil remains but there are still geographic areas in which we know that primates must have existed for millions of years, for which we have no fossil remains. This passage from a paper by Robert Martin, Christophe Soligo and Simon Tavaré describes the gaps in the fossil record of lemurs and callitrichids, which together account for a substantial fraction of primate diversity today.
As there is no convincing fossil evidence for the LCA of euprimates, some increment must obviously be added to the age of the earliest known undoubted fossil representative. With a relatively poor fossil record, the required increment is likely to be large [Martin, 1986, 1990]. The earliest known fossil euprimates are dated at about 55 Ma, but there are undeniable major gaps in the record. Madagascar lemurs (including the recently extinct subfossil species) account for around a quarter of modern primate species, yet not a single fossil relative has so far been found on the island. With the sole exception of Oligocene Bugtilemur in Pakistan, tentatively linked to modern dwarf lemurs in the family Cheirogaleidae [Marivaux et al., 2001; but see Seiffert, 2007], the fossil record of Madagascar lemurs remains totally undocumented. If Bugtilemur is a cheirogaleid, there is a ghost lineage of at least 30 Ma between this fossil and modern cheirogaleids on Madagascar, and older ghost lineages leading to the other extant families of lemurs. If Bugtilemur is not a cheirogaleid but simply an early strepsirrhine, all modern lemurs have a ghost lineage of at least 37 Ma, as fossil members of their sister group (lorisiforms) are now known to date back that far [Martin, 2003; Seiffert et al., 2003, 2005]. Another example of major gaps in the record is provided by the New World monkeys. No convincing relative of marmosets and tamarins (Callitrichidae) has yet been reported, although almost 34% of New World monkeys and more than 11% of extant primates are callitrichids. Undoubted fossil relatives of cebid monkeys indicate that the ghost lineage leading to modern callitrichids extends over at least 20 Ma [McFadden, 1990; Flynn et al., 1995]. It is also noteworthy that there is a gap of several million years in the Oligocene epoch in the primate fossil record as a whole, between the early Oligocene of the Fayum in Egypt and Taqah in Oman, and the late Oligocene deposits of Lothidok, Kenya, and Salla, Bolivia (fig. 1).
Were there no lemurs on Madagascar until 20,000 years ago? Were there no primates in the world during the middle Oligocene? They just had the bad luck of not being fossilized in locations where we currently have a record.
We will eventually find fossil evidence of some of these lineages. Surely we'll find more middle Oligocene sites with primates. Anthropoids must have been diverse by that time, but it seems that much of their Old World diversity became extinct, with the crown catarrhines diverging only by the Late Oligocene. Only the survival of the platyrrhines in the New World, and the Late Oligocene/Early Miocene persistence of groups like propliopithecids inform us about the earlier radiation.
The point is, there are gaps that we know. Is it such a stretch to think that there may be gaps we don't know about? If the tarsiers are found to go back very early, then some stem anthropoid must have existed also.
Next: The (old) new story of anthropoid origins.
Martin RD, Soligo C, Tavaré S. 2007. Primate origins: implications of a Cretaceous ancestry. Folia Primatol 78:277-296. doi:10.1159/000105145