|Eastern barred bandicoot|
This leaves a little over twenty species unaccounted for; species that belong to neither group. All but two of these are bandicoots. These are relatively small animals, with the largest species being about the size of a hare or jackrabbit, and most of the others quite a bit smaller. They're omnivorous, eating things they can find close to the ground, like grass seeds and worms. They share with the herbivorous diprotodonts an unusual feature of the hindfeet whereby the second and third toes are fused into a single digit, albeit one that ends in two claws. Their front teeth, on the other hand, resemble those of, well... every marsupial except the diprotodonts.
In 2003, an analysis showed that bandicoots were closer to the "herbivorous" than to the "carnivorous" Australian marsupials, which is sort of what you'd expect, given the shape of their feet. Unfortunately, the following year, a different one showed a completely different pattern, with bandicoots diverging before the other two groups had parted company from each other. (And therefore, that they're equally related to both).
Over the years, there has also been some dispute as to how the various kinds of bandicoot relate to one another, and how many families they should be placed in. The current view, at least according to a 2012 genetic analysis, has three separate families for the recent species. One of them - consisting only of the pig-footed bandicoot - went extinct in the 1950s, so today we only have two: the bandicoots proper and one other species, the greater bilby, that's strange enough to get a family all to itself. (There was once also such a thing as a lesser bilby, but they, too, went extinct in the 1950s. It's not a great record).
That date is around 26 million years ago, in the late Oligocene epoch. However, we don't have any fossil bandicoots (or bilbies) that are quite that old. Until recently, the very oldest fossil bandicoot was, instead, something like a mere 5 million years old, and the oldest known bilby was about the same. Such gaps are hardly unheard of in the fossil record, but it's still quite a sizeable one, assuming the molecular dating estimate is correct.
And there is good reason to suppose that it is, since we do know of a fossil bandicoot-like animal that's 24 million years old, even though it's too primitive to place in any of the living families, and may well date to shortly before their respective origins. Still, it would be nice to have something from after the split, to confirm quite when it happened.
And now, we have just such a fossil.
The new fossil was recovered from the Riversleigh World Heritage Site in Queensland, a limestone area that is, perhaps, the most significant of all fossil sites on the continent. Exactly how old the fossil is isn't exactly clear, although it's probably around 15 million years old or so, which would place it about half-way between the primitive fossils we already had and the earliest modern ones we had until now. This would be significant, because it likely a time when "modern" bandicoots were crashing onto the scene, replacing older forms as the Australian deserts began to form in earnest, and began to diversify into the various kinds we have today.
The new fossil species, Crash bandicoot, is known only from...
No, seriously, I'm not making this up: that's what it's called. I bet the researchers had been waiting for ages to find an excuse to use that one.
So, anyway, Crash bandicoot is known only from its teeth, which makes it difficult to say too much about it. Assuming that it's teeth were broadly in proportion to the rest of it, it would have been about the size of a typical rabbit. There's also some indication that it was more closely related to the living Australian species than those in New Guinea, although, again, it's difficult to know for sure.
At the time it lived, Riversleigh was a tropical rainforest. Other fossils recovered from the same part of the site confirm this, including the sort of things you'd expect to find in forests, such as koalas, possums, and, um, tree-climbing "marsupial sheep". Not far away, in another part of Riversleigh going by the somewhat unusual name of "Rick's Sausage Site", the researchers identified another fossil relevant to the story: Liyamayi dayi, which may be the oldest known bilby.
Somewhere between the size of the living greater bilby and it's recently extinct lesser cousin, the identification of this animal as belonging to the bilby family is, perhaps, less definitive than the identification of Crash as a modern bandicoot. The fact that nothing else quite like it is known nearby, however, suggests that either it represents the earliest arrival of bilbies (or whatever it is) in the Riversleigh area, or perhaps that it had only just evolved the features that distinguish it as such.
Taken together, the fossils help to confirm the story that we had suspected based on the molecular data, and suggest that "modern" bandicoots and bilbies first appeared in tropical rainforests about 15 million years ago, before rapidly diversifying. The bandicoots stayed broadly in their original habitat, albeit often adapting to sparser woodlands as Australia dried out, while the bilbies took greater advantage of the change, moving into the arid grasslands and scrub of the Outback.
[Photo by JJ Harrison, from Wikimedia Commons. Cladogram adapted from Meredith et al, 2008.]