At some point, the hypothesis that we haven’t found the ancestors becomes incompatible with reasonable hypotheses about the origins of branches by vicariance, dispersal, ecological specialization, or the like. Or reasonable assessments about sampling density and the likelihood of missing the true ancestors at every locality.

I don’t disagree that an “upper bound” is subject to parsimony and credulity. I just don’t think the situation is any better for “lower bounds”. The comments above lay out a 3-million-year fudge factor in the lower bound on the human-chimpanzee divergence, a period during which we have a reasonably good fossil record.

My idea of an upper bound for the human-chimpanzee divergence — based on fossils alone — is between 10 and 13 million years. Also a three-million-year span.

I think we can pin down the orang-human divergence with narrower bounds, because there are biogeographic constraints getting apes into Asia and good phylogenetic signs that the initial dispersal of great apes into Eurasia was accomplished by stem great apes. But the upper bound is only as good as the hypothesis.

I also see the point about any given fossil being ancestral to humans. A priori I would find it extremely unlikely that any fossil we dig up is ancestral to any living human – based just on how unlikely fossilization is in the first place – but I don’t have any good intuition about how likely it is that it was part of a species that eventually led to humans.

For what it’s worth, we do not know if ANY of the known fossil hominins are directly ancestral to living humans, just that (based on morphological analysis) that they are more closely related to humans than chimpanzees – and that’s all we need for the calibration point. That is, any member of the descendant clade after the split; they don’t have to be directly ancestral to living humans to provide a valid lower bound for the human-chimpanzee split.

Many paleontologists would argue that it is somewhere between enormously difficult to effectively impossible to identify whether a particular fossil is directly ancestral to a later lineage or is a closely related side branch. A strict cladist would argue that one should NEVER take the position that the actual ancestor is known.

In addition, the odds of any given fossil group being directly ancestral to a later lineage depends on one’s understanding of the bushiness of the larger group in question. So if you believe the evidence suggests that there have been relatively few hominoid lineages alive at any given time, that increases the odds that any individual fossil belongs to a group that is directly ancestral to later lineages. However, if you believe the evidence suggests that there have often been quite a number of hominoid lineages alive at any given time, then that decreases the likelihood that any particular fossil belongs to a group directly ancestral to later forms.

The latter is something I have a bit of a problem seeing, but then I do not know any paleontology so I will take your word for that. I guess it boils down to what you say about the actual ancestral species vs side branches.

I guess if we find a human-looking (for whatever that means) fossil we don’t know if it is ancestral to humans or a side branch either, but if it is sufficiently close to human we would still expect it to have a shared branch with actual humans not shared with chimps.

Am I getting this right?

It makes sense in my head, so I hope it is not too far off :)

Because I think we’re all reasonably conversant with the broad outlines of the fossil and genetic evidence, I’m sticking with the human-chimpanzee divergence for the example here.

So, first, the lower bound. Yes, 100 Ka is certainly safe, but there are huge numbers of fossils almost universally assigned to the hominin clade found in deposits earlier than 100 Ka. As one pushes that date earlier, the number of near-universally accepted hominin fossils decreases until one gets to a point where the evidence no longer supports an earlier lower bound. You and I may differ as to where that point is, but we are fundamentally basing that lower bound on the first appearance of a fossil for which we accept the hypothesis that it is more closely related to humans than chimpanzees. For you that might be Ardipithecus at ~5 Ma, for me that might be Orrorin at ~6 Ma, and for someone else that might be Sahelanthropus at ~7 Ma; but the fundamental reasoning is the same in all cases.

There is no equivalent for the upper bound. What could it possibly be? The first appearance of a fossil on the gorilla lineage? (I’m not really up on my gorilla fossils, so this will all be “what if?”).
* What if the first near-universally accepted gorilla lineage fossil appears at 2 Ma? This obviously provides no information on the upper bound of the human-chimpanzee divergence – there’s a long absence of fossil information for the gorilla lineage.
* What if the first near-universally accepted gorilla lineage fossil appears at 10 Ma? If you accept this hypothesis, this still provides no significant information on the upper bound. Is that gorilla-lineage fossil from an animal that died and was fossilized IMMEDIATELY after the gorilla-(human,chimpanzee) split or from an animal that lived 8 million years after that split? (as in the first what if?)

If you had a fossil from a species from the human-chimpanzee lineage after it split from the gorilla lineage, it might seem that the date of this fossil could be useful for setting a lower bound, but often is not. If the fossil belongs to the ACTUAL ANCESTRAL SPECIES of humans and chimpanzees, then it could provide an upper bound. However, we generally cannot be certain that this is the case. It certainly could, and odds are, probably does belong to a side branch that left no living descendents (an extinct stem lineage). In that case, it provides no evidence for an upper bound. That stem lineage could have diverged from the common human-chimpanzee-stem lineage ancestor at any time.

Fossils that are on one of the descendant lineages of the split you are trying to date give information on a lower bound for that split, but fossils on separate evolutionary lineages don’t really tell you much.

Of course, the tighter the bounds, the higher the risk that they are wrong, I just don’t see why this is more of a problem for upper bounds than lower bounds?

Sorry if I’m being thick here…

For a human lineage divergence from chimpanzees, anyone with any kind of reasonable knowledge of the evidence could not (in my opinion) possibly push the lower bound of this divergence closer to the present than 4 Ma or so.

However, for an upper bound, I’d say that 20 Ma is probably absolutely safe (but completely useless). So then, how about 15? 10? 7? By the time that number gets near a value that is useful, it is no longer reasonable to use it – and there is no evidence that reasonably supports it (where reasonably = you could convince the majority of your scientific peers). What evidence would you use to support a useful upper bound for the human-chimpanzee divergence?

Why not argue the same for lower bounds? How do you know, really, that the first appearance of a species was not really a collateral? How do you know the polarity of features, really? It’s only the limits of credulity on parsimony that argue for phylogenetic relationships at all.