Messing with history

So, can we know our ancestors? This is the critical question. The answer is, not really. That is to say (and there are always qualifications in this sort of discussion - you need to pay attention to them), you can know what ancestors must have been like, but even if you find a fossil that exactly matches the predictions of the phylogeny, you do not thereby know you have found the ancestor.

Consider a diagram like Darwin's only figure in The Origin of Species:



[While Darwin had small bushes of unsuccessful variant lineages at each stratigraphic level, I have added them continuously.]

Now if an ancestor (for any species, not just humans) existed at point A, but the fossil exists from a species, ultimately unsuccessful as most must be, at point B (blue lineage), it will look like the species that the phylogeny based on the modern taxa C, D and E (red lineages) predicts, but will not be the actual ancestral taxon. However, the classification of these taxa in evolutionary terms will be the same. [The coloured bars represent bedding planes in which fossils have been laid down.]

In case you think this is a strained example, even cases where the speciation has been relatively recent, such as the case of Darwin's finches in the Galapágos Islands, and the likely ancestor is still around, we cannot narrow it down as to which the ancestor must be, at least not easily. Add to this the fact that the ancestor may be extinct and unrecorded, and you can see why pattern cladists are suspicious of ancestors, and why they do not think that a cladogram immediately equals a historical tree.

Are we thus prohibited from knowing the past? Well, it all depends on what you count as "knowing". If our cones of uncertainty restrict the states of the past and the future enough, then we may be able to rule out other putative ancestors and leave only one discovered case. This is, of course, insufficient to show that it is the ancestor. But if we have good reason to think that a species of that kind will have been very likely to be fossilised, then we have increasing certainty.

A concrete case is that of human ancestors immediately prior to our present species. It cannot be argued that humans were a different species from Homo erectus on biological grounds, as morphology is an unreliable guide to species, but we can say that erectus was a very successful and wide-ranging species, so if there were a species that happened to be our actual ancestors in the bedding planes from which we find erectus, it is likely to be erectus. Nothing is certain, but, hey, that's science.

So it all resolves down to how much certainty we actually need before we can say we know a fact in science. It is widely joked about that undergraduates can doubt anything until the alcohol runs out, but in science, doubt has to be reasonable even more so than in law.

This ties in with the nature of scientific explanation. If we can show that an observed species in the fossil record is consistent with our understanding of speciation, ecology, and so forth and the state of species now, are we entitled to say we have explained the present species as descendants of that species? There is no simple answer to this. It boils down to what knowledge is being adduced as background to the inference of the likelihood. If all knowledge is regarded as provisional in science, can we ever say we know the likelihoods of this species then being ancestral to those species now?

There are no rules that I think apply in each and every case. Science is about provisional knowledge, and so what is acceptable inference based on background knowledge will depend largely on the questions being asked. If the question is "Is it likely that erectus is the ancestor of sapiens? the answer depends on how well we think we have described biological realities under the taxon name Homo erectus - if it covers several biological species, then the answer can be "yes" and "no" at the same time; if we think that erectus is in the position of taxon B, then we are adducing more background information to bias our likelihood estimates. If the taxon name covers both A and B, the question is ill-formed, and so on.

So we can see that to ignore or dismiss the pattern cladist position outright is to overlook a good many real epistemological issues. But to reject process cladism on the austere grounds of some idealised pattern cladist purity of inference is equally mistaken.