Dinosaurs that make you go “Hmmm”…
MM has found the extreme instability of cladistic placements for something called Chilesaurus, intriguing. He seems to be thinking: if its place in the tree is so all over the place, how can you trust the method?
As scientists, the next step is to investigate the nature of the phenomenon: the phenomenon concerned is the misbehaviour of family trees under evolutionary parallelism, reversal and convergence. We will need to do a number of things:
– We need to identify and understand the causes of the confusion in such cases – in general terms;
– We need to find what is causing the confusion in particular cases, and presumably simultaneously, to find the true tree.
MM has mentioned how some characters don’t reverse often. This is a first step; it will be necessary to identify which characters are more trustworthy. Other aspects of particular characters also need to be identified and made use of.
Getting to know the personalities of our characters can be done through at least two approaches:
– Spotting patterns amongst the characters, perhaps just in the table before the first cladogram is produced, and also once the first stab has been made;
– Understanding the real-life meaning of the characters, giving us further insights: for example, not just that certain characters behave together as revealed by the simple statistical considerations of the previous point, but why they do.
Another step, this time psychological, is to rise to a plane above the influence of your “friends”. If you don’t you won’t be able to think straight. MM has already crossed the rubicon by considering the likelihood of reversals.
MM took steps at one point to reduce analysis time – i.e. the time the program takes to run. My investigations suggest to me that a heuristic that assumes the sequence of species around the leaves was fixed, having been found by selecting the first at random and then each time selecting the most similar from what was left, can be very fast. I’ve never found out how fast the other cladogenesis programs are, but with a pre-selected sequence of species around the leaves, you can run a tree of 53 species and 400 characters in a couple of minutes. The tree MM shows in his posting is quite similar in size, slightly over 53 species and slightly under 400 characters. The complaint that assuming a fixed sequence of species around the leaves may sometimes prevent the most parsimonious solution from being found is I expect to some extent true (true for any heuristic method of course) but:
– Since each node is free to swivel, the actual sequence effectively automatically adjusts itself as the node structure is generated, at least at the levels nearest the leaves.
– Even a fixed sequence method does actually find a pretty true structure if there is not to much parallelism, reversal and convergence. However if there is, then not even a full search for the most parsimonious solution can be trusted to find the right tree! Any parsimony method will require pre-processing to understand the characters and treat them differently, before it can be trusted. And of course blindly treating all characters the same is only advised by those incompetent in the information science disciplines.
It’s now overdue for someone to do this pivotal work. I’ve decided not to start any new computer programs (by which I really mean projects – I’ve got a few years left I hope!) but MM could. He would be able to get up to speed in about three years; I’ve outlined what I think to be an effective approach at the end of chapter five of the book.
Having looked at Chilesaurus, whose name (from which the place of origin of the fossil could be deduced) was all I knew, and assuming by the look of it it was pre-Cretaceous, I’d say its long back compared to chest-depth suggests closeness to the Coelophysis regions of the dinosaur family tree, and the long tail chevrons look primitive too. That trunk characteristic I’d assign a very high unreversibility. The quite strong albeit short limbs with significant claws, suggests to me basic climbing tendencies, and the reversed pubis suggests herbivory, strongly agreeing with the teeth and jaws. [Some time later…] Having checked it out, I’m happy with that, especially size-wise.