joe@GENETICS.WASHINGTON.EDU (Joe Felsenstein) (04/27/91)
In response to Jim Danoff-Burg's posting in "bionews" (let's move this to the more appropriate groups "population-biology" and "molecular-evolution" rather than "bionews"): > That is: traditionally, systematicians have looked for those > characters that they view as being completely (or nearly so) from the > force of natural selection. Did you mean "completely free from the force of natural selection" or "completely resulting from the force of natural selection"? That was the wrong word to forget to type in the sentence! My impression is that many systematists tried for the former, or said they did. They wanted "conservative" characters and often equated characters whose evolution was affected by natural selection with ones which could change very rapidly in response to natural selection (for example, superficial color patterns or size of the organism). Of course there are also characters affected by natural selection that change with great difficulty and only once in a while. (For example, if the conditions causing or allowing the change are infrequent). > However, in practice this is only weakly followed > at best. Most systematists that I know have openly admitted that they use > whatever character that is most helpful. They in practice are more biologically sensible than their formal prescriptions. In practice they made an intuitive assessment of how rapidly a character could be reversed by local and temporary selection. If so, the character was likely to show parallelism and reversal and was best dropped. But a character like having a placenta was retained even though it was most probably caused by natural selection, because the conditions for getting started on having one are most likely unusual and hence rare. Taking a statistical view this makes sense as the rarer change is the more heavily the character should be weighted if you are (for instance) doing a parsimony analysis. You will find a fairly detailed discussion of this in my 1981 paper in Biological Journal of the Linnean Society, on the logic of character weighting. > a) if the characterization of the traditional approach to > character selection appropriate? See above. In short, sort of. > b) if so: why is it thought that the separation of the causes > from the process necessary? I don't think traditional systematists were really trying for that, as mentioned above, but recent phylogenetic systematists have been explicit about trying for it. As they have argued (I believe incorrectly) that use of the unweighted Wagner parsimony method guaranteed this. They have wanted to escape from having to know the mechanism of evolution which they consider as unknowable. By holding that the only valid method of phylogenetic analysis does not depend on knowing evolutionary mechanisms they have held they have esacped from knowing this, and since they tend to view the same data sets as the only possible source of knowledge about evolutionary mechanisms, they have as a result argued that those mechanisms are intrinsically unkknowable. The extreme of this view are the pattern cladists, but other members of the phylogenetic systematics school have also argued that processes are unknowable and unnecessary to know. More recently (in the last few years) there has been increased interest in the ranks of phylogenetic systematists in character weighting, and this is re-opening the question. > c) lastly, how does this philosophically affect someone who > desires to include behavior into their phylogeny? > -does it mean that we should look for behav. patterns > that presumably evolution doesn't act upon? > -if so, how can we characterize those patterns that > are immune from evolution's action? I think it should be clear from the above that I agree more with the traditional systematists except that I feel that their practice (if not always their prescriptions) can be illuminated by a statistical approach. They would not have absolutely banned behavioral characters. If they are something encoded in the genome we could use them especially if we had some way of assessing or guessing their relative probabilities of change. For instance, most systematists have felt that the fact that crocodiles sit on nests and sing is a valid indication of their having common ancestors with warblers (and other birds). The contemporary phylogenetic systematics school is only now groping its way toward agreement with this view, but if Danoff-Burg adopts it he may still encounter resistance from the folks on the platform at Hennig Society meetings even though many in the audience will quietly agree with it. As he will realize, my own view, although gaining among that audience, is still regarded as heretical (and worse). Nevertheless evolutionary mechanisms and statistical inference frameworks are currently sneaking in the back door of the house of phylogenetic systematics. It would be nice to have a posting here from a more orthodox defender of the views of phylogenetic systematists, as that is the dominant school in contemporary systematics and the view I am presenting is not theirs. ----- Joe Felsenstein, Dept. of Genetics, Univ. of Washington, Seattle, WA 98195 Internet: joe@genetics.washington.edu (IP No. 128.95.12.41) Bitnet/EARN: felsenst@uwavm UUCP: ... uw-beaver!evolution.genetics!joe