IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

[Annelida] Questions about total evidence

Kirk Fitzhugh kfitzhug at nhm.org
Thu Oct 23 17:06:26 EST 2003

Dear Ken Halanych,

Thanks much for your questions. Herein my answers:

1. Where's a "total evidence paper?"

Take a look at Cladistics. Just kidding. More to the point of your question, 
however, is that the phrase "total evidence analysis" or "combined analysis" 
are misnomers and are irrelevant to the notion of the requirement of total 
evidence. As I point out in the slide show, the requirement is applied 
relative to one's responsible action of judging what evidence in need of being 
considered (or explained) has relevance to the consideration (or explanation) 
of other evidence. It is the matter of relevance that dictates what the 
"total" evidence is that must be considered. For any logically sound, non-
deductive inference, there is no need to say one has included the "total" 
evidence since any logically sound inference would by definition include all 
relevant evidence.  

2. "'Moleculoids' (?) often use object criteria to filter the noise from their 
data. Similarly morphological (i.e., non-DNA or Protein) data sets are also 
filtered (either intentionally or unintentionally) by the researcher before 
analysis. That is they do not code features for which they strongly suspect 
rampant homoplasy might exist. This usually is a good thing because it cuts 
out information that does not bear on the issue at hand."  

There are two issues here which must be clearly separated. First, the act of 
"filtering" one's observations is just another way of saying that one makes 
judgements regarding evidential relevance. For instance, my having blue eyes 
would not be taken into consideration in any inference of phylogenetic 
relationships among homonids. The explanation of such a feature is irrelevant 
to the shared similarities in need of explanation among homonid species (we 
can explain my blue eyes at a more exclusive explanatory level). One might 
also give reasons beyond relevance, which pertain to limited observations, 
dubious interpretations of perceptions, etc. The key point, however, is that 
if relevance can be established, then to deny some set of evidence to be 
considered in the act of an inference is to deny the ability to rationally 
believe the conclusion from just the evidence considered. I provide several 
examples of this in the slide show. Thus, the phrase "total evidence" does not 
equate with "throw in every sort of observation." Since phylogenetic 
hypotheses are distinct answers to implicit or explicit causal questions, it 
is the nature of the questions which decide relevance.  

What is more egregious is the act of comparing phylogenetic hypotheses derived 
from different data sets, or the simple mapping of some data onto an existing 
tree. As I point out in the slide show, these comparisons are entirely 
meaningless, and to map characters is to imply that one has provided an 
explanation when in fact there is no basis for that explanation.  

Second, with regard to the notion of "rampant homoplasy," this is one of those 
fascinating conundrums of reasoning which seems to be unique to systematists. 
Given that one believes what they perceive, at least to the point of applying 
the same name to a set of properties (e.g., "There is an A at position 515 in 
all members of species X and Y"), yet have such remarkable ad hoc inferential 
powers to conclude that what they do perceive are not the same properties, 
then one can only resort to one option: they obviously have accounted for 
their observations of shared similarities as due to distinctly separate 
causes. As it is then the case that those observations have been separately 
explained, they then are irrelevant to the explanation of other observations 
which might be accounted for by some common cause. I have no problem with such 
an exercise, but, from an epistemological standpoint, it seems rather 
peculiar. It seems peculiar to accept that all members of X and Y have the 
same property, yet to judge that this is a miraculous coincidence of 
convergence within their respective lineages. Plus, one must systematically 
cull out all these separate effects worthy of distinct explanations from those 
effects which can be explained by some more inclusive common cause, e.g., 
common ancestry.  

To regard observations as the results of "rampant homoplasy" is to misapply 
the term homoplasy. See my slide show on the coining of homoplasy and its 
relation to homology. Since homoplasy (sensu Lankester 1870) is defined as an 
ad hoc hypothesis accounting for shared similarities by events other than 
common ancestry, the only way to apply the term is in the context of an 
inference which has applied common ancestry as the cause - in other words, the 
only way to infer a hypothesis of homoplasy is when all relevant evidence is 

3. "Also do you have any theoretically problems with how to perform such an 
analysis?" I'm not sure I understand your question. If you are asking if I 
have a problem with the notion of inferring a phylogenetic hypothesis where 
the premises of the inference includes nucleotide, histological, behavioral, 
morphological, etc., observations, then my answer is "no." As I note in the 
slide show, the explanatory context for all these observations is the same, 
given that one assumes the explanations of these observations are relevant to 
one another.  

4. " I think we would both agree that a distance approach is not a good 
thing." Quite true. Distance approaches are useless contrivances which have no 
basis in causal explanation. A branching diagram is useless if, as a purported 
hypothesis, it is not based on explicable premises and has no empirical 

5. "Although parsimony is known to work well for morphological data, it has 
been clearly shown to be problematic for some types of molecular data. 
(Unfortunately parsimony is known to have more sever problems with 18S, the 
main gene currently being used for annelids, than other methods.)" How does 
one "know" when they have got the "wrong" answer? How does one judge whether a 
causal explanation is "right" or "wrong?" What theory of truth is being 
applied wherein the basis for judging the merits of a hypothesis only rely on 
the observations for which the hypothesis was inferred to explain? (hint: 
there is no theory of truth [among the six or so out there] on which this 
fallacy can depend). It seems that to keep this myth of the "true tree" going 
one would have to blatantly ignore the requirement of total evidence, and as a 
result, compromise causal explanation. I know of no field of science (other 
than systematics) or any logician where this would be regarded as acceptable.  

6. "This leaves likelihood methods which work better for molecular data, but 
we are no where near developing evolutionary models for morphological. So it 
seems to me that we are currently lacking the methods to do satisfactory 
combined (molecular/nonmolecular) analysis." The only way so-called likelihood 
methods "work better" is because they can give results which meet one's 
preconceived idea of what they want to think happened in the past. In simply 
proceeding from observations to causal hypothesis there are no yardsticks upon 
which to measure the veracity of any hypothesis. More critically, however, 
"likelihood" methods conflate phylogenetic and tokogenetic explanatory 
contexts. A single inferential act cannot deal with such a situation.  

There are no "models" (the proper phrase should be "causal theories") for 
dealing with the explanation of morphological properties? Take a look at the 
slide show: DESCENT WITH MODIFICATION. To apply a "model" which relies on 
branch lengths is to reduce the causal realm down to a within-lineage (i.e., 
tokogenetic) level, rather than a phylogenetic level. Therein the matter of 
relevance comes into play and the explanations of similarities as the results 
of intraspecific events would not be included in the explanations of 
similarities regarded as explicable by common ancestry. So, proper application 
of the requirement of total evidence precludes "likelihood" approaches.  

So, it is not that we lack methods for combining data. This is like saying one 
can not make inferences. The problem is that systematists have not identified 
the ramifications of their desire to explain observations at tokogenetic and 
phylogenetic levels, and the fact that these are entirely different. The 
problem with "likelihood" methods is that they attempt to conflate these 

7. "Lastly --just who is "requiring" total evidence? Is this another 
unilateral decision by George W. Bush?"  

Hmm, I thought I spelled this out in the slide show. Below are some good 
references - I don't know if any of these individuals were/are Republicans.  

Thanks again for the comments, and hope my responses have helped.


Barker, S.F.  (1957).  Induction and hypothesis.  New York: Cornell 
University Press.

Bunge, M.  (1998).  Philosophy of science, volume 2, from explanation to 
justification. New Jersey: Transaction Publishers.

Carnap, R.  (1950).  Logical foundations of probability.  Chicago: 
University of Chicago Press.

Fetzer, J.H.  (1993).  Philosophy of science.  New York: Paragon House.

Fetzer, J.H. & Almeder, R.F.  (1993).  Glossary of epistemology / 
philosophy of science.  New York: Paragon House.

Hacking, I.  (2001).  An introduction to probability and inductive 
logic.  New York: Cambridge University Press.

Hanson, N.R.  (1958).  Patterns of discovery: An inquiry into the 
conceptual foundations of science.  New York: Cambridge University Press.

Harman, G.  (1965).  The inference to the best explanation.  Philosophical
Review, 74, 88-95.

Harré, R.  (1970).  The principles of scientific thinking.  Illinois: The
University of Chicago Press.

Hempel, C.G.  (1965).  Aspects of scientific explanation and other essays in
the philosophy of science.  New York: The Free Press.

Josephson, J.R. & Josephson, S.G.  (1994).  Abductive inference: 
Computation, philosophy, technology.  New York: Cambridge University Press.

Lipton, P.  (1993).  Inference to the best explanation.  New York: Routledge.

Murphey, M.G.  (1994).  Philosophical foundations of historical 
knowledge.  Albany: State University of New York Press.

Rescher, N.  (1970).  Scientific explanation.  New York: The Free Press.

Salmon, M.H.  1995.  Introduction to logic and critical thinking.  New 
York: Harcourt Brace College Publishers.

Salmon, W.C.  (1967).  The foundations of scientific 
inference.  Pittsburgh, Pennsylvania: University of Pittsburgh Press.

Salmon, W.C.  (1984a).  Logic.  Englewood Cliffs, New Jersey: 
Prentice-Hall, Inc.

Salmon, W.C.  (1984b).  Scientific explanation and the causal structure of the
world.  New Jersey: Princeton University Press.

Salmon, W.C.  (1989).  Four decades of scientific explanation.  In P. 
Kitcher & W.C. Salmon (Eds) Volume XIII, scientific explanation.  Minnesota
studies in the philosophy of science (pp. 3-219).  University of Minnesota
Press, Minnesota. Salmon, W.C.  (1998).  Causality and explanation.  New York:
Oxford University Press.

Sober, E.  (1975).  Simplicity.  New York: Oxford University Press.

Thagard, P.  (1988).  Computational philosophy of science.  Cambridge, 
Massachusetts: The MIT Press.

Van Fraassen, B.C.  (1990).  The scientific image.  Oxford: Clarendon Press.

"True science is distinctively the study of useless things. For the
useful things will get studied without the aid of scientific men.
To employ these rare minds on such work is like running a steam
engine by burning diamonds."

Charles Sanders Peirce

J. Kirk Fitzhugh, Ph.D.
Associate Curator of Polychaetes
Research & Collections Branch
Los Angeles County Museum of Natural History
900 Exposition Blvd
Los Angeles CA 90007
Phone:   213-763-3233
FAX:     213-746-2999
e-mail:  kfitzhug at nhm.org

   Discuss  =  <annelida at net.bio.net> = talk to all members
   Server =  <biosci-server at net.bio.net> = un/subscribes
   Archives  = http://www.bio.net/hypermail/annelida/
   Resources = http://www.annelida.net/

More information about the Annelida mailing list

Send comments to us at biosci-help [At] net.bio.net