|Feature Article - September 2016|
|by Do-While Jones|
Cladistics is a useful tool, but it depends on subjective judgment.
This month’s newsletter is a continuation of last month’s discussion, in which a professor of systematics challenged our assertion that the biological classification system is arbitrary. We should again mention that just because something is arbitrary, it isn’t necessarily bad. In the United States it was arbitrarily decided that Thanksgiving Day should be in November. In Canada, Thanksgiving is in October. There is no inherently “right” or “wrong” in either decision. Sometimes decisions just have to be made arbitrarily.
It is good to eliminate human bias, so scientists like to use tools that make decisions according to fixed rules, not subjective feelings. Cladistics is an objective tool scientists use to determine similarity—but it isn’t really as free of subjectivity as some people might believe.
Cladistics is important to evolutionists because they believe that biological similarity is the result of common ancestry. The more closely related two creatures are, the more similar they will be. So, they feel they can recreate the evolutionary tree by organizing creatures according to similarity.
We reject the premise that evolution can be inferred from similarity. The validity of inference in general is addressed in this month’s Email column, so we won’t address it here.
This article addresses the false notion that cladistics is an unbiased way to determine similarity. Yes, cladistics is an objective way to measure similarity, but it isn’t an unbiased way.
If cladistics is a valid tool for determining similarity, it should not be limited to biological creatures. It should be able to be used to determine the similarity of any set of objects. So, let’s try to use it to determine the similarity of electric guitars.
There are many different variations of electric guitars. Some have one, two, or three magnetic pickups (microphones). The length of the strings from the bridge to the nut can be 25.5 inches, 24.75 inches, 22 inches, or some other less common length. The guitar body can be solid, semi-solid, or hollow. The frets can be narrow or jumbo, or some intermediate size. There are many more guitar attributes that could be listed.
There are so many different kinds of guitars, with so many different attributes, that a matrix of all guitars (with a column for each model, and a row for each attribute) would look a lot like this Guitar Center comparison chart of 20 product specifications of four guitars, but with many more rows, and many more columns.
So, imagine you have a huge matrix with 1,000 columns of guitars and 50 rows of attributes. How would a computer determine which guitars are the most similar? It would compare the attributes in each column with the attributes in every other column, looking for the largest number of exact matches. It might wind up with an 857-way tie of guitars with 42 identical attributes. How would it break the tie? You would have to tell the computer which attributes are more important. Most people would think that a difference in color is not nearly as important as the number of pickups, or the length of the neck. So color should be given less weight than the number of pickups or neck length. That’s just an opinion.
Hopefully, it has already occurred to you that you will get different answers depending upon how much weight you give to each attribute. Last month we quoted an explanation of cladistics that included this sentence.
Different datasets and different methods, not to mention violations of the mentioned assumptions, often result in different cladograms. 1
What that means is that we will get different cladograms (that is, relationship diagrams) if we use a different set of 1000 guitars, or a different set of 50 attributes for the original set of 1000 guitars, or give different weights to each attribute. The next sentence in that quote was,
Only scientific investigation can show which is more likely to be correct. 2
That means, by tweaking the weights of each attribute, and/or eliminating guitars and/or attributes, you can eventually come up with a guitar cladogram that “makes sense” from a “scientific” perspective. In other words, you can adjust the program again and again until it finally confirms your prejudice. Then you have firm, scientific proof that you were right all along.
Biologists create huge databases of attributes of living things. There are lots of attributes such as skin covering (fur, feathers, scales, none); type of eye (simple, compound, etc.); number of digits (fingers and toes); etc. Microbiologists like to use databases of genes and proteins and such things instead. Regardless, a computer program compares the columns of creatures with the rows of attributes and computes similarities which depend upon the number of attributes compared and the weights given to each attribute. The only way to tell if the results are “right” is to compare the computer’s answer with the reasonably expected result. If the answer is reasonably close to the prejudged outcome consistent with the scientist’s bias, it is assumed to be right. Otherwise, the parameters are adjusted and the program is run again, hoping for a better outcome.
The cladograms produced by a cladistics program are structured in a particular way. Every branch is a split from an unknown ancestor. Therefore, it can’t possibly create a correct family tree if there are no unknown ancestors. In a previous article we showed that this is why it could not work for the British Royal Family tree. 3 If a computer program doesn’t give the correct answer when you know what the correct answer is, why would you trust it to tell you the correct answer when you don’t know what it is? The answer to that rhetorical question is that you want to believe it is the correct answer.
A cladistics computer can tell you which things are most similar if you tell it which characteristics are most important. Some scientists might think that different things are more important than other scientists might think. That’s why you see so many different, contradictory cladograms in peer-reviewed articles. A significant portion of these technical papers is the justification for why they selected certain parameters, and how much weight they gave to each one.
If we used a computer program to make a cladogram of guitars, its plausibility would depend how much weight we gave to color and how much weight we gave to neck length. But given a cladogram of guitar similarity that seems reasonable, that would not prove that the two very similar guitars are variants of an unknown previous model that blended the characteristics of both guitars. Yes, maybe Fender did start with a sunburst Stratocaster, and then decided to make an all yellow Stratocaster and an all black Stratocaster later; but you can’t know that for sure just from the existence of those guitars. A cladogram can’t tell you that. You need factory production documents to know for sure.
We used the fictional guitar taxonomy example because it is easy to understand. It is easy to find real instances of contradictory examples in the professional literature; but most of them are too technical for the general public to understand (especially the ones based on genetics). We were able to find this real example of two contradictory cladograms published in the journal Nature that isn’t too hard to follow:
FIGURE 2. Continental area cladograms for the Indian Ocean region based on chameleon phylogeny and geological break-up models. 4
In a nutshell (and that is a particularly accurate description) the article tried to infer a geological timeline for the breakup of Gondwana using the inferred evolutionary history of chameleons. It is hard to summarize their belief with a straight face, but we will try.
Evolutionists believe that once upon a time, many millions of years ago, all the continents formed one supercontinent called Gondwana. But then, slowly over time, mysterious (but apparently very powerful) convection currents deep within the earth pulled Gondwana apart, creating the continents we see today.
They also believe there were lizards on Gondwana which evolved into the chameleons of today. The pressing question (in their minds) is, “Did these chameleons evolve before or after the continents separated?”
Chameleons are one of the few extant terrestrial vertebrates thought to have biogeographic patterns that are congruent with the Gondwanan break-up of Madagascar and Africa. Here we show, using molecular and morphological evidence for 52 chameleon taxa, support for a phylogeny and area cladogram that does not fit a simple vicariant history. Oceanic dispersal—not Gondwanan break-up—facilitated species radiation, and the most parsimonious biogeographic hypothesis supports a Madagascan origin for chameleons, with multiple 'out-of-Madagascar' dispersal events to Africa, the Seychelles, the Comoros archipelago, and possibly Reunion Island. … Previous cladistic studies, using 11–24 morphological characters, supported conflicting biogeographic hypotheses that are partly congruent with Gondwanan break-up, , or that suggest a post-Gondwanan, Madagascan origin for chameleons. 5
We don’t care about their results. All we care about is how they got those results. They said,
For phylogenetic analysis, a total of 85 morphological and behavioural characters, and 972 bases of mtDNA (236 3' terminal codons of ND4, tRNASer, tRNAHis and tRNALeu) were used. … For the list of morphological and behavioural characters see Supplementary Information. 6
We don’t have space to list all the characteristics they compared in the Supplementary Information—but here are characteristics 36, 37, and 39 to give you an example of what they were.
In characteristic 39, they divided chameleons into three classes based on rib number. What’s the justification for that? Why not two classes (14 or more, 13 or less)? Why not a different class for every rib number? It probably would have made a difference in the shape of the cladogram.
The authors apparently think we should believe their new cladogram because they used more characteristics. But what if the additional characteristics are less important than the original set? What if the next analysis uses even more characteristics and confirms the original theory?
How can you have any confidence in a method that keeps producing different results depending upon subjective choices you make in what to measure and how to measure it?
|Quick links to|
|Science Against Evolution
|Back issues of
of the Month
3 Disclosure, October 2015, “Caffeine and Insects”
5 Nature, 14 February 2002, Raxworthy, et al., “Chameleon radiation by oceanic dispersal”, pp. 784-787, http://www.nature.com/nature/journal/v415/n6873/full/415784a.html