Taxonomy as a Rigorous Science

[Kotatsu]

I have a vague recollection that parasites can transmit genes to their hosts--I believe they found leach genes in humans, for example. Is that true? I know it's not terribly common (it's one, maybe two, genes, in a few thousand years of people being bitten, as I recall), but it's always possible that the researcher selected the exact worst gene to look at. Retroviruses are another potential vector. If so, it certainly would emphasize the necessity to at least use multiple genes in a genetic study, and would definitely demonstrate that one needs to be VERY careful when selecting genes to analyze!

I don't think I've ever heard anything like that, but fortunately I am working in one of the best parasitology labs in the world, so I'll ask around.

 

[Jodie]

I have skimmed that thread now, and will retain my previous opinion, quoted by EHocking. it is, in any case, nothing that interests me very much, until there as an actual specimen, collected under controlled conditions, so that the purported new bear species could be described properly. A few hairs are as useful and interesting as "Environmental sample #41661562" to me; it does not say anything, regardless of what it may or may not match.

But please feel free to explain how genetic data is less subjective than morphological data, given the concerns I have outlined in several posts now.

Based on your commentary it depends on what you are looking at and whether you have a type specimen. I had said previously that you can't tell by looking at DNA of an unknown and know what it looks like without looking at comparable sequences in other species, which gets back to morphology. However not all morphological traits can be traced back to the same set of genes responsible, ex- Neanderthal's red hair is caused by a different mechanism than how it occurs in HSS.

 

[Jodie]

A note on this, as well.
This story shows the importance of double-checking. If they had sequenced another worm, maybe they wouldn't have got the same contaminated sequence, and this would hopefully have caused them to do it again to settle the matter. And this would not have lead the other PhD student to go and actively look for something to connect the Xenoturbella to molluscs in his morphological samples.

For reasons such as these two (wasps and Xenoturbella; and there are of course many more), I remain very skeptical of extraordinary claims based on extremely little data. Like I said, when they have a whole specimen of this purported new bear/ancient polar bear/whatever, then I will be interested, but until then, there are just too many things that can go wrong...

That was some years ago, wouldn't the process have improved by now to reduce those kinds of errors? Either way, I hear what you are saying about the data.

 

[Kotatsu]

That was some years ago, wouldn't the process have improved by now to reduce those kinds of errors? Either way, I hear what you are saying about the data.

No, because the process that is at fault is allowing human beings to interfere between the collection of the data and the publication of the results. That has not changed noticeably in the last twenty years, and I wouldn't expect it to change noticeably in my lifetime.

 

[Jodie]

So what you are saying is we never see the contradictory evidence in the research so therefore no matter what it is, nothing has any real rigor to it.

I would think that would certainly be a subjective opinion.

 

[Kotatsu]

So what you are saying is we never see the contradictory evidence in the research so therefore no matter what it is, nothing has any real rigor to it.

No, what I am saying is simply that genetic data and morphological data is equally subjective (or objective), and that what matters is the way you treat it. Phylogenies based solely on genetic data are every bit as subjective as phylogenies based solely on morphological data. In both cases, the data has to go through the many potential problems I have outlined before (though of course genetic data have some unique problems, such as alignment, and morphological data has some other unique problems, such as homologization). Changing from the morphology of macroscopic structures to the morphology of molecular structures has not changed the way these things are done enough to remove even the majority of the subjectivity inherent in comparing things.

I would think that would certainly be a subjective opinion.

Of course. I have exactly no objective opinions.

 

[Dinwar]

I would think that would certainly be a subjective opinion.

This is one of the few cases where I'm in disagreement with Kotatsu. I believe that the notion that genetic and morphological data are equally objective/subjective is a testable hypothesis. That test has been run, and the resutls contradict your working hypothesis, Jodie.

When confronted with this evidence you evaded, slandered, and dodged rather than admitting that you were wrong. This does nothing to change the facts, however: What you have presented is a testable hypothesis, and our answers are every bit as objective as any other data supporting a hypothesis in science. Again, you cannot become some empirical that you metamorphose into subjectivism. There is nothing of opinion in this, Jodie, except from your side.

To put it another way: On our side we have two experts providing conclusions drawn from years of experience in these fields, in addition to numerous references supporting their points (meaning you're not relying on just us) and direct evidence supporting our arguments. On your side, we have the mere opinion of someone who demonstrably and admittedly doesn't know the subject areas, much less the data or processes involved. If this were any other discussion, which side do you think would be right?

 

[Jodie]

Me, because it's not opinion, and I have demonstrated repeatedly that what I'm talking about are ancient hominids, but your examples are of sea life. You simply refuse to see my point regarding the sample size of most hominid fossils. Without DNA analysis, which you will probably never have, morphology alone is only one piece of the puzzle.

The subjectivity of genetic analysis that Kotatsu is talking about is based on what his experience was nearly a decade ago, and while he may have a point about the nature of human bias, I still think DNA will give you more information than morphology if you have it. Neither can really give you an accurate picture on their own if dealing with unknowns.

 

[Akri]

Me, because it's not opinion, and I have demonstrated repeatedly that what I'm talking about are ancient hominids, but your examples are of sea life. You simply refuse to see my point regarding the sample size of most hominid fossils. Without DNA analysis, which you will probably never have, morphology alone is only one piece of the puzzle.

The reason we "refuse to see [your] point regarding the sample size of most hominid fossils" is that you've yet to prove that it matters. Just like you've yet to prove that paleontology is subjective. Just as you probably won't make any attempt to prove that DNA provides more information than morphology, or that Kotatsu's experiences are no longer relevant to how things are done today. Because simply asserting things is a lot easier than actually proving them, especially when you're asserting things about a subject you know nothing about.

 

[Kotatsu]

Me, because it's not opinion, and I have demonstrated repeatedly that what I'm talking about are ancient hominids, but your examples are of sea life.

It makes no difference what organisms you use as examples, as the theory underlying morphological and genetic phylogenetics is identical for all organisms. When I taught introductory systematics many years ago, we used screws, nails, and bolts to collect characters from, and the methods work exactly the same. Your apparent contention that the choice of example organisms should make a difference is spurious, not to mention highly unscientific.

The subjectivity of genetic analysis that Kotatsu is talking about is based on what his experience was nearly a decade ago, and while he may have a point about the nature of human bias, I still think DNA will give you more information than morphology if you have it.

Certainly it will give you more information. It will not, however, automatically give you better information. Sequencing a dozen genes is a few days work, and may give you a few thousand characters. Getting a few thousand characters through morphology will take years, so certainly there is an advantage to genetic characters.

However, as I've said above, what you are really looking at when you use genetic material is the morphology of individual molecules associated with that organism. There is no difference. Genetic data is not a separate class of information from morphological data, it is simply morphological data of smaller components of the organism. Here is the progression (roughly):

1. I can look at the overall structure (whole-organism level: number and arrangement of appendages, symmetry, coelomic structure, etc.);
2. I can look for certain diagnostic structures (overarching organ system level: absence/presence of whole organ systems, etc.);
3. I can look at the gross morphology of the structures present (organ system level: chaetotaxy, dental patterns, scutellation, skeletal structure, etc.);
4. I can look into any of these organ systems and focus on single organs (organ level: structure of renal systems, structure of eye, etc.);
5. I can focus on the more detailed structure of these organs (supercellular level: essentially, what cells are present and how to they relate to and interact with each other?);
6. I can zoom in on the individual cells (cellular level: what components exist in the cells of this organ?);
7. I can focus on organelles inside the cells (intracellular level: what proteins are expressed in these cells?);
8. I can focus on part of one of the organelles (the DNA) and study its overall structure (chromosomal level: essentially karyotypes);
9. I can look at gene order inside the chromosomes;
10. I can look at individual genes (genetic analysis).

All of these things are morphology, but focusing on different levels. There is no qualitative difference in studying the morphology of the smallest structure in the organism and studying the largest structure, the data you get is of exactly the same sort, and it is coded in exactly the same way into the programs we use to construct phylogenies. PAUP* does not inherently know what your data is coming from, as long as it is written in a form it can understand. If I construct a morphological data set, I am free to code the various character states as I want. If I want to use ACGT as my annotation instead of 0/1/2/... that is my prerogative, and the programs don't care (1).

ACGT is simply shorthand for a morphological structure. It just happens to be the case that we can normally express the morphology of a molecular in a limited, regular way. The same is not true of most other systems of morphology in organisms, as these are generally more varied. We could certainly make similar annotation systems for other groups of morphological characters, such as chaetotaxy, but they would not normally be as elegant as ACGT.

---
(1) Unless your program reacts on stop codons, in which case you can just rearrange your morphological matrix so you don't get stop codons.

 

[Dinwar]

When I taught introductory systematics many years ago, we used screws, nails, and bolts to collect characters from...

Small world--my first phylogeny was of the same thing! I used the nail as the outgroup.

I've seen phylogenies for everything from dragons to My Little Ponies (that one is complicated by the fact that what constitutes a generation of MLPs is vague at best). The techniques are identical.

The subjectivity of genetic analysis that Kotatsu is talking about is based on what his experience was nearly a decade ago,

Do you have any evidence that these techniques have improved? Remember, we're not talking diagnostic techniques--rather, we're talking about techniques used to identify new creatures.

...I still think DNA will give you more information than morphology if you have it.

Based on nothing--or at least, you've utterly failed to provide any evidence supporting your stance (remember, those new finds for Neanderthals are being examined morphologically as well as genetically, so that's a wash as far as this conversation is concerned). Secondly, more data DOES NOT mean better. Better data means better. If you mishandle the data, or you make poor selections, or the like, your tree is still going to be wrong if you have 10,000,000 datapoints per organism.

You simply refuse to see my point regarding the sample size of most hominid fossils.

Actually, no. I've specifically addressed it. I stated that SOP in paleontology is to assume that organisms have similar intraspecific diversity to other organisms in that taxa, unless evidence dictates otherwise. In other words, we don't assume that the hoofbeats are caused by zebras until we at least see stripes. Humans have experienced events that quite obviously have increased our genetic diversity well beyond what most species have, and therefore we cannot justify assuming that ANY other organism has equal diversity. Your point is that we should assume that diversity is sky-high--in violation of everything we know about how evolution operates (it's a conservative process in most instances) and in violation of everything we know about intraspecific diversity (it's usually not nearly as high as in humans). Your evidence that paleontology is wrong boils down to the fact that in light of new eivdence we revise our stance--which is the very definition of empiricism and is the foundation upon which all of science is built.

You are talking nonsense, demonstrably and irrefutably.

Without DNA analysis, which you will probably never have, morphology alone is only one piece of the puzzle.

No kidding. This is Paleo 101 stuff (actually Geo 106 in my alma matre). NO ONE except you has EVER pretended that morphology was anything but one piece of the puzzle that is paleontological analysis. Unlike you, however, paleontologists don't throw up our hands and say "Since we can't know everything, we may as well walk away." There's a LOT you can learn from morphology. For that matter, there's a lot you can learn from broken up bits of bone too shattered to be identifiable. I recently made that argument in a professional capacity to a client. We have certain data. So that's what we work with. We don't waste our time wishing for things that we don't and can't have; we work with what we've got. That's not subjective; no acknowledgement of hte facts of reality can be termed "subjective" if you wish to maintain utility of the word.

And when two pieces of the puzzle provide the same data--which genetics and morphology demonstrably do, at least in the cases that have been presented here--there's absolutely no justification for demanding we favor one method over the other. None. The only reason to favor one over the other is an a priori bias that has nothing to do with the dataset or question being asked--ie, subjectivism, the very sin you accuse scientists fo committing.

 

[Jodie]

It makes no difference what organisms you use as examples, as the theory underlying morphological and genetic phylogenetics is identical for all organisms. When I taught introductory systematics many years ago, we used screws, nails, and bolts to collect characters from, and the methods work exactly the same. Your apparent contention that the choice of example organisms should make a difference is spurious, not to mention highly unscientific.

Except when the DNA evidence is non existent for most hominids versus the example used by Dinwar. It is a legitimate point.

Certainly it will give you more information. It will not, however, automatically give you better information. Sequencing a dozen genes is a few days work, and may give you a few thousand characters. Getting a few thousand characters through morphology will take years, so certainly there is an advantage to genetic characters.

It will give you better information depending on what you look at , such as hair and eye color, race, even the potential for some diseases.

However, as I've said above, what you are really looking at when you use genetic material is the morphology of individual molecules associated with that organism. There is no difference. Genetic data is not a separate class of information from morphological data, it is simply morphological data of smaller components of the organism. Here is the progression (roughly):

1. I can look at the overall structure (whole-organism level: number and arrangement of appendages, symmetry, coelomic structure, etc.);
2. I can look for certain diagnostic structures (overarching organ system level: absence/presence of whole organ systems, etc.);
3. I can look at the gross morphology of the structures present (organ system level: chaetotaxy, dental patterns, scutellation, skeletal structure, etc.);
4. I can look into any of these organ systems and focus on single organs (organ level: structure of renal systems, structure of eye, etc.);
5. I can focus on the more detailed structure of these organs (supercellular level: essentially, what cells are present and how to they relate to and interact with each other?);
6. I can zoom in on the individual cells (cellular level: what components exist in the cells of this organ?);
7. I can focus on organelles inside the cells (intracellular level: what proteins are expressed in these cells?);
8. I can focus on part of one of the organelles (the DNA) and study its overall structure (chromosomal level: essentially karyotypes);
9. I can look at gene order inside the chromosomes;
10. I can look at individual genes (genetic analysis).

All of these things are morphology, but focusing on different levels. There is no qualitative difference in studying the morphology of the smallest structure in the organism and studying the largest structure, the data you get is of exactly the same sort, and it is coded in exactly the same way into the programs we use to construct phylogenies. PAUP* does not inherently know what your data is coming from, as long as it is written in a form it can understand. If I construct a morphological data set, I am free to code the various character states as I want. If I want to use ACGT as my annotation instead of 0/1/2/... that is my prerogative, and the programs don't care (1).

ACGT is simply shorthand for a morphological structure. It just happens to be the case that we can normally express the morphology of a molecular in a limited, regular way. The same is not true of most other systems of morphology in organisms, as these are generally more varied. We could certainly make similar annotation systems for other groups of morphological characters, such as chaetotaxy, but they would not normally be as elegant as ACGT.

---
(1) Unless your program reacts on stop codons, in which case you can just rearrange your morphological matrix so you don't get stop codons.

My understanding is that you run your sequences through programs, there are several different ones, with statistically defined algorithms to obtain reliable scores that have a 1-2.5% chance of being inaccurate for comparison/alignment. It might be similar to the same systems used for morphology but how accurate can those be if you don't have reliable data to input because it's missing? What am I missing here?

 

[Kotatsu]

Except when the DNA evidence is non existent for most hominids versus the example used by Dinwar. It is a legitimate point.

No, it is not. The methodology is the same, regardless of the data set. The organisms are entirely interchangeable, so it does not matter what organisms you use: if the methodology works for one set of organisms, it works for all of them. Similarly, if the arguments work for one set of organisms, it works for all of them, and for any set of organisms. This is easily understood once you realise that, given an sufficiently powerful (future) computer, long enough time, and a very dense taxon sampling, we could theoretically construct the complete tree of life for all known extant and extinct organisms. This tree could (theoretically) be constructed using the algorithms and phylogenetic theory we have today. The same is true for any subset of this tree, and thus for all subsets of this tree.

It will give you better information depending on what you look at , such as hair and eye color, race, even the potential for some diseases.

Eye colour is determined better by genetic data than with morphological data? Please explain.

I will agree that potential for certain diseases may be obtainable only from genetic data, but that is clearly not a kind of information that is interesting in any way from a phylogenetic point of view, so I don't see how that is relevant. Conversely, length is easily obtainable from morphological data, but impossible to get at from genetic data. The same goes for any character that is determined at least partially by external factors.

My understanding is that you run your sequences through programs, there are several different ones, with statistically defined algorithms to obtain reliable scores that have a 1-2.5% chance of being inaccurate for comparison/alignment. It might be similar to the same systems used for morphology but how accurate can those be if you don't have reliable data to input because it's missing? What am I missing here?

It's not similar to the systems used for morphology, it is exactly the same systems as the ones used for morphology. Many of these algorithms and theories were developed before we could get genetic data, and much of the improvements since have been to cope with problems that are specific to genetic data, such as gene trees vs. species trees, and so on.

And, again, genetic data is in no way more or less reliable than morphological data. Genetic data is morphological data, on a molecular level. There is no difference. Certainly individual scientists may produce more or less reliable data, but that is regardless of if they do genetics or morphology. Of course no algorithm into which you feed unreliable data will give you a reliable result (though as Dinwar has pointed out, it will give a result, as that is how the algorithms work). But, as I pointed out with my examples of "sea life", genetic data is not automatically more reliable than morphological data, as both are collected by humans.

 

[Jodie]

No, it is not. The methodology is the same, regardless of the data set. The organisms are entirely interchangeable, so it does not matter what organisms you use: if the methodology works for one set of organisms, it works for all of them. Similarly, if the arguments work for one set of organisms, it works for all of them, and for any set of organisms. This is easily understood once you realise that, given an sufficiently powerful (future) computer, long enough time, and a very dense taxon sampling, we could theoretically construct the complete tree of life for all known extant and extinct organisms. This tree could (theoretically) be constructed using the algorithms and phylogenetic theory we have today. The same is true for any subset of this tree, and thus for all subsets of this tree.

The methodology is only as good the data you have to work with. If DNA is missing for ancient hominids then you only have part of it, so the conclusion of what belongs where can't be as definitive as say a modern example where DNA would be available.

Eye colour is determined better by genetic data than with morphological data? Please explain.

You can't look at a fossilized skull and tell what color the person's eyes once were. If they are alive, different story, but I'm talking about the fossilized remains of our ancestors.

I will agree that potential for certain diseases may be obtainable only from genetic data, but that is clearly not a kind of information that is interesting in any way from a phylogenetic point of view, so I don't see how that is relevant. Conversely, length is easily obtainable from morphological data, but impossible to get at from genetic data. The same goes for any character that is determined at least partially by external factors.

Well it ought to be if affects bone structure.

It's not similar to the systems used for morphology, it is exactly the same systems as the ones used for morphology. Many of these algorithms and theories were developed before we could get genetic data, and much of the improvements since have been to cope with problems that are specific to genetic data, such as gene trees vs. species trees, and so on.

No it isn't, How is MUSCLE, SMART, BALiBASE, T-Coffee, and Prefab used in morphology when looking at fossilized skulls?

And, again, genetic data is in no way more or less reliable than morphological data. Genetic data is morphological data, on a molecular level. There is no difference. Certainly individual scientists may produce more or less reliable data, but that is regardless of if they do genetics or morphology. Of course no algorithm into which you feed unreliable data will give you a reliable result (though as Dinwar has pointed out, it will give a result, as that is how the algorithms work). But, as I pointed out with my examples of "sea life", genetic data is not automatically more reliable than morphological data, as both are collected by humans.

I thought that's why we developed statistics, to help rule out bias. If the programs for analysis aren't reliable then what is in GenBank?

 

[Kotatsu]

The methodology is only as good the data you have to work with. If DNA is missing for ancient hominids then you only have part of it, so the conclusion of what belongs where can't be as definitive as say a modern example where DNA would be available.

I seem to have missed a step somewhere. Are you talking about placing individual specimens in specific species (i.e. phylogenetic inference)? If so, remember that you cannot do that at all with genetic data alone (except for the few species that are defined entirely by their genetics, such as Enchytraeus crypticus and many bacteria).

You can't look at a fossilized skull and tell what color the person's eyes once were. If they are alive, different story, but I'm talking about the fossilized remains of our ancestors.

If it is fossilized, you cannot get at the eye colour with any method, so this seems to be a poorly chosen example.

Well it ought to be if affects bone structure.

You use too few words to allow me to understand what you mean. Please elaborate.

No it isn't, How is MUSCLE, SMART, BALiBASE, T-Coffee, and Prefab used in morphology when looking at fossilized skulls?

I have absolutely no idea to what extent specific programs are used for anything, but i do know that all the large programs for phylogenetic inference -- PAUP*, MrBayes, MacClade -- can be used, and have been used, for both morphological and genetic data. I will agree that alignment programs are not used for morphological data, as morphological data is sort of self-aligned, so that is an entire field of programs and algorithms that are specific to genetic data, however it is also a field that introduced a whole category of subjectivity that simply doesn't exist in morphological data sets.

I thought that's why we developed statistics, to help rule out bias.

Statistic algorithms cannot necessarily rule out the bias that is in place before you feed in the data, though, nor can it account for how you treat the data after the statistical analysis is done. Again, if nuclear and mitochondrial DNA show different pictures, what do you do? Publish both trees? Get more data? Put emphasis on one tree? Hide the other trees in "Electronic material"? Statistical analysis cannot help you with these kinds of biases.

If the programs for analysis aren't reliable then what is in GenBank?

Well, from a phylogeneticist's and taxonomist's viewpoint, mainly crap. Some good data, but it is too often obscured by misidentification, contaminations, non-identifications, "environmental samples", and so on to be very reliable. There are too many species missing from GenBank for it to be very reliable, unless you combine your analysis with a good understanding of the morphology of the group you are studying. Conversely, for some species there are too many sequences...

Another answer: GenBank has pseudo-raw data, which still needs to be put into some form of phylogenetic program to be of any use.

If I download a set of sequences from GenBank (and yes, I've done this hundreds of times), the main problems I have is that:
1. I have no idea if the taxa have been identified correctly, as there is no quality control like that in GenBank;
2. I have no idea if the sequence is actually from the organism it purports to be from, and not for instance a parasite, unless it is from a group of organisms that has been thoroughly sampled;
3. I still have all the problems of constructing phylogenies from genetic data in front of me.

Probably works much better if you work with mice, Drosophila, Arabidopsis, humans, or some other model organism, I suppose, but that much focus on the biology of a single species seems... well, myopic, to me.

 

[Dinwar]

The methodology is only as good the data you have to work with. If DNA is missing for ancient hominids then you only have part of it, so the conclusion of what belongs where can't be as definitive as say a modern example where DNA would be available.

Here's the dirty little secret you're afraid to tell yourself, Jodie: If you have DNA, you only have part of the story as well. For two reasons. First, DNA is only part of the organism--a rather valuable part in terms of evolutionary biology, but only a part. Since your entire argument boils down to "We don't have everything, therefore all conclusions we draw are subjective", this means that DNA-based phylogenies are every bit as subjective as morphological data, because you're only look at part of the dataset! Unless you're willing to openly argue that some data are intrinsically better.....?

Second, as I've already stated, DNA-based phylogenies are only a subset of that subset. They only ever look at part of the available DNA--specific chromosomes, even specific genes. Kotatsu has demonstrated the quite spectacular failings that often accompany such methods.

You can't look at a fossilized skull and tell what color the person's eyes once were. If they are alive, different story, but I'm talking about the fossilized remains of our ancestors.

True. That said, the fact that we can't gather SOME data does not mean we can't gather ANY data. We have bones. What data can be gathered from bones, we can gather.

No it isn't, How is MUSCLE, SMART, BALiBASE, T-Coffee, and Prefab used in morphology when looking at fossilized skulls?

First, it's "How ARE MUSCLE...".

Second, please examine the math behind them. I've given an overview of the math behind phylogeny. I'll be the first to admit it was incredibly broad, and I left out many details, but I'm certain that if you examine the math behind any program that produces phylogenies, you'll find that they fit the description I provided. They have to; all any phylogeny program is, is a way to make those equations more user-friendly. If these AREN'T phylogeny programs, they're irrelevant to this discussion.

I thought that's why we developed statistics, to help rule out bias.

Completely wrong. I don't blame you--this is a view that's astonishingly pervasive, particularly on thisi forum--but that doesn't negate your error. Stats were developed to help simplify handling large datasets. We found that they can be used to detect certain types of bias. That said, a deep understanding of the system in question is necessary to identify bias; the math can help in some instances, but it misses a lot of them. Take a look at the discussions paleontologists have about collection methods sometime to see what I'm talking about.

Secondly, you need to recognize that all statistics--even those programs you mentioned--are GIGO. If you put in crap data, in the form of misidentified species, poorly sampled materials, contaminated materials, or any number of other reasons, you're going to get crap out. The math will not help you when this happens. It's not enough to sit behind a computer and input numbers into a little black box of a program and see what it spits out; to understand biology, anthropology, or paleontology you need to get your hands dirty. You need actual specimens.

 

[Jodie]

Well that was my point with relying on morphology alone. If you don't have the DNA, yo only have part of the data. The same with DNA, and I said this a few pages back, that you can't determine what someone looks like from looking at their DNA.

Thanks for the lesson in grammar, by the way, I overlooked your faux pas for the sake of staying on topic and not appearing to attack the arguer.

Anyway, I have looked at the math behind those programs and the research regarding the reliability for cross comparison of such programs. It was relevant because Kotatsu said that all you needed to do was change the nomenclature of the data entered into those programs and that they were used for both morphological date and genetic data. At least that was what I understood him to mean, you probably read it differently.

 

[Kotatsu]

Anyway, I have looked at the math behind those programs and the research regarding the reliability for cross comparison of such programs. It was relevant because Kotatsu said that all you needed to do was change the nomenclature of the data entered into those programs and that they were used for both morphological date and genetic data. At least that was what I understood him to mean, you probably read it differently.

I have said no such thing; I am a taxonomist, and to me the word "nomenclature" in a discussion about phylogeny and taxonomy has a very precise meaning, which makes your summary of my position nonsensical. What I have said is that the programs commonly (and traditionally) used for phylogeny reconstruction work equally well with morphological and genetic data, but that genetic data requires the use of alignment programs that add a level of subjectivity to your analysis that is not present in analyses of morphological data. Many of these programs, and the algorithms underlying them, were developed during a period when we didn't have any genetic data at all.

 

[Jodie]

The alignment programs have an error rate of 1-2.5% depending on what you are looking at, how is that subjective? Do you think that kind of percentage is really going to skew the decision to place something in a specific species when it truly belongs in another?

I hear you saying that the results are only as good as the data entered, I don't disagree, but in the case of most DNA you do have a type specimen so that morphology can be compared. That isn't true with ancient hominid skulls. I can't see how it could be as accurate. Not to mention that you only have a small representation of fossilized skulls for these different species of people to begin with.

 

[Kotatsu]

The alignment programs have an error rate of 1-2.5% depending on what you are looking at, how is that subjective? Do you think that kind of percentage is really going to skew the decision to place something in a specific species when it truly belongs in another?

That is entirely incidental to my argument, as you would have known, had you read it. The subjectiveness of alignment of DNA sequence has to do with how you treat mismatches and ambiguities. Do you simply accept the alignment that the program gives you? Do you excise large sections of the DNA that only occur in some taxa? Do you include primer regions or not? In short: when you get your alignment out of whatever program you use, do you accept it as-is, or do you look at it and make sure it looks okay? Both possibilities introduces subjectivity in a manner that does not exist in morphological analyses.

I hear you saying that the results are only as good as the data entered, I don't disagree, but in the case of most DNA you do have a type specimen so that morphology can be compared.

For most DNA sequences, you have a voucher, not a type specimen. A type specimen is something quite specific in taxonomy. That is why I specifically pointed out that:

I seem to have missed a step somewhere. Are you talking about placing individual specimens in specific species (i.e. phylogenetic inference)? If so, remember that you cannot do that at all with genetic data alone (except for the few species that are defined entirely by their genetics, such as Enchytraeus crypticus and many bacteria).

With DNA data alone, you cannot identify more than a handful of metazoan organisms. Most multicellular organisms, at least, have a type specimen that you could potentially compare your voucher with, though this is not always done. And even when the scientist is as thorough as they should be, and actually do compare the voucher with the type (or some part of the type series, at least), there is still the problem with the type only being a clearly preserved (hopefully!) individual, and does not capture the variation present in the species.

That isn't true with ancient hominid skulls. I can't see how it could be as accurate. Not to mention that you only have a small representation of fossilized skulls for these different species of people to begin with.

I think you are confusing the terminology here. Surely, for fossilised hominid skulls, all we would have is precisely type specimens (or other specimens)? I assume that what you mean is that for these specimens, we do not have DNA vouchers. I do not see how that is a problem, though.