The Problem Of Homoplasy Continued

[Bonedigger 42]

As I have said before, and will say many more times as well, homoplasy (convergence and parallelism) is a problem for evolutionary theory on EVERY taxonomic level. I illustrated that on a rather high taxonomic level for mammals in Gilbo's Convergent Evolution Defies Evolution thread. I have also illustrated that at the subfamily level for the Equinae in my Problem of Homoplasy thread. Here is an example at the lowly subgenus level.

In 1990, Stephen W. Taber published a cladistic analysis of North American representatives of the genus Pogonomyrmex, subgenus Pogonomyrmex (the western harvester ant). His paper, titled "Cladistic phylogeny of the North American species complexes of Pogonomyrmex (Hymenoptera: Formicidae)", can be downloaded at this page on Antweb (http://www.antweb.org/description.do?subfamily=myrmicinae&genus=pogonomyrmex&rank=genus&project=allantwebants&project=allantwebants), or directly here (http://antcat.org/documents/2964/2786.pdf). I entered his data into the TNT program (which can be downloaded here at a different page than I have posted before as they have moved it- http://www.lillo.org.ar/phylogeny/tnt/). For those interested in pursuing it further, here is my tnt file for Taber's 1990 analysis (Taber 1990-Pogonomyrmex.tnt), and here is my tre file matching his results (Taber 1990-Pogonomyrmex.tre). Interestingly, I was unable to duplicate his cladogram with the TNT program, although there probably is a way if you tweak enough parameters. However, I was able to manually edit a tree to duplicate his for the purpose of counting homoplasy. In his final cladogram, he recognizes three subgroups, which I have graphically illustrated below; the barbatus group (blue), the occidentalis group (green), and the californicus group (magenta).



And here is the homoplasy count for his cladogram, with a total of 73. The character numbers are given in the row and column headings (tens increments for the rows and ones increments for the columns), and, as I explained before, each count in the table is the number of independent duplicate originations of a character state. Character 0 is a null character used to shift the character numbering to start at 1.



Out of 33 characters (with 78 total character states), that is a lot of homoplasy. In other words, in this evolutionary scenario, 73 times a character state would have to have been acquired a second (or third) time, in addition to it having been acquired by some other line of descent for the first time in the case of parallelisms, or having been present in an earlier ancestor in the case of reversals. In fact, only five (5,9,12,18,33) of the characters are purely synapomorphic (or, in evolutionary terms can be unequivocally attributed to common descent). For example, here are a couple of the character maps for the more extreme instances of homoplasy, character #16:



and character #6



Here is a pdf with all of the character maps for anyone interested (Taber-1990-Character_Maps.pdf). Taber himself recognized the large amount of homoplasy present:

Homoplasies are common in the genus, as indicated by the rather low consistency index (0.369). Parallel evolution occurred in 25 characters, reversals occurred in 18 characters, and both occurred in 14 characters. (Page 316)

Ironically, I have no problem with regarding all "species" of Pogonomyrmex as having descended from a common ancestor. The problem I have is with how the TOE "explains" the origin of those characters as culled genetic accidents acquired by the taxa. This kind of distribution pattern, where characters occur repeatedly in different combinations scattered throughout the taxa, to me at least, just screams that they were present in the original ancestor to begin with. And that, by extension, argues for premeditated design where a wide range of variability can be selected for from an original genome as they disperse and "adapt" to local environments.

Now, the characters being used are species level characters. In other words, they are minor differences that can be virtually indistinguishable except with careful, close study. I pointed that out in the Baraminology Unplugged and Unhindered thread, but I'll illustrate that here. For example, character #25 is the sculpture of the exoskeleton in the head and thorax which varies between "typically coarse", the plesiomorphic state, and "not typically coarse", the derived state. Here is a picture of the head of Pogonomyrmex occidentalis, courtesy of Antweb (http://www.antweb.org/bigPicture.do?name=casent0005718&shot=h&number=1), that illustrates the former state.



The vertical rugae (wrinkles or ridges in the exoskeleton) are relatively widely spaced.

And here is a picture of the head of Pogonomyrmex barbatus, also courtesy of Antweb (http://www.antweb.org/bigPicture.do?name=casent0235329&shot=h&number=1), illustrating the derived state where the rugae are more finely spaced.



These two pictures also allow for a comparison of character #5, one of the few unreversed synapomorphies in Taber's analysis. Character #5 is the relative straightness of the cephalic (head) rugae, with "divergent" being the plesiomorphic state, and "straight or only slightly divergent" being the derived state. In P. occidentalis, you can see how as the rugae travel up to the top of the head, they diverge away from the midline. In P. barbatus, they remain relatively parallel to the midline of the head. You can also see character #6, cephalic aerolation, in the P. occidentalis head. Cephalic aerolation refers to the "pitting" in the spaces between the rugae. P. occidentalis has dense aerolation, but Taber scored the character as ambiguous in P. barbatus because it varies within the species. In the picture above, you can see very few pits in between the rugae in P. barbatus.

As I mentioned earlier, I was unable to duplicate Taber's cladogram with the TNT program. In fact, using the New Technology search with default settings yields a single most parsimonious tree that differs from Taber's in that it separates the occidentalis group (green), but places the californicus group (magenta) as ancestral to the barbatus group (blue) (illustrated below).



A homoplasy count for that arrangement yields a count of 71, which is 2 less than for Taber's analysis.



I'll have more to post on this. Originally I was just going to post this in Gilbo's Convergent Evolution Defies Evolution thread, but I have Taber's book (http://www.amazon.com/gp/product/0890968152/) on the way, in which he does a much larger analysis that includes all Pogonomyrmex species (North, Central, and South American), and utilizes almost three times as many characters. I'm really interested in seeing how homoplasy maps out in that analysis. I also thought it would be better to start a new thread that can examine homoplasy in a wide range of groups, since my original Problem of Homoplasy thread is focused primarily on the Equinae.

 

[Calypsis4 640]

This was very good information, Bonedigger. I think perhaps it did not draw attention because it is so technical and hard to mentally digest. Sort of like my thread "The Mysterious Sixth Power of Separation" which was likewise technical and it covered so many different areas that most readers are not familiar with.

 

But the cladistics argument that our opponents have used is really based on imaginary information. Like always, they take what we know now and connect it with dots...dots that are not justified upon careful consideration.

[Bonedigger 42]

This was very good information, Bonedigger. I think perhaps it did not draw attention because it is so technical and hard to mentally digest. Sort of like my thread "The Mysterious Sixth Power of Separation" which was likewise technical and it covered so many different areas that most readers are not familiar with.

 

But the cladistics argument that our opponents have used is really based on imaginary information. Like always, they take what we know now and connect it with dots...dots that are not justified upon careful consideration.

 

Thanks Cal. I'm about 2/3 done with my next post for this thread. I've created the tnt file character matrix and done the analysis for Taber's 1998 book. I just have to type in all of the character descriptions and print the character maps to file. Just as a preview of the results, he uses 59 taxa (including the outgroup), and 80 characters. The homoplasy count is 293 for the "best fit" evolutionary arrangement. That's 293 times evolution would have had to produce the same character a second (or third) time, if the standard darwinian explanation of chance acquisition and selection is used to "explain" the diversity.

[Calypsis4 640]

 

Thanks Cal. I'm about 2/3 done with my next post for this thread. I've created the tnt file character matrix and done the analysis for Taber's 1998 book. I just have to type in all of the character descriptions and print the character maps to file. Just as a preview of the results, he uses 59 taxa (including the outgroup), and 80 characters. The homoplasy count is 293 for the "best fit" evolutionary arrangement. That's 293 times evolution would have had to produce the same character a second (or third) time, if the standard darwinian explanation of chance acquisition and selection is used to "explain" the diversity.

 

Right, brother. Go for it. I'll be glad to read it.

[nnjamerson 12]

As I have said before, and will say many more times as well, homoplasy (convergence and parallelism) is a problem for evolutionary theory on EVERY taxonomic level. ...Here is an example at the lowly subgenus level.

 

In 1990, Stephen W. Taber published a cladistic analysis ....

 

Interestingly, I was unable to duplicate his cladogram with the TNT program, although there probably is a way if you tweak enough parameters. However, I was able to manually edit a tree to duplicate his for the purpose of counting homoplasy...

Out of 33 characters (with 78 total character states), that is a lot of homoplasy. In other words, in this evolutionary scenario, 73 times a character state would have to have been acquired a second (or third) time,

 

Ironically, I have no problem with regarding all "species" of Pogonomyrmex as having descended from a common ancestor. The problem I have is with how the TOE "explains" the origin of those characters as culled genetic accidents acquired by the taxa. This kind of distribution pattern, where characters occur repeatedly in different combinations scattered throughout the taxa, to me at least, just screams that they were present in the original ancestor to begin with. And that, by extension, argues for premeditated design where a wide range of variability can be selected for from an original genome as they disperse and "adapt" to local environments.

 

As I mentioned earlier, I was unable to duplicate Taber's cladogram with the TNT program. In fact, using the New Technology search with default settings yields a single most parsimonious tree that differs from Taber's ...

 

 

 

 

Sorry, i don't get why this all leads you say "homoplasy (convergence and parallelism) is a problem for evolutionary theory on EVERY taxonomic level. ..."

 

You're looking at one very shoddy dataset from 25 years ago with a high degree of plasticity, which means very poor rigour and little intrinsic robustness, with very few discrete changes supported by few to no others, and actually a vast amount of internal conflict in the character set ...

I'm not in the least surprised you failed to recreate the results using different software that has different algorithms for optimising character states, the polymorphism and conflict seems clearly leading you to find other similarly poor solutions. Bad data in, bad results out, no matter how sophisticated the software. It's simply a bad dataset and shoddy conclusions.

 

I particularly liked your " TOE 'explains' the origin of those characters as culled genetic accidents acquired by the taxa. This kind of distribution pattern, where characters occur repeatedly in different combinations scattered throughout the taxa, to me at least, just screams that they were present in the original ancestor to begin with" I think you're misinterpreting how evolutionary theory explains such character variation. Homoplasy is often simply when characters have experienced convergent selective pressures, and occurs when allowed by a high degree of structural plasticity. Here, in the cladistic framework, homoplasy is inferred as independent derived acquisition of the character state in question, i.e. in different decedents NOT grouped together on the preferred topology... consequently the directly conflicts with them being "present in the original ancestor"... so how do you get from showing multiple examples of the problem of independent derived gains to say "to me at least, just screams that they were present in the original ancestor to begin with". But, i'll grant that many such characters interpreted in such poorly resolved and unsupported cladograms can be instead ancestral shared characters, and simply misinterpreted as derived gains in a bad dataset and shoddy conclusions.

 

Please let me know if there are any nodes that remain resolved after running various analyses to show robustness in the inferred nodes, such as botstrapping, jackknifing or even bremer support. I'd expect none of the nodes can be considered robust, and hence any conclusions are very unstable ... as you hint at. So, ergo as i said before, bad dataset and shoddy conclusions. Case closed?

[Bonedigger 42]

Well, I was waiting until I had Taber's subsequent broader analysis ready so I could include it in my response, but since I still haven't had a chance to get back to finishing that up, it will have to wait for another post...

Sorry, i don't get why this all leads you say "homoplasy (convergence and parallelism) is a problem for evolutionary theory on EVERY taxonomic level. ..."

If you don't understand why I say that, then you obviously didn't bother to read the other thread and other post I hyper-linked to, where I illustrate the same phenomenon in other taxa at higher taxonomic levels. That in itself demonstrates that your point is to dismiss rather than address the problem being raised. To put it simply, ubiquitous homoplasy is not just an artifact of species level characters where the character transformations required are relatively small and at a low taxonomic level.

You're looking at one very shoddy dataset from 25 years ago with a high degree of plasticity, which means very poor rigour and little intrinsic robustness, with very few discrete changes supported by few to no others, and actually a vast amount of internal conflict in the character set ...
I'm not in the least surprised you failed to recreate the results using different software that has different algorithms for optimising character states, the polymorphism and conflict seems clearly leading you to find other similarly poor solutions. Bad data in, bad results out, no matter how sophisticated the software. It's simply a bad dataset and shoddy conclusions.

Perhaps the most amusing thing about this response is the way you manage to shoot yourself in your own foot. The data-set that Taber used, which you dismiss as a "shoddy dataset", is in fact the suite of characters that are used to diagnose the various species of Pogonomyrmex (sensu stricto). If that makes for a bad data-set, then doesn't that just prove my contention about the arbitrariness of genus and species designations in ants, which you derided in another thread? Which is it NN? Bad data, or a wishy-washy bad argument on your part. You can't argue that the data is bad, but the species designations diagnosed by that data are valid.

But, perhaps, the real question is: What do you consider a "good" data-set? One that has been cherry-picked for a preferred topology, with any "cladistically weak" characters eliminated beforehand? And why does the fact that it's 25 years old make it a "shoddy dataset"? Has ant taxonomy changed that much in 25 years?

I particularly liked your " TOE 'explains' the origin of those characters as culled genetic accidents acquired by the taxa. This kind of distribution pattern, where characters occur repeatedly in different combinations scattered throughout the taxa, to me at least, just screams that they were present in the original ancestor to begin with" I think you're misinterpreting how evolutionary theory explains such character variation. Homoplasy is often simply when characters have experienced convergent selective pressures, and occurs when allowed by a high degree of structural plasticity. Here, in the cladistic framework, homoplasy is inferred as independent derived acquisition of the character state in question, i.e. in different decedents NOT grouped together on the preferred topology... consequently the directly conflicts with them being "present in the original ancestor"... so how do you get from showing multiple examples of the problem of independent derived gains to say "to me at least, just screams that they were present in the original ancestor to begin with". But, i'll grant that many such characters interpreted in such poorly resolved and unsupported cladograms can be instead ancestral shared characters, and simply misinterpreted as derived gains in a bad dataset and shoddy conclusions.

And...here is your armchair rationalization which just begs the question. How do you distinguish homologous characters from homoplastic characters? There is nothing intrinsic in the characters themselves to distinguish between the two. The distinction is purely post-analysis. Characters that conform to the resulting phyletic arrangement are synapomorphic (shared derived), and characters that don't conform are deemed to be homoplastic. In fact, characters can often flip flop between the two states, depending on the analysis. Rather than demonstrating the plasticity of the characters, it demonstrates the plasticity of evolutionary explanations for the origin of those characters. The reason it screams to me that they were present in the original ancestor, is because it's far more likely that these characters can be variably expressed or suppressed in different combinations, than that they just "happen" to be acquired multiple times. And that is the crux of the issue. ACQUISITION and inheritance is at the core of any evolutionary explanation for the diversity of life on earth. Without the acquisition of characters, evolution is left with nothing to explain other than the steadfast faith of its devotees.

Please let me know if there are any nodes that remain resolved after running various analyses to show robustness in the inferred nodes, such as botstrapping, jackknifing or even bremer support. I'd expect none of the nodes can be considered robust, and hence any conclusions are very unstable ... as you hint at. So, ergo as i said before, bad dataset and shoddy conclusions.

Feel free to analyze away. I gave you a link to the TNT program (I use the no taxon limit version). You don't even have to do the grunt work of entering the character matrix, character names, and taxon names into the program. I posted my tnt file in the OP. In fact, I challenge you to find an arrangement of the taxa that won't yield an abundance of homoplasy. The problem does not lie in finding just the "right robust" arrangement of the taxa. The problem is that the characters are mosaically distributed, and do not follow a pattern of acquisition and inheritance.

Case closed?

No, case ignored. The core of my argument, in my previous thread and in this one, is that while homology (morphological and genetic similarity) is used as "proof" of common descent, when it comes down to the actual data when dealing with the real world, you have to be completely schizophrenic in how you apply that argument, using it when it is consistent with an a priori assumption of common descent, while dismissing it with ad hoc explanations like homoplasy when it isn't consistent with an assumed pattern of descent. Rather than addressing that argument, you just dismissed the data used in Taber's analysis as "bad". I notice you didn't provide a "better" data-set. I wonder, why is that?