Re: The taxonomy of Sahelanthropus tchadensis from a craniometric perspective

On 8/7/24 6:56 AM, Pandora wrote:

Op 05-08-2024 om 22:49 schreef John Harshman:
 

On 8/5/24 9:14 AM, Pandora wrote:

Op 05-08-2024 om 17:57 schreef John Harshman:
>

On 8/5/24 8:19 AM, Pandora wrote:

Op 05-08-2024 om 16:39 schreef John Harshman:
>

On 8/5/24 7:19 AM, Pandora wrote:

Op 04-08-2024 om 21:19 schreef JTEM:
>

  Pandora wrote:
>

Actually, there's more than one individual of this taxon, from three different localities (TM 247, TM 266 and TM 292). This additional material was announced in Nature in 2005:

>
Where are those localities?  I just did an exhaustive 30 second search
and could only find an actual location associated with 266.
>
And, yes, I did search longer than 30 seconds but it wouldn't have been
nearly as funny if I offered a better time estimate...

>
If it doesn't exceed your attention span you can read the paper at:
>
https://www.researchgate.net/publication/7920249
>
Anyway, the three Sahelanthropus sites are within an area of 0.73 km2.
>

Not too far from where another hominin taxon, Australopithecus bahrelghazali, was discovered in 1995.

>
That appears to be where the 266 was found.

>
No, the Toros-Menalla Sahelanthropus sites are about 150 km west of the Koro-Toro australopithecine site and stratigraphically ~3.5 million years older.
>

If you think that's the wrong place you must have some concept of what is the right place. Where would that be?

>
Well any other day of the week the clown act insists it's South Africa:
>
https://en.wikipedia.org/wiki/Cradle_of_Humankind
>
I would have guessed that you knew.

>
But why do you think South-Africa is the right place?
The phylogenetically most basal and stratigraphically oldest hominins are from East- and North-Africa.
>
See for example:
https://doi.org/10.1016/j.jhevol.2023.103437
>
Their Bayesian inference analysis, with posterior probabilities for nodes given as percentages (fig.6):
>
https://ars.els-cdn.com/content/image/1-s2.0-S0047248423001161-gr6_lrg.jpg
>
That tree topology would refute your hypothesis.

>
To be fair, if those are Bayesian posteriors, many of them are pretty bad. But what is JTEM's hypothesis?

>
That australopithecines are the ancestors of the African apes (Pan and Gorilla) as suggested in this publication (see fig.3 on page 17):
>
https://www.researchgate.net/publication/376650459
>
And apparently also that the human clade originated in South-Africa.
>
Of course, none of it with any substantial support from phylogenetic systematics.

>
I don't believe that the South African origin is JTEM's belief. He seems to be making fun of it. But you may be right about his other hypothesis. It's so hard to tell.
>
The node that puts Sahelanthropus into the human lineage gets only 90% Bayesian support, which is very low, and the 77% for the Pan/human node means it might as well be collapsed, leaving a trichotomy for Gorilla/Pan/hominins.

>
Let's also throw in a little parsimony analysis (majority-rule consensus tree from 10,000 pseudoreplicates, bootstrap support values (%) given as Group-present/Contradicted (GC) frequencies:
>
https://ars.els-cdn.com/content/image/1-s2.0-S0047248423001161-gr3_lrg.jpg
>

Thanks. Notice how bad those bootstrap values are. But something seems odd. How can the great ape node get only 2% bootstrap support? Was the data set chosen so as to omit most of the character support? No, only lots of conflict could give support that low, and there aren't enough possible trees for a tree supported at only 2% to come out on top in a majority rule consensus. Still, anything below 70% might as well be collapsed, leaving a massive polytomy.

 But notice how support for the great ape node in iteration 3 of the parsimony analysis changes to 98% when Papio and Colobus are removed and fossil taxa Victoriapithecus and Ekembo are retained as outgroup:
 https://ars.els-cdn.com/content/image/1-s2.0-S0047248423001161-gr4_lrg.jpg
 Now, how can that be?

I can't see how that would work. Again, it doesn't seem possible that a node supported in only 2% of bootstrap replicates could be the leader in any bootstrap consensus, no matter the data. And changing from 2 to 98? What possible conflict could produce such a result? Something very odd is happening.

Post et al. did three iterations using parsimony and three using Bayesian Analysis. I've reproduced the figures from their table 2 (hoping it formats right in your reader), with iteration 1, 2, and 3 as column 1, 2, and 3 respectively:
 Bootstrap support (%):
 Pongo          21   2  98
Gorilla        74  59  86
Pan            41  40  52
S. tchadensis  30  32  31
A. ramidus     68  62  75
Au. anamensis  74  73  75
Au. afarensis  63  62  63
 Bayesian posteriors (%):
 Pongo         100  97  80
Gorilla        93  96  98
Pan            73  77  73
S. tchadensis  86  90  84
A. ramidus     80  80  81
Au. anamensis  95  94  93
Au. afarensis  97  97  97
 What's your cut-off point (low vs high) with regard to posteriors in Bayesian analysis?

There can't be an absolute cutoff point, but based purely on my experience I would consider anything less than 90% to be garbage and less than 95% dubious. Then again, maybe analyses have improved recently, perhaps there is better mixing than in the past, and maybe morphological analyses are different. The Bayesian posteriors change much less among iterations than do the bootstraps. Is that a good thing or a bad thing?

The Paranthropus node ("robust australopithecines" (not in the table)) gets persistently high support in both parsimony (never less than 96%) and Bayesian analysis (always 100% posterior probability)

Yes, I noticed that was one of the few consistently supported nodes.