A pragmatic redating of our 700 million year ancestry

As part of the second edition of the Ancestor’s Tale, I have been reassessing the dates on the backwards journey from today’s humans to the origin of animals. What is written below is rather technical, and mostly for my own benefit, and also for Richard Dawkins. Nevertheless, perhaps others may be interested in my reasoning, which I hope is seen as a pragmatic assessment of our current beliefs.

Molecular clock papers are generally more accurate (if not more precise) than 10 years ago, since methods have been developed to deal with heterogeneity in the rate of molecular evolution. Hence my revised dates rely heavily on recently published studies, such as by , and particularly  which has a comprehensive coverage of primates and combines multiple fossil calibrations with heterogeneity in the molecular clock rate. For a wider view of molecular clock dates, it is worth looking at the Timetree of Life web site, although their dates do not (in my opinion) take enough account of palaeontology. Other problems with “Timetree of Life” include quoted mean values which include many outdated/inapplicable studies, and “expert opinions” which tend to be entirely from researchers who prefer deep molecular divergence dates. The Timetree project does, however, provide a particularly useful book chapter on fossil constraints , which updates an older interactive version. These provide comprehensive (although sometimes out-of-date) reviews of the paleontological constraints for many vertebrate nodes.

My personal take on the dating changes wrought by the last 10 years of evolutionary research breaks down into 3 time periods, loosely corresponding to very recent, intermediate, and very deep dates. I have not given confidence intervals below, as these are somewhat meaningless for summary dates, which are mostly rounded to the nearest 5 million years anyway (I assume an error of at least ± 5 million years for most estimates > 20MaaI have switch to the standardised convention of using Ma to mean “millions of years ago”). The original data on which my summaries are based, including some of the ranges and confidence intervals, are available in my csv file and displayed in the summary plot at the end of this post.

Time period 1: changes in the recent period

The recent reassessment of the mutation (strictly, substitution) rates and generation times of humans and chimpanzees (as discussed in  and more popularly by John Hawks) leads me to push for older divergence dates on the evolutionary tree of apes. However, there is an increasing realisation that the dates of speciation can be several million years younger than the divergence dates of DNA sequences, and the most recent genealogical ancestor (the “date of last genetic exchange”) can be even more recent still, as I discuss below.

I suspect that some of the slowest mutation rates (e.g. 1.2×10-8/bp/gen = 0.4×10-9 /bp/yr giving an average human/chimp divergence of 13 Ma) are either caused by bias in the whole genome sequencing methodology (e.g. some single nucleotide mutations are not being spotted), or because there is something odd going on with recent mutation or substitution rates in great apes. A paper in Nature the week before last  produces an estimate in line with the average from other studies, approximately 0.5×10-9 substitutions per base pair per yearbMore specifically, 0.44…0.63 × 10-9 by “direct” measurement (subject to errors in different methods of genome sequencing of modern humans), or 0.43×10-9 by inference from similar demographic patterns across the genome between modern and ancient humans.. Since this applies over at least the past 45,000 years, I deem it a reasonable estimate, which provides slightly older dates for the human/chimp and human/gorilla divergences than the ~8 and ~10 million year old estimates from the AUTOsoft methods in , as discussed below.


Figure 3b from Scally et al. (2012), distributed under CC-BY-NC-SA

With a 0.5×10-9/bp/yr mutation rate, and a slight  “hominoid slowdown”, we get average gene divergence dates of approximately 10 Ma for the average chimp/human divergence and perhaps 11 or 12 Ma for the human/gorilla divergence. Following  this corresponds to “speciation times” of about 6-7 Ma and 9-10 Ma respectively. Note that they calculate these values on the basis of speciation by a clean separation between populations, which is rather unlikely. They also say:

the speciation time estimated by [our method] represents an average weighted by gene flow over the period of separation. This means in some cases it can be substantially older than the date of most recent exchange.

Hence the last shared genealogical ancestor (the concestor), is likely to be even younger than this estimated “speciation time”, although by an amount which may be difficult to predict, depending as it does on the odd instance of hybridization occurring well after the main speciation event.

10 Ma (concestor 1 @ 5-7 Ma): Human – chimp average genetic divergence. The AUTOsoft method in   gives 7.93 Ma, which is considerably older than most molecular clock dates from the 1980s and 90s. Even so, recent whole-genome estimates of mutation and substitution rates imply that the overall rate of molecular evolution in the apes has slowed even more than accounted for in the  analysis. In the extreme, these whole-genome estimates produce average genetic divergence dates around 11-13 Ma. Pragmatically, I take a somewhat intermediate value of 10 Ma, with the speciation process perhaps 2-3 million years laterc estimate this as more like 3-4 million years. The concestor date will be younger than this: broadly 5-7 Ma. In a happy accident, this coincides with the dates quoted in the first edition of the Ancestor’s Tale.

12 Ma (concestor 2 @ 8 Ma, was 7 Ma): Human – gorilla average genetic divergence gives a date of ~ 10 Ma (strictly AUTOsoft gives 9.75), versus dates from  of approx 15 Ma under current mutation rates, or approx 12 Ma with a hominid slowdown. An approximately 1-2 million year separation between this and the human-chimp date would account for the observed degree of incomplete lineage sorting. The date of “speciation” is quoted by  as 9-11 Ma, and the date of the concestor is likely to be more recent still, perhaps 8 Ma.


Concestor 3

18 Ma (concestor 3 @14 Ma): Human – orang-utan average genetic divergence.   calculate 18 Ma using their AUTOsoft method (the average of all methods is 15.1 Ma). do an interesting analysis of ILS, and likewise find an 18 Ma date for genetic divergence, but surprisingly recent dates of speciation, from 9-13 Ma. However, these estimates use an obsolete generation time (20 years) and mutation rate (0.8×10-9 to 1.2×10-9/bp/yr), and do not account for a hominid slowdown. Taking this into account gives perhaps 14-15 Ma as a separation date. A marginally more recent concestor date seems reasonable, hence 14 Ma is not unlikely.

Time period 2: Changes over most of the evolutionary timespan

In the next tranche of estimates I only give a single date, since the values are in the tens or hundreds of millions of years and the difference between mean genetic divergence and separation or even concestor dates becomes relatively insignificant. Moreover, this difference will become smaller as generation times decrease in tandem with smaller ancestral body size (although this may be partially offset by a concomitant increase in ancestral population size). In general, I think it is reasonable to assume <3MY between mean genetic divergence and the time of the most recent common genealogical ancestor (the concestor).

Pleasingly, most of the dates between 20 and 500 million years ago remain basically unchanged from my guesses 10 years ago. I’m inclined to make small changes to the primate root, slightly large changes to the placental/mammalian roots, and minor tweaks to the lobe-fin divergences. Here are my estimates. Paragraphs in grey are unchanged from the 1st edition of the Ancestor’s Tale, with justification for these dates not necessarily given here.

Concestor 4 @ 18 Ma: Human – gibbon split. As we go backwards up the primate tree, the impact of a hominid slowdown is decreased and dates are less dependent on a single fossil calibration point. With reference to , it seems pragmatic to start switching gradually from their AUTOsoft date to the average of their 4 methods. In the case of gibbons, their AUTOsoft value is 20.9, and the mean 17.35 Ma. Given perhaps two or three million years between average genetic divergence and the concestor point, I see no reason to change this from my previous 2004 estimate of 18 Ma.


Concestor 5

Concestor 5 @ 25 Ma: Human – old-world-monkey split. The fossil record constrains this between 23.5 and 34 Ma.  give an AUTOsoft value of 28.8 and a mean over all methods of 25.07 Ma. We previously had 25 Ma, ’nuff said.

Concestor 6 @ 40 Ma: Human – new-world-monkey split. This is the first instance on the journey that the AUTOsoft value (40.6) from is more recent than their mean from all methods (44.5). I am prepared to accept a mean gene divergence date of perhaps 41 to 44 Ma. When subtracting a few million years from this to measure the concestor time, we obtain our previous estimate of 40 Ma (which remains a nice round number…)

Concestor 7 @ 60 Ma (was 58): Human – tarsier split.  give 61-63 Ma, which is a fraction older than my previous estimate of 58 Ma. On balance, I reckon a round 60 Ma gives the right idea.

Note that the combined analysis of  places the next 6 dates in the Cenozoic, on the basis that there are no undisputed crown group placental mammals identified from the Cretaceous. I am unwilling to take their view that the divergence of placental mammals was entirely post K/Pg . Their omission of Juramaia  is also problematic. The dates I use are consistent with a “short fuse” model.


Concestor 8

Concestor 8 @ 65 Ma (was 63): Human – lemur split (origin of crown group primates). As we say in the book chapter, this could be either before or after the K/Pg boundary at 66 Ma. More recent molecular clock dates push this just into the Cretaceous, albeit with substantial error (: mean = 67.8, AUTOsoft = 66.7, : 68). Accounting for million years or so between genetic and speciation dates could easily push this after the boundary, and just into the range of 55.5 to 65.8 Ma from .

Concestors 9 & 10 @ 70 Ma: Human – colugo/tree shrew split. Evidence is starting to point to this being two rendezvous points, relatively evenly spaced between the primate and rodent divergences , so perhaps 69 & 71 Ma for each.

Concestor 11 @ 75 Ma: Human – glires (rodent + lagomorph) split. Since there are no undisputed Cretaceous placental fossils (unless e.g. Zalambdalestids at approx 95 Ma belong in Glires), we rely entirely on molecular clock dating for this point. The summary of methods in  gives approximately 69 / 88 / 75 Ma for 3 different calibration strategies, so I am disinclined to change the date from my guess of 75 Ma in the first edition of the Ancestor’s Tale (which had this as Rendezvous 10).

Concestor 12 @ 85 Ma: Human – laurasiathere split. The same arguments go for this point as for Concestor 11, but with the 3 calibration strategies giving 75 / 95 / 82 Ma, again supporting the previous date of 85 Ma given to what was then Concestor 11.

The next 2 points used to be placed concurrent with the continental split of Africa / S. America / Laurasia, assumed to be around 105 Ma. However, that date has been pushed back to 120 Ma . While some have argued this means the speciation dates should be similarly pushed back, I suspect that this severs that the direct link between continental separation and gene divergence (although the possibility of dispersal to already-slightly-separated continents is still plausible).


Concestor 12 or 13, also see my previous post

Concestor 13 @ 90 Ma (was 95 & 105): Human – xenarthra split. I place this slightly earlier than previously, on the basis of  who give approx 89 Ma, and also an inclination for slightly shallower points on the basis of arguments in . There are strong arguments that the afrothere / xenarthra / etc split was almost simultaneous (see  and my previous blog post, which also explains why I am following the current slight trend of grouping xenarthra & afrotheria together into the “Atlantogenata”.

Concestor 14 @ 160 Ma (was 140): Human – marsupial split. Most molecular clock studies put the marsupial divergence relatively close in time to the monotreme divergence. The recent finding of Juramaia , a fossil mammal from 160 Ma and claimed as a crown group eutherian, supports this deeper placing of the marsupial divergence.

Concestor 15 @ 180 Ma: Human – monotreme split. This accords well with 185 Ma from  and their comparisons in a later paper . It also fits with the fossil constraints of 162.5 to 191.1 Ma , so there seems little reason to change it.


Concestor 16

Concestor 16 @ 315 320 Ma (was 310) Human – bird/reptile split. There is a large body of literature on this, as it is perhaps the most commonly used calibration point. I have moved it back a little from 310 to 315 320 Ma on the basis of (see comments to post). Since this is largely fossil-based, the expected gene divergence is probably a little earlier in time, nearer 325 Ma.

Concestor 17 @ 340 Ma: Human – amphibian split. This accords well with fossil constraints of 330.4-350.1 from and the molecular clock estimate of 330-345 Ma from the DeepFin analysis 


Concestor 18

Concestor 18 @ 415 Ma (was 417): Human – lungfish split. The previous date of 417 gives a false sense of accuracy. Note that the DeepFin molecular clock  gives a much younger date, at 375 Ma, which is inconsistent with 390 Ma tetrapod footprints . This also contradicts the argument of a tight fossil constraint of 408-419 Ma . To allow time for the ray-fin divergence between 420 and 425 Ma, the lungfish and coelacanth splits need to be squeezed into the more recent part of the 408-419 Ma range (see comments to post). Additionally, the difficulty disentangling the lungfish and coelacanth divergences argues they diverged from the tetrapod lineage in relatively rapid succession . I thus alter the timing for concestors 18 & 19 to 410 415 and 415 420 Ma.

Concestor 19 @ 420 Ma (was 425): Human – coelacanth split. See above.

Concestor 20 @ 430 Ma (was 440): Human – ray-finned fish split. Almost within the fossil constraint of 416-421.25 Ma The fossil constraint is now thought to be 445-420 Ma. The DeepFin analysis gives 426.6 Ma which I consider to be an underestimate . Nevertheless, it seems reasonable to pull this in to nearer rendezvous 18 and 19 (the DeepFin tree puts this point about 3/4 of the way between Rendezvous 21 and 19). I have gone for 430 Ma, but 435 is probably just as pragmatic.

Concestor 21 @ 460 Ma: Human – shark split. Fossil constraints are 421-462 Ma , with the DeepFin paper giving a molecular clock date of 465 Ma (contrast this with the excessively deep 526 Ma “expert opinion” from timetree). Another possible source of support for this date is the discovery of shark-like scales from ~ 460 Ma . Although these are not concrete proof of this divergence point, I find no contradictory evidence to dispute it.

Time period 3:

I am relatively convinced by the arguments in  that previous attempts to reconcile precombrian molecular and fossil dates are somewhat suspect, and deliberately weighted so as to give little statistical weight to the molecular signal. In other words, I am coming round to the idea that there is indeed real conflict between “rocks” and “clocks” in the precambrian. Generally, I tend to come down more on the side of the rocks, because I suspect that there is some as-yet not understood effect which is modifying the rate of molecular evolution (could it be a change in generation times?). An additional worry of mine is that the deepest of these dates are based on extrapolation from fossil constraints, not interpolation between them; extrapolation produces notoriously overprecise confidence intervals.

In fact, the fossil record paints a relatively consistent story , with undisputed trace fossils (indicative of coelomate animals) around 555 Ma, cnidaria-like fossils from the Doushantuo formation (~ 585 Ma), and sponge  spicules from 630 Ma  with sponge potential biomarkers from 645 Ma dOther evidence has been put forward for early fossil sponges, namely some disputed fossils from the Trezona formation (~640 Ma), and the Namibian Otavia antiqua (~~760 Ma). Those convinced that Otavia is a sponge would need to push concestor 31 back to >760 Ma. In this case, it would not be unreasonable to increase the dates of concestors 27 to 31 by ~ 20%.  However,  dispute any evidence for precambrian sponges. At face value, this potentially reduces the date of concestor 31 to perhaps 555 Ma, and the dates of concestors 26 onwards to within a few tens of millions of years of the cambrian/precambrian boundary.. This suggests a timescale where sponges arose perhaps 650-700 Ma, cnidaria & ctenophores around 600 Ma, and bilaterians around 560 Ma. On the basis of this, a rough timescale is as follows:

Concestor 22 @ 525 Ma (was 530): Human – lamprey/hagfish split. This is no longer constrained as a deep divergence by the fossil record of conodonts .


Concestor 23 or 24

Concestor 23 @ 535 Ma (was 565): Human – tunicate spilt. This must be older than the occurrence of the multiple Haikouichthys fossils and the single specimen of Myllokummingia, both from the Chenjiang biota, and at least 518.5 million years old , probably more like 525 Ma. It has even been suggested that one or other of these are basal vertebrates (which would constrain concestor 22 to >520 Ma, just fitting in to my timescale). In contrast, the exact position of the chordate-like fossils Haikouella and Yunnanozoon (also from the Chenjiang) are now disputed. A conservative placement would have them branch off earlier in the tree, with basal deuterostomes .

Concestor 24 @ 540 Ma (was 560): Human – lancelet split. I have placed this close in time to the previous split, given the previous uncertainty in the ordering of rendezvous 23 and 24 (see my previous post).

Concestor 25 @ 550 Ma (was 570): Human – echinoderm split. Arbitrarily placed halfway between the previous and the next splits, an even placement which is also reflected in .


Concestor 26

Concestor 26 @ 560 Ma (was 590): Human – protostome split. I use this date on the basis of the earliest trace fossils at about 555 Ma , argued to have been caused by coelomate animals (rather than e.g. acoels or cnidarian worms).

Concestor 27 @ 565 Ma (was 630): Human – acoelomorph flatworms.

Concestor 28 @ 600 Ma (was 700): Human – cnidaria split. There are possible cnidarian traces in Ediacaran deposits from 560 Ma , and potential cnidarian embryos in the Doushantuo biota, at approximately 585 Ma give a minimum fossil constraint as 531.5 Ma, and a soft maximum as 581 Ma. A value of 600 Ma treats this maximum as only slightly “soft”. As for molecular clock dates,  give a low estimate of 605 Ma using a rather outdated method. A slightly more recent estimate by the same group gives a range of 709…645 Ma.

Concestor 29 @ 605 Ma (was 750): Human – ctenophore split. Rather arbitrarily placed close to the  cnidarian split on the basis of an uncertain order of concestors 28 & 29.

Concestor 30 @ 620 Ma (was 780): Human – trichoplax split. Trichoplax now has its genome sequenced (Larry Moran has a good discussion). The Bayesian phylogeny in the sequencing paper  has Trichoplax at about 45% of the distance between Cnidaria and sponges. Translating this to my timescales gives (650-600)*0.45+600 = 622.5 Ma.


Concestor 31

Concestor 31 @ 650 Ma (was 800): Human – sponge split. Note that many papers are now starting to suggest or assume sponge paraphyly, with the silicaceous sponges (demosponges & hexactinellida) diverging first (for a contradictory view, see ). The fossil record shows sponge spicules from 630 Ma and the potential sponge biomarker 24-isopropylcholestane in 645 Ma rocks, although both of these are disputed . Even so, these dates are much younger than molecular clock estimates for the divergence time, which are generally >750 Ma. For example, the most recent estimates in  are typically in the range 730-800 Ma. Here I am assuming that the molecular clock dates are indeed too old, but that the evidence of extensive accumulated molecular changes still indicate divergences in the precambrian, perhaps as deep as 100 million years before the cambrian/precambrian boundary.

The figure below provides a graphical representation of my estimates compared to the paleontological constraints and a few molecular clock papers. It should be said that I have chosen the Peterson molecular clock estimates specifically because they are on the low side (compare this to the Timetree of life “expert opinion“ which places the molecular clock dates of concestor 31 at a staggering 1237 Ma).

ATdatesThe data file on which this plot is based is here, and the plot can be generated using the following R code:

dx <- 0.2
RV <- as.numeric(gsub("[^\\d.].*", "", RV, perl=TRUE))
mindate <-as.numeric(gsub("[^\\d.].*", "", Palaeo.notes, perl=TRUE))
from.Benton <- Palaeo.notes == "" | !is.na(mindate)
#by default concestor time is 1 MY more recent than the separation time
d.Sep.Concestor[is.na(d.Sep.Concestor)] <- 1
#by default sep time is 2 MY more recent than av. gene divergence time
d.Sep.Gene[is.na(d.Sep.Gene)] <- 2
plots <- list(list(con=1:15, ylim=c(251,5), mar=c(4, 4, 5, 0) + 0.1, side=c(3,2), 
                   box="7", line=66, names=c("Cenozoic", "Mesozoic")),
              list(con=31:16, ylim=c(1200,251), mar=c(5, 0, 4, 4) + 0.1, side=c(1,4),
                   box="l", line=541, names=c("Proterozoic", "Archaean")))
pdf("ATdates.pdf", 10, 8)
lapply(plots, function(l) {
par(las=3, mar=l$mar)
plot(Anc.Tale.Previous ~ I(RV-dx), pch=16, ylim=l$ylim, xlim=range(l$con), log="y",
     cex=0.5, xlab="", ylab="Date (Ma)", xaxt="n", yaxt="n", bty=l$box)
axis(l$side[1], at=l$con)
abline(v=l$con[-1]+0.5, col="grey85")
abline(h=l$line, col="grey65")
text(min(l$con), l$line * c(0.975,1.025)+c(-3,3), l$names, adj=0.05, col="grey65")
mtext("Concestor", l$side[1], las=1, line=2.5)
segments(RV-dx, Anc.Tale.New, RV-dx, Anc.Tale.New+d.Sep.Concestor, lwd=2, col="red")
segments(RV-dx, Anc.Tale.New+d.Sep.Concestor, RV-dx, Anc.Tale.New+d.Sep.Gene+d.Sep.Concestor,
         lwd=2, lty="11", col="red")
segments(RV, ifelse(is.na(mindate), PalaeoMin, mindate), RV, PalaeoMax, lwd=1,
         col=ifelse(Palaeo.notes=="", "darkgreen", "green"))
segments(RV, mindate, RV, PalaeoMin, lwd=1, lty="11",
         col=ifelse(from.Benton, "darkgreen", "green"))
segments(RV+dx, SpringerMin, RV+dx, SpringerMax, lwd=1, col="blue")
segments(RV+dx-dx/2, SpringerMean, RV+dx, SpringerMean, lwd=1, col="blue")
segments(RV+dx, Springer.AUTOsoft, RV+dx+dx/2, Springer.AUTOsoft, lwd=1, col="blue")
points(RV+dx, dosReis.2012, cex=0.5, col="blue")
segments(RV+dx, Deepfin.Min, RV+dx, Deepfin.Max, lwd=1, col="blue")
points(RV+dx, Deepfin.Mean,pch=4, col="blue", cex=0.5)
points(RV+dx, Peterson.2005.ME, cex=0.6, pch=2, col="blue")
points(RV+dx, Peterson.2005.ML, cex=0.6, pch=6, col="blue")
segments(RV+dx, M1000.Min, RV+dx, M1000.Max, lwd=1, col="blue")
points(RV+dx, Peterson2008.M1000, cex=0.5, pch=23, col="blue", bg="white")
points(RV+dx, Timetree.expert, cex=0.5, pch=16, col="#BB88FF")
#plot legends
legend(15.55,220,c("1st ed",
                   "2nd ed: concestor",
                   "   . . . . : separation",
                   "   . . . . : av. gene"), title="Ancestor's Tale",
       cex=0.75, col=c("black", "red", "red", "red"), pch=c(16,NA, NA, NA),
       bg="white", xpd=TRUE, text.col="grey35", title.col="darkred")
segments(16, 250, 16, 260, col="red", lwd=2)
segments(16, 260, 16, 270, col="red", lwd=2, lty="11")
legend(20.67,220,c("Benton, 2009",
                   "Conservative"), title="Fossil constraints", cex=0.75,
                   bg="white", xpd=TRUE, text.col="grey35", title.col="darkgreen")
segments(21.2, c(234,246), 21.2, c(244, 254), col=c("darkgreen", "green"))
segments(21.2, 257, 21.2, 265, col="darkgreen", lty="11")
legend(24.8,220,c("Springer, 2012: summary",
                  "           . . . . . . : AUTOsoft",
                  "dos Reis, 2012",
                  "Broughton, 2013",
                  "Peterson, 2005: ME",
                  "            . . . . . . : ML",
                  "Peterson, 2008: M1000",
                  "Timetree 'expert'"), title="Molecular clock",
                  pch=c(NA, NA, 1,4,2,6, 23, 16), cex=0.75, bg="white",
                  xpd=TRUE, text.col="grey35", title.col="blue")
px <- 25.23
segments(c(px, px, px, px, px-dx/2, px+dx/2), c(233,266, 301, 308, 240,247),
         px, c(252,275,304, 311, 240,247), col="blue")


Notes   [ + ]

a. I have switch to the standardised convention of using Ma to mean “millions of years ago”
b. More specifically, 0.44…0.63 × 10-9 by “direct” measurement (subject to errors in different methods of genome sequencing of modern humans), or 0.43×10-9 by inference from similar demographic patterns across the genome between modern and ancient humans.
c. estimate this as more like 3-4 million years
d. Other evidence has been put forward for early fossil sponges, namely some disputed fossils from the Trezona formation (~640 Ma), and the Namibian Otavia antiqua (~~760 Ma). Those convinced that Otavia is a sponge would need to push concestor 31 back to >760 Ma. In this case, it would not be unreasonable to increase the dates of concestors 27 to 31 by ~ 20%.  However,  dispute any evidence for precambrian sponges. At face value, this potentially reduces the date of concestor 31 to perhaps 555 Ma, and the dates of concestors 26 onwards to within a few tens of millions of years of the cambrian/precambrian boundary.

7 thoughts on “A pragmatic redating of our 700 million year ancestry

  1. For the next 5 or 6 rendezvous points, a very rough scale can be deduced from fig 1 of the Filozoa paper (http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0002098). Calibrated with the fungal divergence at 1000 Ma, as suggested by Fig 1 in Peterson et al (2008), and using the sponge divergence as above (650Ma), this gives to the nearest 50 MY:
    * 750 Ma for the animal/choanoflagellate divergence
    * 850 for the animal/filasterea divergence
    * 900 for animal/DRIPs, and
    * 1000 for animal/Fungi.
    * 1100 for animal/Amoebozoa.

    On the other hand, Fig 2 in Parfrey et al (2011), suggests a fungal divergence of 1300 Ma, and the following approximate dates
    * 900 Ma for the animal/choanoflagellate divergence
    * 1100 for the animal/Filasterea and animal/DRIPs, so if the branches are separate, perhaps 1050 for Filasterea and 1150 for DRIPs.
    * 1300 for animal/Fungi.
    * 1800 for animal/Amoebozoa (although this is where the tree is rooted).

    In a pragmatic attempt to reconcile these dates, I suspect that a date of 1200 for Fungi is reasonable (certainly compatible with the opisthokont dates in Fig 1B of the Parfrey paper), leading to the following dates: RV32:800; RV33:900; RV34:1000; RV35:1200; RV36:?; RV37:1300.

  2. I’m rather obsessed with 25 & 26, would we not find a hemichordate or Xenoturbella hanging around this point ?

    • Hi – the position of Xenoturbella is still very unclear. I’ve discussed this with Jordi Paps, and provisionally left it where Richard Dawkins and I originally had it as a sister group to the echinoderms and the hemichordates, all grouped together into the “Ambulacraria”. When writing above, I (imprecisely) used “echinoderms” as shorthand for this group, hence rendezvous 25 in my scheme above includes acorn worms (hemichordates) as well as Xenoturbella. But I wouldn’t be surprised if Xenotubella moved in the future.

  3. Very neat post!

    We’ve recently done an analysis of the Metazoa but with the twist that dates were estimated under four different interpretations of the fossil record, so you could look at this if you are curious. I think the point we want to stress in our paper is that levels of uncertainty at such deep divergences are very large. http://dx.doi.org/10.1016/j.cub.2015.09.066

    Regarding the divergence of apes and human, we just published an analysis of neutrally evolving loci under a hierarchical Bayesian model under the coalescent. When calibrating the neutral loci with Benton’s et al. proposed fossil calibration of 10–5.7 Ma, we get an estimate for the neutral mutation rate of 0.41–0.6 x 10^-9/y, which is surprisingly close to Scally and Durbin’s 0.4–0.6 estimate. On the other hand, if we calibrate the phylogeny with the 0.4–0.6 mutation rate estimate, the divergence for human-chimp is then 9.9–5.8 Ma. So I think phylogenetic and experimental mutation rate estimates are converging (see table 4 in http://bit.ly/humanchimp). We also discuss at length the problems of ignoring gene coalescent times when estimating species divergence dates.

    • I’m honoured that you posted on this blog, thanks for that. Yes, I’m aware of your recent paper, and recently recommended it to the DateLife project. I do agree with your assessment of the degree of uncertainty – there are various extra sources of uncertainty that are difficult to incorporate into a full Bayesian analysis (e.g. assignment of fossils to ancient groups, sequence alignment errors, the possibility of correlated changes in substitution rates as a result of ecological or environmental factors. I think there’s still a long way to go before the molecular clock uncertainties are properly bounded.

      Very interested in your ape analysis, and I hadn’t seen the paper, so thanks for the link. I absolutely agree about coalescent vs ‘species’ divergences (it’s also not entirely clear to me what a ‘species’ divergence is, either, but that’s another argument). I think there might be some mileage in incorporating estimates of generation time into the analysis: the assumption is usually made that because wild chimps & hunter gatherer societies have similar generation times (25 & 30 years), then these can be used in clock calibrations. But we have other sources of speed of reproductive maturity of hominid fossils, and it would be interesting to incorporate those into a Bayesian framework.


Leave a Reply

Your email address will not be published. Required fields are marked *