RESEARCH ARTICLE METHODS Burtele tuff.The plagioclase weighted-mean age of sample WM10/B-1 is 40Ar/39Ar dating procedures follow those of ref.2.The mineral separates were predictably less precise than either K-feldspar age,due to the lower potassium irradiated for 5h in two separate batches in the in-core CLICIT facility of the content,but is nevertheless a reasonable result (3.420.03 Myr)that is not Oregon State University TRIGA reactor.Sanidine from the Fish Canyon Tuff of statistically different from the K-feldspar age. Colorado was used as a monitor mineral,with an age of 28.201 Myr2.After irra- The Burtele tuff at the base of the section is dated by the4Ar/Ar method to diation,the feldspar grains were individually analysed under ultra-high vacuum on 3.469+0.008 Myr ago (analytical data are given in Supplementary Information) a MAP 215 Noble-gas mass spectrometer,using a focused CO2 laser as the heating and lies a maximum of about 27m below BRT-VP-2/73,providing a firm device.In all,56 grains were analysed from the two samples(Supplementary Table maximum age constraint of~3.47 Myr ago for the foot specimen.An approximate 7).Most grains(34)provedtobe K-feldspar,asjudgedby theCa/K ratiodetermined age for the foot specimen can be estimated using regional sediment accumulation from the measured argon isotopes,whereas the remainder were relatively low-Ca/K rates.The average rate for older WORMIL strata in the Waki-Mille confluence area plagioclase.Most analyses yielded the anticipated high proportion of radiogenic is 11 cmkyr(ref.2),which yields an estimated age of 3.22 Myr ago for the BRT- r relative to atmosphericAr contamination expected for unaltered feldspars VP-2/73 specimen.This rate is much lower than estimates for the Sidi Hakoma from Pliocene volcanic rocks,but a few exhibited anomalously low radiogenic Member of the Hadar Formation2-,which is closer in age to the BRT ridge content,and were excluded from further analysis;an arbitrary cutoff of 60% section,but is much farther away geographically.Using a Sidi Hakoma accumula- Arwas used,identifying four grains for exclusion.In addition,as is typical for tion rate of 30 cmkyr yields an estimate of 3.38 Myr for BRT-VP-2/73.These East African tephra,a slight tail of the age distribution towards older ages was contrasting rates suggest an age of between 3.2 and 3.4 Myr ago for BRT-VP-2/73. observed.A statistical filter was applied to the sample distributions,using a median For the isotopic analysis of pedogenic carbonate,carbonate nodules were outlier determinant(outliers were classified as falling 1.5 normalized median abso- sampled from peds with slickenside surfaces and clay cutans,within a distinct lute deviations'from the median).Use of this criterion identified three outliers in pedogenic carbonate zone,250 cm below the palaeosol contact with the over- each of WM07/B-1 (K-feldspar)and WM10/B-1 (Plagioclase).The remaining lying silt.C.O and A measurements of carbonate were made using an populations yield simple,unimodal Gaussian-like distributions(Supplementary automated common acid bath peripheral coupled to a Thermo MAT 253 mass Fig.7).Weighted-mean sample ages of the K-feldspar populations from samples spectrometer at Johns Hopkins University,using methods described previously2? WM07/B-1 and WM10/B-I are 3.4840.011 Myr(n 24;1o analytical error, The results are reported in Supplementary Table 8. incorporating error in /the neutron fluence parameter of0.2%)and 3.453+0.011 Myr (n=4),respectively (Supplementary Table 4).An overall weighted-mean of 28.Kuiper,K.F.et al Synchronizing rock clocks of earth history.Science 320, the two K-feldspar ages is 3.469+0.008 Myr,taken as the reference age for the 500-504(2008). 2012 Macmillan Publishers Limited.All rights reservedMETHODS 40Ar/39Ar dating procedures follow those of ref. 2. The mineral separates were irradiated for 5 h in two separate batches in the in-core CLICIT facility of the Oregon State University TRIGA reactor. Sanidine from the Fish Canyon Tuff of Colorado was used as a monitor mineral, with an age of 28.201Myr28. After irra￾diation, the feldspar grains were individually analysed under ultra-high vacuum on a MAP 215 Noble-gas mass spectrometer, using a focused CO2 laser as the heating device. In all, 56 grains were analysed from the two samples (Supplementary Table 7).Most grains (34) proved to be K-feldspar, as judged by the Ca/K ratio determined from the measured argon isotopes, whereas the remainder were relatively low-Ca/K plagioclase. Most analyses yielded the anticipated high proportion of radiogenic 40Ar relative to atmospheric 40Ar contamination expected for unaltered feldspars from Pliocene volcanic rocks, but a few exhibited anomalously low radiogenic content, and were excluded from further analysis; an arbitrary cutoff of 60% 40Ar* was used, identifying four grains for exclusion. In addition, as is typical for East African tephra, a slight tail of the age distribution towards older ages was observed. A statistical filter was applied to the sample distributions, using a median outlier determinant (outliers were classified as falling 1.5 ‘normalized median abso￾lute deviations’ from the median). Use of this criterion identified three outliers in each of WM07/B-1 (K-feldspar) and WM10/B-1 (Plagioclase). The remaining populations yield simple, unimodal Gaussian-like distributions (Supplementary Fig. 7). Weighted-mean sample ages of the K-feldspar populations from samples WM07/B-1 and WM10/B-1 are 3.484 6 0.011 Myr (n 5 24; 1s analytical error, incorporating error in J, the neutron fluence parameter of 0.2%) and 3.453 6 0.011 Myr (n 5 4), respectively (Supplementary Table 4). An overall weighted-mean of the two K-feldspar ages is 3.469 6 0.008Myr, taken as the reference age for the Burtele tuff. The plagioclase weighted-mean age of sample WM10/B-1 is predictably less precise than either K-feldspar age, due to the lower potassium content, but is nevertheless a reasonable result (3.42 6 0.03 Myr) that is not statistically different from the K-feldspar age. The Burtele tuff at the base of the section is dated by the 40Ar/39Ar method to 3.469 6 0.008 Myr ago (analytical data are given in Supplementary Information) and lies a maximum of about 27 m below BRT-VP-2/73, providing a firm maximum age constraint of ,3.47Myr ago for the foot specimen. An approximate age for the foot specimen can be estimated using regional sediment accumulation rates. The average rate for olderWORMIL strata in the Waki-Mille confluence area is 11 cm kyr21 (ref. 2), which yields an estimated age of 3.22Myr ago for the BRT￾VP-2/73 specimen. This rate is much lower than estimates for the Sidi Hakoma Member of the Hadar Formation24–26, which is closer in age to the BRT ridge section, but is much farther away geographically. Using a Sidi Hakoma accumula￾tion rate of 30 cm kyr21 yields an estimate of 3.38Myr for BRT-VP-2/73. These contrasting rates suggest an age of between 3.2 and 3.4 Myr ago for BRT-VP-2/73. For the isotopic analysis of pedogenic carbonate, carbonate nodules were sampled from peds with slickenside surfaces and clay cutans, within a distinct pedogenic carbonate zone, $50 cm below the palaeosol contact with the over￾lying silt. d13C, d18O and D47 measurements of carbonate were made using an automated common acid bath peripheral coupled to a Thermo MAT 253 mass spectrometer at Johns Hopkins University, using methods described previously27. The results are reported in Supplementary Table 8. 28. Kuiper, K. F. et al. Synchronizing rock clocks of earth history. Science 320, 500–504 (2008). RESEARCH ARTICLE ©2012 Macmillan Publishers Limited. All rights reserved