Statistical Estimates of Hominin Origination and Extinction Dates: a Case Study Examining the Australopithecus Anamensiseafarensis Lineage

Home , Australopithecus, Middle Awash, Mille (woreda), Zeresenay Alemseged

Journal of Human Evolution 138 (2020) 102688

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jhevol

Statistical estimates of hominin origination and extinction dates: A case study examining the Australopithecus anamensiseafarensis lineage

* Andrew Du a, b, , John Rowan c, Steve C. Wang d, Bernard A. Wood e, Zeresenay Alemseged b a Department of Anthropology and Geography, Colorado State University, Fort Collins, CO 80523, USA b Department of Organismal Biology and Anatomy, The University of Chicago, Chicago, IL 60637, USA c Organismic and Evolutionary Biology, University of Massachusetts Amherst, Amherst, MA 01003, USA d Department of Mathematics and Statistics, Swarthmore College, Swarthmore, PA 19081, USA e Center for the Advanced Study of Human Paleobiology, Department of Anthropology, The George Washington University, 800 22nd St., NW, Washington, DC 20052, USA article info abstract

Article history: Reliable estimates of when hominin taxa originated and went extinct are central to addressing many Received 5 October 2018 paleoanthropological questions, including those relating to macroevolutionary patterns. The timing of Accepted 9 October 2019 hominin temporal ranges can be used to test chronological predictions generated from phylogenetic Available online xxx hypotheses. For example, hypotheses of phyletic ancestoredescendant relationships, based on morphological data, predict no temporal range overlap between the two taxa. However, a fossil taxon's Keywords: observed temporal range is almost certainly underestimated due to the incompleteness of both the fossil Confidence intervals record itself and its sampling, and this decreases the likelihood of observing temporal overlap. Here, we Origination e Extinction focus on a well-known and widely accepted early hominin lineage, Australopithecus anamensis afarensis, fi Ancestoredescendant and place 95% con dence intervals (CIs) on its origination and extinction dates. We do so to assess Australopithecus anamensis whether its temporal range is consistent with it being a phyletic descendant of Ardipithecus ramidus and/ Australopithecus afarensis or a direct ancestor to the earliest claimed representative of Homo (i.e., Ledi-Geraru). We find that the last appearance of Ar. ramidus falls within the origination CI of Au. anamensiseafarensis, whereas the claimed first appearance of Homo postdates the extinction CI. These results are consistent with Homo evolving from Au. anamensiseafarensis, but temporal overlap between Ar. ramidus and Au. anamensiseafarensis cannot be rejected at this time. Though additional samples are needed, future research should extend our initial analyses to incorporate the uncertainties surrounding the range endpoints of Ar. ramidus and earliest Homo. Overall, our findings demonstrate the need for quantifying the uncertainty surrounding the appearances and disappearances of hominin taxa in order to better understand the timing of evolutionary events in our clade's history. © 2019 Elsevier Ltd. All rights reserved.

1. Introduction at around 2.8 million years ago (Ma) has been implicated in epi- sodes of late Pliocene faunal turnover that included the origin of the Reliable estimates of when hominin taxa originated and went genus Homo (deMenocal, 2004, 2011). Accurate origination and extinct are of critical importance in human origins research. The extinction dates are also needed to properly time-scale hominin timing of external events is often linked with the appearances and phylogenetic trees (e.g., Gomez-Robles et al., 2017) and to infer disappearances of hominin taxa, potentially suggesting a causal evolutionary relationships. For instance, sister taxa should have the relationship between the two. For example, global climate change same origination dates (Paul, 1982; Marshall, 1990), and species proposed to be phyletic ancestors and descendants should have non-overlapping temporal ranges (Paul, 1982, 1992).

* Corresponding author. Despite the longstanding practice of portraying observed homi- E-mail address: [email protected] (A. Du). nin temporal ranges as reality, these data are almost certainly https://doi.org/10.1016/j.jhevol.2019.102688 0047-2484/© 2019 Elsevier Ltd. All rights reserved. 2 A. Du et al. / Journal of Human Evolution 138 (2020) 102688 underestimates due to the incompleteness of both the fossil record (4.15e3.04 Ma using age midpoints; see section 2.2) does not overlap itself and its sampling (Marshall, 2010; Wang and Marshall, 2016; the LAD of Ar. ramidus (4.493e4.300 Ma [Simpson et al., 2019]) or the Wood and Boyle, 2016). Yet, few researchers have statistically esti- FAD of the currently unnamed Homo species from Ledi-Geraru mated hominin origination and extinction dates and quantified the (2.80e2.75 Ma [DiMaggio et al., 2015; Villmoare et al., 2015]), uncertainty surrounding them. Bobe and Leakey (2009) and Bobe which we hereafter refer to as “earliest Homo”. As mentioned pre- and Carvalho (2019) calculated confidence intervals (CIs) for the viously, however, the temporal range of Au. anamensiseafarensis is appearances and disappearances of various hominin taxa (i.e., Par- most likely underestimated. Therefore, based on the current homi- anthropus aethiopicus, P. boisei, Homo habilis sensu stricto, nin fossil record, one cannot rule out that Au. anamensiseafarensis H. rudolfensis, and H. erectus) in the Omo-Turkana Basin of northern did in fact overlap with Ar. ramidus or earliest Homo, which obfus- Kenya and southern Ethiopia. However, the method they use (Koch, cates our understanding of how Au. anamensiseafarensis may have 1987) requires extensive data in the form of number of fossil speci- arisen from or given rise to these species, respectively. mens, not only for the hominin taxon in question but also for the Here, we use the method developed by Strauss and Sadler greater mammalian record. Consequently, this method can be used (1989) to estimate the origination and extinction dates of Au. only for sites where large databases are available (e.g., the Turkana anamensiseafarensis as a way of evaluating whether this species Database [Bobe, 2011]), and it provides CIs for local first and last temporally overlapped with Ar. ramidus and/or earliest Homo. This appearance data (FAD and LAD, respectively), not the taxon's global method is based on the principle that the observed distribution of origination and extinction dates. Maxwell (2018) recently placed CIs fossil horizons and the temporal gaps between them contain in- on the temporal ranges of 13 hominin taxa ranging from the Miocene formation on the gaps between a taxon's FAD and origination date to the Pleistocene, though he acknowledges that the assumptions of and its LAD and extinction date (Paul, 1982). We place 95% confi- his chosen method (Strauss and Sadler, 1989) were violated, thus dence intervals on these estimates to determine whether the LAD rendering his estimated CIs potentially suspect. of Ar. ramidus and the FAD of earliest Homo significantly predate In an attempt to remedy the paucity of estimated dates and CIs and postdate, respectively, the temporal range of Au. ana- for hominin originations and extinctions, we use the Austral- mensiseafarensis. Our analyses offer an initial assessment, based on opithecus anamensiseafarensis lineage (Kimbel et al., 2006; Kimbel temporal grounds, of whether Au. anamensiseafarensis (1) phylet- and Delezene, 2009) from the Pliocene of eastern Africa as a test ically descended from Ar. ramidus and/or (2) gave rise to earliest case. We chose this lineage for two reasons. First, within the Homo via bifurcating cladogenesis. Future work should incorporate context of other early hominins, its fossil record is well sampled, the uncertainties surrounding the LAD and FAD of Ar. ramidus and documented, and studied. The hypodigm comprises over 400 earliest Homo, respectively (see section 4.44.4). specimens, which are well distributed throughout the species' temporal range (Kimbel et al., 2006; Kimbel and Delezene, 2009). 2. Materials and methods Second, the Au. anamensiseafarensis lineage is widely accepted to have been a probable phylogenetic link between early and late 2.1. Data collection Pliocene hominins (White et al., 2006, 2009b, 2015; Villmoare et al., 2015), so more reliable information about when this taxon origi- Information on the temporal distribution of Au. ana- nated and went extinct should help test the phylogenetic hypoth- mensiseafarensis came from fossil horizons, each of which is eses we explore in more detail in the following paragraph. denoted by the presence of at least one identifiable specimen for a Previous phylogenetic hypotheses have proposed an anagenetic given time period. We collected data from the published literature relationship between Ardipithecus ramidus and Au. ana- on both the lower and upper bracketing ages for these horizons at mensiseafarensis (one of three hypotheses proposed by White et al., the finest temporal resolution possible (Table 1). 2009b) and a scenario where Au. anamensiseafarensis cladogeneti- The Strauss and Sadler method (1989) makes the implicit cally bifurcates into Homo and Paranthropus (Wood, 1992). These assumption that depositional environments and sediments hypotheses and modes of speciation generate clear expectations as conducive to the preservation of hominin fossils can be found just to how ancestor and descendant temporal ranges should relate to outside the observed temporal range of Au. anamensiseafarensis each other: specifically, the ranges should not overlap (Fig.1A and B). because otherwise the estimated CIs will be too small (Marshall, Currently, the known temporal range of Au. anamensiseafarensis 1990) (see section 2.4). In order to test this assumption, we collected data from Faith et al. (2018) on the ages and species lists for terrestrial large mammal sites from eastern Africa between 5 Ma to 2 Ma (Table 2). We focused on sites that preserve large-bodied primates (hominins or papionins), which serves as a taphonomic indicator of conditions conducive to the preservation of hominin fossils (Campisano et al., 2004; Maxwell et al., 2018).

2.2. Treatment of fossil horizon ages

We estimated the age of each fossil horizon using the midpoint (i.e., [lower bracketing age þ upper bracketing age]/2), which is equivalent to the mean of the bracketing ages. Using the age midpoint is common practice in paleobiological studies investi- gating fossil patterns through time (e.g., Behrensmeyer et al., 1997; Figure 1. Different modes of speciation. Temporal ranges are expected to not overlap if Alroy et al., 2000; Barry et al., 2002; Hone et al., 2005; Novack- a descendant arose from an ancestor via A) anagenesis or B) bifurcating cladogenesis, Gottshall and Lanier, 2008; Tomiya, 2013; Bibi and Kiessling, whereas ranges are expected to overlap if C) budding cladogenesis is the mode of 2015; Smith and Marcot, 2015; Kaya et al., 2018; Maxwell et al., speciation. “A” and “D” represent the ancestor and descendant lineage, respectively, 2018). Absent any information on a fossil's location between two and “D1” and “D2” denote two different descendant lineages. The x-axis is arbitrary and not meant to signify morphological change. This figure is modified after figure 1 in radiometrically dated beds, the uncertainty surrounding the fossil's Foote (1996). stratigraphic position is uniformly distributed between the two A. Du et al. / Journal of Human Evolution 138 (2020) 102688 3

Table 1 Site, stratigraphic, and age data for Australopithecus anamensiseafarensis fossil horizons.a

Site Country Stratigraphic location Lower age Upper age Midpoint age References (Ma) (Ma) (Ma)b

Hadar Ethiopia BKT-1 complex to BKT-2 complex 3.120 2.950 3.035 Campisano and Feibel (2008) Usno Ethiopia U-12 NAc NA 3.050 Feibel et al. (1989) Hadar Ethiopia Kada Hadar Tuff to BKT-1 complex 3.200 3.120 3.160 Campisano and Feibel (2008) Hadar Ethiopia Triple Tuff 4 to Kada Hadar Tuff 3.240 3.200 3.220 Campisano and Feibel (2008) Woranso- Ethiopia 3.330 3.207 3.269 Haile-Selassie et al. (2016) Mille Hadar Ethiopia KMB to Triple Tuff 4 3.300 3.240 3.270 Campisano and Feibel (2008) Hadar Ethiopia Kada Me'e Tuff to KMB 3.360 3.300 3.330 Campisano and Feibel (2008) Dikika Ethiopia Sidi Hakoma Tuff to Sidi Hakoma graptolite 3.440 3.280 3.360 McDougall and Brown (2008) and Wynn et al. (2006) Koobi Fora Kenya Tulu Bor Tuff to Toroto Tuff 3.440 3.310 3.375 Kimbel (1988) and McDougall and Brown (2008) Hadar Ethiopia Sidi Hakoma Tuff to Kada Me'e Tuff 3.440 3.360 3.400 Campisano and Feibel (2008) and McDougall and Brown (2008) Dikika Ethiopia Basal Member 3.450 3.400 3.425 Alemseged et al. (2005) and Campisano and Feibel (2008) Maka Ethiopia Right below tuff, which is correlated with Sidi NA NA 3.440 McDougall and Brown (2008) and White et al. Hakoma Tuff (1993) Kantis Kenya 3.600 3.300 3.450 Mbua et al. (2016) Bahr-el- Chad NA NA 3.580 Lebatard et al. (2008) Ghazal Woranso- Ethiopia 3.600 3.580 3.590 Haile-Selassie et al. (2010) Mille Laetoli Tanzania Tuff 8 to Yellow Marker Tuff 3.630 3.630 3.630 Harrison (2011a) Laetoli Tanzania Tuff 7 to Tuff 8 3.660 3.630 3.645 Harrison (2011a) Woranso- Ethiopia Mille Tuff to Kilaytoli Tuff 3.760 3.570 3.665 Deino et al. (2010) Mille Laetoli Tanzania Tuff 6 to Tuff 7 3.700 3.660 3.680 Harrison (2011a) Laetoli Tanzania Tuff 5 to Tuff 6 3.790 3.700 3.745 Harrison (2011a) Laetoli Tanzania Tuff 3 to Tuff 4 3.800 3.790 3.795 Harrison (2011a) Laetoli Tanzania Tuff 1 to Tuff 2 3.830 3.810 3.820 Harrison (2011a) Galili Ethiopia Shabeley Laag Member 4.100 3.600 3.850 Kullmer et al. (2008) Belohdelie Ethiopia 3.890 3.860 3.875 White et al. (1993) Allia Bay Kenya in Moiti Tuff NA NA 3.970 Leakey et al. (1995) and McDougall and Brown (2008) Allia Bay Kenya below Moiti Tuff NA NA 4.000 Cofﬁng et al. (1994) and Leakey et al. (1995) Fejej Ethiopia 4.18 4 4.090 Kappelman et al. (1996) Kanapoi Kenya Upper pumiceous tuff to Kanapoi tuff 4.12 4.07 4.095 Leakey et al. (1998) Kanapoi Kenya Lower pumiceous tuff to Upper pumiceous tuff 4.17 4.12 4.145 Leakey et al. (1998) Asa Issie Ethiopia 4.2 4.1 4.15 White et al. (2006)

a Stratigraphic and age data were collected at the finest resolution possible. b Fossil horizons are organized by midpoint age from youngest to oldest. c NA lower and upper ages represent those hominin horizons where only one age estimate was provided. Additional notes can be found in the SOM data file. beds. The expected value of a uniformly distributed random vari- extinction date from the LAD (the white gaps in Fig. 2A and B). able is the midpoint, so this is the best possible age estimate. Because the n fossil horizons are uniformly distributed over the However, estimation of origination and extinction dates requires species' range, these gaps are on average the same size as any other age information on the FAD and LAD only (see section 2.5), so we gap between adjacent fossil horizons. Therefore, the best estimate for also ran our analyses assuming an extreme scenario where the FAD the size of these gaps is the expected value of the age gap distribution, and LAD correspond to the lower and upper bracketing ages for i.e., the mean of all age gaps between the species' FAD and LAD (we these dates, respectively. demonstrate this mathematically in section 2.5). Note that if a species has a smaller number of sampled fossil horizons, this translates to a 2.3. The Strauss and Sadler (1989) method larger mean age gap (Fig. 2C), which offsets the larger gaps between the species' FAD and origination date and its LAD and extinction date The Strauss and Sadler method assumes that the temporal dis- (compare Fig. 2A and B). By estimating the gaps between observed tribution of fossil horizons can be modeled using a stochastic and true range endpoints in this way, we arrive at a point estimate Poisson process. In a Poisson process, points are randomly placed (i.e., a single value) for the true origination and extinction dates. on some mathematical space (Kingman, 1993), which in our case is Confidence intervals for the true origination and extinction dates a one-dimensional timeline. A property of this type of Poisson provide a margin of error on our estimates (Strauss and Sadler, 1989; process is that points are uniformly and independently distributed see also Gingerich and Uhen, 1998; Marshall, 1990). These CIs for through time, i.e., the points are distributed as if they are inde- origination or extinction dates are interpreted in the same way as any pendently sampled from a uniform distribution, bounded by the frequentist confidence interval: if one randomly samples n horizons ends of the timeline (Kingman, 1993). It then follows that the from a species' range many times and calculates 95% CIs for the probability of sampling a point is constant throughout the timeline. range's true endpoint each time, on average 95% of those CIs will In the case of a species' temporal range, n fossil horizons are capture the true endpoint, and 5% will miss it (i.e., the Type I error randomly placed within the species' range according to a Poisson rate is 5%; Fig. 3A). Each CI is one out of many possible outcomes process, thereby producing n þ 1gaps(Fig. 2A and B). The main age (sensu Fig. 3A), and any single CI either includes the true endpoint or gaps of interest for estimating origination and extinction dates are it does not. The CIs will shrink as more fossil horizons are discovered the ones separating the origination date from the FAD and the (Fig. 3B) because a larger sample size decreases the sampling error 4 A. Du et al. / Journal of Human Evolution 138 (2020) 102688

Table 2 Site, stratigraphic, age, and presence of large-bodied primates data for 5e2 Ma terrestrial large mammal sites in eastern Africa.a

Site Country Stratigraphic location Lower age Upper age Midpoint age Large Age references Faunal references (Ma) (Ma) (Ma)b primates?

Hadar (Danauli) Ethiopia Busidima Formation NAc NA 2.000 Y Rowan et al. (in prep.) Geraads et al. (2012, 2013b), Reed (2008), Rowan et al. (in prep.), and Werdelin and Lewis (2013) Omo (Shungura) Ethiopia Member G 2.270 1.870 2.070 Y McDougall et al. (2012) Geraads (2010), Werdelin and Lewis (2013), White and Suwa (2004), and Omo French & American Catalogs West Turkana Kenya Kalochoro 2.330 1.870 2.100 Y McDougall et al. (2012) Fortelius et al. (2016) and Werdelin and Lewis (2013) Omo (Shungura) Ethiopia Member F 2.320 2.270 2.300 Y McDougall et al. (2012) Geraads (2010), White and Suwa (2004), and Omo French & American Catalogs Hadar Ethiopia Busidima Formation NA NA 2.350 Y Kimbel et al. (1996) and Geraads et al. (2012, 2013b), (Maka'amitalu) Campisano (2007) Reed (2008), Rowan et al. (in prep.), and Werdelin and Lewis (2013) Omo (Shungura) Ethiopia Member E 2.400 2.320 2.360 Y McDougall et al. (2012) Geraads (2010), White and Suwa (2004), and Omo French & American Catalogs West Turkana Kenya Lokalalei 2.530 2.330 2.430 Y McDougall et al. (2012) Fortelius et al. (2016) and Geraads (2010) Omo (Shungura) Ethiopia Member D 2.530 2.440 2.490 Y McDougall et al. (2012) Geraads (2010), White and Suwa (2004), and Omo French & American Catalogs Middle Awash Ethiopia Hata 2.500 2.500 2.500 Y de Heinzelin et al. (1999) Bibi et al. (2017), Boisserie and (Bouri) White (2004), de Heinzelin et al. (1999), and White and Suwa (2004) Ledi-Geraru (Lee Ethiopia 2.670 2.580 2.630 Y DiMaggio et al. (2015) Bibi et al. (2017), DiMaggio et al. Adoyta) (2015), Lazagabaster et al. (2018), and Rowan et al. (2017) Laetoli Tanzania Upper Ndolanya 2.660 2.660 2.660 Y Deino (2011) Harrison (2011b) West Turkana Kenya upper Lomekwi 2.820 2.530 2.680 Y McDougall et al. (2012) Fortelius et al. (2016) and Werdelin and Lewis (2013) Ledi Geraru Ethiopia 2.820 2.750 2.790 Y DiMaggio et al. (2015) Bibi et al. (2017), DiMaggio et al. (Gurumaha) (2015), Lazagabaster et al. (2018), and Rowan et al. (2017) Omo (Shungura) Ethiopia Member C 3.070 2.530 2.800 Y McDougall et al. (2012) Geraads (2010), Sanders et al. (2010), Werdelin and Lewis (2013), and Omo French & American Catalogs West Turkana Kenya middle Lomekwi 3.130 2.820 2.980 Y Interpolation (Division of Fortelius et al. (2016), Geraads member into three units) (2010), and Werdelin and Lewis (2013) Hadar Ethiopia Kada Hadar 2 3.120 2.950 3.035 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), Rowan et al. (in prep.), and Werdelin and Lewis (2013) Koobi Fora Kenya Tulu Bor 3.440 2.640 3.040 Y McDougall et al. (2012) Fortelius et al. (2016), Werdelin and Lewis (2013), and Weston and Boisserie (2010) Hadar Ethiopia Kada Hadar 1 3.200 3.120 3.160 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), Rowan et al. (in prep.), and Werdelin and Lewis (2013) Hadar Ethiopia Denen Dora 3 NA NA 3.200 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), and Werdelin and Lewis (2013) Omo (Shungura) Ethiopia Member B 3.440 2.970 3.210 Y McDougall et al. (2012) Geraads (2010), Sanders et al. (2010), Werdelin and Lewis (2013), and Omo French & American Catalogs Hadar Ethiopia Denen Dora 2 NA NA 3.250 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), and Werdelin and Lewis (2013) Lothagam Kenya Kaiyumung 3.500 3.000 3.250 Y Sanders et al. (2010) Boisserie et al. (2014), Geraads (2010), Leakey and Harris (2003), and Werdelin and Peigne (2010) Hadar Ethiopia Sidi Hakoma 4 to NA NA 3.250 Y Reed (2008) Geraads et al. (2012, 2013b), Denen Dora 1 Reed (2008), and Werdelin and Lewis (2013) A. Du et al. / Journal of Human Evolution 138 (2020) 102688 5

Table 2 (continued )

Site Country Stratigraphic location Lower age Upper age Midpoint age Large Age references Faunal references (Ma) (Ma) (Ma)b primates?

West Turkana Kenya lower Lomekwi 3.440 3.130 3.290 Y McDougall et al. (2012) Fortelius et al. (2016), Geraads (2010), and Werdelin and Lewis (2013) Hadar Ethiopia Sidi Hakoma 3 NA NA 3.300 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), and Werdelin and Lewis (2013) Hadar Ethiopia Sidi Hakoma 2 NA NA 3.350 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), and Werdelin and Lewis (2013) Hadar Ethiopia Sidi Hakoma 1 NA NA 3.400 Y Reed (2008) Geraads et al. (2012, 2013b), Reed (2008), and Werdelin and Lewis (2013) Kantis Kenya 3.500 3.400 3.450 Y Mbua et al. (2016) Mbua et al. (2016) Koobi Fora Kenya Lokochot 3.600 3.440 3.520 Y McDougall et al. (2012) Fortelius et al. (2016), Werdelin and Lewis (2013), and Weston and Boisserie (2010) Omo (Shungura) Ethiopia Member A 3.600 3.440 3.520 Y McDougall et al. (2012) Sanders et al. (2010) and Omo French & American Catalogs Woranso-Mille Ethiopia 3.570 3.760 3.665 Y Curran and Haile-Selassie Curran and Haile-Selassie (Aralee Issie) (2016) (2016) Woranso-Mille Ethiopia 3.570 3.760 3.665 Y Curran and Haile-Selassie Curran and Haile-Selassie (Mesgid Dora) (2016) (2016) West Turkana Kenya Kataboi 3.970 3.440 3.710 Y McDougall et al. (2012) Fortelius et al. (2016) Laetoli Tanzania Upper Laetolil 3.850 3.600 3.730 Y Deino (2011) Harrison (2011b) Koobi Fora Kenya Moiti 3.970 3.600 3.790 N McDougall et al. (2012) Fortelius et al. (2016) Koobi Fora Kenya Lonyumun 4.000 3.970 3.990 Y McDougall et al. (2012) Fortelius et al. (2016), Werdelin and Lewis (2013), and Weston and Boisserie (2010) Omo (Mursi) Ethiopia NA NA 4.000 N Drapeau et al. (2014) Drapeau et al. (2014) and Geraads et al. (2015) Laetoli Tanzania Lower Laetolil 4.360 3.850 4.110 N Deino (2011) Harrison (2011b) Asa Issie Ethiopia NA NA 4.120 Y White et al. (2006) White et al. (2006) Kanapoi Kenya Kanapoi 4.170 4.070 4.120 Y Feibel (2003) Geraads (2010), Geraads et al. (2013a), Harris et al. (2003), Sanders et al. (2010), and Werdelin and Lewis (2013) Manonga Valley Tanzania Kiloleli 4.500 4.000 4.250 N Werdelin (2010) Harrison and Baker (1997) and Sanders et al. (2010) Aramis Ethiopia Aramis 4.400 4.400 4.400 Y WoldeGabriel et al. (2009) Bernor et al. (2013), García (2008), Sanders et al. (2010), and White et al., 2009a Lothagam Kenya Apak 5.000 4.200 4.600 Y Sanders et al. (2010) Boisserie et al. (2014), Geraads (2010), Leakey and Harris (2003), and Werdelin and Peigne (2010) Manonga Valley Tanzania Tinde 5.000 4.500 4.750 N Werdelin (2010) Harrison and Baker (1997) and Sanders et al. (2010) Tugen Hills Kenya Chemeron/Mabaget NA NA 5.000 Y Werdelin (2010) Bernor et al. (2010), Geraads (2010), Harrison (2010), Hill et al. (1985), Jablonski and Frost (2010), MacLatchy et al. (2010), Morales et al. (2005), and Sanders et al. (2010)

a The stratigraphic and temporal resolution for these data are coarser than those from Table 1 because faunal lists are usually only published at the geological member level. b Sites are organized by midpoint age from youngest to oldest. c NA lower and upper ages represent those sites where only one age estimate was provided. associated with estimating a parameter. In the extreme case, if an and estimate the stratigraphic position of a mass extinction (Wang entire population is sampled, the CI will be of zero length. et al., 2009). Several studies have confirmed the effectiveness of the Strauss and Sadler method and demonstrated that when the probability of 2.4. Testing the assumptions of the Strauss and Sadler method sampling a species' fossil horizon is uniform throughout its temporal range, this method is the minimum-variance unbiased esti- The assumptions of uniformity and independence are critical, and mator i.e., it is the most precise estimator that can possibly be if they are not met, the average gap size between adjacent horizons calculated among all accurate estimators (Strauss and Sadler, 1989; will not be an accurate estimate of the gaps between the observed Rivadeneira et al., 2009; Wang et al., 2009, 2016). This method has and estimated range endpoints. It is also imperative that the uniform only recently been applied to the hominin fossil record (Maxwell, probability of recovering a species' fossil horizon extends beyond the 2018), but in other clades, it has been used to predict the ages of observed temporal range, so that the gaps between the observed and future fossil finds (Marshall, 1990), test whether hypothesized sis- estimated range endpoints follow the same distribution as the ter species appear simultaneously (Marshall, 1990), test the order of observed gaps within the species' range. We discuss how to test each species appearances predicted from cladograms (Wagner, 1995), of these assumptions in the remainder of this section, but first we 6 A. Du et al. / Journal of Human Evolution 138 (2020) 102688

Figure 2. A hypothetical species' temporal range with (A) 5 or (B) 30 uniformly and independently distributed fossil horizons (vertical lines). Each temporal range was simulated by drawing n points from a uniform distribution bounded by 3.0 and 4.0 Ma (i.e., the species' true extinction and origination dates, respectively). The oldest and youngest fossil horizons (i.e., the FAD and LAD, respectively) define the species' observed range (in gray). The distribution of age gaps between adjacent fossil horizons is used to estimate the gap between the species' FAD (LAD) and origination (extinction) date (gaps in white). (C) The distribution of age gaps between fossil horizons from (A) and (B). The mean of the age gap distribution is the minimum-variance unbiased estimate of the gap between a species' FAD (LAD) and origination (extinction) date. consider how the probability of sampling a fossil horizon can be coin's face, air temperature and movement), the complex in- uniform and random throughout a species' range. teractions between them generate an emergent binomial distri- It is well established that fossil recovery potential is a function of bution with a 50% chance of getting heads or tails (Gotelli and non-random factors such as a species' abundance, rock availability, Ellison, 2013). Thought of another way, patterns are scale- fossil preservation, and collection intensity at any given site dependent in that a strong trend at one scale becomes part of (Behrensmeyer et al., 2000; Holland, 2003). Given this, the some greater noise structure surrounding a different emergent assumption of uniformly and independently distributed fossil ho- pattern at a larger scale (Fig. 4) (see also Du et al., 2018). rizons might be considered unrealistic. However, fossil distribu- Recall that the Strauss and Sadler method assumes the distributions through time are scale-dependent (Wang et al., 2009, 2016), tion of a species' fossil horizons can be modeled with a Poisson and the aforementioned smaller-scale (i.e., site-specific), deter- process, such that within a species' temporal range, its fossil horizons ministic processes interact in complex ways to produce random are uniformly and independently distributed through time (see noise around an emergent pattern at the scale of an entire species' section 2.3). To test if this is the case for the Au. anamensiseafarensis temporal range (Maurer, 1999). The concept of emergence, which fossil horizons, we created a uniform probability plot (Solow et al., comes from complexity science, is defined as a large-scale entity 2006; Vogel et al., 2009; Wang et al., 2009; Oldford, 2016), possessing an orderly structure that cannot be predicted from the wherein the observed ages of fossil horizons are sorted and plotted properties of its constituent parts due to the latter's sheer numbers against quantiles from a standard uniform distribution (which is and complex interactions with each other (Maurer, 1999). For defined from 0 to 1). The linearity of this relationship is used to assess example, although the result of a coin flip is deterministically the assumptions of uniformity and independence. The expected caused by many factors (e.g., flick placement and strength, the quantile values are simply n (the number of observed fossil horizons)

Figure 3. (A) Twenty examples of sampling 30 fossil horizons from the temporal range of a species and calculating a one-sided 95% conﬁdence interval (CI) for the extinction age, which is 3 Ma (dashed line). On average, one 95% CI iteration will exclude the true extinction age, as has happened here (red CI). Fossil horizons were randomly sampled from a uniform distribution bounded by 3.0 Ma and 4.0 Ma (i.e., the extinction and origination ages, respectively). The formula for calculating a one-sided 95% CI, expressed as a proportion =ð Þ of the observed range, is ð1 0:95Þ 1 n 1 1, where n is the number of fossil horizons (equation (20) from Strauss and Sadler [1989]). (B) Scatterplot demonstrating the relationship between the width of 95% CIs versus the number of fossil horizons analyzed. CIs are expressed as the proportion of the observed temporal range. A. Du et al. / Journal of Human Evolution 138 (2020) 102688 7

intervals (Strauss and Sadler, 1989). A two-sided CI is conservative because the estimated interval at either end is larger than the estimate generated by a one-sided CI (i.e., for origination or extinction) (Fig. 3B; Strauss and Sadler, 1989; Marshall, 1990). If fossil horizons are uniformly and independently distributed through time, the estimated origination, O, and extinction, E, ages are equivalent to adding the mean age gap, calculated as ðFAD LADÞ =ðn 1Þ, to the FAD, and subtracting the mean age gap from the LAD, respectively (see section 2.3). To show that this is the case, note that the estimated origination date is given by:

b O ¼ FAD þðFAD LADÞ=ðn 1Þ

b FAD$ðn 1Þ FAD LAD O ¼ þ n 1 n 1

b FAD$n FAD þ FAD LAD O ¼ n 1

b Figure 4. A theoretical time series of some arbitrary variable, illustrating how patterns O ¼ðn $ FAD LADÞ=ðn 1Þ are scale dependent. (A) At some scale, the time series exhibits a strong trend through time. (B) If one zooms out, the signal at the smaller scale (dashed rectangle) becomes which is the estimated origination equation from Strauss and part of some greater random noise structure surrounding a mean value showing zero Sadler (1989) (their equation (8)). Likewise, the estimated extinc- net change through time (solid ordinary least squares line). In the context of the temporal distribution of hominin fossil horizons, non-random processes and patterns tion date is given by at the site scale can become random noise surrounding some emergent pattern at the scale of a species' temporal range. b E ¼ LAD ðFAD LADÞ=ðn 1Þ points evenly distributed between 0 and 1, non-inclusive. To esti- b LAD$ðn 1Þ FAD LAD E ¼ mate 95% confidence bounds for the linear relationship, we gener- n 1 n 1 ated and sorted n independent random values from a standard uniform distribution, repeated this procedure 10,000 times to get b LAD$n LAD FAD þ LAD E ¼ 10,000 vectors with n uniformly distributed random numbers in n 1 each, and then calculated 95% confidence bounds across these vec- b tors for each of the n quantiles. In order to scale the 95% confidence E ¼ðn $ LAD FADÞ=ðn 1Þ bounds to the range of observed fossil horizon ages, we regressed the observed ages against the expected quantile values to get location which is Strauss and Sadler's (1989) estimated extinction equation (intercept) and scale (slope) parameters, and then multiplied the 95% (their equation (8)). confidence bounds by the slope and added the intercept. This is permissible because the uniform distribution is a location-scale family, where shifting the location and/or scaling the width of a given uniform distribution will still result in a distribution being uniform (Casella and Berger, 2002). The Strauss and Sadler method also makes the assumption that the constant rate of fossil recovery extends from within the observed temporal range to immediately outside it (see section 2.3). It is not possible to test this assumption using the observed distribution of Au. anamensiseafarensis fossil horizons, which by definition is bounded by the taxon's FAD and LAD. The best we can do is note whether there are deposits amenable to the preservation of Au. anamensiseafarensis just outside its observed temporal range. This suggests that if the taxon was present then, we would have a good chance of sampling it. To assess the presence of such deposits, we noted whether there are fossil sites conducive to the preservation of hominins e that is, those sites preserving terrestrial large mammals and large-bodied primates (see section 2.1) e immediately older and younger than the FAD and LAD of Au. anamensiseafarensis, respectively.

Figure 5. Uniform probability plot assessing the uniformity and independence of 2.5. Estimating origination and extinction dates Australopithecus anamensiseafarensis fossil horizon ages (n ¼ 30). Observed midpoint ages, along with their bracketing dates as error bars, are plotted against expected quantiles from a standard uniform distribution (see section 2.4). A least squares model We calculated the minimum-variance unbiased estimate of (solid diagonal line) is fit to the data to illustrate linearity. The 95% confidence bounds origination and extinction ages, along with two-sided confidence (dashed lines) are calculated from 10,000 draws from a uniform distribution. 8 A. Du et al. / Journal of Human Evolution 138 (2020) 102688

We estimated a two-sided 95% confidence interval for both (Fig. 6). The mean age gap reflects how much estimated origination origination and extinction ages using equation (19) from Strauss and extinction dates extend the FAD and LAD, respectively, if fossil and Sadler (1989): recovery potential is uniform through time (see section 2.5). e Analyzing midpoint ages, the FAD of the Au. anamensis afarensis m ðn1Þ m ðn1Þ C ¼ 1 2 1 þ þ 1 þ 2 lineage is 4.15 Ma, so the estimated origination date is 4.19 Ma (i.e., FAD LAD FAD LAD 4.15 þ 0.038) (Table 3). Likewise, the LAD of Au. anamensiseafarensis is 3.04 Ma, so the estimated extinction date is fi where C is the selected con dence level (here, 0.95), and m is the 3.00 Ma (i.e., 3.04 0.038) (Table 3). Overall, the estimated tem- fi con dence interval estimated as an extension of the FAD and LAD. poral range extends the observed one by only 7% (Table 3; Fig. 7). fi Here, m was solved using a root- nding algorithm (the uniroot() The estimated 95% CIs for origination and extinction dates extend function in R). the FAD and LAD each by 0.15 Myr to 4.30 and 2.88 Ma, respectively All analyses were performed using R v3.3.3 (R Core Team, 2017), (Table 3; Fig. 7). When the lower (older) bracketing age for the FAD and all of the analyzed data and code are provided as (4.20 Ma) and the upper (younger) bracketing age for the LAD Supplementary Online Material (SOM). (2.95 Ma) are used (an extreme scenario maximizing dating uncertainty), the origination date becomes 4.24 Ma (95% CI: 4.37 Ma) 3. Results and the extinction date is 2.91 Ma (95% CI: 2.78 Ma) (Table 3). There are no trends or large gaps in the temporal distribution of ¼ The observed temporal distribution of Au. anamensiseafarensis terrestrial large mammal sites from 5 Ma to 2 Ma (n 45), and most ¼ fossil horizons (n ¼ 30) closely matches the distribution expected sites preserve large-bodied primates (n 40) (Table 2; Fig. 7). There if horizons are uniformly and independently distributed through are two sites just before the FAD of Au. ana- e d e time (Fig. 5). The uniformity assumption is not violated because mensis afarensis Kiloleli (4.5 4 Ma) and the lower Laetolil Beds the points fall on a straight line and inside the 95% confidence at Laetoli (4.36e3.85 Ma)dand these have no large-bodied pri- bounds. The dating uncertainty surrounding fossil horizons is mates (Table 2). There are three sites slightly younger than the LAD mostly circumscribed within the bounds, suggesting that no of Au. anamensiseafarensisdmiddle Lomekwi at West Turkana e matter where the horizon's true date falls between each set of (3.13 2.82 Ma), Shungura Member C in the lower Omo Valley bracketing ages, the resulting temporal distribution will be (3.07e2.53 Ma), and Gurumaha at Ledi-Geraru (2.82e2.75 Ma)d consistent with uniformity. The independence assumption is not and these all preserve large-bodied primates (Table 2). violated because we do not see large vertical gaps or horizontal “stripes” in the patterning of points (Fig. 5), which would be ex- 4. Discussion pected if non-random preservation and sampling biases created large gaps or clusters of fossil horizons, respectively, at the scale of We estimated origination and extinction dates, along with their e the entire Au. anamensiseafarensis temporal range (Wang et al., 95% CIs, for the Au. anamensis afarensis lineage. By quantifying the 2009). Therefore, the temporal distribution of Au. ana- uncertainty surrounding these dates, we can begin to evaluate mensiseafarensis fossil horizons is consistent with uniformity and whether this lineage's temporal range is consistent with it being a independence, so we can reliably estimate its origination and phyletic descendant of Ar. ramidus (Ward et al., 1999; White et al., extinction ages, as well as their 95% CIs. 2006, 2009b) or a direct ancestor of earliest Homo (Wood, 1992; The observed Au. anamensiseafarensis age gap distribution Villmoare et al., 2015) (via bifurcating cladogenesis). We are not (n ¼ 29) is right-skewed with a mean of 0.038 million years (Myr) claiming that temporal analyses should take precedence over morphological data, but that both criteria need to be satisfied before asserting a particular evolutionary relationship between two taxa (Fig. 1; Paul, 1992; Du and Alemseged, 2019). For example, if species A's LAD is found before the origination CI of species B, it is likely that species B originated after the disappearance of species A, which would be temporally consistent with an anagenetic or bifurcating cladogenetic ancestoredescendant relationship. On the other hand, if species A's LAD falls within the origination CI of species B, then their coexistence cannot be falsified, which is inconsistent with such relationships. We later discuss inferences that incorporate the uncertainty surrounding species A's extinction (see section 4.4).

4.1. Implications for the evolutionary relationship between Ar. ramidus and Au. anamensiseafarensis

Our results show that the LAD of Ar. ramidus (4.493e4.300 Ma [Simpson et al., 2019]) falls within the origination CI for Au. anamensiseafarensis, regardless of whether we used the age midpoint or lower bracketing age for the FAD of Au. anamensiseafarensis (though overlap is miniscule when using midpoint ages). The lower bracketing age (4.20 Ma) might actually be more accurate than the Figure 6. Histogram of the Australopithecus anamensiseafarensis age gap distribution midpoint age (4.15 Ma) because the oldest Au. anamensis specimen (n ¼ 29). Age gaps are deﬁned as the amount of time separating adjacent fossil hori- (ARA-VP-14/1) was found in situ immediately above a paleomag- zons. The asterisk represents the mean age gap, i.e., 0.038 million years. If fossil ho- netic reversal dated to 4.21 Ma (White et al., 2006). Moreover, rizons are uniformly and independently distributed through time, the mean age gap is the minimum-variance unbiased estimate for the gap between the FAD (LAD) and true though there are terrestrial large mammal sites preserved just origination (extinction) date. before the FAD of Au. anamensiseafarensis, the fact that those sites A. Du et al. / Journal of Human Evolution 138 (2020) 102688 9

Table 3 First (FAD) and last appearance data (LAD) of Australopithecus anamensiseafarensis, along with estimated origination and extinction dates and their 95% conﬁdence intervals (CIs).a

FAD Estimated origination date Origination CI LAD Estimated extinction date Extinction CI

Midpoint agesb 4.15 4.19 4.30 3.04 3.00 2.88 Maximizing dating uncertaintyc 4.20 4.24 4.37 2.95 2.91 2.78

a All numbers are in Ma. b “Midpoint ages” results are from using the midpoints of bracketing ages for the FAD and LAD. c “Maximizing dating uncertainty” results are from using the oldest and youngest bracketing ages for the FAD and LAD, respectively.

Figure 7. Observed temporal range of Australopithecus anamensiseafarensis (gray rectangle) along with estimated origination and extinction ages (white rectangle extensions) and their 95% conﬁdence intervals (CIs; error bars). Estimated ages and CIs were calculated using bracketing age midpoints for the FAD and LAD of Au. anamensiseafarensis (see Table 3). The vertical lines inside the gray rectangle show the observed temporal distribution of Au. anamensiseafarensis fossil horizons (n ¼ 30). The LAD of Ardipithecus ramidus (Simpson et al., 2019) and FAD of earliest Homo at Ledi-Geraru, Ethiopia (Villmoare et al., 2015) are presented as error bars representing their dating uncertainty. The points below indicate midpoint ages of sites where terrestrial large mammals are preserved in eastern Africa (n ¼ 45). Gray-ﬁlled points represent those sites where large-bodied primates are found (i.e., hominins or papionins; n ¼ 40).

do not preserve any large-bodied primates suggests the estimated 10 cm/1000 years (Campisano and Feibel, 2007), the age of A.L. origination CI might be too recent. Therefore, our results indicate 444-2 is estimated to be 3.06 Ma, which is slightly older than the that given the current Au. anamensiseafarensis record, the hy- analyzed Au. anamensiseafarensis LAD using age midpoints pothesis of temporal overlap between Ar. ramidus and Au. ana- (3.04 Ma), making the midpoint a more accurate estimate for the mensiseafarensis cannot be falsified, and we should be cautious LAD than the upper bracketing age (2.95 Ma). Together with the about positing a phyletic relationship between them. In fact, it fact that the earliest Homo mandible (LD 350-1) is located 10 m would take the discovery of 19 additional Au. anamensiseafarensis above the Gurumaha Tuff (2.82 Ma) (DiMaggio et al., 2015; fossil horizons to shrink the origination 95% CI in the extreme Villmoare et al., 2015), we can be confident that the species' cur- dating scenario to less than 4.3 Ma, and these additional horizons rent FAD lies outside the extinction CI of Au. anamensiseafarensis would all need to be younger than the FAD of Au. ana- (2.88 Ma). This result is therefore consistent with a bifurcating mensiseafarensis or else the origination CI will be shifted even cladogenetic ancestoredescendant relationship between the two earlier. Furthermore, the 3.4 Ma Burtele foot from Woranso-Mille, taxa. Furthermore, Au. afarensis specimens from nearby Hadar Ethiopia (Haile-Selassie et al., 2012) belongs to a taxon morpho- share many morphological similarities with LD 350-1 (Villmoare logically similar to Ar. ramidus and hence may have descended from et al., 2015), so a potential ancestoredescendant relationship some Ardipithecus species (White et al., 2015; Simpson et al., 2019). should not be controversial. Indeed, based on hominin dental evi- If the Burtele foot belongs to the phyletic descendant of Ar. ramidus, dence from the Shungura Formation, Suwa et al. (1996, pp. the contemporaneity of the Burtele foot and the Au. ana- 270e271) noted that “the circa 2.9 to 2.7 myr time range is likely to mensiseafarensis lineage would falsify a phyletic relationship be- be the transitional period when evolution occurred from an tween Au. anamensiseafarensis and Ar. ramidus, as their temporal A. afarensis-like to a more advanced species … close in overall ranges would overlap substantially. dental morphology to the A. africanus condition but also mostly within the A. afarensis and/or early Homo ranges of variation.” We 4.2. Implications for the evolutionary relationship between Au. note, however, that if an earlier Homo specimen is discovered anamensiseafarensis and earliest Homo within the extinction CI of Au. anamensiseafarensis,an ancestoredescendant relationship via bifurcating cladogenesis The FAD of earliest Homo (2.8e2.75 Ma [DiMaggio et al., 2015; should be viewed with skepticism. The hypothesis would be falsi- Villmoare et al., 2015]) postdates the extinction CI of Au. ana- fied if an earliest Homo specimen is found before the LAD of Au. mensiseafarensis when analyzing age midpoints, but not when anamensiseafarensis, barring any taxonomic misidentifications using the upper bracketing age for the LAD of Au. ana- (e.g., the earlier Homo specimen is a misidentified Au. afarensis). In mensiseafarensis. The youngest published Au. afarensis specimen this case, both morphological and temporal evidence would sug- (A.L. 444-2) is located 10.5 m below the BKT-2 marker tuff gest a budding cladogenetic relationship between the two taxa (2.95 Ma) (Kimbel et al., 1994). Assuming a sedimentation rate of (Fig. 1C). 10 A. Du et al. / Journal of Human Evolution 138 (2020) 102688

4.3. The Au. anamensiseafarensis extinction CI and the post- Ardipithecus kadabba form a lineage (Haile-Selassie, 2004; White 2.95 Ma unconformity in the lower Awash et al., 2009b, 2015), while earliest Homo at Ledi-Geraru might be linked to one or more species of Homo from the early Pleistocene Aside from several tooth crowns from Usno (3.05 Ma), tenta- (e.g., H. habilis sensu stricto, H. rudolfensis)(Villmoare et al., 2015). tively attributed to Au. afarensis (Kimbel and Delezene, 2009), the However, we advise that doing so should await further fossil evi- end of the Au. anamensiseafarensis temporal range is dominated by dence bearing on these potential relationships, given the consid- specimens from one site: Hadar (Table 1). At Hadar, the Hadar erable uncertainties around hominin phylogeny in general (Collard (~3.45e2.95 Ma) and Busidima (~2.7e0.8 Ma) Formations are and Wood, 2000; Wood and Harrison, 2011; Kimbel, 2013). separated by an angular unconformity above the BKT-2 complex (Campisano and Feibel, 2008; Wynn et al., 2008). The post-2.95 Ma 5. Conclusions unconformity resulted from shifts in rift tectonics in the Hadar region that ultimately formed the half-graben in which the Busi- For the Au. anamensiseafarensis lineage, we estimate ages of dima Formation was deposited (Wynn et al., 2008; Campisano, origination and extinction, along with their 95% CIs, to be 4.19 Ma 2012). Thus, it is worth considering whether the HadareBusidima (CI: 4.30 Ma) and 3.00 Ma (CI: 2.88 Ma), respectively. For a scenario unconformity might have caused our estimated Au. ana- where dating uncertainty is maximized, those dates become 4.24 Ma mensiseafarensis extinction CI to be too old. To the northeast away (CI: 4.37 Ma) and 2.91 Ma (CI: 2.78 Ma). We conclude that these dates from the localized unconformity, the 3.0e2.5 Ma period in the are consistent with Au. anamensiseafarensis being a direct ancestor lower Awash is represented by the Lee Adoyta Basin of Ledi-Geraru of earliest Homo (via bifurcating cladogenesis), but temporal overlap (2.82e2.58 Ma), which does not have any Au. afarensis fossils between Ar. ramidus and Au. anamensiseafarensis cannot be ruled despite being only ~30 km away from Hadar (DiMaggio et al., 2015; out at this time. We reiterate that our analysis should be considered Villmoare et al., 2015). Ledi-Geraru's spatiotemporal extent and preliminary, and that future research should incorporate the un- fossil productivity are comparable to that of the youngest member certainties surrounding the LAD and FAD of Ar. ramidus and earliest at Hadar (i.e., Kada Hadar). Ledi-Geraru's deposits represent 105 Homo, respectively, when sample sizes increase for these taxa. We years, are distributed over 10 1 km2, and have produced 103 also emphasize that we are not claiming that temporal tests of vertebrate fossil specimens (DiMaggio et al., 2015; Bibi et al., 2017; phylogenetic hypotheses should supersede morphological ones, but Rowan et al., 2017; Lazagabaster et al., 2018). Similarly, the Kada instead that the former be considered as an additional necessary test Hadar Member represents 105 years, 101 km2, and 103 vertebrate for assessing evolutionary relationships between two taxa. For specimens (Campisano, 2007; Reed, 2008; Hadar Geoinformatics example, hypothesized phyletic ancestors and descendants must be Database: http://gis1.asurite.ad.asu.edu/hadarv3/). Therefore, we sufficiently similar morphologically to the exclusion of other species conclude that the post-2.95 Ma unconformity in the Afar did not in their clade and their temporal ranges must not overlap. Given the appreciably decrease fossil recovery potential after the LAD of Au. insight gained from quantifying the uncertainty surrounding the anamensiseafarensis, and the estimated extinction CI is robust. appearance and disappearance of Au. anamensiseafarensis, similar analyses for other hominin taxa should be a high priority in the study 4.4. Incorporating uncertainty surrounding the range endpoints of of human origins. both ancestor and descendant taxa Acknowledgments A more complete test of temporal range overlap than presented here would take into account the uncertainty surrounding origi- We thank Charles Marshall and Alessio Capobianco for discus- nation and extinction dates for both the proposed ancestral and sions about confidence interval methods. We also thank Mike descendant taxa. Establishing temporal range overlap is most Plavcan and two anonymous reviewers for their helpful comments, straightforward when a candidate ancestor's estimated extinction which greatly improved the manuscript. This research did not date overlaps the estimated origination date of the candidate receive any specific grant from funding agencies in the public, descendant. Inferring overlapping ranges becomes more compli- commercial, or not-for-profit sectors. cated when it is only a taxon's CI that intersects the range of the other taxon, as overlap cannot be demonstrated in these cases. All Appendix A. Supplementary Online Material one can conclude is that range overlap cannot be falsified at this time (as is the case here with Ar. ramidus and Au. ana- Supplementary online material related to this article can be e fi mensis afarensis), since future nds will shrink the CI (Fig. 3B), found online at https://doi.org/10.1016/j.jhevol.2019.102688. which may result in the exclusion of the other taxon's range. To rigorously test range overlap, we need methods for placing CIs on References the gap between the ranges of two taxa (Wang and Everson, 2007). As an analogy, a two-sample t-test assesses if the means of two Alemseged, Z., Wynn, J.G., Kimbel, W.H., Reed, D., Geraads, D., Bobe, R., 2005. A new groups are significantly different not by comparing each mean's hominin from the Basal Member of the Hadar Formation, Dikika, Ethiopia, and fi its geological context. Journal of Human Evolution 49, 499e514. con dence intervals with each other, but by testing whether the Alroy, J., Koch, P.L., Zachos, J.C., 2000. Global climate change and North American confidence intervals surrounding the mean difference between mammalian evolution. Paleobiology 26, 259e288. both groups excludes zero. Barry, J.C., Morgan, M.E., Flynn, L.J., Pilbeam, D., Behrensmeyer, A.K., Raza, S.M., Due to a paucity of data points, robust CIs cannot be estimated Khan, I.A., Badgley, C., Hicks, J., Kelley, J., 2002. Faunal and environmental change in the late Miocene Siwaliks of northern Pakistan. Paleobiology 28, for Ar. ramidus (four fossil horizons [Semaw et al., 2005; Quade 1e71. et al., 2008; White et al., 2009b; Simpson et al., 2019]) and Behrensmeyer, A.K., Kidwell, S.M., Gastaldo, R.A., 2000. Taphonomy and paleobi- e earliest Homo (one fossil horizon [Villmoare et al., 2015]), so we ology. Paleobiology 26, 103 147. Behrensmeyer, A.K., Todd, N.E., Potts, R., McBrinn, G.E., 1997. Late Pliocene faunal suggest we have provided the most rigorous temporal test of Au. turnover in the Turkana Basin, Kenya and Ethiopia. Science 278, 1589e1594. anamensiseafarensis evolutionary relationships currently possible. Bernor, R., Armour-Chelu, M., Gilbert, H., Kaiser, T., Schulz, E., 2010. Equidae. In: Otherwise, one would need to make further assumptions about Ar. Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of California Press, Berkeley, pp. 685e721. ramidus and earliest Homo each sharing a lineage with other Bernor, R.L., Gilbert, H., Semprebon, G.M., Simpson, S., Semaw, S., 2013. Eur- hominin species. For example, one might assume Ar. ramidus and ygnathohippus woldegabrieli, sp. nov. (Perissodactyla, Mammalia), from the A. Du et al. / Journal of Human Evolution 138 (2020) 102688 11

middle Pliocene of Aramis, Ethiopia. Journal of Vertebrate Paleontology 33, Feibel, C.S., Brown, F.H., McDougall, I., 1989. Stratigraphic context of fossil hominids 1472e1485. from the Omo group deposits: Northern Turkana Basin, Kenya and Ethiopia. Bibi, F., Kiessling, W., 2015. Continuous evolutionary change in Plio-Pleistocene American Journal of Physical Anthropology 78, 595e622. mammals of eastern Africa. Proceedings of the National Academy of Sciences Foote, M., 1996. On the probability of ancestors in the fossil record. Paleobiology 22, 112, 10623e10628. 141e151. Bibi, F., Rowan, J., Reed, K., 2017. Late Pliocene Bovidae from Ledi-Geraru (Lower Fortelius, M., Zliobaite,_ I., Kaya, F., Bibi, F., Bobe, R., Leakey, L., Leakey, M., Awash Valley, Ethiopia) and their implications for Afar paleoecology. Journal of Patterson, D., Rannikko, J., Werdelin, L., 2016. An ecometric analysis of the fossil Vertebrate Paleontology 37, e1337639. mammal record of the Turkana Basin. Philosophical Transactions of the Royal Bobe, R., 2011. Fossil mammals and paleoenvironments in the Omo-Turkana Basin. Society B: Biological Sciences 371, 20150232. Evolutionary Anthropology 20, 254e263. García, N., 2008. New Eucyon remains from the Pliocene Aramis Member (Sagantole Bobe, R., Carvalho, S., 2019. Hominin diversity and high environmental variability in Formation), Middle Awash Valley (Ethiopia). Comptes Rendus Palevol 7, the Okote Member, Koobi Fora Formation, Kenya. Journal of Human Evolution 583e590. 126, 91e105. Geraads, D., 2010. Rhinocerotidae. In: Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Bobe, R., Leakey, M.G., 2009. Ecology of Plio-Pleistocene mammals in the Mammals of Africa. University of California Press, Berkeley, pp. 669e683. OmodTurkana Basin and the emergence of Homo. In: Grine, F.E., Fleagle, J.G., Geraads, D., Bobe, R., Manthi, F.K., 2013a. New ruminants (Mammalia) from the Leakey, R.E. (Eds.), The First Humans e Origin and Early Evolution of the Genus Pliocene of Kanapoi, Kenya, and a revision of previous collections, with a note Homo, Vertebrate Paleobiology and Paleoanthropology. Springer Netherlands, on the Suidae. Journal of African Earth Sciences 85, 53e61. Dordrecht, pp. 173e184. Geraads, D., Bobe, R., Reed, K., 2012. Pliocene Bovidae (Mammalia) from the Hadar Boisserie, J.-R., Souron, A., Mackaye, H.T., Likius, A., Vignaud, P., Brunet, M., 2014. Formation of Hadar and Ledi-Geraru, Lower Awash, Ethiopia. Journal of A new species of Nyanzachoerus (Cetartiodactyla: Suidae) from the late Miocene Vertebrate Paleontology 32, 180e197. Toros-Menalla, Chad, Central Africa. PLoS One 9, e103221. Geraads, D., Drapeau, M.S., Bobe, R., Fleagle, J.G., 2015. Vulpes mathisoni, sp. nov., a Boisserie, J.-R., White, T.D., 2004. A new species of Pliocene Hippopotamidae new fox from the Pliocene Mursi Formation of southern Ethiopia and its from the Middle Awash, Ethiopia. Journal of Vertebrate Paleontology 24, contribution to the origin of African foxes. Journal of Vertebrate Paleontology 464e473. 35, e943765. Campisano, C., Behrensmeyer, A., Bobe, R., Levin, N., 2004. High resolution paleo- Geraads, D., Reed, K., Bobe, R., 2013b. Pliocene Giraffidae (Mammalia) from the environmental comparisons between Hadar and Koobi Fora: preliminary re- Hadar Formation of Hadar and Ledi-Geraru, Lower Awash, Ethiopia. Journal of sults of a combined geological and paleontological approach. In: Vertebrate Paleontology 33, 470e481. Paleoanthropology Society Abstracts, p. A35. Gingerich, P.D., Uhen, M.D., 1998. Likelihood estimation of the time of origin of Campisano, C.J., 2007. Tephrostratigraphy and hominin paleoenvironments of the Cetacea and the time of divergence of Cetacea and Artiodactyla. Palaeontologia Hadar Formation, Afar depression, Ethiopia. PhD, Rutgers University, New Electronica 1, 1e47. Brunswick, N.J. Gomez-Robles, A., Smaers, J.B., Holloway, R.L., Polly, P.D., Wood, B.A., 2017. Brain Campisano, C.J., 2012. Geological summary of the Busidima Formation (Plio-Pleis- enlargement and dental reduction were not linked in hominin evolution. Pro- tocene) at the Hadar paleoanthropological site, Afar Depression, Ethiopia. ceedings of the National Academy of Sciences 114, 468e473. Journal of Human Evolution 62, 338e352. Gotelli, N.J., Ellison, A.M., 2013. A Primer of Ecological Statistics, second ed. Sinauer Campisano, C.J., Feibel, C.S., 2007. Connecting local environmental sequences to Associates, Inc., Publishers, Sunderland, Massachusetts. global climate patterns: evidence from the hominin-bearing Hadar Formation, Haile-Selassie, Y., 2004. Late Miocene teeth from Middle Awash, Ethiopia, and early Ethiopia. Journal of Human Evolution 53, 515e527. hominid dental evolution. Science 303, 1503e1505. Campisano, C.J., Feibel, C.S., 2008. Depositional environments and stratigraphic Haile-Selassie, Y., Latimer, B.M., Alene, M., Deino, A.L., Gibert, L., Melillo, S.M., summary of the Pliocene Hadar Formation at Hadar, Afar Depression, Ethiopia. Saylor, B.Z., Scott, G.R., Lovejoy, C.O., 2010. An early Australopithecus afarensis In: Quade, J., Wynn, J.G. (Eds.), The Geology of Early Humans in the Horn of postcranium from Woranso-Mille, Ethiopia. Proceedings of the National Acad- Africa. Geological Society of America. emy of Sciences 107, 12121e12126. Casella, G., Berger, R.L., 2002. Statistical Inference, second ed. Thomson Learning, Haile-Selassie, Y., Melillo, S.M., Ryan, T.M., Levin, N.E., Saylor, B.Z., Deino, A., Australia; Pacific Grove, CA. Mundil, R., Scott, G., Alene, Mulugeta, Gibert, L., 2016. Dentognathic remains of Coffing, K., Feibel, C., Leakey, M., Walker, A., 1994. Four-million-year-old hominids Australopithecus afarensis from Nefuraytu (Woranso-Mille, Ethiopia): compar- from East Lake Turkana, Kenya. American Journal of Physical Anthropology 93, ative description, geology, and paleoecological context. Journal of Human 55e65. Evolution 100, 35e53. Collard, M., Wood, B., 2000. How reliable are human phylogenetic hypotheses? Haile-Selassie, Y., Saylor, B.Z., Deino, A., Levin, N.E., Alene, M., Latimer, B.M., 2012. Proceedings of the National Academy of Sciences 97, 5003e5006. A new hominin foot from Ethiopia shows multiple Pliocene bipedal adapta- Curran, S.C., Haile-Selassie, Y., 2016. Paleoecological reconstruction of hominin- tions. Nature 483, 565e569. bearing middle Pliocene localities at Woranso-Mille, Ethiopia. Journal of Hu- Harris, J.M., Leakey, M.G., Cerling, T.E., Winkler, A., 2003. Early Pliocene tetrapod man Evolution 96, 97e112. remains from Kanapoi, Lake Turkana Basin, Kenya. Contributions in Science de Heinzelin, J., Clark, J.D., White, T., Hart, W., Renne, P., WoldeGabriel, G., Beyene, Y., 498, 39e114. Vrba, E., 1999. Environment and behavior of 2.5-million-year-old Bouri homi- Harrison, T., 2010. Later Tertiary Lorisiformes. In: Werdelin, L., Sanders, W.J. (Eds.), nids. Science 284, 625e629. Cenozoic Mammals of Africa. University of California Press, Berkeley, Deino, A.L., 2011. 40Ar/39Ar dating of Laetoli, Tanzania. In: Paleontology and Geology pp. 333e349. of Laetoli: Human Evolution in Context. Springer, Dordrecht, pp. 77e97. Harrison, T., 2011a. Hominins from the Upper Laetolil and Upper Ndolanya Beds, Deino, A.L., Scott, G.R., Saylor, B., Alene, M., Angelini, J.D., Haile-Selassie, Y., 2010. Laetoli. In: Paleontology and Geology of Laetoli: Human Evolution in Context, 40Ar/39Ar dating, paleomagnetism, and tephrochemistry of Pliocene strata of Vertebrate Paleobiology and Paleoanthropology Series. Springer, Dordrecht, the hominid-bearing Woranso-Mille area, west-central Afar Rift, Ethiopia. pp. 141e188. Journal of Human Evolution 58, 111e126. Harrison, T. (Ed.), 2011b. Paleontology and Geology of Laetoli: Human Evolution in deMenocal, P.B., 2004. African climate change and faunal evolution during the Context, Vertebrate Paleobiology and Paleoanthropology. Springer Netherlands, Pliocene-Pleistocene. Earth and Planetary Science Letters 220, 3e24. Dordrecht. deMenocal, P.B., 2011. Climate and human evolution. Science 331, 540e542. Harrison, T., Baker, E., 1997. Paleontology and biochronology of fossil localities in the DiMaggio, E.N., Campisano, C.J., Rowan, J., Dupont-Nivet, G., Deino, A.L., Bibi, F., Manonga Valley, Tanzania. In: Neogene Paleontology of the Manonga Valley, Lewis, M.E., Souron, A., Garello, D., Werdelin, L., 2015. Late Pliocene fossiliferous Tanzania, Topics in Geobiology. Springer, Boston, pp. 361e393. sedimentary record and the environmental context of early Homo from Afar, Hill, A., Drake, R., Tauxe, L., Monaghan, M., Barry, J., Behrensmeyer, A., Curtis, G., Ethiopia. Science 347, 1355e1359. Jacobs, B.F., Jacobs, L., Johnson, N., 1985. Neogene palaeontology and geochro- Drapeau, M.S., Bobe, R., Wynn, J.G., Campisano, C.J., Dumouchel, L., Geraads, D., nology of the Baringo Basin, Kenya. Journal of Human Evolution 14, 759e773. 2014. The Omo Mursi Formation: a window into the East African Pliocene. Holland, S.M., 2003. Confidence limits on fossil ranges that account for facies Journal of Human Evolution 75, 64e79. changes. Paleobiology 29, 468e479. Du, A., Alemseged, Z., 2019. Temporal evidence shows Australopithecus sediba is Hone, D.W.E., Keesey, T.M., Pisani, D., Purvis, A., 2005. Macroevolutionary trends in unlikely to be the ancestor of Homo. Science Advances 5, eaav9038. the Dinosauria: Cope's rule: Dinosaurs obey Cope's rule. Journal of Evolutionary Du, A., Zipkin, A.M., Hatala, K.G., Renner, E., Baker, J.L., Bianchi, S., Bernal, K.H., Biology 18, 587e595. Wood, B.A., 2018. Pattern and process in hominin brain size evolution are scale- Jablonski, N.G., Frost, S., 2010. Cercopithecoidea. In: Werdelin, L., Sanders, W.J. dependent. Proceedings of the Royal Society B: Biological Sciences 285, (Eds.), Cenozoic Mammals of Africa. University of California Press, Berkeley, 20172738. pp. 393e428. Faith, J.T., Rowan, J., Du, A., Koch, P.L., 2018. Plio-Pleistocene decline of African Kappelman, J., Swisher III, C.C., Fleagle, J.G., Yirga, S., Bown, T.M., Feseha, M., 1996. megaherbivores: no evidence for ancient hominin impacts. Science 362, Age of Australopithecus afarensis from Fejej, Ethiopia. Journal of Human Evo- 938e941. lution 30, 139e146. Feibel, C.S., 2003. Stratigraphy and depositional history of the Lothagam sequence. Kaya, F., Bibi, F., Zliobaite,_ I., Eronen, J.T., Hui, T., Fortelius, M., 2018. The rise and fall In: Lothagam: the Dawn of Humanity in Eastern Africa. Columbia University of the Old World savannah fauna and the origins of the African savannah biome. Press, New York, pp. 17e29. Nature Ecology & Evolution 2, 241e246. 12 A. Du et al. / Journal of Human Evolution 138 (2020) 102688

Kimbel, W., Lockwood, C., Ward, C., Leakey, M., Rak, Y., Johanson, D., 2006. Was Rivadeneira, M.M., Hunt, G., Roy, K., 2009. The use of sighting records to infer Australopithecus anamensis ancestral to A. afarensis? A case of anagenesis in the species extinctions: an evaluation of different methods. Ecology 90, hominin fossil record. Journal of Human Evolution 51, 134e152. 1291e1300. Kimbel, W.H., 1988. Identification of a partial cranium of Australopithecus afarensis Rowan, J., Lazagabaster, I.A., Campisano, C.J., Bibi, F., Boisserie, J.-R., Frost, S.R., from the Koobi Fora Formation, Kenya. Journal of Human Evolution 17, Getachew, T., Kimbel, W.H., Lewis, M.E., Scott, E., Selasie, S., Seyoum, C., 647e656. Souron, A., Werdelin, L., Reed, K.E., 2019. Early Pleistocene large mammals from Kimbel, W.H., 2013. Palaeoanthropology: hesitation on hominin history. Nature 497, the Maka'amitalu Basin of Hadar, Ethiopia, and the paleoecology of early Homo 573e574. in the lower Awash Valley (in prep.). Kimbel, W.H., Delezene, L.K., 2009. “Lucy” redux: a review of research on Austral- Rowan, J., Locke, E.M., Robinson, J.R., Campisano, C.J., Wynn, J.G., Reed, K.E., 2017. opithecus afarensis. American Journal of Physical Anthropology 140, 2e48. Fossil Giraffidae (Mammalia, Artiodactyla) from Lee Adoyta, Ledi-Geraru, and Kimbel, W.H., Johanson, D.C., Rak, Y., 1994. The first skull and other new discoveries Late Pliocene dietary evolution in Giraffids from the Lower Awash Valley, of Australopithecus afarensis at Hadar, Ethiopia. Nature 368, 449e451. Ethiopia. Journal of Mammalian Evolution 24, 359e371. Kimbel, W.H., Walter, R.C., Johanson, D.C., Reed, K.E., Aronson, J.L., Assefa, Z., Sanders, W.J., Gheerbrant, E., Harris, J.M., Saegusa, H., Delmer, C., 2010. Proboscidea. Marean, C.W., Eck, G., Bobe, R., Hovers, E., 1996. Late Pliocene Homo and Old- In: Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of owan tools from the Hadar Formation (Kada Hadar Member), Ethiopia. Journal California Press, Berkeley, pp. 161e251. of Human Evolution 31, 549e561. Semaw, S., Simpson, S.W., Quade, J., Renne, P.R., Butler, R.F., McIntosh, W.C., Kingman, J.F.C., 1993. Poisson Processes, Oxford Studies in Probability. Clarendon Levin, N., Dominguez-Rodrigo, M., Rogers, M.J., 2005. Early Pliocene hominids Press, Oxford. from Gona, Ethiopia. Nature 433, 301e305. Koch, C.F., 1987. Prediction of sample size effects on the measured temporal and Simpson, S.W., Levin, N.E., Quade, J., Rogers, M.J., Semaw, S., 2019. Ardipithecus geographic distribution patterns of species. Paleobiology 13, 100e107. ramidus postcrania from the Gona Project area, Afar Regional State, Ethiopia. Kullmer, O., Sandrock, O., Viola, T.B., Hujer, W., Said, H., Seidler, H., 2008. Suids, Journal of Human Evolution 129, 1e45. elephantoids, paleochronology, and paleoecology of the Pliocene hominid site Smith, D.M., Marcot, J.D., 2015. The fossil record and macroevolutionary history of Galili, Somali Region, Ethiopia. Palaios 23, 452e464. the beetles. Proceedings of the Royal Society B: Biological Sciences 282, Lazagabaster, I.A., Souron, A., Rowan, J., Robinson, J.R., Campisano, C.J., Reed, K.E., 20150060. 2018. Fossil Suidae (Mammalia, Artiodactyla) from Lee Adoyta, Ledi-Geraru, Solow, A.R., Roberts, D.L., Robbirt, K.M., 2006. On the Pleistocene extinctions of lower Awash Valley, Ethiopia: implications for late Pliocene turnover and Alaskan mammoths and horses. Proceedings of the National Academy of Sci- paleoecology. Palaeogeography, Palaeoclimatology, Palaeoecology 504, ences 103, 7351e7353. 186e200. Strauss, D., Sadler, P.M., 1989. Classical confidence intervals and Bayesian probability Leakey, M.G., Feibel, C.S., McDougall, I., Walker, A., 1995. New four-million-year-old estimates for ends of local taxon ranges. Mathematical Geology 21, 411e427. hominid species from Kanapoi and Allia Bay, Kenya. Nature 376, 565. Suwa, G., White, T.D., Howell, F.C., 1996. Mandibular postcanine dentition from the Leakey, M.G., Feibel, C.S., McDougall, I., Ward, C., Walker, A., 1998. New specimens Shungura Formation, Ethiopia: crown morphology, taxonomic allocations, and and confirmation of an early age for Australopithecus anamensis. Nature 393, 62. Plio-Pleistocene hominid evolution. American Journal of Physical Anthropology Leakey, M.G., Harris, J.M. (Eds.), 2003. Lothagam: the Dawn of Humanity in Eastern 101, 247e282. Africa. Columbia University Press, New York. Tomiya, S., 2013. Body size and extinction risk in terrestrial mammals above the Lebatard, A.-E., Bourles, D.L., Duringer, P., Jolivet, M., Braucher, R., Carcaillet, J., species level. The American Naturalist 182, E196eE214. Schuster, M., Arnaud, N., Monie, P., Lihoreau, F., Likius, A., Mackaye, H.T., Villmoare, B., Kimbel, W.H., Seyoum, C., Campisano, C.J., DiMaggio, E.N., Rowan, J., Vignaud, P., Brunet, M., 2008. Cosmogenic nuclide dating of Sahelanthropus Braun, D.R., Arrowsmith, J.R., Reed, K.E., 2015. Early Homo at 2.8 Ma from Ledi- tchadensis and Australopithecus bahrelghazali: Mio-Pliocene hominids from Geraru, Afar, Ethiopia. Science 347, 1352e1355. Chad. Proceedings of the National Academy of Sciences 105, 3226e3231. Vogel, R.M., Hosking, J.R.M., Elphick, C.S., Roberts, D.L., Reed, J.M., 2009. Goodness of MacLatchy, L.M., DeSilva, J.M., Sanders, W.J., Wood, B.A., 2010. Hominini. In: fit of probability distributions for sightings as species approach extinction. Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of Bulletin of Mathematical Biology 71, 701e719. California Press, Berkeley, pp. 471e540. Wagner, P.J., 1995. Stratigraphic tests of cladistic hypotheses. Paleobiology 21, Marshall, C.R., 1990. Confidence intervals on stratigraphic ranges. Paleobiology 16, 153e178. 1e10. Wang, S.C., Chudzicki, D.J., Everson, P.J., 2009. Optimal estimators of the position of Marshall, C.R., 2010. Using confidence intervals to quantify the uncertainty in the a mass extinction when recovery potential is uniform. Paleobiology 35, end-points of stratigraphic ranges. In: Alroy, J., Hunt, G. (Eds.), Quantitative 447e459. Methods in Paleobiology, Paleontological Society Papers. Paleontological Soci- Wang, S.C., Everson, P.J., 2007. Confidence intervals for pulsed mass extinction ety, Boulder, pp. 291e316. events. Paleobiology 33, 324e336. Maurer, B.A., 1999. Untangling Ecological Complexity: the Macroscopic Perspective. Wang, S.C., Everson, P.J., Zhou, H.J., Park, D., Chudzicki, D.J., 2016. Adaptive credible University of Chicago Press, Chicago, Ill. intervals on stratigraphic ranges when recovery potential is unknown. Paleo- Maxwell, S.J., 2018. The quality of the early hominin fossil record: Implications for biology 42, 240e256. evolutionary analyses. PhD, University College London - Birkbeck, London. Wang, S.C., Marshall, C.R., 2016. Estimating times of extinction in the fossil record. Maxwell, S.J., Hopley, P.J., Upchurch, P., Soligo, C., 2018. Sporadic sampling, not Biology Letters 12, 20150989. climatic forcing, drives observed early hominin diversity. Proceedings of the Ward, C., Leakey, M., Walker, A., 1999. The new hominid species Australopithecus National Academy of Sciences 115, 4891e4896, 201721538. anamensis. Evolutionary Anthropology 7, 197e205. Mbua, E., Kusaka, S., Kunimatsu, Y., Geraads, D., Sawada, Y., Brown, F.H., Sakai, T., Werdelin, L., 2010. Chronology of Neogene mammal localities. In: Werdelin, Lars, Boisserie, J.-R., Saneyoshi, M., Omuombo, C., 2016. Kantis: a new Austral- Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of California Press, opithecus site on the shoulders of the Rift Valley near Nairobi, Kenya. Journal of Berkeley, pp. 27e43. Human Evolution 94, 28e44. Werdelin, L., Lewis, M.E., 2013. The Carnivora, Koobi Fora Research Project. Cali- McDougall, I., Brown, F.H., 2008. Geochronology of the pre-KBS Tuff sequence, Omo fornia Academy of Sciences, San Francisco. Group, Turkana Basin. Journal of the Geological Society 165, 549e562. Werdelin, Lars, Peigne, S., 2010. Carnivora. In: Werdelin, L., Sanders, W.J. (Eds.), McDougall, I., Brown, F.H., Vasconcelos, P.M., Cohen, B.E., Thiede, D.S., Cenozoic Mammals of Africa. University of California Press, Berkeley, Buchanan, M.J., 2012. New single crystal 40Ar/39Ar ages improve time scale for pp. 603e657. deposition of the Omo Group, OmoeTurkana Basin, East Africa. Journal of the Weston, E., Boisserie, J.-R., 2010. Hippopotamidae. In: Werdelin, L., Sanders, W.J. Geological Society 169, 213e226. (Eds.), Cenozoic Mammals of Africa. University of California Press, Berkeley, Morales, J., Pickford, M., Soria, D., 2005. Carnivores from the late Miocene and basal pp. 853e871. Pliocene of the Tugen Hills, Kenya. Revista de la Sociedad Geologica de Espana~ White, Tim D., Ambrose, S.H., Suwa, G., Su, D.F., DeGusta, D., Bernor, R.L., 18, 39e61. Boisserie, J.-R., Brunet, M., Delson, E., Frost, S., 2009a. Macrovertebrate pale- Novack-Gottshall, P.M., Lanier, M.A., 2008. Scale-dependence of Cope's rule in body ontology and the Pliocene habitat of Ardipithecus ramidus. Science 326, 67e93. size evolution of Paleozoic brachiopods. Proceedings of the National Academy White, T.D., Asfaw, B., Beyene, Y., Haile-Selassie, Y., Lovejoy, C.O., Suwa, G., of Sciences 105, 5430e5434. WoldeGabriel, G., 2009b. Ardipithecus ramidus and the paleobiology of early Oldford, R.W., 2016. Self-calibrating quantileequantile plots. The American Statis- hominids. Science 326, 64e86. tician 70, 74e90. White, T.D., Lovejoy, C.O., Asfaw, B., Carlson, J.P., Suwa, G., 2015. Neither chimpanzee Paul, C.R.C., 1982. The adequacy of the fossil record. In: Joysey, K.A., Friday, A.E. nor human, Ardipithecus reveals the surprising ancestry of both. Proceedings of (Eds.), Problems of Phylogenetic Reconstruction. Academic Press, London, the National Academy of Sciences 112, 4877e4884. pp. 75e117. White, T.D., Suwa, G., 2004. A new species of Notochoerus (Artiodactyla, Suidae) Paul, C.R.C., 1992. The recognition of ancestors. Historical Biology 6, 239e250. from the Pliocene of Ethiopia. Journal of Vertebrate Paleontology 24, 474e480. Quade, J., Levin, N.E., Simpson, S.W., Butler, R., McIntosh, W.C., Semaw, S., White, T.D., Suwa, G., Hart, W.K., Walter, R.C., WoldeGabriel, G., de Heinzelin, J., Kleinsasser, L., Dupont-Nivet, G., Renne, P., Dunbar, N., 2008. The geology of Clark, J.D., Asfaw, B., Vrba, E., 1993. New discoveries of Australopithecus at Maka Gona, Afar, Ethiopia. In: Quade, J., Wynn, J.G. (Eds.), The Geology of Early in Ethiopia. Nature 366, 261. Humans in the Horn of Africa. Geological Society of America, pp. 1e31. White, T.D., WoldeGabriel, G., Asfaw, B., Ambrose, S., Beyene, Y., Bernor, R.L., R Core Team, 2017. R: a Language and Environment for Statistical Computing. R Boisserie, J.-R., Currie, B., Gilbert, H., Haile-Selassie, Y., Hart, W.K., Hlusko, L.J., Foundation for Statistical Computing, Vienna, Austria. Howell, F.C., Kono, R.T., Lehmann, T., Louchart, A., Lovejoy, C.O., Renne, P.R., Reed, K.E., 2008. Paleoecological patterns at the Hadar hominin site, Afar regional Saegusa, H., Vrba, E.S., Wesselman, H., Suwa, G., 2006. Asa Issie, Aramis and the state, Ethiopia. Journal of Human Evolution 54, 743e768. origin of Australopithecus. Nature 440, 883e889. A. Du et al. / Journal of Human Evolution 138 (2020) 102688 13

WoldeGabriel, G., Ambrose, S.H., Barboni, D., Bonneﬁlle, R., Bremond, L., Currie, B., Wynn, J.G., Alemseged, Z., Bobe, R., Geraads, D., Reed, D., Roman, D.C., 2006. DeGusta, D., Hart, W.K., Murray, A.M., Renne, P.R., 2009. The geological, isotopic, Geological and palaeontological context of a Pliocene juvenile hominin at botanical, invertebrate, and lower vertebrate surroundings of Ardipithecus Dikika, Ethiopia. Nature 443, 332e336. ramidus. Science 326, 65e65e5. Wynn, J.G., Roman, D.C., Alemseged, Z., Reed, D., Geraads, D., Munro, S., 2008. Wood, B., 1992. Origin and evolution of the genus Homo. Nature 355, 783e790. Stratigraphy, depositional environments, and basin structure of the Hadar and Wood, B., Boyle, E.K., 2016. Hominin taxic diversity: fact or fantasy? American Busidima Formations at Dikika, Ethiopia. Geological Society of America Special Journal of Physical Anthropology 159, 37e78. paper, 446, 87e118. Wood, B., Harrison, T., 2011. The evolutionary context of the ﬁrst hominins. Nature 470, 347e352.