Complex and changing patterns of natural selection explain the evolution of the human hip.

Journal of Human Evolution xxx (2015) 1e17

Contents lists available at ScienceDirect

Journal of Human Evolution journal homepage: www.elsevier.com/locate/jhevol

Complex and changing patterns of natural selection explain the evolution of the human hip Mark Grabowski a, b, *, Charles C. Roseman c a

Centre for Ecological and Evolutionary Synthesis (CEES), Department of Biosciences, University of Oslo, 0316 Oslo, Norway Center for the Advanced Study of Human Paleobiology, Department of Anthropology, The George Washington University, Washington, DC 20052, USA c Department of Anthropology, 179 Davenport Hall, University of Illinois, Urbana, IL 61820, USA b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 30 December 2013 Accepted 13 May 2015 Available online xxx

Causal explanations for the dramatic changes that occurred during the evolution of the human hip focus largely on selection for bipedal function and locomotor efficiency. These hypotheses rest on two critical assumptions. The firstdthat these anatomical changes served functional roles in bipedalismdhas been supported in numerous analyses showing how postcranial changes likely affected locomotion. The seconddthat morphological changes that did play functional roles in bipedalism were the result of selection for that behaviordhas not been previously explored and represents a major gap in our understanding of hominin hip evolution. Here we use evolutionary quantitative genetic models to test the hypothesis that strong directional selection on many individual aspects of morphology was responsible for the large differences observed across a sample of fossil hominin hips spanning the Plio-Pleistocene. Our approach uses covariance among traits and the differences between relatively complete fossils to estimate the net selection pressures that drove the major transitions in hominin hip evolution. Our findings show a complex and changing pattern of natural selection drove hominin hip evolution, and that many, but not all, traits hypothesized to play functional roles in bipedalism evolved as a direct result of natural selection. While the rate of evolutionary change for all transitions explored here does not exceed the amount expected if evolution was occurring solely through neutral processes, it was far above rates of evolution for morphological traits in other mammalian groups. Given that stasis is the norm in the mammalian fossil record, our results suggest that large shifts in the adaptive landscape drove hominin evolution. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Hominin evolution Bipedalism Selection Correlated evolution

1. Introduction Hip anatomy has changed drastically over the course of human evolution. Short hip bones, laterally rotated iliac blades, and expansion of the retroauricular region are some of the most important morphological differences distinguishing the hominin (species more closely related to modern humans than any other living taxon) hip from that of other great apes (Dart, 1949a; Lovejoy et al., 1973; Stern and Susman, 1983). Other more subtle differences have been used to distinguish between hominin species. For example, when compared to australopiths, Homo pelves generally have a taller posterior ilium, shorter anterior ilium, larger

* Corresponding author. E-mail address: [email protected] (M. Grabowski).

retroauricular area, more robust and projecting anterior inferior iliac spine, shorter tuberoacetabular sulcus, longer pubic symphyses, and an anteroposteriorly broader birth canal (Simpson et al., 2008, 2014). Most of the differences in hominin hip morphology are hypothesized to play functional roles in bipedal locomotion (see Table 1, Fig. 1). Accordingly, causal explanations for these morphological shifts largely focus on selection for bipedal competence and efficiency (e.g., Stern, 2000; Bramble and Lieberman, 2004). Hypotheses that morphological changes were the result of selection for bipedalism rest on two critical assumptions. The first is that those changes serve functional roles in bipedalism. Comparative and functional morphology can help to reveal the biological roles of traits by determining the functions they likely affected. Biomechanical modeling can be used to test these hypotheses (Ward, 2002), and numerous analyses show how hominin

http://dx.doi.org/10.1016/j.jhevol.2015.05.008 0047-2484/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Grabowski, M., Roseman, C.C., Complex and changing patterns of natural selection explain the evolution of the human hip, Journal of Human Evolution (2015), http://dx.doi.org/10.1016/j.jhevol.2015.05.008

2

M. Grabowski, C.C. Roseman / Journal of Human Evolution xxx (2015) 1e17

Table 1a Hypothesized function for measurements believed to play major roles in bipedalism in modern humans compared to homologous traits in inferred chimpanzee-like LCA. Distancea

Hypothesized functions of derived features in modern humans related to bipedalism

Anterior margin of iliac blade: H1

Posterior margin of iliac blade: H2 Auricular surface length: H3 Retroauricular height: H4

Lateral iliac breadth: H5 Lower iliac height: H6

Pubis length: H7 Pubic symphysis length: H8 Biomechanical moment arm of the ischium: H9

Femur Maximum length: F10 Biomechanical neck length: F11 Femoral Head diameter: F12

Bi-ilac breadth: O13

Bi-acetabular breadth: O14

a

Along with the iliac blades rotating laterally, shortening this distance in modern humans places the lesser gluteal mm. in position to function as abductors and provide pelvic stabilization during single support phase (Sigmon, 1975). Shortening this distance in modern humans reduces lumbar entrapment, allowing lumbar lordosis and placing the body’s center of mass above the hips (see McCollum et al., 2010). Expanded in modern humans reflecting adaptation for weight bearing during bipedal behavior (Berge and Kazmierczak, 1986). Expansion in modern humans leads to increased area for attachment of interosseous sacroiliac ligaments, the size of which is significant for maintaining upright posture (Dart, 1949a,b); places the expanded gluteus maximus m. in a suitable position from which to control heel strike (Lovejoy, 2005), leads to posterior displacement of sacral articular surface relative to acetabulum allowing line of gravity of vertically oriented trunk to pass close to femoral heads instead of in front of them (Stern and Susman, 1983); provides a larger area of attachment for erector spinae mm. that control trunk flexion (Bramble and Lieberman, 2004). e Shorter distance in modern humans places trunk center of gravity closer to hip joints and diminishes muscular torques needed to control balance (Stern and Susman, 1983); allows lumbar mobility not found in non-human African great apes (Lovejoy et al., 2009a,b). e e A shorter distance and dorsal placement of the ischial tuberosity in modern humans allows effective hamstring function on an extended limb, and when combined with a longer lower limb (Sockol et al., 2007), leads to a complex adapted for speed rather than power as seen in great apes (Robinson, 1972). Longer limbs in modern humans increase locomotor efficiency by increasing step length and thus, reducing cost of traveling the same distance (Pontzer et al., 2009). Combined with laterally placed ilia, longer necks in modern humans provide improved mechanical advantage of hip abductors and thus, requires less force production from these muscles (Ruff, 1995). Larger heads in modern humans reflects greater proportion of body mass supported by the lower limbs in bipedal locomotion (Jungers, 1988), are required to withstand elevated forces produced by less optimally placed abductors (Lovejoy, 1988), and are a consequence of body size increase from earlier hominins (Ruff, 1998). Combined with more laterally placed iliac blades, increasing this distance in modern humans increases the muscle moment arm at the hip that combined with gluteal muscle force counters torque generated by body mass and greater bi-acetabuar breadth during upright walking (Ruff, 1995). A wider pelvis in modern humans would allow for greater effects of transverse pelvic rotation, increasing step length and decreasing the center of mass vertical displacement and thus, increasing efficiency (Rak, 1991) and/ or would also leading to greater torque at the hip which must be countered by greater muscle force, a longer biomechanical neck length, or wider bi-iliac breadth (Ruff, 1995; but see Warrener, 2011).

Numbers correspond to traits shown in Fig. 1.

Table 1b Character state in inferred LCA and currently described fossil hominins compared to modern humans.a Distance

Inferred LCA morphology

Australopiths

Anterior margin of iliac blade: H1 Posterior margin of iliac blade: H2 Auricular surface length: H3

Much longer Much longer Shorter

Longer Similar Shorter

Retroauricular height: H4

Much shorter

Lateral iliac breadth: H5 Lower iliac height: H6

Slightly narrower Slightly longer*

Shorter (e.g., A.L. 288-1, Sts-14); Similar (e.g., KSD-VP-1/1, MH2) Wider Shorter

Pubis length: H7 Pubic symphysis length: H8 Biomechanical moment arm of the ischium: H9 Femur Maximum length: F10

Longer Longer Much longer Very short

Biomechanical neck length: F11 Femoral Head diameter: F12 Bi-ilac breadth: O13 Bi-acetabular breadth: O14

Shorter Smaller Slightly narrower Narrower

Longer Shorter Longer (Stern and Susman, 1983); Similar (Kibii et al., 2011) Shorter (e.g., A.L. 288-1); Similar (e.g., KSD-VP-1/1) Longer Smaller Much wider Much wider

H. erectus

Later Homo

H. neanderthalensis

NA Similar Slightly shorter Similar

Similar Similar Similar

Slightly longer Longer Similar

Similar

Similar

NA Slightly shorter Longer Similar Similar

Slightly wider Similar

Similar Shorter

Longer Slightly longer Similar

Longer Longer Similar

Similar

Similar

Similar

Longer Similar Wider Wider

Longer Slightly larger Much wider Slightly wider

NA Slightly larger Wider Slightly wider

*Slightly longer in terms of total lengththe Pan-like LCA has greatly increased height, but the modern human has expanded its bi-acetabular breadth, along with reducing height. a Numbers correspond to traits shown in Fig. 1. Refer to SOM Excel Sheets 1 and 2 for list of fossils in each comparison and full references.



3

Figure 1. Lateral (A) and medial (B) view of right hip bone, posterior view of right femur (C), and superior view of articulated pelvis (D) with linear dimensions used in this analysis. Solid lines denote traits calculated from 3D landmarks, dotted lines denote traits taken with calipers and an osteometric board.

postcranial changes likely affected locomotion (Stern, 2000; Sockol et al., 2007; Crompton et al., 2008; Warrener, 2011; Warrener et al., 2015). One recent study (Pontzer et al., 2009) argued that the most important anatomical changes during hominin evolution relating to locomotor efficiency were the increase in lower limb length and changes to the hip that allowed for movement with a more extended lower limb, including the reduction in size of the ischium and dorsal placement of the ischial tuberosity. This result implies that many of the traits traditionally hypothesized as affecting locomotor performance in early hominins likely played only relatively small roles (Pontzer et al., 2009). Relating these results to hypotheses of selection, Pontzer (2012) argued that changes in other areas of postcranial anatomy between australopiths and early Homo (non-erectus early Homo and Homo erectus) could have been the result of neutral evolution through genetic driftdmorphological changes due to random changes in allele frequenciesdrather than natural selection for locomotor efficiency. This leads to the second assumption in the chain linking morphological change to function to selection: that morphological changes that do play functional roles in bipedalism were the result of selection for this behavior. Changes in the hip throughout the hominin lineage (Table 1) are generally explained through the action of complex patterns of selection for bipedalism focused on individual traits (Sigmon, 1971; Robinson, 1972; Stern and Susman, 1983; Harcourt-Smith, 2007), with a smaller number of character differences believed to be the result of selection for differing obstetric demands (Berge et al., 1984; Rosenberg and Trevathan, 2002; Simpson et al., 2008). But playing a functional role in a

behavior does not mean that a character evolved as a result of selection for that behavior (Gould and Vrba, 1982), or even that it was under any selection at all (Gould and Lewontin, 1979). While directions and magnitudes of morphological change are commonly believed to reflect directions and magnitudes of selection that produced those changes (Simpson, 1953; Latimer, 1991), this is only true for traits that evolve independently of one another (Lovejoy et al., 1999). Because of underlying genetic covariance among traits, selection on one trait can lead to a correlated response in another (Lande, 1979; Cheverud et al., 1983; Lande and Arnold, 1983), and thus, even if they served functional roles in a behavior, morphological differences among taxa may not be a reliable indicator of past selection for that behavior. This second assumption remains a major roadblock in our understanding of hominin hip evolution. While the combination of phylogeny, comparative and functional anatomy, and biomechanical modeling can be used to construct hypotheses about past selective forces (Ward, 2002), we do not actually know how natural selection shaped most of hominin morphology (but see Ackermann and Cheverud, 2004; Rolian et al., 2010; Schroeder et al., 2014), including for the hip. Researchers have made headway in determining how morphological changes would have affected locomotion, but which traits were under natural selection, which were the focus of selection, and how the patterns of selection changed over time remain open questions. These questions are vitally important for understanding hominin hip evolution because they have the potential to reveal which aspects of morphology were the greatest contributors to fitness differences. This is of key relevance to both


4


phylogenetic and functional studies of fossil humans, and how and why hominin hip anatomy evolved. Here we apply the theories and methods of evolutionary quantitative genetics to test the hypothesis that strong directional selection on many individual aspects of morphology was responsible for the changes seen in the hip during hominin evolution. Our approach (Lande, 1979; Lande and Arnold, 1983; Ackermann and Cheverud, 2004; Rolian et al., 2010) takes into account covariance among traits and differences between relatively complete fossils to estimate the selection pressures that drove the major transitions in hominin hip evolution. Estimates of selection reveal which traits were under natural selection and which likely evolved as a correlated response to selection on other traits. In addition, we test the hypothesis that traits previously shown to have significant effects on locomotor performance (Pontzer et al., 2009) were the focus of selection or distinct in any way from other hip traits. We then estimate the overall directions and magnitudes of selection required for each transition and show how selection pressures changed in a broader sense over the course of hominin evolution. Finally, we calculate the rate of morphological change during major transitions of the hominin hip and ask whether the amount of change is consistent with the expectation under neutral evolution. Overall, our approach is a critical step in testing the second assumption of the form-function-selection chaindunderstanding the link between function and selection in the evolution of bipedalism. 2. Theoretical background for our approach Lovejoy et al. (1999) suggested a model for morphological change where selection on one or a small number of traits of major functional importance led to multiple correlated responses in other traits related by underlying developmental processes. This model posits relatively strong developmental interactions between traits that are similar among species, and that evolution occurs through selection on one or a few traits of major functional importance with many correlated responses (Lovejoy et al., 1999). Using the transition between a chimpanzee-like last common ancestor (LCA) of modern humans and chimpanzees to A.L. 288-1 (attributed to Australopithecus afarensis), Lovejoy and colleagues (1990) suggested the reduced distance between sacrum and femoral heads (measured here as lower iliac height; Table 1) was the prime focus of selection, with numerous other changes occurring as a result of correlated responses to selection on this trait. Though Lovejoy et al. (1999) formulated their hypotheses in terms of evolutionary developmental theory, the idea that underlying relationships between traits could influence evolution can be traced back to Darwin (1859). Its modern formdintegrationdwas introduced by Olson and Miller (1958) and realized by Cheverud (e.g., 1982), building on the work of Lande (1979; Lande and Arnold, 1983). The term integration has been used in different waysdboth to describe covariation among traits within populations (Klingenberg, 2008, 2009) and to describe the propensity for a developmental system to produce covariation in populations of organisms (Hallgrimsson et al., 2009). In the most widely accepted definition of integration today (Hallgrimsson et al., 2009), genetic and environmental influences channeled through developmental processes that influence multiple traits can lead to covariation among traits in a population. Integration of developmental processes is a feature of individuals, where pleiotropy (the effects of single loci on multiple traits) and the resulting shared developmental effects are present regardless of variation among individuals in a population. Covariation is a feature of populations and results from variation in integrated developmental systems that produce covariation. As natural

selection acts on variation and covariation within populations, these patterns are vitally important for understanding evolution because both theoretical and empirical studies have found that they can constrain, facilitate, and bias evolutionary trajectories. As such, patterns of covariance can lead to differences between patterns of evolutionary change and the evolutionary processes that produced them (Lande and Arnold, 1983; Hansen and Houle, 2008; Marroig et al., 2012; Goswami et al., 2014). Lande (1979) and colleagues (1983) provided a means to quantify the amount of selection responsible for observed change by deconstructing evolutionary change (i.e., the mean response to selection) as the product of the selection gradient (b) and the pattern of variance and covariance between traits in a population. As the other two components can be estimated from the data, calculating b estimates the forces of directional selection on individual traits that were required for a particular evolutionary transition, taking into account relationships between traits (see Methods section below for more on this approach). The metric of natural selection used here is defined as the increase in relative fitness for a proportional change in the trait of interest, or the regression slope of relative fitness on the trait (Hereford et al., 2004). Relative fitness is the expected fitness of individuals that posses a certain trait valuedi.e., the expected contribution of offspring to the next generation relative to the average of the population. To put it simply, estimates of selection as described here reveal how observed changes in traits impacted the overall fitness of the individuals that possessed them. Here b is a vector of values that reveal the change in relative fitness for a change in each trait while all other traits are held constantda partial regression coefficient. Multivariate estimates of selection show the contribution of changes in each trait to relative fitness, independent of other traits. The most important point of this fitness-based definition is that if a trait has evolved as a correlated response to selection on another trait (i.e., non-significant values of b), changes observed in this trait had little to no effect on fitness. Such a finding would contradict hypotheses that interpret anatomical changes as adaptations for particular functional goals, as these changes cannot be adaptive if they have no effect on fitness. While this approach was originally devised to quantify shortterm microevolutionary change, this analysis follows previous studies that have applied this approach to macroevolutionary questions in hominin evolution (Ackermann and Cheverud, 2004; Rolian et al., 2010; Schroeder et al., 2014) and primate evolution (Marroig and Cheverud, 2004). In this study, differences between taxa are the overall or net effect of natural selection, and changes observed in traits that appear to have been the result of substantial amounts of selection mean that these changes had large fitness effects. It is important to note that estimates of b over long evolutionary time scales reflect the net selection gradient. A near zero value of an entry in b might mean that a trait was under no directional selection for that transition, or it might mean that the trait was under directional selection in one direction for one segment of time and in the opposite direction at another segment, canceling one another out. Though either scenario is possible, if evolutionary change was in one direction but there is no evidence of selection, the end result of an oscillating direction of natural selection is the same as if there was no selection on the trait at all. Observed overall change was caused by a correlated response to selection on other traits that covaried with fitness to a larger extent, not direct selection on that trait. 3. Materials and methods 3.1. Patterns of variation and covariation The genetic basis of covariation between traits that is important for evolution is quantified in the additive genetic variance/



covariance matrix, or G matrix. G is notoriously difficult to estimate and accurate estimation of this matrix generally requires data from hundreds of individuals of known relatedness (Steppan et al., 2002 and references contained within). Following multiple other analyses (e.g., Ackermann and Cheverud, 2004; Marroig et al., 2009; Porto et al., 2009; Rolian et al., 2010), we used the relatively easy to estimate phenotypic variance/covariance matrix (P) as a substitute for the G matrix. As opposed to G, accurate estimates of P do not require pedigreed individuals or experimental organismsdaccuracy is only a result of sample size and the number of traits. The substitution of P for G in evolutionary quantitative genetic studies of morphology is a common (Cheverud, 1995; Roff, 1995, 1996; Marroig et al., 2009; Porto et al., 2009; Garcia et al., 2014) but not universally accepted practice (e.g., Willis et al., 1991; Kruuk et al., 2008). Cheverud (1988) noted that as the effective number of individuals increases across studies (equivalent to the number of perfectly estimated breeding values needed to explain the genetic variation), the more P and G are proportional in a manner that suggests P is a scalar multiple of G. Other studies have replicated these results precisely (Roff, 1995; Roseman, 2012), indicating that P and G are probably very nearly proportional for morphological phenotypes. Moreover, many of the criticisms of this assumption are likely misplaced (e.g., Kruuk et al., 2008) because they fail to account for sampling (Roseman, 2012). In this analysis, as is the case in many applications of evolutionary quantitative genetic theory to problems in mammalian evolution, we take P as our best possible estimate of G and there is every reason to believe that P will approximate G to within an acceptable range of tolerances. As fossil P matrices cannot be constructed because of small sample sizes (Ackermann, 2009), our analysis substituted modern human P matrices for all fossil transitions included here, and both modern human and chimpanzee P matrices for the first transition as stand-ins for fossil variation (Ackermann and Cheverud, 2004). We also assumed a constant P over evolutionary time (Lande, 1980). Both of these assumptions are supported for the hip, as patterns are extremely similar in all non-human great apes and the differences seen when compared to modern humans appear to be the result of

5

selection early in the hominin lineage (Grabowski et al., 2011; Grabowski, 2013). 3.2. Measurements and samples used to quantify P matrices To estimate P matrices, 3D landmark data from the hip of common chimpanzees (n ¼ 139; males ¼ 68, females ¼ 71) and modern humans (n ¼ 140; males ¼ 81, females ¼ 59) were collected using a Microscribe digitizer (Immersion, San Jose) for homologous bony landmarks. The collections were housed at the National Museum of Natural History (Washington, D.C.), the American Museum of Natural History (New York, N.Y.), the Museum of Comparative Zoology (Cambridge, M.A.), the Cleveland Museum of Natural History (Cleveland, O.H.), the Anthropological Institute and Museum at the University of Zurich (Zurich, Switzerland), the Museum fur Naturkunde (Berlin, Germany), the Royal Museum for Central Africa (Tervuren, Belgium), and the Royal Belgian Institute for Natural Sciences (Brussels, Belgium). The right hip bone and right femur were used when possible. The pelvis was articulated using thin strips of masking tape and rubber bands with the pubes meeting at midline; no compensation for soft tissue at the pubic symphysis was included in this analysis following Tague and Lovejoy (1986; see also Weaver, 2002). From these 3D coordinates, interlandmark distances were calculated following Grabowski et al. (2011) based on previously published measurements (Fig. 1, Table 2). We concentrated on non-overlapping traits (i.e., traits that do not share points of origin or span similar parts of anatomy) when possible, contrary to previous studies, to paint a clear picture of how natural selection acted to change individual traits related to function. A small subset of these distances was taken using digital calipers (Mitutoyo Series 500-, 19X-, 20, Plymouth, MI) and a GPM anthropometer (Siber Hegner and Co., Zurich, Switzerland), as these were difficult or impossible to take using 3D methods (Fig. 1). To quantify measurement error, we took repeated measurements on a subset of each species in this analysis and calculated the among-group variance component where groups were each round of repeated measurements (Sokal and Rohlf, 1995). All measurements were at least 93% repeatable, with

Table 2 Description of each measurement used in this study with references. Measurementa Anterior margin of iliac blade: H1 Posterior margin of iliac blade: H2 Auricular surface length: H3 Retroauricular height: H4 Lateral iliac breadth: H5 Lower iliac height: H6 Pubis length: H7 Pubic symphysis length: H8 Biomechanical moment arm of the ischium: H9

Femur maximum length: F10 Biomechanical neck length: F11

Femoral head diameter: F12 Bi-iliac breadth: O13 Bi-acetabular breadth: O14 a

Description Anterior superior iliac spine to anterior inferior iliac spine Posterior superior iliac spine to posterior inferior iliac spine Inferior-medial border auricular surface to superior-medial border auricular surface Most lateral point on auricular surface to medial edge of ilium Most lateral point on auricular surface to ilium's lateral margin taken at the location of maximum breadth Center acetabulum to most caudal point on auricular surface Center acetabulum to superior-medial border pubic symphysis Superior-medial border pubic symphysis to inferior-medial border pubic symphysis Center acetabulum to distance between center of the hamstring surface of ischial tuberosity in homininsdas the latter feature is not discernible in chimpanzees, it was modified to the overall length of the ischium taken at the point farthest away from the center of the acetabulum along the angle of ischial shank. Most proximal point on the head to the distal extremity measured parallel to the shaft Distance from the lateral aspect of the greater trochanter to the medial aspect of the femoral head, minus 1/2 super-inferior diameter of the femoral head Supero-inferior diameter of the femoral head Maximum distance between iliac crests. In modern humans this is also known as bi-cristal breadth Minimum distance between the centers of the acetabula

Reference Robinson, 1972 McHenry, 1975 Robinson, 1972 Ward, 1991; Lovejoy et al., 2009a,b Ward, 1991; Lovejoy et al., 2009a,b Ward, 1991; Lovejoy et al., 2009a,b Robinson, 1972 Robinson, 1972 Lovejoy et al., 1973; Stern and Susman, 1983

Martin and Saller, 1957 Richmond and Jungers, 2008

Richmond and Jungers, 2008 Tague, 1991 Tague, 1991

Letter/number combination refers to traits shown in Fig. 1.


6


the majority above 95%. In addition to estimates of P, these data provide well-estimated means of both taxa for the traits in Table 1. 3.3. Our fossil sample The number of relatively well-preserved fossil pelves (with one or both hip bones and a sacrum) is in the single digits, and only a few of these have associated femora that are also reasonably well preserved. Fossil pelves that meet the first criteria for completeness include A.L. 288-1 (Au. afarensis), Sts 14 (Australopithecus africanus), MH 2 (Australopithecus sediba), BSN-49/P27 (H. erectus, but see below), Sima de los Huesos Pelvis 1 (SH Pelvis 1, Homo heidelbergensis), and Kebara 2 (Homo neanderthalensis, for expanded list of less complete fossils and complete references see Supplementary Online Material [SOM] Excel File). This study focused on A.L. 288-1, SH Pelvis 1, and Kebara 2. This sample was chosen for completeness and availability, and for the large span of geologic time separating these fossils (see Table 3 and SOM Excel File). Taken together, the fossils reveal a set of traits that distinguish them as hominins, but also diverge from the morphology of modern humans (Stern and Susman, 1983; Rak and Arensburg, 1987; Arsuaga et al., 1999; Weaver and Hublin, 2009; Bonmatí et al., 2010). Whether these differences had behavioral implications is a topic of much research (Lovejoy et al., 1973; Stern and Susman, 1983). Here, differences in behavior, whether subtle or significant, will not affect the results as this analysis concentrates on evolutionary processes that led to similarities and differences in morphology. The relatively complete fossil KNM-WT 15000, attributed to H. erectus, might seem an obvious choice for inclusion here, but this pelvis is quite fragmentary, completely lacks a pubis, and is a juvenile. Another issue with including KNM-WT 15000 in the present analysis is the fairly complete but quite diminutive 0.9e1.4 Ma BOU-VP-1/1 H. erectus pelvis (but see Ruff, 2010; Simpson et al., 2014) from Gona, Ethiopia. This pelvis has a suite of traits shared with earlier australopiths (e.g., A.L. 288-1, Sts 14), as well other early Homo individuals (e.g., KNM-ER 3228, OH 28), middle Pleistocene Homo (SH Pelvis 1), and later Neandertals (Kebara 2), including laterally flaring iliac blades and long pubic rami, in contrast to the most well-known reconstruction of KNM-WT 15000 (Walker and Ruff, 1993). Based on these findings, some researchers have suggested that these and other ancestral hip traits were present up to and including H. erectus (Ruff, 1995), and only evolved to the form found in modern humans after our lineage split from our common ancestor with the Neandertals (Arsuaga et al., 1999; Simpson et al., 2008; Weaver and Hublin, 2009). The evolutionary transitions included here are between a Pan troglodytes-like LCA to A.L. 288-1, A.L. 288-1 to SH Pelvis 1, SH Pelvis 1 to modern humans (MHS), and SH Pelvis 1 to Kebara 2. We note that while SH Pelvis 1 may in fact be from a group ancestral to Neandertals and not to modern humans (Arsuaga et al., 2014), we believe the substantial similarities between this fossil and early Homo pelves (Arsuaga et al., 1999) make it currently the best candidate for the morphology of the common ancestor of modern humans and Neandertals. For the purpose of this analysis,

chimpanzee hip bone morphology will be substituted for that of the LCA and will be referred to as the LCA here. Though chimpanzees have evolved independently since their LCA with hominins, African apes to the exclusion of humans share a large number of morphological and locomotor similarities (Pilbeam, 1996; Rolian et al., 2010), including in the pelvis (Robinson, 1972; Sigmon, 1974, 1975). Because of this, morphological analyses regularly take a non-human great ape-like LCA as the starting point of comparison for the presence, absence, and expression levels of traits hypothesized as playing functional roles in bipedalism (e.g., Stern and Susman, 1983; Lovejoy et al., 1999). 3.4. Fossil measurements Complementary fossil measurements used in this analysis were taken from the literature, combined with a small number taken by the lead author from high-quality casts that have not been previously published. Additional measurements were kindly provided by A. Bonmatí and J.L. Arsuaga on the reconstruction of the original SH Pelvis 1 (Table 4, see SOM Excel File Sheet 4 for complete citations). Even though SH Pelvis 1 is astonishingly complete, no femora were recovered in articulation with this fossil, though many femora have been recovered from the site (Bonmatí et al., 2010). Bonmatí et al. (2010) discussed three femora from Sima de los Huesos that are compatible with SH Pelvis 1. Here, we took the conservative approach and used the shortest in length (Femur X, Table 4), with its published femoral head diameter, in this analysis. As biomechanical neck length for this sample of hominins has not yet been published (Trait F11, Fig. 1 and Table 1), the results using this fossil as beginning and end points were accomplished without the inclusion of this trait. Along these lines, there is no published material for the lower limb associated with Kebara 2, though some partial lower postcranial remains were recovered as mentioned in the original description of this material (Bar-Yosef et al., 1986). To include the complete set of hip traits we included the measurement of femoral supero-inferior head diameter and maximum length from the La Ferrassie I Neandertal (Rosenberg et al., 1988). This individual was chosen based on its extremely similar acetabular diameter to Kebara 2 (Rosenberg et al., 1988 and Grabowski, pers. obs.), the size of which was shown to have a strong relationship with overall body size (McHenry, 1992) as well as femoral length in fossil hominins (Simpson et al., 2008), and we feel the value of including the hip of this individual outweighs any possible error caused by over- or under-estimating femur length and femoral head diameter. 3.5. Using individual fossils as species means It should be noted that though the means of the extant species, chimpanzees and modern humans, were calculated based on a large number of individuals (see below), the fossil specimens served as stand-ins for their population or species means (Table 4). While an individual fossil is an unbiased estimate of the mean of the population from which it is drawn, it is an imprecise estimate. If the fossil individuals were outliers with regards to their groups, this

Table 3 Fossil sample included in analysis to complement published measurements. Fossil

Species

Age (Ma)

Sex?

Location of discovery

Reconstruction, location of casta IHO CastdLovejoy (1979), UM Bonmatí et al. (2010) reconstruction, UCM-ISCIII Institute for Human Evolution, Spain Rak (1991), UM

A.L. 288-1 SH Pelvis 1

Au. afarensis H. heidelbergensis

3.18 0.43

F M

Hadar Formation, central Afar, Ethiopia Sima de los Huesos, Spain

Kebara 2

H. neanderthalensis

0.06

M

Kebara Cave, Mt. Carmel, Israel

a

IHO ¼ Institute of Human Origins, UM ¼ University of Missouri.



7

Table 4 Measurement averages for our sample of common chimpanzees, modern humans, and measurements for fossil hominins.a Ilium Sample Chimpanzee average (n ¼ 139) Chimpanzee range e lowest value for each trait Chimpanzee range e highest value for each trait Chimpanzee female (smallest in sample) Modern human average (n ¼ 140) Modern human range e lowest value for each trait Modern human range e highest value for each trait A.L. 288-1 A.L. 288-1 (scaled up using modern human model)c A.L. 288-1 (scaled up using chimpanzee model)c SH Pelvis 1d Kebara 2

Pubis

H2

H3

H4

H5

H6

H7

H8

H9

107.97 86.97 133.52 88.47 41.56 28.24 55.24 51.22 53.71 55.05 43.60 57.46

91.88 71.62 121.65 76.46 48.94 27.57 66.67 33.24 36.86 36.88 36.60 73.75

52.73 35.77 75.63 48.59 59.26 43.31 79.19 39.18 48.02 47.17 65.80 64.70

29.92 22.29 39.63 28.77 69.65 53.77 86.04 42.61 51.66 47.31 66.20 62.60

92.79 69.03 108.24 77.34 91.33 74.24 107.78 86.85 95.29 106.05 104.00 95.89

88.87 67.94 111.71 79.54 77.68 58.21 101.35 49.06 51.75 51.86 66.90 56.12

69.02 57.98 83.78 57.98 84.13 62.89 102.87 67.60 70.47 69.41 98.60 99.06

49.66 34.32 67.59 43.74 37.89 25.46 53.29 25.20b 32.33 29.15 53.80 53.85

80.91 67.85 94.66 69.80 62.06 45.63 80.14 47.70b 59.05 54.57 70.50 54.71

Femoral Sample Chimpanzee average (n ¼ 139) Chimpanzee range e lowest value for each trait Chimpanzee range e highest value for each trait Chimpanzee female (smallest in sample) Modern human average (n ¼ 140) Modern human range e lowest value for each trait Modern human range e highest value for each trait A.L. 288-1 A.L. 288-1 (scaled up using modern human model)c A.L. 288-1 (scaled up using chimpanzee model)c SH Pelvis 1d Kebara 2

Ischium

H1

Pelvic

F10

F11

F12

O13

O14

294.96 253.00 328.20 253.00 444.62 362.00 537.00 280.00b 318.57 312.22 458.00e 458.00g

66.52 53.02 79.06 53.02 90.16 74.19 113.84 47.30b 56.67 58.07 e e

33.23 28.54 38.97 29.10 45.40 37.77 54.56 28.60b 35.86 35.86 51.80f 54.50g

254.31 196.53 289.47 224.30 257.77 217.88 300.43 268.30b 292.81 307.78 335.00b 313.00b

96.25 74.75 121.77 90.43 119.81 93.36 143.52 118.00b 121.14 128.38 134.90b 129.00b

a

Trait numbers refer to traits in Fig. 1 and Tables 1 and 2. Measurements marked were taken from the literature (see SOM Excel File Sheet 4 for references). A.L. 288-1 scaled up to the size of the average Au. afarensis based on currently described fossil sample average of femoral head diameter given a modern human or chimpanzee model of scalingdsee text. d Measurements for SH Pelvis 1 from A. Bonmati and J.L. Arsuaga (pers. comm.) except where indicated. e From SH Femur Xdsee text. f Calculated using regression from Plavcan et al. (2014) based on acetabular diameter from Bonmatí et al. (2010). g From La Ferrassie Idsee text. All other measurements were taken for this analysis (see SOM Excel File Sheets 3 and 4 for additional measurements on a wider fossil sample and full references). b c

could affect the results of this analysis. Building on this, A.L. 288-1 appears to be relatively small for Au. afarensis (McHenry, 1992) with an estimated body mass of 27.3 kg, far below the estimated species mean (39.1 kg, Grabowski et al., in press), as well as the female average for Pan troglodytes troglodytes (45.8 kg, Smith and Jungers, 1997). To reduce the effect that using the probable outlier A.L. 2881 as the species mean for Au. afarensis could have on our results, we took two different approaches. First, we allometrically scaled the hip of A.L. 288-1 up to that of our currently known sample mean of Au. afarensis based on femoral head diameter (Trait F12). This was accomplished by first regressing all other hip traits on femoral head diameter in our modern human comparative sample. We used this regression equation to predict values for the other hip traits in a modern human given the femoral head diameter (28.6 mm) of A.L. 288-1. This produces a scaled modern human given a femoral head diameter of the same size as A.L. 288-1. We then took the difference between the true trait values of A.L. 288-1 (besides femoral head diameter) and the predicted values and added this difference to the intercept terms of our modern human regression. This produces an “Au. afarensis regression” that assumes that Au. afarensis scaled in a similar fashion as modern humans within its population but differed in intercept terms. We determined the sample average femoral head diameter for Au. afarensis based on the currently described fossils (35.86 mm, data from Grabowski et al., in press) and used our new Au. afarensis regression to predict the other hip dimensions, given that the fossil now sits at that femoral head diameter average. As this method assumes that Au. afarensis had a similar pattern of

allometric scaling as in modern humans, we also checked the effect that a different pattern of scaling might have on our final estimates of selection by repeating the same procedure but using the common chimpanzee comparative sample. For our second approach, we chose the smallest female chimpanzee as our LCA based on femoral length but kept A.L 288-1 at original size. While this might remove issues related to size being an important component of the first transition, there remains an apparently large increase in overall size between A.L. 288-1 and SH Pelvis 1, which is noted below. In addition to there being issues with using single fossils as stand-ins for population means; sex differences, taphonomic changes, as well as reconstruction errors may affect the results. The point should be made that these same factors would affect comparable analyses that use fossil materials similarly and as such are unavoidable. 3.6. Importance of standardization when estimating natural selection As has been pointed out previously (Houle et al., 2011), how data are treated can have dramatic effects on whether the results of analyses are biologically meaningful. As this study seeks to compare amounts of direct selection on and the amount of change in traits within individual transitions, and then between transitions, standardization is vital to put all the results on the same scale. The benefits of standardizing variance/covariance matrices and selection pressures by the means of the traits were demonstrated when comparing the strengths of natural selection across


8


studies (Hereford et al., 2004). Mean standardizing solves the problem that different traits may be on different scales, and allows for comparison of proportional changes between traits. But fossil analyses present unique problems. For one, it is not immediately clear which mean to standardize bydstandardizing the mean of the species from which the P matrices are calculated, the ancestral species mean in each transition, or everything by one set of means affects the results. Logging the data has a similar effect as mean standardization (Hereford et al., 2004; Hansen and Houle, 2008) and is arguably more valid for evolutionary transitions such as those included here. Standardization of selection pressures comes down to a philosophicalebiological issuedhow should we compare amounts of selection over large evolutionary transitions and traits that differ substantially in size? If everything was kept on an absolute scale, there are three related issues. For one, the b results of Equation (1) would assume that 1 unit of increase in a trait for one transition is equivalent to 1 unit of increase in another transition. Thus, a doubling in size for a mouse (from 10 to 20 g) would be equivalent to a 10 g change in size for an elephant. Another related issue is that while variances and covariances increase with the size of a trait, they do not usually scale in a linear fashion. This means that it would be difficult to predict whether the doubling of size of an elephant would result in a larger b than the doubling of a mouse when things are kept on an absolute scale. The third issue is biologicaldwhat is more likely to be biologically important, a doubling in the size of a mouse or a change of 10 g in an elephant? Using a proportional scale, whether that proportion is to a set mean or logging the data, solves these issues as it puts evolutionary changes on a scale where change is proportional to something else. The question is what is that “something else.” In mean standardization, changes are proportional to one thing, such as the mean of the traits of the LCA. In log standardization, the bs are now proportional to means that change over the course of the evolutionary transition. In other words, what differs is the proportion of what valuedis one set of means more appropriate than a proportion with a mean that is continually updating? Do we think that the amount of selection required to cause a femur to decrease by 25% is equivalent to the amount of selection required to decrease that new length by another 25%, or is it only equivalent to selection that causes it to decrease in proportion to a set absolute value? In an extreme example, would the amount of selection required to increase a mouse by 100% be equivalent to that required to increase an elephant by 100%? In the log scaled case they would be, given that the log-scaled P matrices were the same. On the other hand, standardizing by either the mouse or elephant means would lead to the amount of change, and thus the selection pressures, of one of the two transitions being substantially smaller than the other, and thus not comparable. For example, if the elephant means were used to standardize, the bs for both the evolutionary transition that resulted in the doubling of the mouse and elephant would be the relative change in fitness given the proportional change in 1 unit of elephant. Moving to the biological meaning of b, would increasing an elephant by a factor of two have a more drastic change on its fitness than increasing a mouse by two? While many would say yes, this depends on the scale, and we are thinking of these size changes in reference to ourselves. How a doubled mouse interacts with their world when compared to an originally sized mouse might be more different than how a doubled elephant interacts with theirs when compared to an originally sized elephant, and this could have a larger impact on the fitness of the mouse. In order to make the comparisons of selection comparable both within and between transitions, we believe logging, or looking at amounts of selection standardized by a continually updating mean,

is the correct way to show proportional change and compare relative fitness. Taking the natural log of all data before calculation of P matrices as well as the response to selection has two results. P matrices calculated from logged data are approximately equivalent to P matrices scaled by the mean of the species that was used in their constructiondi.e., mean standardized P matrices. The big difference comes when using logged data to calculate the Dzdthe difference between the fossil and its ancestor. As mentioned above, in the mean standardization case, the difference between the fossil and its ancestor would be scaled by the ancestor or whatever one chooses to scale by. This results in a term Dz that shows change as the proportion of that mean. In the log scaled case, the difference between the logged fossil and log ancestor is equivalent to breaking the evolutionary transition into a multitude of small steps, calculating the amount of change for each step, dividing by the mean at the beginning of that step, and then taking the sum of those proportions. This is equivalent to mean standardization but the mean is constantly updating along the length of the evolutionary transition (Hansen, pers. comm.). Note that the estimates of selection pressures after log scaling the data have no units. This is true for both the mean-standardized and log-scaled cases, as this enables comparison across different traits, species, or scales of measurement. To compare both the amount of selection and amount of change among traits and among transitions, we took the natural log of the data as the first step in our analysis. This places the direct selection pressures and the amount of evolutionary change on the same scale, making comparisons of pattern (i.e., evolutionary change) and process (i.e., natural selection) possible.

3.7. Data preparation This analysis focuses on within-population variation and covariation, and therefore, it is necessary to remove sources of variation unrelated to the underlying genetic architecture before estimating these patterns. This requires holding constant sources of variation related to sex in the case of chimpanzees, and sex and population in the case of humans, as our broad geographic range has consequences for body shape and postcranial morphology (Ruff, 1995; Pearson, 2000; Weaver, 2003). Variation caused by sex and/or sex and population was controlled for before P matrices were calculated using the residual covariance from a MANOVA with the 14 traits as the dependent variable and sex and/or population as the independent variables (Ackermann and Cheverud, 2000; Marroig et al., 2009; Porto et al., 2009). P matrices will be provided on request to MG.

4. Analyses 4.1. Estimating past selection pressures The average evolutionary response of a population to selection is given by the Lande equation (Lande and Arnold, 1983):

Dz ¼ Gb

(1)

Here Dz is the mean evolutionary response of a population or species to selection (or in this case the difference in means between species or fossils), G is the genetic variance/covariance matrix, and b is the vector of directional selection gradients. As discussed above, P was substituted for G in this analysis. This calculation was done for each of the evolutionary transitions included here.



4.2. Comparing overall directions and magnitudes of selection Selection vectors between the taxa in this analysis are 14 dimensional (i.e., the number of traits), but their magnitudes (or lengths) can be calculated as the norm of the vector. Notation follows that of above.

kbk ¼

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi b21 þ … þ b2k

(2)

This formula results in a single number representing the combined magnitudes of each of the traits. In addition, each vector has a direction, which represents the direction selection is pushing the mean of taxa one to evolve into taxa two. We can compare the directions of vectors between subsequent evolutionary transitions using the formula for the angle between two vectors (Blows and Walsh, 2009). Notation follows that of above.

9

would be represented as a vertical line. The second transition, between Fossils 2 and 3, would be represented as a line that diverges from the direction of selection required to move the morphology between Fossils 1 and 2. The amount it diverges is based on the angle between vectors 1 and 2. In our representation of selection, the end points of each segment mark a particular fossil. The length of each segment expresses the overall relative magnitude of selection for that transitiondthe lengths of all segments are scaled so the longest segment given either the chimpanzee or human P has a value of one to allow comparison both between individual transitions within the same P, as well as between Ps. One important point with this method is that the direction of selection for one transition can only be compared to the transition directly before and after it, except in the case where both end points originate in the same fossil ancestor (see Fig. 4). 4.3. Estimating rates of morphological evolution

1

q ¼ cos

Dzt b kDzk*kbk

(3)

Here we use the combination of length and angular data to graphically represent evolution, showing how both the magnitude and direction of selection changed over time for each evolutionary transition. An example will help clarify our approach. Let us say we have three fossils, Fossils 1, 2, and 3, in order from oldest to youngest. In our formulation of this representation, the magnitude and direction of selection for the transition between Fossils 1 and 2

Rates of evolution for phenotypic traits were calculated following the equation (Lynch, 1990):

D ¼ varB ðlnzÞ=½tvarW ðlnzÞ

(4)

where varB ðlnzÞ and varW ðlnzÞ are the observed between and within-species phenotypic variance for the log transformed hip data, t is the estimated number of generations separating the two species, and D is an estimate of the rate of morphological evolution.

Figure 2. Evolutionary transitions: LCA to A.L. 288-1 (A, B), A.L. 288-1 to SH Pelvis 1 (C, D). Percent change in individual traits in response to selection is shown on the left (A, C). Direct selection pressures required to produce changes on the left with standard error bars is shown on the right (B, D). Numbers on the x-axis correspond to traits in Fig. 1. Gray bars reveal traits that were under a significant amount of selection, white bars an amount of selection that was not significant, and dark gray bars traits where the direction of selection is opposite the direction of change. Note that bs are on the log scale to allow for comparisons and thus, do not have units.


10


In the simplest terms, this equation first scales the amount of variation between populations by the amount of variation within populations and then standardizes this by time. This model is directly related to population genetics as is can roughly estimate how the calculated rate of evolution compares to the expectation under neutral evolutiondwhere drift and neutral mutation are the only evolutionary forces acting on a population (Lynch, 1990). This can be accomplished because under the neutral expectation the expected between-species variance is equal to the rate of input of genetic variance due to mutation (Vm) multiplied by t, the number of generations separating descendent species (Lynch and Hill, 1986). This means that the results of Equation (4) are comparable to the rate of input of genetic variance due to mutation (Vm) scaled by the phenotypic variance (Vp)din other words, Vm/Vp. Based on a broad array of taxa and character types, Lynch (1988) found that Vm/Ve generally falls into the range of 104 e 102. Based on the assumption that heritability is equal to 0.5 (Lynch, 1990), Va ¼ Ve, and the lower limit for the amount of change expected under neutral evolution is Vm/Vp ¼ Vm/(Ve*2) ¼ 104/2 ¼ 5 105. The upper limit determined in the same fashion would be 5 103. A D statistic within this range suggests evolution is no faster (or slower) than expected under neutral evolution. Lower D values could suggest stabilizing selection opposed divergence, higher D values could suggest diversifying natural selection. Equations for varB ðlnzÞ and varW ðlnzÞ can be found in Lynch (1990). As t requires an estimate of the total number of generations separating the two species, we assumed a chimpanzee generation

time of 25 years (Langergraber et al., 2012) for the LCA and australopiths, with a modern human generation time of 29 years (Langergraber et al., 2012) assumed for all Homo groups. Given two different generation times for a transition, the mean generation time is used. Accordingly, estimates of within-species variances for each transition follow chimpanzees for the LCA and A.L. 288-1, and modern humans for all Homo taxa. These results assume that the last common ancestor of humans and chimpanzees lived seven million years ago, the low end of the range suggested by revisions in Langergraber et al. (2012). In addition to the evolutionary transitions explored here, we include comparative data from Lynch (1990) on evolutionary rates for other mammalian groups and cranial diversification in modern humans. All standard errors and tests for significant difference were estimated using a parametric bootstrap technique (Efron and Tibshirani, 1993). All analyses were conducted using programs written in R (R Development Core Team, 2011) by MG.

5. Results and preliminary discussion 5.1. Estimating past selection pressures Reconstructing past selection pressures for each of the transitions included here (Figs. 2 and 3, SOM Figs. 1 and 2) show that the majority of traits have changed due to strong natural selection and not a correlated response to selection on other traits.

Figure 3. Evolutionary transitions: SH Pelvis 1 to MHS (A, B), SH Pelvis 1 to Kebara 2 (C, D). Percent change in individual traits in response to selection is shown on the left (A, C). Direct selection pressures required to produce changes on the left with standard error bars is shown on the right (B, D). Numbers on the x-axis correspond to traits in Fig. 1. Gray bars reveal traits that were under a significant amount of selection, white bars an amount of selection that was not significant, and dark gray bars traits where the direction of selection is opposite the direction of change. Note that bs are on the log scale to allow for comparisons and thus, do not have units.



5.1.1. Last Common Ancestor to A.L. 288-1 (Fig. 2A, B) The pattern of evolutionary change and the pattern of selection are quite similar here. The trait under the strongest selection during this transition was Retroauricular Height (Trait 4), and the trait that evolved proportionally the most was the same trait (Trait 4: 74% increase). Only Bi-iliac Breadth (Trait 13: 15% increase) appears to have changed a substantial amount but does not show signatures of strong natural selection. Comparing the results of this transition to one using a chimpanzee P and A.L. 288-1 scaled up following the chimpanzee model (SOM Fig. 1A, B) reveals similarity in overall patterns of selection, with a few additional traits that changed substantially but show no signature of strong direct selection: Lateral Iliac Breadth (Trait 5: 15% increase), Pubic Symphysis Length (Trait 8: 41% decrease), and Bi-acetabular Breadth (Trait 14: 34% increase). 5.1.2. A.L. 288-1 to SH Pelvis 1 (Fig. 2C, D) Here, a number of traits changed a substantial amount but do not show signatures of strong natural selection (unshaded bars). These were the Auricular Surface Length (Trait 3: 37% increase), Retroauricular Height (Trait 4: 28% increase), Pubic Symphysis Length (Trait 8: 66% increase), Biomechanical Moment Arm of the Ischium (Trait 9: 19% increase), Bi-iliac Breadth (Trait 13: 14% increase), and Bi-acetabular Breadth (Trait 14: 11% increase). In addition, this is the one transition where the direction of change and the direction of selection do not coincide (dark gray trait in Fig. 2C, D). Lateral Iliac Breadth (Trait 5: 9% increase) is increasing in size, but there actually is substantial direct selection in the opposite direction. The traits under about equally the strongest selection during this transition were Femur Maximum Length (Trait 10) and Femoral Head Diameter (Trait 12), with Pubis Length (Trait 7) falling slightly below these. The trait that evolved the most by proportion was Pubic Symphysis Length (Trait 8: 66% increase). 5.1.3. SH Pelvis 1 to MHS (Fig. 3A, B) Though the overall pattern of evolutionary change and selection are similar, there were a fair

11

number of traits that changed substantially but do not show signatures of strong directional selection. These include Lateral Iliac Breadth (Trait 5: 12% decrease), Pubis Length (Trait 7: 15% decrease), Biomechanical Moment Arm of the Ischium (Trait 9: 13% decrease), and Bi-acetabular Breadth (Trait 14: 12% decrease). The trait under the strongest selection for this transition was Bi-iliac Breadth (Trait 13), while the trait that evolved the most was the Posterior Margin of the Iliac Blade (Trait 2: 32% increase). 5.1.4. SH Pelvis 1 to Kebara 2 (Fig. 3C, D) Comparing this transition to the SH Pelvis 1 to Modern Humans (Fig. 3A, B) reveals distinct differences in both the patterns of selection and the patterns of change. Here the amount of evolutionary change is dominated by an increase in the length of the Posterior Margin of the Iliac Blade (Trait 2), which increased over 100%. On the other hand the absolute amount of selection responsible for this change was no more than that seen for a number of other traits here, including Retroauricular Height (Trait 4: 5% decrease) or Biomechanical Moment Arm of the Ischium (Trait 9: 22% decrease). The trait under the strongest selection for this transition was Retroauricular Height (Trait 4). The b results for the first two transitions using the smallest chimpanzee in our sample as a stand-in for the LCA and A.L. 288-1 at the fossil's original size (SOM Figs. 1C, D and 2) are similar in pattern and generally only differ in magnitude of selection from previous results (Fig. 2). 5.2. Overall magnitudes and directions of selection The overall direction of selection for each evolutionary transition changed dramatically when compared to the previous transition (Fig. 4). This result was true regardless of whether a chimpanzee or modern human model of variation was used (SOM Fig. 3), though the overall magnitude of selection required for each transition appears to be greater given a chimpanzee model. This result is likely due to the lower magnitude of evolutionary constraint and greater evolvability in hip anatomy of modern humans than chimpanzees (Grabowski et al., 2011; Grabowski, 2013). It is clear that there was a reduction in the overall strength or magnitude of selection required for each transition after the first, with a significant reduction for the transitions from SH Pelvis 1 to modern humans and Kebara 2. Those same transitions are similar in the overall amount of selection, but dramatically diverge in the direction of selection. This result complements our findings above that show substantial differences in the patterns of selection on individual traits for these two transitions. Statistics used in construction of these graphics are given in SOM Table 1. 5.3. Rates of evolution

Figure 4. Figure comparing overall magnitude and direction of selection required for each evolutionary transition included here. Length of vector shows magnitude of selection for each transition. Angles show angle between hypothetical continuation of selection vector shown in dotted line and second selection vector. Results are given using a modern human model of variation. For the SH Pelvis 1 to modern human (MHS) and SH Pelvis 1 to Kebara 2 transitions, inner angles show difference between the hypothetical vector and the direction for each new species, the outer angle shows the difference between MHS and Kebara 2. Note that the direction of selection for one transition can only be compared to the transition directly before and after it, except in the case where both end points originate in the same fossil ancestor.

Although the transition between the LCA and A.L. 288-1 was the most dramatic morphologically, these two points span the longest period of timedhere assumed to be 4 Madand therefore it appears slowest overall in terms of the rate of evolution (Fig. 5). It was also the only transition included here that fell below the minimum neutral expectation (horizontal gray bar in Fig. 5 with the minimum at 5.05 105), but as the value of this transition is 5.03 105, this difference was likely not significant. However, it is important to underscore here that the known fossil record currently lacks complete and undistorted published pelves prior to A.L. 288-1. Thus, rates may well prove to be much faster depending on the currently unknown nature of hominin hip morphological change between the LCA and Au. afarensis. Additionally, less weight should be put on this result as debate surrounds the date of the LCA of humans and chimpanzees, with estimates ranging from as little as


12


Figure 5. Average rate of evolution for evolutionary transitions including A.L. 288-1 to “Homo erectus” that assumes that the morphology of SH Pelvis 1 is consistent with H. erectus at 1.9 Ma compared with rates of morphological evolution in various mammals from Lynch (1990). Gray box shows lower limits of expected range of values under one neutral model of evolution.

4 Ma (Hobolth et al., 2011) to at least 7e8 Ma (Langergraber et al., 2012), as well as the morphology of the LCA (Lovejoy et al., 2009a; Morgan et al., 2015). A similarly slow rate of change was found for the A.L. 288-1 to SH Pelvis 1 transition, but if we assume that SH Pelvis 1 displays morphology consistent with H. erectus at 1.9 Ma (see above), the rate of change is more or less consistent with that of the later hominin evolutionary transitions included here. All other transitions fall between 1.2 and 1.6 104, which is comfortably within the rate of change given our model of neutral evolution. On the other hand, the rate of evolutionary change for these transitions was far greater than any other mammalian group included here, with the exception of cranial diversification in modern humans. 6. Discussion Our results represent the first quantitative attempt to test the relationship between function and selection as it relates to bipedalism. Overall, our findings argue that a complex and changing pattern of natural selection drove hominin hip evolution, and many, but not all, traits hypothesized to play functional roles in bipedalism evolved as a result of natural selection. The results of this analysis make a strong case that character changes that have been specifically linked to locomotor performance through biomechanical and/or experimental analyses (Sockol et al., 2007; Pontzer et al., 2009) have evolved as a result of selection for bipedalism, though selection pressures for other behaviors were undoubtedly present at all points in hominin evolution (see below). On the other hand, the rate of evolutionary change for all transitions explored here does not exceed the amount expected if evolution was occurring solely through neutral processes. 6.1. Selection on individual traits Our results do not support the picture of natural selection predicted by Lovejoy and colleagues (e.g., 1999). Instead of selection on a few traits leading to many correlated responses, a complex and changing pattern of selection on numerous traits was apparently responsible for the majority of hip evolution (Figs. 2 and 3). Neutral evolutionary processes cannot be ruled out either. Notably lower iliac

height, the focus of selection in Lovejoy's (1999) evolutionary model of morphological evolution, was never the character under the largest amount of selection in any of the transitions included here. In fact, for the first transition (Fig. 2A, B), this character was under a relatively smaller amount of selection than many other traits, including pubic symphysis length. Further support for an evolutionary model where different pelvic regions evolved with some independence from each other (Grabowski et al., 2011) comes from the apparently mosaic nature of the recently described A. sediba pelvis (Kibii et al., 2011). The pelves from the two individuals currently described (MH1 and MH2) demonstrate some morphological elements consistent with australopiths and some with early Homo. While the direction of evolutionary change often matches the direction of selection across the transitions included here, the magnitude of selection frequently does not correspond to the magnitude of change. This can be seen in the differences between the traits that were under the strongest selection for each transition and which traits evolved the most as discussed in the results section. Additionally, in each transition after the first there are a number of traits that apparently changed substantially by a correlated response to selection on other traits. The cause of these differences between pattern and process is the underlying relationships among traits. The implications of these results for paleoanthropology are that though morphological changes can inform hypotheses of past selective forces, underlying relationships between traits must be taken into account when attempting to understand the role of selection in hominin evolution. Ignoring these relationships can have major implications for hypotheses built on the second link of the form-function-selection chain. For example, bi-acetabular breadth changed substantially in both the A.L. 288-1 to SH Pelvis 1 and SH Pelvis 1 to modern human transitions (increasing in the first, decreasing in the second), but was not under strong natural selection in either. While one could argue that this result was because of error in our estimates of selection, shown in the size of our standard error bars, there is often little similarity between the amounts of selection and the amounts of evolutionary change. There are numerous traits that changed proportionally less than bi-acetabular breadth, but the amount of selection that caused those changes is still significant. Focusing on one particular transition, though the decrease in bi-acetabular



breadth between SH Pelvis 1 and modern humans was substantial (a 12% decrease), there is no signature of strong natural selection (i.e., trait 14 in Fig. 3A, B). It appears most likely here that biacetabular breadth changed as a correlated response to selection to change bi-iliac breadth, which was both the focus of substantial amounts of selection and substantial amounts of change during this transition, in conjunction with correlated selection from other traits. Taken together, these results contrast with previous suggestions that the reduction in bi-acetabular breadth seen in the shift from australopiths to Homo was the result of selection for less costly locomotion (Berge, 1994), endurance running (Bramble and Lieberman, 2004), or to allow birth of a more encephalized neonate (Tague and Lovejoy, 1986; Ruff, 1995). This change may have influenced all of those behaviors, but it evolved as a result of selection on other traits, rather than direct selection for a particular function. Other important morphological changes between taxa appear to have been the result of correlated evolution. During the transition from A.L. 288-1 to SH Pelvis 1, both auricular surface length and retroauricular height increased in size but do not show the signature of strong natural selectiondthey apparently evolved as a correlated effect of selection on other traits. This result contrasts with previous suggestions that the expansion of the auricular surface and retroauricular region in Homo when compared to earlier australopiths was the result of selection for increased trunk stabilization during endurance walking and running (Bramble and Lieberman, 2004). These traits may serve to diagnose Homo pelves from australopiths (Simpson et al., 2014), but they apparently evolved as a result of selection on other traits. If these changes, along with changes in bi-acetabular breadth, were not caused by selection, by definition they cannot be considered adaptations (Williams, 1996). Thus, these results could have major implications for the hypothesis that selection for endurance running coincided with the origins of Homo (Bramble and Lieberman, 2004), as it is based on the assumption of a link between function and selection in all of these traits. Finally, during the transition from SH Pelvis 1 to modern humans, lateral iliac breadth, pubis length, and biomechanical moment arm of the ischium all decreased in size but do not show the signature of strong natural selection. While researchers generally agree that hominins from H. erectus onwards did not differ in locomotor mechanics, it is intriguing that these substantial changes in hip morphology were the result of correlated effects of selection on other traits. 6.2. Traits associated with locomotor efficiency Pontzer et al. (2009) argued that the majority of postcranial changes seen between early hominins that were hypothesized to relate to locomotor efficiency played relatively small roles. This was based on the argument that almost all interspecific variation could be accounted for by changes in lower limb length, effective mechanical advantage, and muscle fascicle length. When we look at the two traits in our analysis that correspond to those discussed in Pontzer et al. (2009)dlower limb length, represented here as femoral length, and ischium lengthdwe see a slight relationship between natural selection and traits thought to be overwhelmingly important for locomotor efficiency. Lower limb length shows a signature of strong natural selection for the one transition included here where the lower limb changed substantiallydA.L. 288-1 to SH Pelvis 1 (Fig. 2). On the other hand, the amount of selection on this trait is equal to that of the amount of selection on femoral head diameter, and only slightly more than that on pubis length. A similar finding can be seen for the length of the ischium for the LCA to A.L. 288-1 transitiondthis trait was under strong, but not the

13

strongest, selection to reduce in size, and numerous other traits were under strong selection. While these results do not argue against using experimental approaches to build selective hypotheses, or against the hypothesis that selection for less costly locomotion was one of the driving forces for the adoption and persistence of bipedalism, they do point out the error in thinking only traits that play a significant role in a particular behavior were under strong natural selection. For the hip, selective forces other than locomotor efficiency are undoubtedly at play. Pontzer (2012) suggested that the strongest candidate for postcranial change resulting from natural selection during the evolution of early Homo was the increase in overall body mass. Ignoring recent findings that suggest that some early australopiths were nearly as large as the largest H. erectus (Haile-Selassie et al., 2010; Grabowski et al., in press), we can say that A.L. 288-1 was substantially smaller than later hominins, and likely other hominins that lived at the same time (McHenry, 1992; Grabowski et al., in press). This was the reason we scaled A.L. 288-1 up to the Au. afarensis sample mean in our main results, and used the original A.L. 288-1 measurements but our smallest female chimpanzee as a stand-in for the LCA in the supplemental. Our results show that all but two traits are increasing in size during the A.L. 288-1 to SH Pelvis 1 transition in either case (Fig. 2C; SOM Fig. 2C). This pattern might be expected if size was the focus of selection. On the other hand, the pattern of selection that produced those changes is distinct from all other transitions explored here, with a number of traits evolving substantial amounts but not showing the signature of strong direct selection and others evolving opposite the direction of selection. Thus, even if overall size was the object of natural selection during this transition, this change was the result of complex patterns of selection working within the evolutionary constraints of existing patterns of variance and covariance. 6.3. Overall magnitudes and directions of selection Our results show that over the course of hominin evolution, the overall direction of selection changed substantially (Fig. 4). The low similarity between the directions of selection among different evolutionary transitions suggests that the ultimate causes of selectiondi.e., different functional requirements related to fitness in response to new conditionsdchanged in a dramatic and frequent fashion during hominin evolution. The initial selection pressures that led from the LCA to A.L. 288-1 are almost exactly orthogonal to those that led from A.L. 288-1 to SH Pelvis 1. In other words, somewhere between 3.18 Ma and 0.43 Ma, the age of these fossils, hominins faced a new set of environmental challenges and/or underwent a behavioral shift within a static environment. Because this is such a long period of time, these results are not surprising, but if we extend the argument that the SH Pelvis 1 reveals a comparable morphology to early Homo, these results take on new meaning. They suggest that a drastic shift in the selective regime occurred at some point during the time period of around 3e2 Ma, including a dramatic increase in the magnitude of selection and a rapid rate of evolutionary change. Such a shift could reflect new selection pressures related to food acquisition, reflecting the shift to more extensive carnivory (Domínguez-Rodrigo, 1997; Braun et al., 2010) or dealing with a changing environment (Demenocal, 2004) or a different mode of locomotion (Bramble and Lieberman, 2004). When evolving from SH Pelvis 1 to modern humans or Kebara 2, the direction of selection was once again nearly orthogonal for both of these transitions when compared to the previous transition. Notably, the directions of selection that produced these transitions substantially differed, meaning that the selective pressures that separated our two lineages were quite distinct. These results are


14


not surprising given that modern humans appear to show body proportions of a species adapted to a warmer climate, including pelvic breadth, and Neandertals are described as being coldadapted (Ruff, 1994; Weaver, 2009). Large changes in the direction of selection across the hominin transitions included here are consistent with changes seen in hominin hip morphology. Changes include moving from the tall and narrow pelvis of a non-human great ape-like LCA to the short but broad pelvis with widely flared iliac blades, widely spaced acetabula, and the long femoral neck of the australopiths (Robinson, 1972; Lovejoy et al., 1973). These changes might have persisted in early and later Homo (Arsuaga et al., 1999; Simpson et al., 2008; Weaver and Hublin, 2009), and contrast with the reduction in iliac flare, reduction in bi-acetabular breadth, wider anteroposterior pelvis, and shorter femoral neck of modern humans. This non-linear and erratic pattern of change is also consistent with changes seen in other anatomical elements, including in the distal humerus (McHenry and Brown, 2008), proximal femur (Richmond and Jungers, 2008), and limb proportions (Hartwig-Scherer and Martin, 1991; Richmond et al., 2002; Haeusler and McHenry, 2004, 2007). While the patterns of change seen between fossil species suggests such a result, our approach provides the first quantitative evidence that complex and temporally changing patterns of selection led to some of the morphological changes seen in hominin evolution. It should be noted that an alternative interpretation of the large changes in the direction of the selection vectors over the shorter evolutionary transitions (i.e., between SH Pelvis 1 and modern humans and SH Pelvis 1 and Kebara 2) is that they may be the result of random genetic drift or sampling error, both of which would result in essentially random transitions. This interpretation concurs with findings that cranial differences between modern humans and Neandertals are consistent with a model of evolution due to drift (Weaver et al., 2007).

6.4. Neutral evolution Pontzer (2012) suggested the relatively small amount of postcranial change between australopiths and early Homo might have been the result of drift and not natural selection. Our results suggest that given the speed at which these transitions are occurring and the model of neutral evolution we use here, evolution by neutral mutations fixing through random genetic drift cannot necessarily be ruled out for any of the evolutionary transitions included here (Fig. 5). This is true even for the dramatic amount of change seen in the first transition, with the rate of evolution only slightly less than predicted given drift alone. Thus, our results suggest the dramatic amount of change that occurred over the course of hominin hip evolution was no faster than expected if the only evolutionary forces at work were neutral mutation and random genetic drift. While this result may seem surprising, the finding that large-scale evolutionary changes occurred at a rate slower than predicted by neutral evolutionary models is the norm (Lynch, 1990), and here the rate of hominin hip evolution was comparatively quite high. As has been noted previously (Estes and Arnold, 2007; Hansen, 2012), patterns of evolutionary change most often do not fit neutral models, with change predicted by some models to occur at a substantially slower rate than that observed on short (microevolutionary) time scales and substantially faster than observed on longer (macroevolutionary) time scales (Hansen, 2012). It appears most likely that our results also do not fit with a model that suggests evolution by neutral processes alone. Thus, rather than testing for evolution by neutral mutation, what we can say is that the rate of hominin hip evolution was far above all other

mammalian groups included here (shown in Fig. 5dsee Lynch, 1988, 1990 for others). There are at least two possible reasons for these findings. The first is that hominins evolved at an accelerated rate compared to other taxa. Counteracting this suggestion is that the rate of evolution for craniodental traits in early to late H. erectus, which Lynch (1990) found to be 2.1 105, is similar to other mammalian groups included here. This argues that perhaps only certain morphological regions in hominins evolve at a faster rate than other groups. The second possible reason is that, as frequently observed (Gingerich, 1983; Lynch, 1990; Estes and Arnold, 2007), the longer the time-span of the comparison, the slower the evolutionary rate. This effect was suggested by Lynch (1990) as one possible reason for the elevated rates in modern human cranial traits when compared to other mammals in his study, which had substantially longer divergence times. This view is in line with views on morphological evolution across a wide range of organisms that emphasize stasis as the predominant mode (e.g., Gould and Eldredge, 1977; Gingerich, 1983). Counteracting this suggestion is that the low rate for early to late H. erectus evolution given by Lynch (1990) was calculated based on 1 Myr (million years) of separation time, at or less than a number of transitions included here. In addition, recent work by Uyeda et al. (2011) provides insight into how patterns of evolution diverge from predictions of neutral models. Combining data from historical studies, the fossil record, and interspecific comparative data, the researchers found that for intervals shorter than 1 Myr, rapid evolutionary changes occurred, but the changes were constrained and did not accumulate over evolutionary time. Only after 1 Myr do lineages begin accumulating directional changes, with the mean time between changes of 27 Myr. This pattern is not predicted by neutral evolutionary models, where the amount of divergence is proportional to the amount of time separating lineages. One explanation Uyeda et al. (2011) suggested for this result is that evolutionary changes on shorter time scales may have been the result of populations tracking with local environmental variation, while on longer scales, large rare changes reflect permanent changes in adaptive zones. Natural selection to track the changing adaptive zone might have been quite strong during these periods. Rather than being the result of temporal scaling, the elevated rates seen in the hip may be signs of shifts in the adaptive landscape that drove hominin evolution. This suggestion is consistent with the dramatic shifts in the overall directions of selection seen above (Fig. 4). Though we were unable to include any early Homo (i.e., H. habilis sensu stricto, H. rudolfensis, H. erectus, H. ergaster) in this analysis because recovered fossils are too fragmentary to use with our approach, we believe current evidence supports the hypothesis that the SH Pelvis 1 reveals a morphology that is consistent with early Homo (Arsuaga et al., 1999; see also Holliday, 2012). Based on this assumption, we hypothesize that the rapid rate of change for this transition (Fig. 5) is a strong indicator that natural selection played a particularly large role at this point, as well as a major role throughout the evolution of the hominin hip. 7. Conclusion The classical view of morphological evolution in paleontology is that change equals selection, and a larger amount of change signifies a larger amount of selection. This view was critiqued by Gould and Lewontin (1979) who labeled this line of thinking as the “adaptationist programme.” While the results of this analysis by no means contradict this viewpoint, they do suggest that natural selection was the prime motivator for much of hominin hip evolution. Pattern (i.e., change) may not always reflect process (i.e., natural selection), but interspecific differences in the hominin hip appear



to generally provide a window onto past selection pressures, though a very cloudy one. Acknowledgments We thank Sarah Elton, the associate editor Becky Ackermann, Tim Weaver, and two anonymous reviewers for comments on this manuscript that greatly improved its content. We thank Neil Roach, Scott Williams, Brian Richmond, and Bernard Wood for comments on this manuscript. We are grateful to Judy Chupasko, Linda Gordon, Dave Hunt, Lyman Jellema, Doug Owsley, Kari Bruwelheide, n, Marco Eileen Westwig, Christoph Zollikofer, Marcia Ponce de Leo Milella, Emmanuel Gilissen, Wim Wendelen, Frieder Mayer, Saskia Jancke, and Georges Lenglet for the access to the hominid material used in this analysis. We particularly thank Alejandro Bonmatí and Juan-Luis Arsuaga for graciously providing the new measurements on SH Pelvis 1. We also thank Carol Ward, Scott Williams, and Jennifer Clark for access to the hominin fossil casts. We thank Gabriel Marroig and Diogo Melo for assistance with their matrix error correction method, Thomas Hansen for assistance with the standardization approach, and Kjetil Voje for help designing the allometric scaling method and conversations regarding scaling. We thank C. Keating for creating the image used in Fig. 1. This work was supported by a National Science Foundation Doctoral Dissertation Improvement Grant (BCS-1028699), National Science Foundation Grant BCS-0962903, a Sigma Xi Grants-in-Aid of Research grant, a Beckman Institute Cognitive Science/AI award, a University of Illinois Summer Research Assistance award, The George Washington University Signature Fund, and the Fulbright U.S. Scholar Program. Supplementary Online Material Supplementary online material related to this article can be found at http://dx.doi.org/10.1016/j.jhevol.2015.05.008. References Ackermann, R.R., 2009. Morphological Integration and the Interpretation of Fossil Hominin Diversity. Evol. Biol. 36, 149e156. Ackermann, R.R., Cheverud, J.M., 2000. Phenotypic covariance structure in tamarins (genus Saguinus): a comparison of variation patterns using matrix correlation and common principal component analysis. Am. J. Phys. Anthropol. 111, 489. Ackermann, R.R., Cheverud, J.M., 2004. Detecting genetic drift versus selection in human evolution. Proc. Natl. Acad. Sci. 101, 17946e17951. Arsuaga, J.L., Lorenzo, C., Carretero, J.M., Gracia, A., Martínez, I., García, N., Bermúdez de Castro, J.M., Carbonell, E., 1999. A complete human pelvis from the Middle Pleistocene of Spain. Nature 399, 255e258. llez, A., Sharp, W.D., Arsuaga, J.L., Martínez, I., Arnold, L.J., Aranburu, A., Gracia-Te res, C., Pantoja-Pe rez, A., Bischoff, J., Poza-Rey, E., Pare s, J.M., Quam, R.M., Falgue n-Torres, M., García, N., Carretero, J.M., Demuro, M., Lorenzo, C., Sala, N., Martino zar de Velasco, A., Cuenca-Besco s, G., Go mez-Olivencia, A., Moreno, D., Alca Pablos, A., Shen, C.-C., Rodríguez, L., Ortega, A.I., García, R., Bonmatí, A., Bermúdez de Castro, J.M., Carbonell, E., 2014. Neandertal roots: Cranial and chronological evidence from Sima de los Huesos. Science 344, 1358e1363. Bar-Yosef, O., Vandermeersch, B., Arensburg, B., Goldberg, P., Laville, H., Meignen, L., Rak, Y., Tchernov, E., Tillier, A.M., 1986. New Data on the Origin of Modern Man in the Levant. Curr. Anthropol. 27, 63e64. Berge, C., 1994. How did the australopithecines walk? A biomechanical study of the hip and thigh of Australopithecus afarensis. J. Hum. Evol. 26, 259e273. Berge, C., Kazmierczak, J.B., 1986. Effects of size and locomotor adaptations on the hominid pelvis: evaluation of australopithecine bipedality with a new multivariate method. Folia Primatol. (Basel) 46, 185e204. Berge, C., Orban-Segebarth, R., Schmid, P., 1984. Obstetrical interpretation of the australopithecine pelvic cavity. J. Hum. Evol. 13, 573e587. Blows, M., Walsh, B., 2009. Spherical cows grazing in flatland: constraints to selection and adaptation. In: van der Werf, J., et al. (Eds.), Adaptation and Fitness in Animal Populations. Springer, pp. 83e101. mez-Olivencia, A., Arsuaga, J.L., Carretero, J.M., Gracia, A., Bonmatí, A., Go rmudez de Castro, J.M., Carbonell, E., 2010. Middle Martínez, I., Lorenzo, C., Be Pleistocene lower back and pelvis from an aged human individual from the Sima de los Huesos site, Spain. Proc. Natl. Acad. Sci. 107, 18386e18391.

15

Bramble, D.M., Lieberman, D.E., 2004. Endurance running and the evolution of Homo. Nature 432, 345e352. Braun, D.R., Harris, J.W., Levin, N.E., McCoy, J.T., Herries, A.I., Bamford, M.K., Bishop, L.C., Richmond, B.G., Kibunjia, M., 2010. Early hominin diet included diverse terrestrial and aquatic animals 1.95 Ma in East Turkana, Kenya. Proc. Natl. Acad. Sci. 107, 10002e10007. Cheverud, J.M., 1982. Phenotypic, Genetic, and Environmental Morphological Integration in the Cranium. Evolution 36, 499e516. Cheverud, J.M., 1988. A comparison of genetic and phenotypic correlations. Evolution 42, 958e968. Cheverud, J.M., 1995. Morphological Integration in the Saddle-Back Tamarin (Saguinus fuscicollis) Cranium. Am. Nat. 145, 63e89. Cheverud, J.M., Rutledge, J.J., Atchley, W.R., 1983. Quantitative Genetics of Development: Genetic Correlations Among Age-Specific Trait Values and the Evolution of Ontogeny. Evolution 37, 895e905. Crompton, R.H., Vereecke, E.E., Thorpe, S.K.S., 2008. Locomotion and posture from the common hominoid ancestor to fully modern hominins, with special reference to the last common panin/hominin ancestor. J. Anat. 212, 501e543. Dart, R.A., 1949a. The first pelvic bones of Australopithecus prometheus; preliminary note. Am. J. Phys. Anthropol. 7, 255e257. Dart, R.A., 1949b. Innominate fragments of Australopithecus prometheus. Am. J. Phys. Anthropol. 7, 301e333. Darwin, C., 1859. On the origins of species by means of natural selection. John Murray Publishers, London, London. Demenocal, P.B., 2004. African climate change and faunal evolution during the PlioceneePleistocene. Earth and Planetary Science Letters 220, 3e24. Domínguez-Rodrigo, M., 1997. Meat-eating by early hominids at the FLK 22 Zinjanthropus site, Olduvai Gorge (Tanzania): an experimental approach using cutmark data. J. Hum. Evol. 33, 669e690. Efron, B., Tibshirani, R., 1993. An introduction to the bootstrap. Chapman & Hall, New York. Estes, S., Arnold, S.J., 2007. Resolving the paradox of stasis: models with stabilizing selection explain evolutionary divergence on all timescales. American Naturalist 169, 227e244. Garcia, G., Hingst-Zaher, E., Cerqueira, R., Marroig, G., 2014. Quantitative Genetics and Modularity in Cranial and Mandibular Morphology of Calomys expulsus. Evol. Biol. 41, 619e636. Gingerich, P.D., 1983. Rates of Evolution: Effects of Time and Temporal Scaling. Science 222, 159e161. Goswami, A., Smaers, J.B., Soligo, C., Polly, P.D., 2014. The macroevolutionary consequences of phenotypic integration: from development to deep time. Philos. Trans. R. Soc. Lond. B Biol. Sci. 369, 20130254e20130254. Gould, S.J., Eldredge, N., 1977. Punctuated equilibria: the tempo and mode of evolution reconsidered. Paleobiology 115e151. Gould, S.J., Lewontin, R.C., 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proc. R. Soc. Lond. B Biol. Sci. 205, 581e598. Gould, S.J., Vrba, E.S., 1982. Exaptation-A Missing Term in the Science of Form. Paleobiology 8, 4e15. Grabowski, M.W., 2013. Hominin obstetrics and the evolution of constraints. Evol. Biol. 40, 57e75. Grabowski, M.W., Polk, J.D., Roseman, C.C., 2011. Divergent patterns of integration and reduced constraint in the human hip and the origins of bipedalism. Evolution 65, 1336e1356. Grabowski, M., Hatala, K.G., Jungers, W.L., Richmond, B.G., 2015. Body mass estimates of hominin fossils and the evolution of human body size (in press). J. Hum. Evol. Haeusler, M., McHenry, H.M., 2004. Body proportions of Homo habilis reviewed. J. Hum. Evol. 46, 433e465. Haeusler, M., McHenry, H.M., 2007. Evolutionary reversals of limb proportions in early hominids? Evidence from KNM-ER 3735 (Homo habilis). J. Hum. Evol. 53, 383e405. Haile-Selassie, Y., Latimer, B.M., Alene, M., Deino, A.L., Gibert, L., Melillo, S.M., Saylor, B.Z., Scott, G.R., Lovejoy, C.O., 2010. An early Australopithecus afarensis postcranium from Woranso-Mille, Ethiopia. Proc. Natl. Acad. Sci. 107, 12121e12126. Hallgrimsson, B., Jamniczky, H., Young, N.M., Rolian, C., Parsons, T.E., Boughner, J.C., Marcucio, R., 2009. Deciphering the Palimpsest: Studying the Relationship Between Morphological Integration and Phenotypic Covariation. Evol. Biol. 36, 355e376. Hansen, T.F., 2012. Adaptive landscapes and macroevolutionary dynamics. In: The Adaptive Landscape in Evolutionary Biology. The adaptive landscape in evolutionary biology. Oxford University Press, New York, pp. 205e226. Hansen, T.F., Houle, D., 2008. Measuring and comparing evolvability and constraint in multivariate characters. J. Exp. Biol. 21, 1201e1219. Harcourt-Smith, W.E., 2007. The Origins of Bipedal Locomotion. In: Henke, W., Tattersall, I., Hardt, T. (Eds.), Handbook of Paleoanthropology. Springer, pp. 1483e1518. Hartwig-Scherer, S., Martin, R.D., 1991. Was “Lucy” more human than her ‘child’? Observations on early hominid postcranial skeletons. J. Hum. Evol. 21, 439e449. Hereford, J., Hansen, T.F., Houle, D., 2004. Comparing strengths of directional selection: how strong is strong? Evolution 58, 2133e2143. Hobolth, A., Dutheil, J.Y., Hawks, J., Schierup, M.H., Mailund, T., 2011. Incomplete lineage sorting patterns among human, chimpanzee, and orangutan suggest


16


recent orangutan speciation and widespread selection. Genome Research 21, 349e356. Holliday, T.W., 2012. Body Size, Body Shape, and the Circumscription of the Genus Homo. Curr. Anthropol. 53, S330eS345. labon, C., Wagner, G.P., Hansen, T.F., 2011. Measurement and Meaning Houle, D., Pe in Biology. The Quarterly Review of Biology 86, 3e34. Jungers, W.L., 1988. Relative joint size and hominoid locomotor adaptations with implications for the evolution of hominid bipedalism. J. Hum. Evol. 17, 247e265. Kibii, J.M., Churchill, S.E., Schmid, P., Carlson, K.J., Reed, N.D., de Ruiter, D.J., Berger, L.R., 2011. A partial pelvis of Australopithecus sediba. Science 333, 1407e1411. Klingenberg, C.P., 2008. Morphological integration and developmental modularity. Annual Review of Ecology, Evolution, and Systematics 39, 115e132. Klingenberg, C.P., 2009. Morphometric integration and modularity in configurations of landmarks: tools for evaluating a priori hypotheses. Evol. Dev. 11, 405e421. Kruuk, L.E.B., Slate, J., Wilson, A.J., 2008. New answers for old questions: the evolutionary quantitative genetics of wild animal populations. Annual Review of Ecology, Evolution, and Systematics 39, 525e548. Lande, R., 1979. Quantitative genetic analysis of multivariate evolution, applied to brain: body size allometry. Evolution 33, 402e416. Lande, R., 1980. The Genetic Covariance between Characters Maintained by Pleiotropic Mutations. Genetics 94, 203e215. Lande, R., Arnold, S.J., 1983. The measurement of selection on correlated characters. Evolution 37, 1210e1226. Langergraber, K.E., Prufer, K., Rowney, C., Boesch, C., Crockford, C., Fawcett, K., Inoue, E., Inoue-Muruyama, M., Mitani, J.C., Muller, M.N., Robbins, M.M., Schubert, G., Stoinski, T.S., Viola, B., Watts, D., Wittig, R.M., Wrangham, R.W., Zuberbühler, K., Paabo, S., Vigilant, L., 2012. Generation times in wild chimpanzees and gorillas suggest earlier divergence times in great ape and human evolution. Proc. Natl. Acad. Sci. 109, 15716e15721. Latimer, B., 1991. Locomotor adaptations in Australopithecus afarensis: the issue of arboreality. Origine(s) de la Bipedie Chez les Hominides, pp. 169e176. Lovejoy, C.O., 1979. A reconstruction of the pelvis of A.L.-288-1 (Hadar Formation, Ethiopia). Am. J. Phys. Anthropol. 413. Lovejoy, C.O., 1988. Evolution of human walking. Scientific American 259, 118e125. Lovejoy, C.O., 2005. The natural history of human gait and posture. Part 1. Spine and pelvis. Gait Posture 21, 95e112. Lovejoy, C.O., Cohn, M.J., White, T.D., 1999. Morphological analysis of the mammalian postcranium: a developmental perspective. Proc. Natl. Acad. Sci. 96, 13247e13252. Lovejoy, C.O., Heiple, K.G., Burstein, A.H., 1973. The gait of Australopithecus. Am. J. Phys. Anthropol. 38, 757e779. Lovejoy, C.O., Suwa, G., Simpson, S.W., Matternes, J.H., White, T.D., 2009a. The Great Divides: Ardipithecus ramidus Reveals the Postcrania of Our Last Common Ancestors with African Apes. Science 326 (73e73), 100e106. Lovejoy, C.O., Suwa, G., Spurlock, L., Asfaw, B., White, T.D., 2009b. The Pelvis and Femur of Ardipithecus ramidus: The Emergence of Upright Walking. Science 326, 71e1e71e6. Lynch, M., 1988. The rate of polygenic mutation. Genetics Research 51, 137e148. Lynch, M., 1990. The rate of morphological evolution in mammals from the standpoint of the neutral expectation. Am. Nat. 136, 727e741. Lynch, M., Hill, W.G., 1986. Phenotypic evolution by neutral mutation. Evolution 40, 915e935. Marroig, G., Cheverud, J.M., 2004. Did natural selection or genetic drift produce the cranial diversification of neotropical monkeys? Am. Nat. 163, 417e428. Marroig, G., Melo, D.A.R., Garcia, G., 2012. Modularity, noise, and natural selection. Evolution 66, 1506e1524. Marroig, G., Shirai, L., Porto, A., de Oliveira, F., De Conto, V., 2009. The Evolution of Modularity in the Mammalian Skull II: Evolutionary Consequences. Evol. Biol. 36, 136e148. Martin, R., Saller, K., 1957. Lehrbuch der Anthropologie. G. Fischer, Stuttgart. McCollum, M.A., Rosenman, B.A., Suwa, G., Meindl, R.S., Lovejoy, C.O., 2010. The vertebral formula of the last common ancestor of African apes and humans. J. Exp. Zool. B Mol. Dev. Evol. 314B, 123e134. McHenry, H.M., 1975. A new pelvic fragment from Swartkrans and the relationship between the robust and gracile australopithecines. Am. J. Phys. Anthropol. 43, 245e262. McHenry, H.M., 1992. Body size and proportions in early hominids. Am. J. Phys. Anthropol. 87, 407e431. McHenry, H.M., Brown, C.C., 2008. Side steps: the erratic pattern of hominin postcranial change through time. J. Hum. Evol. 55, 639e651. Morgan, M.E., Lewton, K.L., Kelley, J., Ot arola-Castillo, E., Barry, J.C., Flynn, L.J., Pilbeam, D., 2015. A partial hominoid innominate from the Miocene of Pakistan: Description and preliminary analyses. Proc. Natl. Acad. Sci. 112, 82e87. Olson, E.C., Miller, R.L., 1958. Morphological integration. University of Chicago Press, Chicago. Pearson, O.M., 2000. Activity, climate, and postcranial robusticity: implications for modern human origins and scenarios of adaptive change. Curr. Anthropol. 41, 569e607. Pilbeam, D., 1996. Genetic and morphological records of the Hominoidea and hominid origins: a synthesis. Mol. Phylogenet. Evol. 5, 155e168. Plavcan, J.M., Hammond, A.S., Ward, C.V., 2014. Brief Communication: Calculating hominin and nonhuman anthropoid femoral head diameter from acetabular size. Am. J. Phys. Anthropol. 155, 469e475. Pontzer, H., 2012. Ecological Energetics in Early Homo. Curr. Anthropol. 53, S346eS358.

Pontzer, H., Raichlen, D.A., Sockol, M.D., 2009. The metabolic cost of walking in humans, chimpanzees, and early hominins. J. Hum. Evol. 56, 43e54. Porto, A., de Oliveira, F., Shirai, L., De Conto, V., Marroig, G., 2009. The Evolution of Modularity in the Mammalian Skull I: Morphological Integration Patterns and Magnitudes. Evol. Biol. 36, 118e135. R Development Core Team, 2011. R: A Language and Environment for Statistical Computing. Rak, Y., 1991. Lucy's pelvic anatomy: its role in bipedal gait. J. Hum. Evol. 20, 283e290. Rak, Y., Arensburg, B., 1987. Kebara 2 Neanderthal pelvis: first look at a complete inlet. Am. J. Phys. Anthropol. 73, 227e231. Richmond, B.G., Aiello, L.C., Wood, B.A., 2002. Early hominin limb proportions. J. Hum. Evol. 43, 529e548. Richmond, B.G., Jungers, W.L., 2008. Orrorin tugenensis femoral morphology and the evolution of hominin bipedalism. Science 319, 1662e1665. Robinson, J.T., 1972. Early Hominid Posture and Locomotion. The University of Chicago Press, Chicago. Roff, D.A., 1995. The estimation of genetic correlations from phenotypic correlations: a test of Cheverud's conjecture. Heredity (Edinb) 74, 481e490. Roff, D.A., 1996. The evolution of genetic correlations: An analysis of patterns. Evolution 50, 1392e1403. Rolian, C., Lieberman, D.E., Hallgrimsson, B., 2010. The coevolution of human hands and feet. Evolution 64, 1558e1568. Roseman, C.C., 2012. The soundness of the Cheverud Conjecture and its implications for the study of human evolution. Am. J. Phys. Anthropol. 147, 252. Rosenberg, K., Trevathan, W., 2002. Birth, obstetrics and human evolution. BJOG 109, 1199e1206. Rosenberg, K.R., Brace, C.L., Frayer, D.W., Geise, M.C., Greene, D.L., Tague, R.G., Tappen, N.C., Trevathan, W.R., Trinkaus, E., 1988. The Functional Significance of Neandertal Public Length [and Comments and Reply]. Curr. Anthropol. 29, 595e617. Ruff, C., 1998. Evolution of the Hominid Hip. In: Strasser, et al. (Eds.), Primate Locomotion. Springer US, Boston, MA, pp. 449e469. Ruff, C., 2010. Body size and body shape in early hominins e implications of the Gona pelvis. J. Hum. Evol. 58, 166e178. Ruff, C.B., 1994. Morphological adaptation to climate in modern and fossil hominids. Am. J. Phys. Anthropol. 37, 65e107. Ruff, C.B., 1995. Biomechanics of the hip and birth in early Homo. Am. J. Phys. Anthropol. 98, 527e574. Schroeder, L., Roseman, C.C., Cheverud, J.M., Ackermann, R.R., 2014. Characterizing the Evolutionary Path(s) to Early Homo. PloS One 9, e114307. Sigmon, B.A., 1971. Bipedal behavior and the emergence of erect posture in man. Am. J. Phys. Anthropol. 34, 55e60. Sigmon, B.A., 1974. A functional analysis of pongid hip and thigh musculature. J. Hum. Evol. 3, 161e185. Sigmon, B.A., 1975. Functions and evolution of hominid hip and thigh musculature. In: Tuttle, R. (Ed.), Primate Functional Morphology and Evolution. The Hague & Paris, pp. 235e262. Simpson, G.G., 1953. The Baldwin Effect. Evolution 7, 110e117. Simpson, S.W., Quade, J., Levin, N.E., Butler, R., Dupont-Nivet, G., Everett, M., Semaw, S., 2008. A Female Homo erectus Pelvis from Gona, Ethiopia. Science 322, 1089e1092. Simpson, S.W., Quade, J., Levin, N.E., Semaw, S., 2014. The female Homo pelvis from Gona: response to Ruff (2010). J. Hum. Evol. 68, 32e35. Smith, R.J., Jungers, W.L., 1997. Body mass in comparative primatology. J. Hum. Evol. 32, 523e559. Sockol, M.D., Raichlen, D.A., Pontzer, H., 2007. Chimpanzee locomotor energetics and the origin of human bipedalism. Proc. Natl. Acad. Sci. 104, 12265e12269. Sokal, R.R., Rohlf, F.J., 1995. Biometry: the principles and practice of statistics in biological research. W.H. Freeman, New York. Steppan, S.J., Phillips, P.C., Houle, D., 2002. Comparative quantitative genetics: evolution of the G matrix. TREE 17, 320e327. Stern, J.T., 2000. Climbing to the top: A personal memoir of Australopithecus afarensis. Evol. Antropol. 9, 113e133. Stern, J.T., Susman, R.L., 1983. The locomotor anatomy of Australopithecus afarensis. Am. J. Phys. Anthropol. 60, 279e317. Tague, R.G., 1991. Commonalities in dimorphism and variability in the anthropoid pelvis, with implications for the fossil record. J. Hum. Evol. 21, 153e176. Tague, R.G., Lovejoy, C.O., 1986. The obstetric pelvis of AL 288-1 (Lucy). J. Hum. Evol. 15, 237e255. Uyeda, J.C., Hansen, T.F., Arnold, S.J., Pienaar, J., 2011. The million-year wait for macroevolutionary bursts. Proc. Nat. Acad. Sci. 108, 15908e15913. Walker, A., Ruff, C.B., 1993. The reconstruction of the pelvis. In: Walker, A., Leakey, R.E. (Eds.), The Nariokotome Homo erectus Skeleton. Harvard University Press, Cambridge, pp. 221e233. Ward, C.V., 1991. Functional Anatomy of the Lower Back and Pelvis of the Miocene Hominoid Proconsul nyanzae from Mfangano Island, Kenya. PhD Dissertation. Johns Hopkins University. Ward, C.V., 2002. Interpreting the posture and locomotion of Australopithecus afarensis: Where do we stand? Am. J. Phys. Anthropol. 45, 185e215. Warrener, A.G., 2011. Pelvic Shape, Hip Abductor Mechanics and Locomotor Energetics in Extinct Hominins and Modern Humans. PhD Dissertation. Washington University in St. Louis, St. Louis. Warrener, A.G., Lewton, K.L., Pontzer, H., Lieberman, D.E., 2015. A Wider Pelvis Does Not Increase Locomotor Cost in Humans, with Implications for the Evolution of Childbirth. PloS One 10 e0118903e14.


M. Grabowski, C.C. Roseman / Journal of Human Evolution xxx (2015) 1e17 Weaver, T.D., 2002. A multi-causal functional analysis of hominid hip morphology. Stanford University. Weaver, T.D., 2003. The shape of the Neandertal femur is primarily the consequence of a hyperpolar body form. Proc. Natl. Acad. Sci. 100, 6926e6929. Weaver, T.D., 2009. The meaning of Neandertal skeletal morphology. Proc. Natl. Acad. Sci. 106, 16028e16033. Weaver, T.D., Hublin, J.-J., 2009. Neandertal birth canal shape and the evolution of human childbirth. Proc. Natl. Acad. Sci. 106, 8151e8156.

17

Weaver, T.D., Roseman, C.C., Stringer, C.B., 2007. Were neandertal and modern human cranial differences produced by natural selection or genetic drift? J. Hum. Evol. 53, 135e145. Williams, G.C., 1996. Adaptation and natural selection: a critique of some current evolutionary thought, reprint, illustrated. ed. Princeton University Press, Princeton, N.J. Willis, J.H., Coyne, J.A., Kirkpatrick, M., 1991. Can one predict the evolution of quantitative characters without genetics? Evolution 45, 441e444.


Constraint, natural selection, and the evolution of human body form.

Complexity, Natural Selection and the Evolution of Life and Humans.

Experimental evolution can unravel the complex causes of natural selection in clinical infections.

Evolution and the Growth Process: Natural Selection of Entrepreneurial Traits.

Natural selection and the evolution of reproductive effort.

The role of natural selection in human evolution - insights from Latin America.

Classic reaction kinetics can explain complex patterns of antibiotic action.

The major histocompatibility complex and human evolution.

Natural evolution and human consciousness.

The genotype-phenotype map of yeast complex traits: basic parameters and the role of natural selection.

Evolution of behavior by density-dependent natural selection.

Cultural selection drives the evolution of human communication systems.

Evolution of class-I MHC genes and proteins: from natural selection to thymic selection.

Intrahost dynamics of antiviral resistance in influenza A virus reflect complex patterns of segment linkage, reassortment, and natural selection.

Characterizing the roles of changing population size and selection on the evolution of flux control in metabolic pathways.

Natural selection and the maximization of fitness.

The changing patterns of systemic fluoride intake.

Holding it together: rapid evolution and positive selection in the synaptonemal complex of Drosophila.

Patterns of natural selection acting on the mitochondrial genome of a locally adapted fish species.

Models inconsistent with altruism cannot explain the evolution of human cooperation.

Changing patterns of work.

Altered Ca2+ kinetics associated with α-actinin-3 deficiency may explain positive selection for ACTN3 null allele in human evolution.

Does natural selection explain the fine scale genetic structure at the nuclear exon Glu-5' in blue mussels from Kerguelen?

Changing patterns of work.