The evolution of human and ape hand proportions

Human hands are distinguished from apes by possessing longer thumbs relative to fingers. However, this simple ape-human dichotomy fails to provide an adequate framework for testing competing hypotheses of human evolution and for reconstructing the morphology of the last common ancestor (LCA) of humans and chimpanzees. We inspect human and ape hand-length proportions using phylogenetically informed morphometric analyses and test alternative models of evolution along the anthropoid tree of life, including fossils like the plesiomorphic ape Proconsul heseloni and the hominins Ardipithecus ramidus and Australopithecus sediba. Our results reveal high levels of hand disparity among modern hominoids, which are explained by different evolutionary processes: autapomorphic evolution in hylobatids (extreme digital and thumb elongation), convergent adaptation between chimpanzees and orangutans (digital elongation) and comparatively little change in gorillas and hominins. The human (and australopith) high thumb-to-digits ratio required little change since the LCA, and was acquired convergently with other highly dexterous anthropoids.


Supplementary Figure 1 | Extrinsic hand proportions of humans, apes and other anthropoid primates.
Species means are displayed for all extant hominid species, selected hylobatids, and one species representative of each non-hominoid anthropoid genus. Each element length (in mm) has been adjusted by the known or estimated cube root of body mass (in kg) of the individual. This plot demonstrates the huge disparity in hand proportions among extant hominoids. In comparison to humans, chimpanzees have longer digits and slightly shorter thumbs; contrarily, gorillas exhibit similar digital length but shorter thumbs; orangutans display longer fingers (longer than chimpanzees) and slightly shorter thumbs; finally, hylobatids exhibit much longer digits and thumbs. Other relevant observations in non-hominoids: Theropithecus approaches the human intrinsic hand proportions (Fig. 1b) by a different mechanism (longer pollical and digital metacarpals but much shorter phalanges). Nasalis exhibits chimpanzee-like digital length with a shorter thumb, but African colobines (not included) would exhibit even shorter ones (they display vestigial thumb elements). Fossil species are indicated ( †).

Supplementary Figure 4 | Allometric relationships of thumb (a) and fourth ray (b) lengths relative to body mass (BM) in humans and other anthropoid primates.
In both cases, regression lines are fitted to hylobatids (purple), orangutans (light green), gorillas (red), chimpanzees (orange) and modern humans (light blue) and extended over the remaining comparative sample. Analyses of covariance (ANCOVA) show that there are not significant differences in the slopes of the thumb (F=1.954; P=0.107) and fourth ray (F=1.131; P=0.343) regressions. However, in both cases, modern humans show regression slopes that are not statistically different from zero (Supplementary Table 5), indicating that in humans thumb and digital lengths are not dependent on body size. Comparisons between pollical marginal means (evaluated at lnBM=3.796) reveal that modern humans and hylobatids (not statistically different) exhibit longer thumbs than great apes (P<0.001). Gorillas display even shorter thumbs (P=0.001) than chimpanzees and orangutans (the latter two showing no differences). Comparisons between fourth ray marginal means (evaluated at lnBM=3.828) reveal differences between each ape group and humans (P≤0.008), with the exception of chimpanzees and hylobatids (P=0.332). When accounting for allometric relationships, orangutans exhibit longer fourth ray than hylobatids/chimpanzees, which in turn, are longer than gorillas and humans respectively.

Supplementary Figure 5 | Alternative multivariate multi-regime Ornstein-Uhlenbeck (OU) hypotheses tested for the evolution of extrinsic hand proportions (EHP).
Starting with Brownian motion as a baseline, we compared the relative fit (using AIC c ) of increasingly complex OU models with "ouch": one single regime (OU1), two regimes (OU2, hominoids vs. non-hominoids), four regimes (OU4; platyrrines, cercopithecids, non-human hominoids plus Ardipithecus, Australopithecus-Homo), the five regimes revealed by "surface" (OU5 'surface'; platyrrhines, Papio-Theropithecus, hylobatids, Pan-Pongo, rest of catarrhines). We further designed an alternative version of the previous model (OU5 'alt') in which Pan and Pongo were considered to reflect the plesiomorphic great ape condition. The OU5 'surface' model represented the best fit model, irrespective of the body mass estimate used for Ardipithecus, and the inclusion or not of Ardipithecus and Proconsul (see results in Supplementary Table 8). Figure 6 | Sensitivity test of species sample size in alternative multivariate multi-regime Ornstein-Uhlenbeck (OU) models of extrinsic hand proportions evolution. We compared the fit of different OU models after dropping all hylobatid species and Proconsul. Alternative hypotheses included the OU1, OU2, and OU4 models described in Supplementary Figure 5, plus the three regime output revealed by "surface" (OU3 'surface'), and an alternative version of OU3 'surface' (OU4 'alt') based on the best fit model obtained for the full sample ( Supplementary  Fig. 5). In this case, the best fit model is represented by the OU3 'surface' output, in which upon an anthropoid regime baseline, hominins and gorillas share an optimum (convergent with baboons), whereas Pan and Pongo are again convergent (see results in Supplementary Table 8).

Supplementary Figure 7 | Reconstructed evolutionary history of human and ape
hand proportions by excluding contentious fossils. Same approach as in Figure 4, but excluding Proconsul heseloni and Ardipithecus ramidus. Taxa are color-coded as in the phylogenetic tree (Fig. 3); internal nodes (i.e., ancestral-state reconstruction) are also indicated, highlighting the positions in shape-space of the great ape-human and chimpanzee-human LCA (plus 95% confident intervals for the latter estimate). The overall evolutionary pattern is comparable to that found in previous iterations including more fossils. Again, species of macaques were not labelled due to space restrictions.

Supplementary Figure 8 | Evolution of intrinsic hand proportions (IHP) in
humans and other anthropoid primates. The observed (Fig. 1) and reconstructedstate values are mapped along the branches and nodes of the anthropoid phylogeny. The ancestral state values for the great ape-human and chimpanzee-human last common ancestors (LCA) are highlighted with arrows. The IHP (relative long thumb) of humans, geladas and capuchin monkeys, as well as the IHP (different degrees of relative short thumb) of modern apes and Nasalis are reconstructed as having evolved (independently) from moderate proportions similar to Proconsul. Inset drawing represents a modern human performing a "pad-to-pad" precision grasping 1 . The length of the color legend at the bottom provides scale for the branches of the tree.

Supplementary Figure 8 | Continued.
This method to visualize trait evolution in a tree is explained in detail elsewhere 2 . Basically, ancestral characters are first estimated at the internal nodes again using ML 3,4 and Brownian motion [5][6][7] . Next, all edges along the tree are fractionated, and state estimates are computed at the midpoint of each fraction via interpolation using equation [3] of Felsenstein 7 . This creates the visual appearance of continuous color change along the edges of the tree.

Supplementary Figure 9 | Alternative multi-regime Ornstein-Uhlenbeck (OU) hypotheses tested for the evolution of intrinsic hand proportions (IHP).
In total, we compared the relative fit of eight different OU multi-regime models: Starting with Brownian motion, we followed with the increasingly complex models OU1, OU2, OU4, as well as the best fit model for extrinsic hand proportions described in Supplementary Figure 5. Furthermore we incorporated three extra models (this figure) based on the IHP results revealed by Figure 1 and Supplementary Figure 8 (as expected, "surface" did not perform well with univariate data 8 ). The four adaptive regimes OU4 'IHP' represented the best fit model: Australopithecus-Homo share an adaptive regime with Theropithecus and Cebus (i.e., they are convergent for a relative long thumb); Pan and Pongo are convergent for a relative short thumb (as in the case of extrinsic hand proportions; Supplementary Fig. 5), as well as Nasalis; hylobatids, gorillas and Ardipithecus share the inferred plesiomorphic condition for crown apes, whereas the rest of the cercopithecid and platyrrhine monkeys share a more generalized regime. Alternative models in which Pan, Pongo and Nasalis share the same regime as other hominoids (OU3 'IHP alt1') or where Theropithecus and Cebus are not convergent with Australopithecus-Homo (OU4 'IHP alt2') exhibited an inferior fit (see results in Supplementary

Supplementary Table 3 | Results of the principal component analyses (PCA).
Results are shown for the individuals-based extrinsic hand proportions in our full ( Fig. 2A, Supplementary Fig. 2) and great apes only samples ( Supplementary Fig. 3). *The "small human" regression is based in ten sex-specific population means, including: Eastern and Western African pygmies, Khoe-San, Aeta and Andaman Islanders.

Supplementary Table 7 | Results of the principal component analyses (PCA)
using the covariance matrix between species means. Results are provided for the full set of extrinsic hand proportions (Fig. 4a, b), and for the fourth ray only (Fig.  6a, Figures 5, 6, 9 for descriptions of each model.

Supplementary Note 1: What is "pad-to-pad" precision grasping?
Among the vast array of grips that the human and ape hand are capable of, Napier 10 defined the term "precision grip" to describe instances in which the object being manipulated (with precision) was held between the palmar aspects of the fingers and the opposing thumb. In contrast, "power grip" refers to situations in which the object is held in a "clamp fashion" between the flexed fingers and the palm, and the thumb only plays a subsidiary role by directing the force being applied (as when using a hammer). The term "human-like precision grasping" is commonly used in the literature although sometimes misunderstood: chimpanzees and orangutans can efficiently manipulate objects via different forms of precision grasping (e.g., thumb and index finger tip-to-tip and pad-to-side) 1,11,12 . However, the characteristic human "pad-to-pad" precision grip (i.e., flat contact between the proximal pulps of the thumb and one or more fingers) 13 is precluded in modern apes due to the disproportionate length of their digits II-V relative to the thumb 1,11,14,15 (Fig. 1), as well as by restricted passive hyperextension of the distal phalanges 12,15 .
With the exception of hylobatids, a group that constitutes the exception to many rules in hand morphology 16,17 , a clear trend is revealed within each anthropoid lineage: the more arboreal species exhibit functionally shorter thumbs relative to the fingers. As an example, within extant great apes the highly arboreal orangutans, followed by chimpanzees, display relatively shorter thumbs than the more terrestrial gorillas, which exhibit more generalized proportions (Fig. 1). This has been related to the capability of performing an effective "hook grasp" during below-branch suspension 1,11,16 . Among catarrhines, only Theropithecus gelada approaches the human condition in terms of IHP as computed in our analysis (Fig. 1), but since geladas exhibit an extremely shortened index finger 18 , this "opposability index" would surpass the human condition if the index finger was the denominator instead of the fourth ray. The IHP in this species are explained as a specific feeding adaptation in primates that spend 70% of their daily activity collecting food (blades of grass, seeds and rhizomes) using precision grips 19 . The special adaptation of the hands of geladas is also evident in a special differentiation of the flexor digitorium profundus, as well as other thumb muscles 20 , which is also reflected in their pollical distal phalanx morphology 17 . Capuchins (Cebus) monkeys are the only non-hominoid primates known to use tools habitually 21 . Although platyrrhines lack a "true opposable thumb" 11 , capuchins (unlike other New World primates) commonly display both precision and power grips to manipulate objects such as use of stones as nut cracking tools, and stone flakes as cutting tools [22][23][24] . Thus, these behaviours are consistent with our results of intrinsic hand proportions (Fig.1), which we find to be convergent with humans ( Supplementary Fig. 9).

Supplementary Note 2: Evolutionary scenarios supported by the results of this work
Our results show that contrarily to the idea assumed by some, extant great apes constitute a heterogeneous group in terms of hand and thumb proportions (Figs. 1-2, Supplementary Figs. 1-3). Furthermore, our evolutionary modelling unambiguously shows that the chimpanzee-human LCA exhibited a moderate hand length (relative to overall body size), more similar to humans than to chimpanzees (Figs. 3-4). Of special relevance is the fact that even using different phylogenetic hypotheses (Figs. 5-6, Supplementary Figs. 6-7) our results indicate that digital elongation has been achieved to different degrees and independently in the different extant and fossil ape lineages. Although the evidence presented here is restricted to the hand, broader implications can be reasonably drawn in terms of human and ape evolution:

-Mosaicism and Parallelism in Ape Evolution
Together with previous analyses of limb proportions 25 and skull morphology 26 , these results falsify the view that extant apes, and particularly African apes constitute a homogeneous group with subtle deviations from a common allometric pattern 27 . Furthermore, the degree of heterogeneity in hand proportions revealed here is congruent with a mosaic evolution of the hominoid postcranial skeleton as inferred before from the fossil record [28][29][30][31][32][33][34] . Our results, indicating parallel evolution for digital elongation (with Pan and Pongo sharing convergent similarities, Fig. 3), match previous observations in other anatomical regions of modern apes [35][36][37][38][39][40] . In general, the current evidence reinforces the view that specialized arboreal adaptations exhibited by the living apes are not identical because they evolved independently as biomechanical solutions to largely similar but far from identical positional and locomotor behaviours 16,29,35 , and parallelism was facilitated by their common genetic and developmental base 37,41 . One of the consequences of this hypothesis is that no extant ape will properly represent a living analogue for a given hypothetical ancestor 29,42 .

-Extant Hominoids Are Survivors
As pointed out before, extant apes represent a very decimated expression of a highly diversified group during the Miocene 40,[42][43][44] . What explains their decay? And why there are no fossil apes showing all the derived features of the living lineages? A possible explanation is that offered by Pilbeam and colleagues [45][46][47] , who argue that we have not yet found any bona fide crown great ape in the fossil record. Another hypothesis that we favour is that a select few hominoid lineages (living representatives) survived because they were adapted to specialized lifestyles: e.g., enhanced antipronogrady and frugivory in hylobatids, orangutans and chimpanzees; large body size and folivory in gorillas; and finally, bipedalism and novel manual foraging strategies in hominins 14,48,49 , and were able to compete with the radiation of the more generalized cercopithecids starting in the late Miocene 41,50 . If that were the case, it is striking that the European late Miocene Hispanopithecus/Rudapithecus lineage (Fig. 5), with clearly specialized suspensory adaptations in the hand and other anatomical regions 28,30,34,51 , became extinct at the end of the Miocene. It seems that in this case the specialized lifestyle that allowed the survival of most extant ape lineages became an evolutionary trap for Hispanopithecus during to the "Vallesian crisis" (ca. 9.5 Ma), which caused its extinction -as well as that of other forest-adapted faunaas a consequence of paleoenvironmental changes associated with increased aridification and seasonality that caused the demise of the warm temperate forests (and year-round availability of fruit) in Western and Central Europe 52,53 . Furthermore, the results of this work indicate that suspensory behaviours in Hispanopithecus laietanus (as indicated by finger lengthening) evolved independently from other ape lineages (Fig. 6), reinforcing the view that the West European Miocene apes constituted an independent evolutionary radiation.

-Implications for Knuckle Walking
Humeral length relative to body mass is surprisingly similar in African apes and modern humans 54 , but it is relatively longer in orangutans and lesser apes. These latter two suspensory hominoids also possess higher brachial indices (i.e., 100 x radius length/humerus length), whereas modern humans and gorillas have the lowest brachial indices among extant hominoids. Proconsul, Ardipithecus and australopiths (Au. afarensis, Au. garhi, and Au. sediba) all have intermediate brachial indices that overlap with chimpanzees 55 , suggesting this to be the plesiomorphic proportionality for the upper limb of the African ape-human last common ancestor (LCA). However, our results on extrinsic hand proportions (EHP) favour the hypothesis that gorillas and early hominins are the most conservative in terms of overall hand shape (Figs. 3-4), in agreement with previous observations of Schultz 56 . This has implications for understanding the evolution of knuckle walking. Classically, the hands of great apes were seen as anatomical "hooks" designed for below-branch suspension, so they would be forced when on the ground to walk on the dorsal surfaces of their hooked hands 36,56 . However, this locomotor behaviour is currently seen as a compromise solution between the biomechanical requirements of advanced climbing and terrestrial digitigrady -requiring long versus short fingers, respectively 57,58 . Based on the terrestrial fist-walking of orangutans, Tuttle interpreted it as an intermediate stage between advanced arboreal suspension and terrestrial locomotion 15,58,59 . Whereas fistwalking allowed the hand to be used as a supporting structure, knuckle walking would further allow the manual phalanges to act as a propulsive lever during terrestrial quadrupedalism 58 . In our analyses (Fig. 4) the EHP of the African ape LCA are reconstructed as moderate in digital length (i.e., most similar to the chimpanzeehuman LCA). Thus, irrespective of whether knuckle walking evolved only once at the base of the African ape lineage 45,57,60 , or independently in gorillas and chimpanzees 55,58,[61][62][63] , our results imply that it was not related causally to the possession of especially long digits like those present in Pan or Pongo. Contrarily, origins of knuckle walking should probably be interpreted only in the light of an adaptive complex that would reduce the compressive stresses, as well as the torques generated by the ground reaction force during hyperextension of the metacarpophalangeal joints during terrestrial quadrupedalism while still preserving a powerful grasping hand 64,65 . Among other bony features, it would be associated with: short phalanges relative to metacarpals, high dorsopalmar diameter of the metacarpal heads, as well as pronounced dorsal ridges and large epicondyles on the metacarpal heads. The question remains that if knuckle walking is such an efficient form of terrestrial quadrupedalism, why has it not evolved in other primates too? (as it has outside primates) 66 . The answer to this question is no doubt very complex; "regular" (monkey-like) digitigrady might be restricted by a certain threshold of absolute digital length and body mass within an ancestral terrestrial setting. Evolution of knucklewalking was probably facilitated in African apes instead by their arboreal heritage 15 ; having short tendons for the extrinsic flexor muscles is one of several limiting factors 12,15,67 , and we hypothesize that the possession of an orthograde body plan, as well as long forelimbs relative to hindlimbs, combine to dictate the unusual way in which African apes can perform quadrupedalism.

-Implications for Early Hominin Locomotion
In relation to the longstanding debate on the climbing capabilities of early hominins 68-71 , our results -showing similar digital length in gorillas and modern humans (Figs. 2, 4; Supplementary Fig. 1)-imply that in terms of digital length there is no reason to think that climbing behaviours observed in gorillas 72 were precluded in australopithecines. In fact, trained modern humans are excellent climbers 73 , even exceeding gorillas in acrobatic capabilities 74 . Relevant to the origins of bipedalism, the preserved portions of the thorax and hand of the fossil great ape Pierolapithecus indicate that the acquisition of an orthograde body plan can be decoupled from specialized climbing and suspensory adaptations 29 . This evidence opens the possibility of human bipedalism having originated as a direct exaptation of arboreal orthogrady, without an intermediate stage of advanced suspension or specialized knuckle walking.

-Origins of the Human Hand
In terms of modern human hand proportions, most of the evolutionary change is concentrated in digital elongation/reduction (specifically metacarpal and proximal phalanx) whereas the thumb itself has remained more conservative, with just slight thumb elongation in humans (especially via proximal pollical phalanx; Supplementary  Table 7). Therefore, within living apes (and anthropoids) modern humans do not exhibit the shortest hands nor the longest thumbs, but rather a useful combination that has been selected to allow enhanced thumb to fingers opposition (Fig. 1), as it is revealed by our convergence results with Theropithecus and Cebus ( Supplementary  Fig. 9). Furthermore, these optimal intrinsic proportions evolved from a moderate ratio, as inferred for the chimpanzee-human LCA estimation, with less shape change than by assuming a chimpanzee-like LCA (Fig. 4, Supplementary Fig. 8). This confirms previous hypotheses based on observations of extant taxa, fossil apes and early hominins 17,31,63,75 , and favors classic views of human evolution that preceded the molecular resolution of hominid phylogeny 15,56,67,[76][77][78] . This and previous works indicate that enhanced thumb-to-digits opposition was present in australopiths sensu lato 14,49,[79][80][81][82][83][84] , but see Rolian and Gordon for a different opinion on Au. afarensis 85,86 . This would not be the case of the early Pliocene (4.4 Ma) Ar. ramidus, that exhibits a shorter thumb relative to fingers (i.e., IHP; in the gorilla and hylobatid range, but longer than chimpanzees; Fig. 1 A long thumb relative to fingers (i.e., high IHP; Fig. 1) facilitates enhanced pad-topad opposition and advanced manipulative skills in humans 1 and other non-hominoid primates 18,19,[22][23][24] . But, did this high human IHP ratio evolve specifically for stone tool making? There is archaeological evidence indicating that stone tool use was part of the chimpanzee-human LCA behavioural repertoire 21,96 , and thus not surprisingly also of Au. afarensis 97 . Furthermore, the thumb of O. tugenensis suggests human-like IHP at 6 Ma disassociated of stone tools 48 . We hypothesize that both human-like IHP and stone tool using behaviours evolved prior to the widespread appearance of systematic stone tool making, around 2.5 Ma 98,99 , probably when the derived manual traits distinctive of modern humans and Neandertals first evolved 100,101 . More recently, the newly-described lithic artifacts from Lomekwi 3 (West Turkana, Kenya) push back the earliest evidence of intentional stone tool production at 3.3 Ma 102 , which is consistent with human-like manual dexterity being an ancient adaptation amongst hominins. Harmand et al. argue that the decisive adaptation enabling "Lomekwian" stone knapping most likely related to a reorganization of the central nervous system in yet unidentified hominins 102 .
Thus, among the many features characterizing the human hand 1,10,11 such as a high IHP would not have necessarily evolved originally as a specific adaptation to stone tool making. Instead, they probably evolved as a new foraging/feeding adaptive complex, in the context of habitual bipedalism 14,48,74,103 . The relevance of bipedalism for the emergence of advanced manipulative skills in humans has been recognized ever since Darwin 104 , but also in more recent works 14,74,105,106 . These authors share the same basic idea: regular bipedalism allowed some degree of relaxation of the locomotor selective pressures acting in the upper extremity, facilitating the manipulative selective pressures already present in all primates 1 to refine hand length proportions for advanced manipulative tasks. However, although foot-hand coevolution could have occurred via shared developmental pathways (i.e., pleiotropic effects) 107 , our results indicate that these changes were relatively subtle (human manual hand proportions evolved from moderate -plesiomorphic-proportions, not from a chimp-like ancestor; Fig. 4 and Supplementary Fig. 8). Finally, we agree with idea that human hand length proportions are largely plesiomorphic for the hominin clade, and it was not until later in time, when these proportions were co-opted 108 for purposive and systematic stone tool making in hominins with more advanced cognitive capabilities 14,48,49,102,109 .