The anatomy of the living hominoids, the extant primates most closely related to modern humans (Table 1), has long attracted the attention of researchers (e.g. Tulp, 1641; Tyson, 1699; Camper, 1782, 1799). The close similarities between modern human anatomy and the anatomy of chimpanzees (Pan), gorillas (Gorilla), orangutans (Pongo) and gibbons (Hylobates), and the particularly detailed similarities between modern humans and the African apes, have been noted by researchers for more than 150 years (e.g. Huxley, 1864). However, these observations made little impact on the taxonomy of primates, which continued to reflect the prevailing wisdom that modern humans differed so fundamentally from their closest non-human relatives that they deserved recognition at a high level in the Linnaean hierarchy (e.g. Order Bimanus [Blumenbach, 1795] Family Hominidae [Gray, 1825]).

Table 1 An example of a taxonomy of the living higher primates that recognizes the close genetic links between Pan and Homo. Note that the meanings of ‘hominid’, ‘hominin’ and ‘hominine’ differ from those used in (more ...)

Technical advances in the last 100 years have made available new types of evidence for consideration by primate taxonomists. First, came molecular evidence about the differences among higher primates (e.g. Nuttall, 1904; Zuckerkandl et al. 1960; Goodman, 1963; Zuckerkandl, 1963; Sarich, 1967, 1968). In the past few decades this has been supplemented by comparative evidence about sequence differences at the level of the genome (e.g. Goodman et al. 1994; Ruvolo, 1997). Both these classes of evidence have reinforced the integrity of a group that includes the African apes and modern humans. However, it is only relatively recently that a cadre of researchers has been willing to promote, and adopt, a taxonomy that recognizes a particularly close relationship between Homo and Pan, and between these taxa and Gorilla (e.g. Goodman, 1963; Goodman et al. 1994; Shoshani et al. 1996) (Table 1).

Until the advent of molecular and DNA sequence data, nearly all the evidence taken into account by those studying hominoid systematics came from the hard tissues, and especially the hard tissues of the skull. Evidence from soft tissues has been incorporated into some systematic reviews (e.g. Groves, 1986; Shoshani et al. 1996), but in all cases soft-tissue data were substantially outnumbered by skeletal and dental characters. This near total reliance on skeletal and dental evidence is unfortunate for at least three reasons. First, it equates ‘morphology’ with ‘hard tissue’ or ‘skeletal and dental’ morphology. Second, recent studies have cast doubt on the effectiveness of traditional craniodental hard-tissue evidence for reconstructing hominoid phylogeny (Hartman, 1988; Harrison, 1993; Pilbeam, 1996; Collard & Wood, 2000). Third, opportunities to collect information about hominoid soft-tissue anatomy by dissecting animals sampled from populations in their original locations and habitat are dwindling. Deforestation is leading to the attrition of hominoid habitats at an unprecedented rate. When the ravages of deforestation are combined with the associated threat posed by the bushmeat trade (e.g. Bowen-Jones, 1998), the elimination of chimpanzees from their natural habitats is a real possibility within the next decade (Baillie & Groombridge, 1996). The seemingly inexorable progress of deforestation at other locations in Africa and Asia also threatens the long-term survival of gorillas, orangutans and gibbons in the wild (Baillie & Groombridge, 1996).

The living non-human hominoids have traditionally attracted the attention of hunters, collectors, naturalists and scientists. These individuals have, for a wide range of motives, assembled collections, both large and small, of extant hominoids. The most comprehensive collections include skins and complete skeletons, but many others comprise only skeletal evidence, and of these the majority are dominated by craniodental specimens. Providing resources are made available to curate and conserve these collections appropriately, they will continue to allow researchers to collect information about gross morphology, both external and internal, as well as providing opportunities to collect data about skeletal and dental microstructure. In addition, the skins, depending on the preservation medium (Hall et al. 1995), may also retain sufficient DNA to allow segments of the genome to be characterized. Some of the comparative collections include detailed information about the location, condition, size and weight of the carcass immediately after the animal was trapped and killed. In many cases these data are sufficiently precise to enable skeletal and dental variation to be studied at the level of the species and subspecies, and in some cases also at the level of the deme. However, because of the severely diminished size of hominoid populations in the wild, opportunities to collect comparable data for soft-tissue anatomy are effectively at an end.

Given this context, our study comprised three activities. First, we collated and reviewed evidence in the literature about the soft-tissue anatomy of the living hominoids. Second, we summarized these data to draw attention to the anatomical regions and the systems that are under represented, or not represented at all, in this data set. Third, where appropriate we converted the data that do exist into character states. These were then used in a phylogenetic analysis to test whether hominoid soft tissues are capable of recovering the hypothesis of hominoid relationships that is supported by a large number of independent molecular data sets (Fig. 1).

Fig. 1 Hominoid molecular relationships.

It is difficult to tease out the earliest references to the anatomy of individual great and lesser ape species because in the 17th century the terms ‘Satyr’ and ‘Orang-Outang’ were used more or less indiscriminately. With hindsight, it is clear that at various times these terms have been applied in different geographical regions to aboriginal modern humans as well as to genuine non-human hominoid primates. For example, although the ‘Ourang Outang’ anthropoid creature from Java referred to by Bontius (1658) is remarkably anthropomorphic in appearance, the context of the description suggests that the description, if not the illustration, was based on the Bornean orangutan (Yerkes & Yerkes, 1929). Likewise, despite the anthropomorphic nature of the figure in the famous engraved frontispiece of Tyson's (1699) monograph, the animal he described as ‘Orang-Outang, sive Homo Sylvestris’ was a juvenile chimpanzee, the skeleton of which is on display in the Natural History Museum, London. Conversely, it is clear from the details provided by Buffon (1780) that the orang-utan cadaver dissected by Tyson and Cowper was actually that of a modern human.

Despite being entitled Observations on the Anatomy of the Orang Outang, Traill (1821) deserves the distinction of being the first anatomical description of the chimpanzee. Traill (1818) is sometimes referenced in this context, but this citation refers to an abstract not to a description. There is general agreement that it was the collaboration between the Protestant missionary, Thomas Savage, and the anatomist, Jeffries Wyman, that brought the second African great ape, the gorilla, to the attention of the Western scientific community. Savage & Wyman (1847) were apparently the first researchers to distinguish it clearly from the chimpanzee. Thereafter, Richard Owen – who was one of the first scientists after Traill to dissect a chimpanzee (Owen, 1846) – elaborated on the distinctiveness of the gorilla (Owen, 1849, 1859, 1865). The distinction of introducing the larger of the Asian apes, the orangutan, to Western science clearly falls to Peter Camper. His writings make it very clear that he had dissected at least one specimen of Pongo prior to his published commentaries (Camper, 1779, 1782). Yet again, Richard Owen was one of the pioneers who generated additional information about the orangutan from his own dissections (Owen, 1843). The first sound recognition of the gibbons should be attributed to Le Comte (1697). Buffon (1780) consolidated the case for their distinctiveness, and Keith (1896), in a review of the gibbons, refers to providing ‘… incomplete descriptions of the anatomy of five animals’ (p. 372). Keith credits Kohlbrügge (1890/91) with the distinction of providing the first systematic description of gibbons based on dissection.

Although more than a century has elapsed since the publication of the last of these pioneering ape dissections, the amount of information about the soft tissues of hominoids that has been accumulated from subsequent dissection studies has been meager. The numbers of animals that have been systematically dissected is relatively small. For example, Henry Raven's (1950) anatomical researches on the anatomy of the gorilla were based on the dissection of a single adult male Gorilla gorilla carcass collected from southern Cameroon. Likewise, the observations on the thoracic and abdominal viscera made by Washburn (1950) and Elftman & Atkinson (1950) in the same volume are mainly based on information from the dissection of a single young adult female. These data were supplemented by observations from an adult male, but it seems likely that this was the cadaver Raven used. Swindler & Wood (1973) based their description of the soft-tissue anatomy of Pan on six individuals, and together with the four gibbon dissections reported by Kohlbrügge (1890/91), these are probably the largest comparative hominoid dissection series to have been reported in the literature. The same small sample sizes also apply to Pongo. For example, the primary data in Anderton's (1988) review of the appendicular myology of Pongo came from a single animal. The study of Thorpe et al. (1999) is one of the few recent investigations to involve the systematic dissection of a non-human hominoid, but whilst their sample comprised three Pan cadavers, the published information is confined to the muscular system.

There have been relatively few previous attempts to consolidate information about the soft tissues of the hominoids. Perhaps the most notable is the monumental multi-author Handbuch der Primatenkunde (Hofer et al. 1956) which includes data for hominoids together with information from other primate groups. Sadly Osman Hill did not live long enough to expand the coverage of his extraordinary monograph series Primates: Comparative Anatomy and Taxonomy to include the Hominoidea.

This study used the modern human soft-tissue structures listed in the Nomina Anatomica (NA) as a reference tool for taking stock of the published data about non-human hominoid soft-tissue morphology. Clearly this list omits a few structures not normally found in modern humans. However, it has the advantage that, because it is a list that has been developed over time by experienced human morphologists, if it errs then it does so on the side of being conservative and comprehensive. Only a very few of the entries are too generalized to be useful (see the references to the skin below). With the minimum of modification it was possible to match observations in the literature on non-human hominoids with the structures listed in the NA. Thus, the total number of relevant NA soft-tissue structures – 1783 – is a sensible denominator to use in order to assess the coverage of information about the non-human hominoids, both by system and by anatomical region.

The organization of the information was based on the scheme used in the Sixth Edition of the NA (Warwick & Brookes, 1989). Information from the literature was organized initially by system, or major system component (e.g. ‘arteries’, ‘veins’, ‘lymphatics’ within the vascular system), and then it was cross-referenced by region where appropriate (i.e. for muscles, nerves, arteries and veins). Four relatively crude regional categories were recognized, the ‘Head’ (H), ‘Forelimb’ (F), ‘Trunk’ (T) and ‘Hindlimb’ (HL). Information about the limb girdles was included in the respective limb categories, and neck structures were included in the ‘Trunk’ category. Vessels and nerves were dealt with by region rather than by system, so that, for example, the vasculature of the gut is dealt with under the vessel type, and then assigned to the ‘Trunk’ regional category, rather than to the ‘Alimentary System’.

To help the reader comprehend the large amount of information in Appendix 1, the data have been summarized in Table 2. The system categories, and when appropriate their regional subcategories, are set out in the rows of Table 2. The first column (NA) lists the total number of structures listed in the NA within that category, or subcategory. The second column (N-HH) gives the number of structures within any NA category, or subcategory, for which there is information for one, or more, non-human hominoid genus. Column three (NA%) provides the percentage, within each category and subcategory, of the NA structures for which information is available for at least one non-human hominoid. Column four (N-HH%) gives the cumulative percentage of the NA structures, for each category, or subcategory, for which there are data for at least one non-human hominoid. Column five (PA) provides the number of structures in the NA category, or subcategory, for which there are data that satisfy the criteria (see below) for inclusion in the phylogenetic analyses that form the second part of this contribution. Column six (NA%) gives the percentage of the PA structures in each of the NA categories, or subcategories. The final column (PA%) provides the cumulative percentage of PA structures, for each of the NA categories, and subcategories.

Table 2 System and regional distribution of soft-tissue structures sampled in at least one non-human hominoid. System categories and regional subcategories form the rows. The columns are as follows: NA = Numbers of soft-tissue structures in each category, or (more ...)

It is evident from Table 2 that the global figure of 35% of NA soft-tissue structures represented in the literature by information from more than one non-human hominoid obscures major differences in the representation of systems, tissues and regions. There are three general levels of sampling intensity. Muscles are sampled most intensively, with information being available for more than one non-human hominoid for nearly 90% of the muscles listed in the NA. Among the larger categories of structures the next level of sampling intensity, c. 40–50% of the NA structures, applies to the arteries, the heart and the nerves. The remaining numerically large NA categories are sampled at substantially lower levels of intensity. Of these, the best represented is the alimentary system with 27% of the NA structures represented in the literature. The venous component of the vascular system is the least well represented, at 12%. Among the categories with smaller numbers of structures listed in the NA, the endocrine glands and the skin are relatively well represented, at 43% and 44%, respectively. There appear to be discrepancies between the information under the ‘Skin’ system category given in Tables 2 and 3 and Appendix 1, and the zero score in this category in Table 4. This is because although there are data for the skin in the literature, these data do not correspond to any of the major structural skin subcategories given in the NA.

Table 3 System and regional distribution of soft-tissue structures sampled by all four non-human hominoids. System categories, and their regional subcategories, form the rows. The columns are as follows: NA = Numbers of soft-tissue structures in each of the system (more ...)

Table 4 Soft-tissue structure information broken down by system category, and subcategories, and genus. System categories and subcategories form the rows. The columns are the total number of taxonomic appearances for that system, together with the system rank-order (more ...)

Regional differences in sampling intensity are also noteworthy, and will be referred to again in the ‘Discussion’ section. When the major system categories, or subcategories, are broken down into the four major regions, the forelimb is always either the most intensively sampled region, or, in the case of the muscles, it shares that distinction with the hindlimb. In contrast, the head is always the region least intensively sampled in the existing literature.

If the sampling criterion is altered to consider the NA structures for which information is available for all four of the non-human hominoid primates (Table 3), the dominance of evidence about muscles, and the more intensive sampling of the forelimb, are themes that are repeated. The organization of Table 3 follows that of Table 1, except that the N-HH column in the former refers to structures for which there is information for all four non-human hominoids. The muscle category comprises c. 48% of the structures thus sampled, and just less than half of these – 48 out of 112 – are forelimb muscles. An even more marked forelimb regional dominance – 23 out of 40 – is seen in the artery category. Comparable levels of forelimb dominance – 4 out of 9 – are also seen in the vein subcategory, and in the nerves, where 17 out of a total of 35 come from the forelimb. The head is consistently the least well sampled region. In the case of arteries, muscles and nerves, the head is the region with the poorest representation, and in the venous vascular subcategory it ties with the hindlimb as the region with the poorest sample. The bias in favour of the limbs in general, and the forelimb in particular, is even more remarkable when it is realised that the limbs generally contribute a relatively small percentage of the structures in the NA system categories that can be broken down into regional subsets (i.e. vessels, muscles and nerves).

The soft-tissue data are sorted by taxon in Table 4. The rows are the main NA categories. The first column is the total number of taxon occurrences in that category, and the second column gives the rank order of those occurrences. The remaining columns provide the number of occurrences for that taxon in each NA category, followed by the percentage of the total number. Overall, the non-human hominoid for which information is most abundant is Pan. This taxon has data recorded in the literature for almost a third, 32%, of the soft-tissue structures listed in the NA. Gorilla and Pongo have equal representation, with 26% of the NA structures sampled. Hylobates, at 16%, is the least well sampled living hominoid. For all but two (the pericardium and the urogenital system) of the major NA categories Pan is the best sampled hominoid. In both of the two exceptions Gorilla takes the place of Pan as the most intensively sampled taxon. Within the largest NA category, muscles, Pan and Gorilla are equally well sampled. The sampling intensity in Hylobates never exceeds that in the two non-human African apes, but in two of the major NA soft-tissue categories, the peritoneum and the urogenital system, Hylobates is more intensively sampled than Pongo.

Phylogenetic analyses of traditional cranial and dental morphological data have generally supported hypotheses of relationships for Homo and the living apes that conflict with the consensus molecular phylogeny for the group. The latter links Homo and Pan in a clade to the exclusion of Gorilla, positions Pongo as the sister taxon of the Homo and African apes, and locates Hylobates as the basal extant hominoid (Ruvolo, 1997; Fig. 1). In contrast, some of the analyses using traditional hard-tissue data have suggested that Homo and Pongo form a clade to the exclusion of Gorilla and Pan (Schwartz, 1984a, 1984b, 1988). Others suggest that the African apes, Gorilla and Pan, form a clade to the exclusion of Homo and Pongo, and that Homo and the African apes form a clade to the exclusion of Pongo (Andrews, 1987). Still other studies suggest that the Asian apes, Hylobates and Pongo, are more closely related to one another than either is to any of the African apes or to humans (e.g. Oxnard, 1987, p. 217). Yet more studies have produced phylogenies in which the three great apes are shown to be more closely related to each other than any of them is to Homo (Kluge, 1983; Collard & Wood, 2000). So far, the only morphological analysis to support the same hypothesis of relationship as the molecular data is Shoshani et al. (1996). However, a recent bootstrap analysis of the data used by Shoshani et al. has shown that their data set does not provide statistically significant support for the (Homo, Pan) clade (Gibbs, 1999). Thus, none of the morphological analyses of the extant hominoids carried out so far have can be said to support the same phylogeny as the molecular data. Rather, they have generally suggested relationships that conflict with the molecular phylogeny, or in the one case in which the resulting phylogeny is consistent with the molecular evidence, little confidence can be placed on the result. In view of the foregoing, we have used the soft-tissue data discussed in the first part of this paper as the basis of a new phylogenetic analysis.

One hundred and seventy-one characters conformed with the three criteria. This is 26 fewer than the number of characters analysed by Gibbs et al. (2000). Since the publication of that study the character list has been further refined to eliminate redundancy, maximise the number of ordered characters, and to exclude characters where differences in sample size might have been influencing the choice of character states. We stress that, whilst we have made every effort to maximise the reliability of the data set, it should nevertheless be treated as a ‘work in progress’. In particular, there is a pressing need for studies that will shed further light on variation in the 171 characters within each of the four extant ape genera.

Brief descriptions of the characters, their states and distribution, and the references from which the data were taken are given in Appendix 2. To facilitate further analysis of the characters, they have been organized into slightly different regional and system groups than those used in the NA and Table 2. For example, the characters relating to the neck and tongue, including the surface features of the latter, are included in the ‘Head’ region. Muscles originating in the trunk, but which attach distally to the lower limb, are included in the ‘Trunk’ region. Striated muscles of the male external genitalia are included in the ‘Genito-Urinary’ system, and not with ‘Muscles’. The character state data were additively coded, and a taxon-by-character matrix was compiled.

The data matrix was used to perform two tests of the hypothesis that soft-tissue characters can be relied upon to reconstruct the phylogenetic relationships of the hominoids. The first test was based on parsimony analysis, which identifies the cladogram/s requiring the smallest number of ad hoc hypotheses of character state change to account for the distribution of character states among the taxa. The matrix was subjected to parsimony analysis using PAUP^* 4 (Swofford, 1998), and the shortest cladogram compared to the consensus molecular cladogram for the extant hominoids (Fig. 1). Because parsimony analysis cannot discriminate ‘true’ and ‘false’ clades, we judged the hypothesis to be supported if the analysis favoured either a fully resolved cladogram that was consistent with the molecular cladogram, or a partially resolved cladogram that comprised only molecular clades. We also considered the hypothesis supported if the analysis produced several equally parsimonious cladograms whose strict consensus comprised only clades that were compatible with the molecular cladogram.

The second test of the hypothesis used the phylogenetic bootstrap. This methodology assesses the confidence interval associated with a clade (Felsenstein, 1985; Sanderson, 1995). Using PAUP^* 4, 10 000 matrices were derived from each matrix by sampling with replacement. The new matrices were subjected to parsimony analysis, and a consensus of the most parsimonious cladograms was computed using a confidence region of 70% (Hillis & Bull, 1993). Thereafter, the clades of the consensus cladogram were compared to the molecular cladogram (Fig. 1). In this test the best supported clades should not be ‘false’ clades, since it is commonly assumed in primate phylogenetics that the better the bootstrap support for a clade, the more likely the clade is to be ‘true’ (cf. Corruccini, 1994).

In both the parsimony and the bootstrap analyses, characters were given equal weights. Where obvious transformation series could be identified (e.g. Extent of costal origin of serratus anterior: 0 = ribs 1–9 and occasionally rib 10, 1 = ribs 1–11, 2 = ribs 1–11 and last rib), characters were treated as ordered variables. Otherwise they were treated as unordered variables. Appendix 2 indicates whether a character was treated as an ordered or an unordered variable. Significantly, the results of an analysis in which all the characters were treated as unordered variables produced comparable results to the one described here. No a priori judgements were made as to the primitive or derived condition of characters. Instead, Hylobates was assumed to be the basal hominoid genus and the cladograms were rooted accordingly. The cladograms were obtained using the branch and bound search routine of PAUP^* 4.0.

The bootstrap analysis also supported the hypothesis that hominoid soft-tissue characters are reliable for phylogenetic reconstruction. The (Homo, Pan) clade was supported by 95% of the bootstrap replicates, and the (Gorilla, Pan, Homo) clade by 96%. Alternative groupings, including the traditional (Gorilla, Pan and Pongo) clade and the (Homo, Pongo) clade promoted by Schwartz (1984a, 1984b, 1988) received less than 5% support.

This study used soft-tissue structures listed in the Nomina Anatomica to summarise the published data about non-human hominoid soft-tissue morphology. The taxon coverage is summarized in Table 4. The predominance of information about Pan is intriguing, especially when it is realised that the vast majority of these observations about soft-tissue morphology were made and published well before it was realised that there is a particularly close relationship between Pan and modern humans. It is also noteworthy, for the same reason, that there is as much information about Pongo as there is about Gorilla. The gibbons come a poor fourth in the list, with information for Hylobates (16% of the total) only being available for half the number of NA structures for which data exist for Pan.

The rank order of the total taxon occurrences by system categories and subcategories is also given in Table 4. This rank order, at least for the six best represented NA categories and subcategories, is generally consistent across the four non-human hominoid taxa. The numerical pre-eminence of information about muscles, arteries and nerves is perhaps unsurprising given that across the years these structures have attracted the interest of comparative and clinical anatomists. However, the consistently higher rank for urogenital system structures compared to those from the alimentary system is unexpected, and not easily explained.

When we consider the pattern of regional representation of system categories and subcategories for the structures for which data exist for all four non-human hominoids (Table 3), it is evident that there are substantial regional biases. The most obvious bias is in favour of the limbs, and in particular the forelimb. This latter bias is particularly striking for the subcategories of the vascular system. The extent of the over representation of the limbs has to be considered in relation to the relative numbers of soft-tissue structures in the four anatomical regions in each of the major NA system categories. So, for example, whereas the limbs contribute 28% and 30% of the arteries and nerves in the relevant NA category (Table 5), they make up 80% and 95% of the respective structure categories in the PA (Table 6). The systematic under representation of the soft tissues of the head is in marked contrast to the situation for hard tissues. In the latter case, and probably because of the influence of taphonomy on the palaeontological record, information about the teeth and the skull for the non-human hominids far exceeds the hard-tissue data that are available for the rest of the body (Shoshani et al. 1996; Collard & Wood, 2000). What is remarkable is that interest in functional analysis has not stimulated researchers to more gather comparative information about the soft tissues of the head and neck. There is, for example, no information about the major masticatory muscles of the non-human hominids in the list of PA structures (see Appendix 2). There is clearly an urgent need to develop a comprehensive database for the head and neck soft-tissue anatomy of the non-human hominids.

Table 5 Regional distribution of soft-tissue structures for three of the largest system categories in the Nomina Anatomica

Table 6 Regional breakdown and major soft-tissue system categories for the characters used in the phylogenetic analysis

Turning now to the phylogenetic utility of the soft-tissue morphology, the results of the parsimony and bootstrap tests strongly support the hypothesis that soft-tissue characters can be relied upon to reconstruct the phylogenetic relationships of the extant hominoids. The parsimony analysis unambiguously favoured a cladogram with the same topology as the molecular cladogram, and the bootstrap analysis returned high levels of support for clades that correspond to those of the molecular cladogram. The two main alternative hypotheses of relationship that have been suggested for the extant hominoids received extremely low levels of support in the bootstrap test. The (Gorilla, Pan) clade, that until recently was favoured by most morphologists (e.g. Andrews, 1987, 1992; Andrews & Martin, 1987), featured in less than 5% of the bootstrap cladograms, as did the (Homo, Pongo) clade promoted by Schwartz (1984a, 1984b, 1988). Thus, our data set provides unambiguous morphological endorsement for the phylogeny that is overwhelmingly supported by the molecular evidence. Given that the molecular phylogeny is widely considered to be accurate, our analysis suggests that extant hominoid soft-tissue characters have more phylogenetic utility than hominoid craniodental hard tissues, which conspicuously fail to recover the molecular consensus phylogeny (Hartman, 1988; Collard & Wood, 2000). It is worth noting that the analyses provide stronger support for the molecular phylogeny than those carried out by Gibbs et al. (2000) even though the revisions to the data set were made without reference to the molecular phylogeny.

Why do higher primate soft-tissue and hard-tissue characters differ in their phylogenetic utility? A clue may come from the results of experiments that used rhombomere quail-to-chick grafts to investigate the influence of hindbrain segmentation on craniofacial patterning (Köntges & Lumsden, 1996). This experimental study showed that each rhombomeric population remains coherent throughout ontogeny, with rhombomere-specific matching of muscle connective tissue and their attachment sites for all branchial and tongue muscles. If a similar system operates elsewhere in the body, it would help explain how muscle gross morphology is conserved, whereas the shapes of the skeletal elements to which the muscles are attached are susceptible to changes that contrive to obscure phylogeny.

Another contributory factor may be that soft-tissue characters are not as prone to homoiology as skeletal characters. The term homoiology has been used to refer to shared character states that are phylogenetically misleading and which result from similarities in the way that genotypes interact with the environment (Lieberman, 2000). It has been claimed that, because bone is a dynamic tissue, many osseous morphologies may be homoiologous (Lieberman, 2000). We suspect that homoiology plays a minor role in the generation of the phenotypes we use in our soft-tissue data set. Whereas the mass of a muscle may be affected by activity or inactivity, its attachments are unlikely to be. Likewise, mechanical loading is unlikely to affect the branching pattern of an artery, or the number of digits supplied by a given nerve. Nevertheless, homoiology, as interpreted above, cannot be the whole explanation for the difference in phylogenetic utility between the hard and soft tissues. Because dental enamel does not remodel, it is not prone to homoiology. Yet Hartman (1988) found that molar morphology is unreliable for reconstructing the phylogenetic relationships of the extant hominoids. Thus, other factors must also be involved in reducing the phylogenetic utility of teeth relative to that of soft tissues. Some authors have suggested that function may be a cause of phylogeny-obscuring evolutionary change in tooth morphology (Hartman, 1988; Hunter & Jernvall, 1995). However, recent work on the dentition of the Lake Lagoda seal suggests that developmental constraints may also be a reason why tooth morphology is prone to homoplasy and is therefore a poor guide to low-level phylogenetic relationships (Jernvall, 2000).

This study has shown that for the extant hominoids, and by extension for other higher primates, the classic ‘molecules vs. morphology’ conflict (Patterson, 1987) does not hold. Rather, the contrast is apparently between molecules and soft-tissue morphology on the one hand, and cranio-dental hard-tissue morphology on the other. However, it is possible that factors other than the nature of the tissue may be influencing the outcome of this study. The 171 soft-tissue characters are not distributed across the major body systems in proportion to the numbers of structures listed in the NA (Table 5), nor are they distributed evenly across the regions of the body. Muscles (64%) predominate in the 171 PA characters (Table 6), whereas two out of the three vascular subcategories, the veins and the lymphatics, are poorly represented and unrepresented, respectively, in the PA structure list (Table 6). Like the distributions of the structures set out in Tables 2 and 3, the 171 PA characters are affected by very substantial regional biases that favour the limbs. Thus, 141 of the 171, or 82%, of the characters included in the phylogenetic analysis are limb characters (Table 6). In contrast, the head is badly under represented, so that, for example, there are no head and neck arteries or veins in the PA list (Table 6). Thus, there are two major differences between this and previous studies of relationships among the living hominids. First, there is its restriction to soft tissues. Second, because of the nature of the published information about non-human hominoid morphology, the majority of the data used in the study are from the limbs. The obvious next step is to use the consensus hominoid molecular cladogram to examine whether hard-tissue evidence from the limbs performs as well as limb soft-tissue evidence, and to see if soft-tissue evidence from the head performs as poorly as the hard-tissue evidence from the same region.

This study would not have been possible without the award of an Anatomical Society of Great Britain and Ireland studentship to S.G. M.C. was funded by a Wellcome Trust Bioarchaeology Fellowship award. B.W.'s initial participation was supported by The Leverhulme Trust, and is now funded by The Henry Luce Foundation. We thank Michael Günther, John Shaw-Dunn, David Pilbeam and an anonymous referee for their advice and comments. Lastly, we would like to express our gratitude to Matt Cartmill, whose meticulous reviewing did much to improve the paper.